Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Dynamic changes in fecal bacterial microbiota of dairy cattle across the production line

Dynamic changes in fecal bacterial microbiota of dairy cattle across the production line Background: Microbiota play important roles in the gastrointestinal tract (GIT ) of dairy cattle as the communities are responsible for host health, growth, and production performance. However, a systematic characterization and comparison of microbial communities in the GIT of cattle housed in different management units on a modern dairy farm are still lacking. We used 16S rRNA gene sequencing to evaluate the fecal bacterial communities of 90 dairy cat- tle housed in 12 distinctly defined management units on a modern dairy farm. Results: We found that cattle from management units 5, 6, 8, and 9 had similar bacterial communities while the other units showed varying levels of differences. Hutch calves had a dramatically different bacterial community than adult cattle, with at least 10 genera exclusively detected in their samples but not in non-neonatal cattle. Moreover, we compared fecal bacteria of cattle from every pair of the management units and detailed the number and rela- tive abundance of the significantly differential genera. Lastly, we identified 181 pairs of strongly correlated taxa in the community, showing possible synergistic or antagonistic relationships. Conclusions: This study assesses the fecal microbiota of cattle from 12 distinctly defined management units along the production line on a California dairy farm. The results highlight the similarities and differences of fecal microbiota between cattle from each pair of the management units. Especially, the data indicate that the newborn calves host very different gut bacterial communities than non-neonatal cattle, while non-neonatal cattle adopt one of the two distinct types of gut bacterial communities with subtle differences among the management units. The gut microbial communities of dairy cattle change dramatically in bacterial abundances at different taxonomic levels along the pro - duction line. The findings provide a reference for research and practice in modern dairy farm management. Keywords: Dairy cattle, Production line, Microbiome, Gastrointestinal, 16S rRNA, Bioinformatics cells in the human body has been estimated as 10 times Background more than the number of human cells [1] or at roughly It is now well-established that microbiota, especially bac- the same order in a recent study [2], this ratio can rise terial communities, play crucial roles in the physiology to approximately 120 times in ruminants such as cattle and health of all mammals. While the number of bacterial [3]. Cattle depend on their gastrointestinal microbiota to digest and convert the plant mass that cannot be directly *Correspondence: xdli@ucdavis.edu; zcsu@uncc.edu digested into absorbable nutrients necessary for host Department of Bioinformatics and Genomics, the University of North health and development. Thus, a better understanding of Carolina at Charlotte, Charlotte, North Carolina, USA the structure of the gastrointestinal microbiota is instru- Department of Population Health and Reproduction, School mental for both production and scientific inquiry. Par - of Veterinary Medicine, University of California, Davis, CA, USA Full list of author information is available at the end of the article ticularly, in the modern dairy system, calves, heifers, and © The Author(s) 2022. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. The Creative Commons Public Domain Dedication waiver (http:// creat iveco mmons. org/ publi cdoma in/ zero/1. 0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Zhao et al. BMC Microbiology (2022) 22:132 Page 2 of 15 cows at different ages and production stages are raised dairy herd in California starting from the newborn to late and managed in independent yet inter-connected man- lactation and dry cows using 16S rRNA gene sequencing agement units with housing and dietary differences [4]. data. Specifically, male calves are commonly sold at birth while female calves are raised on the source dairy from birth Results to weaning (approximately 70 days of age) in group pens Composition of the bacterial communities in the fecal or more commonly in individual hutch units that may samples be wooden, metal, or plastic of combination of materials We analyzed a total of 90 fecal samples collected on the and where calves are fed milk two to three times a day same day from 90 different dairy cattle in 12 management and offered solid grain mix ad  libitum with the goal of units (Fig.  1, Table  1) in a dairy herd. We generated a transitioning to a solid diet by the age of weaning. Once total of 6,092,309 16S rRNA gene paired-end sequencing weaned, growing calves are moved to group pens where reads with an average of 67,692 ± 7963 reads per sample. they are fed a roughage diet with concentrate formu- Clustering sequence reads from V3/V4 regions of 16S lated to their nutrient requirements until breeding age. rRNA genes into different Operational Taxonomy Units Once bred, pregnant heifers are moved to maternity pens (OTUs) has been a widely adopted strategy. However, closer to their calving date. After calving, the newborn the traditional OTU methods obscure similar sequences calf is fed colostrum and raised as described earlier. Post- by grouping them into a consensus sequence and fail to partum cows are moved to the fresh cow milking pen tell the technical errors from real biological variations before being moved to subsequent milking pens depend- [14]. u Th s, we used a more recent method DADA2 that ing on the dairy’s management practices with respect to records how many times an exact sequence was read breeding (not pregnant and pregnant pens) and level of and infers sequence variants by a statistical model learn- milk production (high and low milking pens). Late preg- ing the sequencing error rate from the samples them- nant cows around 60 days prior to calving are dried off, an selves [14]. Given the common usage of the term, we use industry practice of cessation of milking to allow the dam OTU to represent the inferred sequence variant instead to replenish her body resources and initiate colostrogen- of Amplicon Sequence Variant (ASV) that the DADA2 esis in preparation for giving birth to a calf. Adult cows authors proposed [15]. After quality filtering and chimera at different stages are commonly fed a total mixed ration removal with DADA2 [14], a total of 5,080,044 sequence formulated to meet the specific stage of lactation nutrient reads with an average of 56,444 ± 6633 reads per sample requirements. Given such differences in management, were preserved. OTUs that were taxonomically classified housing, and diets, it is important to understand the as “Archaea/Eukaryote” or appeared in less than 9 sam- dynamics of the gut microbiota changes between/among ples (10% of sample size) were excluded. Although the different management units over the entire production number of sequencing reads varied in our samples, rar- life cycle of dairy cattle. The knowledge gained can help efaction curves showed that our sampling was sufficient design new strategies to improve production as well as to represent the bacterial communities as the curve had the health of both the animals and humans who consume almost plateaued or started to form a plateau (Addi- the produced meat and milk. tional  file  1: Fig. S1). The OTU table was normalized by Since traditional culture-dependent techniques can rarefying with a threshold of 12,637 sequences, which is only recover a small portion of the microbial population the minimum number of reads of all the 90 samples after [5], several research groups have explored the gut micro- processing. All downstream analyses were based on the biota of dairy cattle using the next-generation sequenc- normalized OTU table (Additional file  2: Supplementary ing (NGS) technologies [3, 6–13]. For instance, Mao file 1). et al. analyzed the microbiota of ten gastrointestinal sites We identified a total of 4681 OTUs in the 90 fecal in Holstein cattle using 16S rRNA gene sequencing and samples. These OTUs could be taxonomically assigned found that these cattle hosted microbiota with signifi - to 20 phyla, of which Firmicutes (2573/55.0%), Bac- cant spatial heterogeneity [7]. Dill-McFarland et  al. ana- teroidetes (1291/27.6%), and Tenericutes (267/5.7%) lyzed the succession pattern of bacterial communities in included the highest number of the OTUs (Fig.  2A). dairy cattle from 2-week-old to first lactation [9]. Shanks While only Firmicutes (587,075/51.6% reads), Bacteroi- et al. profiled the structure of fecal bacteria in cattle from detes (412,852/36.3% reads), Patescibacteria (53,402/4.7% various feeding operations [12]. However, a systematic reads), and Proteobacteria (16,379/1.4% reads) were contrast of the gut bacterial microbiota in dairy cattle observed in all 90 samples, they accounted for 94.1% of along all the management units on a modern farm is still the total bacterial communities at the phylum level. Fir- absent. To fill these gaps, we analyzed the gut bacterial micutes had the highest average relative abundance in communities of dairy cattle in management units of a the samples from all the management units except hutch Zhao  et al. BMC Microbiology (2022) 22:132 Page 3 of 15 Fig. 1 Management units and dairy cattle life cycle on a dairy farm. A Schematic diagram of the independent yet inter-connected management units of the dairy herd where the samples were collected. The number around each box represents the management unit ID used in this study. B Dairy cattle life cycle by age and production stage calves, where Bacteroidetes had the highest average rela- at this taxonomic level. Ruminococcaceae (397,075/34.9% tive abundance of 62.3% and Firmicutes only accounted reads) was the most dominant taxon in average rela- for 33.4% (Fig. 2B). tive abundance at this level, followed by Rikenellaceae At the family level, these OTUs could be assigned to (131,673/11.6% reads), Prevotellaceae (96,322/8.5% 99 families, in which Ruminococcaceae (1441/34.9%), reads), Lachnospiraceae (77,143/6.8% reads), Murib- Rikenellaceae (500/11.6%), Lachnospiraceae (319/6.8%), aculaceae (70,454/6.2% reads) and Saccharimonadaceae and Prevotellaceae (273/8.5%) contained the highest (53,121/4.7% reads) (Fig. 2D). number of the OTUs (Fig.  2C). Although only 12 fami- At the genus level, 2169 (46.3%) of the 4681 OTUs lies were detected in all 90 samples, they accounted for could be assigned to 167 known genera, while the 992,097 (87.2%) reads of the total bacterial communities remaining 2512 OTUs (53.7%) could not be classified Zhao et al. BMC Microbiology (2022) 22:132 Page 4 of 15 Table 1 Classification of cattle based on the growth, production stages, and their residential management units of the dairy herd where samples were collected Management Unit No. of Unit ID Description Samples Hutch calves 9 1 From birth to approximately 1–2 weeks after weaning, individually housed (1 to 70 days age approximately) Post weaned heifers 6 2 Group-housed heifers and bull calves (if not sold) fed solid diet Breeding heifers 8 3 Approximately 13 to 15 months old Springer 8 4 Within 1 to 4 weeks from calving, pregnant nulliparous heifers Fresh uniparous cows 8 5 1 to 2 months post-calving, first lactation cows Mid-lactation uniparous cows 8 6 60 to 250 Days in Milk, first lactation Fresh multiparous cows 8 7 1 to 2 months post-calving, second lactation or greater Mid-lactation multiparous cows 8 8 60 to 250 Days in Milk, second lactation or greater Pregnant late 8 9 > 250 Days in Milk lactation cows Far-off dry cows 3 10 21 to 60 days prior to calving, multiparous dry cows Close up dry cows 8 11 within 21 days prior to calving, multiparous dry cows Hospital pen/ 8 12 Lactating cows treated with medication that requires milk withdrawal. May include fresh cows Fresh (post-calving) (post-calving) during their transition from producing colostrum to milk and/or pending milk withdrawal after treatment at drying (prior to entering Far-off dry pen). to known genera, thus annotated by 70 lowest known and B, the fecal bacterial communities in pre-weaned taxonomic ranks (family, order, etc.). These unclassified calves (unit 1) fed primarily milk and housed in indi- taxa might be novel bacteria in the cattle feces or not vidual hutches (Table  1) had significantly lower Chao 1 differentiable solely based on the hypervariable regions and Shannon indexes than those in all other units, except of 16S rRNA genes. Overall, the largest number of the for the post-weaned group-housed calves fed a solid diet OTUs were assigned to [f ]Ruminococcaceae (588/11.9% (unit 2) (Kruskal Wallis, false discovery rate, FDR < 0.05), OTUs), followed by Rikenellaceae_RC9_gut_group suggesting that the hutch calves generally had simpler (288/6.15% OTUs), and Ruminococcaceae_UCG-010 fecal bacterial communities. Post weaned heifers (unit 2), (250/5.3% OTUs) (Fig.  2E). While only 17 genera were while having no significant differences in