DISEASES OF AQUATIC ORGANISMSDis Aquat Org

Vol 124 41ndash54 2017httpsdoiorg103354dao03111

Published March 30

INTRODUCTION

The American lobster Homarus americanus sup-ports a valuable commercial fishery in the Gulf ofMaine with landings totaling over half a billion dol-lars in 2014 (National Marine Fisheries Service 2014)Although landings are at historic highs the southernextent of this fishery has recently come under threatwith epizootic lobster shell disease being one poten-

tial driver (Tlusty et al 2014) This disease occurswhen bacteria consume the lobster cuticle resultingin deep lesions on infected lobsters (Smolowitz et al2005) Most types of shell disease first appear asmelanized spots on the epicuticle (Tlusty amp Metzler2012) In the epizootic form of shell disease thesespots quickly develop into shallow lesions on the exo-cuticle that ultimately penetrate into the endocuticleand eventually cause ulceration (Smolowitz et al

copy Inter-Research 2017 middot wwwint-rescomCorresponding author mtlustyneaqorg

Fine-scale transition to lower bacterial diversity andaltered community composition precedes shell disease

in laboratory-reared juvenile American lobster

Sarah G Feinman1 Andrea Unzueta Martiacutenez25 Jennifer L Bowen15 Michael F Tlusty34

1Biology Department University of Massachusetts Boston Boston MA 02125 USA2Biology Department University of Hawaii at Manoa Honolulu HI 96822 USA

3Anderson Cabot Center for Ocean Life New England Aquarium Boston MA 02110 USA4School for the Environment University of Massachusetts Boston Boston MA 02125 USA

5Present address Department of Marine and Environmental Science Northeastern University Nahant MA 01908 USA

ABSTRACT The American lobster Homarus americanus supports a valuable commercial fisheryin the Northeastern USA and Maritime Canada however stocks in the southern portion of thelobsterrsquos range have shown declines in part due to the emergence of shell disease Epizootic shelldisease is a bacterially induced cuticular erosion that renders even mildly affected lobsters unmar-ketable because of their appearance and in more severe cases can cause mortality Despite theimportance of this disease the associated bacterial communities have not yet been fully character-ized We sampled 2 yr old laboratory-reared lobsters that displayed signs of shell disease at thesite of disease as well as at 05 1 and 15 cm away from the site of disease to determine how thebacterial community changed over this fine spatial scale Illumina sequencing of the 16S rRNAgene revealed a distinct bacterial community at the site of disease with significant reductions inbacterial diversity and richness compared to more distant sampling locations The bacterial com-munity composition 05 cm from the site of disease was also altered and there was an observabledecrease in bacterial diversity and richness even though there were no signs of disease at thatlocation Given the distinctiveness of the bacterial community at the site of disease and 05 cmfrom the site of disease we refer to these communities as affected and transitionary and suggestthat these bacteria including the previously proposed causative agent Aquimarina lsquohomariarsquo areimportant for the initiation and progression of this laboratory model of shell disease

KEY WORDS American lobster middot Shell disease middot Bacterial communities middot Microbiome middot Highthroughput sequencing middot 16S rRNA gene

Resale or republication not permitted without written consent of the publisher

Dis Aquat Org 124 41ndash54 2017

2005) Due to the physical severity of epizootic shelldisease even mildly affected lobsters are generallyconsidered unmarketable due to the unappealingappearance of their shells thus affecting the value ofthis important commercial fishery (Castro et al 2012)

In addition to aesthetic changes associated withepizootic shell disease animals can also suffer fromdecreased health (Tlusty et al 2014) and increasedmortality (Castro et al 2012) Lobsters affected byshell disease are energetically compromised withdisruptions to their metabolism and hormone signal-ing (Tarrant et al 2012) They also tend to have highlevels of the molting hormone ecdysone which candisrupt their reproductive cycle (Laufer et al 2005)These physiological changes may in turn influencepopulation dynamics particularly since disease pre -valence among egg-bearing females can be as highas 70 (Castro amp Somers 2012) The recent declinein the southern New England American lobster pop-ulation has been positively associated with an in creasein epizootic shell disease (Castro et al 2012 Howell2012) highlighting the importance of understandingthe ecological dynamics of this disease

Despite the importance of epizootic shell disease tothe health of the American lobster and the importanceof this fishery to coastal economic sustainability thebacteria associated with epizootic shell disease havenot yet been fully identified Initial investigation ofthe bacteria associated with epizootic shell diseaseidentified a few key members of the shell micro-biome (Chistoserdov et al 2005) however this studywas based on culturing techniques which typicallyassess less than 1 of the bacterial community (Sta-ley amp Konopka 1985) Later studies based on culture-independent fingerprinting methods eliminated thisculture bias however those studies did not fullycharacterize the bacterial community associated withboth healthy and diseased shell throughout the pro-gression of the disease (Chistoserdov et al 2012Quinn et al 2012b Whitten et al 2014) A higher res-olution sequencing study using next-generation se -quencing techniques identified 170 different bacter-ial species associated with the lobster shell an orderof magnitude more than previous studies (Meres etal 2012) This study examined the lobster shell bac-terial community associated with 3 distinct diseasestates diseased cuticle healthy cuticle from diseasedanimals and healthy cuticle from healthy animalsand concluded that the epizootic shell disease maybe due to a dysbiotic shift in the shell bacterial com-munity (Meres et al 2012) These results highlightthe importance of examining the bacterial commu-nity as a whole as this disease may be due to changes

in the community not individual bacteria Further anassessment of temporal changes in shell bacterialcommunities is needed to better understand the pro-gression of shell disease

The importance of microbially induced diseases tocoastal ecosystem health is well documented (Harvellet al 1999 Fey et al 2015) however it is still difficultto link specific bacterial taxa to these diseases In thisstudy we used a laboratory-based model that hasbeen used in numerous studies of lobster shell dis-ease (Quinn et al 2012ab Tlusty amp Metzler 2012Whitten et al 2014) to characterize the bacterial com-munities recovered from lobster shells Our objec-tives were to (1) compare healthy and diseased areasof the lobster shell (2) investigate how these commu-nities changed spatially and temporally as the dis-ease progressed and (3) identify any associated dys-biotic shifts that occurred in the bacterial communityWe used Illumina 16S rRNA gene sequencing onsamples collected over very fine spatial scales fromthe shell of diseased and healthy juvenile lobsters toanalyze the lobster shell bacterial community and weexamined how this community changed with dis-tance from the site of disease The animals used inthis study were not formally diagnosed with epizooticshell disease they instead suffered from a laboratorymodel of shell disease (LMSD) that is common in cap-tivity and likely due to environmental stress Theseresults nonetheless allow closer scrutiny of the pro-cess of lesion initiation and spread which would notbe possible in wild lobster populations Based on pre-vious studies of LMSD (Whitten et al 2014) wehypothesized that the bacterial community wouldgradually change and become less diverse with in -creasing distance from the site of melanization Ourresults indicate shifts in bacterial community compo-sition at the site of disease as well as at 05 cm fromthe site of disease though these shifts were ac -companied by a decrease rather than an increase inbacterial richness and diversity compared to unaf-fected shell

MATERIALS AND METHODS

Sample collection

Samples were collected from 5 juvenile Americanlobsters raised at the New England Aquarium Lob-ster Research and Rearing Facility (Boston MA USA)which features a semi-closed recirculation systemwith water from Boston Harbor (15 water replace-ment daily) Lobsters were housed in conditions sim-

42

Feinman et al Bacterial communities in lobster shell disease

ilar to those described in prior shell disease re search(Chistoserdov et al 2012 Quinn et al 2012b Tlustyamp Metzler 2012 Whitten et al 2014) Briefly all lob-sters were kept in individual 9 cm diameter contain-ers (hereafter referred to as pens) and fed a diet con-sisting of Economac-4trade (Aqua fauna Bio-Marine) acommercially available shrimp feed This diet lacksthe carotenoid astaxanthin causing the lobster shellto have no color a desirable phenotype for this studyas it allows early detection and better visualization ofthe melanization associated with LMSD (Tlusty ampHyland 2005)

The diseased lobsters used in this study were ap -proximately 2 yr old and had all molted approxi-mately 5 to 6 mo before the initiation of sampling withthe exception of a single lobster that molted approxi-mately 3 mo prior to the initiation of sampling Theselobsters were chosen because they all had visiblesigns of LMSD on their left claw though the degreeof melanization varied among lobsters AlthoughLMSD was also present on other parts of the lobstersrsquoshells we focused solely on the left claw of each lob-ster in order to eliminate any potential confoundingdifferences that could arise from comparing differentparts of the lobster shell (Meres et al 2012) We cate-gorized the degree of melanization on each lobsterclaw defining lesions as melanized areas of the shellthat had become soft or pitted and spots as areas thatwere melanized but maintained a rigidity similar tothat of the unaffected lobster shell

We sampled the diseased lobsters approximatelyevery 10 d over the course of the experiment for atotal of 4 time points per lobster Shell microbiomesamples were collected using sterile toothpicks toscrape the bacterial community from the desired areaof each lobster shell Gentle contact was made be -tween the toothpick and the lobster and this contactwas maintained over the entire sampling area Eachtoothpick was then stored in an individual cryovialand transported to the laboratory on dry ice where itwas transferred to a minus20degC freezer During each sam-pling shell scrapings were taken from the center ofthe lesion or spot on the left claw (referred to as 0 cm)as well as at a distance of 05 1 and 15 cm laterallyat an angle away from the dactylus (Fig 1) Sampleswere also taken from each lobster pen at a point justbelow the waterline while the pen was out of thewater on all 4 sampling dates for a total of 100 sam-ples (5 samples per lobster 5 lobsters 4 time points)We kept photographic records of each lobster through-out the course of the experiment and used these pho-tographs to guide our sampling and ensure that sim-ilar shell locations were sampled on each visit Some

of the lobsters molted during the course of the exper-iment but we continued to sample the same locationsas described above after molting

In addition to these samples we also collected shellbacterial community samples from 5 lobsters thatappeared to be free of disease These additional lob-sters were from the same age group and housed inthe same conditions but had all molted approximately1 mo before the initiation of sampling and showed novisible signs of LMSD We sampled these lobsters onthe same dates as the diseased lobsters but only tooka single bacterial community sample from the leftclaw of each lobster These lobsters served as a rep-resentation of what a healthy bacterial communitymight look like on a disease-free lobster so lobstersfrom this group that developed signs of LMSD weresubsequently eliminated from the study

DNA extraction and sequencing

DNA was extracted from toothpicks using the Pow-erSoilreg DNA Extraction Kit (MOBIO) following themanufacturerrsquos instructions Kit negative controlsextractions containing no toothpick were performedperiodically and subsequently amplified to ensurethat there was no kit contamination in the extracts(Salter et al 2014) Additionally DNA was extracteddirectly from a sterile toothpick to ensure there wasno amplification from the toothpick alone DNA ex -tracts from lobster and pen samples were amplifiedin triplicate using bacterial specific (515F 5rsquo-GTGCCA GCM GCC GCG GTA A-3rsquo 806R 5rsquo-GGA CTACHV GGG TWT CTA AT-3rsquo) uniquely barcoded 16SrRNA primers containing adaptors for Il lumina se -quencing (Caporaso et al 2012) Each 25 microl PCRreaction contained 10 microl 5-Prime Hot Master Mix(5 Primereg Thermo Fisher Scientific) and 02 microM ofeach primer In addition 05 to 2 microl of DNA was ad -ded to each reaction depending on the original DNA

43

Fig 1 Main sampling scheme employed in this study Shellscrapings were taken from (S) center of lesion or spot on theleft claw (referred to as 0 cm) and (S) 05 cm (S) 10 cm

and (S) 15 cm from the lesion or spot

Dis Aquat Org 124 41ndash54 2017

concentration PCR products were purified using theQIAquickreg PCR Purification Kit (QIAGEN) followingthe manufacturerrsquos instructions and quantified usinga Qubitreg 20 Fluorometer (Life Technologies ThermoScientific) PCR products were pooled in equimolaramounts and the quantity of the pool was checkedusing the KAPA Library Quantification Kit for IlluminaSequencing Platforms (Kapa Biosystems) The poolwas then loaded onto a MiSeq sequencer (Illumina)and sequenced using a paired-end V2 300 cycle kit

Sequence analysis

Quality filtering and sequence analysis was per-formed in QIIME using default parameters unlessnoted (Caporaso et al 2010) Paired-end reads werejoined specifying at least 10 base pairs in overlap andsequences within a Phred score of 30 or higher wereretained (indicating 999 accuracy in base calling)Chimeric sequences were identified using usearch61an algorithm that performs both de novo and refer-ence-based chimera detection with the GenomesOnLine Database (GOLD) as a reference and chimeraswere subsequently removed (Edgar et al 2011Reddy et al 2015) After these quality-filtering steps2 726 677 sequences remained Operational taxonomicunits (OTUs) were assigned at 97 sequence iden-tity using Swarm (Maheacute et al 2014) OTUs identifiedas Archaea or chloroplasts or that were unassignedat the kingdom level using the Ribosomal DatabaseProject (RDP) classifier (Wang et al 2007) were re -moved from downstream analysis Samples were rar-ified to the lowest sequencing depth 4735 OTUs persample for analysis of community composition andalpha diversity

Statistical analysis

The weighted UniFrac metric was used for analysisof community composition because it considers bothphylogenetic distance and abundance (Lozupone etal 2011) We visualized differences in bacterial com-munity structure among 05 cm increments on eachlobster using principal coordinates analysis and testedfor differences between these communities using per-mutational multivariate analysis of variance (PERM-ANOVA Anderson 2001) with individual lobster as ablocking component To test for pairwise differencesin each 05 cm increment we compared logit-trans-formed weighted UniFrac similarities of each 05 cmcomparison using a Welchrsquos t-test which accounts for

unequal variance again using individual lobster as ablocking component (R Core Team 2014) We alsoused QIIME to calculate 2 alpha diversity metricsphylogenetic distance a richness estimator that takesphylogeny into account (Faith amp Baker 2006) andShannon diversity Significance testing of alpha diver-sity measures was done using analysis of variance onlinear models that included the individual lobstersampled as a random effect

We defined the core microbiome as OTUs thatwere in n minus 1 samples within a given shell conditionThis resulted in 76 OTUs that were considered to belsquocorersquo to 1 or more shell condition We then identifiedrepresentative sequences from each of these OTUsusing the NCBI Basic Local Alignment Search Tool(BLAST) All sequence information produced fromthis study is available from the Sequence ReadArchive under accession no PRJNA307069

RESULTS

Sequence analysis of the 100 bacterial samplestaken from 5 diseased lobsters and their pensyielded 2 726 677 sequences corresponding to 18 204unique OTUs defined at 97 sequence identityAnalysis of samples taken from the left claw of lob-sters affected by LMSD showed that bacterial com-munity composition at the site of disease (distance0 cm) clustered separately from most other samples(Fig 2A) indicating that lobster shell at the siteof disease harbors a distinct bacterial community(PERMANOVA F = 2277 p lt 0001 df = 62) Thebacterial community found on lobster shell 05 cmaway from the site of LMSD (Fig 2A) was betweenthe 0 cm community and the 1 and 15 cm communi-ties and distinct from each (F = 713 p lt 0001 df =31 and F = 618 p lt 0001 df = 46) indicating thatthis may be a transitionary bacterial communityintermediate between affected and unaffected lob-ster shell Bacterial communities from lobster shell 1and 15 cm away from the site of the LMSD clus-tered together and more tightly than the bacterialcommunities from the other 2 sets of samples indi-cating that these communities are more similar(Fig 2A) and may represent an un affected bacterialcommunity distinct from affected and transitionarycommunities (F = 1700 p lt 0001 df = 62) Impor-tantly we did not see a change in bacterial commu-nity composition driven by sampling date (Fig S1 inthe Supplement at www int-res com articles suppld124p041 _ supp pdf) indicating that distance fromthe site of disease was more important than temporal

44

Dis Aquat Org 124 41ndash54 2017

2005) Due to the physical severity of epizootic shelldisease even mildly affected lobsters are generallyconsidered unmarketable due to the unappealingappearance of their shells thus affecting the value ofthis important commercial fishery (Castro et al 2012)

In addition to aesthetic changes associated withepizootic shell disease animals can also suffer fromdecreased health (Tlusty et al 2014) and increasedmortality (Castro et al 2012) Lobsters affected byshell disease are energetically compromised withdisruptions to their metabolism and hormone signal-ing (Tarrant et al 2012) They also tend to have highlevels of the molting hormone ecdysone which candisrupt their reproductive cycle (Laufer et al 2005)These physiological changes may in turn influencepopulation dynamics particularly since disease pre -valence among egg-bearing females can be as highas 70 (Castro amp Somers 2012) The recent declinein the southern New England American lobster pop-ulation has been positively associated with an in creasein epizootic shell disease (Castro et al 2012 Howell2012) highlighting the importance of understandingthe ecological dynamics of this disease

Despite the importance of epizootic shell disease tothe health of the American lobster and the importanceof this fishery to coastal economic sustainability thebacteria associated with epizootic shell disease havenot yet been fully identified Initial investigation ofthe bacteria associated with epizootic shell diseaseidentified a few key members of the shell micro-biome (Chistoserdov et al 2005) however this studywas based on culturing techniques which typicallyassess less than 1 of the bacterial community (Sta-ley amp Konopka 1985) Later studies based on culture-independent fingerprinting methods eliminated thisculture bias however those studies did not fullycharacterize the bacterial community associated withboth healthy and diseased shell throughout the pro-gression of the disease (Chistoserdov et al 2012Quinn et al 2012b Whitten et al 2014) A higher res-olution sequencing study using next-generation se -quencing techniques identified 170 different bacter-ial species associated with the lobster shell an orderof magnitude more than previous studies (Meres etal 2012) This study examined the lobster shell bac-terial community associated with 3 distinct diseasestates diseased cuticle healthy cuticle from diseasedanimals and healthy cuticle from healthy animalsand concluded that the epizootic shell disease maybe due to a dysbiotic shift in the shell bacterial com-munity (Meres et al 2012) These results highlightthe importance of examining the bacterial commu-nity as a whole as this disease may be due to changes

in the community not individual bacteria Further anassessment of temporal changes in shell bacterialcommunities is needed to better understand the pro-gression of shell disease

The importance of microbially induced diseases tocoastal ecosystem health is well documented (Harvellet al 1999 Fey et al 2015) however it is still difficultto link specific bacterial taxa to these diseases In thisstudy we used a laboratory-based model that hasbeen used in numerous studies of lobster shell dis-ease (Quinn et al 2012ab Tlusty amp Metzler 2012Whitten et al 2014) to characterize the bacterial com-munities recovered from lobster shells Our objec-tives were to (1) compare healthy and diseased areasof the lobster shell (2) investigate how these commu-nities changed spatially and temporally as the dis-ease progressed and (3) identify any associated dys-biotic shifts that occurred in the bacterial communityWe used Illumina 16S rRNA gene sequencing onsamples collected over very fine spatial scales fromthe shell of diseased and healthy juvenile lobsters toanalyze the lobster shell bacterial community and weexamined how this community changed with dis-tance from the site of disease The animals used inthis study were not formally diagnosed with epizooticshell disease they instead suffered from a laboratorymodel of shell disease (LMSD) that is common in cap-tivity and likely due to environmental stress Theseresults nonetheless allow closer scrutiny of the pro-cess of lesion initiation and spread which would notbe possible in wild lobster populations Based on pre-vious studies of LMSD (Whitten et al 2014) wehypothesized that the bacterial community wouldgradually change and become less diverse with in -creasing distance from the site of melanization Ourresults indicate shifts in bacterial community compo-sition at the site of disease as well as at 05 cm fromthe site of disease though these shifts were ac -companied by a decrease rather than an increase inbacterial richness and diversity compared to unaf-fected shell

MATERIALS AND METHODS

Sample collection

Samples were collected from 5 juvenile Americanlobsters raised at the New England Aquarium Lob-ster Research and Rearing Facility (Boston MA USA)which features a semi-closed recirculation systemwith water from Boston Harbor (15 water replace-ment daily) Lobsters were housed in conditions sim-

42

Feinman et al Bacterial communities in lobster shell disease

ilar to those described in prior shell disease re search(Chistoserdov et al 2012 Quinn et al 2012b Tlustyamp Metzler 2012 Whitten et al 2014) Briefly all lob-sters were kept in individual 9 cm diameter contain-ers (hereafter referred to as pens) and fed a diet con-sisting of Economac-4trade (Aqua fauna Bio-Marine) acommercially available shrimp feed This diet lacksthe carotenoid astaxanthin causing the lobster shellto have no color a desirable phenotype for this studyas it allows early detection and better visualization ofthe melanization associated with LMSD (Tlusty ampHyland 2005)

The diseased lobsters used in this study were ap -proximately 2 yr old and had all molted approxi-mately 5 to 6 mo before the initiation of sampling withthe exception of a single lobster that molted approxi-mately 3 mo prior to the initiation of sampling Theselobsters were chosen because they all had visiblesigns of LMSD on their left claw though the degreeof melanization varied among lobsters AlthoughLMSD was also present on other parts of the lobstersrsquoshells we focused solely on the left claw of each lob-ster in order to eliminate any potential confoundingdifferences that could arise from comparing differentparts of the lobster shell (Meres et al 2012) We cate-gorized the degree of melanization on each lobsterclaw defining lesions as melanized areas of the shellthat had become soft or pitted and spots as areas thatwere melanized but maintained a rigidity similar tothat of the unaffected lobster shell

We sampled the diseased lobsters approximatelyevery 10 d over the course of the experiment for atotal of 4 time points per lobster Shell microbiomesamples were collected using sterile toothpicks toscrape the bacterial community from the desired areaof each lobster shell Gentle contact was made be -tween the toothpick and the lobster and this contactwas maintained over the entire sampling area Eachtoothpick was then stored in an individual cryovialand transported to the laboratory on dry ice where itwas transferred to a minus20degC freezer During each sam-pling shell scrapings were taken from the center ofthe lesion or spot on the left claw (referred to as 0 cm)as well as at a distance of 05 1 and 15 cm laterallyat an angle away from the dactylus (Fig 1) Sampleswere also taken from each lobster pen at a point justbelow the waterline while the pen was out of thewater on all 4 sampling dates for a total of 100 sam-ples (5 samples per lobster 5 lobsters 4 time points)We kept photographic records of each lobster through-out the course of the experiment and used these pho-tographs to guide our sampling and ensure that sim-ilar shell locations were sampled on each visit Some

of the lobsters molted during the course of the exper-iment but we continued to sample the same locationsas described above after molting

In addition to these samples we also collected shellbacterial community samples from 5 lobsters thatappeared to be free of disease These additional lob-sters were from the same age group and housed inthe same conditions but had all molted approximately1 mo before the initiation of sampling and showed novisible signs of LMSD We sampled these lobsters onthe same dates as the diseased lobsters but only tooka single bacterial community sample from the leftclaw of each lobster These lobsters served as a rep-resentation of what a healthy bacterial communitymight look like on a disease-free lobster so lobstersfrom this group that developed signs of LMSD weresubsequently eliminated from the study

DNA extraction and sequencing

DNA was extracted from toothpicks using the Pow-erSoilreg DNA Extraction Kit (MOBIO) following themanufacturerrsquos instructions Kit negative controlsextractions containing no toothpick were performedperiodically and subsequently amplified to ensurethat there was no kit contamination in the extracts(Salter et al 2014) Additionally DNA was extracteddirectly from a sterile toothpick to ensure there wasno amplification from the toothpick alone DNA ex -tracts from lobster and pen samples were amplifiedin triplicate using bacterial specific (515F 5rsquo-GTGCCA GCM GCC GCG GTA A-3rsquo 806R 5rsquo-GGA CTACHV GGG TWT CTA AT-3rsquo) uniquely barcoded 16SrRNA primers containing adaptors for Il lumina se -quencing (Caporaso et al 2012) Each 25 microl PCRreaction contained 10 microl 5-Prime Hot Master Mix(5 Primereg Thermo Fisher Scientific) and 02 microM ofeach primer In addition 05 to 2 microl of DNA was ad -ded to each reaction depending on the original DNA

43

Fig 1 Main sampling scheme employed in this study Shellscrapings were taken from (S) center of lesion or spot on theleft claw (referred to as 0 cm) and (S) 05 cm (S) 10 cm

and (S) 15 cm from the lesion or spot

Dis Aquat Org 124 41ndash54 2017

concentration PCR products were purified using theQIAquickreg PCR Purification Kit (QIAGEN) followingthe manufacturerrsquos instructions and quantified usinga Qubitreg 20 Fluorometer (Life Technologies ThermoScientific) PCR products were pooled in equimolaramounts and the quantity of the pool was checkedusing the KAPA Library Quantification Kit for IlluminaSequencing Platforms (Kapa Biosystems) The poolwas then loaded onto a MiSeq sequencer (Illumina)and sequenced using a paired-end V2 300 cycle kit

Sequence analysis

Quality filtering and sequence analysis was per-formed in QIIME using default parameters unlessnoted (Caporaso et al 2010) Paired-end reads werejoined specifying at least 10 base pairs in overlap andsequences within a Phred score of 30 or higher wereretained (indicating 999 accuracy in base calling)Chimeric sequences were identified using usearch61an algorithm that performs both de novo and refer-ence-based chimera detection with the GenomesOnLine Database (GOLD) as a reference and chimeraswere subsequently removed (Edgar et al 2011Reddy et al 2015) After these quality-filtering steps2 726 677 sequences remained Operational taxonomicunits (OTUs) were assigned at 97 sequence iden-tity using Swarm (Maheacute et al 2014) OTUs identifiedas Archaea or chloroplasts or that were unassignedat the kingdom level using the Ribosomal DatabaseProject (RDP) classifier (Wang et al 2007) were re -moved from downstream analysis Samples were rar-ified to the lowest sequencing depth 4735 OTUs persample for analysis of community composition andalpha diversity

Statistical analysis

The weighted UniFrac metric was used for analysisof community composition because it considers bothphylogenetic distance and abundance (Lozupone etal 2011) We visualized differences in bacterial com-munity structure among 05 cm increments on eachlobster using principal coordinates analysis and testedfor differences between these communities using per-mutational multivariate analysis of variance (PERM-ANOVA Anderson 2001) with individual lobster as ablocking component To test for pairwise differencesin each 05 cm increment we compared logit-trans-formed weighted UniFrac similarities of each 05 cmcomparison using a Welchrsquos t-test which accounts for

unequal variance again using individual lobster as ablocking component (R Core Team 2014) We alsoused QIIME to calculate 2 alpha diversity metricsphylogenetic distance a richness estimator that takesphylogeny into account (Faith amp Baker 2006) andShannon diversity Significance testing of alpha diver-sity measures was done using analysis of variance onlinear models that included the individual lobstersampled as a random effect

We defined the core microbiome as OTUs thatwere in n minus 1 samples within a given shell conditionThis resulted in 76 OTUs that were considered to belsquocorersquo to 1 or more shell condition We then identifiedrepresentative sequences from each of these OTUsusing the NCBI Basic Local Alignment Search Tool(BLAST) All sequence information produced fromthis study is available from the Sequence ReadArchive under accession no PRJNA307069

RESULTS

Sequence analysis of the 100 bacterial samplestaken from 5 diseased lobsters and their pensyielded 2 726 677 sequences corresponding to 18 204unique OTUs defined at 97 sequence identityAnalysis of samples taken from the left claw of lob-sters affected by LMSD showed that bacterial com-munity composition at the site of disease (distance0 cm) clustered separately from most other samples(Fig 2A) indicating that lobster shell at the siteof disease harbors a distinct bacterial community(PERMANOVA F = 2277 p lt 0001 df = 62) Thebacterial community found on lobster shell 05 cmaway from the site of LMSD (Fig 2A) was betweenthe 0 cm community and the 1 and 15 cm communi-ties and distinct from each (F = 713 p lt 0001 df =31 and F = 618 p lt 0001 df = 46) indicating thatthis may be a transitionary bacterial communityintermediate between affected and unaffected lob-ster shell Bacterial communities from lobster shell 1and 15 cm away from the site of the LMSD clus-tered together and more tightly than the bacterialcommunities from the other 2 sets of samples indi-cating that these communities are more similar(Fig 2A) and may represent an un affected bacterialcommunity distinct from affected and transitionarycommunities (F = 1700 p lt 0001 df = 62) Impor-tantly we did not see a change in bacterial commu-nity composition driven by sampling date (Fig S1 inthe Supplement at www int-res com articles suppld124p041 _ supp pdf) indicating that distance fromthe site of disease was more important than temporal

