Viral coinfection is shaped by host ecology and virus– virus … · 2017. 5. 4. · Viral coinfection is shaped by host ecology and virus– virus interactions across diverse microbial

Viral coinfection is shaped by host ecology and virusndash

virus interactions across diverse microbial taxa and

environmentsSamuel L Dıaz-Mu~noz

Department of Biology Center for Genomics and Systems Biology New York University 12 Waverly PlaceNew York NY 10003 USA

Corresponding author E-mail samdiazmunoznyuedu

Abstract

Infection of more than one virus in a host coinfection is common across taxa and environments Viral coinfection can ena-ble genetic exchange alter the dynamics of infections and change the course of viral evolution Yet a systematic test of thefactors explaining variation in viral coinfection across different taxa and environments awaits completion Here I employthree microbial data sets of virusndashhost interactions covering cross-infectivity culture coinfection and single-cell coinfec-tion (total 6564 microbial hosts 13103 viruses) to provide a broad comprehensive picture of the ecological and biologicalfactors shaping viral coinfection I found evidence that ecology and virusndashvirus interactions are recurrent factors shapingcoinfection patterns Host ecology was a consistent and strong predictor of coinfection across all three data sets cross-infectivity culture coinfection and single-cell coinfection Host phylogeny or taxonomy was a less consistent predictorbeing weak or absent in the cross-infectivity and single-cell coinfection models yet it was the strongest predictor in the cul-ture coinfection model Virusndashvirus interactions strongly affected coinfection In the largest test of superinfection exclusionto date prophage sequences reduced culture coinfection by other prophages with a weaker effect on extrachromosomalvirus coinfection At the single-cell level prophage sequences eliminated coinfection Virusndashvirus interactions also increasedculture coinfection with ssDNAndashdsDNA coinfectionsgt2more likely than ssDNA-only coinfections The presence of CRISPRspacers was associated with a 50 reduction in single-cell coinfection in a marine bacteria despite the absence of exactspacer matches in any active infection Collectively these results suggest the environment bacteria inhabit and the interac-tions among surrounding viruses are two factors consistently shaping viral coinfection patterns These findings highlightthe role of virusndashvirus interactions in coinfection with implications for phage therapy microbiome dynamics and viralinfection treatments

Key words bacteriophages microbial ecology superinfection immunity CRISPR-Cas microbial evolution prophages

1 Introduction

Viruses outnumber hosts by a significant margin (Bergh et al1989 Weinbauer 2004 Suttle 2007 Rohwer and Barott 2012Wigington et al 2016) In this situation infection of more thanone strain or type of virus in a host (coinfection) might beexpected to be a rather frequent occurrence potentially leadingto virusndashvirus interactions (Dıaz-Mu~noz and Koskella 2014)

Across many different viral groups virusndashvirus interactionswithin a host can alter genetic exchange (Turner et al 1999Worobey and Holmes 1999 Dang et al 2004 Cicin-Sain et al2005) modify viral evolution (Dropulic et al 1996 Turner andChao 1998 Ghedin et al 2005 Joseph et al 2009 Refardt 2011Roux et al 2015) and change the fate of the host (Vignuzzi et al2006 Abrahams et al 2009 Li et al 2010) Yet there is little

VC The Author 2017 Published by Oxford University PressThis is an Open Access article distributed under the terms of the Creative Commons Attribution License (httpcreativecommonsorglicensesby40)which permits unrestricted reuse distribution and reproduction in any medium provided the original work is properly cited

1

Virus Evolution 2017 3(1) vex011

doi 101093vevex011Research article

information regarding the ecological and biological factors thatshape coinfection and virusndashvirus interactions Given that mostlaboratory studies of viruses focus on a single virus at a time(DaPalma et al 2010) understanding the drivers of coinfectionand virusndashvirus interactions is a pressing frontier for viralecology

Recent studies of bacteriophages have highlighted just howwidespread coinfection is and have started to shed light on theecology of viral coinfection For example although studies ofphage host range have shown that some bacterial hosts can beinfected by more than one type of phage for some time (Spencer1957 Kaiser and Dworkin 1975 Moebus and Nattkemper 1981)mounting evidence indicates this is a pervasive phenomenon(Flores et al 2011 2013 Koskella and Meaden 2013) The abilityof two or more viruses to independently infect the same hostmdashcross-infectivitymdashrevealed in these studies of host range is anecessary but not sufficient criterion to determine coinfectionThus cross-infectivity or the potential for coinfection repre-sents a baseline shaping coinfection patterns To confirm coin-fection sequence-based approaches have been used to detectviral coinfection at two scalesgt1 virus infecting the same mul-ticellular host (eg a multicellular eukaryote or a colony of bac-terial cells hereafter called culture coinfection) or to thecoinfection of a single cell At the culture coinfection scale astudy mining sequence data to examine virusndashhost relation-ships uncovered widespread coinfection in publicly availablebacterial and archaeal genome sequence data (Roux et al 2015)At the single-cell scale two studies have provided for the firsttime single-cell level information on viruses associated withspecific hosts isolated from the environment in a culture-independent manner (Roux et al 2014 Labonte et al 2015)Collectively these studies suggest that there is a large potentialfor coinfection and that this potential is realized at both thehost culture and single cell scales A summary of these studiessuggests roughly half of hosts have the potential to be infected(ie are cross-infective) or are actually infected by an averageofgt2 viruses (Table 1) Thus there is extensive evidence acrossvarious methodologies taxa environments that coinfection iswidespread and virusndashvirus interactions may be a frequentoccurrence

The aforementioned studies among others establish thatviral coinfection is a frequent occurrence in bacterial and arch-aeal hosts but systematic tests of the factors explaining varia-tion in viral coinfection across different taxa and environmentsare lacking However the literature suggests four factors thatare likely to play a role host ecology host taxonomy or phylog-eny host defense mechanisms and virusndashvirus interactionsThe relevance and importance of these are likely to varyfor cross-infectivity culture coinfection and single-cellcoinfection

Cross-infectivity has been examined in studies of phagehost-range which have provided some insight into the factorsaffecting potential coinfection In a single bacterial speciesthere can be wide variation in phage host range (Holmfeldt

et al 2007) and thus cross-infectivity A larger quantitativestudy of phage-bacteria infection networks in multiple taxaalso found wide variation in cross-infectivity in narrow taxo-nomic ranges (strain or species level) with some hosts suscepti-ble to few viruses and others to many (Flores et al 2011)However at broader host taxonomic scales cross-infectivity fol-lowed a modular pattern suggesting that higher taxonomicranks could influence cross-infectivity (Flores et al 2013) Thushost taxonomy may show a scale dependent effect in shapingcoinfection patterns Flores et al (2013) also highlighted thepotential role of ecology in shaping coinfection patterns as geo-graphic separation played a role in cross-infectivity Thus bac-terial ecology and phylogeny (particularly at taxonomic ranksabove species) are the primary candidates for drivers ofcoinfection

Culture and single cell coinfection require cross-infectivitybut also require both the bacteria and infecting viruses to allowsimultaneous or sequential infection Thus in addition to bac-terial phylogeny and ecology we can expect additional factorsshaping coinfection namely bacterial defense mechanisms andvirusndashvirus interactions An extensive collection of studies pro-vides evidence that bacterial and viral mechanisms may affectcoinfection Bacteria understandably reluctant to welcomeviruses possess a collection of mechanisms of defense againstviral infection (Labrie et al 2010) including restriction enzymes(Linn and Arber 1968 Murray 2002) and CRISPR-Cas systems(Horvath and Barrangou 2010) The latter have been shown to bean adaptive immune system for bacteria protecting from futureinfection by the same phage (Barrangou et al 2007) and preserv-ing the memory of viral infections past (Held and Whitaker2009) Metagenomic studies of CRISPR in natural environmentssuggest rapid coevolution of CRISPR arrays (Tyson and Banfield2008) but little is known regarding the in situ protective effectsof CRISPR on cells which should now be possible to decipherwith single-cell genomics

Viruses also have mechanisms to mediate infection by otherviruses some of which were identified in early lab studies ofbacteriophages (Ellis and Delbruck 1939 Delbruck 1946) Anexample of a well-described phenomenon of virusndashvirus inter-actions is superinfection immunity conferred by lysogenicviruses (Bertani 1953) which can inhibit coinfection of culturesand single cells (Bertani 1954) While this mechanism has beendescribed in several species its frequency at broader taxonomicscales and its occurrence in natural settings is not well knownMost attention in virusndashvirus interactions has focused on mech-anisms limiting coinfection with the assumption that coinfec-tion invariably reduces host fitness (Berngruber et al 2010)However some patterns of nonrandom coinfection suggest ele-vated coinfection (Turner et al 1999 Dang et al 2004 Cicin-Sainet al 2005) and specific viral mechanisms that promote co-infection have been identified including enhanced expressionof the phage attachment site upon infection (Joseph et al 2009)particle aggregation (Altan-Bonnet and Chen 2015 Aguileraet al 2017) and phagendashphage communication (Erez et al 2017)

Table 1 Viral coinfection is prevalent across various methodologies taxa environments and scales of coinfection

Cross-infectivity (potential coinfection) Culture coinfection Single-cell coinfection

Number of viruses 489 (6461) 337761804 2376083Prop of bacteria wmultiple infections 0654 0538 0450Reference (Flores et al 2011) (Roux et al 2014 2015) (Roux et al 2014)

2 | Virus Evolution 2017 Vol 3 No 1

Systematic coinfection has been proposed (Roux et al 2012) toexplain findings of chimeric viruses of mixed nucleic acids inmetagenome reads (Diemer and Stedman 2012 Roux et al2013) This suspicion was confirmed in a study of marine bacte-ria that found highly nonrandom patterns of coinfectionbetween ssDNA and dsDNA viruses in a lineage of marine bacte-ria (Roux et al 2014) but the frequency of this phenomenonacross bacterial taxa remains to be uncovered Thus detailedmolecular studies of coinfection dynamics and viromesequence data are generating questions ripe for testing acrossdiverse taxa and environments

Here I employ virusndashhost interaction data sets to date to pro-vide a broad comprehensive picture of the ecological and bio-logical factors shaping coinfection The data sets are each thelargest of their kind providing an opportunity to examine viralcoinfection at multiple scales from cross-infectivity (1005hosts 499 viruses) to culture coinfection (5492 hosts 12498viruses) and single-cell coinfection (127 hosts 143 viruses) toanswer the following questions

1 How do ecology bacterial phylogenytaxonomy bacterialdefense mechanisms and virusndashvirus interactions explainvariation in estimates of viral coinfection

2 What is the relative importance of each of these factors incross-infectivity culture coinfection and single-cellcoinfection

3 Do prophage sequences limit viral coinfection4 Do ssDNA and dsDNA viruses show evidence of preferential

coinfection5 Do sequences associated with the CRISPR-Cas bacterial

defense mechanism limit coinfection of single cells

The results of this study suggest that microbial host ecologyand virusndashvirus interactions were consistently important medi-ators of the frequency and extent of coinfection Host taxonomyand CRISPR spacers also shaped culture and single-cell coinfec-tion patterns respectively Virusndashvirus interactions served tolimit and promote coinfection

2 Materials and Methods21 Data sets

I assembled data collectively representing 13103 viral infectionsin 6564 bacterial and archaeal hosts from diverse environments(Supplementary Table S1) These data are composed of threedata sets that provide an increasingly fine-grained examinationof coinfection from cross-infectivity (potential coinfection) tocoinfection at the culture (pure cultures or single colonies notnecessarily single cells) and single-cell levels The data setexamining cross-infectivity assessed infection experimentallywith laboratory cultures while other two data sets (culture andsingle-cell coinfection) used sequence data to infer infection

The first data set on cross-infectivity (potential coinfection)is composed of bacteriophage host-range infection matricesdocumenting the results of experimental attempts at lytic infec-tion in cultured phage and hosts (Flores et al 2011) It compilesresults from 38 published studies encompassing 499 phagesand 1005 bacterial hosts Additionally I entered new metadata(ecosystem and ecosystem category) to match the culture coin-fection data set (see description below) to enable comparisonsbetween these two multi-taxon data sets The host-range infec-tion data are matrices of infection success or failure via thelsquospot testrsquo briefly a drop of phage lysate is lsquospottedrsquo on a bacte-rial lawn and lysing of bacteria is noted as presence or absence

This data set represents studies with varying sample composi-tions in terms of bacteria and phage species bacterial energysources source of samples bacterial association and isolationhabitat

The second data set on culture coinfection is derived fromviral sequence information mined from published microbialgenome sequence data on National Center for BiotechnologyInformationrsquos (NCBI) databases (Roux et al 2015) Thus this sec-ond data set provided information on actual (as opposed topotential) coinfection of cultures representing 12498 viralinfections in 5492 bacterial and archaeal hosts The set includesdata on viruses that are incorporated into the host genome (pro-phages) as well as extrachromosomal viruses detected in thegenome assemblies (representing lytic chronic carrier stateand lsquoextrachromosomal prophagersquo infections) Genomes ofmicrobial strains were primarily generated from colonies orpure cultures (except for 27 hosts known to be represented bysingle cells) Thus although these data could represent coinfec-tion at the single-cell level they are more conservativelyregarded as culture coinfections In addition I imported meta-data from the US Department of Energy Joint Genome InstituteGenomes Online Database (GOLD Mukherjee et al 2017) toassess the ecological and host factors (ecosystem ecosystemcategory and energy source) that could influence culture coin-fection I curated these records and added missing metadata(nfrac14 4964 records) by consulting strain databases (eg NCBIBioSample DSMZ BEI Resources PATRIC and ATCC) and theprimary literature

