Equine Veterinary Journal ISSN 0425-1644 DOI: 10.1111/evj.12361
Changes in the faecal microbiota of mares precede the development of post partum colic J. S. WEESE*, S. J. HOLCOMBE†, R. M EMBERTSON†, K. A. KURTZ†, H. A. ROESSNER†, M. JALALI and S. E. WISMER‡ Department of Pathobiology, Ontario Veterinary College, University of Guelph, Ontario, Canada † Department of Large Animal Clinical Sciences, College of Veterinary Medicine, Michigan State University, East Lansing, USA ‡ Rood and Riddle Equine Hospital, Lexington, Kentucky, USA. *Correspondence email: [email protected]; Received: 21.04.14; Accepted: 03.09.14
Summary Reasons for performing study: Disruptions in the gastrointestinal microbiota may trigger development of post partum colic. Objectives: To determine the effects of the periparturient period on the faecal microbiome and identify associations between the faecal microbiota and post partum colic. Study design: Longitudinal case–control study. Methods: Pre- and post partum faecal samples were collected from mares on 3 farms in central Kentucky. Next generation sequencing of the V4 region of the 16S rRNA gene was performed on samples from 13 mares that developed colic, 13 mares that did not display colic and 5 nonpregnant controls. Results: There were 4,523,727 sequences from 85 samples evaluated (mean ± s.d. 53,220 ± 29,160, range 8442–122,535). Twenty-five phyla were identified, although only Firmicutes, Verrucomicrobia, Actinobacteria and Proteobacteria were present at a relative abundance of 1% or greater. The faecal microbiota of late-term mares differed from nonpregnant mares, with differences in microbial community membership and structure but not the relative abundance of major phyla. There was limited impact of foaling and the post partum period on the faecal microbiome. Faecal samples obtained from mares prior to episodes of colic had significantly higher relative abundance of Proteobacteria (8.2%, P = 0.0006) compared with samples from mares that did not display colic (3.7%). All samples with a relative abundance of Firmicutes of ≤50% preceded colic, as did 6/7 (86%) samples with >4% Proteobacteria. Differences in microbiota membership and structure were also present between mares that developed large colon volvulus and matched controls that did not have colic. Sixty-one indicator operational taxon units were identified for the control (vs. volvulus) samples, and these were dominated by Lachnospiraceae (n = 38) and Ruminococcaceae (n = 8). Conclusions: Foaling had minimal effects on the mares’ faecal microbiota. Numerous differences in the faecal microbiota preceded colic. Associations between Firmicutes (particularly Lachnospiraceae and Ruminococcaceae) and Proteobacteria and development of colic could lead to measures to predict and prevent colic. The Summary is available in Chinese – see Supporting information. Keywords: horse; faecal microbiome; Firmicutes; Proteobacteria; colic; post partum
Introduction Colic is an important problem in post partum mares [1–3] and large colon volvulus, one of the most severe forms of colic, is over represented in this group [1,3]. However, the relationships and triggers connecting these kinds of colic with the post partum period are unclear. Snyder and colleagues reported that half of the admissions to University of California at Davis for surgical correction of colonic volvulus were post partum mares that had foaled 2 h and required administration of analgesics. Final diagnoses and treatment (surgical vs. medical) were obtained from veterinarians on each farm and from veterinarians at Rood and Riddle Equine Hospital if the mares were hospitalised.
DNA extraction DNA extraction was performed using a commercial kit that included a bead-beating step (E.Z.N.A. Stool DNA Kita) following the manufacturer’s instructions. Quantity and quality DNA were accessed by spectrophotometryb.