bacterial com - consistently detected in the 90 samples, these commonly munities compared with hutch calves (unit 1), consist- shared genera occupied 72.8% of the total bacterial com- ently differed from breeding heifers (unit 3), springers munities at this taxonomic level. Most of the assigned (unit 4), and pregnant late lactation cows (unit 9), which genera had an average abundance of < 2% in the sam- might be related to the development of the rumen and ples from the management units, while only 15 genera these growing young cattle being full ruminants by the had a relative abundance ≥ 2% including [f] Rumino- time they were bred. coccaceae (11.5%), Ruminococcaceae_UCG-005 (8.8%), Rikenellaceae_RC9_gut_group (8.2%), [f ]Muribaculaceae Similarity and difference of the fecal bacterial communities (6.2%), [f]Lachnospiraceae (5.2%), Candidatus_Saccha- between management units rimonas (4.7%), [o] Bacteroidales (4.3%), Bacteroides To further evaluate the similarities and differences of (3.6%), Ruminococcaceae_UCG-014 (3.5%), Ruminococ- the bacterial communities between different manage- caceae_UCG-013 (3.4%), Christensenellaceae_R-7_group ment units, we calculated the β-diversity (Bray-Curtis (3.0%), Ruminococcaceae_UCG-010 (2.6%), Prevotel- distance) of the samples and visualized the results laceae_UCG-003 (2.5%), [f]Prevotellaceae (2.2%), and using non-metric multidimensional scaling (NMDS). Alistipes (2.2%) (Fig. 2F). As shown in Fig. 4A, the hutch calves’ samples (unit 1) were largely grouped to form a cluster (I), which sepa- Alpha diversity of the cattle fecal bacterial communities rated from the samples from the other management in the management units units, with two post-weaned heifers’ (unit 2) samples We next compared the richness and evenness of the fecal included, suggesting that the fecal bacterial communi- bacterial communities of cattle in different management ties of hutch calves (unit 1) are similar to one another units in the production line using both Chao 1 richness but largely different from those of cattle in the remain- index and Shannon diversity index. As shown in Fig.  3A ing management units. Interestingly, the samples from Zhao  et al. BMC Microbiology (2022) 22:132 Page 5 of 15 Fig. 2 Assignment of the OTUs to different taxonomic levels. A The number of OTUs assigned to different phyla. B Average relative abundances of the phyla in the samples from different management units. C The number of OTUs assigned to different families. D Average relative abundances of the families in the samples from different management units. E The number of OTUs assigned to different genera. F Average relative abundances of the genera in the samples from different management units. In B, D, and F, the length of color-coded bars represents the average relative abundance of the taxa in the samples in the indicated management units. Taxa with < 2% relative abundance in all the units were merged into the “Other/Unassigned” category Zhao et al. BMC Microbiology (2022) 22:132 Page 6 of 15 Fig. 3 Diversity of fecal bacterial communities of cattle in the management units. Boxplots of Chao 1 richness index (A) and Shannon diversity index (B) of fecal bacterial communities of cattle in the management units (x-axis). Each violet-red dot represents the bacterial communities in a fecal sample, and management units are coded by distinct colors. The numbers above each box indicate the management units that are significantly different from the current one with the pairwise Wilcoxon test (P < 0.05) the other units form two rather compact yet distinct significant differences between the two clusters clusters II and III (Fig.  4A), suggesting the samples in (Additional file 3: Fig. S2). each cluster have quite similar bacterial communities. Cluster II contains samples from all the management Shared and unique taxa at the genus level units except unit 1; in contrast, cluster III comprises in the management units samples only from units 3, 5, 6, 7, 8, 11, and 12, sug- For the 237 taxa at the genus level (167 assigned to gesting that non-neonatal cattle (from unit 2 to 12) known genera and 70 assigned to higher taxonomic may only have one of two types of rather uniform bac- ranks), 112 of them had the average relative abundance of terial structures. 0.1% or greater in at least one management unit. Among Moreover, we found that at the phylum level, the these 112 genera, 70 were observed in all management relative abundances of several phyla were signifi- units while the remaining 42 genera existed in a varying cantly different between the samples in clusters number of management units. Particularly, there were II and III (Fig.  4B-E, t-test, P < 0.05). The sample s several genera largely detected only in management unit in cluster II were high in Bacteroidetes (Fig.  4B), 1but barely seen in other units, suggesting these taxa while samples in cluster III were high in Fir- may play important roles in the early stage of life for micutes (Fig.  4C), Patescibacteria (Fig.  4D), and the cattle. For  example, 10 genera, as shown in Table  2 Actinobacteria (Fig.  4E). There were also five addi- listed by the relative abundance from high to low, were tional phyla with low relative abundances showing observed majorly in unit 1 but sparsely seen in other units. In contrast, we found Prevotellaceae_UCG-004, (See figure on next page.) Fig. 4 Similarity and difference of the fecal bacterial communities. A Non-metric multidimensional scaling (NMDS) plot of the Bray-Curtis distance for bacterial communities in different management units. Each point represents a fecal bacterial community and is colored by the management unit from which it was sampled. The communities are grouped into three clusters (I, II, and III) circled in blue, black, and red rings, respectively. Boxplot of four phyla’s relative abundances in clusters II and III for B Bacteroidetes C Firmicutes D Patescibacteria and E Actinobacteria Zhao  et al. BMC Microbiology (2022) 22:132 Page 7 of 15 Fig. 4 (See legend on previous page.) Zhao et al. BMC Microbiology (2022) 22:132 Page 8 of 15 Table 2 Number of occurrences and taxonomic assignment at the genus level for taxa that were either majorly detected in unit 1 or missing in unit 1 but observed in other units Avg. No. of Reads Avg. No. of Sum of Avg. No. of Reads Phylum Family Genus in Unit 1 Reads in Unit 2 in other 10 Units Genera majorly observed in unit 1 330 (2.6%) 4 2 Bacteroidetes Prevotellaceae Prevotella_9 157 (1.2%) 0 0 Bacteroidetes Prevotellaceae Prevotella_2 102 (0.8%) 0 3 Bacteroidetes Prevotellaceae Prevotella 97 (0.8%) 10 5 Clostridiales Ruminococcaceae Faecalibacterium 87 (0.7%) 0 1 Proteobacteria Burkholderiaceae Sutterella 86 (0.7%) 9 0 Bacteroidetes Tannerellaceae Parabacteroides 74 (0.6%) 2 0 Fusobacteria Fusobacteriaceae Fusobacterium 72 (0.6%) 17 0 Firmicutes Ruminococcaceae Subdoligranulum 54 (0.4%) 4 0 Firmicutes Ruminococcaceae Oscillospira 86 (0.7%) 9 0 Bacteroidetes Tannerellaceae Parabacteroides 15 (0.1%) 0 6 Erysipelotrichales Erysipelotrichaceae Erysipelotrichaceae_UCG-004 Genera missing in unit 1 but observed in other units 0 31 1581 Bacteroidales Prevotellaceae Prevotellaceae_UCG-004 0 1 1196 Bacteroidales p-2534-18B5_gut_group Unassigned, annotated as [f ]p-2534-18B5_gut_group 0 16 905 Bacteroidales Rikenellaceae dgA-11_gut_group 0 2 401 Clostridiales Ruminococcaceae Caproiciproducens dgA-11_gut_group, Caproiciproducens, and an unas- potential interactions between different bacteria, we cal - signed genus ([f ]p-2534-18B5_gut_group) missing in ani- culated the correlations of the core bacteria using the mals from unit 1, but in all other 11 management units SparCC (Sparse Correlations for Compositional data) (Table 2). algorithm [16]. By setting ±0.7 as the threshold values for strong positive and negative correlations and pseudo Dynamics of bacterial communities p-values < 0.05 as the significant correlations, we found between the management units 61 OTUs that had strong correlations with at least one From the core bacteria, the 112 genera with an average other OTU (Fig.  6; Additional  file  4: Supplementary minimal abundance of 0.1% in at least one management file  2; Additional  file  5: Supplementary file  3). Most of unit), we found 56 of them displayed significantly differ - these strong correlations are positive, while only one pair ent abundances in the samples from the 12 management between Ruminococcaceae_UCG-005 and Bacteroides is units (ANOVA, FDR < 0.05). We further summarized negative (SparCC = − 0.70). In Fig.  6, all the strong cor- how each of these 56 genera was significantly different relations are shown within the networks. From the 181 in their composition between different pairs of man - strong correlations, 117 are between the OTUs assigned agement units (Tukey HSD, P < 0.05, Fig.  5A and B). For to the same phylum, indicating that strong correlations example, unit 1 had a varying number of significantly dif - are more likely to be within the same phylum than those ferent genera with all other 11 units, from 11 with unit 5 from different phyla. to 27 with unit 7 (Fig. 5A). As shown in Fig. 5B, the rela- tive abundances of these significantly different genera Functional prediction of the bacterial community between unit 1 and other units were in the range of 25.9 To further explore the fecal bacteria of the dairy cat- and 52.6%. Interestingly, there is no significantly different tle, we predicted the functions of the bacterial com- genus between any pair of management units 5, 6, 8, and munity using a marker gene based tool Tax4Fun2 [17]. 9, suggesting the types and relative abundances of the Compared to the initial version Tax4Fun, Tax4Fun2 has genera were similar in these four units. higher accuracy, allows users to customize the reference genomes, and integrates a multifunctional redundancy Synergistic and antagonistic relationships evaluation [17, 18]. Bacteria thriving in a community may interact with one A total of 7726 Kyoto Encyclopedia of Genes and another synergistically or antagonistically. To reveal such Genomes (KEGG) Orthologs (KOs) were predicted in Zhao  et al. BMC Microbiology (2022) 22:132 Page 9 of 15 Fig. 5 Significantly different taxa at genus level between all pairs of management units. A Number of genera significantly different between each pair of the management units. B Relative abundance of significantly different genera between each pair of the management units Zhao et al. BMC Microbiology (2022) 22:132 Page 10 of 15 Fig. 6 Strong correlations in the core bacteria. Each node is an OTU and the number within represents the OTU ID. Nodes are colored based on their taxonomic assignment at the phylum level. The edges in green and red indicate positive and negative correlations the identified bacterial microbiota in the samples while 20 of them were shown in Fig. 7C. Again, energy metab- no single KO had a relative abundance higher than 2% olism related pathways are most abundant, such as gly- in any sample (Additional  file  6: Supplementary file  4). colysis/gluconeogenesis, carbon fixation, and methane Among them, 4001 (51.8%) KOs were present in all 90 metabolism. samples, indicating these molecular functions were essential for the cattle regardless of age, housing, and Discussion production status such as K00001 (alcohol dehydroge- In modern dairy management, it is not uncommon to nase) that mediates the oxidation and reduction of eth- house animals in different management units based on anol [19]. Interestingly, some KOs were only present in their ages, nutrition requirements, reproductive status, the samples from one management unit. For instance, a lactation status, production, and management styles [21]. collection of 9 KOs (K04340, K14194, K16046, K16227, Such a management system optimizes milk production K18254, K18611, K18906, K20218, and K21329) were and management but brings challenges profiling and ana - only observed in some samples of unit 1, though at lyzing the bacterial communities in the GIT because the very low abundances (Additional file  6: Supplementary bacterial communities can be influenced by numerous file  4). These KOs might be involved in the food diges - factors including, but not limited to, diet, animal physiol- tion of hutch calves. For instance, K18611 (4-pyridoxic ogy and status, the farm environment, geographic loca- dehydrogenase) is an enzyme that degrades vitamin B6, tion, antimicrobial use, and management practices of the which is contained in milk, the primary diet for hutch production lifecycle [6, 12, 22–25]. For instance, the bac- calves housed in unit 1 [20]. There were 982 (12.7%) sig - terial communities of cattle housed individually were less nificantly differential KOs (ANOVA, FDR < 0.05) and the likely to be affected by other cattle compared with those top 20 most abundant ones were shown in Fig. 7A. housed together in open pens. Another example is that From the pathway perspective, Tax4Fun2 predicted 6, cattle with diseases may be treated with antimicrobial 45, and 239 pathways at KEGG level 3, 2, and 1, respec- drugs that may directly change the balance of the