The Core and Seasonal Microbiota of Raw Bovine Milk in TankerTrucks and the Impact of Transfer to a Milk Processing Facility

Mary E Kablea Yanin Srisengfaa Miles Lairda Jose Zaragozaa Jeremy McLeodb Jessie Heidenreichb Maria L Marcoa

Department of Food Science and Technology University of California Davis Davis California USAa Hilmar Cheese Company Hilmar California USAb

ABSTRACT Currently the bacterial composition of raw milk in tanker trucks and the outcomes of transfer and storage of thatmilk at commercial processing facilities are not well understood We set out to identify the bacteria in raw milk collected forlarge-scale dairy product manufacturing Raw bovine milk samples from 899 tanker trucks arriving at two dairy processors inSan Joaquin Valley of California during three seasons (spring summer and fall) were analyzed by community 16S rRNA genesequencing This analysis revealed highly diverse bacterial populations which exhibited seasonal differences Raw milk collectedin the spring contained the most diverse bacterial communities with the highest total cell numbers and highest proportions be-ing those of Actinobacteria Even with this complexity a core microbiota was present consisting of 29 taxonomic groups andhigh proportions of Streptococcus and Staphylococcus and unidentified members of Clostridiales Milk samples were also col-lected from five large-volume silos and from 13 to 25 tankers whose contents were unloaded into each of them during 2 days inthe summer Transfer of the milk to storage silos resulted in two community types One group of silos contained a high propor-tion of Streptococcus spp and was similar in that respect to the tankers that filled them The community found in the other groupof silos was distinct and dominated by Acinetobacter Overall despite highly diverse tanker milk community structures distinctmilk bacterial communities were selected within the processing facility environment This knowledge can inform the develop-ment of new sanitation procedures and process controls to ensure the consistent production of safe and high-quality dairy prod-ucts on a global scale

IMPORTANCE Raw milk harbors diverse bacteria that are crucial determinants of the quality and safety of fluid milk and (fer-mented) dairy products These bacteria enter farm milk during transport storage and processing Although pathogens are de-stroyed by pasteurization not all bacteria and their associated enzymes are eliminated Our comprehensive analyses of the bac-terial composition of raw milk upon arrival and shortly after storage at major dairy processors showed that the communities ofmilk microbiota are highly diverse Even with these differences there was a core microbiota that exhibited distinct seasonaltrends Remarkably the effects of the processing facility outweighed those of the raw milk microbiome and the microbial com-position changed distinctly within some but not all silos within a short time after transfer This knowledge can be used to informcleaning and sanitation procedures as well as to enable predictions of the microbial communities in raw milk that result in eitherhigh-quality or defective products

Received 9 May 2016 Accepted 25 July 2016 Published 23 August 2016

Citation Kable ME Srisengfa Y Laird M Zaragoza J McLeod J Heidenreich J Marco ML 2016 The core and seasonal microbiota of raw bovine milk in tanker trucks and theimpact of transfer to a milk processing facility mBio 7(4)e00836-16 doi101128mBio00836-16

Editor Mark J Bailey CEH-Oxford

Copyright copy 2016 Kable et al This is an open-access article distributed under the terms of the Creative Commons Attribution 40 International license

Address correspondence to Maria L Marco mmarcoudavisedu

Bacteria are essential determinants of the shelf life organolepticqualities and safety of fresh and (minimally) processed foods

While the microbial ecology of fresh produce (1ndash4) and animalproducts (5ndash10) has been intensively studied the influence ofstorage transport and processing facilities on the microbial com-position and ecology of those products is less comprehensivelyunderstood (11ndash14)

Bovine milk and manufactured dairy products made from itare among the most frequently consumed foods at global scales(15) Only a small fraction of produced milk is consumed as abeverage with much higher quantities being either fermented tocreate other food products (eg cheese yogurt etc) or processedfor ingredients such as whey or lactose (16) Milk is safe to con-sume after pasteurization but is still susceptible to microbe-induced spoilage and quality defects In contrast some of the sur-

viving microorganisms in milk contributes beneficially to theorganoleptic qualities of fermented dairy products Lactic acidbacteria (LAB) are particularly important because of their positiveand negative impacts on fresh and fermented dairy foods (17ndash20)

A wide variety of bacterial species have been detected in rawand minimally processed milk (6 10) Dairy locations and milkingpractices including housing (indoor versus outdoor) and feedand bedding type alter the bacterial populations present on cowteats on dust and in air in the milking parlor and ultimatelycontribute to the raw milk microbiome (7 10 21ndash24) By com-parison the microbiota present in fresh milk after transport andstorage is not as well understood (25ndash27) Increases in standardplate counts (25) and slightly higher coliform counts (27) werefound upon transfer of raw milk from farm tanks to dairy proces-sor bulk tanks Although on-farm knowledge of raw milk micro-

RESEARCH ARTICLE

crossmark

JulyAugust 2016 Volume 7 Issue 4 e00836-16 reg mbioasmorg 1

on August 23 2018 by guest

httpmbioasm

orgD

ownloaded from

biota is important for identification of the sources of entry thebacterial composition of raw milk as it reaches the site of pasteur-ization and processing is directly relevant to the production ofhigh-quality dairy products with a long shelf life

California is the largest dairy producer in the United Statesproviding the largest amount of fluid milk (42337 million lb peryear 206 of total United States milk production) (28) and thesecond largest quantity of cheese (1899 million lb per year) sur-passed only by Wisconsin (2263 million lb per year) (29) How-ever the microbial community in raw milk samples in Californiajust prior to use for the manufacture of dairy products has not yetto our knowledge been comprehensively examined Given thatboth the location and transportation of raw milk can affect itsmicrobial contents we set out to identify the microbiota of rawmilk collected for large-scale product manufacturing in Califor-nia Specifically we measured the consistency of the milk micro-biota upon arrival at two processing facilities and determined howthat microbiota was affected by large-scale short-term storage atthe manufacturing facility

RESULTSBacterial populations in raw milk in tanker trucks are highlydiverse The bacterial diversity in raw bovine milk after bulktransport was determined for 899 tanker trucks upon arrival attwo different dairy processors in the California central valley Thiscollection included 229 tankers filled in the fall of 2013 and an-other 264 and 406 tankers filled in the spring and summer of 2014respectively The larger set of samples collected in the summerincluded milk collected from two sampling dates 1 week apart Foreach of those tanker truck milk samples we obtained at least

15000 16S rRNA gene reads by high-throughput DNA sequenc-ing that met quality-filtering requirements for further analysis

Phylogenetic and taxonomic assessments of the 16S rRNA V4regions revealed that bacterial populations in the raw milk in thetankers were highly diverse and variable In 354 (394) of the rawmilk samples analyzed taxa detected at less than 1 relative abun-dance accounted for 50 or more of the bacteria present Thevariation in the bacterial populations between the tankers wasexemplified by the finding that while some of the trucks containedhigh (30) proportions of bacteria of certain taxa such as Strep-tococcus Pseudomonas Staphylococcus and Mycoplasma thesesame taxa were detected at very low (1) levels in the milk of theother tankers tested (Fig 1)

Core microbiome of raw milk Despite these differences in theraw milk bacterial communities a core milk microbiome was stillpresent A total of 29 taxa were detected in 100 of the raw milksamples examined (Table 1) This core microbiome encompassedmembers of the Firmicutes Actinobacteria Bacteroidetes Proteo-bacteria and Tenericutes phyla The most abundant bacterial taxawere Streptococcus Staphylococcus and unidentified members oforder Clostridiales and comprised medians of 65 54 and 63 ofthe milk microbiota respectively (Table 1) Within the order Clos-tridiales several identifiable taxa were also present at relativelyhigh median relative abundances including species of the genusClostridium (15) and unidentified members of families Clostri-diaceae (13) and Lachnospiraceae (2) Moreover Corynebac-terium (37) Turicibacter (25) and Acinetobacter (12) werealso abundant Although Mycoplasma composed 75 of the totalcommunity in at least 1 milk sample its median relative abun-dance in all milk tested was relatively low (026) Notably Pseu-

FIG 1 Variations in the proportions of predominant bacteria taxa in raw milk delivered to two dairy production facilities in California Data represent relativeabundances of taxa found at 30 or greater relative abundance in at least one raw tanker milk sample Relative abundances were calculated after rarefaction ofthe OTU table to 15000 sequences per sample (f) family

Kable et al

2 reg mbioasmorg JulyAugust 2016 Volume 7 Issue 4 e00836-16

on August 23 2018 by guest

httpmbioasm

orgD

ownloaded from

domonas was not a part of the core microbiome Although it waspresent in relatively high proportions in some of the milk tested(Fig 1) Pseudomonas was entirely absent from two tankers andwas therefore not included in the core A complete list of all bac-terial taxa identified in the raw tanker milk samples is provided inTable S1 in the supplemental material