16S rRNA gene amplification and sequencing The V4 region of the 16S rRNA gene was amplified using the primers S-D-Bact-0564-a-S-15 (5'-AYTGGGYDTAAAGNG-3') and S-D-Bact-0785b-A-18 (5'-TACNVGGGTATCTAATCC-3') [20]. Primers contained an overlapping region of the forward and reverse Illumina sequencing primers (TCGTCGGCAGCGTCAGATGTGTATAAGAGACAG and GTCTCGTGGGCTCGGA GATGTGTATAAGAGACAG, respectively) to allow them to anneal to primers containing the Illumina adaptors plus the 8 bp identifier indices (forward: AATGATACGGCGACCACCGAGATCTACAC-index-TCGTCGGCAGCGTC; reverse: CAAGCAGAAGACGGCATACGAGAT-index-GTCTCGTGGGCTCGG). A 48 μl reaction was performed with 25 μl of Kapa2G Fast HotStart ReadyMix 2Xc; 1.3 μl of BSAd; 18.9 μl of PCR-grade H2O, 2 μl of DNA template and 0.4 μl of both the forward and the reverse 16S primers (10 pmol/μl). The following PCR conditions were used; 3 min at 94°C for denaturing, and 30 cycles of 45 s at 94°C for denaturing, 60 s at 50°C for annealing and 90 s at 72°C for elongation followed by a final period of 10 min at 72°C and kept at 4°C until purification. Polymerase chain reaction products were purified (Agencourt AMPure XP beads)e, quantified by spectrophotometryb and normalised to a final concentration of 2 nmol/l. Sequencing of the library pool was performed at the University of Guelph’s Advanced Analysis Centre using an Illumina MiSeqf and 2 × 250 chemistry.
Data analysis Mothur v1.32.1 was used for analysis [21]. Paired end reads were aligned and sequence libraries were cleaned by removing sequences not consistent with the target amplicon size (240 bp), those containing any ambiguous base calls or long runs (>8 bp) of holopolymers and those that did not align with the correct 16 s rRNA gene region. Sequences were aligned using the Silva 16S rRNA reference database (http://www.arb -silva.de) [22]. Chimeras were detected using uchime [23] and removed. Taxonomy was assigned using the RDP taxonomy database (http://rdp .cme.msu.edu/index.jsp). Sequences not classified as bacteria were also removed. Coverage was assessed using Good’s coverage and rarefaction curves were generated. Relative abundances were calculated and compared using the Wilcoxon test, logistic regression or Steel–Dwass multiple comparisons test, with a P value ≤0.05 considered significant. Sequences were binned into operational taxon units (OTUs) at a 3% dissimilarity level. Because of the potential impact of differences in sequence numbers of results of various indices, subsampling was performed to normalise sequence number for subsequent analyses [24]. This involved random selection of a number of sequences for each sample that corresponded to the sample size of the sample with the lowest sequence numbers. Population diversity (an assessment of richness and
642
J. S. Weese et al.
relative abundance of different members of the population) and evenness (an assessment of the similarity of relative abundances of different members of the population) were calculated using the inverse Simpson’s and Shannon’s evenness indices, respectively. Dendrograms were developed based on the Yue and Clayton measure of dissimilarity (a measure of community structure, which considers shared OTUs and their relative abundances) and traditional Jaccard index (a measure of community membership, which just considers the number of shared OTUs, not their abundance). Parsimony test was applied to compare microbial populations using both the Yue and Clayton and Jaccard trees. Indicator analysis and principal coordinate analysis (PCoA) were also performed. The core faecal microbiota was assessed through identification of OTUs present in all samples at a minimum relative abundance of 1%. Linear discriminant analysis effect size [25] was also performed to identify genera that were enriched in samples that preceded large colon volvulus vs. controls.