The third data set included single-cell amplified genomesproviding information on coinfection and virusndashvirus interac-tions within single cells (Roux et al 2014) This genomic data setcovers viral infections in 127 single cells of SUP05 marine bacte-ria (sulfur-oxidizing Gammaproteobacteria) isolated from differ-ent depths of the oxygen minimum zone in the Saanich Inletnear Vancouver Island British Columbia (Roux et al 2014)These single-cell data represent a combined 143 viral infectionsincluding past infections (CRISPRs and prophages) and activeinfections (active at the time of isolation eg ongoing lyticinfections)

A list and description of data sources are included inSupplementary Table S1 and the raw data used in this paperare deposited in the FigShare data repository (httpdxdoiorg106084m9figshare2072929)

22 Factors explaining cross-infectivity and coinfection

To test the factors that potentially influence coinfectionmdashecology host taxonomyphylogeny host defense mechanismsand virusndashvirus interactionsmdashI conducted regression analyseson each of the three data sets representing an increasingly finescale analysis of coinfection from cross-infectivity to culturecoinfection and finally to single-cell coinfection The data setsrepresent three distinct phenomena related to coinfection andthus the variables and data in each one are necessarily differ-ent The measures of coinfection used were the following (1)the amount of phage that could infect each host (2) the numberof extrachromosomal infections and (3) the number of activeviral infections respectively for cross-infectivity culture coin-fection and single-cell coinfection data sets Virusndashvirus inter-actions were measured as the association of theseaforementioned ongoing viral infections with the presence ofother types of viruses (eg prophages) or the association ofviruses with different characteristics (eg ssDNA vs dsDNA)Ecological factors and bacterial taxonomy were tested in all

S L Dıaz-Mu~noz | 3

data sets but virusndashvirus interactions for example were notevaluated in the cross-infectivity data set because they do notapply (ie measures potential and not actual coinfection) Table2 details the data used to test each of the factors that potentiallyinfluence coinfection and details on the analysis of each dataset are provided in turn below

First to test the potential influence of these factors on cross-infectivity I conducted a factorial analysis of variance (ANOVA)on the cross-infectivity data set The dependent variable wasthe amount of phage that could infect each host measured spe-cifically as the proportion of tested phage that could infect eachhost because the infection matrices in this data set werederived from different studies testing varying numbers ofphages Thus as the number of phage that could infect eachhost (cross-infectivity) increased the potential for coinfectionwas greater The five independent variables tested were thestudy typesource (natural coevolution artificial) bacterialtaxon (roughly corresponding to Genus) habitat from whichbacteria and phages were isolated bacterial energy source (pho-tosynthetic or heterotrophic) and bacterial association(eg pathogen free-living) Geographic origin and phage taxawere present in original the metadata but were largely incom-plete therefore they were not included in the analyses Modelcriticism suggested that ANOVA assumptions were reasonablymet (Supplementary Fig S1) despite using proportion data(Warton and Hui 2011) ANOVA on the arc-sine transform of theproportions and a binomial regression provided qualitativelysimilar results (data not shown see associated code in FigSharerepository httpdxdoiorg106084m9figshare2072929)

Second to test the factors influencing culture coinfection Iconducted a negative binomial regression on the culture coin-fection data set The number of extrachromosomal virusesinfecting each host culture as detected by sequence data wasthe dependent variable These viruses represent ongoing infec-tions (eg lytic chronic or carrier state) as opposed to viralsequences integrated into the genome that may or may not beactive Thus as more extrachromosomal infections are detectedin a host culture culture coinfection increases The five explan-atory variables tested were the number of prophage sequencesssDNA virus presence energy source (eg heterotrophicauto-trophic) taxonomic rank (Genus was selected as the best pre-dictor over Phylum or Family details in code in FigSharerepository httpdxdoiorg106084m9figshare2072929) andhabitat (environmental host-associated or engineered asdefined by GOLD specifications Mukherjee et al 2017) I con-ducted stepwise model selection using Akaikersquos InformationCriterion (AIC) a routinely used measure of entropy that ranksthe relative quality of alternative models (Akaike 1974) to arriveat a reduced model that minimized the deviance of theregression

Third to test the factors influencing single cell coinfection Iconducted a Poisson regression on single cell data set Thenumber of actively infecting viruses in each cell as detected by

sequence data was the dependent variable This variableexcludes viral sequences integrated into the genome (eg pro-phages) that may or may not be active Thus as the number ofactive viral infections detected increase single cell coinfectionis greater The four explanatory variables tested were phyloge-netic cluster (based on small subunit rRNA amplicon sequenc-ing) ocean depth number of prophage sequences and numberof CRISPR spacers I conducted stepwise model selection withAIC (as done with the culture coinfection data set see above) toarrive at a reduced model that minimized deviance

23 Virusndashvirus interactions (prophages and ssdsDNA)in culture and single cell coinfection

To obtain more detailed information (beyond the regressionanalyses) on the influence of virusndashvirus interactions on the fre-quency and extent of coinfection I analyzed the effect of pro-phage sequences in culture and single-cell coinfections and theinfluence of ssDNA and dsDNA viruses on culture coinfection

The examinations of prophage infections were entirelysequence-based in the culture coinfection and single-cell coin-fection data sets Because only sequence data is used the activ-ity of these prophage sequences cannot be confirmed Forinstance the prophage sequences in the original single-celldata set were conservatively termed lsquoputative defective pro-phagesrsquo (Roux et al 2014) because sequences were too small tobe confidently assigned as complete prophage sequences (Rouxpers comm) Therefore I will hereafter be primarily using thephrase lsquoprophage sequencesrsquo (especially when indicatingresults) synonymously with lsquoprophagesrsquo primarily for gram-matical convenience To determine whether prophage sequen-ces affected the frequency and extent of coinfection in theculture coinfection data set I examined host cultures infectedexclusively by prophage sequences or extrachromosomalviruses (representing chronic carrier state and lsquoextrachromo-somal prophagersquo infections) and all prophage-infected culturesI tested whether prophage-only coinfections were infected by adifferent average number of viruses compared toextrachromosomal-only coinfections using a Wilcoxon RankSum test I tested whether prophage-infected cultures weremore likely than not to be coinfections and whether these coin-fections were more likely to occur with additional prophagesequences or extrachromosomal viruses

To examine the effect of prophage sequences on coinfectionfrequency in the single-cell data set I examined cells with pro-phage sequences and calculated the proportion of those thatalso had active infections To determine the extent of coinfec-tion in cells with prophage sequences I calculated the averagenumber of active viruses infecting these cells I examined differ-ences in the frequency of active infection between bacteria withor without prophage sequences using a proportion test and dif-ferences in the average amount of current viral infections usinga Wilcoxon rank sum test

Table 2 Factors explaining coinfection tested with each of the three data sets and (specific variable tested)

Cross-infectivity (potential coinfection) Culture-level coinfection Single-cell coinfection

Ecology Yes (habitat association) Yes (habitat) Yes (depth)Taxonomic group Yes (rank) Yes (rank) Yes (rRNA cluster)Host defense sequences No No Yes (CRISPR)Virusndashvirus interactions NA Yes (prophage ssDNA) Yes (prophage)


To examine whether ssDNA and dsDNA viruses exhibitednonrandom patterns of culture coinfection as found by Rouxet al (2014) for single cells of SUP05 marine bacteria I examinedthe larger culture coinfection data set (based on Roux et al2015) that is composed of sequence-based viral detections in allbacterial and archaeal NCBI genome sequences I first comparedthe frequency of dsDNAndashssDNA mixed coinfections againstssDNA-only coinfections among all host cultures coinfectedwith at least one ssDNA virus (nfrac14 331) using a binomial test Toprovide context for this finding (because ssDNA viruses were aminority of detected viral infections) I also examined the preva-lence of dsDNA-only and dsDNA-mixed coinfections and com-pared the proportion of ssDNA-mixedssDNA-total coinfectionswith the proportion of dsDNA-mixeddsDNA-total coinfectionsusing a proportion test To examine whether potential biases inthe detectability of ssDNA viruses relative to dsDNA viruses(Roux et al 2016) could affect ssDNAndashdsDNA coinfection detec-tion I also tested the percent detectability of ssDNA viruses inthe overall data set compared to the percentage of coinfectedhost cultures that contained at least one ssDNA virus using aproportion test Finally because the extended persistent infec-tions maintained by some ssDNA viruses (particularly in theInoviridae family) have been proposed to underlie previous pat-terns of enhanced coinfection between dsDNA and ssDNAviruses I examined the proportion of coinfections involvingssDNA viruses that had at least one virus of the Inoviridaeamong ssDNA-only coinfections ssDNA single infections andssDNAndashdsDNA coinfections

24 Effect of CRISPR spacers on single cell coinfection

In order to obtain a more detailed picture of the potential role ofbacterial defense mechanisms in coinfection I investigated theeffect of CRISPR spacers (sequences retained from past viralinfections by CRISPR-Cas defense mechanisms) on the fre-quency of cells undergoing active infections in the single celldata set (nfrac14 127 cells of SUP05 bacteria) I examined cells thatharbored CRISPR spacers in the genome and calculated the pro-portion of those that also had active infections To determinethe extent of coinfection in cells with CRISPR spacers I calcu-lated the average number of active viruses infecting these cellsI examined differences in the frequency of cells undergoingactive infections in cells with or without CRISPR spacers using aproportion test I examined differences in the average numberof current viral infections in cells with or without CRISPRspacers using a Wilcoxon rank sum test

25 Statistical analyses

I conducted all statistical analyses in the R statistical program-ming environment (R Development Core Team 2011) and gener-ated graphs using the ggplot2 package (Wickham 2009) Meansare presented as means 6 SD unless otherwise stated Data andcode for analyses and figures are available in the FigShare datarepository (httpdxdoiorg106084m9figshare2072929)

3 Results31 Host ecology and virusndashvirus interactionsconsistently explain variation in potential (cross-infectivity) culture and single-cell coinfection

To test the factors that influence viral coinfection of microbes Iconducted regression analyses on each of the three datasets (cross-infectivity culture coinfection and single cell

coinfection) The explanatory variables tested in each data setvaried slightly (Table 2) but can be grouped into host ecologyvirusndashvirus interactions host taxonomic or phylogenetic groupand sequences associated with host defense Overall host ecol-ogy and virusndashvirus interactions (association of ongoing infec-tions with other viruses eg prophages or ssDNA viruses)appeared as statistically significant factors that explained asubstantial amount of variation in coinfection in every data setthey were tested (see details for each data set in subheadingsbelow) Host taxonomy was a less consistent predictor acrossthe data sets The taxonomic group of the host explained littlevariation or was not a statistically significant predictor of cross-infectivity and single-cell coinfection respectively Howeverhost taxonomy was the strongest predictor of the amount ofculture coinfection judging by the amount of devianceexplained in the negative binomial regression Sequences asso-ciated with host defense mechanisms specifically CRISPRspacers were tested only in the single cell data set (nfrac14 127 cellsof SUP05 marine bacteria) and were a statistically significantand strong predictor of coinfection

32 Potential for coinfection (cross-infectivity) is shapedby bacterial ecology and taxonomy

To test the viral and host factors that affected cross-infectivityI conducted an analysis of variance on the cross-infectivity dataset a compilation of 38 phage host-range studies that measuredthe ability of different phages (nfrac14 499) to infect different bacte-rial hosts (nfrac14 1005) using experimental infections of culturedmicrobes in the laboratory (Flores et al 2011) In this data setcross-infectivity is the proportion of tested phage that couldinfect each host Stepwise model selection with AIC yielded areduced model that explained 3389 of the variance in cross-infectivity using three factors host association (eg free-livingpathogen) isolation habitat (eg soil animal) and taxonomicgrouping (Fig 1) The two ecological factors together explain-edgt30 of the variance in cross-infectivity while taxonomyexplained only 3 (Fig 1D) Host association explained 841of the variance in cross-infectivity Bacteria that were patho-genic to cows had the highest potential coinfection with gt75of tested phage infecting each bacterial strain (Fig 1B) In abso-lute terms the average host that was a pathogenic to cowscould be infected 15 phages on average The isolation habitatexplained 2436 of the variation in cross-infectivity with clini-cal isolates having the highest median cross-infectivity fol-lowed by sewagedairy products soil sewage and laboratorychemostats All these habitats had gt75 of tested phage infect-ing each host on average (Fig 1A) and in absolute terms theaverage host in each of these habitats could be infected by 3ndash15different phages