bacte- tively. Metabolism (70.4–72.0%) was the most abundant rial communities. pathway at level 3 in every management unit, followed In this study, we sought to characterize the composi- by Environmental Information Processing (12.0–13.2%), tion, diversity, and dynamics of the fecal bacteria in dif- Cellular Processes (6.1–6.7%), Genetic Information Pro- ferent management units over the production lifecycle of cessing (5.3–5.7%), Human Diseases (2.8–3.1%), and dairy cattle on a California dairy farm. We used paired- Organismal Systems (1.2–1.3%, Additional  file  7: Sup- end sequencing of the V3/V4 hypervariable regions of plementary file  5). Further analysis revealed 10 signifi - the 16S rRNA genes to profile the bacterial communi - cantly differential pathways at level 2, with the energy ties in 90 fecal samples collected from 90 cattle in the 12 metabolism pathway being the most abundant (Fig. 7B). management units. Our results suggest high similarity However, the abundances of these pathways combined between any two of the management units 5, 6, 8, and were only 6.6% on average. At level 1, there were 68 sig- 9, but show significant differences in most of the other nificantly differential pathways, and the most abundant management unit pairs in terms of richness, evenness, Zhao  et al. BMC Microbiology (2022) 22:132 Page 11 of 15 Fig. 7 Significantly differential orthologs and pathways in the KEGG database predicted using Tax4Fun2. A Top 20 the most abundant significantly differential KOs. B Significantly differential level 2 pathways. C Top 20 the most abundant significantly differential level 1 pathways. Rows represent KOs/pathways and columns represent management units. Subplots were arranged by the relative abundance from high to low Zhao et al. BMC Microbiology (2022) 22:132 Page 12 of 15 and structure of the gut bacterial communities at the 2 months (70 days on average) to 13-month old, the genus level. microbiota changes were still ongoing towards the bacte- Overall, consistent with several earlier studies [6, 7, 12, rial profiles in adults during this transition period. Using 13, 22], our results reveal Firmicutes and Bacteroidetes the same set of samples, we have previously found that were the two most dominant taxa at the phylum level, E. coli from hutch calves exhibited a wider spectrum of distantly followed by several other phyla. Specifically, our resistance to antimicrobial drugs compared to bacteria results show that, on average, the Firmicutes and Bacte- from other units [21]. Currently, it largely remains unde- roidetes represented 51.6 and 36.3% of the total commu- termined concerning the roles of bacterial communities nities, respectively. Interestingly, the relative abundance on antimicrobial resistance of specific bacterial species, of Bacteroidetes was lower in cows immediately before however, in future studies, it will be interesting to assess and after calving (within two weeks from calving) in units such roles, for example, resistomes on phenotypes of with fresh multiparous cows (18.7%) and close up dry antimicrobial resistance. cows (23.2%), while the relative abundances were in the For the samples from non-neonatal cattle (from unit 2 range of 29.4 and 62.2% in other units (Fig. 2B). to unit 12), we did not find any evidence supporting that For some genera, as Mao et al. [7] demonstrated, Prevo- the bacterial communities could be clustered based on tella had a lower abundance in samples taken from the the management unit membership. This demonstrated small and large intestines, but higher in the forestomach, that the management unit itself is not a determinate force such as the rumen. In our work, the relative abundance of for the structure of the bacterial communities. Instead, as Prevotella was only 0.07%. However, in one of the hutch shown in Fig. 4A, we did observe distinct patterns where calves’ (unit 1) samples, we observed Prevotella had a most of our samples from non-neonatal cattle formed relative abundance of 7.16% (905 reads) while the num- two clusters (II and III), implying they had two major dif- ber of Prevotella reads in the other 89 samples was in ferent structures. As we looked deeper, we found that at the range of 0 to 11. We also found in this sample, Rumi- the phylum level, the relative abundances of several phyla nococcus_1 had a relatively high abundance (5.56%, 703 became very different between the two clusters (Fig.  4B- reads) while this genus was in the range of 0 to 115 reads E). For example. in cluster III, the mean and median of in other samples. So potentially there is some synergistic Patescibacteria relative abundance were 20.9 and 19.7% relationship between Prevotella and Ruminococcus_1 and while those values in cluster II were only 1.9 and 1.1%, the interaction prevent Prevotella from being degraded respectively. What caused the differences in Patescibac - or eliminated when going through the GIT. Such interac- teria in various samples still needs future research, but tion could also be related to the diet of hutch calves being the potential interactions between some taxa classified as primarily milk compared to all the other units fed a solid Firmicutes and Patescibacteria seem to prevent Patesci- diet. Moreover, feeding waste milk from the hospital pen bacteria from decreasing. It has been reported that the or fresh cow pen to calves is a common practice in the rumen bacteria in the 1st and 2nd lactations, which over- dairy industry and was done on this dairy at the time lapped with some of our management units, are dynamic of sampling. Waste milk commonly contains antibiotic yet similar, and the samples from the two lactations can- residues of varying concentrations as it is collected from not be separated by the lactation cycle in PCA visualiza- milking cows in the hospital being treated with antibiot- tion [8]. This demonstrated that the bacterial structure ics or awaiting clearance during their withdrawal period might undergo further shaping going through the GIT before rejoining the pens where milk is harvested for (their rumen samples versus our fecal samples) or the human consumption. formed patterns were just a case-sensitive phenomenon. It is noticeable that bacterial communities in hutch Pitta et  al. reported that significant bacterial population calves were significantly different from those in other change was observed during the transition from 21 days management units except post-weaned heifers (Figs.  2, before calving to 21 days after calving in uniparous and 3, 4A, 5). These differences confirmed the earlier find - multiparous cows, respectively [26]. This shift period cor - ings that the enteric microbiota in neo-natal calves were responded approximately to our management units 4 and different from adult cows and the bacterial communities 5 for uniparous cows and units 10 and 11 for multiparous underwent a dramatic change during the development cows. However, we did not see this significant shift as dem - at an early age [9, 10, 23]. Intriguingly, bacteria from the onstrated earlier [26]. The only significant difference we samples in post-weaned heifers displayed similar patterns observed was Shannon diversity between units 4 and 5. This with the communities from the adult cattle, predomi- could be due to the different sites of the GIT from which nately Firmicutes; but at the same time, they showed the samples were taken (their rumen samples versus our no significant difference from those of hutch calves. As fecal samples). It is also possible that this was due to the four the post weaned heifers were aged from approximately management units in our study not being strictly narrowed Zhao  et al. BMC Microbiology (2022) 22:132 Page 13 of 15 to 21 days prior and post-calving. As the dairy farm environ- sampling population in each management unit can also ment is dynamic, it is not surprising that most taxa were be found at Li et  al. [21]. These samples were also used shared by cattle housed in different management units. for an antimicrobial resistance study that has been pub- Measuring potential bacterial interactions was chal- lished by Li et al. [21]. An aliquot of each sample shipped lenging and barely reported in dairy cattle studies. The in refrigerated conditions to Dr. Su’s laboratory at the classic correlation methods had their limitations when Department of Bioinformatics and Genomics, the Uni- applied to genomic data such as 16S rRNA gene sequenc- versity of North Carolina at Charlotte for 16S rRNA gene ing data which are sparse and compositional. SparCC as a sequencing. method that negates the negative correlation bias of com- positional data [27] and identifies true association missed by others [28], was used here as a way of evaluating bac- Illumina MiSeq sequencing terial interactions. In general, we identified 180 strong DNA from stool samples was extracted with the Qiagen positive correlations and 1 strong negative correlation DNA Stool kit following the manufacturer’s instructions. between 61 OTUs (Fig.  6). These co-occurrent OTUs Two steps of Polymerase Chain Reaction (PCR) pro- may tend to share the same habitats and perform similar cedures were used to generate amplicons from the 16S functions, as most of these co-occurrent patterns spotted RNA genes for sequencing. The first-round PCR was to between OTUs were in the same phyla (Fig. 6). target V3/V4 regions of 16S rRNA genes with the for- ward primer: 5′- CCT ACG GGNGGC WGC AG and the reverse primer: 5′- GAC TAC HVGGG TAT CTA ATC C. Conclusions This step was done with the KAPA Biosciences HiFi PCR In this study, we profiled the structure and dynamics of kit and additional BSA. The protocol consists of initial gut bacterial communities from cattle in 12 independ- denaturation at 95 °C for 3 min, followed by 25 cycles of ent yet inter-connected management units on a mod- denaturation (90 °C for 30 s, 55 °C for 30 s, and 72 °C for ern California dairy farm. To the best of our knowledge, 30 s), and final elongation at 72 °C for 5 min. The PCR this is the first study that describes the bacterial com - products were cleaned up with Ampure XP beads. The munities across all management units and reveals the second-round PCR was performed with Nextera XT structures of gut microbial communities to each of index Primers and sequencing Adaptors with the follow- the well-defined management units. We analyzed the ing setting: initial denaturation at 95 °C for 3 min, fol- changes in gut microbial communities across production lowed by 8 cycles of denaturation (90 °C for 30 s, 55 °C for systems. Though the fecal microbiota were similar in 4 of 30 s, and 72 °C for 30 s), and final elongation at 72 °C for the management units, they showed significant changes 5 min. The PCR products were cleaned up with Ampure between others. It is confirmed that microbial ecology XP beads and paired-end sequenced (2 × 300 bp) on an underwent dramatic changes in the early days of life, as Illumina MiSeq platform at the University of North Car- evidenced by the significantly different bacteria in hutch olina at Charlotte. calves from other adult cattle and revealed dynamics of the bacterial abundance in the later stages of the produc- tion lifecycle. Moreover, we identified at least 10 genera OTU table construction that were detected only in hutch calves but were absent Primers with raw sequences were removed by Cutadapt in all the other cattle in other units. These genera might [29]. We performed quality control using DADA2’s “fil - play crucial roles in the early establishment and develop- terAndTrim” function with “trancLen” equal to 200 bp ment of the GIT. We also dissected potential interactions for forward reads and 150 bp for reverse reads based on among gut bacterial groups, mostly from the species in quality profiles. Technical error rate learning was per - the same phylum. formed with all the sequences in the samples. Sample inference was performed by the “dada” function with the Methods setting optional parameter “pool = TRUE. ” Paired-end Sample collection and study herd reads merger in DADA2 resulted in approximately 50% On a single day in June 2016, a total of 90 fecal samples loss of sequences, thus only forward reads were used in were collected from 90 individual cattle in 12 manage- this study. OTU table and chimera removal were imple- ment units (Fig. 1, Table 1) from a dairy herd in the Cen- mented with default parameters. We used the “assignTax- tral Valley of California, USA. Cattle in each of these onomy” function that provided a native implementation management units were identified based on convenience of the RDP Classifier [30] with minimum bootstrap con - sampling. Trained study personnel collected the fecal fidence of 80 to assign taxonomy from the phylum level samples manually from the rectum of cattle using stand- ard veterinary protocols. Other information such as the Zhao et al. BMC Microbiology (2022) 22:132 Page 14 of 15 to the genus level to each OTU. SILVA database release Additional file 5: Supplementary file 3. Taxonomic table. Table of taxo - 132 [31] was used as the reference database. nomic classification of 61 strongly correlated OTUs. Additional file 6: Supplementary file 4. The table of function predic- tions generated by the Tax4Fun2 program. Function prediction Molecular function prediction was done using Tax- Additional file 7: Supplementary file 5. The table of pathway predic- tions generated by the Tax4Fun2 program. 