Results Initially, 221 mares were enrolled. The pregnant mare group were mean ± s.d. age 10 ± 4 years. All of the mares were clinically normal throughout the study period and none received antimicrobials. Twenty-four of the 221 pregnant mares (11%) developed post partum colic. Faecal samples prior to the colic episode were collected for 19 of the 24 horses, and samples from 13 of these were available for microbiota analysis. Colic cases were matched with pregnant mares from the same farm that foaled during the same month, but did not develop colic, to control for the effects of farm and time of year. Faecal samples from the 5 nonpregnant control mares were selected based on farm and date the sample was collected. The mean and median ages of these mares were 11.9 and 11.5 years, respectively, with an s.d. of 5.5 years. All had foals in previous years, were clinically normal and received no medications during the study period. After all quality control and sequence cleaning steps, 4,523,727 sequences from 85 samples remained (mean ± s.d. 53,220 ± 29,160, range 8442–122,535). There were 43 samples from 13 mares that developed post partum colic (range 1–4 samples/mare, median 4), 37 samples from 13 mares that did not develop post partum colic (range 1–4, median 2) and 5 samples from 5 nonpregnant control mares. The colic group consisted of 4 mares with 360° large colon volvulus (LCV), 3 with ‘gas colic’, 3 with unspecified ‘medical colic’, one nephrosplenic entrapment, one large colon impaction and one with colic attributed to enterocolitis. Preparturition samples were collected 6–27 days prior to foaling (mean ± s.d. 16.0 ± 6.7 days). Sequences were assigned to 25 different phyla, although only 4 (Firmicutes, Verrucomicrobia, Actinobacteria and Proteobacteria) were present at a relative abundance of 1% or greater overall (Fig 1). The 25 phyla were further assigned to 61 classes, 112 orders, 251 families and 765 genera. Overall, 7.6% of sequences were unclassified at the phylum level. The 10 most abundant classes, orders, families and genera are presented in Table 1. A random subsample of 8442 sequences per sample was used to normalise groups for subsequent comparisons. Rarefaction curves for the subsampled population are displayed in Fig 2. Novel OTUs were still being identified at the end of subsampling, indicating lack of complete sampling effort; however, the rate of new OTU discovery was relatively limited at that point.
Comparison of prepartum and control samples The faecal microbiota of mares during late pregnancy was different from nonpregnant control mares, with a significant difference based on Parsimony analysis of both the Jaccard (P = 0.02) and Yue and Clayton (P = 0.02; Fig 3) trees. There was no difference in diversity between the prepartum and control groups (inverse Simpson’s 105 vs. 46.8, respectively, P = 0.09). Similarly, there was no difference in Shannon’s evenness (0.78 vs. 0.72), although this approached significance (P = 0.06). Broader distribution of the prepartum samples can also be visualised by PCoA (Fig 4). There were no differences in the relative abundances of major phyla between groups, with the only phylum difference being an increase in Equine Veterinary Journal 47 (2015) 641–649 © 2014 EVJ Ltd
Faecal microbiota in post partum colic
J. S. Weese et al.
Firmicutes Verrucomicrobia Unclassified Planctomycetes Actinobacteria Proteobacteria Spirochaetes Bacteroidetes Fibrobacteres
1.0
Relative abundance
0.8
0.6
0.4
0.2
Fig 1: Depiction of the relative abundances of the most common phyla in the faecal microbiota of 26 mares prior to parturition pre- and post partum (Samples 1–3), and 5 nonpregnant controls.
0.0
1
Spirochaetes (0.67 vs. 0.20%, P = 0.04) in prepartum samples. There were limited differences in genera, with prepartum samples having more Roseburia (mean relative abundance 0.28 vs. 0.10%, P = 0.01), Treponema (0.66 vs. 0.20%, P = 0.05) and Papillibacter (0.04 vs. 0.02%, P = 0.04).
Comparison of pre- and post partum samples There was limited apparent impact of foaling and the post partum period, with no differences in evenness or diversity between any time points. There were also no statistically significant differences in the relative abundances of any phyla. The limited difference between sample times can be visualised on the Yue and Clayton (Fig 5) and Jaccard (not presented) dendrograms.
Comparison of samples that preceded a colic episode and those that did not Thirteen samples were collected prior to a colic episode by 1–76 days (mean 17.5). Samples from mares that developed colic before the next sampling time had a significantly higher relative abundance of Proteobacteria (8.2 vs. 3.7%, P = 0.0006) compared with samples from mares that did not develop colic by the next sampling time. Eight samples had a relative abundance of Firmicutes of ≤50%, and all of those were from samples that preceded the onset of colic. Additionally, 6 of 7 (86%) samples with a relative abundance of Proteobacteria of ≥4% were precolic samples. However, neither Parsimony nor unifrac tests identified significant differences in community membership or structure (all P = 0.3). There were also no differences in diversity (P = 0.5) or evenness (P = 0.5).