The microbiota in raw milk varies depending on the seasonBecause 16S rRNA gene sequencing can result in nonuniformsample coverage operational taxonomic unit (OTU) count nor-malization methods are necessary prior to any comparative anal-

yses (30 31) Therefore three methods for normalizing OTUcounts were employed cumulative sum scaling (CSS) CSS fol-lowed by batch correction and rarefaction at a depth of 15000sequences per sample Principal coordinate analysis (PCoA) of theweighted UniFrac distance (beta diversity) among the bacteria inthe tanker milk samples was then performed on the normalizeddata These comparisons showed that all three methods resulted inPCoA values with substantial overlap among all milk collected butalso with some clustering according to season (Fig 2 see alsoFig S1 in the supplemental material)

TABLE 1 Core milk microbiota

Phylum Class Order Family GenusRelative median abundance a

Firmicutes Bacilli Bacillales Staphylococcaceae Staphylococcus 545Salinicoccus 062Macrococcus 045

Bacillaceae Unidentified 068Bacillus 051

Planococcaceae Unidentified 109Lactobacillales Aerococcaceae Unidentified 097

Enterococcaceae Enterococcus 081Streptococcaceae Streptococcus 651

Turicibacterales Turicibacteraceae Turicibacter 245Clostridia Clostridiales Lachnospiraceae Unidentified 203

Butyrivibrio 079Dorea 067Coprococcus 036

Clostridiaceae Clostridium 147Unidentified 133

Ruminococcaceae Unidentified 435Ruminococcus 084

Peptostreptococcaceae Unidentified 222Unidentified Unidentified 633

Actinobacteria Actinobacteria Actinomycetales Micrococcaceae Unidentified 031Kocuria 025

Corynebacteriaceae Corynebacterium 370Yaniellaceae Yaniella 049

Bacteroidetes Bacteroidia Bacteroidales Unidentified Unidentified 086Bacteroidaceae 5-7N15 081

Proteobacteria Gammaproteobacteria Enterobacteriales Enterobacteriaceae Unidentified 040Pseudomonadales Moraxellaceae Acinetobacter 119

Tenericutes Mollicutes Mycoplasmatales Mycoplasmataceae Mycoplasma 026a All 899 raw milk samples from tanker trucks including milk tested from three seasons were used to determine the values

FIG 2 PCoA of the weighted UniFrac distances between bacterial communities in raw milk tankers Rarefaction at 15000 sequences per sample precededUniFrac analysis Milk samples are colored by (A) season (spring light green summer purple and fall orange) and (B) processor (A red and B blue)Dimension 2 (Dim2) and dimension 3 (Dim3) show changes in the overall community composition between seasons but not between processors

Raw Milk Microbiota during Transport and Storage

JulyAugust 2016 Volume 7 Issue 4 e00836-16 reg mbioasmorg 3

on August 23 2018 by guest

httpmbioasm

orgD

ownloaded from

According to Adonis a part of the vegan R package wrapped inQIIME the season when the milk was collected explained about5 of the variation in bacterial diversity between the raw milksamples (R2 004619 [rarefaction] 014313 [CSS] and 007911[CSS with batch correction] P 00001) These differences werelikely not due to different sampling dates alone because milk sam-ples collected on 2 different days within the summer season werehighly similar Although alpha diversity explained 9 of the bac-terial variation (R2 009199 [rarefaction] 005538 [CSS] and007341 [CSS with batch correction] P 00001) this variationcorresponded well to seasonal changes in the microbiota (Fig 3Asee also Fig S2 in the supplemental material) By comparisonsequencing depth exerted little or no influence on the variation inbacterial diversity between the samples (R2 001242 [rarefac-tion] 000642 [CSS] and 000873 [CSS with batch correction]P 00001) Similarly the dairy processor where the milk wasdelivered had very little impact on the microbial composition(R2 001052 [rarefaction] 00164 [CSS] and 001332 [CSS withbatch correction] P 00001) (Fig 2B see also Fig S1 in thesupplemental material)

The total estimated bacterial richness per sample differed be-tween seasons Raw milk collected in the spring contained thehighest median species richness according to the breakaway pack-age in R (see Fig S2 in the supplemental material) Similarly thehighest number of OTUs was observed in milk examined in thespring (Fig 3A) The lowest numbers of OTUs were detected inthe fall These differences also corresponded to the total estimatedcell numbers (Fig 3B) Quantitative PCR (qPCR) applied to theenumeration of bacteria in the raw milk indicated that all tankerscontained an average of 14 103 bacterial cellsml Althoughmilk sampled in the spring contained only modestly higher num-bers of cells (average of 16 103 bacterial cellsml) this differ-ence was significant in comparison to the levels observed in the fall(Fig 3B)

Even though bacterial species richness and median observednumbers of OTUs were highest in milk collected in the springthere were no OTUs that were uniquely present in those milksamples and absent in the other seasons Seven OTUs were absentfrom all milk collected in the spring season that were found in thesummer or fall By comparison only two OTUs were absent from

all milk sampled in the summer season and one OTU was notfound in milk collected in the fall Moreover milk transported indifferent tankers in the spring season contained bacterial commu-nities that were more similar to each other than was the case withthose found in milk samples from the other two seasons (Fig 3C)Therefore although milk examined in the spring contained a highnumber of OTUs the whole group of 264 milk samples in springhad fewer total unique OTUs than those collected in the other twoseasons

Consistent with the variations in alpha diversity the relativeabundances of individual bacterial taxa were dependent on theseason examined For example Firmicutes species were mostabundant overall but were present at significantly lower quantitiesin the spring than in the other seasons (Fig 4) In contrast theproportions of species of the Actinobacteria and Chloroflexi phylawere highest in spring (significant associations determined byboth MaAsLin and LefSe) and Bacteroidetes numbers were signif-icantly enriched in the fall (Fig 4) At deeper taxonomic levelsspecies of the genera Streptococcus and Staphylococcus were highestin number in the summer and the populations of unidentifiedmembers of the Lachnospiraceae family were increased in fall (sig-nificant differences determined by both MaAsLin and LefSe)(Fig 5)

Impact of large-volume storage on raw milk microbial com-munities To measure whether the raw milk microbiota was re-tained upon transfer of milk from the tanker truck to short-termlarge-volume storage we examined milk contained in five large-volume silos (silos A B C D and E) and the tankers that filledthem These samples were collected on 2 days 1 week apart in thesummer of 2014 Each silo was tested at least once per week exceptfor silo D which was tested only in the second week

The bacterial composition of raw milk in silos was distinctfrom that in the tankers accounting for 5 of the variation byPCoA of the weighted UniFrac distances (Adonis R2 005147[rarefied] and 00457 [CSS]) (see Fig S3 in the supplemental ma-terial) This difference was consistent with increased proportionsof several taxa in the silos At the order level Lactobacillales andPseudomonadales numbers were significantly enriched in silos rel-ative to tankers (Fig 6A) Streptococcaceae (order Lactobacillales)numbers were significantly enriched in silos (statistically signifi-

FIG 3 Seasonal differences in alpha (within-sample) and beta (between-sample) diversities of raw tanker milk microbial communities Significant differencesare indicated by the presence of different lowercase letters above each box plot (A) The number of observed OTUs at 15000 sequences per sample wassignificantly higher in spring than in summer or fall by the Kruskal-Wallis test (P value 22e-16) followed by Nemenyi test pairwise comparisons (Benjamini-Hochberg-corrected P values 0001 [spring versus fall] 0001 [spring versus summer] and not significant [ns fall versus summer]) (B) The numbers of cellsper milliliter estimated by qPCR for 365 milk samples total (134 from spring 97 from summer and 134 from fall) Total cell numbers differ between milk samplescollected in the fall and spring by the Kruskal-Wallis test (P value 0001929) followed by Nemenyi test pairwise comparisons (Benjamini-Hochberg-correctedP value 00012 [spring versus fall] ns [spring versus summer] and ns [summer versus fall]) (C) Weighted UniFrac distances among raw milk communitieswithin each season The distances between milk samples in spring were the lowest and in summer were the highest among the seasons tested (Kruskal-WallisP value 22e-16 all Nemenyi test pairwise comparison Benjamini-Hochberg-corrected P values were 0001)

Kable et al

4 reg mbioasmorg JulyAugust 2016 Volume 7 Issue 4 e00836-16

on August 23 2018 by guest

httpmbioasm

orgD

ownloaded from

cant by MaAsLin and LefSe) (Fig 6A) The numbers of organismsof the genera Lactococcus and Streptococcus within this familytended to be higher in silos than in tankers but genus-level differ-ences were not significant by both statistical methods used