4Fun2 with the default reference library. The functional profiles were generated with the built-in functions Acknowledgments “runRefBlast” and “makeFunctionalPrediction” with Not applicable. default settings. Authors’ contributions Z.S. and X.L. designed and conceived the experiments. Sample and animal measurements controlled and instructed by X.L., S.S.A., and D.R.W. L.Z. per- Statistical analysis formed microbial data processing and analysis. L.Z., Z.S., and X.L. interpreted Rarefying was performed by the “single_rarefaction” the results. L.Z. and Z.S. wrote the manuscript. S.S.A. prepared the artwork of the dairy farm. X.L., E.R.A., S.S.A., and D.R.W. contributed to the revision of the function with the minimum number of reads of all sam- manuscript. The author(s) read and approved the final manuscript. ples, which is equal to 12,637 in QIIME [32]. We used alpha diversity, including Chao 1 index which evaluates Funding This study was partially supported by the National Institute of Food and Agri- richness, and Shannon index which evaluates diversity culture (NIFA)‘s Exploratory Research program (2015–67030-23892), the NIFA’s to measure the within-sample diversity. Differences in Dairy Herd Health and Food Safety Formula Funds (CALV-DHHFS-0053) from Chao 1 index and Shannon index in different units were the USDA, and National Institutes of Health (R01GM106013). assessed by the Kruskal Wallis test with the Benjamini- Availability of data and materials Hochberg (BH) correction for multi-comparisons, and All DNA sequences have been deposited in NCBI’s Sequence Read Archive pairwise management units’ comparisons were per- (SRA) with the accession number PRJNA607283. Direct link: https:// www. ncbi. nlm. nih. gov/ sra/? term= PRJNA 607283 formed by pairwise Wilcoxon test. The Bray-Curtis dis - tance matrix was employed to perform beta diversity Declarations analysis. Visualization was done by a non-metric mul- tidimensional scaling plot. SparCC correlations were Ethics approval and consent to participate calculated by FastSpar [33], a C++ implementation of Sampling was approved by the Institutional Animal Care and Use Commit- tee (IACUC) of University of California Davis (protocol number 18941) and SparCC algorithm (100 bootstrap samples were gener- all experiments were performed in accordance with relevant guidelines and ated for pseudo p-value calculation). R software [34] regulations. and R packages ggplot2 [35], vegan [36], superheat [37], Consent for publication ggpubr [38], and qgraph [39], were used for calculation Not applicable. and visualization. Competing interests The authors declare that they have no competing interests. Abbreviations GIT: Gastrointestinal tract; NGS: Next-generation sequencing; OTU: Operational Author details taxonomy unit; NMDS: Non-metric multidimensional scaling; SparCC: Sparse Department of Bioinformatics and Genomics, the University of North Carolina correlations for compositional data; KEGG: Kyoto encyclopedia of genes and at Charlotte, Charlotte, North Carolina, USA. Western Institute for Food Safety genomes; KO: KEGG ortholog. and Security, University of California, Davis, California, USA. Depar tment of Population Health and Reproduction, School of Veterinary Medicine, Univer- sity of California, Davis, CA, USA. Veterinary Medicine Teaching and Research Supplementary Information Center, School of Veterinary Medicine, University of California, Davis, CA, USA. The online version contains supplementary material available at https:// doi. org/ 10. 1186/ s12866- 022- 02549-3. Received: 28 September 2021 Accepted: 11 April 2022 Additional file 1: Figure S1. Rarefaction curves showing “Number of Observed OTUs” as a function of “Number of Sequencing Reads” for each sample. References Additional file 2: Supplementary file 1. Normalized OTU table. The table 1. Reid A, Greene S. Faq: human microbiome. Am Acad Microbiol. 2014. includes the information of OTU ID, abundance, and taxonomic assign- https:// asm. org/ Repor ts/ FAQ- Human- Micro biome. ment for each OTU. 2. Sender R, Fuchs S, Milo R. Revised estimates for the number of human Additional file 3: Figure S2. Boxplot of five additional phyla with low and Bacteria cells in the body. PLoS Biol. 2016;14(8):e1002533. https:// doi. relative abundances showing significant difference between clusters II org/ 10. 1371/ journ al. pbio. 10025 33. and III in Fig. 4A. 3. Ross EM, Moate PJ, Marett LC, Cocks BG, Hayes BJ. Metagenomic predic- tions: from microbiome to complex health and environmental pheno- Additional file 4: Supplementary file 2. SparCC table. Table of calculated types in humans and cattle. PLoS One. 2013;8(9):e73056. https:// doi. org/ SparCC correlations and pseudo p-values for 61 OTUs that had strong cor- 10. 1371/ journ al. pone. 00730 56. relations with at least one other OTU. Zhao  et al. BMC Microbiology (2022) 22:132 Page 15 of 15 4. Konboon M, Bani-Yaghoub M, Pithua PO, Rhee N, Aly SS. A nested 25. Smith G. Antimicrobial decision making for enteric diseases of cattle. Vet compartmental model to assess the efficacy of paratuberculosis control Clin N Am Food Anim Pract. 2015;31(1):47–60. https:// doi. org/ 10. 1016/j. measures on US dairy farms. PLoS One. 2018;13(10):e0203190.cvfa. 2014. 11. 004. 5. Janssen PH. Identifying the dominant soil bacterial taxa in libraries of 16S 26. Pitta DW, Indugu N, Kumar S, Vecchiarelli B, Sinha R, Baker LD, et al. rRNA and 16S rRNA genes. Appl Environ Microbiol. 2006;72(3):1719–28. Metagenomic assessment of the functional potential of the rumen https:// doi. org/ 10. 1128/ AEM. 72.3. 1719- 1728. 2006. microbiome in Holstein dairy cows. Anaerobe. 2016;38:50–60. https:// doi. 6. Tang MT, Han H, Yu Z, Tsuruta T, Nishino N. Variability, stability, and resil-org/ 10. 1016/j. anaer obe. 2015. 12. 003. ience of fecal microbiota in dairy cows fed whole crop corn silage. Appl 27. Gloor GB, Macklaim JM, Pawlowsky-Glahn V, Egozcue JJ. Microbiome Microbiol Biotechnol. 2017;101(16):6355–64. https:// doi. org/ 10. 1007/ datasets are compositional: and this is not optional. Front Microbiol. s00253- 017- 8348-8. 2017;8:2224. https:// doi. org/ 10. 3389/ fmicb. 2017. 02224. 7. Mao S, Zhang M, Liu J, Zhu W. Characterising the bacterial microbiota 28. Tsilimigras MC, Fodor AA. Compositional data analysis of the microbiome: across the gastrointestinal tracts of dairy cattle: membership and poten- fundamentals, tools, and challenges. Ann Epidemiol. 2016;26(5):330–5. tial function. Sci Rep. 2015;5:16116. https:// doi. org/ 10. 1038/ srep1 6116.https:// doi. org/ 10. 1016/j. annep idem. 2016. 03. 002. 8. Jewell KA, McCormick CA, Odt CL, Weimer PJ, Suen G. Ruminal bacterial 29. Martin M. Cutadapt removes adapter sequences from high-throughput community composition in dairy cows is dynamic over the course of two sequencing reads. EMBnet journal. 2011;17(1):10–2. lactations and correlates with feed efficiency. Appl Environ Microbiol. 30. Wang Q, Garrity GM, Tiedje JM, Cole JR. Naive Bayesian classifier for rapid 2015;81(14):4697–710. https:// doi. org/ 10. 1128/ AEM. 00720- 15. assignment of rRNA sequences into the new bacterial taxonomy. Appl 9. Dill-McFarland K, Breaker J, Suen G. Microbial succession in the gas- Environ Microbiol. 2007;73(16):5261–7. trointestinal tract of dairy cows from 2 weeks to first lactation. Sci Rep. 31. Quast C, Pruesse E, Yilmaz P, Gerken J, Schweer T, Yarza P, et al. The SILVA 2017;7:40864. https:// doi. org/ 10. 1038/ srep4 0864. ribosomal RNA gene database project: improved data processing and 10. Dias J, Marcondes MI, de Souza SM, e Silva BCM, Noronha MF, Resende web-based tools. Nucleic Acids Res. 2012;41(D1):D590–D6. https:// doi. RT, et al. Bacterial community dynamics across the gastrointestinal tracts org/ 10. 1093/ nar/ gks12 19. of dairy calves during preweaning development. Appl Environ Microbiol. 32. Caporaso JG, Kuczynski J, Stombaugh J, Bittinger K, Bushman FD, Costello 2018;84(9):e02675–17. EK, et al. QIIME allows analysis of high-throughput community sequenc- 11. Jami E, Israel A, Kotser A, Mizrahi I. Exploring the bovine rumen bacterial ing data. Nat Methods. 2010;7(5):335. community from birth to adulthood. ISME J. 2013;7(6):1069–79. https:// 33. Watts SC, Ritchie SC, Inouye M, Holt KE. FastSpar: rapid and scalable corre- doi. org/ 10. 1038/ ismej. 2013.2. lation estimation for compositional data. Bioinformatics. 2019;35(6):1064– 12. Shanks OC, Kelty CA, Archibeque S, Jenkins M, Newton RJ, McLellan SL, 6. https:// doi. org/ 10. 1093/ bioin forma tics/ bty734. et al. Community structures of fecal bacteria in cattle from different ani- 34. R Core Team. R. A language and environment for statistical computing. mal feeding operations. Appl Environ Microbiol. 2011;77(9):2992–3001. R Foundation for statistical computing. Vienna URL https:// www.R- proje https:// doi. org/ 10. 1128/ AEM. 02988- 10. ctorg/; 2018. 13. Hagey JV, Bhatnagar S, Heguy JM, Karle BM, Price PL, Meyer D, et al. Fecal 35. Wickham H. ggplot2: elegant graphics for data analysis: Springer; 2016. microbial communities in a large representative cohort of California dairy 36. Oksanen J, Kindt R, Legendre P, O’Hara B, Stevens MHH, Oksanen MJ, et al. cows. Front Microbiol. 2019;10:1093. https:// doi. org/ 10. 3389/ fmicb. 2019. The vegan package. Commun Ecol Package. 2007;10:631–7. 01093. 37. Barter RL, Yu B. Superheat: an R package for creating beautiful and 14. Callahan BJ, McMurdie PJ, Rosen MJ, Han AW, Johnson AJ, Holmes SP. extendable heatmaps for visualizing complex data. J Computational DADA2: high-resolution sample inference from Illumina amplicon data. Graph Stat. 2018;27(4):910–22. Nat Methods. 2016;13(7):581–3. https:// doi. org/ 10. 1038/ nmeth. 3869. 38. Kassambara A. Ggpubr:“ggplot2” based publication ready plots (version 15. Callahan BJ, McMurdie PJ, Holmes SP. Exact sequence variants should 0.1. 7); 2017. replace operational taxonomic units in marker-gene data analysis. ISME J. 39. Epskamp S, Cramer AO, Waldorp LJ, Schmittmann VD, Borsboom D. 2017;11(12):2639–43. https:// doi. org/ 10. 1038/ ismej. 2017. 119. Qgraph: network visualizations of relationships in psychometric data. J 16. Friedman J, Alm EJ. Inferring correlation networks from genomic survey Stat Softw. 2012;48(4):1–18. data. PLoS Comput Biol. 2012;8(9):e1002687. https:// doi. org/ 10. 1371/ journ al. pcbi. 10026 87. Publisher’s Note 17. Wemheuer F, Taylor JA, Daniel R, Johnston E, Meinicke P, Thomas T, et al. Springer Nature remains neutral with regard to jurisdictional claims in pub- Tax4Fun2: prediction of habitat-specific functional profiles and functional lished maps and institutional affiliations. redundancy based on 16S rRNA gene sequences. Environmental Microbi- ome. 2020;15:1–12. 18. Aßhauer KP, Wemheuer B, Daniel R, Meinicke P. Tax4Fun: predicting functional profiles from metagenomic 16S rRNA data. Bioinformatics. 2015;31(17):2882–4. 19. Jörnvall H, Persson B, Jeffery J. Characteristics of alcohol/polyol dehydro - genases: the zinc-containing long-chain alcohol dehydrogenases. Eur J Biochem. 1987;167(2):195–201. 20. Kanehisa M, Goto S. KEGG: Kyoto encyclopedia of genes and genomes. Nucleic Acids Res. 2000;28(1):27–30. 21. Li X, Aly SS, Su Z, Pereira RV, Williams DR, Rossitto P, et al. Phenotypic Re Read ady y to to submit y submit your our re researc search h ? Choose BMC and benefit fr ? Choose BMC and benefit from om: : antimicrobial resistance profiles of E. coli and Enterococcus from dairy cattle in different management units on a Central California dairy. Clin fast, convenient online submission Microbiol. 2018;7:312. https:// doi. org/ 10. 4172/ 2327- 5073. 10003 12. thorough peer review by experienced researchers in your field 22. Kim M, Kim J, Kuehn L, Bono J, Berry E, Kalchayanand N, et al. Investiga- tion of bacterial diversity in the feces of cattle fed different diets. J Anim rapid publication on acceptance Sci. 2014;92(2):683–94. support for research data, including large and complex data types 23. Durso L, Wells JE, Kim MS. Diversity of microbiomes in beef cattle. In: • gold Open Access which fosters wider collaboration and increased citations Highlander SK, Rodriguez-Valera F, White BA, editors. Encyclopedia of Metagenomics: environmental Metagenomics. Boston: Springer US; 2015. maximum visibility for your research: over 100M website views per year p. 129–38. 24. Malmuthuge N, Guan LL. Understanding the gut microbiome of At BMC, research is always in progress. dairy calves: opportunities to improve early-life gut health. J Dairy Sci. Learn more biomedcentral.com/submissions 2017;100(7):5996–6005. https:// doi. org/ 10. 3168/ jds. 2016- 12239. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png BMC Microbiology Springer Journals

Dynamic changes in fecal bacterial microbiota of dairy cattle across the production line

Loading next page...