44

Feinman et al Bacterial communities in lobster shell disease

ilar to those described in prior shell disease re search(Chistoserdov et al 2012 Quinn et al 2012b Tlustyamp Metzler 2012 Whitten et al 2014) Briefly all lob-sters were kept in individual 9 cm diameter contain-ers (hereafter referred to as pens) and fed a diet con-sisting of Economac-4trade (Aqua fauna Bio-Marine) acommercially available shrimp feed This diet lacksthe carotenoid astaxanthin causing the lobster shellto have no color a desirable phenotype for this studyas it allows early detection and better visualization ofthe melanization associated with LMSD (Tlusty ampHyland 2005)

The diseased lobsters used in this study were ap -proximately 2 yr old and had all molted approxi-mately 5 to 6 mo before the initiation of sampling withthe exception of a single lobster that molted approxi-mately 3 mo prior to the initiation of sampling Theselobsters were chosen because they all had visiblesigns of LMSD on their left claw though the degreeof melanization varied among lobsters AlthoughLMSD was also present on other parts of the lobstersrsquoshells we focused solely on the left claw of each lob-ster in order to eliminate any potential confoundingdifferences that could arise from comparing differentparts of the lobster shell (Meres et al 2012) We cate-gorized the degree of melanization on each lobsterclaw defining lesions as melanized areas of the shellthat had become soft or pitted and spots as areas thatwere melanized but maintained a rigidity similar tothat of the unaffected lobster shell

We sampled the diseased lobsters approximatelyevery 10 d over the course of the experiment for atotal of 4 time points per lobster Shell microbiomesamples were collected using sterile toothpicks toscrape the bacterial community from the desired areaof each lobster shell Gentle contact was made be -tween the toothpick and the lobster and this contactwas maintained over the entire sampling area Eachtoothpick was then stored in an individual cryovialand transported to the laboratory on dry ice where itwas transferred to a minus20degC freezer During each sam-pling shell scrapings were taken from the center ofthe lesion or spot on the left claw (referred to as 0 cm)as well as at a distance of 05 1 and 15 cm laterallyat an angle away from the dactylus (Fig 1) Sampleswere also taken from each lobster pen at a point justbelow the waterline while the pen was out of thewater on all 4 sampling dates for a total of 100 sam-ples (5 samples per lobster 5 lobsters 4 time points)We kept photographic records of each lobster through-out the course of the experiment and used these pho-tographs to guide our sampling and ensure that sim-ilar shell locations were sampled on each visit Some

of the lobsters molted during the course of the exper-iment but we continued to sample the same locationsas described above after molting

In addition to these samples we also collected shellbacterial community samples from 5 lobsters thatappeared to be free of disease These additional lob-sters were from the same age group and housed inthe same conditions but had all molted approximately1 mo before the initiation of sampling and showed novisible signs of LMSD We sampled these lobsters onthe same dates as the diseased lobsters but only tooka single bacterial community sample from the leftclaw of each lobster These lobsters served as a rep-resentation of what a healthy bacterial communitymight look like on a disease-free lobster so lobstersfrom this group that developed signs of LMSD weresubsequently eliminated from the study

DNA extraction and sequencing

DNA was extracted from toothpicks using the Pow-erSoilreg DNA Extraction Kit (MOBIO) following themanufacturerrsquos instructions Kit negative controlsextractions containing no toothpick were performedperiodically and subsequently amplified to ensurethat there was no kit contamination in the extracts(Salter et al 2014) Additionally DNA was extracteddirectly from a sterile toothpick to ensure there wasno amplification from the toothpick alone DNA ex -tracts from lobster and pen samples were amplifiedin triplicate using bacterial specific (515F 5rsquo-GTGCCA GCM GCC GCG GTA A-3rsquo 806R 5rsquo-GGA CTACHV GGG TWT CTA AT-3rsquo) uniquely barcoded 16SrRNA primers containing adaptors for Il lumina se -quencing (Caporaso et al 2012) Each 25 microl PCRreaction contained 10 microl 5-Prime Hot Master Mix(5 Primereg Thermo Fisher Scientific) and 02 microM ofeach primer In addition 05 to 2 microl of DNA was ad -ded to each reaction depending on the original DNA

43

Fig 1 Main sampling scheme employed in this study Shellscrapings were taken from (S) center of lesion or spot on theleft claw (referred to as 0 cm) and (S) 05 cm (S) 10 cm

and (S) 15 cm from the lesion or spot

Dis Aquat Org 124 41ndash54 2017

concentration PCR products were purified using theQIAquickreg PCR Purification Kit (QIAGEN) followingthe manufacturerrsquos instructions and quantified usinga Qubitreg 20 Fluorometer (Life Technologies ThermoScientific) PCR products were pooled in equimolaramounts and the quantity of the pool was checkedusing the KAPA Library Quantification Kit for IlluminaSequencing Platforms (Kapa Biosystems) The poolwas then loaded onto a MiSeq sequencer (Illumina)and sequenced using a paired-end V2 300 cycle kit

Sequence analysis

Quality filtering and sequence analysis was per-formed in QIIME using default parameters unlessnoted (Caporaso et al 2010) Paired-end reads werejoined specifying at least 10 base pairs in overlap andsequences within a Phred score of 30 or higher wereretained (indicating 999 accuracy in base calling)Chimeric sequences were identified using usearch61an algorithm that performs both de novo and refer-ence-based chimera detection with the GenomesOnLine Database (GOLD) as a reference and chimeraswere subsequently removed (Edgar et al 2011Reddy et al 2015) After these quality-filtering steps2 726 677 sequences remained Operational taxonomicunits (OTUs) were assigned at 97 sequence iden-tity using Swarm (Maheacute et al 2014) OTUs identifiedas Archaea or chloroplasts or that were unassignedat the kingdom level using the Ribosomal DatabaseProject (RDP) classifier (Wang et al 2007) were re -moved from downstream analysis Samples were rar-ified to the lowest sequencing depth 4735 OTUs persample for analysis of community composition andalpha diversity

Statistical analysis

The weighted UniFrac metric was used for analysisof community composition because it considers bothphylogenetic distance and abundance (Lozupone etal 2011) We visualized differences in bacterial com-munity structure among 05 cm increments on eachlobster using principal coordinates analysis and testedfor differences between these communities using per-mutational multivariate analysis of variance (PERM-ANOVA Anderson 2001) with individual lobster as ablocking component To test for pairwise differencesin each 05 cm increment we compared logit-trans-formed weighted UniFrac similarities of each 05 cmcomparison using a Welchrsquos t-test which accounts for

unequal variance again using individual lobster as ablocking component (R Core Team 2014) We alsoused QIIME to calculate 2 alpha diversity metricsphylogenetic distance a richness estimator that takesphylogeny into account (Faith amp Baker 2006) andShannon diversity Significance testing of alpha diver-sity measures was done using analysis of variance onlinear models that included the individual lobstersampled as a random effect

We defined the core microbiome as OTUs thatwere in n minus 1 samples within a given shell conditionThis resulted in 76 OTUs that were considered to belsquocorersquo to 1 or more shell condition We then identifiedrepresentative sequences from each of these OTUsusing the NCBI Basic Local Alignment Search Tool(BLAST) All sequence information produced fromthis study is available from the Sequence ReadArchive under accession no PRJNA307069

RESULTS

Sequence analysis of the 100 bacterial samplestaken from 5 diseased lobsters and their pensyielded 2 726 677 sequences corresponding to 18 204unique OTUs defined at 97 sequence identityAnalysis of samples taken from the left claw of lob-sters affected by LMSD showed that bacterial com-munity composition at the site of disease (distance0 cm) clustered separately from most other samples(Fig 2A) indicating that lobster shell at the siteof disease harbors a distinct bacterial community(PERMANOVA F = 2277 p lt 0001 df = 62) Thebacterial community found on lobster shell 05 cmaway from the site of LMSD (Fig 2A) was betweenthe 0 cm community and the 1 and 15 cm communi-ties and distinct from each (F = 713 p lt 0001 df =31 and F = 618 p lt 0001 df = 46) indicating thatthis may be a transitionary bacterial communityintermediate between affected and unaffected lob-ster shell Bacterial communities from lobster shell 1and 15 cm away from the site of the LMSD clus-tered together and more tightly than the bacterialcommunities from the other 2 sets of samples indi-cating that these communities are more similar(Fig 2A) and may represent an un affected bacterialcommunity distinct from affected and transitionarycommunities (F = 1700 p lt 0001 df = 62) Impor-tantly we did not see a change in bacterial commu-nity composition driven by sampling date (Fig S1 inthe Supplement at www int-res com articles suppld124p041 _ supp pdf) indicating that distance fromthe site of disease was more important than temporal

44

Dis Aquat Org 124 41ndash54 2017

concentration PCR products were purified using theQIAquickreg PCR Purification Kit (QIAGEN) followingthe manufacturerrsquos instructions and quantified usinga Qubitreg 20 Fluorometer (Life Technologies ThermoScientific) PCR products were pooled in equimolaramounts and the quantity of the pool was checkedusing the KAPA Library Quantification Kit for IlluminaSequencing Platforms (Kapa Biosystems) The poolwas then loaded onto a MiSeq sequencer (Illumina)and sequenced using a paired-end V2 300 cycle kit

Sequence analysis

Quality filtering and sequence analysis was per-formed in QIIME using default parameters unlessnoted (Caporaso et al 2010) Paired-end reads werejoined specifying at least 10 base pairs in overlap andsequences within a Phred score of 30 or higher wereretained (indicating 999 accuracy in base calling)Chimeric sequences were identified using usearch61an algorithm that performs both de novo and refer-ence-based chimera detection with the GenomesOnLine Database (GOLD) as a reference and chimeraswere subsequently removed (Edgar et al 2011Reddy et al 2015) After these quality-filtering steps2 726 677 sequences remained Operational taxonomicunits (OTUs) were assigned at 97 sequence iden-tity using Swarm (Maheacute et al 2014) OTUs identifiedas Archaea or chloroplasts or that were unassignedat the kingdom level using the Ribosomal DatabaseProject (RDP) classifier (Wang et al 2007) were re -moved from downstream analysis Samples were rar-ified to the lowest sequencing depth 4735 OTUs persample for analysis of community composition andalpha diversity

Statistical analysis

The weighted UniFrac metric was used for analysisof community composition because it considers bothphylogenetic distance and abundance (Lozupone etal 2011) We visualized differences in bacterial com-munity structure among 05 cm increments on eachlobster using principal coordinates analysis and testedfor differences between these communities using per-mutational multivariate analysis of variance (PERM-ANOVA Anderson 2001) with individual lobster as ablocking component To test for pairwise differencesin each 05 cm increment we compared logit-trans-formed weighted UniFrac similarities of each 05 cmcomparison using a Welchrsquos t-test which accounts for

unequal variance again using individual lobster as ablocking component (R Core Team 2014) We alsoused QIIME to calculate 2 alpha diversity metricsphylogenetic distance a richness estimator that takesphylogeny into account (Faith amp Baker 2006) andShannon diversity Significance testing of alpha diver-sity measures was done using analysis of variance onlinear models that included the individual lobstersampled as a random effect

We defined the core microbiome as OTUs thatwere in n minus 1 samples within a given shell conditionThis resulted in 76 OTUs that were considered to belsquocorersquo to 1 or more shell condition We then identifiedrepresentative sequences from each of these OTUsusing the NCBI Basic Local Alignment Search Tool(BLAST) All sequence information produced fromthis study is available from the Sequence ReadArchive under accession no PRJNA307069

RESULTS

Sequence analysis of the 100 bacterial samplestaken from 5 diseased lobsters and their pensyielded 2 726 677 sequences corresponding to 18 204unique OTUs defined at 97 sequence identityAnalysis of samples taken from the left claw of lob-sters affected by LMSD showed that bacterial com-munity composition at the site of disease (distance0 cm) clustered separately from most other samples(Fig 2A) indicating that lobster shell at the siteof disease harbors a distinct bacterial community(PERMANOVA F = 2277 p lt 0001 df = 62) Thebacterial community found on lobster shell 05 cmaway from the site of LMSD (Fig 2A) was betweenthe 0 cm community and the 1 and 15 cm communi-ties and distinct from each (F = 713 p lt 0001 df =31 and F = 618 p lt 0001 df = 46) indicating thatthis may be a transitionary bacterial communityintermediate between affected and unaffected lob-ster shell Bacterial communities from lobster shell 1and 15 cm away from the site of the LMSD clus-tered together and more tightly than the bacterialcommunities from the other 2 sets of samples indi-cating that these communities are more similar(Fig 2A) and may represent an un affected bacterialcommunity distinct from affected and transitionarycommunities (F = 1700 p lt 0001 df = 62) Impor-tantly we did not see a change in bacterial commu-nity composition driven by sampling date (Fig S1 inthe Supplement at www int-res com articles suppld124p041 _ supp pdf) indicating that distance fromthe site of disease was more important than temporal

44

Feinman et al Bacterial communities in lobster shell disease

patterns in determining bacterial community compo-sition in this study

Since the 1 and 15 cm distances appeared to rep-resent unaffected lobster shell we compared thebacterial community found at these locations to thebacterial community found on lobsters that did nothave LMSD including samples from the left claw oflobsters that showed no visible signs disease andsamples from 4 different locations on the left claw oflobsters that had recently molted (Fig 2B) Samplestaken 1 and 15 cm from the site of disease on dis-eased lobster shell clustered together with samplestaken from the left claw of lobsters that showed novisible signs of LMSD (Fig 2B lsquocleanrsquo lobsters com-pared to diseased lobsters F = 613 p gt 095 df = 41)indicating that locations 1 and 15 cm from the site ofLMSD are in fact representative of unaffected dis-ease-free shell

Bacterial communities recovered from the shell ofmolted lobsters (Fig 2B) showed no differences amonglocation on the shell (F = 064 p gt 040 df = 15) sug-gesting that in the area sampled there was no sys-tematic change in bacterial community compositionacross disease-free lobster shell We therefore usedmolted lobsters as a baseline for what might be ex -pected for community change among locations on aclean disease-free lobster claw and compared this to

the bacterial community change of each 05 cm incre-ment on the lobsters (Fig 3) We found that there wasless similarity between bacterial communities at 0and 05 cm (Fig 3) in lobsters with lesions (Welchrsquost-test t = minus389 p lt 0005 df = 7) and spots (t = minus358

45

Fig 2 Principal coordinates analysis of the weighted UniFrac metric comparing bacterial community composition of (A) dis-eased lobster shell and (B) unaffected lobster shell Diseased lobster shell includes samples collected from (red) the site ofLMSD as well as (blue) 05 cm (green) 1 cm and (purple) 15 cm away from the site of the disease Unaffected lobster shell in-cludes (F) samples from those same distances on lobsters affected by LMSD that had recently molted and were therefore con-sidered disease free (n) samples from the left claw of lobsters with no recorded history of disease and samples collected from

diseased lobsters (d) 1 and (d) 15 cm from the site of LMSD

Fig 3 Median weighted UniFrac similarity of each sequen-tial 05 cm increment on the lobster claw Midline indicatesthe median value box indicates the 25th and 75th quartilesand whiskers extend to the smallest or largest value ob-served within 15 times interquartile range beyond the box Values

not within that range are represented as outliers

Dis Aquat Org 124 41ndash54 2017

p lt 001 df = 8) than in molted lobsters indicatingthat the bacterial community at the site of diseasewas more different from adjacent sites than would beexpected based on molted lobsters that showed nosigns of the LMSD

The weighted UniFrac similarities found whencomparing bacterial community composition at 05and 1 cm (Fig 3) were statistically different on lob-sters that had a lesion at 0 cm compared to lobstersthat had molted (t = minus216 p lt 005 df = 16) but werenot statistically different at α = 005 for lobsters thathad a spot at 0 cm (t = minus210 p gt 006 df = 10)Despite this discrepancy on lobsters with a spot at0 cm we still consider the 05 cm distance a transi-tionary zone in LMSD First the average weightedUniFrac similarity among the 05 and 1 cm distanceson lobsters with a spot at 0 cm was lower than onmolted lobsters indicating that the 05 cm bacterialcommunity on lobsters with a spot at 0 cm was lesssimilar to adjacent sites than expected Second theresults of the Welchrsquos t-test were reasonably close(p = 0062) to being significant at an arbitrarilyselected α = 005 Finally PERMANOVA analysisshowed significant support for grouping the data into3 different groups 0 05 and 1 and 15 cm whencomparing all lobsters with a lesion at 0 cm (F = 1630p lt 0001 df = 30) and lobsters with a spot at 0 cm(F = 1306 p lt 0001 df = 30) We therefore definedthe 05 cm distance as a transitionary zone harboringa bacterial community that was differentiated fromthe unaffected lobster shell community before thevisual onset of LMSD

The transitionary zone was limited to the 05 cmdistance as the weighted UniFrac similarity on lob-sters that had lesions or spots on their left claw wasnot statistically different from the same comparisonon the molted lobster shell at a dis-tance of 1 and 15 cm (Fig 3 t = 035p gt 07 df = 6 and t = minus097 p gt 03df = 7) This indicates that the bacterialcommunity was not altered by LMSDat 1 cm away from the site of melaniza-tion (and presumably at more distantlocations) and was therefore represen-tative of what was found on unaffectedlobster shell at this point in the moltcycle Given these results we here-after differentiate 3 categories of shell0 cm affected shell 05 cm transitio -nary shell and 1 cm and 15 cm un -affected shell

We also examined if the bacterialcommunity within a spot could be con-

sidered a transitionary condition from unaffectedshell to the more advanced lesion shell The lack of adistinct separation in the principal coordinates analy-sis between lobsters with lesions at 0 cm and spots at0 cm (Fig 2A) suggests that the categories of lesionand spot may not be meaningful with regard to theassociated bacterial communities To test this wecompared the bacterial community composition oflobsters with lesions to those with spots on both af -fected and transitionary shell and found there wasno statistical evidence to support separating lesionbacterial communities from spot bacterial communi-ties (F = 183 p gt 01 df = 15 and F = 082 p gt 03df = 15) These data suggest that once there wasvisual evidence of LMSD there was little subsequentchange in bacterial community composition and wetherefore no longer distinguish be tween lesion andspot in our analysis

After classifying samples into 3 discrete shell con-dition groups (affected transitionary and unaf-fected) based on bacterial community compositionwe compared differences in diversity within eachgroup using the pen where each lobster was keptas a baseline for maximum bacterial diversity (Fig 4)Regardless of shell condition bacterial diversity wasconsistently lower on the lobster shell than in thepen Alpha diversity varied among shell conditionswith the lowest diversity observed on the affectedshell Bacterial diversity on the transitionary shellwas intermediate between the affected shell andunaffected shell indicating that the drop in bacterialdiversity began before the visual appearance ofLMSD These differences in diversity account forsome of the differences seen in the rank abun-dance curve of each shell condition (Fig S2 in the Supplement)

46

Fig 4 Analysis of alpha diversity associated with LMSD for 3 different shellconditions and the lobster pen Differences in diversity between shell condi-tions were significant at a p lt 0001 for each diversity metric as indicated byletters (lowercase letters are used for comparisons using phylogenetic distanceas a measure uppercase letters are used for comparisons using Shannon

diversity) Error bars represent standard deviation

Feinman et al Bacterial communities in lobster shell disease

We identified the bacteria that were important tochanges in diversity among different shell conditionsby analyzing the core microbiome We defined thecore microbiome as bacteria that were present in n minus1 samples from each shell condition because bacteriathat are truly important to causing a shift to a shellcondition should be present in nearly all samplesfrom that shell condition While this criterion does notcapture all of the most abundant bacteria present(see Fig S2 in the Supplement) it narrows the focusto those that are more likely to be contributing to theprogression of LMSD

There were 76 OTUs considered to be part ofthe core microbiome of 1 or more shell condition

(Fig 5) These 76 OTUs comprised a total of 49 ofthe se quences recovered from affected shell 38of the sequences recovered from transitionary shelland 36 of the sequences recovered from unaf-fected shell Of these OTUs 16 were consideredunique to the core microbiome of unaffected shell(Fig 5AD) yet they only constituted 5 of thesequences from this shell condition Converselythere were 14 OTUs that were unique to the coremicrobiome of affected shell accounting for 34 ofthe sequences from this shell condition (Fig 5AB)This stark contrast is due in part to the significantlylower diversity in the af fected shell compared to theunaffected shell (Fig 4) From the affected core

47

Fig 5 (A) Core microbiome of each shell condition with (BminusG) accompanying bar charts of each operational taxonomic unit(OTU) in that core microbiome and its relative abundance within each shell condition (title text color matches the portion of (A)corresponding to the bar chart) Please note the difference in scale for each bar plot Individual OTUs are designated by theirOTU number along the x-axis Error bars represent standard deviation Numbers within the Venn diagram indicate the number of OTUs that belong to each core microbiome while italics indicate the percent those OTUs account for in each shell

condition

Dis Aquat Org 124 41ndash54 2017

microbiome we identified OTU 0 from the candi-date genus Candidatus Thio globus and OTU 18genus Aquimarina as important to LMSD based ontheir proportional abundances (combined 27Fig 5B Table 1)

There were 13 OTUs that were unique to the tran-sitionary core microbiome accounting for 5 of thesequences recovered from this part of the shell(Fig 5AC) The presence of OTUs that are unique tothe transitionary core microbiome indicates thatthere are members of the transitionary bacterial com-munity beyond what might be expected from a sim-ple mixing of the microbiota from healthy and dis-eased lobster shell OTUs 44 78 and 160 from thegenera Marinobacter Loktanella and CandidatusEndobugula stood out as important members of thetransitionary core microbiome due to their abun-dance in both the affected and transitionary shellconditions (Fig 5C Table 1)

There were also 33 OTUs found in the core micro-biome of more than 1 shell condition Of these 33OTUs 19 were shared between the transitionary andunaffected shell (Fig 5AG) whereas only 4 OTUs

were shared between the transitionary and affectedshell (Fig 5AE) This low number of OTUs sharedbetween the core microbiomes of transitionary andaffected shell provides further evidence that the bac-teria associated with this transition zone are notmerely a result of the outward expansion of the bac-teria inhabiting affected shell There were no OTUsshared between the affected and unaffected coremicrobiome that were not also a part of the transi-tionary core microbiome which also highlights theimportance of the bacterial community found 05 cmfrom the site of LMSD as a transition between these 2shell conditions From the affected transitionary coremicrobiome overlap we identified OTU 47 from thegenus Colwellia and OTU 118 from the genus Areni-cella to be of interest based on their relative abun-dances (Fig 5E Table 1) Finally there were 10 ubi -quitous OTUs shared among the core microbiomes ofall 3 shell conditions (Fig 5AF) OTU 1 from thegenus Perspicuibacter and OTU 19 from the genusLutimonas were of particular importance due to theirabundances on the transitionary and affected shellrespectively (Fig 5F Table 1)

48

OTU Order Family Genus Identity Abundance

Affected Core0 Thioglobusa 91 150 18 Flavobacteriales Flavobacteriaceae Aquimarina 100 118 21 Flavobacteriales Flavobacteriaceae Maritimimonas 96 1129 Alteromonadales Colwelliaceae Colwellia 99 2348 Alteromonadales Colwelliaceae Colwellia 100 0375 Rhodobacterales Rhodobacteraceae Loktanella 99 06121 Rhodobacterales Hyphomonadaceae Robiginitomaculum 96 04135 Rhizobiales Methylobacteriaceae Methylobacterium 100 04159 Oceanospirillales Oceanospirillaceae Oleispira 100 01229 Rhodobacterales Rhodobacteraceae Octadecabacter 100 04238 Rhizobiales Hyphomicrobiaceae Maritalea 99 07275 Flavobacteriales Flavobacteriaceae Cloacibacterium 100 01316 Oceanospirillales Oceanospirillaceae Neptunomonas 100 041306 Flavobacteriales Flavobacteriaceae Lutimonas 98 01

Transitionary Core2 Arenicellales Arenicellaceae Perspicuibacter 99 0739 Arenicellales Arenicellaceae Arenicella 98 0344 Alteromonadales Alteromonadaceae Marinobacter 100 1553 Rhodobacterales Rhodobacteraceae Roseovarius 100 0178 Rhodobacterales Rhodobacteraceae Loktanella 100 0189 Rhizobiales Phyllobacteriaceae Pseudahrensia 98 01107 Rhodobacterales Hyphomonadaceae Fretibacter 96 02160 Cellvibrionales Cellvibrionaceae Endobugulaa 99 10254 Oceanospirillales Alcanivoracaceae Marinicella 99 04338 Chromatiales Thioalkalispiraceae Thioprofundum 94 01

Table 1 Taxonomy for core operational taxonomic units (OTUs) identified in this study Closest cultured representative for allOTUs listed in order from Fig 5 Bold OTUs are those shown in Fig 6 Abundance is the relative abundance in that shell

condition Note OTU 0 has not yet been classified at the order or family level

(Table continued on next page)

Feinman et al Bacterial communities in lobster shell disease 49

Table 1 (continued)

OTU Order Family Genus Identity Abundance

370 Burkholderiales Comamonadaceae Diaphorobacter 100 01387 Rhodothermales Rubricoccaceae Rubrivirga 98 01410 Rhizobiales Phyllobacteriaceae Aminobacter 96 01

Unaffected Core6 Rhodobacterales Rhodobacteraceae Amylibacter 99 047 Rhodobacterales Rhodobacteraceae Halocynthiibacter 99 0123 Thiotrichales Thiotrichaceae Cocleimonas 96 0688 Rhizobiales Phyllobacteriaceae Pseudahrensia 99 0193 Arenicellales Arenicellaceae Arenicella 97 05110 Rhodobacterales Hyphomonadaceae Fretibacter 96 05128 Chromatiales Woeseiaceae Woeseia 96 03177 Acidimicrobiales Acidimicrobiaceae Ilumatobacter 100 02191 Sphingobacteriales Saprospiraceae Lewinella 92 05220 Acidimicrobiales Microthrixaceae Microthrixa 90 03233 Sphingobacteriales Saprospiraceae Lewinella 91 04277 Deferribacterales Deferribacteraceae Caldithrix 91 03294 Nitrosomonadales Nitrosomonadaceae Nitrosomonas 99 02302 Incertaesedis Rhodothermaceae Rhodothermus 89 03339 Rhodobacterales Rhodobacteraceae Labrenzia 97 02360 Rhizobiales Phyllobacteriaceae Aminobacter 95 02

Affected and Transitionary Core47 Alteromonadales Colwelliaceae Colwellia 99 3373 Oceanospirillales Oceanospirillaceae Bacterioplanes 91 0377 Rhodobacterales Rhodobacteraceae Loktanella 100 04118 Arenicellales Arenicellaceae Arenicella 99 17

Ubiquitous Core1 Arenicellales Arenicellaceae Perspicuibacter 99 278 Rhodobacterales Rhodobacteraceae Amylibacter 98 029 Rhodobacterales Rhodobacteraceae Litoreibacter 100 0210 Rhodobacterales Rhodobacteraceae Jannaschia 100 0219 Flavobacteriales Flavobacteriaceae Lutimonas 96 2538 Arenicellales Arenicellaceae Arenicella 99 1652 Rhodobacterales Rhodobacteraceae Sulfitobacter 100 0371 Flavobacteriales Flavobacteriaceae Maribacter 99 0698 Rhodobacterales Hyphomonadaceae Litorimonas 97 03268 Rhizobiales Methylobacteriaceae Methylobacterium 100 02

Transitionary and Unaffected Core20 Flavobacteriales Flavobacteriaceae Lutimonas 97 0322 Thiotrichales Thiotrichaceae Cocleimonas 97 0857 Rhodobacterales Rhodobacteraceae Phaeobacter 99 0174 Verrucomicrobiales Rubritaleaceae Rubritalea 97 0682 Acidimicrobiales Microthrixaceae Microthrixa 91 0287 Rhodobacterales Rhodobacteraceae Loktanella 98 02106 Cellvibrionales Cellvibrionaceae Endobugulaa 97 10125 Planctomycetales Planctomycetaceae Blastopirellula 91 03131 Arenicellales Arenicellaceae Perspicuibacter 93 02132 Rhodobacterales Rhodobacteraceae Maribius 95 02137 Rhizobiales Hyphomicrobiaceae Hyphomicrobium 98 02140 Rhodobacterales Hyphomonadaceae Fretibacter 95 02147 Rhizobiales Hyphomicrobiaceae Filomicrobium 96 03149 Flavobacteriales Flavobacteriaceae Lutimonas 97 01150 Rhodobacterales Rhodobacteraceae Oceaniovalibus 94 02161 Sphingobacteriales Saprospiraceae Phaeodactylibacter 89 02164 Rhodobacterales Rhodobacteraceae Citreicella 93 01174 Chromatiales Thioalkalispiraceae Thioprofundum 93 02262 Chromatiales Granulosicoccaceae Granulosicoccus 97 01aCandidate classification