Within the Pseudomonadales the genera Acinetobacter and Pseu-domonas also tended to be more abundant in silos than in tankersLastly numbers of Mycoplasma (order Mycoplasmatales) a mem-ber of the core milk community were also enriched in silos (sta-tistically significant by MaAsLin and LefSe) (Fig 6A) In contrastmembers of the order Clostridiales and genus Corynebacterium(order Actinomycetales) were detected in significantly lower pro-portions in silos than in tankers (statistically significant by MaAs-Lin and LefSe) (Fig 6A) These changes are notable because thecell density of bacteria in the silos was higher than in the tankersthat filled them (Fig 6B) suggesting that the increased relativeabundances of members of Lactobacillales and Pseudomonadalesorders were due to bacterial growth

Notably not all of the bacterial communities in the silos wereequally distinct from those in the tankers Analysis of the weightedUniFrac distances for silos alone by the unweighted pair groupmethod using average linkages (UPGMA) showed two majorgroups of silo-associated microbiota (Fig 7A see also Fig S4A inthe supplemental material) The first group included silos withbacteria that were significantly different from those in the tankersthat filled them as shown by a relatively large UniFrac distancebetween the tankers and the companion silo (Fig 7B see alsoFig S4B) Acinetobacter numbers were significantly enriched inthis group relative to the second group (statistically significant byMaAsLin and LefSe) Similarly silo B2 an outlier according toUPGMA clustering was also dominated by Acinetobacter The sec-ond group of silo-associated bacterial communities was moresimilar to those in the companion tankers by weighted UniFracand was enriched in numbers of Streptococcus Macrococcus andCorynebacterium (statistically significant by MaAsLin and LefSe)than to those in the first group Clostridium was also more abun-dant in the second group of silos than in the first although thisgenus was present at less than 2 relative abundance in the dataset (data not shown) No metadata collected at the time of sam-pling including clean-in-place (CIP) times and the date of sample

FIG 4 Relative proportions of the bacterial phyla present in raw tanker milk The relative proportions of OTU counts rarefied at 15000 sequences per sampleare shown A star beneath a box plot indicates a significant positive correlation by analysis of rarefied data with MaAslin (correlation coefficient 0 qvalue 005) and CSS-normalized data with LefSe (LDA effect size 2 P value 005) ldquoThermirdquo is a taxonomic assignment based on genome trees and is notofficially recognized (ie Bergeyrsquos Manual of Systematic Bacteriology)

FIG 5 Seasonal variation in the relative proportions of abundant taxa in the rawtanker milk microbiota Taxa present in at least 1 median relative abundanceafter rarefaction at 15000 sequences per sample are shown Those that were pos-itively correlated with a particular season by MaAslin analysis of rarefied data(coefficient 0 q value 005) and LefSe analysis of CSS-normalized data (LDAeffect size 0 P value 005) are marked with a white star

Raw Milk Microbiota during Transport and Storage

JulyAugust 2016 Volume 7 Issue 4 e00836-16 reg mbioasmorg 5

on August 23 2018 by guest

httpmbioasm

orgD

ownloaded from

collection were significantly correlated to these results Howeversilos A and B were present only in the first group whereas silos Dand E were present only in the second Silo C was present in bothclusters The median weighted UniFrac distance from tankers thatfilled each silo was significantly correlated with the number ofbacterial cells per milliliter estimated for each silo (Pearsonrsquosproduct-moment correlation 0706 P value 0007) (Fig 8)

DISCUSSION

We found that the bacterial populations in raw milk arriving atdairy processing facilities are highly diverse and differ according toseason however they still contain a core reproducible microbi-ota This core microbiota was vulnerable to modification as shownby the significant change in some bacterial populations upontransfer of the milk to storage silos

Although a number of studies have identified the bacteria inraw milk on farms and within creameries the microbial content ofmilk as it reaches the site of processing has remained largely un-known Here we found that bacterial populations in tanker milkare highly diverse and heterogeneous with a large proportion oftaxa present at less than 1 relative abundance This finding isconsistent with previous observations of raw milk in general (7)Notably the high level of bacterial diversity in raw milk is distinctfrom the bacterial composition of other host-associated environ-ments For example 20 or less of human fecal communities iscomposed of taxa below 1 relative abundance (unpublished data

from our laboratory and from the American Gut Project [round20] and data from Turnbaugh et al [32] analyzed in QIITAhttpsqiitaucsdedu) Such heterogeneity might be a result ofthe highly nutritive content of milk and the numerous potentialsources of bacteria such as aerosols cattle skin bedding feedhuman handling the microbiota resident on milking equipmenton-site bulk tanks used for storage and tanker trucks used fortransport Notably it was previously concluded that tanker haul-ing and cleaning practices do not significantly impact total bacte-rial or thermophilic spore counts in milk (33) Therefore the wallsof the transport tankers were not likely the primary source ofraw-milk-associated bacteria examined here

There is a possibility that dead cells could have contributed tothe bacterial heterogeneity of the milk examined For the purposeof distinguishing living and dead cells several studies have appliedpropidium monoazide (PMA) to measure the living fraction ofcells in processed or pasteurized milk (5 34 35) However be-cause raw milk has not undergone thermal processing and there-fore likely contains a preponderance of living cells we did not usePMA treatment We expect that any dead cells would not be con-sistently present and when present would be in such low propor-tions that they would not be represented among the core micro-biota

Despite the diversity and heterogeneity of bacteria in milkconsistent trends were also detected among the microbial com-munities according to various metrics (PCoA of the weightedUniFrac distance metric alpha-diversity metrics and relativeabundances of specific taxa) Among the parameters examinedthe season in which the milk was collected best distinguished be-tween the microbiota and had a greater impact than processingfacility sequencing depth or day of sampling Seasonal variationin bacterial community composition has been observed for nu-merous agricultural products including but not limited towheat lettuce radish sprouts and milk (2 7 36 37) A dominantseasonal difference found here was that the milk samples collectedin the spring contained the highest median richness in bacterialdiversity In addition to the increased diversity the total bacterialcell counts were modestly but still significantly higher in thespring Although there are myriad possible explanations for theincreased species richness this finding coincided with a significantincrease in the relative proportion of bacteria in the phylum Acti-nobacteria Because the Actinobacteria species identified here areassociated with soil (38ndash41) it is possible that the increase in spe-cies richness observed in spring was due to seasonal changes suchas increased precipitation promoting growth of Actinobacteria inthe soil and transfer to the cow udder Notably these results werenot necessarily due to increased access to the pasture in the springas pasture rates are low for the California Central Valley (data notshown)

Although the bacterial populations in each of the milk samplescollected in the spring season were more diverse than those in themilk sampled in other seasons they were also the most similar toeach other and contained fewer unique OTUs as a group than wasthe case for the milk from either the summer or fall By compari-son the highest number of unique OTUs was found in the fall Theweighted UniFrac distances between samples were also signifi-cantly greater in the fall than in the other seasons This differencein UniFrac distances might indicate that milk collected from dif-ferent dairies in the fall were exposed to a greater diversity ofbacteria possibly due to differences in feeding and housing

FIG 6 Bacterial community composition and cell number estimates for silosand tankers (A) Relative proportions of bacterial taxa that were present in atleast 2 relative abundance in silos and tankers at a rarefaction depth of 15000sequences All bacterial taxa present at less than 2 relative abundance weregrouped into the ldquootherrdquo classification Significant positive associations aremarked with a star (B) Bacterial cell counts estimated by qPCR are shown forlarge-volume silos and the tankers that filled them There is a trend for in-creased numbers of cells per milliliter in large-volume silos relative to tankerswhich approaches but does not quite reach significance (Welchrsquos t test P value006762)

Kable et al

6 reg mbioasmorg JulyAugust 2016 Volume 7 Issue 4 e00836-16

on August 23 2018 by guest

httpmbioasm

orgD

ownloaded from

throughout the region during that time of year Another possibil-ity is that the milk was undersampled and that DNA sequenceanalysis did not fully measure the distributions of rare taxa Sub-sequent studies could therefore aim for larger sample sizes andgreater sequencing depths to ensure bacterial recovery and iden-tification

Regardless of the sample-to-sample variation and seasonal dif-ferences certain taxa were represented in all tanker milk exam-ined The core microbiota encompassed taxa from 5 phyla and 18

families Among these bacteria Staphylococcus was particularlydominant This genus was previously detected in bovine milk anddescribed as one of the most abundant genera in human breastmilk (42ndash44) Both Staphylococcus and Streptococcus were detectedat the highest levels in milk overall and the populations of bothwere significantly enriched in the summer season This result is inagreement with culture-based studies that have similarly foundhigh numbers of Streptococcus in either the summer or winter(45ndash47)