2
3 Group
Control
Pre
Comparison of samples that preceded a colic episode by
Changes in the faecal microbiota of mares precede the development of post partum colic J. S. WEESE*, S. J. HOLCOMBE†, R. M EMBERTSON†, K. A. KURTZ†, H. A. ROESSNER†, M. JALALI and S. E. WISMER‡ Department of Pathobiology, Ontario Veterinary College, University of Guelph, Ontario, Canada † Department of Large Animal Clinical Sciences, College of Veterinary Medicine, Michigan State University, East Lansing, USA ‡ Rood and Riddle Equine Hospital, Lexington, Kentucky, USA. *Correspondence email: [email protected]; Received: 21.04.14; Accepted: 03.09.14
Summary Reasons for performing study: Disruptions in the gastrointestinal microbiota may trigger development of post partum colic. Objectives: To determine the effects of the periparturient period on the faecal microbiome and identify associations between the faecal microbiota and post partum colic. Study design: Longitudinal case–control study. Methods: Pre- and post partum faecal samples were collected from mares on 3 farms in central Kentucky. Next generation sequencing of the V4 region of the 16S rRNA gene was performed on samples from 13 mares that developed colic, 13 mares that did not display colic and 5 nonpregnant controls. Results: There were 4,523,727 sequences from 85 samples evaluated (mean ± s.d. 53,220 ± 29,160, range 8442–122,535). Twenty-five phyla were identified, although only Firmicutes, Verrucomicrobia, Actinobacteria and Proteobacteria were present at a relative abundance of 1% or greater. The faecal microbiota of late-term mares differed from nonpregnant mares, with differences in microbial community membership and structure but not the relative abundance of major phyla. There was limited impact of foaling and the post partum period on the faecal microbiome. Faecal samples obtained from mares prior to episodes of colic had significantly higher relative abundance of Proteobacteria (8.2%, P = 0.0006) compared with samples from mares that did not display colic (3.7%). All samples with a relative abundance of Firmicutes of ≤50% preceded colic, as did 6/7 (86%) samples with >4% Proteobacteria. Differences in microbiota membership and structure were also present between mares that developed large colon volvulus and matched controls that did not have colic. Sixty-one indicator operational taxon units were identified for the control (vs. volvulus) samples, and these were dominated by Lachnospiraceae (n = 38) and Ruminococcaceae (n = 8). Conclusions: Foaling had minimal effects on the mares’ faecal microbiota. Numerous differences in the faecal microbiota preceded colic. Associations between Firmicutes (particularly Lachnospiraceae and Ruminococcaceae) and Proteobacteria and development of colic could lead to measures to predict and prevent colic. The Summary is available in Chinese – see Supporting information. Keywords: horse; faecal microbiome; Firmicutes; Proteobacteria; colic; post partum
Introduction Colic is an important problem in post partum mares [1–3] and large colon volvulus, one of the most severe forms of colic, is over represented in this group [1,3]. However, the relationships and triggers connecting these kinds of colic with the post partum period are unclear. Snyder and colleagues reported that half of the admissions to University of California at Davis for surgical correction of colonic volvulus were post partum mares that had foaled 2 h and required administration of analgesics. Final diagnoses and treatment (surgical vs. medical) were obtained from veterinarians on each farm and from veterinarians at Rood and Riddle Equine Hospital if the mares were hospitalised.
DNA extraction DNA extraction was performed using a commercial kit that included a bead-beating step (E.Z.N.A. Stool DNA Kita) following the manufacturer’s instructions. Quantity and quality DNA were accessed by spectrophotometryb.