Dis Aquat Org 124 41ndash54 2017

DISCUSSION

High-throughput sequencing results from lobsterswith and without visible signs of LMSD (ie lobsterswith spots lobsters with lesions and lobsters that hadrecently molted and did not exhibit signs of LMSD)indicate the presence of 3 distinct bacterial commu-nities (Fig 6A) We classified these 3 distinct commu-nities as affected transitionary and unaffected shellcommunities based on weighted UniFrac analysis ofcommunity composition (Figs 2 amp 3) The presence ofa distinct bacterial community at the site of diseaseagrees with prior work indicating that bacteria arethe causative agent of shell disease (Chistoserdov etal 2012) However our small spatial scale samplingallowed a more comprehensive characterization of thebacterial community associated with LMSD enabling

us to separate out the bacterial communities associ-ated with the transition from unaffected to affectedshell and thus identifying the presence of a third bac-terial community on shells affected with LMSD thetransitionary community

Interestingly our results indicate no differentiationamong bacterial communities found on lobsters withlesions and spots in either affected or transitionaryshell This finding is counter to previously reportedgenetic fingerprinting results that showed differ-ences in the bacterial communities among lesionsand spots (Quinn et al 2012b) Based on their initialfindings Quinn et al (2012b) suggested that the bac-teria present in a spot may be important in initiatingshell disease in lobsters whereas bacteria found inlesions were secondary opportunistic colonizers thatthrived in the new conditions The lack of differenti-

50

Fig 6 Conceptual diagram of the laboratory model of shell disease (LMSD) Visual summary of the findings of this study show-ing (A) the main observations of this study and (BC) 2 different alternate models of these data though others are possible (A1)The 3 distinct bacterial communities present on diseased lobster shell unaffected transitionary and affected shell communi-ties highlighting the lack of difference between spot and lesion communities The post-molt community is fairly similar to theunaffected shell community though this community devoid of disease is fairly uniform across the sampling area (A2) The ob-served decrease in bacterial diversity as LMSD progresses (A3) Several operational taxonomic units (OTUs) that are of inter-est to LMSD based on their abundances in either the transitionary or affected shell communities (B) Alternate Model 1 Thetransitionary bacterial community though unique is unimportant to the progression of LMSD and simply represents a community that is able to live in the proximity of disease-causing bacteria (C) Alternate Model 2 The transitionary bacterial

community consists of precursory bacteria that are present before visual signs of the disease occur

Feinman et al Bacterial communities in lobster shell disease

ation between lesion and spot communities in ourdata suggests that this may not be the case The tran-sitionary shell bacterial community that we identi-fied however may be important for initiation ofLMSD while the affected shell bacterial communitycould represent secondary colonizers

In addition to identifying 3 distinct bacterial com-munities we observed a positive association be tweenbacterial diversity and distance from the site of LMSD(Figs 4 amp 6A) This is contrary to work done by Whit-ten et al (2014) which showed greater bacterialdiversity in diseased shell compared to healthy shellThese previous results however were based onmethods that recovered an average of 8 to 11 taxa persample The sequencing methods we used allowedfor a much greater depth which revealed a greaterproportion of low abundance taxa thereby increas-ing apparent diversity The reduction in bacteria atthe site of LMSD indicates that the diseased shellenvironment may be highly selective for specific bac-terial taxa perhaps due to the host response and theantibacterial effects of melanin Alternatively op -portunistic bacteria associated with the newly ex -posed chitin matrix may outcompete other bacteriathereby excluding them Since bacterial diversitywas also reduced on the transitionary shell comparedto the unaffected shell we suggest that competitionrather than environmental factors lowered bacterialdiversity at the site of LMSD It is also possible thatalong with changes in the bacterial communitychanges occurred in the physical environment of thetransitionary shell prior to the onset of disease thathave yet to be identified

The presence of a transitionary community sug-gests that it is not just the bacteria at the site of dis-ease that are altered during LMSD but that there isalso a change in the bacterial community composi-tion at distances up to 05 cm away from the site ofmelanization The transitionary community couldrepresent several possibilities the 2 most parsimo-nious being (1) an outer edge of the disease zone thatis fouled by disease causing bacteria or (2) a novelcommunity that is a precursor for pathogenic bacteriathat later infect the lobster shell These 2 alternatemodels for disease transmission depicted in Fig 6BCprovide possible explanations for the changes in bothbacterial diversity and community structure we ob -served over space and time In Alternate Model 1(Fig 6B) the transitionary and affected bacterialcommunities both appear when spots are first ob -servable on the lobster shell Here the transitionarycommunity is not involved as a causative agent ofLMSD but instead is present due to fouling from the

disease-causing community and changes in the shellenvironment that occur due to the presence of dis-ease-causing bacteria As LMSD progresses theaffected area of the shell increases (transition fromspot to lesion) until the lobster eventually molts Thismodel is in agreement with previous reports regard-ing the progression of shell disease (Chistoserdov etal 2012 Davies et al 2014 Whitten et al 2014)though the sampling schemes in these previous stud-ies did not allow for the detection of a transitionarycommunity Alternate Model 2 (Fig 6C) interpretsthe transitionary bacterial community as a precursorcommunity arriving prior to the onset of visual signsof LMSD In this model the transitionary communityrepresents initial colonizers that may cause a shift inthe lobster shell bacterial community and lower bac-terial diversity prior to visual signs of disease Thisinitial invasion creates an environment that allowsfor secondary colonization that advances the diseasestate at which time the melanization associated withshell disease becomes observable The fact that thebacterial community at 05 cm is distinct from thecommunity at 0 cm (Figs 2 amp 3) and that there arebacteria that are unique to the core microbiome at05 cm (Fig 5) offers support for this second modelThis model hinges on the existence of a currentlyunobserved state where the transitionary and un -affected communities are present prior to visual signsof disease This state will be difficult to observe infuture studies as discovery would only come fromresearchers predicting the exact location where shelldisease might occur and sampling that predictedlocation prior to visual signs of disease While both ofthese alternate models differ from the currently pro-posed model for the progression of epizootic shelldisease both are in agreement with the previouslyproposed polymicrobial nature of shell disease thatis either caused by or amplified by a dysbiosis in bacterial communities on unaffected shell (Meres etal 2012) Culture studies examining growth dynamicsof potential pathogens in the presence and absenceof members of the transitionary community couldoffer additional insights into the location and initia-tion of LMSD and could refine the models proposedhere

Given the potential importance of the transitionaryand affected shell communities to the initiation andprogression of LMSD we identified key bacterialmembers of these shell conditions from the coremicrobiome (Figs 5 amp 6A) and evaluated them interms of the 2 hypothesized models The affectedshell was dominated by just 2 OTUs 0 and 18 whichaccounted for nearly 27 of the sequences from this

51

Dis Aquat Org 124 41ndash54 2017

shell condition (Fig 5B) Given their proportionaldominance these taxa appeared to be very importantto the affected shell community The closest culturedrepresentative to OTU 0 was from Candidatus Thio -globus singularis a mixotrophic bacterium belong-ing to the SUP05 clade known for its potential to oxi-dize sulfur and fix carbon in the dark ocean and inlow oxygen zones (Marshall amp Morris 2015) This cul-tured representative however only shared 91sequence identity with OTU 0 Therefore this OTUlikely represents a new uncultured species As thiswas the most abundant OTU from both the affectedand transitionary shell (see Table S1 in the Supple-ment) better identification and classification of thisbacterium is important for understanding its role inthe progression of LMSD OTU 18 the second mostabundant OTU on the affected shell (see Table S1)shared 100 sequence identity with a representa-tive from the genus Aquimarina (Esteves et al 2013)and 99 sequence identity with bacteria re coveredfrom lobster shells after having been exposed toAquimarina lsquohomariarsquo (Quinn et al 2012a) Aquima-rina lsquohomariarsquo has been identified by several studiesas a potential causative agent of epizootic shell dis-ease as it can readily be recovered from lesions(Chistoserdov et al 2005 2012 Quinn et al 2012aWhitten et al 2014) The abundance of OTU 18 onthe affected shell in this study (12 Fig 5B) con-firms the potential importance of Aquimarina lsquohomariarsquoto LMSD though its lack of dominance on the transi-tionary shell (2 Fig 5B) suggests that it is not assuccessful outside of melanized areas of the shell (seeTable S1)

In the transitionary core microbiome (Fig 5C)OTUs 44 and 160 appeared to be particularly impor-tant OTU 44 was one of the most abundant OTUs inall shell conditions however it was only consideredcore to the transitionary shell (Table 1 see Table S1)This OTU was most closely related to Marinobactersalsuginis (100 sequence identity S V Law et alunpubl data accession no KT921331) Other speciesfrom this genus have proteins for the deacetylation ofchitin (Song et al 2013) which suggests that OTU 44may be important to initiation of disease creatingconditions under which opportunistic colonizers asso-ciated with LMSD might thrive (Malloy 1978 Chisto -serdov et al 2005) OTU 160 was most closely relatedto the candidate species Endobugula sertula (99sequence identity McGovern amp Hellberg 2003) abryozoan symbiont that produces polyketides to pro-tect the host larvae from predation (Lindquist amp Hay1996 Lopanik et al 2004) In addition to OTUs 44and 160 OTU 78 was part of the transitionary core

microbiome though quantitatively it appeared to bemore important to the affected shell Although thisOTU was not a member of the affected shell coremicrobiome it was present in 13 of the 16 affectedshell samples and was present at relatively highabundance in 7 of those samples This OTU genusLoktanella which shared 100 sequence identitywith bacteria recovered from other diseased lobstershells (Quinn et al 2012a) may therefore be in fectingthe shell alongside early colonizers and in some in -stances growing in abundance as LMSD progresses

Of the 4 OTUs shared between the affected andtransitionary core microbiomes (Fig 5E) OTUs 47and 118 were numerically dominant in both commu-nities OTU 47 belonged to the genus Colwelliawhich Meres et al (2012) reported as one of the manygroups of species that helped to differentiate dis-eased and healthy lobster shell In the current studythis bacterium was abundant on transitionary shellan area of the shell that would have previously re -mained unsampled indicating that members of thisgenus may be more important in promoting LMSDthan previously thought if the transitionary shellindeed reflects an initial invasion OTU 118 was mostclosely aligned with Arenicella chitinivorans (99sequence identity) a bacterium that differs from oth-ers in the same genus by its ability to degrade chitin(Nedashkovskaya et al 2013) Given that previouswork by Whitten et al (2014) also recovered mem-bers of this genus from lesioned areas of the lobstershell it appears that the ability to degrade chitin maybe an important characteristic allowing bacteria tothrive on both the affected and transitionary shell

Many studies have reported the incidence of ubiq-uitous bacteria on the lobster shell both on diseasedand healthy shell (Chistoserdov et al 2005 Meres etal 2012 Whitten et al 2014) In this study there were10 OTUs present in the core microbiome of all 3 shellconditions (Fig 5F) and OTUs 1 and 19 were propor-tionally dominant on the transitionary and affectedshell respectively though they were highly abun-dant in all 3 shell conditions (see Table S1) OTU 1 ismost likely Perspicuibacter marinus (99 sequenceidentity) an aerobic marine bacterium recently iso-lated from surface seawater in Japan (Teramoto et al2015) At this time not much is known about thisorganism but it has been cultured and thereforecould be used in future studies exploring the dynam-ics of the transitionary lobster shell community OTU19 an important shell bacterium found on affectedshell was most similar to other bacteria found inassociation with sponges (99 sequence identitySipkema et al 2011) It is likely that this bacterium is

52

Feinman et al Bacterial communities in lobster shell disease

from the genus Lutimonas (96 sequence identityBuerger et al 2012) which is part of the same familyas Aquimarina and may represent another oppor-tunistic colonizer of diseased lobster shell

In addition to these core microbiome memberswhich appear to be important due to their relativeabundance on either the transitionary or affectedshell there were a number of other core microbiomeOTUs that had identities similar to taxa identified inprevious studies of epizootic shell disease Severalstudies have reported Aquimarina lsquohomariarsquo as apotential causative agent in shell disease including abacterial challenge study that was able to recoverthis bacterium from lobsters with lesions (Quinn et al2012a) and a next-generation sequencing study thatexamined healthy and diseased shell (Meres et al2012) Although we only recovered a single OTU thatwas associated with the genus Aquimarina (OTU 18)we recovered several other OTUs including Lutimo -nas that were from the Flavobacteria family (Table 1)Several studies have reported that members of thisfamily aside from Aquimarina are important to epi-zootic shell disease (Chistoserdov et al 2012 Mereset al 2012 Quinn et al 2012b) In addition to Aqui -marina the lsquoThalassobiusrsquo genus has been associatedwith epizootic shell disease (Meres et al 2012 Quinnet al 2012a) Although we did not recover any bacte-ria from this genus we recovered several bacteriafrom the Rhodobacteraceae family including previ-ously observed bacteria from the Loktanella and Jan-naschia genera (Chistoserdov et al 2012 Meres et al2012) In addition to Aquimarina Loktanella andJannaschia Meres et al (2012) also reported bacteriafrom the genus Cardiobacterium and suborder Micro -coccineae as some of the most abundant bacteria intheir study We did not recover any bacteria fromeither of these groups nor their respective ordersThese differences could be due to the taxonomicidentification tools used in different studies but mayalso reflect differences in wild versus captive-raisedlobsters

In conclusion using a fine spatial scale we charac-terized changes in bacterial community compositionthat occur during the progression of LMSD Wedemonstrated that affected areas of lobster shell havealtered bacterial community composition and reducedbacterial diversity compared to unaffected shell Wealso demonstrated that areas close to the site of dis-ease but not yet melanized have altered bacterialcommunities and reduced bacterial diversity in com-parison to unaffected shell and represent a transi-tionary community that may be important in the initi-ation of LMSD We identified key bacteria associated

with each stage of the disease and proposed 2 differ-ent conceptual models one that suggests that thetransitionary bacterial community arises due to foul-ing and another that suggests the transitionary shellharbors bacteria important to the initiation of LMSDUnderstanding the full complement of bacteria asso-ciated with affected transitionary and asymptomaticlobster shells is the first step toward understandingthe role of the transitionary bacterial community andhelping us develop mechanisms to diminish thespread and severity of this disease

Acknowledgements We thank Anita Kim from the NewEngland Aquarium Lobster Research and Rearing Facilityfor caring for the animals used in this study and Jeff Dusen-berry for the use of the supercomputing facilities managedby the Research Computing Department at the University ofMassachusetts Boston We also thank Rachel Vincent anundergraduate assistant on this research as well as PatrickKearns John Angell and all the members of the BowenLaboratory Funding for this project was provided by theSanofi Genzyme Doctoral Research Fellowship (SGF) theUniversity of Massachusetts Boston Doctoral DissertationResearch Grant (SGF) and the National Science Founda-tion Research Experience for Undergraduates Award DBI-1359241 to Dr Rachel Skvirsky

LITERATURE CITED

Anderson MJ (2001) A new method for non-parametric mul-tivariate analysis of variance Austral Ecol 26 32minus46

Buerger S Spoering A Gavrish E Leslin C Ling L EpsteinSS (2012) Microbial scout hypothesis and microbial dis-covery Appl Environ Microbiol 78 3229minus3233

Caporaso JG Kuczynski J Stombaugh J Bittinger K andothers (2010) QIIME allows analysis of high-throughputcommunity sequencing data Nat Methods 7 335minus336

Caporaso JG Lauber CL Walters WA Berg-Lyons D andothers (2012) Ultra-high-throughput microbial commu-nity analysis on the Illumina HiSeq and MiSeq platformsISME J 6 1621minus1624

Castro KM Somers BA (2012) Observations of epizooticshell disease in American lobsters Homarus americanusin southern New England J Shellfish Res 31 423minus430

Castro KM Cobb JS Gomez-Chiarri M Tlusty M (2012)Epizootic shell disease in American lobsters Homarusamericanus in southern New England past present andfuture Dis Aquat Org 100 149minus158

Chistoserdov AY Smolowitz R Mirasol F Hsu A (2005) Cul-ture-dependent characterization of the microbial com-munity associated with epizootic shell disease lesions inAmerican lobster Homarus americanus J Shellfish Res24 741minus747

Chistoserdov AY Quinn RA Gubbala SL Smolowitz R(2012) Bacterial communities associated with lesions ofshell disease in the American lobster Homarus ameri-canus Milne-Edwards J Shellfish Res 31 449minus462

Davies CE Whitten M Kim A Wootton EC and others (2014)A comparison of the structure of American (Homarusamericanus) and European (Homarus gammarus) lobster

53

Dis Aquat Org 124 41ndash54 2017

p lt 001 df = 8) than in molted lobsters indicatingthat the bacterial community at the site of diseasewas more different from adjacent sites than would beexpected based on molted lobsters that showed nosigns of the LMSD

The weighted UniFrac similarities found whencomparing bacterial community composition at 05and 1 cm (Fig 3) were statistically different on lob-sters that had a lesion at 0 cm compared to lobstersthat had molted (t = minus216 p lt 005 df = 16) but werenot statistically different at α = 005 for lobsters thathad a spot at 0 cm (t = minus210 p gt 006 df = 10)Despite this discrepancy on lobsters with a spot at0 cm we still consider the 05 cm distance a transi-tionary zone in LMSD First the average weightedUniFrac similarity among the 05 and 1 cm distanceson lobsters with a spot at 0 cm was lower than onmolted lobsters indicating that the 05 cm bacterialcommunity on lobsters with a spot at 0 cm was lesssimilar to adjacent sites than expected Second theresults of the Welchrsquos t-test were reasonably close(p = 0062) to being significant at an arbitrarilyselected α = 005 Finally PERMANOVA analysisshowed significant support for grouping the data into3 different groups 0 05 and 1 and 15 cm whencomparing all lobsters with a lesion at 0 cm (F = 1630p lt 0001 df = 30) and lobsters with a spot at 0 cm(F = 1306 p lt 0001 df = 30) We therefore definedthe 05 cm distance as a transitionary zone harboringa bacterial community that was differentiated fromthe unaffected lobster shell community before thevisual onset of LMSD

The transitionary zone was limited to the 05 cmdistance as the weighted UniFrac similarity on lob-sters that had lesions or spots on their left claw wasnot statistically different from the same comparisonon the molted lobster shell at a dis-tance of 1 and 15 cm (Fig 3 t = 035p gt 07 df = 6 and t = minus097 p gt 03df = 7) This indicates that the bacterialcommunity was not altered by LMSDat 1 cm away from the site of melaniza-tion (and presumably at more distantlocations) and was therefore represen-tative of what was found on unaffectedlobster shell at this point in the moltcycle Given these results we here-after differentiate 3 categories of shell0 cm affected shell 05 cm transitio -nary shell and 1 cm and 15 cm un -affected shell

We also examined if the bacterialcommunity within a spot could be con-

sidered a transitionary condition from unaffectedshell to the more advanced lesion shell The lack of adistinct separation in the principal coordinates analy-sis between lobsters with lesions at 0 cm and spots at0 cm (Fig 2A) suggests that the categories of lesionand spot may not be meaningful with regard to theassociated bacterial communities To test this wecompared the bacterial community composition oflobsters with lesions to those with spots on both af -fected and transitionary shell and found there wasno statistical evidence to support separating lesionbacterial communities from spot bacterial communi-ties (F = 183 p gt 01 df = 15 and F = 082 p gt 03df = 15) These data suggest that once there wasvisual evidence of LMSD there was little subsequentchange in bacterial community composition and wetherefore no longer distinguish be tween lesion andspot in our analysis

After classifying samples into 3 discrete shell con-dition groups (affected transitionary and unaf-fected) based on bacterial community compositionwe compared differences in diversity within eachgroup using the pen where each lobster was keptas a baseline for maximum bacterial diversity (Fig 4)Regardless of shell condition bacterial diversity wasconsistently lower on the lobster shell than in thepen Alpha diversity varied among shell conditionswith the lowest diversity observed on the affectedshell Bacterial diversity on the transitionary shellwas intermediate between the affected shell andunaffected shell indicating that the drop in bacterialdiversity began before the visual appearance ofLMSD These differences in diversity account forsome of the differences seen in the rank abun-dance curve of each shell condition (Fig S2 in the Supplement)

46

Fig 4 Analysis of alpha diversity associated with LMSD for 3 different shellconditions and the lobster pen Differences in diversity between shell condi-tions were significant at a p lt 0001 for each diversity metric as indicated byletters (lowercase letters are used for comparisons using phylogenetic distanceas a measure uppercase letters are used for comparisons using Shannon

diversity) Error bars represent standard deviation

Feinman et al Bacterial communities in lobster shell disease

We identified the bacteria that were important tochanges in diversity among different shell conditionsby analyzing the core microbiome We defined thecore microbiome as bacteria that were present in n minus1 samples from each shell condition because bacteriathat are truly important to causing a shift to a shellcondition should be present in nearly all samplesfrom that shell condition While this criterion does notcapture all of the most abundant bacteria present(see Fig S2 in the Supplement) it narrows the focusto those that are more likely to be contributing to theprogression of LMSD

There were 76 OTUs considered to be part ofthe core microbiome of 1 or more shell condition

(Fig 5) These 76 OTUs comprised a total of 49 ofthe se quences recovered from affected shell 38of the sequences recovered from transitionary shelland 36 of the sequences recovered from unaf-fected shell Of these OTUs 16 were consideredunique to the core microbiome of unaffected shell(Fig 5AD) yet they only constituted 5 of thesequences from this shell condition Converselythere were 14 OTUs that were unique to the coremicrobiome of affected shell accounting for 34 ofthe sequences from this shell condition (Fig 5AB)This stark contrast is due in part to the significantlylower diversity in the af fected shell compared to theunaffected shell (Fig 4) From the affected core

47

Fig 5 (A) Core microbiome of each shell condition with (BminusG) accompanying bar charts of each operational taxonomic unit(OTU) in that core microbiome and its relative abundance within each shell condition (title text color matches the portion of (A)corresponding to the bar chart) Please note the difference in scale for each bar plot Individual OTUs are designated by theirOTU number along the x-axis Error bars represent standard deviation Numbers within the Venn diagram indicate the number of OTUs that belong to each core microbiome while italics indicate the percent those OTUs account for in each shell

condition

Dis Aquat Org 124 41ndash54 2017

microbiome we identified OTU 0 from the candi-date genus Candidatus Thio globus and OTU 18genus Aquimarina as important to LMSD based ontheir proportional abundances (combined 27Fig 5B Table 1)

There were 13 OTUs that were unique to the tran-sitionary core microbiome accounting for 5 of thesequences recovered from this part of the shell(Fig 5AC) The presence of OTUs that are unique tothe transitionary core microbiome indicates thatthere are members of the transitionary bacterial com-munity beyond what might be expected from a sim-ple mixing of the microbiota from healthy and dis-eased lobster shell OTUs 44 78 and 160 from thegenera Marinobacter Loktanella and CandidatusEndobugula stood out as important members of thetransitionary core microbiome due to their abun-dance in both the affected and transitionary shellconditions (Fig 5C Table 1)

There were also 33 OTUs found in the core micro-biome of more than 1 shell condition Of these 33OTUs 19 were shared between the transitionary andunaffected shell (Fig 5AG) whereas only 4 OTUs

were shared between the transitionary and affectedshell (Fig 5AE) This low number of OTUs sharedbetween the core microbiomes of transitionary andaffected shell provides further evidence that the bac-teria associated with this transition zone are notmerely a result of the outward expansion of the bac-teria inhabiting affected shell There were no OTUsshared between the affected and unaffected coremicrobiome that were not also a part of the transi-tionary core microbiome which also highlights theimportance of the bacterial community found 05 cmfrom the site of LMSD as a transition between these 2shell conditions From the affected transitionary coremicrobiome overlap we identified OTU 47 from thegenus Colwellia and OTU 118 from the genus Areni-cella to be of interest based on their relative abun-dances (Fig 5E Table 1) Finally there were 10 ubi -quitous OTUs shared among the core microbiomes ofall 3 shell conditions (Fig 5AF) OTU 1 from thegenus Perspicuibacter and OTU 19 from the genusLutimonas were of particular importance due to theirabundances on the transitionary and affected shellrespectively (Fig 5F Table 1)

48

OTU Order Family Genus Identity Abundance

Affected Core0 Thioglobusa 91 150 18 Flavobacteriales Flavobacteriaceae Aquimarina 100 118 21 Flavobacteriales Flavobacteriaceae Maritimimonas 96 1129 Alteromonadales Colwelliaceae Colwellia 99 2348 Alteromonadales Colwelliaceae Colwellia 100 0375 Rhodobacterales Rhodobacteraceae Loktanella 99 06121 Rhodobacterales Hyphomonadaceae Robiginitomaculum 96 04135 Rhizobiales Methylobacteriaceae Methylobacterium 100 04159 Oceanospirillales Oceanospirillaceae Oleispira 100 01229 Rhodobacterales Rhodobacteraceae Octadecabacter 100 04238 Rhizobiales Hyphomicrobiaceae Maritalea 99 07275 Flavobacteriales Flavobacteriaceae Cloacibacterium 100 01316 Oceanospirillales Oceanospirillaceae Neptunomonas 100 041306 Flavobacteriales Flavobacteriaceae Lutimonas 98 01

Transitionary Core2 Arenicellales Arenicellaceae Perspicuibacter 99 0739 Arenicellales Arenicellaceae Arenicella 98 0344 Alteromonadales Alteromonadaceae Marinobacter 100 1553 Rhodobacterales Rhodobacteraceae Roseovarius 100 0178 Rhodobacterales Rhodobacteraceae Loktanella 100 0189 Rhizobiales Phyllobacteriaceae Pseudahrensia 98 01107 Rhodobacterales Hyphomonadaceae Fretibacter 96 02160 Cellvibrionales Cellvibrionaceae Endobugulaa 99 10254 Oceanospirillales Alcanivoracaceae Marinicella 99 04338 Chromatiales Thioalkalispiraceae Thioprofundum 94 01

Table 1 Taxonomy for core operational taxonomic units (OTUs) identified in this study Closest cultured representative for allOTUs listed in order from Fig 5 Bold OTUs are those shown in Fig 6 Abundance is the relative abundance in that shell

condition Note OTU 0 has not yet been classified at the order or family level

(Table continued on next page)

Feinman et al Bacterial communities in lobster shell disease 49

Table 1 (continued)

OTU Order Family Genus Identity Abundance

370 Burkholderiales Comamonadaceae Diaphorobacter 100 01387 Rhodothermales Rubricoccaceae Rubrivirga 98 01410 Rhizobiales Phyllobacteriaceae Aminobacter 96 01

Unaffected Core6 Rhodobacterales Rhodobacteraceae Amylibacter 99 047 Rhodobacterales Rhodobacteraceae Halocynthiibacter 99 0123 Thiotrichales Thiotrichaceae Cocleimonas 96 0688 Rhizobiales Phyllobacteriaceae Pseudahrensia 99 0193 Arenicellales Arenicellaceae Arenicella 97 05110 Rhodobacterales Hyphomonadaceae Fretibacter 96 05128 Chromatiales Woeseiaceae Woeseia 96 03177 Acidimicrobiales Acidimicrobiaceae Ilumatobacter 100 02191 Sphingobacteriales Saprospiraceae Lewinella 92 05220 Acidimicrobiales Microthrixaceae Microthrixa 90 03233 Sphingobacteriales Saprospiraceae Lewinella 91 04277 Deferribacterales Deferribacteraceae Caldithrix 91 03294 Nitrosomonadales Nitrosomonadaceae Nitrosomonas 99 02302 Incertaesedis Rhodothermaceae Rhodothermus 89 03339 Rhodobacterales Rhodobacteraceae Labrenzia 97 02360 Rhizobiales Phyllobacteriaceae Aminobacter 95 02