 
/lp/springer-journals/dynamic-changes-in-fecal-bacterial-microbiota-of-dairy-cattle-across-AZm5d7Uoe3

References (38)

Publisher
Springer Journals
Copyright
Copyright © The Author(s) 2022
eISSN
1471-2180
DOI
10.1186/s12866-022-02549-3
Publisher site
See Article on Publisher Site

Abstract

Background: Microbiota play important roles in the gastrointestinal tract (GIT ) of dairy cattle as the communities are responsible for host health, growth, and production performance. However, a systematic characterization and comparison of microbial communities in the GIT of cattle housed in different management units on a modern dairy farm are still lacking. We used 16S rRNA gene sequencing to evaluate the fecal bacterial communities of 90 dairy cat- tle housed in 12 distinctly defined management units on a modern dairy farm. Results: We found that cattle from management units 5, 6, 8, and 9 had similar bacterial communities while the other units showed varying levels of differences. Hutch calves had a dramatically different bacterial community than adult cattle, with at least 10 genera exclusively detected in their samples but not in non-neonatal cattle. Moreover, we compared fecal bacteria of cattle from every pair of the management units and detailed the number and rela- tive abundance of the significantly differential genera. Lastly, we identified 181 pairs of strongly correlated taxa in the community, showing possible synergistic or antagonistic relationships. Conclusions: This study assesses the fecal microbiota of cattle from 12 distinctly defined management units along the production line on a California dairy farm. The results highlight the similarities and differences of fecal microbiota between cattle from each pair of the management units. Especially, the data indicate that the newborn calves host very different gut bacterial communities than non-neonatal cattle, while non-neonatal cattle adopt one of the two distinct types of gut bacterial communities with subtle differences among the management units. The gut microbial communities of dairy cattle change dramatically in bacterial abundances at different taxonomic levels along the pro - duction line. The findings provide a reference for research and practice in modern dairy farm management. Keywords: Dairy cattle, Production line, Microbiome, Gastrointestinal, 16S rRNA, Bioinformatics cells in the human body has been estimated as 10 times Background more than the number of human cells [1] or at roughly It is now well-established that microbiota, especially bac- the same order in a recent study [2], this ratio can rise terial communities, play crucial roles in the physiology to approximately 120 times in ruminants such as cattle and health of all mammals. While the number of bacterial [3]. Cattle depend on their gastrointestinal microbiota to digest and convert the plant mass that cannot be directly *Correspondence: xdli@ucdavis.edu; zcsu@uncc.edu digested into absorbable nutrients necessary for host Department of Bioinformatics and Genomics, the University of North health and development. Thus, a better understanding of Carolina at Charlotte, Charlotte, North Carolina, USA the structure of the gastrointestinal microbiota is instru- Department of Population Health and Reproduction, School mental for both production and scientific inquiry. Par - of Veterinary Medicine, University of California, Davis, CA, USA Full list of author information is available at the end of the article ticularly, in the modern dairy system, calves, heifers, and © The Author(s) 2022. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. The Creative Commons Public Domain Dedication waiver (http:// creat iveco mmons. org/ publi cdoma in/ zero/1. 0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Zhao et al. BMC Microbiology (2022) 22:132 Page 2 of 15 cows at different ages and production stages are raised dairy herd in California starting from the newborn to late and managed in independent yet inter-connected man- lactation and dry cows using 16S rRNA gene sequencing agement units with housing and dietary differences [4]. data. Specifically, male calves are commonly sold at birth while female calves are raised on the source dairy from birth Results to weaning (approximately 70 days of age) in group pens Composition of the bacterial communities in the fecal or more commonly in individual hutch units that may samples be wooden, metal, or plastic of combination of materials We analyzed a total of 90 fecal samples collected on the and where calves are fed milk two to three times a day same day from 90 different dairy cattle in 12 management and offered solid grain mix ad  libitum with the goal of units (Fig.  1, Table  1) in a dairy herd. We generated a transitioning to a solid diet by the age of weaning. Once total of 6,092,309 16S rRNA gene paired-end sequencing weaned, growing calves are moved to group pens where reads with an average of 67,692 ± 7963 reads per sample. they are fed a roughage diet with concentrate formu- Clustering sequence reads from V3/V4 regions of 16S lated to their nutrient requirements until breeding age. rRNA genes into different Operational Taxonomy Units Once bred, pregnant heifers are moved to maternity pens (OTUs) has been a widely adopted strategy. However, closer to their calving date. After calving, the newborn the traditional OTU methods obscure similar sequences calf is fed colostrum and raised as described earlier. Post- by grouping them into a consensus sequence and fail to partum cows are moved to the fresh cow milking pen tell the technical errors from real biological variations before being moved to subsequent milking pens depend- [14]. u Th s, we used a more recent method DADA2 that ing on the dairy’s management practices with respect to records how many times an exact sequence was read breeding (not pregnant and pregnant pens) and level of and infers sequence variants by a statistical model learn- milk production (high and low milking pens). Late preg- ing the sequencing error rate from the samples them- nant cows around 60 days prior to calving are dried off, an selves [14]. Given the common usage of the term, we use industry practice of cessation of milking to allow the dam OTU to represent the inferred sequence variant instead to replenish her body resources and initiate colostrogen- of Amplicon Sequence Variant (ASV) that the DADA2 esis in preparation for giving birth to a calf. Adult cows authors proposed [15]. After quality filtering and chimera at different stages are commonly fed a total mixed ration removal with DADA2 [14], a total of 5,080,044 sequence formulated to meet the specific stage of lactation nutrient reads with an average of 56,444 ± 6633 reads per sample requirements. Given such differences in management, were preserved. OTUs that were taxonomically classified housing, and diets, it is important to understand the as “Archaea/Eukaryote” or appeared in less than 9 sam- dynamics of the gut microbiota changes between/among ples (10% of sample size) were excluded. Although the different management units over the entire production number of sequencing reads varied in our samples, rar- life cycle of dairy cattle. The knowledge gained can help efaction curves showed that our sampling was sufficient design new strategies to improve production as well as to represent the bacterial communities as the curve had the health of both the animals and humans who consume almost plateaued or started to form a plateau (Addi- the produced meat and milk. tional  file  1: Fig. S1). The OTU table was normalized by Since traditional culture-dependent techniques can rarefying with a threshold of 12,637 sequences, which is only recover a small portion of the microbial population the minimum number of reads of all the 90 samples after [5], several research groups have explored the gut micro- processing. All downstream analyses were based on the biota of dairy cattle using the next-generation sequenc- normalized OTU table (Additional file  2: Supplementary ing (NGS) technologies [3, 6–13]. For instance, Mao file 1). et al. analyzed the microbiota of ten gastrointestinal sites We identified a total of 4681 OTUs in the 90 fecal in Holstein cattle using 16S rRNA gene sequencing and samples. These OTUs could be taxonomically assigned found that these cattle hosted microbiota with signifi - to 20 phyla, of which Firmicutes (2573/55.0%), Bac- cant spatial heterogeneity [7]. Dill-McFarland et  al. ana- teroidetes (1291/27.6%), and Tenericutes (267/5.7%) lyzed the succession pattern of bacterial communities in included the highest number of the OTUs (Fig.  2A). dairy cattle from 2-week-old to first lactation [9]. Shanks While only Firmicutes (587,075/51.6% reads), Bacteroi- et al. profiled the structure of fecal bacteria in cattle from detes (412,852/36.3% reads), Patescibacteria (53,402/4.7% various feeding operations [12]. However, a systematic reads), and Proteobacteria (16,379/1.4% reads) were contrast of the gut bacterial microbiota in dairy cattle observed in all 90 samples, they accounted for 94.1% of along all the management units on a modern farm is still the total bacterial communities at the phylum level. Fir- absent. To fill these gaps, we analyzed the gut bacterial micutes had the highest average relative abundance in communities of dairy cattle in management units of a the samples from all the management units except hutch Zhao  et al. BMC Microbiology (2022) 22:132 Page 3 of 15 Fig. 1 Management units and dairy cattle life cycle on a dairy farm. A Schematic diagram of the independent yet inter-connected management units of the dairy herd where the samples were collected. The number around each box represents the management unit ID used in this study. B Dairy cattle life cycle by age and production stage calves, where Bacteroidetes had the highest average rela- at this taxonomic level. Ruminococcaceae (397,075/34.9% tive abundance of 62.3% and Firmicutes only accounted reads) was the most dominant taxon in average rela- for 33.4% (Fig. 2B). tive abundance at this level, followed by Rikenellaceae At the family level, these OTUs could be assigned to (131,673/11.6% reads), Prevotellaceae (96,322/8.5% 99 families, in which Ruminococcaceae (1441/34.9%), reads), Lachnospiraceae (77,143/6.8% reads), Murib- Rikenellaceae (500/11.6%), Lachnospiraceae (319/6.8%), aculaceae (70,454/6.2% reads) and Saccharimonadaceae and Prevotellaceae (273/8.5%) contained the highest (53,121/4.7% reads) (Fig. 2D). number of the OTUs (Fig.  2C). Although only 12 fami- At the genus level, 2169 (46.3%) of the 4681 OTUs lies were detected in all 90 samples, they accounted for could be assigned to 167 known genera, while the 992,097 (87.2%) reads of the total bacterial communities remaining 2512 OTUs (53.7%) could not be classified Zhao et al. BMC Microbiology (2022) 22:132 Page 4 of 15 Table 1 Classification of cattle based on the growth, production stages, and their residential management units of the dairy herd where samples were collected Management Unit No. of Unit ID Description Samples Hutch calves 9 1 From birth to approximately 1–2 weeks after weaning, individually housed (1 to 70 days age approximately) Post weaned heifers 6 2 Group-housed heifers and bull calves (if not sold) fed solid diet Breeding heifers 8 3 Approximately 13 to 15 months old Springer 8 4 Within 1 to 4 weeks from calving, pregnant nulliparous heifers Fresh uniparous cows 8 5 1 to 2 months post-calving, first lactation cows Mid-lactation uniparous cows 8 6 60 to 250 Days in Milk, first lactation Fresh multiparous cows 8 7 1 to 2 months post-calving, second lactation or greater Mid-lactation multiparous cows 8 8 60 to 250 Days in Milk, second lactation or greater Pregnant late 8 9 > 250 Days in Milk lactation cows Far-off dry cows 3 10 21 to 60 days prior to calving, multiparous dry cows Close up dry cows 8 11 within 21 days prior to calving, multiparous dry cows Hospital pen/ 8 12 Lactating cows treated with medication that requires milk withdrawal. May include fresh cows Fresh (post-calving) (post-calving) during their transition from producing colostrum to milk and/or pending milk withdrawal after treatment at drying (prior to entering Far-off dry pen). to known genera, thus annotated by 70 lowest known and B, the fecal bacterial communities in pre-weaned taxonomic ranks (family, order, etc.). These unclassified calves (unit 1) fed primarily milk and housed in indi- taxa might be novel bacteria in the cattle feces or not vidual hutches (Table  1) had significantly lower Chao 1 differentiable solely based on the hypervariable regions and Shannon indexes than those in all other units, except of 16S rRNA genes. Overall, the largest number of the for the post-weaned group-housed calves fed a solid diet OTUs were assigned to [f ]Ruminococcaceae (588/11.9% (unit 2) (Kruskal Wallis, false discovery rate, FDR < 0.05), OTUs), followed by Rikenellaceae_RC9_gut_group suggesting that the hutch calves generally had simpler (288/6.15% OTUs), and Ruminococcaceae_UCG-010 fecal bacterial communities. Post weaned heifers (unit 2), (250/5.3% OTUs) (Fig.  2E). While only 17 genera were