16S rRNA gene amplification and sequencing The V4 region of the 16S rRNA gene was amplified using the primers S-D-Bact-0564-a-S-15 (5'-AYTGGGYDTAAAGNG-3') and S-D-Bact-0785b-A-18 (5'-TACNVGGGTATCTAATCC-3') [20]. Primers contained an overlapping region of the forward and reverse Illumina sequencing primers (TCGTCGGCAGCGTCAGATGTGTATAAGAGACAG and GTCTCGTGGGCTCGGA GATGTGTATAAGAGACAG, respectively) to allow them to anneal to primers containing the Illumina adaptors plus the 8 bp identifier indices (forward: AATGATACGGCGACCACCGAGATCTACAC-index-TCGTCGGCAGCGTC; reverse: CAAGCAGAAGACGGCATACGAGAT-index-GTCTCGTGGGCTCGG). A 48 μl reaction was performed with 25 μl of Kapa2G Fast HotStart ReadyMix 2Xc; 1.3 μl of BSAd; 18.9 μl of PCR-grade H2O, 2 μl of DNA template and 0.4 μl of both the forward and the reverse 16S primers (10 pmol/μl). The following PCR conditions were used; 3 min at 94°C for denaturing, and 30 cycles of 45 s at 94°C for denaturing, 60 s at 50°C for annealing and 90 s at 72°C for elongation followed by a final period of 10 min at 72°C and kept at 4°C until purification. Polymerase chain reaction products were purified (Agencourt AMPure XP beads)e, quantified by spectrophotometryb and normalised to a final concentration of 2 nmol/l. Sequencing of the library pool was performed at the University of Guelph’s Advanced Analysis Centre using an Illumina MiSeqf and 2 × 250 chemistry.
Data analysis Mothur v1.32.1 was used for analysis [21]. Paired end reads were aligned and sequence libraries were cleaned by removing sequences not consistent with the target amplicon size (240 bp), those containing any ambiguous base calls or long runs (>8 bp) of holopolymers and those that did not align with the correct 16 s rRNA gene region. Sequences were aligned using the Silva 16S rRNA reference database (http://www.arb -silva.de) [22]. Chimeras were detected using uchime [23] and removed. Taxonomy was assigned using the RDP taxonomy database (http://rdp .cme.msu.edu/index.jsp). Sequences not classified as bacteria were also removed. Coverage was assessed using Good’s coverage and rarefaction curves were generated. Relative abundances were calculated and compared using the Wilcoxon test, logistic regression or Steel–Dwass multiple comparisons test, with a P value ≤0.05 considered significant. Sequences were binned into operational taxon units (OTUs) at a 3% dissimilarity level. Because of the potential impact of differences in sequence numbers of results of various indices, subsampling was performed to normalise sequence number for subsequent analyses [24]. This involved random selection of a number of sequences for each sample that corresponded to the sample size of the sample with the lowest sequence numbers. Population diversity (an assessment of richness and
642
J. S. Weese et al.
relative abundance of different members of the population) and evenness (an assessment of the similarity of relative abundances of different members of the population) were calculated using the inverse Simpson’s and Shannon’s evenness indices, respectively. Dendrograms were developed based on the Yue and Clayton measure of dissimilarity (a measure of community structure, which considers shared OTUs and their relative abundances) and traditional Jaccard index (a measure of community membership, which just considers the number of shared OTUs, not their abundance). Parsimony test was applied to compare microbial populations using both the Yue and Clayton and Jaccard trees. Indicator analysis and principal coordinate analysis (PCoA) were also performed. The core faecal microbiota was assessed through identification of OTUs present in all samples at a minimum relative abundance of 1%. Linear discriminant analysis effect size [25] was also performed to identify genera that were enriched in samples that preceded large colon volvulus vs. controls.