Affected and Transitionary Core47 Alteromonadales Colwelliaceae Colwellia 99 3373 Oceanospirillales Oceanospirillaceae Bacterioplanes 91 0377 Rhodobacterales Rhodobacteraceae Loktanella 100 04118 Arenicellales Arenicellaceae Arenicella 99 17

Ubiquitous Core1 Arenicellales Arenicellaceae Perspicuibacter 99 278 Rhodobacterales Rhodobacteraceae Amylibacter 98 029 Rhodobacterales Rhodobacteraceae Litoreibacter 100 0210 Rhodobacterales Rhodobacteraceae Jannaschia 100 0219 Flavobacteriales Flavobacteriaceae Lutimonas 96 2538 Arenicellales Arenicellaceae Arenicella 99 1652 Rhodobacterales Rhodobacteraceae Sulfitobacter 100 0371 Flavobacteriales Flavobacteriaceae Maribacter 99 0698 Rhodobacterales Hyphomonadaceae Litorimonas 97 03268 Rhizobiales Methylobacteriaceae Methylobacterium 100 02

Transitionary and Unaffected Core20 Flavobacteriales Flavobacteriaceae Lutimonas 97 0322 Thiotrichales Thiotrichaceae Cocleimonas 97 0857 Rhodobacterales Rhodobacteraceae Phaeobacter 99 0174 Verrucomicrobiales Rubritaleaceae Rubritalea 97 0682 Acidimicrobiales Microthrixaceae Microthrixa 91 0287 Rhodobacterales Rhodobacteraceae Loktanella 98 02106 Cellvibrionales Cellvibrionaceae Endobugulaa 97 10125 Planctomycetales Planctomycetaceae Blastopirellula 91 03131 Arenicellales Arenicellaceae Perspicuibacter 93 02132 Rhodobacterales Rhodobacteraceae Maribius 95 02137 Rhizobiales Hyphomicrobiaceae Hyphomicrobium 98 02140 Rhodobacterales Hyphomonadaceae Fretibacter 95 02147 Rhizobiales Hyphomicrobiaceae Filomicrobium 96 03149 Flavobacteriales Flavobacteriaceae Lutimonas 97 01150 Rhodobacterales Rhodobacteraceae Oceaniovalibus 94 02161 Sphingobacteriales Saprospiraceae Phaeodactylibacter 89 02164 Rhodobacterales Rhodobacteraceae Citreicella 93 01174 Chromatiales Thioalkalispiraceae Thioprofundum 93 02262 Chromatiales Granulosicoccaceae Granulosicoccus 97 01aCandidate classification

Dis Aquat Org 124 41ndash54 2017

DISCUSSION

High-throughput sequencing results from lobsterswith and without visible signs of LMSD (ie lobsterswith spots lobsters with lesions and lobsters that hadrecently molted and did not exhibit signs of LMSD)indicate the presence of 3 distinct bacterial commu-nities (Fig 6A) We classified these 3 distinct commu-nities as affected transitionary and unaffected shellcommunities based on weighted UniFrac analysis ofcommunity composition (Figs 2 amp 3) The presence ofa distinct bacterial community at the site of diseaseagrees with prior work indicating that bacteria arethe causative agent of shell disease (Chistoserdov etal 2012) However our small spatial scale samplingallowed a more comprehensive characterization of thebacterial community associated with LMSD enabling

us to separate out the bacterial communities associ-ated with the transition from unaffected to affectedshell and thus identifying the presence of a third bac-terial community on shells affected with LMSD thetransitionary community

Interestingly our results indicate no differentiationamong bacterial communities found on lobsters withlesions and spots in either affected or transitionaryshell This finding is counter to previously reportedgenetic fingerprinting results that showed differ-ences in the bacterial communities among lesionsand spots (Quinn et al 2012b) Based on their initialfindings Quinn et al (2012b) suggested that the bac-teria present in a spot may be important in initiatingshell disease in lobsters whereas bacteria found inlesions were secondary opportunistic colonizers thatthrived in the new conditions The lack of differenti-

50

Fig 6 Conceptual diagram of the laboratory model of shell disease (LMSD) Visual summary of the findings of this study show-ing (A) the main observations of this study and (BC) 2 different alternate models of these data though others are possible (A1)The 3 distinct bacterial communities present on diseased lobster shell unaffected transitionary and affected shell communi-ties highlighting the lack of difference between spot and lesion communities The post-molt community is fairly similar to theunaffected shell community though this community devoid of disease is fairly uniform across the sampling area (A2) The ob-served decrease in bacterial diversity as LMSD progresses (A3) Several operational taxonomic units (OTUs) that are of inter-est to LMSD based on their abundances in either the transitionary or affected shell communities (B) Alternate Model 1 Thetransitionary bacterial community though unique is unimportant to the progression of LMSD and simply represents a community that is able to live in the proximity of disease-causing bacteria (C) Alternate Model 2 The transitionary bacterial

community consists of precursory bacteria that are present before visual signs of the disease occur

Feinman et al Bacterial communities in lobster shell disease

ation between lesion and spot communities in ourdata suggests that this may not be the case The tran-sitionary shell bacterial community that we identi-fied however may be important for initiation ofLMSD while the affected shell bacterial communitycould represent secondary colonizers

In addition to identifying 3 distinct bacterial com-munities we observed a positive association be tweenbacterial diversity and distance from the site of LMSD(Figs 4 amp 6A) This is contrary to work done by Whit-ten et al (2014) which showed greater bacterialdiversity in diseased shell compared to healthy shellThese previous results however were based onmethods that recovered an average of 8 to 11 taxa persample The sequencing methods we used allowedfor a much greater depth which revealed a greaterproportion of low abundance taxa thereby increas-ing apparent diversity The reduction in bacteria atthe site of LMSD indicates that the diseased shellenvironment may be highly selective for specific bac-terial taxa perhaps due to the host response and theantibacterial effects of melanin Alternatively op -portunistic bacteria associated with the newly ex -posed chitin matrix may outcompete other bacteriathereby excluding them Since bacterial diversitywas also reduced on the transitionary shell comparedto the unaffected shell we suggest that competitionrather than environmental factors lowered bacterialdiversity at the site of LMSD It is also possible thatalong with changes in the bacterial communitychanges occurred in the physical environment of thetransitionary shell prior to the onset of disease thathave yet to be identified

The presence of a transitionary community sug-gests that it is not just the bacteria at the site of dis-ease that are altered during LMSD but that there isalso a change in the bacterial community composi-tion at distances up to 05 cm away from the site ofmelanization The transitionary community couldrepresent several possibilities the 2 most parsimo-nious being (1) an outer edge of the disease zone thatis fouled by disease causing bacteria or (2) a novelcommunity that is a precursor for pathogenic bacteriathat later infect the lobster shell These 2 alternatemodels for disease transmission depicted in Fig 6BCprovide possible explanations for the changes in bothbacterial diversity and community structure we ob -served over space and time In Alternate Model 1(Fig 6B) the transitionary and affected bacterialcommunities both appear when spots are first ob -servable on the lobster shell Here the transitionarycommunity is not involved as a causative agent ofLMSD but instead is present due to fouling from the

disease-causing community and changes in the shellenvironment that occur due to the presence of dis-ease-causing bacteria As LMSD progresses theaffected area of the shell increases (transition fromspot to lesion) until the lobster eventually molts Thismodel is in agreement with previous reports regard-ing the progression of shell disease (Chistoserdov etal 2012 Davies et al 2014 Whitten et al 2014)though the sampling schemes in these previous stud-ies did not allow for the detection of a transitionarycommunity Alternate Model 2 (Fig 6C) interpretsthe transitionary bacterial community as a precursorcommunity arriving prior to the onset of visual signsof LMSD In this model the transitionary communityrepresents initial colonizers that may cause a shift inthe lobster shell bacterial community and lower bac-terial diversity prior to visual signs of disease Thisinitial invasion creates an environment that allowsfor secondary colonization that advances the diseasestate at which time the melanization associated withshell disease becomes observable The fact that thebacterial community at 05 cm is distinct from thecommunity at 0 cm (Figs 2 amp 3) and that there arebacteria that are unique to the core microbiome at05 cm (Fig 5) offers support for this second modelThis model hinges on the existence of a currentlyunobserved state where the transitionary and un -affected communities are present prior to visual signsof disease This state will be difficult to observe infuture studies as discovery would only come fromresearchers predicting the exact location where shelldisease might occur and sampling that predictedlocation prior to visual signs of disease While both ofthese alternate models differ from the currently pro-posed model for the progression of epizootic shelldisease both are in agreement with the previouslyproposed polymicrobial nature of shell disease thatis either caused by or amplified by a dysbiosis in bacterial communities on unaffected shell (Meres etal 2012) Culture studies examining growth dynamicsof potential pathogens in the presence and absenceof members of the transitionary community couldoffer additional insights into the location and initia-tion of LMSD and could refine the models proposedhere

Given the potential importance of the transitionaryand affected shell communities to the initiation andprogression of LMSD we identified key bacterialmembers of these shell conditions from the coremicrobiome (Figs 5 amp 6A) and evaluated them interms of the 2 hypothesized models The affectedshell was dominated by just 2 OTUs 0 and 18 whichaccounted for nearly 27 of the sequences from this

51

Dis Aquat Org 124 41ndash54 2017

shell condition (Fig 5B) Given their proportionaldominance these taxa appeared to be very importantto the affected shell community The closest culturedrepresentative to OTU 0 was from Candidatus Thio -globus singularis a mixotrophic bacterium belong-ing to the SUP05 clade known for its potential to oxi-dize sulfur and fix carbon in the dark ocean and inlow oxygen zones (Marshall amp Morris 2015) This cul-tured representative however only shared 91sequence identity with OTU 0 Therefore this OTUlikely represents a new uncultured species As thiswas the most abundant OTU from both the affectedand transitionary shell (see Table S1 in the Supple-ment) better identification and classification of thisbacterium is important for understanding its role inthe progression of LMSD OTU 18 the second mostabundant OTU on the affected shell (see Table S1)shared 100 sequence identity with a representa-tive from the genus Aquimarina (Esteves et al 2013)and 99 sequence identity with bacteria re coveredfrom lobster shells after having been exposed toAquimarina lsquohomariarsquo (Quinn et al 2012a) Aquima-rina lsquohomariarsquo has been identified by several studiesas a potential causative agent of epizootic shell dis-ease as it can readily be recovered from lesions(Chistoserdov et al 2005 2012 Quinn et al 2012aWhitten et al 2014) The abundance of OTU 18 onthe affected shell in this study (12 Fig 5B) con-firms the potential importance of Aquimarina lsquohomariarsquoto LMSD though its lack of dominance on the transi-tionary shell (2 Fig 5B) suggests that it is not assuccessful outside of melanized areas of the shell (seeTable S1)

In the transitionary core microbiome (Fig 5C)OTUs 44 and 160 appeared to be particularly impor-tant OTU 44 was one of the most abundant OTUs inall shell conditions however it was only consideredcore to the transitionary shell (Table 1 see Table S1)This OTU was most closely related to Marinobactersalsuginis (100 sequence identity S V Law et alunpubl data accession no KT921331) Other speciesfrom this genus have proteins for the deacetylation ofchitin (Song et al 2013) which suggests that OTU 44may be important to initiation of disease creatingconditions under which opportunistic colonizers asso-ciated with LMSD might thrive (Malloy 1978 Chisto -serdov et al 2005) OTU 160 was most closely relatedto the candidate species Endobugula sertula (99sequence identity McGovern amp Hellberg 2003) abryozoan symbiont that produces polyketides to pro-tect the host larvae from predation (Lindquist amp Hay1996 Lopanik et al 2004) In addition to OTUs 44and 160 OTU 78 was part of the transitionary core

microbiome though quantitatively it appeared to bemore important to the affected shell Although thisOTU was not a member of the affected shell coremicrobiome it was present in 13 of the 16 affectedshell samples and was present at relatively highabundance in 7 of those samples This OTU genusLoktanella which shared 100 sequence identitywith bacteria recovered from other diseased lobstershells (Quinn et al 2012a) may therefore be in fectingthe shell alongside early colonizers and in some in -stances growing in abundance as LMSD progresses

Of the 4 OTUs shared between the affected andtransitionary core microbiomes (Fig 5E) OTUs 47and 118 were numerically dominant in both commu-nities OTU 47 belonged to the genus Colwelliawhich Meres et al (2012) reported as one of the manygroups of species that helped to differentiate dis-eased and healthy lobster shell In the current studythis bacterium was abundant on transitionary shellan area of the shell that would have previously re -mained unsampled indicating that members of thisgenus may be more important in promoting LMSDthan previously thought if the transitionary shellindeed reflects an initial invasion OTU 118 was mostclosely aligned with Arenicella chitinivorans (99sequence identity) a bacterium that differs from oth-ers in the same genus by its ability to degrade chitin(Nedashkovskaya et al 2013) Given that previouswork by Whitten et al (2014) also recovered mem-bers of this genus from lesioned areas of the lobstershell it appears that the ability to degrade chitin maybe an important characteristic allowing bacteria tothrive on both the affected and transitionary shell

Many studies have reported the incidence of ubiq-uitous bacteria on the lobster shell both on diseasedand healthy shell (Chistoserdov et al 2005 Meres etal 2012 Whitten et al 2014) In this study there were10 OTUs present in the core microbiome of all 3 shellconditions (Fig 5F) and OTUs 1 and 19 were propor-tionally dominant on the transitionary and affectedshell respectively though they were highly abun-dant in all 3 shell conditions (see Table S1) OTU 1 ismost likely Perspicuibacter marinus (99 sequenceidentity) an aerobic marine bacterium recently iso-lated from surface seawater in Japan (Teramoto et al2015) At this time not much is known about thisorganism but it has been cultured and thereforecould be used in future studies exploring the dynam-ics of the transitionary lobster shell community OTU19 an important shell bacterium found on affectedshell was most similar to other bacteria found inassociation with sponges (99 sequence identitySipkema et al 2011) It is likely that this bacterium is

52

Feinman et al Bacterial communities in lobster shell disease

from the genus Lutimonas (96 sequence identityBuerger et al 2012) which is part of the same familyas Aquimarina and may represent another oppor-tunistic colonizer of diseased lobster shell

In addition to these core microbiome memberswhich appear to be important due to their relativeabundance on either the transitionary or affectedshell there were a number of other core microbiomeOTUs that had identities similar to taxa identified inprevious studies of epizootic shell disease Severalstudies have reported Aquimarina lsquohomariarsquo as apotential causative agent in shell disease including abacterial challenge study that was able to recoverthis bacterium from lobsters with lesions (Quinn et al2012a) and a next-generation sequencing study thatexamined healthy and diseased shell (Meres et al2012) Although we only recovered a single OTU thatwas associated with the genus Aquimarina (OTU 18)we recovered several other OTUs including Lutimo -nas that were from the Flavobacteria family (Table 1)Several studies have reported that members of thisfamily aside from Aquimarina are important to epi-zootic shell disease (Chistoserdov et al 2012 Mereset al 2012 Quinn et al 2012b) In addition to Aqui -marina the lsquoThalassobiusrsquo genus has been associatedwith epizootic shell disease (Meres et al 2012 Quinnet al 2012a) Although we did not recover any bacte-ria from this genus we recovered several bacteriafrom the Rhodobacteraceae family including previ-ously observed bacteria from the Loktanella and Jan-naschia genera (Chistoserdov et al 2012 Meres et al2012) In addition to Aquimarina Loktanella andJannaschia Meres et al (2012) also reported bacteriafrom the genus Cardiobacterium and suborder Micro -coccineae as some of the most abundant bacteria intheir study We did not recover any bacteria fromeither of these groups nor their respective ordersThese differences could be due to the taxonomicidentification tools used in different studies but mayalso reflect differences in wild versus captive-raisedlobsters

In conclusion using a fine spatial scale we charac-terized changes in bacterial community compositionthat occur during the progression of LMSD Wedemonstrated that affected areas of lobster shell havealtered bacterial community composition and reducedbacterial diversity compared to unaffected shell Wealso demonstrated that areas close to the site of dis-ease but not yet melanized have altered bacterialcommunities and reduced bacterial diversity in com-parison to unaffected shell and represent a transi-tionary community that may be important in the initi-ation of LMSD We identified key bacteria associated

with each stage of the disease and proposed 2 differ-ent conceptual models one that suggests that thetransitionary bacterial community arises due to foul-ing and another that suggests the transitionary shellharbors bacteria important to the initiation of LMSDUnderstanding the full complement of bacteria asso-ciated with affected transitionary and asymptomaticlobster shells is the first step toward understandingthe role of the transitionary bacterial community andhelping us develop mechanisms to diminish thespread and severity of this disease

Acknowledgements We thank Anita Kim from the NewEngland Aquarium Lobster Research and Rearing Facilityfor caring for the animals used in this study and Jeff Dusen-berry for the use of the supercomputing facilities managedby the Research Computing Department at the University ofMassachusetts Boston We also thank Rachel Vincent anundergraduate assistant on this research as well as PatrickKearns John Angell and all the members of the BowenLaboratory Funding for this project was provided by theSanofi Genzyme Doctoral Research Fellowship (SGF) theUniversity of Massachusetts Boston Doctoral DissertationResearch Grant (SGF) and the National Science Founda-tion Research Experience for Undergraduates Award DBI-1359241 to Dr Rachel Skvirsky

LITERATURE CITED

Anderson MJ (2001) A new method for non-parametric mul-tivariate analysis of variance Austral Ecol 26 32minus46

Buerger S Spoering A Gavrish E Leslin C Ling L EpsteinSS (2012) Microbial scout hypothesis and microbial dis-covery Appl Environ Microbiol 78 3229minus3233

Caporaso JG Kuczynski J Stombaugh J Bittinger K andothers (2010) QIIME allows analysis of high-throughputcommunity sequencing data Nat Methods 7 335minus336

Caporaso JG Lauber CL Walters WA Berg-Lyons D andothers (2012) Ultra-high-throughput microbial commu-nity analysis on the Illumina HiSeq and MiSeq platformsISME J 6 1621minus1624

Castro KM Somers BA (2012) Observations of epizooticshell disease in American lobsters Homarus americanusin southern New England J Shellfish Res 31 423minus430

Castro KM Cobb JS Gomez-Chiarri M Tlusty M (2012)Epizootic shell disease in American lobsters Homarusamericanus in southern New England past present andfuture Dis Aquat Org 100 149minus158

Chistoserdov AY Smolowitz R Mirasol F Hsu A (2005) Cul-ture-dependent characterization of the microbial com-munity associated with epizootic shell disease lesions inAmerican lobster Homarus americanus J Shellfish Res24 741minus747

Chistoserdov AY Quinn RA Gubbala SL Smolowitz R(2012) Bacterial communities associated with lesions ofshell disease in the American lobster Homarus ameri-canus Milne-Edwards J Shellfish Res 31 449minus462

Davies CE Whitten M Kim A Wootton EC and others (2014)A comparison of the structure of American (Homarusamericanus) and European (Homarus gammarus) lobster

53

Feinman et al Bacterial communities in lobster shell disease

We identified the bacteria that were important tochanges in diversity among different shell conditionsby analyzing the core microbiome We defined thecore microbiome as bacteria that were present in n minus1 samples from each shell condition because bacteriathat are truly important to causing a shift to a shellcondition should be present in nearly all samplesfrom that shell condition While this criterion does notcapture all of the most abundant bacteria present(see Fig S2 in the Supplement) it narrows the focusto those that are more likely to be contributing to theprogression of LMSD

There were 76 OTUs considered to be part ofthe core microbiome of 1 or more shell condition

(Fig 5) These 76 OTUs comprised a total of 49 ofthe se quences recovered from affected shell 38of the sequences recovered from transitionary shelland 36 of the sequences recovered from unaf-fected shell Of these OTUs 16 were consideredunique to the core microbiome of unaffected shell(Fig 5AD) yet they only constituted 5 of thesequences from this shell condition Converselythere were 14 OTUs that were unique to the coremicrobiome of affected shell accounting for 34 ofthe sequences from this shell condition (Fig 5AB)This stark contrast is due in part to the significantlylower diversity in the af fected shell compared to theunaffected shell (Fig 4) From the affected core

47

Fig 5 (A) Core microbiome of each shell condition with (BminusG) accompanying bar charts of each operational taxonomic unit(OTU) in that core microbiome and its relative abundance within each shell condition (title text color matches the portion of (A)corresponding to the bar chart) Please note the difference in scale for each bar plot Individual OTUs are designated by theirOTU number along the x-axis Error bars represent standard deviation Numbers within the Venn diagram indicate the number of OTUs that belong to each core microbiome while italics indicate the percent those OTUs account for in each shell

condition

Dis Aquat Org 124 41ndash54 2017

microbiome we identified OTU 0 from the candi-date genus Candidatus Thio globus and OTU 18genus Aquimarina as important to LMSD based ontheir proportional abundances (combined 27Fig 5B Table 1)

There were 13 OTUs that were unique to the tran-sitionary core microbiome accounting for 5 of thesequences recovered from this part of the shell(Fig 5AC) The presence of OTUs that are unique tothe transitionary core microbiome indicates thatthere are members of the transitionary bacterial com-munity beyond what might be expected from a sim-ple mixing of the microbiota from healthy and dis-eased lobster shell OTUs 44 78 and 160 from thegenera Marinobacter Loktanella and CandidatusEndobugula stood out as important members of thetransitionary core microbiome due to their abun-dance in both the affected and transitionary shellconditions (Fig 5C Table 1)

There were also 33 OTUs found in the core micro-biome of more than 1 shell condition Of these 33OTUs 19 were shared between the transitionary andunaffected shell (Fig 5AG) whereas only 4 OTUs

were shared between the transitionary and affectedshell (Fig 5AE) This low number of OTUs sharedbetween the core microbiomes of transitionary andaffected shell provides further evidence that the bac-teria associated with this transition zone are notmerely a result of the outward expansion of the bac-teria inhabiting affected shell There were no OTUsshared between the affected and unaffected coremicrobiome that were not also a part of the transi-tionary core microbiome which also highlights theimportance of the bacterial community found 05 cmfrom the site of LMSD as a transition between these 2shell conditions From the affected transitionary coremicrobiome overlap we identified OTU 47 from thegenus Colwellia and OTU 118 from the genus Areni-cella to be of interest based on their relative abun-dances (Fig 5E Table 1) Finally there were 10 ubi -quitous OTUs shared among the core microbiomes ofall 3 shell conditions (Fig 5AF) OTU 1 from thegenus Perspicuibacter and OTU 19 from the genusLutimonas were of particular importance due to theirabundances on the transitionary and affected shellrespectively (Fig 5F Table 1)

48

OTU Order Family Genus Identity Abundance

Affected Core0 Thioglobusa 91 150 18 Flavobacteriales Flavobacteriaceae Aquimarina 100 118 21 Flavobacteriales Flavobacteriaceae Maritimimonas 96 1129 Alteromonadales Colwelliaceae Colwellia 99 2348 Alteromonadales Colwelliaceae Colwellia 100 0375 Rhodobacterales Rhodobacteraceae Loktanella 99 06121 Rhodobacterales Hyphomonadaceae Robiginitomaculum 96 04135 Rhizobiales Methylobacteriaceae Methylobacterium 100 04159 Oceanospirillales Oceanospirillaceae Oleispira 100 01229 Rhodobacterales Rhodobacteraceae Octadecabacter 100 04238 Rhizobiales Hyphomicrobiaceae Maritalea 99 07275 Flavobacteriales Flavobacteriaceae Cloacibacterium 100 01316 Oceanospirillales Oceanospirillaceae Neptunomonas 100 041306 Flavobacteriales Flavobacteriaceae Lutimonas 98 01

Transitionary Core2 Arenicellales Arenicellaceae Perspicuibacter 99 0739 Arenicellales Arenicellaceae Arenicella 98 0344 Alteromonadales Alteromonadaceae Marinobacter 100 1553 Rhodobacterales Rhodobacteraceae Roseovarius 100 0178 Rhodobacterales Rhodobacteraceae Loktanella 100 0189 Rhizobiales Phyllobacteriaceae Pseudahrensia 98 01107 Rhodobacterales Hyphomonadaceae Fretibacter 96 02160 Cellvibrionales Cellvibrionaceae Endobugulaa 99 10254 Oceanospirillales Alcanivoracaceae Marinicella 99 04338 Chromatiales Thioalkalispiraceae Thioprofundum 94 01

Table 1 Taxonomy for core operational taxonomic units (OTUs) identified in this study Closest cultured representative for allOTUs listed in order from Fig 5 Bold OTUs are those shown in Fig 6 Abundance is the relative abundance in that shell

condition Note OTU 0 has not yet been classified at the order or family level

(Table continued on next page)

Feinman et al Bacterial communities in lobster shell disease 49

Table 1 (continued)

OTU Order Family Genus Identity Abundance

370 Burkholderiales Comamonadaceae Diaphorobacter 100 01387 Rhodothermales Rubricoccaceae Rubrivirga 98 01410 Rhizobiales Phyllobacteriaceae Aminobacter 96 01

Unaffected Core6 Rhodobacterales Rhodobacteraceae Amylibacter 99 047 Rhodobacterales Rhodobacteraceae Halocynthiibacter 99 0123 Thiotrichales Thiotrichaceae Cocleimonas 96 0688 Rhizobiales Phyllobacteriaceae Pseudahrensia 99 0193 Arenicellales Arenicellaceae Arenicella 97 05110 Rhodobacterales Hyphomonadaceae Fretibacter 96 05128 Chromatiales Woeseiaceae Woeseia 96 03177 Acidimicrobiales Acidimicrobiaceae Ilumatobacter 100 02191 Sphingobacteriales Saprospiraceae Lewinella 92 05220 Acidimicrobiales Microthrixaceae Microthrixa 90 03233 Sphingobacteriales Saprospiraceae Lewinella 91 04277 Deferribacterales Deferribacteraceae Caldithrix 91 03294 Nitrosomonadales Nitrosomonadaceae Nitrosomonas 99 02302 Incertaesedis Rhodothermaceae Rhodothermus 89 03339 Rhodobacterales Rhodobacteraceae Labrenzia 97 02360 Rhizobiales Phyllobacteriaceae Aminobacter 95 02

Affected and Transitionary Core47 Alteromonadales Colwelliaceae Colwellia 99 3373 Oceanospirillales Oceanospirillaceae Bacterioplanes 91 0377 Rhodobacterales Rhodobacteraceae Loktanella 100 04118 Arenicellales Arenicellaceae Arenicella 99 17

Ubiquitous Core1 Arenicellales Arenicellaceae Perspicuibacter 99 278 Rhodobacterales Rhodobacteraceae Amylibacter 98 029 Rhodobacterales Rhodobacteraceae Litoreibacter 100 0210 Rhodobacterales Rhodobacteraceae Jannaschia 100 0219 Flavobacteriales Flavobacteriaceae Lutimonas 96 2538 Arenicellales Arenicellaceae Arenicella 99 1652 Rhodobacterales Rhodobacteraceae Sulfitobacter 100 0371 Flavobacteriales Flavobacteriaceae Maribacter 99 0698 Rhodobacterales Hyphomonadaceae Litorimonas 97 03268 Rhizobiales Methylobacteriaceae Methylobacterium 100 02

Transitionary and Unaffected Core20 Flavobacteriales Flavobacteriaceae Lutimonas 97 0322 Thiotrichales Thiotrichaceae Cocleimonas 97 0857 Rhodobacterales Rhodobacteraceae Phaeobacter 99 0174 Verrucomicrobiales Rubritaleaceae Rubritalea 97 0682 Acidimicrobiales Microthrixaceae Microthrixa 91 0287 Rhodobacterales Rhodobacteraceae Loktanella 98 02106 Cellvibrionales Cellvibrionaceae Endobugulaa 97 10125 Planctomycetales Planctomycetaceae Blastopirellula 91 03131 Arenicellales Arenicellaceae Perspicuibacter 93 02132 Rhodobacterales Rhodobacteraceae Maribius 95 02137 Rhizobiales Hyphomicrobiaceae Hyphomicrobium 98 02140 Rhodobacterales Hyphomonadaceae Fretibacter 95 02147 Rhizobiales Hyphomicrobiaceae Filomicrobium 96 03149 Flavobacteriales Flavobacteriaceae Lutimonas 97 01150 Rhodobacterales Rhodobacteraceae Oceaniovalibus 94 02161 Sphingobacteriales Saprospiraceae Phaeodactylibacter 89 02164 Rhodobacterales Rhodobacteraceae Citreicella 93 01174 Chromatiales Thioalkalispiraceae Thioprofundum 93 02262 Chromatiales Granulosicoccaceae Granulosicoccus 97 01aCandidate classification