Bacterial energy source (heterotrophic autotrophic) and thetype of study (natural coevolution artificial) were not selectedby AIC in final model An alternative categorization of habitatthat matched the culture coinfection data set (ecosystem host-associated engineered environmental) was not a statisticallysignificant predictor and was dropped from the model Resultsfrom this alternative categorization (Supplementary Fig S1 andTable S2) show that bacteria-phage groups isolated from host-associated ecosystems (eg plants humans) had the highestcross-infectivity followed by engineered ecosystems (18lower) and environmental ecosystems (44 lower) Details ofthe full reduced and alternative models are provided inSupplementary Figure S2 and associated code in the FigSharerepository (httpdxdoiorg106084m9figshare2072929)


33 Culture coinfection is influenced by host ecologytaxonomy and virusndashvirus interactions

To test the viral and host factors that affected culture coinfec-tion I conducted a negative binomial regression on the culturecoinfection data set which documented 12498 viral infectionsin 5492 microbial hosts using NCBI-deposited sequencedgenomes (Roux et al 2015) supplemented with metadata col-lected from the GOLD database (Mukherjee et al 2017) Culturecoinfection was measured as the number of extrachromosomalvirus sequences (representing lytic chronic or carrier stateinfections) detected in each host culture Stepwise model selec-tion with AIC resulted in a reduced model that used three fac-tors to explain the number of extrachromosomal infectionshost taxonomy (Genus) number of prophages and host ecosys-tem The genera with the most coinfection (meangt25extrachromosomal infections) were Avibacterium ShigellaSelenomonas Faecalibacterium Myxococcus and Oenococcuswhereas Enterococcus Serratia Helicobacter Pantoea EnterobacterPectobacterium Shewanella Edwardsiella and Dickeya were rarely(meanlt 045 extrachromosomal infections) coinfected (Fig 2A)

Microbes that were host-associated had the highest meanextrachromosomal infections (139 6 162 nfrac14 4282) followed byengineered (132 6 144 nfrac14 281) and environmental (095 6 097nfrac14 450) ecosystems (Fig 2B) The model predicts that each addi-tional prophage sequence yields 79 the amount of extrachro-mosomal infections or a 21 reduction (Fig 2C)

Host energy source (heterotrophic autotrophic) and ssDNAvirus presence were not statistically significant predictors asjudged by AIC model selection Details of the full reduced andalternative models are provided in the Supplementary Materialsand all associated code and data is deposited in FigShare reposi-tory (httpdxdoiorg106084m9figshare2072929)

34 Single-cell coinfection is influenced by bacterialecology virusndashvirus interactions and sequencesassociated with the CRISPR-Cas defense mechanism

To test the viral and host factors that affected viral coinfectionof single cells I constructed a Poisson regression on thesingle-cell coinfection data set which characterized 143 viral

barley rootsbovine and ovine feces

clinical isolatesdairy products

fecesfood fresh water soil sewage

fresh watergastroenteritis patients

hospital sewage fresh waterhuman and animal

human and animal fecal isolatesIndustrial cheese plants

lab batch culturelab chemostatlab agar plates

marineplant soil

poultry intestinesauerkraut

sewagesewage dairy products

silage (fermented bovine feed)soil

soil freshwater plantswine lagoon

yogurt industrial

000 025 050 075 100Prop of tested phage infecting

Isol

atio

n H

abita

t

bovine pathogen

fish pathogen

free

human pathogen

human pathogen oysters

human symbiont

mycorrhizal

plant symbiont


Bac

teria

l Ass

ocia

tion

BurkholderiaCampylobacter

CellulophagabalticaEnterobacteriaceae

EscherichiaEscherichia and Salmonella

Escherichia coliFlavobacteriaceae

LactobacillusLactococcus lactis

LeuconostocMycobacterium

PseudoalteromonasPseudomonas

RhizobiumSalmonella

StaphylococcusStreptococcus

Synechococcus and AnacystisSynechococcus and Prochlorococcus

Vibrio


Bac

teria

l Tax

aBA

C D

BacterialAssociation

Isolation_Habitat_new

Bacteria_new

Residuals

Df

7

22

5

970

Sum Sq

106497

308534

4259

809121

Mean Sq

15214

14024

08518

00834

F value

182389

168127

102115

NA

Pr(gtF)

0

0

0

NA

Figure 1 Bacterial-phage ecology and taxonomy explain most of the variation in cross-infectivity The data set is a compilation of 38 phage host range studies that

measured the ability of different phages (nfrac14499) to infect different bacterial hosts (nfrac141005) using experimental infections of cultured microbes in the laboratory (ie

the lsquospot testrsquo) Cross-infectivity or potential coinfection is the number of phages that can infect a given bacterial host here measured as the proportion of tested

phages infecting each host (represented by points) Cross-infectivity was the response variable in an analysis of variance (ANOVA) that examined the effect of five fac-

tors (study type bacterial taxon habitat from which bacteria and phages were isolated bacterial energy source and bacterial association) on variation in the response

variable Those factors selected after stepwise model selection using Akaikersquos Information Criterion (AIC) are depicted in panels AndashC In these panels each point repre-

sents a bacterial host hosts with a value of zero on the x-axis could be infected by none of the tested phage whereas those with a value of one could be infected by all

the tested phages Thus the x-axis is a scale of the potential for coinfection Point colors correspond to hosts that were tested in the same study Note data points are

offset by a random and small amount (jittered) to enhance visibility and reduce overplotting The ANOVA table is presented in panel D


infections in SUP-05 marine bacteria (sulfur-oxidizingGammaproteobacteria) isolated as single cells (nfrac14 127) directlyfrom their habitat (Roux et al 2014) Single cell coinfection wasmeasured as the number of actively infecting viruses detected

by sequence data in each cell Stepwise model selection withAIC led to a model that included three factors explaining thenumber of actively infecting viruses in single cells number ofprophages number of CRISPR spacers and ocean depth at

(Intercept)

ECOSYSTEMEnvironmental

GenBacteroides

GenDickeya

GenEdwardsiella

GenEnterobacter

GenEnterococcus

GenHelicobacter

GenListeria

GenNeisseria

GenPantoea

GenSerratia

GenStaphylococcus

GenStenotrophomonas

GenStreptococcus

GenVibrio

GenXanthomonas

prophage

Coef

082

023

095

238

217

142

143

149

099

084

149

146

095

148

087

086

085

023

Std_Error

037

01

045

081

109

054

039

048

047

039

064

06

038

065

038

038

04

002

zvalue

22

228

211

293

198

262

363

311

212

215

233

243

25

229

229

225

214

1387

p

003

002

004

0

005

001

0

0

003

003

002

002

001

002

002

002

003

0

Exp_Coef

227

08

039

009

011

024

024

022

037

043

022

023

039

023

042

042

043

079DickeyaEdwardsiella

PectobacteriumShewanella

EnterobacterPantoea

HelicobacterSerratia

EnterococcusStenotrophomonas

AtopobiumAzospirillum

CollinsellaDorea

ErwiniaParaprevotella

PhascolarctobacteriumPhotobacterium

ProteusRoseburia

SodalisRalstonia

HalorubrumListeria

BordetellaCandidatus Arthromitus

GluconobacterSulfolobus

ThaueraBifidobacterium

BurkholderiaStaphylococcus

VibrioCronobacterLeptotrichia

NovosphingobiumWolbachia

StreptococcusBacteroides

ActinomycesHaloferax

AeromonasXanthomonas

NeisseriaPseudoalteromonas

PseudomonasPrevotella

AggregatibacterAlicyclobacillus

AlistipesBradyrhizobium

CapnocytophagaCarnobacterium

CitrobacterComamonas

CrocosphaeraDesulfovibrioEubacteriumGardnerella

GordoniaLeuconostoc

LoktanellaMethanobrevibacter

PediococcusPelosinus

PorphyromonasSaccharopolyspora

SalinisporaSUP05

ThermococcusCampylobacter

SalmonellaMycobacterium

ActinobacillusRhodobacter

CorynebacteriumPaenibacillusStreptomyces

VeillonellaMannheimia

SphingomonasXylella

ClostridiumFrankia

KomagataeibacterOscillibacter

FusobacteriumRuminococcus

LactococcusAgrobacterium

AliivibrioDialister

MethylobacteriumMoraxella

PeptostreptococcusBacillus

MesorhizobiumMegasphaera

NatrialbaProvidencia

LactobacillusBlautia

Candidatus LiberibacterKurthia

LeptolyngbyaLysinibacillus

MagnetospirillumMorganella

NitratireductorRiemerellaLeptospira

unknownBrucella

DelftiaOchrobactrumRhodococcusHaemophilusSphingobiumBrevibacillusNocardiopsis

BartonellaBrachyspira

GallibacteriumPasteurella

PhaeospirillumPhotorhabdus

EscherichiaKlebsiella

SinorhizobiumAcetobacter

AcinetobacterYersinia

FaecalibacteriumMyxococcusOenococcus

SelenomonasShigella

Avibacterium

0 1 2 3 4Mean extrachromosomal viruses infecting

Gen

us

Engineered

Environmental

Host associated

0 1 2 3 4 5 6 7 8 9 10 11 12 13Extrachromosomal viruses infecting

Eco

syst

em

0123456789

10111213

0 3 6 9Number of prophages in host

Ext

rach

rom

osom

al

viru

ses

infe

ctin

g

BA

C

D

Figure 2 Host taxonomy ecology and number of prophage sequences predict variation in culture coinfection The data set is composed of viral infections detected

using sequence data (nfrac1412498) in all bacterial and archaeal hosts (nfrac145492) with sequenced genomes in the National Center for Biotechnology Informationrsquos (NCBI)

databases supplemented with data on host habitat and energy source collected primarily from the Joint Genome Institutersquos Genomes Online Database (JGI-GOLD)

Extrachromosomal viruses represent ongoing acute or persistent infections in the microbial culture that was sequenced Thus increases in the axes labeled

lsquo extrachromosomal viruses infectingrsquo represent increasing viral coinfection in the host culture (not necessarily in single cells within that culture) The number of

extrachromosomal viruses infecting a host was the response variable in a negative binomial regression that tested the effect of five variables (the number of prophage

sequences ssDNA virus presence host taxon host energy source and habitat ecosystem) on variation in the response variable Plots AndashC depict all variables retained

after stepwise model selection using Akaikersquos Information Criterion (AIC) In Panel A each point represents a microbial genus with its corresponding mean number of

extrachromosomal virus infections (only genera withgt1 host sampled and nonzero means are included) The colors of each point correspond to each genus In Panels

B and C each point represents a unique host with a corresponding number of extrachromosomal virus infections Panel B groups hosts according to their habitat eco-

system as defined by JGI-GOLD database schema the point colors correspond to each of these three ecosystem categories Panel C groups hosts according to the num-

ber of prophage sequences detected in host sequences the color shades of points become lighter in hosts with more detected prophages Panels B and C have data

points are offset by a random and small amount (jittered) to enhance visibility and reduce overplotting The regression table is presented in Panel D and only includes

variables with Plt005


which cells were collected (Fig 3) The model estimated eachadditional prophage would result in 9 of the amount ofactive infections (ie a 91 reduction) while each additionalCRISPR spacer would result in a 54 reduction Depth had apositive influence on coinfection every 50-m increase in depthwas predicted to result in 80 more active infections

The phylogenetic cluster of the SUP-05 bacterial cells (basedon small subunit rRNA amplicon sequencing) was not a statisti-cally significant predictor selected using AIC model selectionDetails of the full reduced and alternative models are providedin the Supplementary Materials and all associated code anddata is deposited in FigShare repository (httpdxdoiorg106084m9figshare2072929)

35 Prophages limit culture and single-cell coinfection

Based on the results of the regression analyses showing thatprophage sequences limited coinfection and aiming to test thephenomenon of superinfection exclusion in a large data set Iconducted a more detailed analysis of the influence of pro-phages on coinfection in microbial cultures I also conducted asimilar analysis using the single cell data set which is relatively

smaller (nfrac14 127 host cells) but provides better single-cell levelresolution

First because virusndashvirus interactions can occur in culturesof cells I tested whether prophage-infected host culturesreduced the probability of other viral infections in the entire cul-ture coinfection data set A majority of prophage-infected hostcultures were coinfections (5648 of nfrac14 3134) a modest butstatistically distinguishable increase over a 05 null (BinomialExact Pfrac14 4344e13 Fig 4A) Of these coinfected host cultures(nfrac14 1770) cultures with more than one prophage sequence(3254) were 2 times less frequent than those with both pro-phage sequences and extrachromosomal viruses (BinomialExact Plt 22e16 Fig 4B) Therefore integrated prophagesappear to reduce the chance of the culture being infected withadditional prophage sequences but not additional extrachro-mosomal viruses Accordingly host cultures co-infected exclu-sively by extrachromosomal viruses (nfrac14 675) were infected by325 6 134 viruses compared to 254 6 102 prophage sequences(nfrac14 575) these quantities showed a statistically significant dif-ference (Wilcoxon Rank Sum Wfrac14 1253985 Plt 22e16 Fig 4C)

Second to test whether prophage sequences could affectcoinfection of single cells in a natural environment I examineda single cell amplified genomics data set of SUP05 marine

0

1

2

3

4

5

100 150 185Ocean depth (m)

Num

ber

of v

iruse

s ac

tivel

y in

fect

ing

cell

0

1

2

3

4

5

0 1 2Number of putative defective prophages

Num

ber

of v

iruse

s ac

tivel

y in

fect

ing

cell

0

1

2

3

4

5

0 1CRISPR spacers

Num

ber

of v

iruse

s ac

tivel

y in

fect

ing

cell

(Intercept)