FIG 7 Variations in silo microbiota (A) UPGMA cluster dendrogram of weighted UniFrac distances between raw milk silo communities (B) Box plot of theweighted UniFrac distances between each silo on the x axis and the tankers that filled it (C) Relative proportions of the taxa that were present in at least 2 relativeabundance in the silo milk samples Silos are labeled with a letter designating a physical silo and a number indicating either the first or second sampling week Silosamples with the same designation (eg A1) constitute the same silo tested at different times on the same day All analyses were performed using an OTU tablethat was rarefied to 15000 sequences per sample Significantly enriched taxa are indicated with a star

Raw Milk Microbiota during Transport and Storage

JulyAugust 2016 Volume 7 Issue 4 e00836-16 reg mbioasmorg 7

on August 23 2018 by guest

httpmbioasm

orgD

ownloaded from

Endospore-forming bacteria including Bacillus and Clostrid-ium were also members of the core raw milk microbiota Thesetaxa are common on the dairy farm and are known to be associ-ated with dairy processing environments (10 48 49) Bacillus andClostridium also encompass species associated with spoilage ofpasteurized milk and milk products Clostridium species espe-cially Clostridium tyrobutyricum are known to cause late blowingdefects in hard and semihard cheeses (50) Bacillus species causetexture defects and off flavors in pasteurized and refrigerated fluidmilk products (51) Clostridium unidentified members of Clos-tridiales and unidentified members of Bacillaceae tended to bepresent at lowest relative abundance in the summer The lowerproportion of these endospore-forming bacteria contradicts pre-vious studies showing that the prevalence of spore formers in rawmilk is high in the summer (52 53) Notably those studies wereperformed in locations with higher relative humidities than theCentral Valley CA which has a Mediterranean or semiarid cli-mate (54) Moreover feeding practices differ depending on geo-graphic location and silage in particular can serve as a source ofthermoduric bacteria and endospore former contamination (4852 53 55)

Corynebacterium and Acinetobacter and taxa in the Enterobac-teriaceae family were also members of the core microbiota and arebacteria frequently detected in raw milk (7) Of these genera Co-rynebacterium is regarded as both thermoduric and psychotropic(56) Members of this genus contribute beneficially to flavor de-velopment of smear-ripened cheeses (57) Although Corynebacte-rium was found in all tankers examined numbers of this memberof the Actinobacteria phylum tended to be enriched in springsuggesting that product flavor outcomes might be impacted byseasonal changes in the raw milk microbial community In con-trast Gammaproteobacteria such as both Acinetobacter and Enter-obacteriaceae are associated with spoilage (58) Acinetobacter isalso a psychrotroph however it is generally unable to survive

pasteurization (59) Notably Acinetobacter and Enterobacteriaceaetended to be more abundant in the spring and summer whereasPseudomonas another member of the Gammaproteobacteria wasmore prevalent in the fall Overall these results indicate a seasonalcomponent with respect to the dominant spoilage and nonstarterbacteria in milk Although pasteurization eliminates the majorityof impacts of these organisms on dairy products these bacteriaand their cellular components including heat-stable extracellularproteases and lipases (60) can sometimes survive thermal treat-ment or reenter through processing lines (5 59 61ndash63) The met-abolic and stress tolerance distinctions between different bacterialspecies could provide new opportunities to develop different hy-giene measures targeting the most problematic (or desirable) spe-cies on a seasonal basis

Lastly Streptococcus was found in the highest relative abun-dance among all the identified genera and was a member of thecore milk microbiota Both Streptococcus and Enterococcus a LABrelative of Streptococcus and member of the core microbiota werepreviously shown to be among the most abundant LAB in rawbovine milk collected directly from dairy farms (10) and dairyproduction facilities (5 7 64 65) Compared to other LAB generaStreptococcus and Enterococcus are recognized to be highly ther-moduric (66) and therefore might survive pasteurization BothStreptococcus and Enterococcus are typically detected together withother LAB such as Lactococcus Lactobacillus and Leuconostoc inraw milk (6 7) Taken together these bacteria are generally im-portant in fermented dairy product processing because they cansignificantly alter early acidification stages of cheese ripening andflavor development (67ndash70) Notably Lactococcus Lactobacillusand Leuconostoc were present at very low median relative abun-dances of 027 015 and 003 respectively in our raw milksamples and were not present in the core Although the reasons forthis are not clear it is notable that Lactococcus was present at highrelative proportions in some of the milk contained in large-volume holding silos (Fig 7) suggesting that these LAB mightgrow from low starting quantities in milk at the time of collection

By comparing the bacterial populations in milk in tankertrucks and silos examined in the summer season we establishedthat there was a rapid transformation in the microbial composi-tion in milk upon entry into dairy processing facilities Milk fromthe silos contained significant increases in the relative abundancesof Lactobacillales and Pseudomonadales compared to the subset oftankers that filled them Within these phyla bacteria such as Lac-tococcus Streptococcus Acinetobacter and Pseudomonas tended tobe present in relatively higher proportions in the silos than in thetanker trucks It is technically possible that some of this differencecould be due to incomplete mixing within the silos Howevermixing practices were the same for all silos and the enrichmentswere not evenly distributed throughout all silos Approximatelyhalf of the silos contained bacterial populations that were verysimilar to those in the tankers used to fill them These silos tendedto have a higher relative abundance of Streptococcus The othersilos were populated with a microbiota that was clearly distinctand separate from that in the tankers These silos containedgreater proportions of Acinetobacter and Lactococcus Moreoveralthough Lactococcus can be used as a starter culture in cheesefermentations because this organism was not consistently en-riched in the silos the recovery of this organism was likely not theresult of cross-contamination due to aerosols or other transfermechanisms on site Our findings in combination with increased

FIG 8 Median weighted UniFrac distances between silos and tankers com-pared to bacterial load The median weighted UniFrac distances between silosand the tankers that filled them are shown on the y axis The log10 valuescorresponding to the number of cells per milliliter estimated for each silo usingqPCR are shown on the x axis These values are linearly correlated (Pearsonrsquosproduct-moment correlation 0706 P value 0007) Points are depicted as textwith a letter designating a physical silo and a number indicating either the firstor second sampling week The text is colored by UPGMA cluster (1 black2 blue 3 purple) The tendency toward both higher bacterial load andlarger UniFrac distance tends to be silo specific For example silo A has thegreatest distance from corresponding tankers and E has the smallest distanceregardless of time since CIP or collection date

Kable et al

8 reg mbioasmorg JulyAugust 2016 Volume 7 Issue 4 e00836-16

on August 23 2018 by guest

httpmbioasm

orgD

ownloaded from

bacterial cell count estimates within those silos suggest that mem-bers of those species grew during cold storage

Similarly to our findings bacterial numbers were previouslyfound to increase during low-temperature containment (25) Therelative proportions of Acinetobacter were also found to increase(71) The notion that these specific genera were enriched by coldstorage is supported by the fact that species of Acinetobacter Lac-tococcus and Streptococcus are regarded to be psychrotrophs (7273) However the development of two distinct silo communitiesindicates that processes other than just cold storage might havedetermined the community development No taxa were signifi-cantly associated with tankers that filled a given set of silos indi-cating that tanker communities are not a strong predictor of silocommunity composition Equipment-associated persistent bio-films or other external contamination sources might instead pre-dispose individual silos toward specific microbial communitystructures Regardless the observed differences between bacterialpopulations before and after transfer and short-term (3-to-6-h)storage in different containment vessels show that these popula-tions are dynamic and can change quickly in response to newconditions within food processing facilities

In conclusion we have shown that the raw milk microbialcommunities that we examined were similar to each other despitebeing collected from different farms transported to different lo-cations and sampled at different times of the year Beyond thecore milk microbiota in California the bacterial content changeddepending on season and containment equipment (truck or silo)Importantly the bacterial composition of raw milk can be dra-matically yet variably impacted during low-temperature short-term storage This knowledge is crucial to identifying the bacteriathat are responsible for sporadic but consistent defects in cheeseand other dairy products and developing improved methods forthe treatment and handling of raw and processed milk to ensurethe production of consistently high-quality products