Dis Aquat Org 124 41ndash54 2017

DISCUSSION

High-throughput sequencing results from lobsterswith and without visible signs of LMSD (ie lobsterswith spots lobsters with lesions and lobsters that hadrecently molted and did not exhibit signs of LMSD)indicate the presence of 3 distinct bacterial commu-nities (Fig 6A) We classified these 3 distinct commu-nities as affected transitionary and unaffected shellcommunities based on weighted UniFrac analysis ofcommunity composition (Figs 2 amp 3) The presence ofa distinct bacterial community at the site of diseaseagrees with prior work indicating that bacteria arethe causative agent of shell disease (Chistoserdov etal 2012) However our small spatial scale samplingallowed a more comprehensive characterization of thebacterial community associated with LMSD enabling

us to separate out the bacterial communities associ-ated with the transition from unaffected to affectedshell and thus identifying the presence of a third bac-terial community on shells affected with LMSD thetransitionary community

Interestingly our results indicate no differentiationamong bacterial communities found on lobsters withlesions and spots in either affected or transitionaryshell This finding is counter to previously reportedgenetic fingerprinting results that showed differ-ences in the bacterial communities among lesionsand spots (Quinn et al 2012b) Based on their initialfindings Quinn et al (2012b) suggested that the bac-teria present in a spot may be important in initiatingshell disease in lobsters whereas bacteria found inlesions were secondary opportunistic colonizers thatthrived in the new conditions The lack of differenti-

50

Fig 6 Conceptual diagram of the laboratory model of shell disease (LMSD) Visual summary of the findings of this study show-ing (A) the main observations of this study and (BC) 2 different alternate models of these data though others are possible (A1)The 3 distinct bacterial communities present on diseased lobster shell unaffected transitionary and affected shell communi-ties highlighting the lack of difference between spot and lesion communities The post-molt community is fairly similar to theunaffected shell community though this community devoid of disease is fairly uniform across the sampling area (A2) The ob-served decrease in bacterial diversity as LMSD progresses (A3) Several operational taxonomic units (OTUs) that are of inter-est to LMSD based on their abundances in either the transitionary or affected shell communities (B) Alternate Model 1 Thetransitionary bacterial community though unique is unimportant to the progression of LMSD and simply represents a community that is able to live in the proximity of disease-causing bacteria (C) Alternate Model 2 The transitionary bacterial

community consists of precursory bacteria that are present before visual signs of the disease occur

Feinman et al Bacterial communities in lobster shell disease

ation between lesion and spot communities in ourdata suggests that this may not be the case The tran-sitionary shell bacterial community that we identi-fied however may be important for initiation ofLMSD while the affected shell bacterial communitycould represent secondary colonizers

In addition to identifying 3 distinct bacterial com-munities we observed a positive association be tweenbacterial diversity and distance from the site of LMSD(Figs 4 amp 6A) This is contrary to work done by Whit-ten et al (2014) which showed greater bacterialdiversity in diseased shell compared to healthy shellThese previous results however were based onmethods that recovered an average of 8 to 11 taxa persample The sequencing methods we used allowedfor a much greater depth which revealed a greaterproportion of low abundance taxa thereby increas-ing apparent diversity The reduction in bacteria atthe site of LMSD indicates that the diseased shellenvironment may be highly selective for specific bac-terial taxa perhaps due to the host response and theantibacterial effects of melanin Alternatively op -portunistic bacteria associated with the newly ex -posed chitin matrix may outcompete other bacteriathereby excluding them Since bacterial diversitywas also reduced on the transitionary shell comparedto the unaffected shell we suggest that competitionrather than environmental factors lowered bacterialdiversity at the site of LMSD It is also possible thatalong with changes in the bacterial communitychanges occurred in the physical environment of thetransitionary shell prior to the onset of disease thathave yet to be identified

The presence of a transitionary community sug-gests that it is not just the bacteria at the site of dis-ease that are altered during LMSD but that there isalso a change in the bacterial community composi-tion at distances up to 05 cm away from the site ofmelanization The transitionary community couldrepresent several possibilities the 2 most parsimo-nious being (1) an outer edge of the disease zone thatis fouled by disease causing bacteria or (2) a novelcommunity that is a precursor for pathogenic bacteriathat later infect the lobster shell These 2 alternatemodels for disease transmission depicted in Fig 6BCprovide possible explanations for the changes in bothbacterial diversity and community structure we ob -served over space and time In Alternate Model 1(Fig 6B) the transitionary and affected bacterialcommunities both appear when spots are first ob -servable on the lobster shell Here the transitionarycommunity is not involved as a causative agent ofLMSD but instead is present due to fouling from the

disease-causing community and changes in the shellenvironment that occur due to the presence of dis-ease-causing bacteria As LMSD progresses theaffected area of the shell increases (transition fromspot to lesion) until the lobster eventually molts Thismodel is in agreement with previous reports regard-ing the progression of shell disease (Chistoserdov etal 2012 Davies et al 2014 Whitten et al 2014)though the sampling schemes in these previous stud-ies did not allow for the detection of a transitionarycommunity Alternate Model 2 (Fig 6C) interpretsthe transitionary bacterial community as a precursorcommunity arriving prior to the onset of visual signsof LMSD In this model the transitionary communityrepresents initial colonizers that may cause a shift inthe lobster shell bacterial community and lower bac-terial diversity prior to visual signs of disease Thisinitial invasion creates an environment that allowsfor secondary colonization that advances the diseasestate at which time the melanization associated withshell disease becomes observable The fact that thebacterial community at 05 cm is distinct from thecommunity at 0 cm (Figs 2 amp 3) and that there arebacteria that are unique to the core microbiome at05 cm (Fig 5) offers support for this second modelThis model hinges on the existence of a currentlyunobserved state where the transitionary and un -affected communities are present prior to visual signsof disease This state will be difficult to observe infuture studies as discovery would only come fromresearchers predicting the exact location where shelldisease might occur and sampling that predictedlocation prior to visual signs of disease While both ofthese alternate models differ from the currently pro-posed model for the progression of epizootic shelldisease both are in agreement with the previouslyproposed polymicrobial nature of shell disease thatis either caused by or amplified by a dysbiosis in bacterial communities on unaffected shell (Meres etal 2012) Culture studies examining growth dynamicsof potential pathogens in the presence and absenceof members of the transitionary community couldoffer additional insights into the location and initia-tion of LMSD and could refine the models proposedhere

Given the potential importance of the transitionaryand affected shell communities to the initiation andprogression of LMSD we identified key bacterialmembers of these shell conditions from the coremicrobiome (Figs 5 amp 6A) and evaluated them interms of the 2 hypothesized models The affectedshell was dominated by just 2 OTUs 0 and 18 whichaccounted for nearly 27 of the sequences from this

51

Dis Aquat Org 124 41ndash54 2017

shell condition (Fig 5B) Given their proportionaldominance these taxa appeared to be very importantto the affected shell community The closest culturedrepresentative to OTU 0 was from Candidatus Thio -globus singularis a mixotrophic bacterium belong-ing to the SUP05 clade known for its potential to oxi-dize sulfur and fix carbon in the dark ocean and inlow oxygen zones (Marshall amp Morris 2015) This cul-tured representative however only shared 91sequence identity with OTU 0 Therefore this OTUlikely represents a new uncultured species As thiswas the most abundant OTU from both the affectedand transitionary shell (see Table S1 in the Supple-ment) better identification and classification of thisbacterium is important for understanding its role inthe progression of LMSD OTU 18 the second mostabundant OTU on the affected shell (see Table S1)shared 100 sequence identity with a representa-tive from the genus Aquimarina (Esteves et al 2013)and 99 sequence identity with bacteria re coveredfrom lobster shells after having been exposed toAquimarina lsquohomariarsquo (Quinn et al 2012a) Aquima-rina lsquohomariarsquo has been identified by several studiesas a potential causative agent of epizootic shell dis-ease as it can readily be recovered from lesions(Chistoserdov et al 2005 2012 Quinn et al 2012aWhitten et al 2014) The abundance of OTU 18 onthe affected shell in this study (12 Fig 5B) con-firms the potential importance of Aquimarina lsquohomariarsquoto LMSD though its lack of dominance on the transi-tionary shell (2 Fig 5B) suggests that it is not assuccessful outside of melanized areas of the shell (seeTable S1)

In the transitionary core microbiome (Fig 5C)OTUs 44 and 160 appeared to be particularly impor-tant OTU 44 was one of the most abundant OTUs inall shell conditions however it was only consideredcore to the transitionary shell (Table 1 see Table S1)This OTU was most closely related to Marinobactersalsuginis (100 sequence identity S V Law et alunpubl data accession no KT921331) Other speciesfrom this genus have proteins for the deacetylation ofchitin (Song et al 2013) which suggests that OTU 44may be important to initiation of disease creatingconditions under which opportunistic colonizers asso-ciated with LMSD might thrive (Malloy 1978 Chisto -serdov et al 2005) OTU 160 was most closely relatedto the candidate species Endobugula sertula (99sequence identity McGovern amp Hellberg 2003) abryozoan symbiont that produces polyketides to pro-tect the host larvae from predation (Lindquist amp Hay1996 Lopanik et al 2004) In addition to OTUs 44and 160 OTU 78 was part of the transitionary core

microbiome though quantitatively it appeared to bemore important to the affected shell Although thisOTU was not a member of the affected shell coremicrobiome it was present in 13 of the 16 affectedshell samples and was present at relatively highabundance in 7 of those samples This OTU genusLoktanella which shared 100 sequence identitywith bacteria recovered from other diseased lobstershells (Quinn et al 2012a) may therefore be in fectingthe shell alongside early colonizers and in some in -stances growing in abundance as LMSD progresses

Of the 4 OTUs shared between the affected andtransitionary core microbiomes (Fig 5E) OTUs 47and 118 were numerically dominant in both commu-nities OTU 47 belonged to the genus Colwelliawhich Meres et al (2012) reported as one of the manygroups of species that helped to differentiate dis-eased and healthy lobster shell In the current studythis bacterium was abundant on transitionary shellan area of the shell that would have previously re -mained unsampled indicating that members of thisgenus may be more important in promoting LMSDthan previously thought if the transitionary shellindeed reflects an initial invasion OTU 118 was mostclosely aligned with Arenicella chitinivorans (99sequence identity) a bacterium that differs from oth-ers in the same genus by its ability to degrade chitin(Nedashkovskaya et al 2013) Given that previouswork by Whitten et al (2014) also recovered mem-bers of this genus from lesioned areas of the lobstershell it appears that the ability to degrade chitin maybe an important characteristic allowing bacteria tothrive on both the affected and transitionary shell

Many studies have reported the incidence of ubiq-uitous bacteria on the lobster shell both on diseasedand healthy shell (Chistoserdov et al 2005 Meres etal 2012 Whitten et al 2014) In this study there were10 OTUs present in the core microbiome of all 3 shellconditions (Fig 5F) and OTUs 1 and 19 were propor-tionally dominant on the transitionary and affectedshell respectively though they were highly abun-dant in all 3 shell conditions (see Table S1) OTU 1 ismost likely Perspicuibacter marinus (99 sequenceidentity) an aerobic marine bacterium recently iso-lated from surface seawater in Japan (Teramoto et al2015) At this time not much is known about thisorganism but it has been cultured and thereforecould be used in future studies exploring the dynam-ics of the transitionary lobster shell community OTU19 an important shell bacterium found on affectedshell was most similar to other bacteria found inassociation with sponges (99 sequence identitySipkema et al 2011) It is likely that this bacterium is

52

Feinman et al Bacterial communities in lobster shell disease

from the genus Lutimonas (96 sequence identityBuerger et al 2012) which is part of the same familyas Aquimarina and may represent another oppor-tunistic colonizer of diseased lobster shell

In addition to these core microbiome memberswhich appear to be important due to their relativeabundance on either the transitionary or affectedshell there were a number of other core microbiomeOTUs that had identities similar to taxa identified inprevious studies of epizootic shell disease Severalstudies have reported Aquimarina lsquohomariarsquo as apotential causative agent in shell disease including abacterial challenge study that was able to recoverthis bacterium from lobsters with lesions (Quinn et al2012a) and a next-generation sequencing study thatexamined healthy and diseased shell (Meres et al2012) Although we only recovered a single OTU thatwas associated with the genus Aquimarina (OTU 18)we recovered several other OTUs including Lutimo -nas that were from the Flavobacteria family (Table 1)Several studies have reported that members of thisfamily aside from Aquimarina are important to epi-zootic shell disease (Chistoserdov et al 2012 Mereset al 2012 Quinn et al 2012b) In addition to Aqui -marina the lsquoThalassobiusrsquo genus has been associatedwith epizootic shell disease (Meres et al 2012 Quinnet al 2012a) Although we did not recover any bacte-ria from this genus we recovered several bacteriafrom the Rhodobacteraceae family including previ-ously observed bacteria from the Loktanella and Jan-naschia genera (Chistoserdov et al 2012 Meres et al2012) In addition to Aquimarina Loktanella andJannaschia Meres et al (2012) also reported bacteriafrom the genus Cardiobacterium and suborder Micro -coccineae as some of the most abundant bacteria intheir study We did not recover any bacteria fromeither of these groups nor their respective ordersThese differences could be due to the taxonomicidentification tools used in different studies but mayalso reflect differences in wild versus captive-raisedlobsters

In conclusion using a fine spatial scale we charac-terized changes in bacterial community compositionthat occur during the progression of LMSD Wedemonstrated that affected areas of lobster shell havealtered bacterial community composition and reducedbacterial diversity compared to unaffected shell Wealso demonstrated that areas close to the site of dis-ease but not yet melanized have altered bacterialcommunities and reduced bacterial diversity in com-parison to unaffected shell and represent a transi-tionary community that may be important in the initi-ation of LMSD We identified key bacteria associated

with each stage of the disease and proposed 2 differ-ent conceptual models one that suggests that thetransitionary bacterial community arises due to foul-ing and another that suggests the transitionary shellharbors bacteria important to the initiation of LMSDUnderstanding the full complement of bacteria asso-ciated with affected transitionary and asymptomaticlobster shells is the first step toward understandingthe role of the transitionary bacterial community andhelping us develop mechanisms to diminish thespread and severity of this disease

Acknowledgements We thank Anita Kim from the NewEngland Aquarium Lobster Research and Rearing Facilityfor caring for the animals used in this study and Jeff Dusen-berry for the use of the supercomputing facilities managedby the Research Computing Department at the University ofMassachusetts Boston We also thank Rachel Vincent anundergraduate assistant on this research as well as PatrickKearns John Angell and all the members of the BowenLaboratory Funding for this project was provided by theSanofi Genzyme Doctoral Research Fellowship (SGF) theUniversity of Massachusetts Boston Doctoral DissertationResearch Grant (SGF) and the National Science Founda-tion Research Experience for Undergraduates Award DBI-1359241 to Dr Rachel Skvirsky

LITERATURE CITED

Anderson MJ (2001) A new method for non-parametric mul-tivariate analysis of variance Austral Ecol 26 32minus46

Buerger S Spoering A Gavrish E Leslin C Ling L EpsteinSS (2012) Microbial scout hypothesis and microbial dis-covery Appl Environ Microbiol 78 3229minus3233

Caporaso JG Kuczynski J Stombaugh J Bittinger K andothers (2010) QIIME allows analysis of high-throughputcommunity sequencing data Nat Methods 7 335minus336

Caporaso JG Lauber CL Walters WA Berg-Lyons D andothers (2012) Ultra-high-throughput microbial commu-nity analysis on the Illumina HiSeq and MiSeq platformsISME J 6 1621minus1624

Castro KM Somers BA (2012) Observations of epizooticshell disease in American lobsters Homarus americanusin southern New England J Shellfish Res 31 423minus430

Castro KM Cobb JS Gomez-Chiarri M Tlusty M (2012)Epizootic shell disease in American lobsters Homarusamericanus in southern New England past present andfuture Dis Aquat Org 100 149minus158

Chistoserdov AY Smolowitz R Mirasol F Hsu A (2005) Cul-ture-dependent characterization of the microbial com-munity associated with epizootic shell disease lesions inAmerican lobster Homarus americanus J Shellfish Res24 741minus747

Chistoserdov AY Quinn RA Gubbala SL Smolowitz R(2012) Bacterial communities associated with lesions ofshell disease in the American lobster Homarus ameri-canus Milne-Edwards J Shellfish Res 31 449minus462

Davies CE Whitten M Kim A Wootton EC and others (2014)A comparison of the structure of American (Homarusamericanus) and European (Homarus gammarus) lobster

53

Dis Aquat Org 124 41ndash54 2017

microbiome we identified OTU 0 from the candi-date genus Candidatus Thio globus and OTU 18genus Aquimarina as important to LMSD based ontheir proportional abundances (combined 27Fig 5B Table 1)

There were 13 OTUs that were unique to the tran-sitionary core microbiome accounting for 5 of thesequences recovered from this part of the shell(Fig 5AC) The presence of OTUs that are unique tothe transitionary core microbiome indicates thatthere are members of the transitionary bacterial com-munity beyond what might be expected from a sim-ple mixing of the microbiota from healthy and dis-eased lobster shell OTUs 44 78 and 160 from thegenera Marinobacter Loktanella and CandidatusEndobugula stood out as important members of thetransitionary core microbiome due to their abun-dance in both the affected and transitionary shellconditions (Fig 5C Table 1)

There were also 33 OTUs found in the core micro-biome of more than 1 shell condition Of these 33OTUs 19 were shared between the transitionary andunaffected shell (Fig 5AG) whereas only 4 OTUs

were shared between the transitionary and affectedshell (Fig 5AE) This low number of OTUs sharedbetween the core microbiomes of transitionary andaffected shell provides further evidence that the bac-teria associated with this transition zone are notmerely a result of the outward expansion of the bac-teria inhabiting affected shell There were no OTUsshared between the affected and unaffected coremicrobiome that were not also a part of the transi-tionary core microbiome which also highlights theimportance of the bacterial community found 05 cmfrom the site of LMSD as a transition between these 2shell conditions From the affected transitionary coremicrobiome overlap we identified OTU 47 from thegenus Colwellia and OTU 118 from the genus Areni-cella to be of interest based on their relative abun-dances (Fig 5E Table 1) Finally there were 10 ubi -quitous OTUs shared among the core microbiomes ofall 3 shell conditions (Fig 5AF) OTU 1 from thegenus Perspicuibacter and OTU 19 from the genusLutimonas were of particular importance due to theirabundances on the transitionary and affected shellrespectively (Fig 5F Table 1)

48

OTU Order Family Genus Identity Abundance

Affected Core0 Thioglobusa 91 150 18 Flavobacteriales Flavobacteriaceae Aquimarina 100 118 21 Flavobacteriales Flavobacteriaceae Maritimimonas 96 1129 Alteromonadales Colwelliaceae Colwellia 99 2348 Alteromonadales Colwelliaceae Colwellia 100 0375 Rhodobacterales Rhodobacteraceae Loktanella 99 06121 Rhodobacterales Hyphomonadaceae Robiginitomaculum 96 04135 Rhizobiales Methylobacteriaceae Methylobacterium 100 04159 Oceanospirillales Oceanospirillaceae Oleispira 100 01229 Rhodobacterales Rhodobacteraceae Octadecabacter 100 04238 Rhizobiales Hyphomicrobiaceae Maritalea 99 07275 Flavobacteriales Flavobacteriaceae Cloacibacterium 100 01316 Oceanospirillales Oceanospirillaceae Neptunomonas 100 041306 Flavobacteriales Flavobacteriaceae Lutimonas 98 01

Transitionary Core2 Arenicellales Arenicellaceae Perspicuibacter 99 0739 Arenicellales Arenicellaceae Arenicella 98 0344 Alteromonadales Alteromonadaceae Marinobacter 100 1553 Rhodobacterales Rhodobacteraceae Roseovarius 100 0178 Rhodobacterales Rhodobacteraceae Loktanella 100 0189 Rhizobiales Phyllobacteriaceae Pseudahrensia 98 01107 Rhodobacterales Hyphomonadaceae Fretibacter 96 02160 Cellvibrionales Cellvibrionaceae Endobugulaa 99 10254 Oceanospirillales Alcanivoracaceae Marinicella 99 04338 Chromatiales Thioalkalispiraceae Thioprofundum 94 01

Table 1 Taxonomy for core operational taxonomic units (OTUs) identified in this study Closest cultured representative for allOTUs listed in order from Fig 5 Bold OTUs are those shown in Fig 6 Abundance is the relative abundance in that shell

condition Note OTU 0 has not yet been classified at the order or family level

(Table continued on next page)

Feinman et al Bacterial communities in lobster shell disease 49

Table 1 (continued)

OTU Order Family Genus Identity Abundance

370 Burkholderiales Comamonadaceae Diaphorobacter 100 01387 Rhodothermales Rubricoccaceae Rubrivirga 98 01410 Rhizobiales Phyllobacteriaceae Aminobacter 96 01

Unaffected Core6 Rhodobacterales Rhodobacteraceae Amylibacter 99 047 Rhodobacterales Rhodobacteraceae Halocynthiibacter 99 0123 Thiotrichales Thiotrichaceae Cocleimonas 96 0688 Rhizobiales Phyllobacteriaceae Pseudahrensia 99 0193 Arenicellales Arenicellaceae Arenicella 97 05110 Rhodobacterales Hyphomonadaceae Fretibacter 96 05128 Chromatiales Woeseiaceae Woeseia 96 03177 Acidimicrobiales Acidimicrobiaceae Ilumatobacter 100 02191 Sphingobacteriales Saprospiraceae Lewinella 92 05220 Acidimicrobiales Microthrixaceae Microthrixa 90 03233 Sphingobacteriales Saprospiraceae Lewinella 91 04277 Deferribacterales Deferribacteraceae Caldithrix 91 03294 Nitrosomonadales Nitrosomonadaceae Nitrosomonas 99 02302 Incertaesedis Rhodothermaceae Rhodothermus 89 03339 Rhodobacterales Rhodobacteraceae Labrenzia 97 02360 Rhizobiales Phyllobacteriaceae Aminobacter 95 02

Affected and Transitionary Core47 Alteromonadales Colwelliaceae Colwellia 99 3373 Oceanospirillales Oceanospirillaceae Bacterioplanes 91 0377 Rhodobacterales Rhodobacteraceae Loktanella 100 04118 Arenicellales Arenicellaceae Arenicella 99 17

Ubiquitous Core1 Arenicellales Arenicellaceae Perspicuibacter 99 278 Rhodobacterales Rhodobacteraceae Amylibacter 98 029 Rhodobacterales Rhodobacteraceae Litoreibacter 100 0210 Rhodobacterales Rhodobacteraceae Jannaschia 100 0219 Flavobacteriales Flavobacteriaceae Lutimonas 96 2538 Arenicellales Arenicellaceae Arenicella 99 1652 Rhodobacterales Rhodobacteraceae Sulfitobacter 100 0371 Flavobacteriales Flavobacteriaceae Maribacter 99 0698 Rhodobacterales Hyphomonadaceae Litorimonas 97 03268 Rhizobiales Methylobacteriaceae Methylobacterium 100 02

Transitionary and Unaffected Core20 Flavobacteriales Flavobacteriaceae Lutimonas 97 0322 Thiotrichales Thiotrichaceae Cocleimonas 97 0857 Rhodobacterales Rhodobacteraceae Phaeobacter 99 0174 Verrucomicrobiales Rubritaleaceae Rubritalea 97 0682 Acidimicrobiales Microthrixaceae Microthrixa 91 0287 Rhodobacterales Rhodobacteraceae Loktanella 98 02106 Cellvibrionales Cellvibrionaceae Endobugulaa 97 10125 Planctomycetales Planctomycetaceae Blastopirellula 91 03131 Arenicellales Arenicellaceae Perspicuibacter 93 02132 Rhodobacterales Rhodobacteraceae Maribius 95 02137 Rhizobiales Hyphomicrobiaceae Hyphomicrobium 98 02140 Rhodobacterales Hyphomonadaceae Fretibacter 95 02147 Rhizobiales Hyphomicrobiaceae Filomicrobium 96 03149 Flavobacteriales Flavobacteriaceae Lutimonas 97 01150 Rhodobacterales Rhodobacteraceae Oceaniovalibus 94 02161 Sphingobacteriales Saprospiraceae Phaeodactylibacter 89 02164 Rhodobacterales Rhodobacteraceae Citreicella 93 01174 Chromatiales Thioalkalispiraceae Thioprofundum 93 02262 Chromatiales Granulosicoccaceae Granulosicoccus 97 01aCandidate classification

Dis Aquat Org 124 41ndash54 2017

DISCUSSION

High-throughput sequencing results from lobsterswith and without visible signs of LMSD (ie lobsterswith spots lobsters with lesions and lobsters that hadrecently molted and did not exhibit signs of LMSD)indicate the presence of 3 distinct bacterial commu-nities (Fig 6A) We classified these 3 distinct commu-nities as affected transitionary and unaffected shellcommunities based on weighted UniFrac analysis ofcommunity composition (Figs 2 amp 3) The presence ofa distinct bacterial community at the site of diseaseagrees with prior work indicating that bacteria arethe causative agent of shell disease (Chistoserdov etal 2012) However our small spatial scale samplingallowed a more comprehensive characterization of thebacterial community associated with LMSD enabling

us to separate out the bacterial communities associ-ated with the transition from unaffected to affectedshell and thus identifying the presence of a third bac-terial community on shells affected with LMSD thetransitionary community

Interestingly our results indicate no differentiationamong bacterial communities found on lobsters withlesions and spots in either affected or transitionaryshell This finding is counter to previously reportedgenetic fingerprinting results that showed differ-ences in the bacterial communities among lesionsand spots (Quinn et al 2012b) Based on their initialfindings Quinn et al (2012b) suggested that the bac-teria present in a spot may be important in initiatingshell disease in lobsters whereas bacteria found inlesions were secondary opportunistic colonizers thatthrived in the new conditions The lack of differenti-

50

Fig 6 Conceptual diagram of the laboratory model of shell disease (LMSD) Visual summary of the findings of this study show-ing (A) the main observations of this study and (BC) 2 different alternate models of these data though others are possible (A1)The 3 distinct bacterial communities present on diseased lobster shell unaffected transitionary and affected shell communi-ties highlighting the lack of difference between spot and lesion communities The post-molt community is fairly similar to theunaffected shell community though this community devoid of disease is fairly uniform across the sampling area (A2) The ob-served decrease in bacterial diversity as LMSD progresses (A3) Several operational taxonomic units (OTUs) that are of inter-est to LMSD based on their abundances in either the transitionary or affected shell communities (B) Alternate Model 1 Thetransitionary bacterial community though unique is unimportant to the progression of LMSD and simply represents a community that is able to live in the proximity of disease-causing bacteria (C) Alternate Model 2 The transitionary bacterial

community consists of precursory bacteria that are present before visual signs of the disease occur

Feinman et al Bacterial communities in lobster shell disease

ation between lesion and spot communities in ourdata suggests that this may not be the case The tran-sitionary shell bacterial community that we identi-fied however may be important for initiation ofLMSD while the affected shell bacterial communitycould represent secondary colonizers

In addition to identifying 3 distinct bacterial com-munities we observed a positive association be tweenbacterial diversity and distance from the site of LMSD(Figs 4 amp 6A) This is contrary to work done by Whit-ten et al (2014) which showed greater bacterialdiversity in diseased shell compared to healthy shellThese previous results however were based onmethods that recovered an average of 8 to 11 taxa persample The sequencing methods we used allowedfor a much greater depth which revealed a greaterproportion of low abundance taxa thereby increas-ing apparent diversity The reduction in bacteria atthe site of LMSD indicates that the diseased shellenvironment may be highly selective for specific bac-terial taxa perhaps due to the host response and theantibacterial effects of melanin Alternatively op -portunistic bacteria associated with the newly ex -posed chitin matrix may outcompete other bacteriathereby excluding them Since bacterial diversitywas also reduced on the transitionary shell comparedto the unaffected shell we suggest that competitionrather than environmental factors lowered bacterialdiversity at the site of LMSD It is also possible thatalong with changes in the bacterial communitychanges occurred in the physical environment of thetransitionary shell prior to the onset of disease thathave yet to be identified