Depth

CRISPR

Defectiveprophage

Exp_Coef

0111

1016

0463

009

Robust SE

0091

0005

013

0085

LL

0022

1006

0267

0014

UL

0553

1026

0803

0571

p

0007

0001

0006

0011

DC

A B

Figure 3 Host ecology number of prophage sequences and CRISPR spacers predict variation in single-cell coinfection The data set is composed of viral infections

(nfrac14143) detected using genome sequence data in a set of SUP-05 (sulfur-oxidizing Gammaproteobacteria) marine bacteria isolated as single cells (nfrac14127) from the

Saanich Inlet in British Columbia Canada In Panels AndashC each point represents a single-cell amplified genome (SAG) The y-axis depicts the number of viruses involved

in active infections (active at the time of isolation eg ongoing lytic infections) in each cell Thus increases in the y-axis represent increasing active viral coinfection of

single cells The number of viruses actively infecting each cell was the response variable in a Poisson regression that tested the effect of four variables (small subunit

rRNA phylogenetic cluster ocean depth number of prophage sequences and number of CRISPR spacers) on variation in the response variable Panels AndashC depict all

variables retained after stepwise model selection using Akaikersquos Information Criterion (AIC) and group cells according to the ocean depth at which the cells were iso-

lated (A) the number of prophage sequences detected (B conservatively termed putative defective prophages here as in the original published data set because

sequences were too small to be confidently assigned as complete prophages) and the number of CRISPR spacers detected (C) point colors correspond to the categories

of each variable in each plot Each point is offset by a random and small amount (jittered) to enhance visibility and reduce overplotting The regression table is pre-

sented in panel D


bacteria Cells with prophage sequences (conservatively termedputative defective prophages in the original published data setas in Fig 3B because sequences were too small to be confi-dently assigned as complete prophages Roux pers comm)were less likely to have current infections 909 of cells withprophage sequences had active viral infections compared to7455 of cells that had no prophage sequences (v2frac14 142607dffrac14 1 P valuefrac14 000015) Bacteria with prophage sequenceswere undergoing active infection by an average of 009 6 030phages whereas bacteria without prophages were infected by122 6 105 phages (Fig 3B) No host with prophages had currentcoinfections (ie gt1 active virus infection)

36 Nonrandom coinfection of host cultures byssDNA and dsDNA viruses suggests mechanismsenhancing coinfection

The presence of ssDNA viruses in host cultures was not a statis-tically significant predictor of extrachromosomal infections inthe regression analysis of the culture coinfection data set (basedon Roux et al 2015) that is composed of sequence-based viraldetections in all bacterial and archaeal NCBI genome sequencesHowever to test the hypothesis that ssDNAndashdsDNA viral infec-tions exhibit nonrandom coinfection patterns (as found by Rouxet al 2014 for single SUP05 bacterial cells) I conducted a morefocused analysis on the culture coinfection data set Coinfectedhost cultures containing ssDNA viruses (nfrac14 331) were morelikely to have dsDNA or unclassified viruses (7069) than mul-tiple ssDNA infections (exact binomial Pfrac14 3314e14) Thesecoinfections weregt2 times more likely to involve at least onedsDNA virus than none (exact binomial Pfrac14 2559e11 Fig 5A)To examine whether differential detectability of ssDNA virusescould affect this result and place it in context of all the infec-tions recorded in the data set I conducted further tests Single-stranded DNA viruses represented 682 of the viruses detectedin host cultures in the overall data set (nfrac14 12490) Coinfectedhost cultures that contained at least one ssDNA virus repre-sented 619 of hosts in line with and not statistically differentfrom the aforementioned overall detectability of ssDNA viruses(v2frac14 322 dffrac14 1 P valuefrac14 007254) The most numerous coinfec-tions in the culture coinfection data set were dsDNA-only coin-fections (nfrac14 1898) The proportion of ssDNA-mixedtotal-ssDNA coinfections (071) was higher than the proportion ofdsDNA-mixeddsDNA-total (019) coinfections (Fig 5Bv2frac14 39855 dffrac14 1 Plt 22e16) Of all the ssDNA viruses detected(nfrac14 852) in the data set a majority belonged to the Inoviridaefamily (9096) with the remainder in the Microviridae (070) ortheir family was unknown (833) The proportion of hostsinfected with Inoviridae was highest in dsDNAndashssDNA coinfec-tions (096) higher than hosts with ssDNA-only coinfections(090) or than hosts with ssDNA single infections (086) the dif-ference in these three proportions was statistically differentfrom zero (v2frac14 1098 dffrac14 2 Pfrac14 000413)

37 CRISPR spacers limit coinfection at single cell levelwithout spacer matches

The regression analysis of the single-cell data set (nfrac14 127 cellsof SUP-05 marine bacterial cells) revealed that the presence ofCRISPR spacers had a significant overall effect on coinfection ofsingle cells therefore I examined the effects of CRISPR spacersmore closely Cells with CRISPR spacers were less likely to haveactive viral infections than those without spacers (v2frac14 1403dffrac14 1 P valuefrac14 000018) 3200 of cells with CRISPRs had active

00

01

02

03

04

05

Co infected Single Infections

Pro

port

ion

prop

hage

infe

cted

hos

ts

00

01

02

03

04

05

06

07

Mixed Prophage Only

Pro

port

ion

prop

hage

infe

cted

hos

ts

2

3

4

5

6

7

8

9

10

11

Extrachromosomal Only Prophage Only

Num

ber

of v

iruse

s in

eac

h ho

st c

ultu

re

B

C

A

Figure 4 Host cultures infected with prophages limit coinfection by other pro-

phages but not extrachromosomal viruses The initial data set was composed

of viral infections detected using sequence data in sequenced genomes of

microbial hosts available in National Center for Biotechnology Information

(NCBI) databases A slight but statistically significant majority of host cultures

with prophage sequences (nfrac143134) were coinfected ie had more than one

extrachromosomal virus or prophage detected (Panel A blue bar) Of these host

cultures containing multiple prophages were less frequent than those contain-

ing a mix of prophages and extrachromosomal (eg acute or persistent) infec-

tions (Panel B) On average (black horizontal bars) extrachromosomal-only

coinfections involved more viruses than prophage-only coinfections (Panel C)

In Panel C each point represents a bacterial or archaeal host and is offset by a

random and small amount to enhance visibility


viral infections compared to 8095 percent of bacteria withoutCRISPR spacers Bacterial cells with CRISPR spacers had068 6 107 current phage infections compared to 121 6 100 forthose without spacers (Fig 3C) Bacterial cells with CRISPRspacers could have active infections and coinfections with up tothree phages None of the CRISPR spacers matched any of theactively infecting viruses (Roux et al 2014) using an exact matchsearch (Roux pers comm)

4 Discussion41 Summary of findings

The compilation of multiple large-scale data sets of virusndashhostinteractions (gt6000 hostsgt13000 viruses) allowed a system-atic test of the ecological and biological factors that influenceviral coinfection rates in microbes across a broad range of taxaand environments I found evidence for the importance andconsistency of host ecology and virusndashvirus interactions inshaping potential (cross-infectivity) culture and single-cellcoinfection In contrast host taxonomy was a less consistentpredictor of viral coinfection being a weak predictor of cross-infectivity and single cell coinfection but representing thestrongest predictor of host culture coinfections In the mostcomprehensive test of the phenomenon of superinfectionimmunity conferred by prophages (nfrac14 3134 hosts) I found thatprophage sequences were predictive of limited coinfection ofcultures by other prophages but less strongly predictive of coin-fection by extrachromosomal viruses Furthermore in a smallersample of single cells of SUP-05 marine bacteria (nfrac14 127 cells)prophage sequences completely excluded coinfection (by pro-phages or extrachromosomal viruses) In contrast I found evi-dence of increased culture coinfection by ssDNA and dsDNAphages suggesting mechanisms that may enhance coinfectionAt a fine-scale single-cell data revealed that CRISPR spacerslimit coinfection of single cells in a natural environmentdespite the absence of exact spacer matches in the infectingviruses suggesting a potential role for bacterial defense mecha-nisms In light of the increasing awareness of the widespreadoccurrence of viral coinfection this study provides the

foundation for future work on the frequency mechanisms anddynamics of viral coinfection and its ecological and evolution-ary consequences

42 Host correlates of coinfection

Host ecology stood out as an important predictor of coinfectionacross all three data sets cross-infectivity culture coinfectionand single cell coinfection The specific ecological variables dif-fered in the data sets (bacterial association isolation habitatand ocean depth) but ecological factors were retained as statis-tically significant predictors in all three models Moreoverwhen ecological variables were standardized between thecross-infectivity and culture coinfection data sets the differen-ces in cross-infectivity and coinfection among ecosystem cate-gories were remarkably similar (Supplementary Fig S5 andTable S2) This result implies that the ecological factors thatpredict whether a host can be coinfected are the same that pre-dict whether a host will be coinfected These results lend fur-ther support to the consistency of host ecology as a predictor ofcoinfection On the other hand host taxonomy was a less con-sistent predictor It was weak or absent in the cross-infectivityand single-cell coinfection models respectively yet it was thestrongest predictor of culture coinfection This difference couldbe because the hosts in the cross-infectivity and single-celldata sets varied predominantly at the strain level (Flores et al2011 Roux et al 2014) whereas infection patterns are less vari-able at the Genus level and higher taxonomic ranks (Floreset al 2013 Roux et al 2015) as seen with the culture coinfectionmodel Collectively these findings suggest that the diverse andcomplex patterns of cross-infectivity (Holmfeldt et al 2007)and coinfection observed at the level of bacterial strains maybe best explained by local ecological factors while at highertaxonomic ranks the phylogenetic origin of hosts increases inimportance (Flores et al 2011 2013 Roux et al 2015) Particularbacterial lineages can exhibit dramatic differences in cross-infectivity (Koskella and Meaden 2013 Liu et al 2015) and asthis study shows coinfection Thus further studies with eco-logical data and multi-scale phylogenetic information will be

ssDNA dsDNA

ssDNA only

ssDNA + unclassified

0 50 100 150 200Number of Hosts

Com

posi

tion

of C

oinf

ectio

ns

dsDNA mixed

dsDNA only

ssDNA only

ssDNA mixed

0 500 1000 1500Number of Hosts

Com

posi

tion

of C

oinf

ectio

ns

BA

Figure 5 Culture coinfections between single-stranded DNA (ssDNA) and double stranded DNA (dsDNA) viruses are more common than expected by chance The initial

data set was composed of viral infections detected using sequence data in sequenced genomes of microbial hosts available in National Center for Biotechnology

Information (NCBI) databases Panel A Composition of coinfections in all host cultures coinfected with at least one ssDNA virus (nfrac14331) at the time of sequencing

Blue bars are mixed coinfections composed of at least one ssDNA virus and at least one dsDNA or unclassified virus The grey bar depicts coinfections exclusively com-

posed of ssDNA viruses (ssDNA-only) Host culture coinfections involving ssDNA viruses were more likely to occur with dsDNA or unclassified viruses than with multi-

ple ssDNA viruses (exact binomial Pfrac143314e14) Panel B The majority of host culture coinfections in the data set exclusively involved dsDNA viruses (nfrac141898

dsDNA-only red bar) There were more mixed coinfections involving ssDNA viruses than ssDNA-only coinfections (compare blue and grey bar) while dsDNA-only coin-

fections were more prevalent than mixed coinfections involving dsDNA viruses (compare red bars) Accordingly the proportion of ssDNA-mixedtotal-ssDNA coinfec-

tions was higher than the proportion of dsDNA-mixeddsDNA-total coinfections (proportion test Plt22e16)


necessary to test the relative influence of bacterial phylogenyon coinfection

Sequences associated with the CRISPR-Cas bacterialdefense mechanism were another important factor influencingcoinfection patterns but were only tested in the comparativelysmaller single-cell data set (nfrac14 127 cells of SUP05 marine bacte-ria) The presence of CRISPR spacers reduced the extent ofactive viral infections even though these spacers had no exactmatches (Roux pers comm) to any of the infecting virusesidentified (Roux et al 2014) These results provide some of thefirst evidence from a natural environment that CRISPRrsquos pro-tective effects could extend beyond viruses with exact matchesto the particular spacers within the cell (Semenova et al 2011Fineran et al 2014) Although very specific sequence matchingis thought to be required for CRISPR-Cas-based immunity(Mojica et al 2005 Barrangou et al 2007 Brouns et al 2008) thesystem can tolerate mismatches in protospacers (within andoutside of the seed region Semenova et al 2011 Fineran et al2014) enabling protection (interference) against related phagesby a mechanism of enhanced spacer acquisition termed pri-ming (Fineran et al 2014) The seemingly broader protectiveeffect of CRISPR-Cas beyond specific sequences may helpexplain continuing effectiveness of CRISPR-Cas (Fineran et al2014) in the face of rapid viral coevolution for escape(Andersson and Banfield 2008 Tyson and Banfield 2008Heidelberg et al 2009) However caution is warranted asCRISPR spacers sequences alone cannot indicate whetherCRISPR-Cas is active Some bacterial taxa have inactive CRISPR-Cas systems notably Escherichia coli K-12 (Westra et al 2010)and phages also possess anti-CRISPR mechanisms (Seed et al2013 Bondy-Denomy et al 2015) that can counter bacterialdefenses In this case the unculturability of SUP05 bacteria pre-clude laboratory confirmation of CRISP-Cas activity (but seeShah et al 2017) but given the strong statistical association ofCRISPR spacer sequences with a reduction in active viral infec-tions (after controlling for other factors) it is most parsimoni-ous to tentatively conclude that the CRISPR-Cas mechanism isthe likeliest culprit behind these reductions To elucidate andconfirm the roles of bacterial defense systems in shaping coin-fection more data on CRISPR-Cas and other viral infectiondefense mechanisms will be required across different taxa andenvironments