while having no significant differences in bacterial com - consistently detected in the 90 samples, these commonly munities compared with hutch calves (unit 1), consist- shared genera occupied 72.8% of the total bacterial com- ently differed from breeding heifers (unit 3), springers munities at this taxonomic level. Most of the assigned (unit 4), and pregnant late lactation cows (unit 9), which genera had an average abundance of < 2% in the sam- might be related to the development of the rumen and ples from the management units, while only 15 genera these growing young cattle being full ruminants by the had a relative abundance ≥ 2% including [f] Rumino- time they were bred. coccaceae (11.5%), Ruminococcaceae_UCG-005 (8.8%), Rikenellaceae_RC9_gut_group (8.2%), [f ]Muribaculaceae Similarity and difference of the fecal bacterial communities (6.2%), [f]Lachnospiraceae (5.2%), Candidatus_Saccha- between management units rimonas (4.7%), [o] Bacteroidales (4.3%), Bacteroides To further evaluate the similarities and differences of (3.6%), Ruminococcaceae_UCG-014 (3.5%), Ruminococ- the bacterial communities between different manage- caceae_UCG-013 (3.4%), Christensenellaceae_R-7_group ment units, we calculated the β-diversity (Bray-Curtis (3.0%), Ruminococcaceae_UCG-010 (2.6%), Prevotel- distance) of the samples and visualized the results laceae_UCG-003 (2.5%), [f]Prevotellaceae (2.2%), and using non-metric multidimensional scaling (NMDS). Alistipes (2.2%) (Fig. 2F). As shown in Fig. 4A, the hutch calves’ samples (unit 1) were largely grouped to form a cluster (I), which sepa- Alpha diversity of the cattle fecal bacterial communities rated from the samples from the other management in the management units units, with two post-weaned heifers’ (unit 2) samples We next compared the richness and evenness of the fecal included, suggesting that the fecal bacterial communi- bacterial communities of cattle in different management ties of hutch calves (unit 1) are similar to one another units in the production line using both Chao 1 richness but largely different from those of cattle in the remain- index and Shannon diversity index. As shown in Fig.  3A ing management units. Interestingly, the samples from Zhao  et al. BMC Microbiology (2022) 22:132 Page 5 of 15 Fig. 2 Assignment of the OTUs to different taxonomic levels. A The number of OTUs assigned to different phyla. B Average relative abundances of the phyla in the samples from different management units. C The number of OTUs assigned to different families. D Average relative abundances of the families in the samples from different management units. E The number of OTUs assigned to different genera. F Average relative abundances of the genera in the samples from different management units. In B, D, and F, the length of color-coded bars represents the average relative abundance of the taxa in the samples in the indicated management units. Taxa with < 2% relative abundance in all the units were merged into the “Other/Unassigned” category Zhao et al. BMC Microbiology (2022) 22:132 Page 6 of 15 Fig. 3 Diversity of fecal bacterial communities of cattle in the management units. Boxplots of Chao 1 richness index (A) and Shannon diversity index (B) of fecal bacterial communities of cattle in the management units (x-axis). Each violet-red dot represents the bacterial communities in a fecal sample, and management units are coded by distinct colors. The numbers above each box indicate the management units that are significantly different from the current one with the pairwise Wilcoxon test (P < 0.05) the other units form two rather compact yet distinct significant differences between the two clusters clusters II and III (Fig.  4A), suggesting the samples in (Additional file 3: Fig. S2). each cluster have quite similar bacterial communities. Cluster II contains samples from all the management Shared and unique taxa at the genus level units except unit 1; in contrast, cluster III comprises in the management units samples only from units 3, 5, 6, 7, 8, 11, and 12, sug- For the 237 taxa at the genus level (167 assigned to gesting that non-neonatal cattle (from unit 2 to 12) known genera and 70 assigned to higher taxonomic may only have one of two types of rather uniform bac- ranks), 112 of them had the average relative abundance of terial structures. 0.1% or greater in at least one management unit. Among Moreover, we found that at the phylum level, the these 112 genera, 70 were observed in all management relative abundances of several phyla were signifi- units while the remaining 42 genera existed in a varying cantly different between the samples in clusters number of management units. Particularly, there were II and III (Fig.  4B-E, t-test, P < 0.05). The sample s several genera largely detected only in management unit in cluster II were high in Bacteroidetes (Fig.  4B), 1but barely seen in other units, suggesting these taxa while samples in cluster III were high in Fir- may play important roles in the early stage of life for micutes (Fig.  4C), Patescibacteria (Fig.  4D), and the cattle. For  example, 10 genera, as shown in Table  2 Actinobacteria (Fig.  4E). There were also five addi- listed by the relative abundance from high to low, were tional phyla with low relative abundances showing observed majorly in unit 1 but sparsely seen in other units. In contrast, we found Prevotellaceae_UCG-004, (See figure on next page.) Fig. 4 Similarity and difference of the fecal bacterial communities. A Non-metric multidimensional scaling (NMDS) plot of the Bray-Curtis distance for bacterial communities in different management units. Each point represents a fecal bacterial community and is colored by the management unit from which it was sampled. The communities are grouped into three clusters (I, II, and III) circled in blue, black, and red rings, respectively. Boxplot of four phyla’s relative abundances in clusters II and III for B Bacteroidetes C Firmicutes D Patescibacteria and E Actinobacteria Zhao  et al. BMC Microbiology (2022) 22:132 Page 7 of 15 Fig. 4 (See legend on previous page.) Zhao et al. BMC Microbiology (2022) 22:132 Page 8 of 15 Table 2 Number of occurrences and taxonomic assignment at the genus level for taxa that were either majorly detected in unit 1 or missing in unit 1 but observed in other units Avg. No. of Reads Avg. No. of Sum of Avg. No. of Reads Phylum Family Genus in Unit 1 Reads in Unit 2 in other 10 Units Genera majorly observed in unit 1 330 (2.6%) 4 2 Bacteroidetes Prevotellaceae Prevotella_9 157 (1.2%) 0 0 Bacteroidetes Prevotellaceae Prevotella_2 102 (0.8%) 0 3 Bacteroidetes Prevotellaceae Prevotella 97 (0.8%) 10 5 Clostridiales Ruminococcaceae Faecalibacterium 87 (0.7%) 0 1 Proteobacteria Burkholderiaceae Sutterella 86 (0.7%) 9 0 Bacteroidetes Tannerellaceae Parabacteroides 74 (0.6%) 2 0 Fusobacteria Fusobacteriaceae Fusobacterium 72 (0.6%) 17 0 Firmicutes Ruminococcaceae Subdoligranulum 54 (0.4%) 4 0 Firmicutes Ruminococcaceae Oscillospira 86 (0.7%) 9 0 Bacteroidetes Tannerellaceae Parabacteroides 15 (0.1%) 0 6 Erysipelotrichales Erysipelotrichaceae Erysipelotrichaceae_UCG-004 Genera missing in unit 1 but observed in other units 0 31 1581 Bacteroidales Prevotellaceae Prevotellaceae_UCG-004 0 1 1196 Bacteroidales p-2534-18B5_gut_group Unassigned, annotated as [f ]p-2534-18B5_gut_group 0 16 905 Bacteroidales Rikenellaceae dgA-11_gut_group 0 2 401 Clostridiales Ruminococcaceae Caproiciproducens dgA-11_gut_group, Caproiciproducens, and an unas- potential interactions between different bacteria, we cal - signed genus ([f ]p-2534-18B5_gut_group) missing in ani- culated the correlations of the core bacteria using the mals from unit 1, but in all other 11 management units SparCC (Sparse Correlations for Compositional data) (Table 2). algorithm [16]. By setting ±0.7 as the threshold values for strong positive and negative correlations and pseudo Dynamics of bacterial communities p-values < 0.05 as the significant correlations, we found between the management units 61 OTUs that had strong correlations with at least one From the core bacteria, the 112 genera with an average other OTU (Fig.  6; Additional  file  4: Supplementary minimal abundance of 0.1% in at least one management file  2; Additional  file  5: Supplementary file  3). Most of unit), we found 56 of them displayed significantly differ - these strong correlations are positive, while only one pair ent abundances in the samples from the 12 management between Ruminococcaceae_UCG-005 and Bacteroides is units (ANOVA, FDR < 0.05). We further summarized negative (SparCC = − 0.70). In Fig.  6, all the strong cor- how each of these 56 genera was significantly different relations are shown within the networks. From the 181 in their composition between different pairs of man - strong correlations, 117 are between the OTUs assigned agement units (Tukey HSD, P < 0.05, Fig.  5A and B). For to the same phylum, indicating that strong correlations example, unit 1 had a varying number of significantly dif - are more likely to be within the same phylum than those ferent genera with all other 11 units, from 11 with unit 5 from different phyla. to 27 with unit 7 (Fig. 5A). As shown in Fig. 5B, the rela- tive abundances of these significantly different genera Functional prediction of the bacterial community between unit 1 and other units were in the range of 25.9 To further explore the fecal bacteria of the dairy cat- and 52.6%. Interestingly, there is no significantly different tle, we predicted the functions of the bacterial com- genus between any pair of management units 5, 6, 8, and munity using a marker gene based tool Tax4Fun2 [17]. 9, suggesting the types and relative abundances of the Compared to the initial version Tax4Fun, Tax4Fun2 has genera were similar in these four units. higher accuracy, allows users to customize the reference genomes, and integrates a multifunctional redundancy Synergistic and antagonistic relationships evaluation [17, 18]. Bacteria thriving in a community may interact with one A total of 7726 Kyoto Encyclopedia of Genes and another synergistically or antagonistically. To reveal such Genomes (KEGG) Orthologs (KOs) were predicted in Zhao  et al. BMC Microbiology (2022) 22:132 Page 9 of 15 Fig. 5 Significantly different taxa at genus level between all pairs of management units. A Number of genera significantly different between each pair of the management units. B Relative abundance of significantly different genera between each pair of the management units Zhao et al. BMC Microbiology (2022) 22:132 Page 10 of 15 Fig. 6 Strong correlations in the core bacteria. Each node is an OTU and the number within represents the OTU ID. Nodes are colored based on their taxonomic assignment at the phylum level. The edges in green and red indicate positive and negative correlations the identified bacterial microbiota in the samples while 20 of them were shown in Fig. 7C. Again, energy metab- no single KO had a relative abundance higher than 2% olism related pathways are most abundant, such as gly- in any sample (Additional  file  6: Supplementary file  4). colysis/gluconeogenesis, carbon fixation, and methane Among them, 4001 (51.8%) KOs were present in all 90 metabolism. samples, indicating these molecular functions were essential for the cattle regardless of age, housing, and Discussion production status such as K00001 (alcohol dehydroge- In modern dairy management, it is not uncommon to nase) that mediates the oxidation and reduction of eth- house animals in different management units based on anol [19]. Interestingly, some KOs were only present in their ages, nutrition requirements, reproductive status, the samples from one management unit. For instance, a lactation status, production, and management styles [21]. collection of 9 KOs (K04340, K14194, K16046, K16227, Such a management system optimizes milk production K18254, K18611, K18906, K20218, and K21329) were and management but brings challenges profiling and ana - only observed in some samples of unit 1, though at lyzing the bacterial communities in the GIT because the very low abundances (Additional file  6: Supplementary bacterial communities can be influenced by numerous file  4). These KOs might be involved in the food diges - factors including, but not limited to, diet, animal physiol- tion of hutch calves. For instance, K18611 (4-pyridoxic ogy and status, the farm environment, geographic loca- dehydrogenase) is an enzyme that degrades vitamin B6, tion, antimicrobial use, and management practices of the which is contained in milk, the primary diet for hutch production lifecycle [6, 12, 22–25]. For instance, the bac- calves housed in unit 1 [20]. There were 982 (12.7%) sig - terial communities of cattle housed individually were less nificantly differential KOs (ANOVA, FDR < 0.05) and the likely to be affected by other cattle compared with those top 20 most abundant ones were shown in Fig. 7A. housed together in open pens. Another example is that From the pathway perspective, Tax4Fun2 predicted 6, cattle with diseases may be treated with antimicrobial 45, and 239 pathways at KEGG level 3, 2, and 1, respec- drugs that may directly