The presence of a transitionary community sug-gests that it is not just the bacteria at the site of dis-ease that are altered during LMSD but that there isalso a change in the bacterial community composi-tion at distances up to 05 cm away from the site ofmelanization The transitionary community couldrepresent several possibilities the 2 most parsimo-nious being (1) an outer edge of the disease zone thatis fouled by disease causing bacteria or (2) a novelcommunity that is a precursor for pathogenic bacteriathat later infect the lobster shell These 2 alternatemodels for disease transmission depicted in Fig 6BCprovide possible explanations for the changes in bothbacterial diversity and community structure we ob -served over space and time In Alternate Model 1(Fig 6B) the transitionary and affected bacterialcommunities both appear when spots are first ob -servable on the lobster shell Here the transitionarycommunity is not involved as a causative agent ofLMSD but instead is present due to fouling from the

disease-causing community and changes in the shellenvironment that occur due to the presence of dis-ease-causing bacteria As LMSD progresses theaffected area of the shell increases (transition fromspot to lesion) until the lobster eventually molts Thismodel is in agreement with previous reports regard-ing the progression of shell disease (Chistoserdov etal 2012 Davies et al 2014 Whitten et al 2014)though the sampling schemes in these previous stud-ies did not allow for the detection of a transitionarycommunity Alternate Model 2 (Fig 6C) interpretsthe transitionary bacterial community as a precursorcommunity arriving prior to the onset of visual signsof LMSD In this model the transitionary communityrepresents initial colonizers that may cause a shift inthe lobster shell bacterial community and lower bac-terial diversity prior to visual signs of disease Thisinitial invasion creates an environment that allowsfor secondary colonization that advances the diseasestate at which time the melanization associated withshell disease becomes observable The fact that thebacterial community at 05 cm is distinct from thecommunity at 0 cm (Figs 2 amp 3) and that there arebacteria that are unique to the core microbiome at05 cm (Fig 5) offers support for this second modelThis model hinges on the existence of a currentlyunobserved state where the transitionary and un -affected communities are present prior to visual signsof disease This state will be difficult to observe infuture studies as discovery would only come fromresearchers predicting the exact location where shelldisease might occur and sampling that predictedlocation prior to visual signs of disease While both ofthese alternate models differ from the currently pro-posed model for the progression of epizootic shelldisease both are in agreement with the previouslyproposed polymicrobial nature of shell disease thatis either caused by or amplified by a dysbiosis in bacterial communities on unaffected shell (Meres etal 2012) Culture studies examining growth dynamicsof potential pathogens in the presence and absenceof members of the transitionary community couldoffer additional insights into the location and initia-tion of LMSD and could refine the models proposedhere

Given the potential importance of the transitionaryand affected shell communities to the initiation andprogression of LMSD we identified key bacterialmembers of these shell conditions from the coremicrobiome (Figs 5 amp 6A) and evaluated them interms of the 2 hypothesized models The affectedshell was dominated by just 2 OTUs 0 and 18 whichaccounted for nearly 27 of the sequences from this

51

Dis Aquat Org 124 41ndash54 2017

shell condition (Fig 5B) Given their proportionaldominance these taxa appeared to be very importantto the affected shell community The closest culturedrepresentative to OTU 0 was from Candidatus Thio -globus singularis a mixotrophic bacterium belong-ing to the SUP05 clade known for its potential to oxi-dize sulfur and fix carbon in the dark ocean and inlow oxygen zones (Marshall amp Morris 2015) This cul-tured representative however only shared 91sequence identity with OTU 0 Therefore this OTUlikely represents a new uncultured species As thiswas the most abundant OTU from both the affectedand transitionary shell (see Table S1 in the Supple-ment) better identification and classification of thisbacterium is important for understanding its role inthe progression of LMSD OTU 18 the second mostabundant OTU on the affected shell (see Table S1)shared 100 sequence identity with a representa-tive from the genus Aquimarina (Esteves et al 2013)and 99 sequence identity with bacteria re coveredfrom lobster shells after having been exposed toAquimarina lsquohomariarsquo (Quinn et al 2012a) Aquima-rina lsquohomariarsquo has been identified by several studiesas a potential causative agent of epizootic shell dis-ease as it can readily be recovered from lesions(Chistoserdov et al 2005 2012 Quinn et al 2012aWhitten et al 2014) The abundance of OTU 18 onthe affected shell in this study (12 Fig 5B) con-firms the potential importance of Aquimarina lsquohomariarsquoto LMSD though its lack of dominance on the transi-tionary shell (2 Fig 5B) suggests that it is not assuccessful outside of melanized areas of the shell (seeTable S1)

In the transitionary core microbiome (Fig 5C)OTUs 44 and 160 appeared to be particularly impor-tant OTU 44 was one of the most abundant OTUs inall shell conditions however it was only consideredcore to the transitionary shell (Table 1 see Table S1)This OTU was most closely related to Marinobactersalsuginis (100 sequence identity S V Law et alunpubl data accession no KT921331) Other speciesfrom this genus have proteins for the deacetylation ofchitin (Song et al 2013) which suggests that OTU 44may be important to initiation of disease creatingconditions under which opportunistic colonizers asso-ciated with LMSD might thrive (Malloy 1978 Chisto -serdov et al 2005) OTU 160 was most closely relatedto the candidate species Endobugula sertula (99sequence identity McGovern amp Hellberg 2003) abryozoan symbiont that produces polyketides to pro-tect the host larvae from predation (Lindquist amp Hay1996 Lopanik et al 2004) In addition to OTUs 44and 160 OTU 78 was part of the transitionary core

microbiome though quantitatively it appeared to bemore important to the affected shell Although thisOTU was not a member of the affected shell coremicrobiome it was present in 13 of the 16 affectedshell samples and was present at relatively highabundance in 7 of those samples This OTU genusLoktanella which shared 100 sequence identitywith bacteria recovered from other diseased lobstershells (Quinn et al 2012a) may therefore be in fectingthe shell alongside early colonizers and in some in -stances growing in abundance as LMSD progresses

Of the 4 OTUs shared between the affected andtransitionary core microbiomes (Fig 5E) OTUs 47and 118 were numerically dominant in both commu-nities OTU 47 belonged to the genus Colwelliawhich Meres et al (2012) reported as one of the manygroups of species that helped to differentiate dis-eased and healthy lobster shell In the current studythis bacterium was abundant on transitionary shellan area of the shell that would have previously re -mained unsampled indicating that members of thisgenus may be more important in promoting LMSDthan previously thought if the transitionary shellindeed reflects an initial invasion OTU 118 was mostclosely aligned with Arenicella chitinivorans (99sequence identity) a bacterium that differs from oth-ers in the same genus by its ability to degrade chitin(Nedashkovskaya et al 2013) Given that previouswork by Whitten et al (2014) also recovered mem-bers of this genus from lesioned areas of the lobstershell it appears that the ability to degrade chitin maybe an important characteristic allowing bacteria tothrive on both the affected and transitionary shell

Many studies have reported the incidence of ubiq-uitous bacteria on the lobster shell both on diseasedand healthy shell (Chistoserdov et al 2005 Meres etal 2012 Whitten et al 2014) In this study there were10 OTUs present in the core microbiome of all 3 shellconditions (Fig 5F) and OTUs 1 and 19 were propor-tionally dominant on the transitionary and affectedshell respectively though they were highly abun-dant in all 3 shell conditions (see Table S1) OTU 1 ismost likely Perspicuibacter marinus (99 sequenceidentity) an aerobic marine bacterium recently iso-lated from surface seawater in Japan (Teramoto et al2015) At this time not much is known about thisorganism but it has been cultured and thereforecould be used in future studies exploring the dynam-ics of the transitionary lobster shell community OTU19 an important shell bacterium found on affectedshell was most similar to other bacteria found inassociation with sponges (99 sequence identitySipkema et al 2011) It is likely that this bacterium is

52

Feinman et al Bacterial communities in lobster shell disease

from the genus Lutimonas (96 sequence identityBuerger et al 2012) which is part of the same familyas Aquimarina and may represent another oppor-tunistic colonizer of diseased lobster shell

In addition to these core microbiome memberswhich appear to be important due to their relativeabundance on either the transitionary or affectedshell there were a number of other core microbiomeOTUs that had identities similar to taxa identified inprevious studies of epizootic shell disease Severalstudies have reported Aquimarina lsquohomariarsquo as apotential causative agent in shell disease including abacterial challenge study that was able to recoverthis bacterium from lobsters with lesions (Quinn et al2012a) and a next-generation sequencing study thatexamined healthy and diseased shell (Meres et al2012) Although we only recovered a single OTU thatwas associated with the genus Aquimarina (OTU 18)we recovered several other OTUs including Lutimo -nas that were from the Flavobacteria family (Table 1)Several studies have reported that members of thisfamily aside from Aquimarina are important to epi-zootic shell disease (Chistoserdov et al 2012 Mereset al 2012 Quinn et al 2012b) In addition to Aqui -marina the lsquoThalassobiusrsquo genus has been associatedwith epizootic shell disease (Meres et al 2012 Quinnet al 2012a) Although we did not recover any bacte-ria from this genus we recovered several bacteriafrom the Rhodobacteraceae family including previ-ously observed bacteria from the Loktanella and Jan-naschia genera (Chistoserdov et al 2012 Meres et al2012) In addition to Aquimarina Loktanella andJannaschia Meres et al (2012) also reported bacteriafrom the genus Cardiobacterium and suborder Micro -coccineae as some of the most abundant bacteria intheir study We did not recover any bacteria fromeither of these groups nor their respective ordersThese differences could be due to the taxonomicidentification tools used in different studies but mayalso reflect differences in wild versus captive-raisedlobsters

In conclusion using a fine spatial scale we charac-terized changes in bacterial community compositionthat occur during the progression of LMSD Wedemonstrated that affected areas of lobster shell havealtered bacterial community composition and reducedbacterial diversity compared to unaffected shell Wealso demonstrated that areas close to the site of dis-ease but not yet melanized have altered bacterialcommunities and reduced bacterial diversity in com-parison to unaffected shell and represent a transi-tionary community that may be important in the initi-ation of LMSD We identified key bacteria associated

with each stage of the disease and proposed 2 differ-ent conceptual models one that suggests that thetransitionary bacterial community arises due to foul-ing and another that suggests the transitionary shellharbors bacteria important to the initiation of LMSDUnderstanding the full complement of bacteria asso-ciated with affected transitionary and asymptomaticlobster shells is the first step toward understandingthe role of the transitionary bacterial community andhelping us develop mechanisms to diminish thespread and severity of this disease

Acknowledgements We thank Anita Kim from the NewEngland Aquarium Lobster Research and Rearing Facilityfor caring for the animals used in this study and Jeff Dusen-berry for the use of the supercomputing facilities managedby the Research Computing Department at the University ofMassachusetts Boston We also thank Rachel Vincent anundergraduate assistant on this research as well as PatrickKearns John Angell and all the members of the BowenLaboratory Funding for this project was provided by theSanofi Genzyme Doctoral Research Fellowship (SGF) theUniversity of Massachusetts Boston Doctoral DissertationResearch Grant (SGF) and the National Science Founda-tion Research Experience for Undergraduates Award DBI-1359241 to Dr Rachel Skvirsky

LITERATURE CITED

Anderson MJ (2001) A new method for non-parametric mul-tivariate analysis of variance Austral Ecol 26 32minus46

Buerger S Spoering A Gavrish E Leslin C Ling L EpsteinSS (2012) Microbial scout hypothesis and microbial dis-covery Appl Environ Microbiol 78 3229minus3233

Caporaso JG Kuczynski J Stombaugh J Bittinger K andothers (2010) QIIME allows analysis of high-throughputcommunity sequencing data Nat Methods 7 335minus336

Caporaso JG Lauber CL Walters WA Berg-Lyons D andothers (2012) Ultra-high-throughput microbial commu-nity analysis on the Illumina HiSeq and MiSeq platformsISME J 6 1621minus1624

Castro KM Somers BA (2012) Observations of epizooticshell disease in American lobsters Homarus americanusin southern New England J Shellfish Res 31 423minus430

Castro KM Cobb JS Gomez-Chiarri M Tlusty M (2012)Epizootic shell disease in American lobsters Homarusamericanus in southern New England past present andfuture Dis Aquat Org 100 149minus158

Chistoserdov AY Smolowitz R Mirasol F Hsu A (2005) Cul-ture-dependent characterization of the microbial com-munity associated with epizootic shell disease lesions inAmerican lobster Homarus americanus J Shellfish Res24 741minus747

Chistoserdov AY Quinn RA Gubbala SL Smolowitz R(2012) Bacterial communities associated with lesions ofshell disease in the American lobster Homarus ameri-canus Milne-Edwards J Shellfish Res 31 449minus462

Davies CE Whitten M Kim A Wootton EC and others (2014)A comparison of the structure of American (Homarusamericanus) and European (Homarus gammarus) lobster

53

Feinman et al Bacterial communities in lobster shell disease 49

Table 1 (continued)

OTU Order Family Genus Identity Abundance

370 Burkholderiales Comamonadaceae Diaphorobacter 100 01387 Rhodothermales Rubricoccaceae Rubrivirga 98 01410 Rhizobiales Phyllobacteriaceae Aminobacter 96 01

Unaffected Core6 Rhodobacterales Rhodobacteraceae Amylibacter 99 047 Rhodobacterales Rhodobacteraceae Halocynthiibacter 99 0123 Thiotrichales Thiotrichaceae Cocleimonas 96 0688 Rhizobiales Phyllobacteriaceae Pseudahrensia 99 0193 Arenicellales Arenicellaceae Arenicella 97 05110 Rhodobacterales Hyphomonadaceae Fretibacter 96 05128 Chromatiales Woeseiaceae Woeseia 96 03177 Acidimicrobiales Acidimicrobiaceae Ilumatobacter 100 02191 Sphingobacteriales Saprospiraceae Lewinella 92 05220 Acidimicrobiales Microthrixaceae Microthrixa 90 03233 Sphingobacteriales Saprospiraceae Lewinella 91 04277 Deferribacterales Deferribacteraceae Caldithrix 91 03294 Nitrosomonadales Nitrosomonadaceae Nitrosomonas 99 02302 Incertaesedis Rhodothermaceae Rhodothermus 89 03339 Rhodobacterales Rhodobacteraceae Labrenzia 97 02360 Rhizobiales Phyllobacteriaceae Aminobacter 95 02

Affected and Transitionary Core47 Alteromonadales Colwelliaceae Colwellia 99 3373 Oceanospirillales Oceanospirillaceae Bacterioplanes 91 0377 Rhodobacterales Rhodobacteraceae Loktanella 100 04118 Arenicellales Arenicellaceae Arenicella 99 17

Ubiquitous Core1 Arenicellales Arenicellaceae Perspicuibacter 99 278 Rhodobacterales Rhodobacteraceae Amylibacter 98 029 Rhodobacterales Rhodobacteraceae Litoreibacter 100 0210 Rhodobacterales Rhodobacteraceae Jannaschia 100 0219 Flavobacteriales Flavobacteriaceae Lutimonas 96 2538 Arenicellales Arenicellaceae Arenicella 99 1652 Rhodobacterales Rhodobacteraceae Sulfitobacter 100 0371 Flavobacteriales Flavobacteriaceae Maribacter 99 0698 Rhodobacterales Hyphomonadaceae Litorimonas 97 03268 Rhizobiales Methylobacteriaceae Methylobacterium 100 02

Transitionary and Unaffected Core20 Flavobacteriales Flavobacteriaceae Lutimonas 97 0322 Thiotrichales Thiotrichaceae Cocleimonas 97 0857 Rhodobacterales Rhodobacteraceae Phaeobacter 99 0174 Verrucomicrobiales Rubritaleaceae Rubritalea 97 0682 Acidimicrobiales Microthrixaceae Microthrixa 91 0287 Rhodobacterales Rhodobacteraceae Loktanella 98 02106 Cellvibrionales Cellvibrionaceae Endobugulaa 97 10125 Planctomycetales Planctomycetaceae Blastopirellula 91 03131 Arenicellales Arenicellaceae Perspicuibacter 93 02132 Rhodobacterales Rhodobacteraceae Maribius 95 02137 Rhizobiales Hyphomicrobiaceae Hyphomicrobium 98 02140 Rhodobacterales Hyphomonadaceae Fretibacter 95 02147 Rhizobiales Hyphomicrobiaceae Filomicrobium 96 03149 Flavobacteriales Flavobacteriaceae Lutimonas 97 01150 Rhodobacterales Rhodobacteraceae Oceaniovalibus 94 02161 Sphingobacteriales Saprospiraceae Phaeodactylibacter 89 02164 Rhodobacterales Rhodobacteraceae Citreicella 93 01174 Chromatiales Thioalkalispiraceae Thioprofundum 93 02262 Chromatiales Granulosicoccaceae Granulosicoccus 97 01aCandidate classification

Dis Aquat Org 124 41ndash54 2017

DISCUSSION

High-throughput sequencing results from lobsterswith and without visible signs of LMSD (ie lobsterswith spots lobsters with lesions and lobsters that hadrecently molted and did not exhibit signs of LMSD)indicate the presence of 3 distinct bacterial commu-nities (Fig 6A) We classified these 3 distinct commu-nities as affected transitionary and unaffected shellcommunities based on weighted UniFrac analysis ofcommunity composition (Figs 2 amp 3) The presence ofa distinct bacterial community at the site of diseaseagrees with prior work indicating that bacteria arethe causative agent of shell disease (Chistoserdov etal 2012) However our small spatial scale samplingallowed a more comprehensive characterization of thebacterial community associated with LMSD enabling

us to separate out the bacterial communities associ-ated with the transition from unaffected to affectedshell and thus identifying the presence of a third bac-terial community on shells affected with LMSD thetransitionary community

Interestingly our results indicate no differentiationamong bacterial communities found on lobsters withlesions and spots in either affected or transitionaryshell This finding is counter to previously reportedgenetic fingerprinting results that showed differ-ences in the bacterial communities among lesionsand spots (Quinn et al 2012b) Based on their initialfindings Quinn et al (2012b) suggested that the bac-teria present in a spot may be important in initiatingshell disease in lobsters whereas bacteria found inlesions were secondary opportunistic colonizers thatthrived in the new conditions The lack of differenti-

50

Fig 6 Conceptual diagram of the laboratory model of shell disease (LMSD) Visual summary of the findings of this study show-ing (A) the main observations of this study and (BC) 2 different alternate models of these data though others are possible (A1)The 3 distinct bacterial communities present on diseased lobster shell unaffected transitionary and affected shell communi-ties highlighting the lack of difference between spot and lesion communities The post-molt community is fairly similar to theunaffected shell community though this community devoid of disease is fairly uniform across the sampling area (A2) The ob-served decrease in bacterial diversity as LMSD progresses (A3) Several operational taxonomic units (OTUs) that are of inter-est to LMSD based on their abundances in either the transitionary or affected shell communities (B) Alternate Model 1 Thetransitionary bacterial community though unique is unimportant to the progression of LMSD and simply represents a community that is able to live in the proximity of disease-causing bacteria (C) Alternate Model 2 The transitionary bacterial

community consists of precursory bacteria that are present before visual signs of the disease occur

Feinman et al Bacterial communities in lobster shell disease

ation between lesion and spot communities in ourdata suggests that this may not be the case The tran-sitionary shell bacterial community that we identi-fied however may be important for initiation ofLMSD while the affected shell bacterial communitycould represent secondary colonizers

In addition to identifying 3 distinct bacterial com-munities we observed a positive association be tweenbacterial diversity and distance from the site of LMSD(Figs 4 amp 6A) This is contrary to work done by Whit-ten et al (2014) which showed greater bacterialdiversity in diseased shell compared to healthy shellThese previous results however were based onmethods that recovered an average of 8 to 11 taxa persample The sequencing methods we used allowedfor a much greater depth which revealed a greaterproportion of low abundance taxa thereby increas-ing apparent diversity The reduction in bacteria atthe site of LMSD indicates that the diseased shellenvironment may be highly selective for specific bac-terial taxa perhaps due to the host response and theantibacterial effects of melanin Alternatively op -portunistic bacteria associated with the newly ex -posed chitin matrix may outcompete other bacteriathereby excluding them Since bacterial diversitywas also reduced on the transitionary shell comparedto the unaffected shell we suggest that competitionrather than environmental factors lowered bacterialdiversity at the site of LMSD It is also possible thatalong with changes in the bacterial communitychanges occurred in the physical environment of thetransitionary shell prior to the onset of disease thathave yet to be identified

The presence of a transitionary community sug-gests that it is not just the bacteria at the site of dis-ease that are altered during LMSD but that there isalso a change in the bacterial community composi-tion at distances up to 05 cm away from the site ofmelanization The transitionary community couldrepresent several possibilities the 2 most parsimo-nious being (1) an outer edge of the disease zone thatis fouled by disease causing bacteria or (2) a novelcommunity that is a precursor for pathogenic bacteriathat later infect the lobster shell These 2 alternatemodels for disease transmission depicted in Fig 6BCprovide possible explanations for the changes in bothbacterial diversity and community structure we ob -served over space and time In Alternate Model 1(Fig 6B) the transitionary and affected bacterialcommunities both appear when spots are first ob -servable on the lobster shell Here the transitionarycommunity is not involved as a causative agent ofLMSD but instead is present due to fouling from the

disease-causing community and changes in the shellenvironment that occur due to the presence of dis-ease-causing bacteria As LMSD progresses theaffected area of the shell increases (transition fromspot to lesion) until the lobster eventually molts Thismodel is in agreement with previous reports regard-ing the progression of shell disease (Chistoserdov etal 2012 Davies et al 2014 Whitten et al 2014)though the sampling schemes in these previous stud-ies did not allow for the detection of a transitionarycommunity Alternate Model 2 (Fig 6C) interpretsthe transitionary bacterial community as a precursorcommunity arriving prior to the onset of visual signsof LMSD In this model the transitionary communityrepresents initial colonizers that may cause a shift inthe lobster shell bacterial community and lower bac-terial diversity prior to visual signs of disease Thisinitial invasion creates an environment that allowsfor secondary colonization that advances the diseasestate at which time the melanization associated withshell disease becomes observable The fact that thebacterial community at 05 cm is distinct from thecommunity at 0 cm (Figs 2 amp 3) and that there arebacteria that are unique to the core microbiome at05 cm (Fig 5) offers support for this second modelThis model hinges on the existence of a currentlyunobserved state where the transitionary and un -affected communities are present prior to visual signsof disease This state will be difficult to observe infuture studies as discovery would only come fromresearchers predicting the exact location where shelldisease might occur and sampling that predictedlocation prior to visual signs of disease While both ofthese alternate models differ from the currently pro-posed model for the progression of epizootic shelldisease both are in agreement with the previouslyproposed polymicrobial nature of shell disease thatis either caused by or amplified by a dysbiosis in bacterial communities on unaffected shell (Meres etal 2012) Culture studies examining growth dynamicsof potential pathogens in the presence and absenceof members of the transitionary community couldoffer additional insights into the location and initia-tion of LMSD and could refine the models proposedhere

Given the potential importance of the transitionaryand affected shell communities to the initiation andprogression of LMSD we identified key bacterialmembers of these shell conditions from the coremicrobiome (Figs 5 amp 6A) and evaluated them interms of the 2 hypothesized models The affectedshell was dominated by just 2 OTUs 0 and 18 whichaccounted for nearly 27 of the sequences from this

51

Dis Aquat Org 124 41ndash54 2017

shell condition (Fig 5B) Given their proportionaldominance these taxa appeared to be very importantto the affected shell community The closest culturedrepresentative to OTU 0 was from Candidatus Thio -globus singularis a mixotrophic bacterium belong-ing to the SUP05 clade known for its potential to oxi-dize sulfur and fix carbon in the dark ocean and inlow oxygen zones (Marshall amp Morris 2015) This cul-tured representative however only shared 91sequence identity with OTU 0 Therefore this OTUlikely represents a new uncultured species As thiswas the most abundant OTU from both the affectedand transitionary shell (see Table S1 in the Supple-ment) better identification and classification of thisbacterium is important for understanding its role inthe progression of LMSD OTU 18 the second mostabundant OTU on the affected shell (see Table S1)shared 100 sequence identity with a representa-tive from the genus Aquimarina (Esteves et al 2013)and 99 sequence identity with bacteria re coveredfrom lobster shells after having been exposed toAquimarina lsquohomariarsquo (Quinn et al 2012a) Aquima-rina lsquohomariarsquo has been identified by several studiesas a potential causative agent of epizootic shell dis-ease as it can readily be recovered from lesions(Chistoserdov et al 2005 2012 Quinn et al 2012aWhitten et al 2014) The abundance of OTU 18 onthe affected shell in this study (12 Fig 5B) con-firms the potential importance of Aquimarina lsquohomariarsquoto LMSD though its lack of dominance on the transi-tionary shell (2 Fig 5B) suggests that it is not assuccessful outside of melanized areas of the shell (seeTable S1)

In the transitionary core microbiome (Fig 5C)OTUs 44 and 160 appeared to be particularly impor-tant OTU 44 was one of the most abundant OTUs inall shell conditions however it was only consideredcore to the transitionary shell (Table 1 see Table S1)This OTU was most closely related to Marinobactersalsuginis (100 sequence identity S V Law et alunpubl data accession no KT921331) Other speciesfrom this genus have proteins for the deacetylation ofchitin (Song et al 2013) which suggests that OTU 44may be important to initiation of disease creatingconditions under which opportunistic colonizers asso-ciated with LMSD might thrive (Malloy 1978 Chisto -serdov et al 2005) OTU 160 was most closely relatedto the candidate species Endobugula sertula (99sequence identity McGovern amp Hellberg 2003) abryozoan symbiont that produces polyketides to pro-tect the host larvae from predation (Lindquist amp Hay1996 Lopanik et al 2004) In addition to OTUs 44and 160 OTU 78 was part of the transitionary core

microbiome though quantitatively it appeared to bemore important to the affected shell Although thisOTU was not a member of the affected shell coremicrobiome it was present in 13 of the 16 affectedshell samples and was present at relatively highabundance in 7 of those samples This OTU genusLoktanella which shared 100 sequence identitywith bacteria recovered from other diseased lobstershells (Quinn et al 2012a) may therefore be in fectingthe shell alongside early colonizers and in some in -stances growing in abundance as LMSD progresses

Of the 4 OTUs shared between the affected andtransitionary core microbiomes (Fig 5E) OTUs 47and 118 were numerically dominant in both commu-nities OTU 47 belonged to the genus Colwelliawhich Meres et al (2012) reported as one of the manygroups of species that helped to differentiate dis-eased and healthy lobster shell In the current studythis bacterium was abundant on transitionary shellan area of the shell that would have previously re -mained unsampled indicating that members of thisgenus may be more important in promoting LMSDthan previously thought if the transitionary shellindeed reflects an initial invasion OTU 118 was mostclosely aligned with Arenicella chitinivorans (99sequence identity) a bacterium that differs from oth-ers in the same genus by its ability to degrade chitin(Nedashkovskaya et al 2013) Given that previouswork by Whitten et al (2014) also recovered mem-bers of this genus from lesioned areas of the lobstershell it appears that the ability to degrade chitin maybe an important characteristic allowing bacteria tothrive on both the affected and transitionary shell

Many studies have reported the incidence of ubiq-uitous bacteria on the lobster shell both on diseasedand healthy shell (Chistoserdov et al 2005 Meres etal 2012 Whitten et al 2014) In this study there were10 OTUs present in the core microbiome of all 3 shellconditions (Fig 5F) and OTUs 1 and 19 were propor-tionally dominant on the transitionary and affectedshell respectively though they were highly abun-dant in all 3 shell conditions (see Table S1) OTU 1 ismost likely Perspicuibacter marinus (99 sequenceidentity) an aerobic marine bacterium recently iso-lated from surface seawater in Japan (Teramoto et al2015) At this time not much is known about thisorganism but it has been cultured and thereforecould be used in future studies exploring the dynam-ics of the transitionary lobster shell community OTU19 an important shell bacterium found on affectedshell was most similar to other bacteria found inassociation with sponges (99 sequence identitySipkema et al 2011) It is likely that this bacterium is

52

Feinman et al Bacterial communities in lobster shell disease

from the genus Lutimonas (96 sequence identityBuerger et al 2012) which is part of the same familyas Aquimarina and may represent another oppor-tunistic colonizer of diseased lobster shell