This study revealed the strong predictive power of severalhost factors in explaining viral coinfection yet there was stillsubstantial unexplained variation in the regression models (seeResults) Thus the host factors tested herein should be regardedas starting points for future experimental examinations Forinstance host energy source was not a statistically significantpredictor in the cross-infectivity and culture coinfection mod-els perhaps due to the much smaller sample sizes of autotro-phic hosts However both data sets yield similar summarystatistics with heterotrophic hosts having 59 higher cross-infectivity and 77 higher culture coinfection (SupplementaryFig S6) Moreover other factors not examined such as geogra-phy could plausibly affect coinfection The geographic origin ofstrains can affect infection specificity such that bacteria iso-lated from one location are likely to be infected by more phageisolated from the same location as observed with marinemicrobes (Flores et al 2013) This pattern could be due to theinfluence of local adaptation of phages to their hosts (Koskellaet al 2011) and represents an interesting avenue for furtherresearch

43 The role of virusndashvirus interactions in coinfection

The results of this study suggest that virusndashvirus interactionsplay a role in limiting and enhancing coinfection First host cul-tures and single cells with prophage sequences showed limitedcoinfection In what is effectively the largest test of viral super-infection exclusion (nfrac14 3134 hosts) host cultures with pro-phage sequences exhibited limited coinfection by otherprophages but not as much by extrachromosomal virusesAlthough this focused analysis showed that prophage sequen-ces limited extrachromosomal infection less strongly than addi-tional prophage infections the regression analysis suggestedthat they did have an effect a statistically significant 21reduction in extrachromosomal infections for each additionalprophage detected As these were culture coinfections and notnecessarily single-cell coinfections these results are consistentwith a single-cell study of Salmonella cultures showing that lyso-gens can activate cell subpopulations to be transiently immunefrom viral infection (Cenens et al 2015) Prophage sequenceshad a more dramatic impact at the single-cell level in SUP05marine bacteria in a natural environment severely limitingactive viral infection and completely excluding coinfection Theresults on culture-level and single-cell coinfection come fromvery different data sets which should be examined carefullybefore drawing general patterns The culture coinfection dataset is composed of an analysis of all publicly available bacterialand archaeal genome sequences in NCBI databases that showevidence of a viral infection These sequences show a biastowards particular taxonomic groups (eg Proteobacteria modelstudy species) and those that are easy to grow in pure cultureThe single cell data set is limited to 127 cells of just one hosttype (SUP05 marine bacteria) isolated in a particular environ-ment as opposed to the 5492 hosts in the culture coinfectiondata set This limitation prohibits taxonomic generalizationsabout the effects on prophage sequences on single cells butextends laboratory findings to a natural environmentAdditionally both of these data sets use sequences to detect thepresence of prophages which means the activity of these pro-phage sequences cannot be confirmed as with any study usingsequence data alone Along these lines in the original singlecell coinfection data set (Roux et al 2014) the prophage sequen-ces were conservatively termed lsquoputative defective prophagesrsquo(as in Fig 3 here) because sequences were too small to be confi-dently assigned as complete prophage sequences (Roux perscomm) If prophages in either data set are not active this couldmean that bacterial domestication of phage functions(Asadulghani et al 2009 Bobay et al 2014) rather than phagendashphage interactions in a strict sense would explain protectionfrom infection conferred by these prophage sequences in hostcultures and single cells In view of these current limitations awider taxonomic and ecological range of culture and single-cellsequence data together with laboratory studies when possibleshould confirm and elucidate the role of prophage sequences inaffecting coinfection dynamics Interactions in coinfectionbetween temperate bacteriophages can affect viral fitness(Dulbecco 1952 Refardt 2011) suggesting latent infections are aprofitable avenue for future research on virusndashvirusinteractions

Second another virusndashvirus interaction examined in thisstudy appeared to increase the chance of coinfection Whileprophage sequences strongly limited coinfection in single cellsRoux et alrsquos (2014) original analysis of this same data set foundstrong evidence of enhanced coinfection (ie higher than


expected by random chance) between dsDNA and ssDNAMicroviridae phages in bacteria from the SUP05_03 cluster I con-ducted a larger test of this hypothesis using a different diverseset of 331 bacterial and archaeal hosts infected by ssDNAviruses included in the culture coinfection data set (Roux et al2015) This analysis provides evidence that ssDNAndashdsDNA cul-ture coinfections occur more frequently than would be expectedby chance extending the taxonomic applicability of this resultfrom one SUP05 bacterial lineage to hundreds of taxa Thusenhanced coinfection perhaps due to the long persistent infec-tion cycles of some ssDNA viruses particularly the Inoviridae(Rakonjac et al 2011) might be a major factor explaining find-ings of phages with chimeric genomes composed of differenttypes of nucleic acids (Diemer and Stedman 2012 Roux et al2013) Filamentous or rod-shaped ssDNA phages (familyInoviridae Szekely and Breitbart 2016) often conduct multiplereplications without killing their bacterial hosts (Rakonjac et al2011 Mai-Prochnow et al 2015 Szekely and Breitbart 2016)allowing more time for other viruses to coinfect the cell andproviding a potential mechanistic basis for the enhanceddsDNAndashssDNA coinfections In line with this prediction over96 of dsDNAndashssDNA host culture coinfections contained atleast one virus from the Inoviridae a greater (and statisticallydistinguishable) percentage than contained in ssDNA-only coin-fections or ssDNA single infections Notwithstanding caution iswarranted due to known biases against the detectability ofssDNA compared to dsDNA viruses in high-throughputsequencing library preparation (Roux et al 2016) In this studythis concern can be addressed by noting that not all thesequencing data in the culture coinfection data set was gener-ated by high-throughput sequencing the prevalence of hostcoinfections containing ssDNA viruses was virtually identical(within 063) to the detectability of ssDNA viruses in the over-all data set and the prevalence of ssDNA viruses in this data set(6) is in line with validated unbiased estimates of ssDNAvirus prevalence from natural (marine) habitats (Roux et al2016) However given that the ssDNA virus sample is compara-tively smaller than the dsDNA sample future studies shouldfocus on examining ssDNAndashdsDNA coinfection prevalence andmechanisms

Collectively these results highlight the importance of virusndashvirus interactions as part of the suite of evolved viral strategiesto mediate frequent interactions with other viruses from limit-ing to promoting coinfection depending on the evolutionaryand ecological context (Turner and Duffy 2008)

44 Implications and applications

In sum the results of this study suggest microbial host ecologyand virusndashvirus interactions are important drivers of the wide-spread phenomenon of viral coinfection An important implica-tion is that virusndashvirus interactions will constitute an importantselective pressure on viral evolution The importance of virusndashvirus interactions may have been underappreciated because ofan overestimation of the importance of superinfection exclu-sion (Dulbecco 1952) Paradoxically superinfection avoidancemay actually highlight the selective force of virusndashvirus interac-tions In an evolutionary sense this viral trait exists preciselybecause the potential for coinfection is high If this is correctvariability in potential and realized coinfection as found in thisstudy suggests that the manifestation of superinfection exclu-sion will vary across viral groups according to their ecologicalcontext Accordingly there are specific viral mechanisms thatpromote coinfection as found in this study with ssDNAdsDNA

coinfections and in other studies (Turner et al 1999 Dang et al2004 Cicin-Sain et al 2005 Altan-Bonnet and Chen 2015Aguilera et al 2017 Erez et al 2017) I found substantial varia-tion in cross-infectivity culture coinfection and single-cellcoinfection and in the analyses herein ecology was always astatistically significant and strong predictor of coinfection sug-gesting that the selective pressure for coinfection is going tovary across local ecologies This is in agreement with observa-tions of variation in viral genetic exchange (which requirescoinfection) rates across different geographic localities in a vari-ety of viruses (Held and Whitaker 2009 Trifonov et al 2009Dıaz-Mu~noz et al 2013)

These results have clear implications not only for the studyof viral ecology in general but for practical biomedical and agri-cultural applications of phages and bacteriaarchaea Phagetherapy is often predicated on the high host specificity ofphages but intentional coinfection could be an important partof the arsenal as part of combined or cocktail phage therapyThis study also suggests that viral coinfection in the micro-biome should be examined as part of the influence of thevirome on the larger microbiome (Reyes et al 2010 Minot et al2011 Pride et al 2012) Finally if these results apply in eukary-otic viruses and their hosts variation in viral coinfection ratesshould be considered in the context of treating and preventinginfections as coinfection likely represents the default conditionof human hosts (Wylie et al 2014) Coinfection and virusndashvirusinteractions have been implicated in changing disease out-comes for hosts (Vignuzzi et al 2006) altering epidemiology ofviral diseases (Nelson et al 2008) and impacting antimicrobialtherapies (Birger et al 2015) In sum the results of this studysuggest that the ecological context mechanisms and evolu-tionary consequences of virusndashvirus interactions should be con-sidered as an important subfield in the study of viruses

Supplementary data

Supplementary data are available at Virus Evolution online

Acknowledgements

Simon Roux kindly provided extensive assistance with pre-viously published data sets TBK Reddy kindly provideddatabase records from Joint Genome Institutersquos GenomesOnline Database I am indebted to Joshua Weitz BrittKoskella Jay Lennon and Nicholas Matzke for providinghelpful critiques and advice on an earlier version of thismanuscript A Faculty Fellowship to SLDM from New YorkUniversity supported this work

Data availability

A list and description of data sources are included inSupplementary Table S1 and the raw data and code used inthis paper are deposited in the FigShare data repository(httpdxdoiorg106084m9figshare2072929)

Conflict of interest None declared

ReferencesAbrahams M R et al (2009) lsquoQuantitating the Multiplicity of

Infection with Human Immunodeficiency Virus Type 1Subtype C Reveals a Non-Poisson Distribution of TransmittedVariantsrsquo Journal of Virology 83 3556ndash67


Aguilera E R et al (2017) lsquoPlaques formed by mutagenized viralpopulations have elevated coinfection frequenciesrsquo mBio8e02020ndash16

Akaike H (1974) lsquoA New Look at the Statistical ModelIdentificationrsquo IEEE Transactions on Automatic Control 196716ndash23

Altan-Bonnet N and Chen Y-H (2015) lsquoIntercellularTransmission of Viral Populations with Vesiclesrsquo Journal ofVirology 89 12242ndash4

Andersson A F and Banfield J F (2008) lsquoVirus PopulationDynamics and Acquired Virus Resistance in Natural MicrobialCommunitiesrsquo Science 320 1047ndash50

Asadulghani M et al (2009) lsquoThe Defective Prophage Pool ofEscherichia coli O157 Prophage-Prophage InteractionsPotentiate Horizontal Transfer of Virulence DeterminantsrsquoPLoS Pathogen 5 e1000408

Barrangou R et al (2007) lsquoCRISPR Provides Acquired ResistanceAgainst Viruses in Prokaryotesrsquo Science 315 1709ndash12

Bergh Oslash et al (1989) lsquoHigh Abundance of Viruses Found inAquatic Environmentsrsquo Nature 340 467ndash8

Berngruber T W Weissing F J and Gandon S (2010)lsquoInhibition of Superinfection and the Evolution of ViralLatencyrsquo Journal of Virology 84 10200ndash8

Bertani G (1953) lsquoLysogenic Versus Lytic Cycle of PhageMultiplicationrsquo Cold Spring Harbor Symposia on QuantitativeBiology 18 65ndash70

(1954) lsquoStudies on Lysogenesis III Superinfection ofLysogenic Shigella dysenteriae with Temperate Mutants of theCarried Phagersquo Journal of Bacteriology 67 696ndash707

Birger R B et al (2015) lsquoThe Potential Impact of Coinfection onAntimicrobial Chemotherapy and Drug Resistancersquo Trends inMicrobiology 23 537ndash44

Bobay L-M Touchon M and Rocha E P C (2014) lsquoPervasiveDomestication of Defective Prophages by Bacteriarsquo Proceedingsof the National Academy of Sciences of the United States of America111 12127ndash32

Bondy-Denomy J et al (2015) lsquoMultiple Mechanisms for CRISPRndashCas Inhibition by anti-CRISPR Proteinsrsquo Nature 526 136ndash9

Brouns S J J et al (2008) lsquoSmall CRISPR RNAs Guide AntiviralDefense in PROKARYOTESrsquo Science 321 960ndash4

Cenens W et al (2015) lsquoViral Transmission Dynamics at Single-Cell Resolution Reveal Transiently Immune SubpopulationsCaused by a Carrier State Associationrsquo PLoS Genetics 11e1005770