MATERIALS AND METHODSMilk source and sampling Milk was collected with a stainless steel dipperfrom the inlet at the top of tanker trucks with 6000-gal (22712-liter)capacity upon arrival at two dairy processors on 7 October 2013 (fall) 13October 2013 (fall) 5 March 2014 (spring) 11 March 2014 (spring) 26August 2014 (summer) 27 August 2014 (summer) and 1 September 2014(summer) For the summer samples the local weather conditions on thetwo collection dates were very similar (80degF with no precipitation andwind speeds of 5 mph) Samples were collected from individual trucksover a 20-h period on each sampling date After collection milk sampleswere placed in sterile 50-ml tubes or 400-ml bags and kept at 4degC At theend of the 20-h sampling period samples were shipped overnight on ice toDavis CA where they were processed for bacterial DNA extraction im-mediately upon arrival (a total of 12 to 32 h after collection) The tankerscontained milk of variable grades from between one and three differentdairy farms of a total of 200 farms located in Merced Stanislaus TulareKings Fresno Madera Kern andor San Joaquin county in CaliforniaDairy size and milking practices were varied although they were consis-tent insofar as silage and alfalfa were the primary feed types and milk washeld no longer than 48 h between pickup and delivery Although there wasnot a single cleaning protocol for the tankers valid wash tags indicatingthat tankers had been washed within the previous 24 h were checked priorto unloading at both facilities A total of 974 raw tanker milk sampleswere collected including 273 from spring (148 from processor A and 125from processor B) 451 from summer (300 from processor A and 151 fromprocessor B) and 244 from fall (126 from processor A and 118 fromprocessor B) On 26 August 2014 and 1 September 2014 milk was also

sampled from five large-volume-capacity silos at processor A The siloswere cleaned when empty (approximately every 48 h) They were kept attemperatures below 72degC and the samples were collected from a valvelocated at the silo base immediately after they were filled For each silo 13to 25 corresponding tanker milk trucks were measured throughout thefilling period and labeled with the silo lot identifier (ID) number Mixingwithin the silos was a result of the filling process and mechanical agi-tation Silo fill times were variable but filling was typically completedwithin 3 to 6 h

Bacterial genomic DNA extraction Bacterial cells were collectedfrom 25 to 30 ml raw milk by centrifugation at 13000 g at 4degC for 5 minThe cells were then suspended in phosphate-buffered saline (PBS pH 72 137 mM NaCl 268 mM KCl 101 mM Na2HPO4 176 mM KH2PO4)and centrifuged a second time at 13000 g for 2 min Bacterial pelletswere stored at 20degC until DNA extraction A PowerFood microbialDNA isolation kit (Mo Bio Laboratories Inc Carlsbad CA) was used forDNA extraction and purification according to the manufacturerrsquos proto-col with the exception that instead of subjecting MicroBead tubes (MoBio Laboratories Inc Carlsbad CA) to vortex mixing the cell suspen-sions were shaken twice for 1 min at 65 ms on a FastPrep-24 instrument(MP Biomedicals LLC)

Bacterial count estimates by quantitative real-time PCR Bacte-rial cell numbers were estimated using quantitative PCR (qPCR) Eachreaction mixture contained SsoFast Evagreen Supermix with Low ROX(Bio-Rad Laboratories Inc USA) and 400 nM UniF (5=-GTGSTGCAYGGYYGTCGTCA-3=) and UniR (5=-ACGTCRTCCMCNCCTTCCTC-3=) (74) qPCR was performed in a 7500 Fast Real TimePCR system (Applied Biosystems Foster City CA) with initiation at95degC for 20 s followed by 40 cycles of 95degC for 3 s and 60degC for 30 sNonspecific amplification was evaluated by melting curve analysisBacteria were enumerated by comparisons of average threshold cycle(CT) values (n 2) to a genomic DNA standard curve that was run onthe same qPCR plate The standard curve was prepared from DNAextracted from Lactobacillus casei BL23 grown to exponential phase(A600 05 to 07) at 37degC in MRS broth (Becton Dickinson andCompany Sparks MD) L casei cell numbers were determined byplating serial dilutions of the exponential-phase cells onto MRS agarfor colony enumeration

16S rRNA gene sequencing The V4 region of 16S rRNA genes wasamplified with primers F515 and R806 with a random 8-bp bar code onthe 5= end of F515 (75 76) PCR amplification was performed using ExTaq DNA polymerase (TaKaRa Otsu Japan) with 35 cycles of 94degC for45 s 54degC for 60 s and 72degC for 30 s The PCR products were pooled andthen purified with a Wizard SV Gel and PCR cleanup system (PromegaMadison WI) NEXTflex adapters (Bioo Scientific Austin TX) were li-gated to the 16S rRNA amplicons prior to 250-bp paired-end sequencingperformed on an Illumina MiSeq instrument at the University of Califor-nia Davis (httpdnatechgenomecenterucdavisedu)

16S rRNA gene sequence analysis FASTQ files were analyzed usingQIIME version 191 (77) Paired-end sequences were joined using fastq-join (78) with 175 bp overlap and a 1 maximum difference requiredbetween all paired sequences The split_libraries_fastqpy script was usedfor demultiplexing and quality filtering of the resulting joined sequencesDemultiplexing was performed only with bar codes containing no se-quencing errors and quality filtering was performed at a Phred qualitythreshold of 30 Chimeric sequences were identified with USEARCH (7980) and removed The remaining DNA sequences were grouped intoOTUs with 97 matched sequence identity by the use of the ldquoopen refer-encerdquo OTU picking method in QIIME with default settings Greengenes13_8 was used as the reference database (81) for both chimera checkingand OTU picking Sequences were aligned by the use of PyNAST (81 82)and taxonomy was assigned using the taxonomy database in Greengenes(83 84) The resulting OTU counts were filtered to remove any OTU thatoccurred at less than 0005 relative abundance in the raw tanker milkdata set (85) OTUs occurring at less than 0016 in the silos-versus-

Raw Milk Microbiota during Transport and Storage

JulyAugust 2016 Volume 7 Issue 4 e00836-16 reg mbioasmorg 9

on August 23 2018 by guest

httpmbioasm

orgD

ownloaded from

tankers data set (10521986 sequences) and less than 0288 in the silos-versus-silos data set (583436 sequences) were also removed to focus ouranalysis on the more abundant bacteria within the processing facility

Statistics OTU counts were adjusted by rarefaction at 15000 se-quences per sample or by cumulative sum scaling (CSS) (30) in QIIMEDNA from five milk samples collected in the fall season was PCR amplifiedand used in each of five MiSeq runs These samples were labeled as con-trols and principal coordinate analysis in the QIIME package was used todetermine the presence of a batch effect due to PCR amplification andsequencing A confounding batch effect was present for raw tanker milksamples between sequencing runs in examinations performed using CSSnormalization or unweighted UniFrac Therefore unweighted UniFracvalues were not used in this study and batch correction was performed onCSS-normalized data using the ComBat function in the sva package inbioconductor (86 87) Illustrations of these analyses were generated usingthe R vegan package (88)

The statistical significance of differences in community compositionwas determined by analyzing the weighted UniFrac distances with Adonisfrom the R vegan package wrapped in QIIME with 9999 permutationsPrior to performing alpha diversity or differential abundance testing thesequencing controls were removed from the OTU table after normaliza-tion (CSS with batch correction or rarefaction) was performed This al-lowed correction of batch effects without skewing the differential abun-dance results due to the presence of replicate samples

The alpha diversity of each sample was determined in two ways Firstthe relative number of OTUs observed per sample in each season wasevaluated at a sequencing depth of 15000 by the use of the Kruskal-Wallistest followed by the Nemenyi test with the Tukey method for pairwisecomparisons and P values were adjusted using Benjamini-Hochberg false-discovery-rate correction Second the breakaway Species Richness Esti-mation and Modeling R package (89) was used to estimate the total speciesrichness of each sample and the results were compared in the same way asthe rarefied alpha diversity values

To determine differences in the relative abundances of specific taxo-nomic classifications between experimental groups OTU counts from therarefied OTU table were summed by the most specific taxonomic classi-fication identified for each OTU and analyzed using MaAsLin (multivar-iate analysis with linear modeling) version 003 MaAsLin is a statisticalsoftware package that applies removal of low-abundance values boostingand a multivariate linear model followed by Bonferroni false-discovery-rate correction to identify taxa that are significantly associated with par-ticular metadata (httphuttenhowersphharvardedumaaslin) (90 91)The data from the CSS-normalized OTU table with batch correction weresimilarly summed and analyzed using LefSe (92) LefSe was selected overmetagenomeSeq because the results were more conservative (30) Differ-ences were considered significant by analysis in MaAslin if the q value(corrected P value) was less than 005 and in LefSe if the P value was lessthan 005 and the linear discriminatory analysis (LDA) effect size wasgreater than 20 Bacterial features are reported in the body of the text asdifferentially abundant between groups if the LefSe analysis of the CSS-normalized OTU table with batch correction agreed with the MaAsLinanalysis of the rarefied OTU table results for that particular feature

The silo milk microbial communities and those in the companiontanker milk samples were sequenced in two sequencing runs No batcheffect was detected for these two runs and therefore batch correction wasnot necessary prior to analysis of the impact of silo storage on raw milkmicrobial communities In order to evaluate only the most abundantbacterial taxa in the silo communities only the taxa present at greater that2 relative abundance were evaluated using the statistical methods de-scribed above