Results Initially, 221 mares were enrolled. The pregnant mare group were mean ± s.d. age 10 ± 4 years. All of the mares were clinically normal throughout the study period and none received antimicrobials. Twenty-four of the 221 pregnant mares (11%) developed post partum colic. Faecal samples prior to the colic episode were collected for 19 of the 24 horses, and samples from 13 of these were available for microbiota analysis. Colic cases were matched with pregnant mares from the same farm that foaled during the same month, but did not develop colic, to control for the effects of farm and time of year. Faecal samples from the 5 nonpregnant control mares were selected based on farm and date the sample was collected. The mean and median ages of these mares were 11.9 and 11.5 years, respectively, with an s.d. of 5.5 years. All had foals in previous years, were clinically normal and received no medications during the study period. After all quality control and sequence cleaning steps, 4,523,727 sequences from 85 samples remained (mean ± s.d. 53,220 ± 29,160, range 8442–122,535). There were 43 samples from 13 mares that developed post partum colic (range 1–4 samples/mare, median 4), 37 samples from 13 mares that did not develop post partum colic (range 1–4, median 2) and 5 samples from 5 nonpregnant control mares. The colic group consisted of 4 mares with 360° large colon volvulus (LCV), 3 with ‘gas colic’, 3 with unspecified ‘medical colic’, one nephrosplenic entrapment, one large colon impaction and one with colic attributed to enterocolitis. Preparturition samples were collected 6–27 days prior to foaling (mean ± s.d. 16.0 ± 6.7 days). Sequences were assigned to 25 different phyla, although only 4 (Firmicutes, Verrucomicrobia, Actinobacteria and Proteobacteria) were present at a relative abundance of 1% or greater overall (Fig 1). The 25 phyla were further assigned to 61 classes, 112 orders, 251 families and 765 genera. Overall, 7.6% of sequences were unclassified at the phylum level. The 10 most abundant classes, orders, families and genera are presented in Table 1. A random subsample of 8442 sequences per sample was used to normalise groups for subsequent comparisons. Rarefaction curves for the subsampled population are displayed in Fig 2. Novel OTUs were still being identified at the end of subsampling, indicating lack of complete sampling effort; however, the rate of new OTU discovery was relatively limited at that point.
Comparison of prepartum and control samples The faecal microbiota of mares during late pregnancy was different from nonpregnant control mares, with a significant difference based on Parsimony analysis of both the Jaccard (P = 0.02) and Yue and Clayton (P = 0.02; Fig 3) trees. There was no difference in diversity between the prepartum and control groups (inverse Simpson’s 105 vs. 46.8, respectively, P = 0.09). Similarly, there was no difference in Shannon’s evenness (0.78 vs. 0.72), although this approached significance (P = 0.06). Broader distribution of the prepartum samples can also be visualised by PCoA (Fig 4). There were no differences in the relative abundances of major phyla between groups, with the only phylum difference being an increase in Equine Veterinary Journal 47 (2015) 641–649 © 2014 EVJ Ltd
Faecal microbiota in post partum colic
J. S. Weese et al.
Firmicutes Verrucomicrobia Unclassified Planctomycetes Actinobacteria Proteobacteria Spirochaetes Bacteroidetes Fibrobacteres
1.0
Relative abundance
0.8
0.6
0.4
0.2
Fig 1: Depiction of the relative abundances of the most common phyla in the faecal microbiota of 26 mares prior to parturition pre- and post partum (Samples 1–3), and 5 nonpregnant controls.
0.0
1
Spirochaetes (0.67 vs. 0.20%, P = 0.04) in prepartum samples. There were limited differences in genera, with prepartum samples having more Roseburia (mean relative abundance 0.28 vs. 0.10%, P = 0.01), Treponema (0.66 vs. 0.20%, P = 0.05) and Papillibacter (0.04 vs. 0.02%, P = 0.04).
Comparison of pre- and post partum samples There was limited apparent impact of foaling and the post partum period, with no differences in evenness or diversity between any time points. There were also no statistically significant differences in the relative abundances of any phyla. The limited difference between sample times can be visualised on the Yue and Clayton (Fig 5) and Jaccard (not presented) dendrograms.
Comparison of samples that preceded a colic episode and those that did not Thirteen samples were collected prior to a colic episode by 1–76 days (mean 17.5). Samples from mares that developed colic before the next sampling time had a significantly higher relative abundance of Proteobacteria (8.2 vs. 3.7%, P = 0.0006) compared with samples from mares that did not develop colic by the next sampling time. Eight samples had a relative abundance of Firmicutes of ≤50%, and all of those were from samples that preceded the onset of colic. Additionally, 6 of 7 (86%) samples with a relative abundance of Proteobacteria of ≥4% were precolic samples. However, neither Parsimony nor unifrac tests identified significant differences in community membership or structure (all P = 0.3). There were also no differences in diversity (P = 0.5) or evenness (P = 0.5).
2
3 Group
Control
Pre
Comparison of samples that preceded a colic episode by