Cicin-Sain L et al (2005) lsquoFrequent Coinfection of Cells ExplainsFunctional In Vivo Complementation betweenCytomegalovirus Variants in the Multiply Infected HostrsquoJournal of Virology 79 9492ndash502

Dang Q et al (2004) lsquoNonrandom HIV-1 Infection and DoubleInfection Via Direct and Cell-Mediated Pathwaysrsquo Proceedingsof the National Academy of Sciences of the United States of America101 632ndash7

DaPalma T et al (2010) lsquoA Systematic Approach to Virus-VirusInteractionsrsquo Virus Research 149 1ndash9

Delbruck M (1946) lsquoBacterial Viruses or BacteriophagesrsquoBiological Reviews 21 30ndash40

Dıaz-Mu~noz S L and Koskella B (2014) lsquoBacteriandashPhageInteractions in Natural Environmentsrsquo Advances in AppliedMicrobiology 89 135ndash83

et al (2013) lsquoElectrophoretic Mobility ConfirmsReassortment Bias Among Geographic Isolates of SegmentedRNA Phagesrsquo BMC Evolutionary Biology 13 206

Diemer G S and Stedman K M (2012) lsquoA Novel Virus GenomeDiscovered in an Extreme Environment Suggests

Recombination Between Unrelated Groups of RNA and DNAVirusesrsquo Biology Direct 7 13

Dropulic B Hermankova M and Pitha P M (1996) lsquoAConditionally Replicating HIV-1 Vector Interferes with Wild-Type HIV-1 Replication and Spreadrsquo Proceedings of the NationalAcademy of Sciences of the United States of America 93 11103ndash8

Dulbecco R (1952) lsquoMutual Exclusion Between Related PhagesrsquoJournal of Bacteriology 63 209ndash17

Ellis E L and Delbruck M (1939) lsquoThe Growth of BacteriophagersquoJournal of General Physiology 22 365ndash84

Erez Z et al (2017) lsquoCommunication Between Viruses GuidesLysis-Lysogeny Decisionsrsquo Nature 541 488ndash93

Fineran P C et al (2014) lsquoDegenerate Target Sites Mediate RapidPrimed CRISPR Adaptationrsquo Proceedings of the National Academyof Sciences of the United States of America 111 E1629ndash38

Flores C O et al (2011) lsquoStatistical Structure of HostndashPhageInteractionsrsquo Proceedings of the National Academy of Sciences ofthe United States of America 108 E288ndash97

Valverde S and Weitz J S (2013) lsquoMulti-Scale Structureand Geographic Drivers of Cross-Infection Within MarineBacteria and Phagesrsquo The ISME Journal 7 520ndash32

Ghedin E et al (2005) lsquoLarge-Scale Sequencing of HumanInfluenza Reveals the Dynamic Nature of Viral GenomeEvolutionrsquo Nature 437 1162ndash6

Heidelberg J F et al (2009) lsquoGerm Warfare in a Microbial MatCommunity CRISPRs Provide Insights into the Co-Evolution ofHost and Viral Genomesrsquo Plos One 4 e4169

Held N L and Whitaker R J (2009) lsquoViral BiogeographyRevealed by Signatures in Sulfolobus islandicus GenomesrsquoEnvironmental Microbiology 11 457ndash66

Holmfeldt K et al (2007) lsquoLarge Variabilities in HostStrain Susceptibility and Phage Host Range GovernInteractions Between Lytic Marine Phages andTheir Flavobacterium Hostsrsquo Applied and EnvironmentalMicrobiology 73 6730ndash9

Horvath P and Barrangou R (2010) lsquoCRISPRCas The ImmuneSystem of Bacteria and Archaearsquo Science 327 167ndash70

Joseph S B et al (2009) lsquoCoinfection Rates in Phi 6Bacteriophage are Enhanced by Virus-Induced Changes inHost Cellsrsquo Evolutionary Applications 2 24ndash31

Kaiser D and Dworkin M (1975) lsquoGene Transfer toMyxobacterium by Escherichia coli Phage P1rsquo Science 187 653ndash4

Koskella B and Meaden S (2013) lsquoUnderstandingBacteriophage Specificity in Natural Microbial CommunitiesrsquoViruses 5 806ndash23

et al (2011) lsquoLocal Biotic Environment Shapes the SpatialScale of Bacteriophage Adaptation to Bacteriarsquo The AmericanNaturalist 177 440ndash51

Labonte J M et al (2015) lsquoSingle-Cell Genomics-Based Analysisof Virus-Host Interactions in Marine SurfaceBacterioplanktonrsquo The ISME Journal 9 2386ndash99

Labrie S J Samson J E and Moineau S (2010) lsquoBacteriophageResistance Mechanismsrsquo Nature 8 317ndash27

Li C et al (2010) lsquoReassortment Between Avian H5N1 andHuman H3N2 Influenza Viruses Creates Hybrid Viruses withSubstantial Virulencersquo Proceedings of the National Academy ofSciences of the United States of America 107 4687ndash92

Linn S and Arber W (1968) lsquoHost Specificity of DNA Producedby Escherichia coli X In Vitro Restriction of Phage fd ReplicativeFormrsquo Proceedings of the National Academy of Sciences of the UnitedStates of America 59 1300ndash6

Liu J et al (2015) lsquoThe Diversity and Host Interactions ofPropionibacterium acnes Bacteriophages on Human Skinrsquo TheISME Journal 9 2078ndash93


Mai-Prochnow A et al (2015) lsquoBig Things in Small Packages TheGenetics of Filamentous Phage and Effects on Fitness of TheirHostrsquorsquo FEMS Microbiology Reviews 39 465ndash87

Minot S et al (2011) lsquoThe Human Gut Virome Inter-IndividualVariation and Dynamic Response to Dietrsquo Genome Research 211616ndash25

Moebus K and Nattkemper H (1981) lsquoBacteriophage SensitivityPatterns Among Bacteria Isolated from Marine WatersrsquoHelgolander Meeresuntersuchungen 34 375ndash85

Mojica F J M et al (2005) lsquoIntervening Sequences of RegularlySpaced Prokaryotic Repeats Derive from Foreign GeneticElementsrsquo Journal of Molecular Evolution 60 174ndash82

Mukherjee S et al (2017) lsquoGenomes OnLine Database (GOLD) v6Data Updates and Feature Enhancementsrsquo Nucleic AcidsResearch 45 D446ndash56

Murray N E (2002) lsquo2001 Fred Griffith Review LectureImmigration Control of DNA in Bacteria Self Versus Non-SelfrsquoMicrobiology (Reading Engl) 148 3ndash20

Nelson M I et al (2008) lsquoMultiple Reassortment Events in theEvolutionary History of H1N1 Influenza A Virus Since 1918rsquoPLoS Pathogens 4 e1000012

Pride D T et al (2012) lsquoEvidence of a Robust ResidentBacteriophage Population Revealed Through Analysis of theHuman Salivary Viromersquo The ISME Journal 6 915ndash26

R Development Core Team (2011) R A language and environmentfor statistical computing R Foundation for Statistical ComputingVienna Austria ISBN 3-900051-07-0 httpwwwR-projectorg

Rakonjac J et al (2011) lsquoFilamentous Bacteriophage BiologyPhage Display and Nanotechnology Applicationsrsquo CurrentIssues in Molecular Biology 13 51ndash76

Refardt D (2011) lsquoWithin-Host Competition DeterminesReproductive Success of Temperate Bacteriophagesrsquo The ISMEJournal 5 1451ndash60

Reyes A et al (2010) lsquoViruses in the Faecal Microbiota ofMonozygotic Twins and Their Mothersrsquo Nature 466 334ndash8

Rohwer F and Barott K (2012) lsquoViral Informationrsquo Biology andPhilosophy 28 283ndash97

Roux S et al (2012) lsquoEvolution and Diversity of the MicroviridaeViral Family Through A Collection of 81 New CompleteGenomes Assembled from Virome Readsrsquo PLoS One 7 e40418

et al (2013) lsquoChimeric Viruses Blur the Borders Between theMajor Groups of Eukaryotic Single-Stranded DNA VirusesrsquoNature Communications 4 2700

et al (2014) lsquoEcology and Evolution of Viruses InfectingUncultivated SUP05 Bacteria as Revealed by Single-Cell- andMeta-Genomicsrsquo eLife 3 e03125

et al (2015) lsquoViral Dark Matter and Virus-Host InteractionsResolved from Publicly Available Microbial Genomesrsquo eLife 4e08490

et al (2016) lsquoTowards Quantitative Viromics for BothDouble-Stranded and Single-Stranded DNA Virusesrsquo PeerJ 4e2777

Seed K D et al (2013) lsquoA Bacteriophage Encodes Its OwnCRISPRCas Adaptive Response to Evade Host InnateImmunityrsquo Nature 494 489ndash91

Semenova E et al (2011) lsquoInterference by Clustered RegularlyInterspaced Short Palindromic Repeat (CRISPR) RNA is

Governed by a Seed Sequencersquo Proceedings of theNational Academy of Sciences of the United States of America 10810098ndash103

Shah V Chang B X and Morris R M (2017) lsquoCultivation of aChemoautotroph from the SUP05 Clade of Marine Bacteria thatProduces Nitrite and Consumes Ammoniumrsquo The ISME Journal11 263ndash71

Spencer R (1957) lsquoA Possible Example of Geographical Variationin Bacteriophage Sensitivityrsquo Journal of General Microbiology 17xindashxii

Suttle C A (2007) lsquoMarine VirusesndashMajor Players in the GlobalEcosystemrsquo Nature Reviews Microbiology 5 801ndash12

Szekely A J and Breitbart M (2016) lsquoSingle-stranded DNAphages from early molecular biology tools to recent revolu-tions in environmental microbiologyrsquo FEMS Microbiol Lett363(6)fnw027 doi 101093femslefnw027

Trifonov V Khiabanian H and Rabadan R (2009)lsquoGeographic Dependence Surveillance and Origins of the 2009Influenza A (H1N1) Virusrsquo New England Journal of Medicine 361115ndash9

Turner P and Chao L (1998) lsquoSex and the Evolution ofIntrahost Competition in RNA Virus phi 6rsquo Genetics 150523ndash32

Turner P et al (1999) lsquoHybrid Frequencies Confirm Limit toCoinfection in the RNA Bacteriophage phi 6rsquo Journal of Virology73 2420ndash4

Turner P E and Duffy S (2008) lsquoEvolutionary ecology of multi-ple phage adsorption and infectionrsquo In S Abedon (Ed)Bacteriophage Ecology Population Growth Evolution and Impact ofBacterial Viruses (Advances in Molecular and CellularMicrobiology pp 195ndash216) Cambridge Cambridge UniversityPress doi101017CBO9780511541483011

Tyson G W and Banfield J F (2008) lsquoRapidly Evolving CRISPRsImplicated in Acquired Resistance of Microorganisms toVirusesrsquo Environmental Microbiology 10 200ndash7

Vignuzzi M et al (2006) lsquoQuasispecies Diversity DeterminesPathogenesis Through Cooperative Interactions in a ViralPopulationrsquo Nature 439 344ndash8

Warton D I and Hui F K C (2011) lsquoThe Arcsine isAsinine The Analysis of Proportions in Ecologyrsquo Ecology92 3ndash10

Weinbauer M G (2004) lsquoEcology of Prokaryotic Virusesrsquo FEMSMicrobiology Reviews 28 127ndash81

Westra E R et al (2010) lsquoH-NS-Mediated Repression of CRISPR-Based Immunity in Escherichia coli K12 Can Be Relieved by theTranscription Activator LeuOrsquo Molecular Microbiology 771380ndash93

Wickham H (2009) ggplot2 Elegant Graphics for Data AnalysisNew York Springer-Verlag

Wigington C H et al (2016) lsquoRe-Examination of the RelationshipBetween Marine Virus and Microbial Cell Abundancesrsquo NatureMicrobiology 1 15024

Worobey M and Holmes E C (1999) lsquoEvolutionary aspects ofrecombination in RNA Virusesrsquo Journal of General Virology 80Pt10 2535ndash43

Wylie K M et al (2014) lsquoMetagenomic Analysis ofDouble-Stranded DNA Viruses in Healthy Adultsrsquo BMC Biology12 71


information regarding the ecological and biological factors thatshape coinfection and virusndashvirus interactions Given that mostlaboratory studies of viruses focus on a single virus at a time(DaPalma et al 2010) understanding the drivers of coinfectionand virusndashvirus interactions is a pressing frontier for viralecology