Accession numbers Joined and demultiplexed DNA sequences weredeposited in the Qiita database (httpsqiitaucsdedu) under study ID10485 and in the European Nucleotide Archive (ENA) under accessionnumber ERP015209

SUPPLEMENTAL MATERIALSupplemental material for this article may be found at httpmbioasmorglookupsuppldoi101128mBio00836-16-DCSupplemental

Figure S1 TIF file 27 MBFigure S2 TIF file 26 MBFigure S3 TIF file 25 MBFigure S4 TIF file 27 MBTable S1 PDF file 02 MB

ACKNOWLEDGMENTS

We thank Ray Lum and Betty Lumbres for performing and supervisingthe milk sample collection necessary for this study

The California Dairy Research Foundation grant ldquoBacterial signaturesof milk safety and qualityrdquo funded this study The funding agency did notparticipate in study design data collection or interpretation of the data

Hilmar Cheese Company employs JM and JH

FUNDING INFORMATIONThis work including the efforts of Maria L Marco was funded by Califor-nia Dairy Research Foundation (CDRF)

REFERENCES1 Ottesen AR Gonzaacutelez Pentildea A White JR Pettengill JB Li C Allard S

Rideout S Allard M Hill T Evans P Strain E Musser S Knight RBrown E 2013 Baseline survey of the anatomical microbial ecology of animportant food plant Solanum lycopersicum (tomato) BMC Microbiol13114 httpdxdoiorg1011861471-2180-13-114

2 Williams TR Moyne AL Harris LJ Marco ML 2013 Season irrigationleaf age and Escherichia coli inoculation influence the bacterial diversityin the lettuce phyllosphere PLoS One 8e68642 httpdxdoiorg101371journalpone0068642

3 Williams TR Marco ML 2014 Phyllosphere microbiota compositionand microbial community transplantation on lettuce plants grown in-doors mBio 5e01564-14 httpdxdoiorg101128mBio01564-14

4 Leff JW Fierer N 2013 Bacterial communities associated with the sur-faces of fresh fruits and vegetables PLoS One 8e59310 httpdxdoiorg101371journalpone0059310

5 Quigley L McCarthy R OrsquoSullivan O Beresford TP Fitzgerald GFRoss RP Stanton C Cotter PD 2013 The microbial content of raw andpasteurized cow milk as determined by molecular approaches J Dairy Sci964928 ndash 4937 httpdxdoiorg103168jds2013-6688

6 Quigley L OrsquoSullivan O Beresford TP Ross RP Fitzgerald GF CotterPD 2012 High-throughput sequencing for detection of subpopulationsof bacteria not previously associated with artisanal cheeses Appl EnvironMicrobiol 785717ndash5723 httpdxdoiorg101128AEM00918-12

7 Quigley L OrsquoSullivan O Stanton C Beresford TP Ross RP FitzgeraldGF Cotter PD 2013 The complex microbiota of raw milk FEMS Micro-biol Rev 37664 ndash 698 httpdxdoiorg1011111574-697612030

8 Dunkley CS McReynolds JL Dunkley KD Njongmeta LN BerghmanLR Kubena LF Nisbet DJ Ricke SC 2007 Molting in Salmonellaenteritidis-challenged laying hens fed alfalfa crumbles IV Immune andstress protein response Poult Sci 862502ndash2508 httpdxdoiorg103382ps2006-00401

9 Krause DO Nagaraja TG Wright AD Callaway TR 2013 Board-invited review rumen microbiology leading the way in microbial ecologyJ Anim Sci 91331ndash341 httpdxdoiorg102527jas2012-5567

10 Vacheyrou M Normand AC Guyot P Cassagne C Piarroux R BoutonY 2011 Cultivable microbial communities in raw cow milk and potentialtransfers from stables of sixteen French farms Int J Food Microbiol 146253ndash262 httpdxdoiorg101016jijfoodmicro201102033

11 Delbegraves C Ali-Mandjee L Montel MC 2007 Monitoring bacterial com-munities in raw milk and cheese by culture-dependent and -independent16S rRNA gene-based analyses Appl Environ Microbiol 731882ndash1891httpdxdoiorg101128AEM01716-06

12 Machado SG da Silva FL Bazzolli DM Heyndrickx M Costa PMVanetti MC 2015 Pseudomonas spp and Serratia liquefaciens as predom-inant spoilers in cold raw milk J Food Sci 80M1842ndashM1849 httpdxdoiorg1011111750-384112957

13 Ercolini D Russo F Torrieri E Masi P Villani F 2006 Changes in thespoilage-related microbiota of beef during refrigerated storage under dif-

Kable et al

10 reg mbioasmorg JulyAugust 2016 Volume 7 Issue 4 e00836-16

on August 23 2018 by guest

httpmbioasm

orgD

ownloaded from

ferent packaging conditions Appl Environ Microbiol 724663ndash 4671httpdxdoiorg101128AEM00468-06

14 Ferrocino I Greppi A La Storia A Rantsiou K Ercolini D Cocolin L2016 Impact of nisin-activated packaging on microbiota of beef burgersduring storage Appl Environ Microbiol 82549 ndash559 httpdxdoiorg101128AEM03093-15

15 Food and Agriculture Organization of the United Nations (FAOSTAT)2013 Food balance domestic utilization data httpfaostat3faoorgcompareE

16 Hoskin R Cessna J 2016 From raw milk to dairy products United StatesDepartment of Agriculture Economic Research Service Washington DCh t t p w w w e r s u s d a g o v t o p i c s a n i m a l - p r o d u c t s d a i r y backgroundaspx

17 Corroler D Mangin I Desmasures N Gueguen M 1998 An ecologicalstudy of lactococci isolated from raw milk in the Camembert cheese reg-istered designation of origin area Appl Environ Microbiol 644729 ndash 4735

18 Wolfe BE Button JE Santarelli M Dutton RJ 2014 Cheese rind com-munities provide tractable systems for in situ and in vitro studies of mi-crobial diversity Cell 158422ndash 433 httpdxdoiorg101016jcell201405041

19 Sandine WE Daly C Elliker PR Vedamuthu ER 1972 Causes andcontrol of culture-related flavor defects in cultured dairy-products JDairy Sci 551031ndash1039 httpdxdoiorg103168jdsS0022-0302(72)85617-0

20 Ledenbach LH Marshall RT 2009 Microbiological spoilage of dairyproducts p 41ndash 67 In Sperber WH Doyle MP (ed) Compendium of themicrobiological spoilage of foods and beverages Springer New York NYhttpdxdoiorg101007978-1-4419-0826-1_2

21 Coorevits A De Jonghe V Vandroemme J Reekmans R Heyrman JMessens W De Vos P Heyndrickx M 2008 Comparative analysis of thediversity of aerobic spore-forming bacteria in raw milk from organic andconventional dairy farms Syst Appl Microbiol 31126 ndash140 httpdxdoiorg101016jsyapm200803002

22 White DG Harmon RJ Matos JE Langlois BE 1989 Isolation andidentification of coagulase-negative Staphylococcus species from bovinebody sites and streak canals of nulliparous heifers J Dairy Sci 721886 ndash1892 httpdxdoiorg103168jdsS0022-0302(89)79307-3

23 Lejeune JT Rajala-Schultz PJ 2009 Food safety unpasteurized milk acontinued public health threat Clin Infect Dis 4893ndash100 httpdxdoiorg101086595007

24 Bonizzi I Buffoni JN Feligini M Enne G 2009 Investigating the rela-tionship between raw milk bacterial composition as described by inter-genic transcribed spacer-PCR fingerprinting and pasture altitude J ApplMicrobiol 1071319 ndash1329 httpdxdoiorg101111j 1365-2672200904311x

25 Costello M Rhee MS Bates MP Clark S Luedecke LO Kang DH 2003Eleven-year trends of microbiological quality in bulk tank milk FoodProtect Trends 23393ndash 400 httpopenagricolanalusdagovRecordIND44658844

26 Pantoja JC Reinemann DJ Ruegg PL 2009 Associations among milkquality indicators in raw bulk milk J Dairy Sci 924978 ndash 4987 httpdxdoiorg103168jds2009-2329

27 Pantoja JC Reinemann DJ Ruegg PL 2011 Factors associated withcoliform count in unpasteurized bulk milk J Dairy Sci 942680 ndash2691httpdxdoiorg103168jds2010-3721

28 United States Department of Agriculture Economic Research Service(ERS) 2015 Milk cows and production by state and region (annual)USDA ERS Washington DC httpwwwersusdagovdata-productsdairy-dataaspx