change the balance of the bacte- tively. Metabolism (70.4–72.0%) was the most abundant rial communities. pathway at level 3 in every management unit, followed In this study, we sought to characterize the composi- by Environmental Information Processing (12.0–13.2%), tion, diversity, and dynamics of the fecal bacteria in dif- Cellular Processes (6.1–6.7%), Genetic Information Pro- ferent management units over the production lifecycle of cessing (5.3–5.7%), Human Diseases (2.8–3.1%), and dairy cattle on a California dairy farm. We used paired- Organismal Systems (1.2–1.3%, Additional  file  7: Sup- end sequencing of the V3/V4 hypervariable regions of plementary file  5). Further analysis revealed 10 signifi - the 16S rRNA genes to profile the bacterial communi - cantly differential pathways at level 2, with the energy ties in 90 fecal samples collected from 90 cattle in the 12 metabolism pathway being the most abundant (Fig. 7B). management units. Our results suggest high similarity However, the abundances of these pathways combined between any two of the management units 5, 6, 8, and were only 6.6% on average. At level 1, there were 68 sig- 9, but show significant differences in most of the other nificantly differential pathways, and the most abundant management unit pairs in terms of richness, evenness, Zhao  et al. BMC Microbiology (2022) 22:132 Page 11 of 15 Fig. 7 Significantly differential orthologs and pathways in the KEGG database predicted using Tax4Fun2. A Top 20 the most abundant significantly differential KOs. B Significantly differential level 2 pathways. C Top 20 the most abundant significantly differential level 1 pathways. Rows represent KOs/pathways and columns represent management units. Subplots were arranged by the relative abundance from high to low Zhao et al. BMC Microbiology (2022) 22:132 Page 12 of 15 and structure of the gut bacterial communities at the 2 months (70 days on average) to 13-month old, the genus level. microbiota changes were still ongoing towards the bacte- Overall, consistent with several earlier studies [6, 7, 12, rial profiles in adults during this transition period. Using 13, 22], our results reveal Firmicutes and Bacteroidetes the same set of samples, we have previously found that were the two most dominant taxa at the phylum level, E. coli from hutch calves exhibited a wider spectrum of distantly followed by several other phyla. Specifically, our resistance to antimicrobial drugs compared to bacteria results show that, on average, the Firmicutes and Bacte- from other units [21]. Currently, it largely remains unde- roidetes represented 51.6 and 36.3% of the total commu- termined concerning the roles of bacterial communities nities, respectively. Interestingly, the relative abundance on antimicrobial resistance of specific bacterial species, of Bacteroidetes was lower in cows immediately before however, in future studies, it will be interesting to assess and after calving (within two weeks from calving) in units such roles, for example, resistomes on phenotypes of with fresh multiparous cows (18.7%) and close up dry antimicrobial resistance. cows (23.2%), while the relative abundances were in the For the samples from non-neonatal cattle (from unit 2 range of 29.4 and 62.2% in other units (Fig. 2B). to unit 12), we did not find any evidence supporting that For some genera, as Mao et al. [7] demonstrated, Prevo- the bacterial communities could be clustered based on tella had a lower abundance in samples taken from the the management unit membership. This demonstrated small and large intestines, but higher in the forestomach, that the management unit itself is not a determinate force such as the rumen. In our work, the relative abundance of for the structure of the bacterial communities. Instead, as Prevotella was only 0.07%. However, in one of the hutch shown in Fig. 4A, we did observe distinct patterns where calves’ (unit 1) samples, we observed Prevotella had a most of our samples from non-neonatal cattle formed relative abundance of 7.16% (905 reads) while the num- two clusters (II and III), implying they had two major dif- ber of Prevotella reads in the other 89 samples was in ferent structures. As we looked deeper, we found that at the range of 0 to 11. We also found in this sample, Rumi- the phylum level, the relative abundances of several phyla nococcus_1 had a relatively high abundance (5.56%, 703 became very different between the two clusters (Fig.  4B- reads) while this genus was in the range of 0 to 115 reads E). For example. in cluster III, the mean and median of in other samples. So potentially there is some synergistic Patescibacteria relative abundance were 20.9 and 19.7% relationship between Prevotella and Ruminococcus_1 and while those values in cluster II were only 1.9 and 1.1%, the interaction prevent Prevotella from being degraded respectively. What caused the differences in Patescibac - or eliminated when going through the GIT. Such interac- teria in various samples still needs future research, but tion could also be related to the diet of hutch calves being the potential interactions between some taxa classified as primarily milk compared to all the other units fed a solid Firmicutes and Patescibacteria seem to prevent Patesci- diet. Moreover, feeding waste milk from the hospital pen bacteria from decreasing. It has been reported that the or fresh cow pen to calves is a common practice in the rumen bacteria in the 1st and 2nd lactations, which over- dairy industry and was done on this dairy at the time lapped with some of our management units, are dynamic of sampling. Waste milk commonly contains antibiotic yet similar, and the samples from the two lactations can- residues of varying concentrations as it is collected from not be separated by the lactation cycle in PCA visualiza- milking cows in the hospital being treated with antibiot- tion [8]. This demonstrated that the bacterial structure ics or awaiting clearance during their withdrawal period might undergo further shaping going through the GIT before rejoining the pens where milk is harvested for (their rumen samples versus our fecal samples) or the human consumption. formed patterns were just a case-sensitive phenomenon. It is noticeable that bacterial communities in hutch Pitta et  al. reported that significant bacterial population calves were significantly different from those in other change was observed during the transition from 21 days management units except post-weaned heifers (Figs.  2, before calving to 21 days after calving in uniparous and 3, 4A, 5). These differences confirmed the earlier find - multiparous cows, respectively [26]. This shift period cor - ings that the enteric microbiota in neo-natal calves were responded approximately to our management units 4 and different from adult cows and the bacterial communities 5 for uniparous cows and units 10 and 11 for multiparous underwent a dramatic change during the development cows. However, we did not see this significant shift as dem - at an early age [9, 10, 23]. Intriguingly, bacteria from the onstrated earlier [26]. The only significant difference we samples in post-weaned heifers displayed similar patterns observed was Shannon diversity between units 4 and 5. This with the communities from the adult cattle, predomi- could be due to the different sites of the GIT from which nately Firmicutes; but at the same time, they showed the samples were taken (their rumen samples versus our no significant difference from those of hutch calves. As fecal samples). It is also possible that this was due to the four the post weaned heifers were aged from approximately management units in our study not being strictly narrowed Zhao  et al. BMC Microbiology (2022) 22:132 Page 13 of 15 to 21 days prior and post-calving. As the dairy farm environ- sampling population in each management unit can also ment is dynamic, it is not surprising that most taxa were be found at Li et  al. [21]. These samples were also used shared by cattle housed in different management units. for an antimicrobial resistance study that has been pub- Measuring potential bacterial interactions was chal- lished by Li et al. [21]. An aliquot of each sample shipped lenging and barely reported in dairy cattle studies. The in refrigerated conditions to Dr. Su’s laboratory at the classic correlation methods had their limitations when Department of Bioinformatics and Genomics, the Uni- applied to genomic data such as 16S rRNA gene sequenc- versity of North Carolina at Charlotte for 16S rRNA gene ing data which are sparse and compositional. SparCC as a sequencing. method that negates the negative correlation bias of com- positional data [27] and identifies true association missed by others [28], was used here as a way of evaluating bac- Illumina MiSeq sequencing terial interactions. In general, we identified 180 strong DNA from stool samples was extracted with the Qiagen positive correlations and 1 strong negative correlation DNA Stool kit following the manufacturer’s instructions. between 61 OTUs (Fig.  6). These co-occurrent OTUs Two steps of Polymerase Chain Reaction (PCR) pro- may tend to share the same habitats and perform similar cedures were used to generate amplicons from the 16S functions, as most of these co-occurrent patterns spotted RNA genes for sequencing. The first-round PCR was to between OTUs were in the same phyla (Fig. 6). target V3/V4 regions of 16S rRNA genes with the for- ward primer: 5′- CCT ACG GGNGGC WGC AG and the reverse primer: 5′- GAC TAC HVGGG TAT CTA ATC C. Conclusions This step was done with the KAPA Biosciences HiFi PCR In this study, we profiled the structure and dynamics of kit and additional BSA. The protocol consists of initial gut bacterial communities from cattle in 12 independ- denaturation at 95 °C for 3 min, followed by 25 cycles of ent yet inter-connected management units on a mod- denaturation (90 °C for 30 s, 55 °C for 30 s, and 72 °C for ern California dairy farm. To the best of our knowledge, 30 s), and final elongation at 72 °C for 5 min. The PCR this is the first study that describes the bacterial com - products were cleaned up with Ampure XP beads. The munities across all management units and reveals the second-round PCR was performed with Nextera XT structures of gut microbial communities to each of index Primers and sequencing Adaptors with the follow- the well-defined management units. We analyzed the ing setting: initial denaturation at 95 °C for 3 min, fol- changes in gut microbial communities across production lowed by 8 cycles of denaturation (90 °C for 30 s, 55 °C for systems. Though the fecal microbiota were similar in 4 of 30 s, and 72 °C for 30 s), and final elongation at 72 °C for the management units, they showed significant changes 5 min. The PCR products were cleaned up with Ampure between others. It is confirmed that microbial ecology XP beads and paired-end sequenced (2 × 300 bp) on an underwent dramatic changes in the early days of life, as Illumina MiSeq platform at the University of North Car- evidenced by the significantly different bacteria in hutch olina at Charlotte. calves from other adult cattle and revealed dynamics of the bacterial abundance in the later stages of the produc- tion lifecycle. Moreover, we identified at least 10 genera OTU table construction that were detected only in hutch calves but were absent Primers with raw sequences were removed by Cutadapt in all the other cattle in other units. These genera might [29]. We performed quality control using DADA2’s “fil - play crucial roles in the early establishment and develop- terAndTrim” function with “trancLen” equal to 200 bp ment of the GIT. We also dissected potential interactions for forward reads and 150 bp for reverse reads based on among gut bacterial groups, mostly from the species in quality profiles. Technical error rate learning was per - the same phylum. formed with all the sequences in the samples. Sample inference was performed by the “dada” function with the Methods setting optional parameter “pool = TRUE. ” Paired-end Sample collection and study herd reads merger in DADA2 resulted in approximately 50% On a single day in June 2016, a total of 90 fecal samples loss of sequences, thus only forward reads were used in were collected from 90 individual cattle in 12 manage- this study. OTU table and chimera removal were imple- ment units (Fig. 1, Table 1) from a dairy herd in the Cen- mented with default parameters. We used the “assignTax- tral Valley of California, USA. Cattle in each of these onomy” function that provided a native implementation management units were identified based on convenience of the RDP Classifier [30] with minimum bootstrap con - sampling. Trained study personnel collected the fecal fidence of 80 to assign taxonomy from the phylum level samples manually from the rectum of cattle using stand- ard veterinary protocols. Other information such as the Zhao et al. BMC Microbiology (2022) 22:132 Page 14 of 15 to the genus level to each OTU. SILVA database release Additional file 5: Supplementary file 3. Taxonomic table. Table of taxo - 132 [31] was used as the reference database. nomic classification of 61 strongly correlated OTUs. Additional file 6: Supplementary file 4. The table of function predic- tions generated by the Tax4Fun2 program. Function prediction Molecular function prediction was done using Tax- Additional file 7: Supplementary file 5. The table of pathway predic- tions generated by the Tax4Fun2 program. 