Recent studies of bacteriophages have highlighted just howwidespread coinfection is and have started to shed light on theecology of viral coinfection For example although studies ofphage host range have shown that some bacterial hosts can beinfected by more than one type of phage for some time (Spencer1957 Kaiser and Dworkin 1975 Moebus and Nattkemper 1981)mounting evidence indicates this is a pervasive phenomenon(Flores et al 2011 2013 Koskella and Meaden 2013) The abilityof two or more viruses to independently infect the same hostmdashcross-infectivitymdashrevealed in these studies of host range is anecessary but not sufficient criterion to determine coinfectionThus cross-infectivity or the potential for coinfection repre-sents a baseline shaping coinfection patterns To confirm coin-fection sequence-based approaches have been used to detectviral coinfection at two scalesgt1 virus infecting the same mul-ticellular host (eg a multicellular eukaryote or a colony of bac-terial cells hereafter called culture coinfection) or to thecoinfection of a single cell At the culture coinfection scale astudy mining sequence data to examine virusndashhost relation-ships uncovered widespread coinfection in publicly availablebacterial and archaeal genome sequence data (Roux et al 2015)At the single-cell scale two studies have provided for the firsttime single-cell level information on viruses associated withspecific hosts isolated from the environment in a culture-independent manner (Roux et al 2014 Labonte et al 2015)Collectively these studies suggest that there is a large potentialfor coinfection and that this potential is realized at both thehost culture and single cell scales A summary of these studiessuggests roughly half of hosts have the potential to be infected(ie are cross-infective) or are actually infected by an averageofgt2 viruses (Table 1) Thus there is extensive evidence acrossvarious methodologies taxa environments that coinfection iswidespread and virusndashvirus interactions may be a frequentoccurrence

The aforementioned studies among others establish thatviral coinfection is a frequent occurrence in bacterial and arch-aeal hosts but systematic tests of the factors explaining varia-tion in viral coinfection across different taxa and environmentsare lacking However the literature suggests four factors thatare likely to play a role host ecology host taxonomy or phylog-eny host defense mechanisms and virusndashvirus interactionsThe relevance and importance of these are likely to varyfor cross-infectivity culture coinfection and single-cellcoinfection

Cross-infectivity has been examined in studies of phagehost-range which have provided some insight into the factorsaffecting potential coinfection In a single bacterial speciesthere can be wide variation in phage host range (Holmfeldt

et al 2007) and thus cross-infectivity A larger quantitativestudy of phage-bacteria infection networks in multiple taxaalso found wide variation in cross-infectivity in narrow taxo-nomic ranges (strain or species level) with some hosts suscepti-ble to few viruses and others to many (Flores et al 2011)However at broader host taxonomic scales cross-infectivity fol-lowed a modular pattern suggesting that higher taxonomicranks could influence cross-infectivity (Flores et al 2013) Thushost taxonomy may show a scale dependent effect in shapingcoinfection patterns Flores et al (2013) also highlighted thepotential role of ecology in shaping coinfection patterns as geo-graphic separation played a role in cross-infectivity Thus bac-terial ecology and phylogeny (particularly at taxonomic ranksabove species) are the primary candidates for drivers ofcoinfection

Culture and single cell coinfection require cross-infectivitybut also require both the bacteria and infecting viruses to allowsimultaneous or sequential infection Thus in addition to bac-terial phylogeny and ecology we can expect additional factorsshaping coinfection namely bacterial defense mechanisms andvirusndashvirus interactions An extensive collection of studies pro-vides evidence that bacterial and viral mechanisms may affectcoinfection Bacteria understandably reluctant to welcomeviruses possess a collection of mechanisms of defense againstviral infection (Labrie et al 2010) including restriction enzymes(Linn and Arber 1968 Murray 2002) and CRISPR-Cas systems(Horvath and Barrangou 2010) The latter have been shown to bean adaptive immune system for bacteria protecting from futureinfection by the same phage (Barrangou et al 2007) and preserv-ing the memory of viral infections past (Held and Whitaker2009) Metagenomic studies of CRISPR in natural environmentssuggest rapid coevolution of CRISPR arrays (Tyson and Banfield2008) but little is known regarding the in situ protective effectsof CRISPR on cells which should now be possible to decipherwith single-cell genomics

Viruses also have mechanisms to mediate infection by otherviruses some of which were identified in early lab studies ofbacteriophages (Ellis and Delbruck 1939 Delbruck 1946) Anexample of a well-described phenomenon of virusndashvirus inter-actions is superinfection immunity conferred by lysogenicviruses (Bertani 1953) which can inhibit coinfection of culturesand single cells (Bertani 1954) While this mechanism has beendescribed in several species its frequency at broader taxonomicscales and its occurrence in natural settings is not well knownMost attention in virusndashvirus interactions has focused on mech-anisms limiting coinfection with the assumption that coinfec-tion invariably reduces host fitness (Berngruber et al 2010)However some patterns of nonrandom coinfection suggest ele-vated coinfection (Turner et al 1999 Dang et al 2004 Cicin-Sainet al 2005) and specific viral mechanisms that promote co-infection have been identified including enhanced expressionof the phage attachment site upon infection (Joseph et al 2009)particle aggregation (Altan-Bonnet and Chen 2015 Aguileraet al 2017) and phagendashphage communication (Erez et al 2017)

Table 1 Viral coinfection is prevalent across various methodologies taxa environments and scales of coinfection

Cross-infectivity (potential coinfection) Culture coinfection Single-cell coinfection

Number of viruses 489 (6461) 337761804 2376083Prop of bacteria wmultiple infections 0654 0538 0450Reference (Flores et al 2011) (Roux et al 2014 2015) (Roux et al 2014)


























































marineplant soil





yogurt industrial


Isol

atio

n H

abita

t

bovine pathogen

fish pathogen

free

human pathogen


human symbiont

mycorrhizal

plant symbiont


Bac

teria

l Ass

ocia

tion








RhizobiumSalmonella



Vibrio


Bac

teria

l Tax

aBA

C D



Bacteria_new

Residuals

Df

7

22

5

970

Sum Sq

106497

308534

4259

809121

Mean Sq

15214

14024

08518

00834

F value

182389

168127

102115

NA

Pr(gtF)

0

0

0

NA













(Intercept)


GenBacteroides

GenDickeya

GenEdwardsiella

GenEnterobacter

GenEnterococcus

GenHelicobacter

GenListeria

GenNeisseria

GenPantoea

GenSerratia

GenStaphylococcus

GenStenotrophomonas

GenStreptococcus

GenVibrio

GenXanthomonas

prophage

Coef

082

023

095

238

217

142

143

149

099

084

149

146

095

148

087

086

085

023

Std_Error

037

01

045

081

109

054

039

048

047

039

064

06

038

065

038

038

04

002

zvalue

22

228

211

293

198

262

363

311

212

215

233

243

25

229

229

225

214

1387

p

003

002

004

0

005

001

0

0

003

003

002

002

001

002

002

002

003

0

Exp_Coef

227

08

039

009

011

024

024

022

037

043

022

023

039

023

042

042

043



EnterobacterPantoea




CollinsellaDorea



ProteusRoseburia

SodalisRalstonia

HalorubrumListeria

















GordoniaLeuconostoc




SalinisporaSUP05






SphingomonasXylella

ClostridiumFrankia




AliivibrioDialister










unknownBrucella









SelenomonasShigella

Avibacterium


Gen

us

Engineered

Environmental

Host associated


Eco

syst

em

0123456789

10111213


Ext

rach

rom

osom

al

viru

ses

infe

ctin

g

BA

C

D























0

1

2

3

4

5


Num

ber

of v

iruse

s ac

tivel

y in

fect

ing

cell

0

1

2

3

4

5


Num

ber

of v

iruse

s ac

tivel

y in

fect

ing

cell

0

1

2

3

4

5

0 1CRISPR spacers

Num

ber

of v

iruse

s ac

tivel

y in

fect

ing

cell

(Intercept)

Depth

CRISPR

Defectiveprophage

Exp_Coef

0111

1016

0463

009

Robust SE

0091

0005

013

0085

LL

0022

1006

0267

0014

UL

0553

1026

0803

0571

p

0007

0001

0006

0011

DC

A B











sented in panel D







00

01

02

03

04

05


Pro

port

ion

prop

hage

infe

cted

hos

ts

00

01

02

03

04

05

06

07

Mixed Prophage Only

Pro

port

ion

prop

hage

infe

cted

hos

ts

2

3

4

5

6

7

8

9

10

11


Num

ber

of v

iruse

s in

eac

h ho

st c

ultu

re

B

C

A





















ssDNA dsDNA

ssDNA only



Com

posi

tion

of C

oinf

ectio

ns

dsDNA mixed

dsDNA only

ssDNA only

ssDNA mixed


Com

posi

tion

of C

oinf

ectio

ns

BA
























Supplementary data


Acknowledgements


Data availability


















































































































































marineplant soil





yogurt industrial


Isol

atio

n H

abita

t

bovine pathogen

fish pathogen

free

human pathogen


human symbiont

mycorrhizal

plant symbiont


Bac

teria

l Ass

ocia

tion








RhizobiumSalmonella



Vibrio


Bac

teria

l Tax

aBA

C D



Bacteria_new

Residuals

Df

7

22

5

970

Sum Sq

106497

308534

4259

809121

Mean Sq

15214

14024

08518

00834

F value

182389

168127

102115

NA

Pr(gtF)

0

0

0

NA













(Intercept)


GenBacteroides

GenDickeya

GenEdwardsiella

GenEnterobacter

GenEnterococcus

GenHelicobacter

GenListeria

GenNeisseria

GenPantoea

GenSerratia

GenStaphylococcus

GenStenotrophomonas

GenStreptococcus

GenVibrio

GenXanthomonas

prophage

Coef

082

023

095

238

217

142

143

149

099

084

149

146

095

148

087

086

085

023

Std_Error

037

01

045

081

109

054

039

048

047

039

064

06

038

065

038

038

04

002

zvalue

22

228

211

293

198

262

363

311

212

215

233

243

25

229

229

225

214

1387

p

003

002

004

0

005

001

0

0

003

003

002

002

001

002

002

002

003

0

Exp_Coef

227

08

039

009

011

024

024

022

037

043

022

023

039

023

042

042

043



EnterobacterPantoea




CollinsellaDorea



ProteusRoseburia

SodalisRalstonia

HalorubrumListeria

















GordoniaLeuconostoc




SalinisporaSUP05






SphingomonasXylella

ClostridiumFrankia




AliivibrioDialister










unknownBrucella









SelenomonasShigella

Avibacterium


Gen

us

Engineered

Environmental

Host associated


Eco

syst

em

0123456789

10111213


Ext

rach

rom

osom

al

viru

ses

infe

ctin

g

BA

C

D























0

1

2

3

4

5


Num

ber

of v

iruse

s ac

tivel

y in

fect

ing

cell

0

1

2

3

4

5


Num

ber

of v

iruse

s ac

tivel

y in

fect

ing

cell

0

1

2

3

4

5

0 1CRISPR spacers

Num

ber

of v

iruse

s ac

tivel

y in

fect

ing

cell

(Intercept)

Depth

CRISPR

Defectiveprophage

Exp_Coef

0111

1016

0463

009

Robust SE

0091

0005

013

0085

LL

0022

1006

0267

0014

UL

0553

1026

0803

0571

p

0007

0001

0006

0011

DC

A B











sented in panel D







00

01

02

03

04

05


Pro

port

ion

prop

hage

infe

cted

hos

ts

00

01

02

03

04

05

06

07

Mixed Prophage Only

Pro

port

ion

prop

hage

infe

cted

hos

ts

2

3

4

5

6

7

8

9

10

11


Num

ber

of v

iruse

s in

eac

h ho

st c

ultu

re

B

C

A





















ssDNA dsDNA

ssDNA only



Com

posi

tion

of C

oinf

ectio

ns

dsDNA mixed

dsDNA only

ssDNA only

ssDNA mixed


Com

posi

tion

of C

oinf

ectio

ns

BA
























Supplementary data


Acknowledgements


Data availability
































































































































marineplant soil





yogurt industrial


Isol

atio

n H

abita

t

bovine pathogen

fish pathogen

free

human pathogen


human symbiont

mycorrhizal

plant symbiont


Bac

teria

l Ass

ocia

tion








RhizobiumSalmonella



Vibrio


Bac

teria

l Tax

aBA

C D



Bacteria_new

Residuals

Df

7

22

5

970

Sum Sq

106497

308534

4259

809121

Mean Sq

15214

14024

08518

00834

F value

182389

168127

102115

NA

Pr(gtF)

0

0

0

NA













(Intercept)


GenBacteroides

GenDickeya

GenEdwardsiella

GenEnterobacter

GenEnterococcus

GenHelicobacter

GenListeria

GenNeisseria

GenPantoea

GenSerratia

GenStaphylococcus

GenStenotrophomonas

GenStreptococcus

GenVibrio

GenXanthomonas

prophage

Coef

082

023

095

238

217

142

143

149

099

084

149

146

095

148

087

086

085

023

Std_Error

037

01

045

081

109

054

039

048

047

039

064

06

038

065

038

038

04

002

zvalue

22

228

211

293

198

262

363

311

212

215

233

243

25

229

229

225

214

1387

p

003

002

004

0

005

001

0

0

003

003

002

002

001

002

002

002

003

0

Exp_Coef

227

08

039

009

011

024

024

022

037

043

022

023

039

023

042

042

043



EnterobacterPantoea




CollinsellaDorea



ProteusRoseburia

SodalisRalstonia

HalorubrumListeria

















GordoniaLeuconostoc




SalinisporaSUP05






SphingomonasXylella

ClostridiumFrankia




AliivibrioDialister










unknownBrucella









SelenomonasShigella

Avibacterium


Gen

us

Engineered

Environmental

Host associated


Eco

syst

em

0123456789

10111213


Ext

rach

rom

osom

al

viru

ses

infe

ctin

g

BA

C

D























0

1

2

3

4

5


Num

ber

of v

iruse

s ac

tivel

y in

fect

ing

cell

0

1

2

3

4

5


Num

ber

of v

iruse

s ac

tivel

y in

fect

ing

cell

0

1

2

3

4

5

0 1CRISPR spacers

Num

ber

of v

iruse

s ac

tivel

y in

fect

ing

cell

(Intercept)