29 NASS 2015 Dairy Products Annual Summary National Agricultural Sta-tistics Service Agricultural Statistics Board United States Department ofAgr icu l ture h t tp usda mannl ib corne l l eduMannUsdaviewDocumentInfododocumentID1052

30 Joseph N Paulson CS Corrada Bravo H Pop M 2013 Robust methodsfor differential abundance analysis in marker gene surveys Nat Methods101200 ndash1202 httpdxdoiorg101038nmeth2658

31 Hughes JB Hellman JJ Ricketts TH Bohannan BJM 2001 Countingthe uncountable statistical approaches to estimating microbial diversityAppl Environ Microbiol 674399 ndash 4406 httpdxdoiorg101128AEM67104399-44062001

32 Turnbaugh PJ Hamady M Yatsunenko T Cantarel BL Duncan A LeyRE Sogin ML Jones WJ Roe BA Affourtit JP Egholm M Henrissat BHeath AC Knight R Gordon JI 2009 A core gut microbiome in obese

and lean twins Nature 457480 ndash 484 httpdxdoiorg101038nature07540

33 Darchuk EM Waite-Cusic J Meunier-Goddik L 2015 Effect of com-mercial hauling practices and tanker cleaning treatments on raw milkmicrobiological quality J Dairy Sci 987384 ndash7393 httpdxdoiorg103168jds2015-9746

34 Zhang Z Liu W Xu H Aguilar ZP Shah NP Wei H 2015 Propidiummonoazide combined with real-time PCR for selective detection of viableStaphylococcus aureus in milk powder and meat products J Dairy Sci 981625ndash1633 httpdxdoiorg103168jds2014-8938

35 Soejima T Minami J Iwatsuki K 2012 Rapid propidium monoazidePCR assay for the exclusive detection of viable Enterobacteriaceae cells inpasteurized milk J Dairy Sci 953634 ndash3642 httpdxdoiorg103168jds2012-5360

36 Smit E Leeflang P Gommans S van den Broek J van Mil S WernarsK 2001 Diversity and seasonal fluctuations of the dominant members ofthe bacterial soil community in a wheat field as determined by cultivationand molecular methods Appl Environ Microbiol 672284 ndash2291 httpdxdoiorg101128AEM6752284-22912001

37 Asakura H Tachibana M Taguchi M Hiroi T Kurazono H Makino SKasuga F Igimi S 2016 Seasonal and growth-dependent dynamics ofbacterial community in radish sprouts J Food Saf 36392ndash 401 httpdxdoiorg101111jfs1225610

38 Edmonds-Wilson SL Nurinova NI Zapka CA Fierer N Wilson M2015 Review of human hand microbiome research J Dermatol Sci 803ndash12 httpdxdoiorg101016jjdermsci201507006

39 Piao Z Yang L Zhao L Yin S 2008 Actinobacterial community struc-ture in soils receiving long-term organic and inorganic amendments ApplEnviron Microbiol 74526 ndash530 httpdxdoiorg101128AEM00843-07

40 Eisenlord SD Zak DR Upchurch RA 2012 Dispersal limitation and theassembly of soil Actinobacteria communities in a long-term chronose-quence Ecol Evol 2538 ndash549 httpdxdoiorg101002ece3210

41 Barnard RL Osborne CA Firestone MK 2013 Responses of soil bacte-rial and fungal communities to extreme desiccation and rewetting ISME J72229 ndash2241 httpdxdoiorg101038ismej2013104

42 Martiacuten R Heilig HG Zoetendal EG Jimeacutenez E Fernaacutendez L Smidt HRodriacuteguez JM 2007 Cultivation-independent assessment of the bacterialdiversity of breast milk among healthy women Res Microbiol 15831ndash37httpdxdoiorg101016jresmic200611004

43 Heikkilauml MP Saris PE 2003 Inhibition of Staphylococcus aureus by thecommensal bacteria of human milk J Appl Microbiol 95471ndash 478 httpdxdoiorg101046j1365-2672200302002x

44 Hunt KM Foster JA Forney LJ Schuumltte UM Beck DL Abdo Z FoxLK Williams JE McGuire MK McGuire MA 2011 Characterization ofthe diversity and temporal stability of bacterial communities in humanm i l k P L o S O n e 6 e 2 1 3 1 3 h t t p d x d o i o r g 1 0 1 3 7 1 journalpone0021313

45 Olde Riekerink RG Barkema HW Stryhn H 2007 The effect of seasonon somatic cell count and the incidence of clinical mastitis J Dairy Sci901704 ndash1715 httpdxdoiorg103168jds2006-567

46 Reference deleted47 Gillespie BE Lewis MJ Boonyayatra S Maxwell ML Saxton A Oliver

SP Almeida RA 2012 Short communication evaluation of bulk tankmilk microbiological quality of nine dairy farms in Tennessee J Dairy Sci954275ndash 4279 httpdxdoiorg103168jds2011-4881

48 Julien MC Dion P Lafreniegravere C Antoun H Drouin P 2008 Sources ofclostridia in raw milk on farms Appl Environ Microbiol 746348 ndash 6357httpdxdoiorg101128AEM00913-08

49 Huck JR Hammond BH Murphy SC Woodcock NH Boor KJ 2007Tracking spore-forming bacterial contaminants in fluid milk-processingsystems J Dairy Sci 904872ndash 4883 httpdxdoiorg103168jds2007-0196

50 Klijn N Nieuwenhof FF Hoolwerf JD van der Waals CB WeerkampAH 1995 Identification of Clostridium tyrobutyricum as the causativeagent of late blowing in cheese by species-specific PCR amplification ApplEnviron Microbiol 612919 ndash2924

51 Gopal N Hill C Ross PR Beresford TP Fenelon MA Cotter PD 2015The prevalence and control of Bacillus and related spore-forming bacteriain the dairy industry Front Microbiol 61418 httpdxdoiorg103389fmicb201501418

52 Bermuacutedez J Gonzaacutelez MJ Olivera JA Burguentildeo JA Juliano P Fox EMReginensi SM 2016 Seasonal occurrence and molecular diversity of clos-

Raw Milk Microbiota during Transport and Storage

JulyAugust 2016 Volume 7 Issue 4 e00836-16 reg mbioasmorg 11

on August 23 2018 by guest

httpmbioasm

orgD

ownloaded from

tridia species spores along cheesemaking streams of 5 commercial dairyplants J Dairy Sci 993358 ndash3366 httpdxdoiorg103168jds2015-10079

53 Buehner KP Anand S Djira GD Garcia A 2014 Prevalence of thermo-duric bacteria and spores on 10 Midwest dairy farms J Dairy Sci 976777ndash 6784 httpdxdoiorg103168jds2014-8342

54 Peel MC Finlayson BL McMahon TA 2007 Updated world map of theKoppen-Geiger climate classification Hydrol Earth Syst Sci 111633ndash1644 httpdxdoiorg105194hess-11-1633-2007

55 Christiansson A Bertilsson J Svensson B 1999 Bacillus cereus spores inraw milk factors affecting the contamination of milk during the grazingperiod J Dairy Sci 82305ndash314 httpdxdoiorg103168jdsS0022-0302(99)75237-9

56 Washam CJ Olson HC Vedamuthu ER 1977 Heat-resistant psychro-trophic bacteria isolated from pasteurized milk J Food Protect 40101ndash108

57 Corsetti A Rossi J Gobbetti M 2001 Interactions between yeasts andbacteria in the smear surface-ripened cheeses Int J Food Microbiol 691ndash10 httpdxdoiorg101016S0168-1605(01)00567-0

58 Cousin MA 1982 Presence and activity of psychrotrophic microorgan-isms in milk and dairy-productsmdasha review J Food Protect 45172ndash207

59 Wang CY Wu HD Lee LN Chang HT Hsu YL Yu CJ Yang PCHsueh PR 2006 Pasteurization is effective against multidrug-resistantbacteria Am J Infect Control 34320 ndash322 httpdxdoiorg101016jajic200511006

60 Dogan B Boor KJ 2003 Genetic diversity and spoilage potentials amongPseudomonas spp isolated from fluid milk products and dairy processingplants Appl Environ Microbiol 69130 ndash138 httpdxdoiorg101128AEM691130-1382003

61 Gunasekera TS Soslashrensen A Attfield PV Soslashrensen SJ Veal DA 2002Inducible gene expression by nonculturable bacteria in milk after pasteur-ization Appl Environ Microbiol 681988 ndash1993 httpdxdoiorg101128AEM6841988-19932002

62 Cleto S Matos S Kluskens L Vieira MJ 2012 Characterization ofcontaminants from a sanitized milk processing plant PLoS One 7e40189httpdxdoiorg101371journalpone0040189