4Fun2 with the default reference library. The functional profiles were generated with the built-in functions Acknowledgments “runRefBlast” and “makeFunctionalPrediction” with Not applicable. default settings. Authors’ contributions Z.S. and X.L. designed and conceived the experiments. Sample and animal measurements controlled and instructed by X.L., S.S.A., and D.R.W. L.Z. per- Statistical analysis formed microbial data processing and analysis. L.Z., Z.S., and X.L. interpreted Rarefying was performed by the “single_rarefaction” the results. L.Z. and Z.S. wrote the manuscript. S.S.A. prepared the artwork of the dairy farm. X.L., E.R.A., S.S.A., and D.R.W. contributed to the revision of the function with the minimum number of reads of all sam- manuscript. The author(s) read and approved the final manuscript. ples, which is equal to 12,637 in QIIME [32]. We used alpha diversity, including Chao 1 index which evaluates Funding This study was partially supported by the National Institute of Food and Agri- richness, and Shannon index which evaluates diversity culture (NIFA)‘s Exploratory Research program (2015–67030-23892), the NIFA’s to measure the within-sample diversity. Differences in Dairy Herd Health and Food Safety Formula Funds (CALV-DHHFS-0053) from Chao 1 index and Shannon index in different units were the USDA, and National Institutes of Health (R01GM106013). assessed by the Kruskal Wallis test with the Benjamini- Availability of data and materials Hochberg (BH) correction for multi-comparisons, and All DNA sequences have been deposited in NCBI’s Sequence Read Archive pairwise management units’ comparisons were per- (SRA) with the accession number PRJNA607283. Direct link: https:// www. ncbi. nlm. nih. gov/ sra/? term= PRJNA 607283 formed by pairwise Wilcoxon test. The Bray-Curtis dis - tance matrix was employed to perform beta diversity Declarations analysis. Visualization was done by a non-metric mul- tidimensional scaling plot. SparCC correlations were Ethics approval and consent to participate calculated by FastSpar [33], a C++ implementation of Sampling was approved by the Institutional Animal Care and Use Commit- tee (IACUC) of University of California Davis (protocol number 18941) and SparCC algorithm (100 bootstrap samples were gener- all experiments were performed in accordance with relevant guidelines and ated for pseudo p-value calculation). R software [34] regulations. and R packages ggplot2 [35], vegan [36], superheat [37], Consent for publication ggpubr [38], and qgraph [39], were used for calculation Not applicable. and visualization. Competing interests The authors declare that they have no competing interests. Abbreviations GIT: Gastrointestinal tract; NGS: Next-generation sequencing; OTU: Operational Author details taxonomy unit; NMDS: Non-metric multidimensional scaling; SparCC: Sparse Department of Bioinformatics and Genomics, the University of North Carolina correlations for compositional data; KEGG: Kyoto encyclopedia of genes and at Charlotte, Charlotte, North Carolina, USA. Western Institute for Food Safety genomes; KO: KEGG ortholog. and Security, University of California, Davis, California, USA. Depar tment of Population Health and Reproduction, School of Veterinary Medicine, Univer- sity of California, Davis, CA, USA. Veterinary Medicine Teaching and Research Supplementary Information Center, School of Veterinary Medicine, University of California, Davis, CA, USA. The online version contains supplementary material available at https:// doi. org/ 10. 1186/ s12866- 022- 02549-3. Received: 28 September 2021 Accepted: 11 April 2022 Additional file 1: Figure S1. Rarefaction curves showing “Number of Observed OTUs” as a function of “Number of Sequencing Reads” for each sample. References Additional file 2: Supplementary file 1. Normalized OTU table. The table 1. Reid A, Greene S. Faq: human microbiome. Am Acad Microbiol. 2014. includes the information of OTU ID, abundance, and taxonomic assign- https:// asm. org/ Repor ts/ FAQ- Human- Micro biome. ment for each OTU. 2. Sender R, Fuchs S, Milo R. Revised estimates for the number of human Additional file 3: Figure S2. Boxplot of five additional phyla with low and Bacteria cells in the body. PLoS Biol. 2016;14(8):e1002533. https:// doi. relative abundances showing significant difference between clusters II org/ 10. 1371/ journ al. pbio. 10025 33. and III in Fig. 4A. 3. Ross EM, Moate PJ, Marett LC, Cocks BG, Hayes BJ. Metagenomic predic- tions: from microbiome to complex health and environmental pheno- Additional file 4: Supplementary file 2. SparCC table. Table of calculated types in humans and cattle. PLoS One. 2013;8(9):e73056. https:// doi. org/ SparCC correlations and pseudo p-values for 61 OTUs that had strong cor- 10. 1371/ journ al. pone. 00730 56. relations with at least one other OTU. Zhao  et al. BMC Microbiology (2022) 22:132 Page 15 of 15 4. Konboon M, Bani-Yaghoub M, Pithua PO, Rhee N, Aly SS. A nested 25. Smith G. Antimicrobial decision making for enteric diseases of cattle. Vet compartmental model to assess the efficacy of paratuberculosis control Clin N Am Food Anim Pract. 2015;31(1):47–60. https:// doi. org/ 10. 1016/j. measures on US dairy farms. PLoS One. 2018;13(10):e0203190.cvfa. 2014. 11. 004. 5. Janssen PH. Identifying the dominant soil bacterial taxa in libraries of 16S 26. Pitta DW, Indugu N, Kumar S, Vecchiarelli B, Sinha R, Baker LD, et al. rRNA and 16S rRNA genes. Appl Environ Microbiol. 2006;72(3):1719–28. Metagenomic assessment of the functional potential of the rumen https:// doi. org/ 10. 1128/ AEM. 72.3. 1719- 1728. 2006. microbiome in Holstein dairy cows. Anaerobe. 2016;38:50–60. https:// doi. 6. Tang MT, Han H, Yu Z, Tsuruta T, Nishino N. Variability, stability, and resil-org/ 10. 1016/j. anaer obe. 2015. 12. 003. ience of fecal microbiota in dairy cows fed whole crop corn silage. Appl 27. Gloor GB, Macklaim JM, Pawlowsky-Glahn V, Egozcue JJ. Microbiome Microbiol Biotechnol. 2017;101(16):6355–64. https:// doi. org/ 10. 1007/ datasets are compositional: and this is not optional. Front Microbiol. s00253- 017- 8348-8. 2017;8:2224. https:// doi. org/ 10. 3389/ fmicb. 2017. 02224. 7. Mao S, Zhang M, Liu J, Zhu W. Characterising the bacterial microbiota 28. Tsilimigras MC, Fodor AA. Compositional data analysis of the microbiome: across the gastrointestinal tracts of dairy cattle: membership and poten- fundamentals, tools, and challenges. Ann Epidemiol. 2016;26(5):330–5. tial function. Sci Rep. 2015;5:16116. https:// doi. org/ 10. 1038/ srep1 6116.https:// doi. org/ 10. 1016/j. annep idem. 2016. 03. 002. 8. Jewell KA, McCormick CA, Odt CL, Weimer PJ, Suen G. Ruminal bacterial 29. Martin M. Cutadapt removes adapter sequences from high-throughput community composition in dairy cows is dynamic over the course of two sequencing reads. EMBnet journal. 2011;17(1):10–2. lactations and correlates with feed efficiency. Appl Environ Microbiol. 30. Wang Q, Garrity GM, Tiedje JM, Cole JR. Naive Bayesian classifier for rapid 2015;81(14):4697–710. https:// doi. org/ 10. 1128/ AEM. 00720- 15. assignment of rRNA sequences into the new bacterial taxonomy. Appl 9. Dill-McFarland K, Breaker J, Suen G. Microbial succession in the gas- Environ Microbiol. 2007;73(16):5261–7. trointestinal tract of dairy cows from 2 weeks to first lactation. Sci Rep. 31. Quast C, Pruesse E, Yilmaz P, Gerken J, Schweer T, Yarza P, et al. The SILVA 2017;7:40864. https:// doi. org/ 10. 1038/ srep4 0864. ribosomal RNA gene database project: improved data processing and 10. Dias J, Marcondes MI, de Souza SM, e Silva BCM, Noronha MF, Resende web-based tools. Nucleic Acids Res. 2012;41(D1):D590–D6. https:// doi. RT, et al. Bacterial community dynamics across the gastrointestinal tracts org/ 10. 1093/ nar/ gks12 19. of dairy calves during preweaning development. Appl Environ Microbiol. 32. Caporaso JG, Kuczynski J, Stombaugh J, Bittinger K, Bushman FD, Costello 2018;84(9):e02675–17. EK, et al. QIIME allows analysis of high-throughput community sequenc- 11. Jami E, Israel A, Kotser A, Mizrahi I. Exploring the bovine rumen bacterial ing data. Nat Methods. 2010;7(5):335. community from birth to adulthood. ISME J. 2013;7(6):1069–79. https:// 33. Watts SC, Ritchie SC, Inouye M, Holt KE. FastSpar: rapid and scalable corre- doi. org/ 10. 1038/ ismej. 2013.2. lation estimation for compositional data. Bioinformatics. 2019;35(6):1064– 12. Shanks OC, Kelty CA, Archibeque S, Jenkins M, Newton RJ, McLellan SL, 6. https:// doi. org/ 10. 1093/ bioin forma tics/ bty734. et al. Community structures of fecal bacteria in cattle from different ani- 34. R Core Team. R. A language and environment for statistical computing. mal feeding operations. Appl Environ Microbiol. 2011;77(9):2992–3001. R Foundation for statistical computing. Vienna URL https:// www.R- proje https:// doi. org/ 10. 1128/ AEM. 02988- 10. ctorg/; 2018. 13. Hagey JV, Bhatnagar S, Heguy JM, Karle BM, Price PL, Meyer D, et al. Fecal 35. Wickham H. ggplot2: elegant graphics for data analysis: Springer; 2016. microbial communities in a large representative cohort of California dairy 36. Oksanen J, Kindt R, Legendre P, O’Hara B, Stevens MHH, Oksanen MJ, et al. cows. Front Microbiol. 2019;10:1093. https:// doi. org/ 10. 3389/ fmicb. 2019. The vegan package. Commun Ecol Package. 2007;10:631–7. 01093. 37. Barter RL, Yu B. Superheat: an R package for creating beautiful and 14. Callahan BJ, McMurdie PJ, Rosen MJ, Han AW, Johnson AJ, Holmes SP. extendable heatmaps for visualizing complex data. J Computational DADA2: high-resolution sample inference from Illumina amplicon data. Graph Stat. 2018;27(4):910–22. Nat Methods. 2016;13(7):581–3. https:// doi. org/ 10. 1038/ nmeth. 3869. 38. Kassambara A. Ggpubr:“ggplot2” based publication ready plots (version 15. Callahan BJ, McMurdie PJ, Holmes SP. Exact sequence variants should 0.1. 7); 2017. replace operational taxonomic units in marker-gene data analysis. ISME J. 39. Epskamp S, Cramer AO, Waldorp LJ, Schmittmann VD, Borsboom D. 2017;11(12):2639–43. https:// doi. org/ 10. 1038/ ismej. 2017. 119. Qgraph: network visualizations of relationships in psychometric data. J 16. Friedman J, Alm EJ. Inferring correlation networks from genomic survey Stat Softw. 2012;48(4):1–18. data. PLoS Comput Biol. 2012;8(9):e1002687. https:// doi. org/ 10. 1371/ journ al. pcbi. 10026 87. Publisher’s Note 17. Wemheuer F, Taylor JA, Daniel R, Johnston E, Meinicke P, Thomas T, et al. Springer Nature remains neutral with regard to jurisdictional claims in pub- Tax4Fun2: prediction of habitat-specific functional profiles and functional lished maps and institutional affiliations. redundancy based on 16S rRNA gene sequences. Environmental Microbi- ome. 2020;15:1–12. 18. Aßhauer KP, Wemheuer B, Daniel R, Meinicke P. Tax4Fun: predicting functional profiles from metagenomic 16S rRNA data. Bioinformatics. 2015;31(17):2882–4. 19. Jörnvall H, Persson B, Jeffery J. Characteristics of alcohol/polyol dehydro - genases: the zinc-containing long-chain alcohol dehydrogenases. Eur J Biochem. 1987;167(2):195–201. 20. Kanehisa M, Goto S. KEGG: Kyoto encyclopedia of genes and genomes. Nucleic Acids Res. 2000;28(1):27–30. 21. Li X, Aly SS, Su Z, Pereira RV, Williams DR, Rossitto P, et al. Phenotypic Re Read ady y to to submit y submit your our re researc search h ? Choose BMC and benefit fr ? Choose BMC and benefit from om: : antimicrobial resistance profiles of E. coli and Enterococcus from dairy cattle in different management units on a Central California dairy. Clin fast, convenient online submission Microbiol. 2018;7:312. https:// doi. org/ 10. 4172/ 2327- 5073. 10003 12. thorough peer review by experienced researchers in your field 22. Kim M, Kim J, Kuehn L, Bono J, Berry E, Kalchayanand N, et al. Investiga- tion of bacterial diversity in the feces of cattle fed different diets. J Anim rapid publication on acceptance Sci. 2014;92(2):683–94. support for research data, including large and complex data types 23. Durso L, Wells JE, Kim MS. Diversity of microbiomes in beef cattle. In: • gold Open Access which fosters wider collaboration and increased citations Highlander SK, Rodriguez-Valera F, White BA, editors. Encyclopedia of Metagenomics: environmental Metagenomics. Boston: Springer US; 2015. maximum visibility for your research: over 100M website views per year p. 129–38. 24. Malmuthuge N, Guan LL. Understanding the gut microbiome of At BMC, research is always in progress. dairy calves: opportunities to improve early-life gut health. J Dairy Sci. Learn more biomedcentral.com/submissions 2017;100(7):5996–6005. https:// doi. org/ 10. 3168/ jds. 2016- 12239.

Journal

BMC MicrobiologySpringer Journals

Published: May 14, 2022

Keywords: Dairy cattle; Production line; Microbiome; Gastrointestinal; 16S rRNA; Bioinformatics

There are no references for this article.