Depth

CRISPR

Defectiveprophage

Exp_Coef

0111

1016

0463

009

Robust SE

0091

0005

013

0085

LL

0022

1006

0267

0014

UL

0553

1026

0803

0571

p

0007

0001

0006

0011

DC

A B











sented in panel D







00

01

02

03

04

05


Pro

port

ion

prop

hage

infe

cted

hos

ts

00

01

02

03

04

05

06

07

Mixed Prophage Only

Pro

port

ion

prop

hage

infe

cted

hos

ts

2

3

4

5

6

7

8

9

10

11


Num

ber

of v

iruse

s in

eac

h ho

st c

ultu

re

B

C

A





















ssDNA dsDNA

ssDNA only



Com

posi

tion

of C

oinf

ectio

ns

dsDNA mixed

dsDNA only

ssDNA only

ssDNA mixed


Com

posi

tion

of C

oinf

ectio

ns

BA
























Supplementary data


Acknowledgements


Data availability



















































































































marineplant soil





yogurt industrial


Isol

atio

n H

abita

t

bovine pathogen

fish pathogen

free

human pathogen


human symbiont

mycorrhizal

plant symbiont


Bac

teria

l Ass

ocia

tion








RhizobiumSalmonella



Vibrio


Bac

teria

l Tax

aBA

C D



Bacteria_new

Residuals

Df

7

22

5

970

Sum Sq

106497

308534

4259

809121

Mean Sq

15214

14024

08518

00834

F value

182389

168127

102115

NA

Pr(gtF)

0

0

0

NA













(Intercept)


GenBacteroides

GenDickeya

GenEdwardsiella

GenEnterobacter

GenEnterococcus

GenHelicobacter

GenListeria

GenNeisseria

GenPantoea

GenSerratia

GenStaphylococcus

GenStenotrophomonas

GenStreptococcus

GenVibrio

GenXanthomonas

prophage

Coef

082

023

095

238

217

142

143

149

099

084

149

146

095

148

087

086

085

023

Std_Error

037

01

045

081

109

054

039

048

047

039

064

06

038

065

038

038

04

002

zvalue

22

228

211

293

198

262

363

311

212

215

233

243

25

229

229

225

214

1387

p

003

002

004

0

005

001

0

0

003

003

002

002

001

002

002

002

003

0

Exp_Coef

227

08

039

009

011

024

024

022

037

043

022

023

039

023

042

042

043



EnterobacterPantoea




CollinsellaDorea



ProteusRoseburia

SodalisRalstonia

HalorubrumListeria

















GordoniaLeuconostoc




SalinisporaSUP05






SphingomonasXylella

ClostridiumFrankia




AliivibrioDialister










unknownBrucella









SelenomonasShigella

Avibacterium


Gen

us

Engineered

Environmental

Host associated


Eco

syst

em

0123456789

10111213


Ext

rach

rom

osom

al

viru

ses

infe

ctin

g

BA

C

D























0

1

2

3

4

5


Num

ber

of v

iruse

s ac

tivel

y in

fect

ing

cell

0

1

2

3

4

5


Num

ber

of v

iruse

s ac

tivel

y in

fect

ing

cell

0

1

2

3

4

5

0 1CRISPR spacers

Num

ber

of v

iruse

s ac

tivel

y in

fect

ing

cell

(Intercept)

Depth

CRISPR

Defectiveprophage

Exp_Coef

0111

1016

0463

009

Robust SE

0091

0005

013

0085

LL

0022

1006

0267

0014

UL

0553

1026

0803

0571

p

0007

0001

0006

0011

DC

A B











sented in panel D







00

01

02

03

04

05


Pro

port

ion

prop

hage

infe

cted

hos

ts

00

01

02

03

04

05

06

07

Mixed Prophage Only

Pro

port

ion

prop

hage

infe

cted

hos

ts

2

3

4

5

6

7

8

9

10

11


Num

ber

of v

iruse

s in

eac

h ho

st c

ultu

re

B

C

A





















ssDNA dsDNA

ssDNA only



Com

posi

tion

of C

oinf

ectio

ns

dsDNA mixed

dsDNA only

ssDNA only

ssDNA mixed


Com

posi

tion

of C

oinf

ectio

ns

BA
























Supplementary data


Acknowledgements


Data availability







































































































marineplant soil





yogurt industrial


Isol

atio

n H

abita

t

bovine pathogen

fish pathogen

free

human pathogen


human symbiont

mycorrhizal

plant symbiont


Bac

teria

l Ass

ocia

tion








RhizobiumSalmonella



Vibrio


Bac

teria

l Tax

aBA

C D



Bacteria_new

Residuals

Df

7

22

5

970

Sum Sq

106497

308534

4259

809121

Mean Sq

15214

14024

08518

00834

F value

182389

168127

102115

NA

Pr(gtF)

0

0

0

NA













(Intercept)


GenBacteroides

GenDickeya

GenEdwardsiella

GenEnterobacter

GenEnterococcus

GenHelicobacter

GenListeria

GenNeisseria

GenPantoea

GenSerratia

GenStaphylococcus

GenStenotrophomonas

GenStreptococcus

GenVibrio

GenXanthomonas

prophage

Coef

082

023

095

238

217

142

143

149

099

084

149

146

095

148

087

086

085

023

Std_Error

037

01

045

081

109

054

039

048

047

039

064

06

038

065

038

038

04

002

zvalue

22

228

211

293

198

262

363

311

212

215

233

243

25

229

229

225

214

1387

p

003

002

004

0

005

001

0

0

003

003

002

002

001

002

002

002

003

0

Exp_Coef

227

08

039

009

011

024

024

022

037

043

022

023

039

023

042

042

043



EnterobacterPantoea




CollinsellaDorea



ProteusRoseburia

SodalisRalstonia

HalorubrumListeria

















GordoniaLeuconostoc




SalinisporaSUP05






SphingomonasXylella

ClostridiumFrankia




AliivibrioDialister










unknownBrucella









SelenomonasShigella

Avibacterium


Gen

us

Engineered

Environmental

Host associated


Eco

syst

em

0123456789

10111213


Ext

rach

rom

osom

al

viru

ses

infe

ctin

g

BA

C

D























0

1

2

3

4

5


Num

ber

of v

iruse

s ac

tivel

y in

fect

ing

cell

0

1

2

3

4

5


Num

ber

of v

iruse

s ac

tivel

y in

fect

ing

cell

0

1

2

3

4

5

0 1CRISPR spacers

Num

ber

of v

iruse

s ac

tivel

y in

fect

ing

cell

(Intercept)

Depth

CRISPR

Defectiveprophage

Exp_Coef

0111

1016

0463

009

Robust SE

0091

0005

013

0085

LL

0022

1006

0267

0014

UL

0553

1026

0803

0571

p

0007

0001

0006

0011

DC

A B











sented in panel D







00

01

02

03

04

05


Pro

port

ion

prop

hage

infe

cted

hos

ts

00

01

02

03

04

05

06

07

Mixed Prophage Only

Pro

port

ion

prop

hage

infe

cted

hos

ts

2

3

4

5

6

7

8

9

10

11


Num

ber

of v

iruse

s in

eac

h ho

st c

ultu

re

B

C

A





















ssDNA dsDNA

ssDNA only



Com

posi

tion

of C

oinf

ectio

ns

dsDNA mixed

dsDNA only

ssDNA only

ssDNA mixed


Com

posi

tion

of C

oinf

ectio

ns

BA
























Supplementary data


Acknowledgements


Data availability




























































































(Intercept)


GenBacteroides

GenDickeya

GenEdwardsiella

GenEnterobacter

GenEnterococcus

GenHelicobacter

GenListeria

GenNeisseria

GenPantoea

GenSerratia

GenStaphylococcus

GenStenotrophomonas

GenStreptococcus

GenVibrio

GenXanthomonas

prophage

Coef

082

023

095

238

217

142

143

149

099

084

149

146

095

148

087

086

085

023

Std_Error

037

01

045

081

109

054

039

048

047

039

064

06

038

065

038

038

04

002

zvalue

22

228

211

293

198

262

363

311

212

215

233

243

25

229

229

225

214

1387

p

003

002

004

0

005

001

0

0

003

003

002

002

001

002

002

002

003

0

Exp_Coef

227

08

039

009

011

024

024

022

037

043

022

023

039

023

042

042

043



EnterobacterPantoea




CollinsellaDorea



ProteusRoseburia

SodalisRalstonia

HalorubrumListeria

















GordoniaLeuconostoc




SalinisporaSUP05






SphingomonasXylella

ClostridiumFrankia




AliivibrioDialister










unknownBrucella









SelenomonasShigella

Avibacterium


Gen

us

Engineered

Environmental

Host associated


Eco

syst

em

0123456789

10111213


Ext

rach

rom

osom

al

viru

ses

infe

ctin

g

BA

C

D























0

1

2

3

4

5


Num

ber

of v

iruse

s ac

tivel

y in

fect

ing

cell

0

1

2

3

4

5


Num

ber

of v

iruse

s ac

tivel

y in

fect

ing

cell

0

1

2

3

4

5

0 1CRISPR spacers

Num

ber

of v

iruse

s ac

tivel

y in

fect

ing

cell

(Intercept)

Depth

CRISPR

Defectiveprophage

Exp_Coef

0111

1016

0463

009

Robust SE

0091

0005

013

0085

LL

0022

1006

0267

0014

UL

0553

1026

0803

0571

p

0007

0001

0006

0011

DC

A B











sented in panel D







00

01

02

03

04

05


Pro

port

ion

prop

hage

infe

cted

hos

ts

00

01

02

03

04

05

06

07

Mixed Prophage Only

Pro

port

ion

prop

hage

infe

cted

hos

ts

2

3

4

5

6

7

8

9

10

11


Num

ber

of v

iruse

s in

eac

h ho

st c

ultu

re

B

C

A





















ssDNA dsDNA

ssDNA only



Com

posi

tion

of C

oinf

ectio

ns

dsDNA mixed

dsDNA only

ssDNA only

ssDNA mixed


Com

posi

tion

of C

oinf

ectio

ns

BA
























Supplementary data


Acknowledgements


Data availability

































































































0

1

2

3

4

5


Num

ber

of v

iruse

s ac

tivel

y in

fect

ing

cell

0

1

2

3

4

5


Num

ber

of v

iruse

s ac

tivel

y in

fect

ing

cell

0

1

2

3

4

5

0 1CRISPR spacers

Num

ber

of v

iruse

s ac

tivel

y in

fect

ing

cell

(Intercept)

Depth

CRISPR

Defectiveprophage

Exp_Coef

0111

1016

0463

009

Robust SE

0091

0005

013

0085

LL

0022

1006

0267

0014

UL

0553

1026

0803

0571

p

0007

0001

0006

0011

DC

A B











sented in panel D







00

01

02

03

04

05


Pro

port

ion

prop

hage

infe

cted

hos

ts

00

01

02

03

04

05

06

07

Mixed Prophage Only

Pro

port

ion

prop

hage

infe

cted

hos

ts

2

3

4

5

6

7

8

9

10

11


Num

ber

of v

iruse

s in

eac

h ho

st c

ultu

re

B

C

A





















ssDNA dsDNA

ssDNA only



Com

posi

tion

of C

oinf

ectio

ns

dsDNA mixed

dsDNA only

ssDNA only

ssDNA mixed


Com

posi

tion

of C

oinf

ectio

ns

BA
























Supplementary data


Acknowledgements


Data availability































































































00

01

02

03

04

05


Pro

port

ion

prop

hage

infe

cted

hos

ts

00

01

02

03

04

05

06

07

Mixed Prophage Only

Pro

port

ion

prop

hage

infe

cted

hos

ts

2

3

4

5

6

7

8

9

10

11


Num

ber

of v

iruse

s in

eac

h ho

st c

ultu

re

B

C

A





















ssDNA dsDNA

ssDNA only



Com

posi

tion

of C

oinf

ectio

ns

dsDNA mixed

dsDNA only

ssDNA only

ssDNA mixed


Com

posi

tion

of C

oinf

ectio

ns

BA
























Supplementary data


Acknowledgements


Data availability
































































































ssDNA dsDNA

ssDNA only



Com

posi

tion

of C

oinf

ectio

ns

dsDNA mixed

dsDNA only

ssDNA only

ssDNA mixed


Com

posi

tion

of C

oinf

ectio

ns

BA
























Supplementary data


Acknowledgements


Data availability







































































































Supplementary data


Acknowledgements


Data availability
































































































Supplementary data


Acknowledgements


Data availability





















































































































































































































Documents

Viral coinfection is shaped by host ecology and virus– virus … · 2017. 5. 4. · Viral coinfection is shaped by host ecology and virus– virus interactions across diverse microbial