In addition to these core microbiome memberswhich appear to be important due to their relativeabundance on either the transitionary or affectedshell there were a number of other core microbiomeOTUs that had identities similar to taxa identified inprevious studies of epizootic shell disease Severalstudies have reported Aquimarina lsquohomariarsquo as apotential causative agent in shell disease including abacterial challenge study that was able to recoverthis bacterium from lobsters with lesions (Quinn et al2012a) and a next-generation sequencing study thatexamined healthy and diseased shell (Meres et al2012) Although we only recovered a single OTU thatwas associated with the genus Aquimarina (OTU 18)we recovered several other OTUs including Lutimo -nas that were from the Flavobacteria family (Table 1)Several studies have reported that members of thisfamily aside from Aquimarina are important to epi-zootic shell disease (Chistoserdov et al 2012 Mereset al 2012 Quinn et al 2012b) In addition to Aqui -marina the lsquoThalassobiusrsquo genus has been associatedwith epizootic shell disease (Meres et al 2012 Quinnet al 2012a) Although we did not recover any bacte-ria from this genus we recovered several bacteriafrom the Rhodobacteraceae family including previ-ously observed bacteria from the Loktanella and Jan-naschia genera (Chistoserdov et al 2012 Meres et al2012) In addition to Aquimarina Loktanella andJannaschia Meres et al (2012) also reported bacteriafrom the genus Cardiobacterium and suborder Micro -coccineae as some of the most abundant bacteria intheir study We did not recover any bacteria fromeither of these groups nor their respective ordersThese differences could be due to the taxonomicidentification tools used in different studies but mayalso reflect differences in wild versus captive-raisedlobsters

In conclusion using a fine spatial scale we charac-terized changes in bacterial community compositionthat occur during the progression of LMSD Wedemonstrated that affected areas of lobster shell havealtered bacterial community composition and reducedbacterial diversity compared to unaffected shell Wealso demonstrated that areas close to the site of dis-ease but not yet melanized have altered bacterialcommunities and reduced bacterial diversity in com-parison to unaffected shell and represent a transi-tionary community that may be important in the initi-ation of LMSD We identified key bacteria associated

with each stage of the disease and proposed 2 differ-ent conceptual models one that suggests that thetransitionary bacterial community arises due to foul-ing and another that suggests the transitionary shellharbors bacteria important to the initiation of LMSDUnderstanding the full complement of bacteria asso-ciated with affected transitionary and asymptomaticlobster shells is the first step toward understandingthe role of the transitionary bacterial community andhelping us develop mechanisms to diminish thespread and severity of this disease

Acknowledgements We thank Anita Kim from the NewEngland Aquarium Lobster Research and Rearing Facilityfor caring for the animals used in this study and Jeff Dusen-berry for the use of the supercomputing facilities managedby the Research Computing Department at the University ofMassachusetts Boston We also thank Rachel Vincent anundergraduate assistant on this research as well as PatrickKearns John Angell and all the members of the BowenLaboratory Funding for this project was provided by theSanofi Genzyme Doctoral Research Fellowship (SGF) theUniversity of Massachusetts Boston Doctoral DissertationResearch Grant (SGF) and the National Science Founda-tion Research Experience for Undergraduates Award DBI-1359241 to Dr Rachel Skvirsky

LITERATURE CITED

Anderson MJ (2001) A new method for non-parametric mul-tivariate analysis of variance Austral Ecol 26 32minus46

Buerger S Spoering A Gavrish E Leslin C Ling L EpsteinSS (2012) Microbial scout hypothesis and microbial dis-covery Appl Environ Microbiol 78 3229minus3233

Caporaso JG Kuczynski J Stombaugh J Bittinger K andothers (2010) QIIME allows analysis of high-throughputcommunity sequencing data Nat Methods 7 335minus336

Caporaso JG Lauber CL Walters WA Berg-Lyons D andothers (2012) Ultra-high-throughput microbial commu-nity analysis on the Illumina HiSeq and MiSeq platformsISME J 6 1621minus1624

Castro KM Somers BA (2012) Observations of epizooticshell disease in American lobsters Homarus americanusin southern New England J Shellfish Res 31 423minus430

Castro KM Cobb JS Gomez-Chiarri M Tlusty M (2012)Epizootic shell disease in American lobsters Homarusamericanus in southern New England past present andfuture Dis Aquat Org 100 149minus158

Chistoserdov AY Smolowitz R Mirasol F Hsu A (2005) Cul-ture-dependent characterization of the microbial com-munity associated with epizootic shell disease lesions inAmerican lobster Homarus americanus J Shellfish Res24 741minus747

Chistoserdov AY Quinn RA Gubbala SL Smolowitz R(2012) Bacterial communities associated with lesions ofshell disease in the American lobster Homarus ameri-canus Milne-Edwards J Shellfish Res 31 449minus462

Davies CE Whitten M Kim A Wootton EC and others (2014)A comparison of the structure of American (Homarusamericanus) and European (Homarus gammarus) lobster

53

Dis Aquat Org 124 41ndash54 2017

DISCUSSION

High-throughput sequencing results from lobsterswith and without visible signs of LMSD (ie lobsterswith spots lobsters with lesions and lobsters that hadrecently molted and did not exhibit signs of LMSD)indicate the presence of 3 distinct bacterial commu-nities (Fig 6A) We classified these 3 distinct commu-nities as affected transitionary and unaffected shellcommunities based on weighted UniFrac analysis ofcommunity composition (Figs 2 amp 3) The presence ofa distinct bacterial community at the site of diseaseagrees with prior work indicating that bacteria arethe causative agent of shell disease (Chistoserdov etal 2012) However our small spatial scale samplingallowed a more comprehensive characterization of thebacterial community associated with LMSD enabling

us to separate out the bacterial communities associ-ated with the transition from unaffected to affectedshell and thus identifying the presence of a third bac-terial community on shells affected with LMSD thetransitionary community

Interestingly our results indicate no differentiationamong bacterial communities found on lobsters withlesions and spots in either affected or transitionaryshell This finding is counter to previously reportedgenetic fingerprinting results that showed differ-ences in the bacterial communities among lesionsand spots (Quinn et al 2012b) Based on their initialfindings Quinn et al (2012b) suggested that the bac-teria present in a spot may be important in initiatingshell disease in lobsters whereas bacteria found inlesions were secondary opportunistic colonizers thatthrived in the new conditions The lack of differenti-

50

Fig 6 Conceptual diagram of the laboratory model of shell disease (LMSD) Visual summary of the findings of this study show-ing (A) the main observations of this study and (BC) 2 different alternate models of these data though others are possible (A1)The 3 distinct bacterial communities present on diseased lobster shell unaffected transitionary and affected shell communi-ties highlighting the lack of difference between spot and lesion communities The post-molt community is fairly similar to theunaffected shell community though this community devoid of disease is fairly uniform across the sampling area (A2) The ob-served decrease in bacterial diversity as LMSD progresses (A3) Several operational taxonomic units (OTUs) that are of inter-est to LMSD based on their abundances in either the transitionary or affected shell communities (B) Alternate Model 1 Thetransitionary bacterial community though unique is unimportant to the progression of LMSD and simply represents a community that is able to live in the proximity of disease-causing bacteria (C) Alternate Model 2 The transitionary bacterial

community consists of precursory bacteria that are present before visual signs of the disease occur

Feinman et al Bacterial communities in lobster shell disease

ation between lesion and spot communities in ourdata suggests that this may not be the case The tran-sitionary shell bacterial community that we identi-fied however may be important for initiation ofLMSD while the affected shell bacterial communitycould represent secondary colonizers

In addition to identifying 3 distinct bacterial com-munities we observed a positive association be tweenbacterial diversity and distance from the site of LMSD(Figs 4 amp 6A) This is contrary to work done by Whit-ten et al (2014) which showed greater bacterialdiversity in diseased shell compared to healthy shellThese previous results however were based onmethods that recovered an average of 8 to 11 taxa persample The sequencing methods we used allowedfor a much greater depth which revealed a greaterproportion of low abundance taxa thereby increas-ing apparent diversity The reduction in bacteria atthe site of LMSD indicates that the diseased shellenvironment may be highly selective for specific bac-terial taxa perhaps due to the host response and theantibacterial effects of melanin Alternatively op -portunistic bacteria associated with the newly ex -posed chitin matrix may outcompete other bacteriathereby excluding them Since bacterial diversitywas also reduced on the transitionary shell comparedto the unaffected shell we suggest that competitionrather than environmental factors lowered bacterialdiversity at the site of LMSD It is also possible thatalong with changes in the bacterial communitychanges occurred in the physical environment of thetransitionary shell prior to the onset of disease thathave yet to be identified

The presence of a transitionary community sug-gests that it is not just the bacteria at the site of dis-ease that are altered during LMSD but that there isalso a change in the bacterial community composi-tion at distances up to 05 cm away from the site ofmelanization The transitionary community couldrepresent several possibilities the 2 most parsimo-nious being (1) an outer edge of the disease zone thatis fouled by disease causing bacteria or (2) a novelcommunity that is a precursor for pathogenic bacteriathat later infect the lobster shell These 2 alternatemodels for disease transmission depicted in Fig 6BCprovide possible explanations for the changes in bothbacterial diversity and community structure we ob -served over space and time In Alternate Model 1(Fig 6B) the transitionary and affected bacterialcommunities both appear when spots are first ob -servable on the lobster shell Here the transitionarycommunity is not involved as a causative agent ofLMSD but instead is present due to fouling from the

disease-causing community and changes in the shellenvironment that occur due to the presence of dis-ease-causing bacteria As LMSD progresses theaffected area of the shell increases (transition fromspot to lesion) until the lobster eventually molts Thismodel is in agreement with previous reports regard-ing the progression of shell disease (Chistoserdov etal 2012 Davies et al 2014 Whitten et al 2014)though the sampling schemes in these previous stud-ies did not allow for the detection of a transitionarycommunity Alternate Model 2 (Fig 6C) interpretsthe transitionary bacterial community as a precursorcommunity arriving prior to the onset of visual signsof LMSD In this model the transitionary communityrepresents initial colonizers that may cause a shift inthe lobster shell bacterial community and lower bac-terial diversity prior to visual signs of disease Thisinitial invasion creates an environment that allowsfor secondary colonization that advances the diseasestate at which time the melanization associated withshell disease becomes observable The fact that thebacterial community at 05 cm is distinct from thecommunity at 0 cm (Figs 2 amp 3) and that there arebacteria that are unique to the core microbiome at05 cm (Fig 5) offers support for this second modelThis model hinges on the existence of a currentlyunobserved state where the transitionary and un -affected communities are present prior to visual signsof disease This state will be difficult to observe infuture studies as discovery would only come fromresearchers predicting the exact location where shelldisease might occur and sampling that predictedlocation prior to visual signs of disease While both ofthese alternate models differ from the currently pro-posed model for the progression of epizootic shelldisease both are in agreement with the previouslyproposed polymicrobial nature of shell disease thatis either caused by or amplified by a dysbiosis in bacterial communities on unaffected shell (Meres etal 2012) Culture studies examining growth dynamicsof potential pathogens in the presence and absenceof members of the transitionary community couldoffer additional insights into the location and initia-tion of LMSD and could refine the models proposedhere

Given the potential importance of the transitionaryand affected shell communities to the initiation andprogression of LMSD we identified key bacterialmembers of these shell conditions from the coremicrobiome (Figs 5 amp 6A) and evaluated them interms of the 2 hypothesized models The affectedshell was dominated by just 2 OTUs 0 and 18 whichaccounted for nearly 27 of the sequences from this

51

Dis Aquat Org 124 41ndash54 2017

shell condition (Fig 5B) Given their proportionaldominance these taxa appeared to be very importantto the affected shell community The closest culturedrepresentative to OTU 0 was from Candidatus Thio -globus singularis a mixotrophic bacterium belong-ing to the SUP05 clade known for its potential to oxi-dize sulfur and fix carbon in the dark ocean and inlow oxygen zones (Marshall amp Morris 2015) This cul-tured representative however only shared 91sequence identity with OTU 0 Therefore this OTUlikely represents a new uncultured species As thiswas the most abundant OTU from both the affectedand transitionary shell (see Table S1 in the Supple-ment) better identification and classification of thisbacterium is important for understanding its role inthe progression of LMSD OTU 18 the second mostabundant OTU on the affected shell (see Table S1)shared 100 sequence identity with a representa-tive from the genus Aquimarina (Esteves et al 2013)and 99 sequence identity with bacteria re coveredfrom lobster shells after having been exposed toAquimarina lsquohomariarsquo (Quinn et al 2012a) Aquima-rina lsquohomariarsquo has been identified by several studiesas a potential causative agent of epizootic shell dis-ease as it can readily be recovered from lesions(Chistoserdov et al 2005 2012 Quinn et al 2012aWhitten et al 2014) The abundance of OTU 18 onthe affected shell in this study (12 Fig 5B) con-firms the potential importance of Aquimarina lsquohomariarsquoto LMSD though its lack of dominance on the transi-tionary shell (2 Fig 5B) suggests that it is not assuccessful outside of melanized areas of the shell (seeTable S1)

In the transitionary core microbiome (Fig 5C)OTUs 44 and 160 appeared to be particularly impor-tant OTU 44 was one of the most abundant OTUs inall shell conditions however it was only consideredcore to the transitionary shell (Table 1 see Table S1)This OTU was most closely related to Marinobactersalsuginis (100 sequence identity S V Law et alunpubl data accession no KT921331) Other speciesfrom this genus have proteins for the deacetylation ofchitin (Song et al 2013) which suggests that OTU 44may be important to initiation of disease creatingconditions under which opportunistic colonizers asso-ciated with LMSD might thrive (Malloy 1978 Chisto -serdov et al 2005) OTU 160 was most closely relatedto the candidate species Endobugula sertula (99sequence identity McGovern amp Hellberg 2003) abryozoan symbiont that produces polyketides to pro-tect the host larvae from predation (Lindquist amp Hay1996 Lopanik et al 2004) In addition to OTUs 44and 160 OTU 78 was part of the transitionary core

microbiome though quantitatively it appeared to bemore important to the affected shell Although thisOTU was not a member of the affected shell coremicrobiome it was present in 13 of the 16 affectedshell samples and was present at relatively highabundance in 7 of those samples This OTU genusLoktanella which shared 100 sequence identitywith bacteria recovered from other diseased lobstershells (Quinn et al 2012a) may therefore be in fectingthe shell alongside early colonizers and in some in -stances growing in abundance as LMSD progresses

Of the 4 OTUs shared between the affected andtransitionary core microbiomes (Fig 5E) OTUs 47and 118 were numerically dominant in both commu-nities OTU 47 belonged to the genus Colwelliawhich Meres et al (2012) reported as one of the manygroups of species that helped to differentiate dis-eased and healthy lobster shell In the current studythis bacterium was abundant on transitionary shellan area of the shell that would have previously re -mained unsampled indicating that members of thisgenus may be more important in promoting LMSDthan previously thought if the transitionary shellindeed reflects an initial invasion OTU 118 was mostclosely aligned with Arenicella chitinivorans (99sequence identity) a bacterium that differs from oth-ers in the same genus by its ability to degrade chitin(Nedashkovskaya et al 2013) Given that previouswork by Whitten et al (2014) also recovered mem-bers of this genus from lesioned areas of the lobstershell it appears that the ability to degrade chitin maybe an important characteristic allowing bacteria tothrive on both the affected and transitionary shell

Many studies have reported the incidence of ubiq-uitous bacteria on the lobster shell both on diseasedand healthy shell (Chistoserdov et al 2005 Meres etal 2012 Whitten et al 2014) In this study there were10 OTUs present in the core microbiome of all 3 shellconditions (Fig 5F) and OTUs 1 and 19 were propor-tionally dominant on the transitionary and affectedshell respectively though they were highly abun-dant in all 3 shell conditions (see Table S1) OTU 1 ismost likely Perspicuibacter marinus (99 sequenceidentity) an aerobic marine bacterium recently iso-lated from surface seawater in Japan (Teramoto et al2015) At this time not much is known about thisorganism but it has been cultured and thereforecould be used in future studies exploring the dynam-ics of the transitionary lobster shell community OTU19 an important shell bacterium found on affectedshell was most similar to other bacteria found inassociation with sponges (99 sequence identitySipkema et al 2011) It is likely that this bacterium is

52

Feinman et al Bacterial communities in lobster shell disease

from the genus Lutimonas (96 sequence identityBuerger et al 2012) which is part of the same familyas Aquimarina and may represent another oppor-tunistic colonizer of diseased lobster shell

In addition to these core microbiome memberswhich appear to be important due to their relativeabundance on either the transitionary or affectedshell there were a number of other core microbiomeOTUs that had identities similar to taxa identified inprevious studies of epizootic shell disease Severalstudies have reported Aquimarina lsquohomariarsquo as apotential causative agent in shell disease including abacterial challenge study that was able to recoverthis bacterium from lobsters with lesions (Quinn et al2012a) and a next-generation sequencing study thatexamined healthy and diseased shell (Meres et al2012) Although we only recovered a single OTU thatwas associated with the genus Aquimarina (OTU 18)we recovered several other OTUs including Lutimo -nas that were from the Flavobacteria family (Table 1)Several studies have reported that members of thisfamily aside from Aquimarina are important to epi-zootic shell disease (Chistoserdov et al 2012 Mereset al 2012 Quinn et al 2012b) In addition to Aqui -marina the lsquoThalassobiusrsquo genus has been associatedwith epizootic shell disease (Meres et al 2012 Quinnet al 2012a) Although we did not recover any bacte-ria from this genus we recovered several bacteriafrom the Rhodobacteraceae family including previ-ously observed bacteria from the Loktanella and Jan-naschia genera (Chistoserdov et al 2012 Meres et al2012) In addition to Aquimarina Loktanella andJannaschia Meres et al (2012) also reported bacteriafrom the genus Cardiobacterium and suborder Micro -coccineae as some of the most abundant bacteria intheir study We did not recover any bacteria fromeither of these groups nor their respective ordersThese differences could be due to the taxonomicidentification tools used in different studies but mayalso reflect differences in wild versus captive-raisedlobsters

In conclusion using a fine spatial scale we charac-terized changes in bacterial community compositionthat occur during the progression of LMSD Wedemonstrated that affected areas of lobster shell havealtered bacterial community composition and reducedbacterial diversity compared to unaffected shell Wealso demonstrated that areas close to the site of dis-ease but not yet melanized have altered bacterialcommunities and reduced bacterial diversity in com-parison to unaffected shell and represent a transi-tionary community that may be important in the initi-ation of LMSD We identified key bacteria associated

with each stage of the disease and proposed 2 differ-ent conceptual models one that suggests that thetransitionary bacterial community arises due to foul-ing and another that suggests the transitionary shellharbors bacteria important to the initiation of LMSDUnderstanding the full complement of bacteria asso-ciated with affected transitionary and asymptomaticlobster shells is the first step toward understandingthe role of the transitionary bacterial community andhelping us develop mechanisms to diminish thespread and severity of this disease

Acknowledgements We thank Anita Kim from the NewEngland Aquarium Lobster Research and Rearing Facilityfor caring for the animals used in this study and Jeff Dusen-berry for the use of the supercomputing facilities managedby the Research Computing Department at the University ofMassachusetts Boston We also thank Rachel Vincent anundergraduate assistant on this research as well as PatrickKearns John Angell and all the members of the BowenLaboratory Funding for this project was provided by theSanofi Genzyme Doctoral Research Fellowship (SGF) theUniversity of Massachusetts Boston Doctoral DissertationResearch Grant (SGF) and the National Science Founda-tion Research Experience for Undergraduates Award DBI-1359241 to Dr Rachel Skvirsky

LITERATURE CITED

Anderson MJ (2001) A new method for non-parametric mul-tivariate analysis of variance Austral Ecol 26 32minus46

Buerger S Spoering A Gavrish E Leslin C Ling L EpsteinSS (2012) Microbial scout hypothesis and microbial dis-covery Appl Environ Microbiol 78 3229minus3233

Caporaso JG Kuczynski J Stombaugh J Bittinger K andothers (2010) QIIME allows analysis of high-throughputcommunity sequencing data Nat Methods 7 335minus336

Caporaso JG Lauber CL Walters WA Berg-Lyons D andothers (2012) Ultra-high-throughput microbial commu-nity analysis on the Illumina HiSeq and MiSeq platformsISME J 6 1621minus1624

Castro KM Somers BA (2012) Observations of epizooticshell disease in American lobsters Homarus americanusin southern New England J Shellfish Res 31 423minus430

Castro KM Cobb JS Gomez-Chiarri M Tlusty M (2012)Epizootic shell disease in American lobsters Homarusamericanus in southern New England past present andfuture Dis Aquat Org 100 149minus158

Chistoserdov AY Smolowitz R Mirasol F Hsu A (2005) Cul-ture-dependent characterization of the microbial com-munity associated with epizootic shell disease lesions inAmerican lobster Homarus americanus J Shellfish Res24 741minus747

Chistoserdov AY Quinn RA Gubbala SL Smolowitz R(2012) Bacterial communities associated with lesions ofshell disease in the American lobster Homarus ameri-canus Milne-Edwards J Shellfish Res 31 449minus462

Davies CE Whitten M Kim A Wootton EC and others (2014)A comparison of the structure of American (Homarusamericanus) and European (Homarus gammarus) lobster

53

Feinman et al Bacterial communities in lobster shell disease

ation between lesion and spot communities in ourdata suggests that this may not be the case The tran-sitionary shell bacterial community that we identi-fied however may be important for initiation ofLMSD while the affected shell bacterial communitycould represent secondary colonizers

In addition to identifying 3 distinct bacterial com-munities we observed a positive association be tweenbacterial diversity and distance from the site of LMSD(Figs 4 amp 6A) This is contrary to work done by Whit-ten et al (2014) which showed greater bacterialdiversity in diseased shell compared to healthy shellThese previous results however were based onmethods that recovered an average of 8 to 11 taxa persample The sequencing methods we used allowedfor a much greater depth which revealed a greaterproportion of low abundance taxa thereby increas-ing apparent diversity The reduction in bacteria atthe site of LMSD indicates that the diseased shellenvironment may be highly selective for specific bac-terial taxa perhaps due to the host response and theantibacterial effects of melanin Alternatively op -portunistic bacteria associated with the newly ex -posed chitin matrix may outcompete other bacteriathereby excluding them Since bacterial diversitywas also reduced on the transitionary shell comparedto the unaffected shell we suggest that competitionrather than environmental factors lowered bacterialdiversity at the site of LMSD It is also possible thatalong with changes in the bacterial communitychanges occurred in the physical environment of thetransitionary shell prior to the onset of disease thathave yet to be identified

The presence of a transitionary community sug-gests that it is not just the bacteria at the site of dis-ease that are altered during LMSD but that there isalso a change in the bacterial community composi-tion at distances up to 05 cm away from the site ofmelanization The transitionary community couldrepresent several possibilities the 2 most parsimo-nious being (1) an outer edge of the disease zone thatis fouled by disease causing bacteria or (2) a novelcommunity that is a precursor for pathogenic bacteriathat later infect the lobster shell These 2 alternatemodels for disease transmission depicted in Fig 6BCprovide possible explanations for the changes in bothbacterial diversity and community structure we ob -served over space and time In Alternate Model 1(Fig 6B) the transitionary and affected bacterialcommunities both appear when spots are first ob -servable on the lobster shell Here the transitionarycommunity is not involved as a causative agent ofLMSD but instead is present due to fouling from the

disease-causing community and changes in the shellenvironment that occur due to the presence of dis-ease-causing bacteria As LMSD progresses theaffected area of the shell increases (transition fromspot to lesion) until the lobster eventually molts Thismodel is in agreement with previous reports regard-ing the progression of shell disease (Chistoserdov etal 2012 Davies et al 2014 Whitten et al 2014)though the sampling schemes in these previous stud-ies did not allow for the detection of a transitionarycommunity Alternate Model 2 (Fig 6C) interpretsthe transitionary bacterial community as a precursorcommunity arriving prior to the onset of visual signsof LMSD In this model the transitionary communityrepresents initial colonizers that may cause a shift inthe lobster shell bacterial community and lower bac-terial diversity prior to visual signs of disease Thisinitial invasion creates an environment that allowsfor secondary colonization that advances the diseasestate at which time the melanization associated withshell disease becomes observable The fact that thebacterial community at 05 cm is distinct from thecommunity at 0 cm (Figs 2 amp 3) and that there arebacteria that are unique to the core microbiome at05 cm (Fig 5) offers support for this second modelThis model hinges on the existence of a currentlyunobserved state where the transitionary and un -affected communities are present prior to visual signsof disease This state will be difficult to observe infuture studies as discovery would only come fromresearchers predicting the exact location where shelldisease might occur and sampling that predictedlocation prior to visual signs of disease While both ofthese alternate models differ from the currently pro-posed model for the progression of epizootic shelldisease both are in agreement with the previouslyproposed polymicrobial nature of shell disease thatis either caused by or amplified by a dysbiosis in bacterial communities on unaffected shell (Meres etal 2012) Culture studies examining growth dynamicsof potential pathogens in the presence and absenceof members of the transitionary community couldoffer additional insights into the location and initia-tion of LMSD and could refine the models proposedhere

Given the potential importance of the transitionaryand affected shell communities to the initiation andprogression of LMSD we identified key bacterialmembers of these shell conditions from the coremicrobiome (Figs 5 amp 6A) and evaluated them interms of the 2 hypothesized models The affectedshell was dominated by just 2 OTUs 0 and 18 whichaccounted for nearly 27 of the sequences from this

51

Dis Aquat Org 124 41ndash54 2017

shell condition (Fig 5B) Given their proportionaldominance these taxa appeared to be very importantto the affected shell community The closest culturedrepresentative to OTU 0 was from Candidatus Thio -globus singularis a mixotrophic bacterium belong-ing to the SUP05 clade known for its potential to oxi-dize sulfur and fix carbon in the dark ocean and inlow oxygen zones (Marshall amp Morris 2015) This cul-tured representative however only shared 91sequence identity with OTU 0 Therefore this OTUlikely represents a new uncultured species As thiswas the most abundant OTU from both the affectedand transitionary shell (see Table S1 in the Supple-ment) better identification and classification of thisbacterium is important for understanding its role inthe progression of LMSD OTU 18 the second mostabundant OTU on the affected shell (see Table S1)shared 100 sequence identity with a representa-tive from the genus Aquimarina (Esteves et al 2013)and 99 sequence identity with bacteria re coveredfrom lobster shells after having been exposed toAquimarina lsquohomariarsquo (Quinn et al 2012a) Aquima-rina lsquohomariarsquo has been identified by several studiesas a potential causative agent of epizootic shell dis-ease as it can readily be recovered from lesions(Chistoserdov et al 2005 2012 Quinn et al 2012aWhitten et al 2014) The abundance of OTU 18 onthe affected shell in this study (12 Fig 5B) con-firms the potential importance of Aquimarina lsquohomariarsquoto LMSD though its lack of dominance on the transi-tionary shell (2 Fig 5B) suggests that it is not assuccessful outside of melanized areas of the shell (seeTable S1)

In the transitionary core microbiome (Fig 5C)OTUs 44 and 160 appeared to be particularly impor-tant OTU 44 was one of the most abundant OTUs inall shell conditions however it was only consideredcore to the transitionary shell (Table 1 see Table S1)This OTU was most closely related to Marinobactersalsuginis (100 sequence identity S V Law et alunpubl data accession no KT921331) Other speciesfrom this genus have proteins for the deacetylation ofchitin (Song et al 2013) which suggests that OTU 44may be important to initiation of disease creatingconditions under which opportunistic colonizers asso-ciated with LMSD might thrive (Malloy 1978 Chisto -serdov et al 2005) OTU 160 was most closely relatedto the candidate species Endobugula sertula (99sequence identity McGovern amp Hellberg 2003) abryozoan symbiont that produces polyketides to pro-tect the host larvae from predation (Lindquist amp Hay1996 Lopanik et al 2004) In addition to OTUs 44and 160 OTU 78 was part of the transitionary core

microbiome though quantitatively it appeared to bemore important to the affected shell Although thisOTU was not a member of the affected shell coremicrobiome it was present in 13 of the 16 affectedshell samples and was present at relatively highabundance in 7 of those samples This OTU genusLoktanella which shared 100 sequence identitywith bacteria recovered from other diseased lobstershells (Quinn et al 2012a) may therefore be in fectingthe shell alongside early colonizers and in some in -stances growing in abundance as LMSD progresses