63 Dogan B Boor KJ 2003 Genetic diversity and spoilage potentials amongPseudomonas spp isolated from fluid milk products and dairy processingplants Appl Environ Microbiol 69130 ndash138 httpdxdoiorg101128AEM691130-1382003

64 Masoud W Vogensen FK Lillevang S Abu Al-Soud W Soslashrensen SJJakobsen M 2012 The fate of indigenous microbiota starter culturesEscherichia coli Listeria innocua and Staphylococcus aureus in Danish rawmilk and cheeses determined by pyrosequencing and quantitative realtime (qRT)-PCR Int J Food Microbiol 153192ndash202 httpdxdoiorg101016jijfoodmicro201111014

65 Hou Q Xu H Zheng Y Xi X Kwok LY Sun Z Zhang H Zhang W2015 Evaluation of bacterial contamination in raw milk ultra-high tem-perature milk and infant formula using single molecule real-time se-quencing technology J Dairy Sci 988464 ndash 8472 httpdxdoiorg103168jds2015-9886

66 McAuley CM Gobius KS Britz ML Craven HM 2012 Heat resistanceof thermoduric enterococci isolated from milk Int J Food Microbiol 154162ndash168 httpdxdoiorg101016jijfoodmicro201112033

67 Gelsomino R Vancanneyt M Cogan TM Condon S Swings J 2002Source of enterococci in a farmhouse raw-milk cheese Appl Environ Mi-crobiol 683560 ndash3565 httpdxdoiorg101128AEM6873560-35652002

68 Marth EH 1963 Microbiological and chemical aspects of Cheddar cheeseripening A review J Dairy Sci 46869 ndash 890 httpdxdoiorg103168jdsS0022-0302(63)89174-2

69 Settanni L Moschetti G 2010 Non-starter lactic acid bacteria used toimprove cheese quality and provide health benefits Food Microbiol 27691ndash 697 httpdxdoiorg101016jfm201005023

70 Franciosi E Settanni L Cavazza A Poznanski E 2009 Biodiversity andtechnological potential of wild lactic acid bacteria from raw cowsrsquo milk IntDairy J 193ndash11 httpdxdoiorg101016jidairyj200807008

71 Raats D Offek M Minz D Halpern M 2011 Molecular analysis ofbacterial communities in raw cow milk and the impact of refrigeration onits structure and dynamics Food Microbiol 28465ndash 471 httpdxdoiorg101016jfm201010009

72 Pothakos V Taminiau B Huys G Nezer C Daube G Devlieghere F2014 Psychrotrophic lactic acid bacteria associated with production batch

recalls and sporadic cases of early spoilage in Belgium between 2010 and2014 Int J Food Microbiol 191157ndash163 httpdxdoiorg101016jijfoodmicro201409013

73 Hantsis-Zacharov E Halpern M 2007 Culturable psychrotrophic bac-terial communities in raw milk and their proteolytic and lipolytic traitsAppl Environ Microbiol 737162ndash7168 httpdxdoiorg101128AEM00866-07

74 Belenguer A Duncan SH Calder AG Holtrop G Louis P Lobley GEFlint HJ 2006 Two routes of metabolic cross-feeding between Bifidobac-terium adolescentis and butyrate-producing anaerobes from the humangut Appl Environ Microbiol 723593ndash3599 httpdxdoiorg101128AEM7253593-35992006

75 Bokulich NA Joseph CM Allen G Benson AK Mills DA 2012 Next-generation sequencing reveals significant bacterial diversity of botrytizedw i n e P L o S O n e 7 e 3 6 3 5 7 h t t p d x d o i o r g 1 0 1 3 7 1 journalpone0036357

76 McInnis EA Kalanetra KM Mills DA Maga EA 2015 Analysis of rawgoat milk microbiota impact of stage of lactation and lysozyme on micro-bial diversity Food Microbiol 46121ndash131 httpdxdoiorg101016jfm201407021

77 Caporaso JG Kuczynski J Stombaugh J Bittinger K Bushman FDCostello EK Fierer N Pentildea AG Goodrich JK Gordon JI Huttley GAKelley ST Knights D Koenig JE Ley RE Lozupone CA McDonald DMuegge BD Pirrung M Reeder J Sevinsky JR Tumbaugh PJ WaltersWA Widmann J Yatsunenko T Zaneveld J Knight R 2010 QIIMEallows analysis of high-throughput community sequencing data NatMethods 7335ndash336 httpdxdoiorg101038nmethf303

78 Aronesty E 2011 Ea-utils command-line tools for processing biologicalsequencing data httpexpressionanalysisgithubioea-utils

79 Edgar RC Haas BJ Clemente JC Quince C Knight R 2011 UCHIMEimproves sensitivity and speed of chimera detection Bioinformatics 272194 ndash2200 httpdxdoiorg101093bioinformaticsbtr381

80 Edgar RC 2010 Search and clustering orders of magnitude faster thanBLAST Bioinformatics 262460 ndash2461 httpdxdoiorg101093bioinformaticsbtq461

81 DeSantis TZ Hugenholtz P Larsen N Rojas M Brodie EL Keller KHuber T Dalevi D Hu P Andersen GL 2006 Greengenes a chimera-checked 16S rRNA gene database and workbench compatible with ARBAppl Environ Microbiol 725069 ndash5072 httpdxdoiorg101128AEM03006-05

82 Caporaso JG Bittinger K Bushman FD DeSantis TZ Andersen GLKnight R 2010 PyNAST a flexible tool for aligning sequences to a tem-plate alignment Bioinformatics 26266 ndash267 httpdxdoiorg101093bioinformaticsbtp636

83 McDonald D Price MN Goodrich J Nawrocki EP DeSantis TZ ProbstA Andersen GL Knight R Hugenholtz P 2012 An improved Green-genes taxonomy with explicit ranks for ecological and evolutionary anal-yses of bacteria and archaea ISME J 6610 ndash 618 httpdxdoiorg101038ismej2011139

84 Werner JJ Koren O Hugenholtz P DeSantis TZ Walters WA Capo-raso JG Angenent LT Knight R Ley RE 2012 Impact of training sets onclassification of high-throughput bacterial 16S rRNA gene surveys ISMEJ 694 ndash103 httpdxdoiorg101038ismej201182

85 Bokulich NA Subramanian S Faith JJ Gevers D Gordon JI Knight RMills DA Caporaso JG 2013 Quality-filtering vastly improves diversityestimates from Illumina amplicon sequencing Nat Methods 1057ndashU11httpdxdoiorg101038nmeth2276

86 Leek JT Johnson WE Parker HS Jaffe AE Storey JD 2012 The svapackage for removing batch effects and other unwanted variation in high-throughput experiments Bioinformatics 28882ndash 883 httpdxdoiorg101093bioinformaticsbts034

87 Johnson WE Li C Rabinovic A 2007 Adjusting batch effects in microar-ray expression data using empirical Bayes methods Biostatistics8118 ndash127 httpdxdoiorg101093biostatisticskxj037

88 Jari Oksanen FGB Kindt R Legendre P Minchin PR OrsquoHara RBSimpson GL Solymos P Henry M Stevens H Wagner H 2015 vegancommunity ecology package R package version 23-0 httpscranr-projectorgwebpackagesveganindexhtml

89 Willis A Bunge J 2015 Estimating diversity via frequency ratios Bio-metrics 711042ndash1049 httpdxdoiorg101111biom12332

90 Morgan XC Tickle TL Sokol H Gevers D Devaney KL Ward DVReyes JA Shah SA LeLeiko N Snapper SB Bousvaros A Korzenik J

Kable et al

12 reg mbioasmorg JulyAugust 2016 Volume 7 Issue 4 e00836-16

on August 23 2018 by guest

httpmbioasm

orgD

ownloaded from

Sands BE Xavier RJ Huttenhower C 2012 Dysfunction of the intestinalmicrobiome in inflammatory bowel disease and treatment Genome Biol13R79 httpdxdoiorg101186gb-2012-13-9-r79

91 Morgan XC Kabakchiev B Waldron L Tyler AD Tickle TL MilgromR Stempak JM Gevers D Xavier RJ Silverberg MS Huttenhower C2015 Associations between host gene expression the mucosal micro-

biome and clinical outcome in the pelvic pouch of patients with inflam-matory bowel disease Genome Biol 1667 httpdxdoiorg101186s13059-015-0637-x

92 Segata N Izard J Waldron L Gevers D Miropolsky L Garrett WSHuttenhower C 2011 Metagenomic biomarker discovery and explana-tion Genome Biol 12R60 httpdxdoiorg101186gb-2011-12-6-r60

Raw Milk Microbiota during Transport and Storage

JulyAugust 2016 Volume 7 Issue 4 e00836-16 reg mbioasmorg 13

on August 23 2018 by guest

httpmbioasm

orgD

ownloaded from