Of the 4 OTUs shared between the affected andtransitionary core microbiomes (Fig 5E) OTUs 47and 118 were numerically dominant in both commu-nities OTU 47 belonged to the genus Colwelliawhich Meres et al (2012) reported as one of the manygroups of species that helped to differentiate dis-eased and healthy lobster shell In the current studythis bacterium was abundant on transitionary shellan area of the shell that would have previously re -mained unsampled indicating that members of thisgenus may be more important in promoting LMSDthan previously thought if the transitionary shellindeed reflects an initial invasion OTU 118 was mostclosely aligned with Arenicella chitinivorans (99sequence identity) a bacterium that differs from oth-ers in the same genus by its ability to degrade chitin(Nedashkovskaya et al 2013) Given that previouswork by Whitten et al (2014) also recovered mem-bers of this genus from lesioned areas of the lobstershell it appears that the ability to degrade chitin maybe an important characteristic allowing bacteria tothrive on both the affected and transitionary shell

Many studies have reported the incidence of ubiq-uitous bacteria on the lobster shell both on diseasedand healthy shell (Chistoserdov et al 2005 Meres etal 2012 Whitten et al 2014) In this study there were10 OTUs present in the core microbiome of all 3 shellconditions (Fig 5F) and OTUs 1 and 19 were propor-tionally dominant on the transitionary and affectedshell respectively though they were highly abun-dant in all 3 shell conditions (see Table S1) OTU 1 ismost likely Perspicuibacter marinus (99 sequenceidentity) an aerobic marine bacterium recently iso-lated from surface seawater in Japan (Teramoto et al2015) At this time not much is known about thisorganism but it has been cultured and thereforecould be used in future studies exploring the dynam-ics of the transitionary lobster shell community OTU19 an important shell bacterium found on affectedshell was most similar to other bacteria found inassociation with sponges (99 sequence identitySipkema et al 2011) It is likely that this bacterium is

52

Feinman et al Bacterial communities in lobster shell disease

from the genus Lutimonas (96 sequence identityBuerger et al 2012) which is part of the same familyas Aquimarina and may represent another oppor-tunistic colonizer of diseased lobster shell

In addition to these core microbiome memberswhich appear to be important due to their relativeabundance on either the transitionary or affectedshell there were a number of other core microbiomeOTUs that had identities similar to taxa identified inprevious studies of epizootic shell disease Severalstudies have reported Aquimarina lsquohomariarsquo as apotential causative agent in shell disease including abacterial challenge study that was able to recoverthis bacterium from lobsters with lesions (Quinn et al2012a) and a next-generation sequencing study thatexamined healthy and diseased shell (Meres et al2012) Although we only recovered a single OTU thatwas associated with the genus Aquimarina (OTU 18)we recovered several other OTUs including Lutimo -nas that were from the Flavobacteria family (Table 1)Several studies have reported that members of thisfamily aside from Aquimarina are important to epi-zootic shell disease (Chistoserdov et al 2012 Mereset al 2012 Quinn et al 2012b) In addition to Aqui -marina the lsquoThalassobiusrsquo genus has been associatedwith epizootic shell disease (Meres et al 2012 Quinnet al 2012a) Although we did not recover any bacte-ria from this genus we recovered several bacteriafrom the Rhodobacteraceae family including previ-ously observed bacteria from the Loktanella and Jan-naschia genera (Chistoserdov et al 2012 Meres et al2012) In addition to Aquimarina Loktanella andJannaschia Meres et al (2012) also reported bacteriafrom the genus Cardiobacterium and suborder Micro -coccineae as some of the most abundant bacteria intheir study We did not recover any bacteria fromeither of these groups nor their respective ordersThese differences could be due to the taxonomicidentification tools used in different studies but mayalso reflect differences in wild versus captive-raisedlobsters

In conclusion using a fine spatial scale we charac-terized changes in bacterial community compositionthat occur during the progression of LMSD Wedemonstrated that affected areas of lobster shell havealtered bacterial community composition and reducedbacterial diversity compared to unaffected shell Wealso demonstrated that areas close to the site of dis-ease but not yet melanized have altered bacterialcommunities and reduced bacterial diversity in com-parison to unaffected shell and represent a transi-tionary community that may be important in the initi-ation of LMSD We identified key bacteria associated

with each stage of the disease and proposed 2 differ-ent conceptual models one that suggests that thetransitionary bacterial community arises due to foul-ing and another that suggests the transitionary shellharbors bacteria important to the initiation of LMSDUnderstanding the full complement of bacteria asso-ciated with affected transitionary and asymptomaticlobster shells is the first step toward understandingthe role of the transitionary bacterial community andhelping us develop mechanisms to diminish thespread and severity of this disease

Acknowledgements We thank Anita Kim from the NewEngland Aquarium Lobster Research and Rearing Facilityfor caring for the animals used in this study and Jeff Dusen-berry for the use of the supercomputing facilities managedby the Research Computing Department at the University ofMassachusetts Boston We also thank Rachel Vincent anundergraduate assistant on this research as well as PatrickKearns John Angell and all the members of the BowenLaboratory Funding for this project was provided by theSanofi Genzyme Doctoral Research Fellowship (SGF) theUniversity of Massachusetts Boston Doctoral DissertationResearch Grant (SGF) and the National Science Founda-tion Research Experience for Undergraduates Award DBI-1359241 to Dr Rachel Skvirsky

LITERATURE CITED

Anderson MJ (2001) A new method for non-parametric mul-tivariate analysis of variance Austral Ecol 26 32minus46

Buerger S Spoering A Gavrish E Leslin C Ling L EpsteinSS (2012) Microbial scout hypothesis and microbial dis-covery Appl Environ Microbiol 78 3229minus3233

Caporaso JG Kuczynski J Stombaugh J Bittinger K andothers (2010) QIIME allows analysis of high-throughputcommunity sequencing data Nat Methods 7 335minus336

Caporaso JG Lauber CL Walters WA Berg-Lyons D andothers (2012) Ultra-high-throughput microbial commu-nity analysis on the Illumina HiSeq and MiSeq platformsISME J 6 1621minus1624

Castro KM Somers BA (2012) Observations of epizooticshell disease in American lobsters Homarus americanusin southern New England J Shellfish Res 31 423minus430

Castro KM Cobb JS Gomez-Chiarri M Tlusty M (2012)Epizootic shell disease in American lobsters Homarusamericanus in southern New England past present andfuture Dis Aquat Org 100 149minus158

Chistoserdov AY Smolowitz R Mirasol F Hsu A (2005) Cul-ture-dependent characterization of the microbial com-munity associated with epizootic shell disease lesions inAmerican lobster Homarus americanus J Shellfish Res24 741minus747

Chistoserdov AY Quinn RA Gubbala SL Smolowitz R(2012) Bacterial communities associated with lesions ofshell disease in the American lobster Homarus ameri-canus Milne-Edwards J Shellfish Res 31 449minus462

Davies CE Whitten M Kim A Wootton EC and others (2014)A comparison of the structure of American (Homarusamericanus) and European (Homarus gammarus) lobster

53

Dis Aquat Org 124 41ndash54 2017

shell condition (Fig 5B) Given their proportionaldominance these taxa appeared to be very importantto the affected shell community The closest culturedrepresentative to OTU 0 was from Candidatus Thio -globus singularis a mixotrophic bacterium belong-ing to the SUP05 clade known for its potential to oxi-dize sulfur and fix carbon in the dark ocean and inlow oxygen zones (Marshall amp Morris 2015) This cul-tured representative however only shared 91sequence identity with OTU 0 Therefore this OTUlikely represents a new uncultured species As thiswas the most abundant OTU from both the affectedand transitionary shell (see Table S1 in the Supple-ment) better identification and classification of thisbacterium is important for understanding its role inthe progression of LMSD OTU 18 the second mostabundant OTU on the affected shell (see Table S1)shared 100 sequence identity with a representa-tive from the genus Aquimarina (Esteves et al 2013)and 99 sequence identity with bacteria re coveredfrom lobster shells after having been exposed toAquimarina lsquohomariarsquo (Quinn et al 2012a) Aquima-rina lsquohomariarsquo has been identified by several studiesas a potential causative agent of epizootic shell dis-ease as it can readily be recovered from lesions(Chistoserdov et al 2005 2012 Quinn et al 2012aWhitten et al 2014) The abundance of OTU 18 onthe affected shell in this study (12 Fig 5B) con-firms the potential importance of Aquimarina lsquohomariarsquoto LMSD though its lack of dominance on the transi-tionary shell (2 Fig 5B) suggests that it is not assuccessful outside of melanized areas of the shell (seeTable S1)

In the transitionary core microbiome (Fig 5C)OTUs 44 and 160 appeared to be particularly impor-tant OTU 44 was one of the most abundant OTUs inall shell conditions however it was only consideredcore to the transitionary shell (Table 1 see Table S1)This OTU was most closely related to Marinobactersalsuginis (100 sequence identity S V Law et alunpubl data accession no KT921331) Other speciesfrom this genus have proteins for the deacetylation ofchitin (Song et al 2013) which suggests that OTU 44may be important to initiation of disease creatingconditions under which opportunistic colonizers asso-ciated with LMSD might thrive (Malloy 1978 Chisto -serdov et al 2005) OTU 160 was most closely relatedto the candidate species Endobugula sertula (99sequence identity McGovern amp Hellberg 2003) abryozoan symbiont that produces polyketides to pro-tect the host larvae from predation (Lindquist amp Hay1996 Lopanik et al 2004) In addition to OTUs 44and 160 OTU 78 was part of the transitionary core

microbiome though quantitatively it appeared to bemore important to the affected shell Although thisOTU was not a member of the affected shell coremicrobiome it was present in 13 of the 16 affectedshell samples and was present at relatively highabundance in 7 of those samples This OTU genusLoktanella which shared 100 sequence identitywith bacteria recovered from other diseased lobstershells (Quinn et al 2012a) may therefore be in fectingthe shell alongside early colonizers and in some in -stances growing in abundance as LMSD progresses

Of the 4 OTUs shared between the affected andtransitionary core microbiomes (Fig 5E) OTUs 47and 118 were numerically dominant in both commu-nities OTU 47 belonged to the genus Colwelliawhich Meres et al (2012) reported as one of the manygroups of species that helped to differentiate dis-eased and healthy lobster shell In the current studythis bacterium was abundant on transitionary shellan area of the shell that would have previously re -mained unsampled indicating that members of thisgenus may be more important in promoting LMSDthan previously thought if the transitionary shellindeed reflects an initial invasion OTU 118 was mostclosely aligned with Arenicella chitinivorans (99sequence identity) a bacterium that differs from oth-ers in the same genus by its ability to degrade chitin(Nedashkovskaya et al 2013) Given that previouswork by Whitten et al (2014) also recovered mem-bers of this genus from lesioned areas of the lobstershell it appears that the ability to degrade chitin maybe an important characteristic allowing bacteria tothrive on both the affected and transitionary shell

Many studies have reported the incidence of ubiq-uitous bacteria on the lobster shell both on diseasedand healthy shell (Chistoserdov et al 2005 Meres etal 2012 Whitten et al 2014) In this study there were10 OTUs present in the core microbiome of all 3 shellconditions (Fig 5F) and OTUs 1 and 19 were propor-tionally dominant on the transitionary and affectedshell respectively though they were highly abun-dant in all 3 shell conditions (see Table S1) OTU 1 ismost likely Perspicuibacter marinus (99 sequenceidentity) an aerobic marine bacterium recently iso-lated from surface seawater in Japan (Teramoto et al2015) At this time not much is known about thisorganism but it has been cultured and thereforecould be used in future studies exploring the dynam-ics of the transitionary lobster shell community OTU19 an important shell bacterium found on affectedshell was most similar to other bacteria found inassociation with sponges (99 sequence identitySipkema et al 2011) It is likely that this bacterium is

52

Feinman et al Bacterial communities in lobster shell disease

from the genus Lutimonas (96 sequence identityBuerger et al 2012) which is part of the same familyas Aquimarina and may represent another oppor-tunistic colonizer of diseased lobster shell

In addition to these core microbiome memberswhich appear to be important due to their relativeabundance on either the transitionary or affectedshell there were a number of other core microbiomeOTUs that had identities similar to taxa identified inprevious studies of epizootic shell disease Severalstudies have reported Aquimarina lsquohomariarsquo as apotential causative agent in shell disease including abacterial challenge study that was able to recoverthis bacterium from lobsters with lesions (Quinn et al2012a) and a next-generation sequencing study thatexamined healthy and diseased shell (Meres et al2012) Although we only recovered a single OTU thatwas associated with the genus Aquimarina (OTU 18)we recovered several other OTUs including Lutimo -nas that were from the Flavobacteria family (Table 1)Several studies have reported that members of thisfamily aside from Aquimarina are important to epi-zootic shell disease (Chistoserdov et al 2012 Mereset al 2012 Quinn et al 2012b) In addition to Aqui -marina the lsquoThalassobiusrsquo genus has been associatedwith epizootic shell disease (Meres et al 2012 Quinnet al 2012a) Although we did not recover any bacte-ria from this genus we recovered several bacteriafrom the Rhodobacteraceae family including previ-ously observed bacteria from the Loktanella and Jan-naschia genera (Chistoserdov et al 2012 Meres et al2012) In addition to Aquimarina Loktanella andJannaschia Meres et al (2012) also reported bacteriafrom the genus Cardiobacterium and suborder Micro -coccineae as some of the most abundant bacteria intheir study We did not recover any bacteria fromeither of these groups nor their respective ordersThese differences could be due to the taxonomicidentification tools used in different studies but mayalso reflect differences in wild versus captive-raisedlobsters

In conclusion using a fine spatial scale we charac-terized changes in bacterial community compositionthat occur during the progression of LMSD Wedemonstrated that affected areas of lobster shell havealtered bacterial community composition and reducedbacterial diversity compared to unaffected shell Wealso demonstrated that areas close to the site of dis-ease but not yet melanized have altered bacterialcommunities and reduced bacterial diversity in com-parison to unaffected shell and represent a transi-tionary community that may be important in the initi-ation of LMSD We identified key bacteria associated

with each stage of the disease and proposed 2 differ-ent conceptual models one that suggests that thetransitionary bacterial community arises due to foul-ing and another that suggests the transitionary shellharbors bacteria important to the initiation of LMSDUnderstanding the full complement of bacteria asso-ciated with affected transitionary and asymptomaticlobster shells is the first step toward understandingthe role of the transitionary bacterial community andhelping us develop mechanisms to diminish thespread and severity of this disease

Acknowledgements We thank Anita Kim from the NewEngland Aquarium Lobster Research and Rearing Facilityfor caring for the animals used in this study and Jeff Dusen-berry for the use of the supercomputing facilities managedby the Research Computing Department at the University ofMassachusetts Boston We also thank Rachel Vincent anundergraduate assistant on this research as well as PatrickKearns John Angell and all the members of the BowenLaboratory Funding for this project was provided by theSanofi Genzyme Doctoral Research Fellowship (SGF) theUniversity of Massachusetts Boston Doctoral DissertationResearch Grant (SGF) and the National Science Founda-tion Research Experience for Undergraduates Award DBI-1359241 to Dr Rachel Skvirsky

LITERATURE CITED

Anderson MJ (2001) A new method for non-parametric mul-tivariate analysis of variance Austral Ecol 26 32minus46

Buerger S Spoering A Gavrish E Leslin C Ling L EpsteinSS (2012) Microbial scout hypothesis and microbial dis-covery Appl Environ Microbiol 78 3229minus3233

Caporaso JG Kuczynski J Stombaugh J Bittinger K andothers (2010) QIIME allows analysis of high-throughputcommunity sequencing data Nat Methods 7 335minus336

Caporaso JG Lauber CL Walters WA Berg-Lyons D andothers (2012) Ultra-high-throughput microbial commu-nity analysis on the Illumina HiSeq and MiSeq platformsISME J 6 1621minus1624

Castro KM Somers BA (2012) Observations of epizooticshell disease in American lobsters Homarus americanusin southern New England J Shellfish Res 31 423minus430

Castro KM Cobb JS Gomez-Chiarri M Tlusty M (2012)Epizootic shell disease in American lobsters Homarusamericanus in southern New England past present andfuture Dis Aquat Org 100 149minus158

Chistoserdov AY Smolowitz R Mirasol F Hsu A (2005) Cul-ture-dependent characterization of the microbial com-munity associated with epizootic shell disease lesions inAmerican lobster Homarus americanus J Shellfish Res24 741minus747

Chistoserdov AY Quinn RA Gubbala SL Smolowitz R(2012) Bacterial communities associated with lesions ofshell disease in the American lobster Homarus ameri-canus Milne-Edwards J Shellfish Res 31 449minus462

Davies CE Whitten M Kim A Wootton EC and others (2014)A comparison of the structure of American (Homarusamericanus) and European (Homarus gammarus) lobster

53

Feinman et al Bacterial communities in lobster shell disease

from the genus Lutimonas (96 sequence identityBuerger et al 2012) which is part of the same familyas Aquimarina and may represent another oppor-tunistic colonizer of diseased lobster shell

In addition to these core microbiome memberswhich appear to be important due to their relativeabundance on either the transitionary or affectedshell there were a number of other core microbiomeOTUs that had identities similar to taxa identified inprevious studies of epizootic shell disease Severalstudies have reported Aquimarina lsquohomariarsquo as apotential causative agent in shell disease including abacterial challenge study that was able to recoverthis bacterium from lobsters with lesions (Quinn et al2012a) and a next-generation sequencing study thatexamined healthy and diseased shell (Meres et al2012) Although we only recovered a single OTU thatwas associated with the genus Aquimarina (OTU 18)we recovered several other OTUs including Lutimo -nas that were from the Flavobacteria family (Table 1)Several studies have reported that members of thisfamily aside from Aquimarina are important to epi-zootic shell disease (Chistoserdov et al 2012 Mereset al 2012 Quinn et al 2012b) In addition to Aqui -marina the lsquoThalassobiusrsquo genus has been associatedwith epizootic shell disease (Meres et al 2012 Quinnet al 2012a) Although we did not recover any bacte-ria from this genus we recovered several bacteriafrom the Rhodobacteraceae family including previ-ously observed bacteria from the Loktanella and Jan-naschia genera (Chistoserdov et al 2012 Meres et al2012) In addition to Aquimarina Loktanella andJannaschia Meres et al (2012) also reported bacteriafrom the genus Cardiobacterium and suborder Micro -coccineae as some of the most abundant bacteria intheir study We did not recover any bacteria fromeither of these groups nor their respective ordersThese differences could be due to the taxonomicidentification tools used in different studies but mayalso reflect differences in wild versus captive-raisedlobsters

In conclusion using a fine spatial scale we charac-terized changes in bacterial community compositionthat occur during the progression of LMSD Wedemonstrated that affected areas of lobster shell havealtered bacterial community composition and reducedbacterial diversity compared to unaffected shell Wealso demonstrated that areas close to the site of dis-ease but not yet melanized have altered bacterialcommunities and reduced bacterial diversity in com-parison to unaffected shell and represent a transi-tionary community that may be important in the initi-ation of LMSD We identified key bacteria associated

with each stage of the disease and proposed 2 differ-ent conceptual models one that suggests that thetransitionary bacterial community arises due to foul-ing and another that suggests the transitionary shellharbors bacteria important to the initiation of LMSDUnderstanding the full complement of bacteria asso-ciated with affected transitionary and asymptomaticlobster shells is the first step toward understandingthe role of the transitionary bacterial community andhelping us develop mechanisms to diminish thespread and severity of this disease

Acknowledgements We thank Anita Kim from the NewEngland Aquarium Lobster Research and Rearing Facilityfor caring for the animals used in this study and Jeff Dusen-berry for the use of the supercomputing facilities managedby the Research Computing Department at the University ofMassachusetts Boston We also thank Rachel Vincent anundergraduate assistant on this research as well as PatrickKearns John Angell and all the members of the BowenLaboratory Funding for this project was provided by theSanofi Genzyme Doctoral Research Fellowship (SGF) theUniversity of Massachusetts Boston Doctoral DissertationResearch Grant (SGF) and the National Science Founda-tion Research Experience for Undergraduates Award DBI-1359241 to Dr Rachel Skvirsky

LITERATURE CITED

Anderson MJ (2001) A new method for non-parametric mul-tivariate analysis of variance Austral Ecol 26 32minus46

Buerger S Spoering A Gavrish E Leslin C Ling L EpsteinSS (2012) Microbial scout hypothesis and microbial dis-covery Appl Environ Microbiol 78 3229minus3233

Caporaso JG Kuczynski J Stombaugh J Bittinger K andothers (2010) QIIME allows analysis of high-throughputcommunity sequencing data Nat Methods 7 335minus336

Caporaso JG Lauber CL Walters WA Berg-Lyons D andothers (2012) Ultra-high-throughput microbial commu-nity analysis on the Illumina HiSeq and MiSeq platformsISME J 6 1621minus1624

Castro KM Somers BA (2012) Observations of epizooticshell disease in American lobsters Homarus americanusin southern New England J Shellfish Res 31 423minus430

Castro KM Cobb JS Gomez-Chiarri M Tlusty M (2012)Epizootic shell disease in American lobsters Homarusamericanus in southern New England past present andfuture Dis Aquat Org 100 149minus158

Chistoserdov AY Smolowitz R Mirasol F Hsu A (2005) Cul-ture-dependent characterization of the microbial com-munity associated with epizootic shell disease lesions inAmerican lobster Homarus americanus J Shellfish Res24 741minus747

Chistoserdov AY Quinn RA Gubbala SL Smolowitz R(2012) Bacterial communities associated with lesions ofshell disease in the American lobster Homarus ameri-canus Milne-Edwards J Shellfish Res 31 449minus462

Davies CE Whitten M Kim A Wootton EC and others (2014)A comparison of the structure of American (Homarusamericanus) and European (Homarus gammarus) lobster

53

Dis Aquat Org 124 41ndash54 2017

cuticle with particular reference to shell disease J Inver-tebr Pathol 117 33minus41

Edgar RC Haas BJ Clemente JC Quince C Knight R (2011)UCHIME improves sensitivity and speed of chimeradetection Bioinformatics 27 2194minus2200

Esteves AIS Hardoim CCP Xavier JR Gonccedilalves JMSCosta R (2013) Molecular richness and biotechnologicalpotential of bacteria cultured from Irciniidae sponges inthe north-east Atlantic FEMS Microbiol Ecol 85 519minus536

Faith DP Baker AM (2006) Phylogenetic diversity (PD) andbiodiversity conservation some bioinformatics chal-lenges Evol Bioinform Online 2 121minus128

Fey SB Siepielski AM Nussleacute S Cervantes-Yoshida K andothers (2015) Recent shifts in the occurrence cause andmagnitude of animal mass mortality events Proc NatlAcad Sci USA 112 1083minus1088

Harvell CD Kim K Burkholder JM Colwell RR and others(1999) Emerging marine diseases climate links andanthropogenic factors Science 285 1505minus1510

Howell P (2012) The status of the southern New Englandlobster stock J Shellfish Res 31 573minus579

Laufer H Demir N Biggers WJ (2005) Response of theAmerican lobster to the stress of shell disease J ShellfishRes 24 757minus760

Lindquist N Hay ME (1996) Palatability and chemicaldefense of marine invertebrate larvae Ecol Monogr 66 431minus450

Lopanik N Lindquist N Targett N (2004) Potent cytotoxinsproduced by a microbial symbiont protect host larvaefrom predation Oecologia 139 131minus139

Lozupone C Lladser ME Knights D Stombaugh J Knight R(2011) UniFrac an effective distance metric for microbialcommunity comparison ISME J 5 169minus172

Maheacute F Rognes T Quince C de Vargas C (2014) Swarm robust and fast clustering method for amplicon-basedstudies PeerJ 2 e593

Malloy SC (1978) Bacteria induced shell disease of lobsters(Homarus americanus) J Wildl Dis 14 2minus10

Marshall KT Morris RM (2015) Genome sequence of lsquoCan-didatus Thioglobus singularisrsquo strain PS1 a mixotrophfrom the SUP05 clade of marine gammaproteobacteriaGenome Announc 3 e01155-15

McGovern TM Hellberg ME (2003) Cryptic species crypticendosymbionts and geographical variation in chemicaldefences in the bryozoan Bugula neritina Mol Ecol 12 1207minus1215

Meres NJ Ajuzie CC Sikaroodi M Vemulapalli M ShieldsJD Gillevet PM (2012) Dysbiosis in epizootic shell dis-ease of the American lobster (Homarus americanus)J Shellfish Res 31 463minus472

National Marine Fisheries Service (2014) NOAA CurrentFishery Statistics Fisheries of the United States 2014 USDepartment of Commerce

Nedashkovskaya OI Cleenwerck I Zhukova NV Kim SBdeVos P (2013) Arenicella chitinivorans sp nov a gamma -proteobacterium isolated from the sea urchin Strongylo-centrotus intermedius Int J Syst Evol Microbiol 63 4124minus4129

Quinn RA Metzler A Smolowitz RM Tlusty M Chistoser-dov AY (2012a) Exposures of Homarus americanus shellto three bacteria isolated from naturally occurring epi-

zootic shell disease lesions J Shellfish Res 31 485minus493Quinn RA Metzler A Tlusty M Smolowitz RM Leberg P

Chistoserdov AY (2012b) Lesion bacterial communitiesin American lobsters with diet-induced shell diseaseDis Aquat Org 98 221minus233

R Core Team (2014) R A language and environment for sta-tistical computing R Foundation for Statistical Comput-ing Vienna

Reddy TBK Thomas AD Stamatis D Bertsch J and others(2015) The Genomes OnLine Database (GOLD) v 5 ametadata management system based on a four level(meta)genome project classification Nucleic Acids Res43 D1099-D1106

Salter SJ Cox MJ Turek EM Calus ST and others (2014)Reagent and laboratory contamination can criticallyimpact sequence-based microbiome analyses BMC Biol12 87

Sipkema D Schippers K Maalcke WJ Yang Y Salim SBlanch HW (2011) Multiple approaches to enhance thecultivability of bacteria associated with the marinesponge Haliclona (gellius) sp Appl Environ Microbiol 77 2130minus2140

Smolowitz R Chistoserdov AY Hsu A (2005) A descriptionof the pathology of epizootic shell disease in the Ameri-can lobster Homarus americanus H Milne Edwards1837 J Shellfish Res 24 749minus756

Song L Ren L Li X Yu D Yu Y Wang X Liu G (2013) Com-plete genome sequence of Marinobacter sp BSs20148Genome Announc 1 e00236-13

Staley JT Konopka A (1985) Measurement of in situ activi-ties of nonphotosynthetic microorganisms in aquatic andterrestrial habitats Annu Rev Microbiol 39 321minus346

Tarrant AM Franks DG Verslycke T (2012) Gene expres-sion in American lobster (Homarus americanus) with epi-zootic shell disease J Shellfish Res 31 505minus513

Teramoto M Yagyu K Nishijima M (2015) Perspicuibactermarinus gen nov sp nov a semi-transparent bacteriumisolated from surface seawater and description of Areni-cellaceae fam nov and Arenicellales ord nov Int J SystEvol Microbiol 65 353minus358

Tlusty M Hyland C (2005) Astaxanthin deposition in thecuticle of juvenile American lobster (Homarus ameri-canus) implications for phenotypic and genotypic col-oration Mar Biol 147 113minus119

Tlusty MF Metzler A (2012) Relationship between tempera-ture and shell disease in laboratory populations of juve-nile American lobsters (Homarus americanus) J Shell-fish Res 31 533minus541

Tlusty MF Kim A Castro KM (2014) Modeling shell diseasein American lobster (Homarus americanus) as individ-ual-based health trajectories Can J Fish Aquat Sci 71 808minus813

Wang Q Garrity GM Tiedje JM Cole JR (2007) NaiveBayesian classifier for rapid assignment of rRNA se -quences into the new bacterial taxonomy Appl EnvironMicrobiol 73 5261minus5267

Whitten MMA Davies CE Kim A Tlusty M Wootton ECChistoserdov AY Rowley AF (2014) Cuticles of Europeanand American lobsters harbor diverse bacterial speciesand differ in disease susceptibility MicrobiologyOpen 3 395minus409

54

Editorial responsibility Jeffrey Shields Gloucester Point Virginia USA

Submitted May 17 2016 Accepted February 6 2017Proofs received from author(s) March 21 2017