Comparison of Strategies for Isolating Anaerobic Bacteria from the Porcine Intestine

This work determined that using a combination of isolation methods is necessary to increase the diversity of bacteria recovered from the intestines of monogastric mammals. Direct plating methods have traditionally been used to isolate enteric bacteria, and recent methods (e.g., diffusion methods [i.e., ichip] or differential isolation of endospore-forming bacteria) have been suggested to be superior at increasing diversity, including the recovery of previously uncultured taxa.

ABSTRACT

The isolation of bacteria that represent the diversity of autochthonous taxa in the gastrointestinal tract is necessary to fully ascertain their function, but the majority of bacterial species inhabiting the intestines of mammals are fastidious and thus challenging to isolate. The goal of the current study was to isolate a diverse assemblage of anaerobic bacteria from the intestine of pigs as a model animal and to comparatively examine various novel and traditional isolation strategies. Methods used included long-term enrichments, direct plating, a modified ichip method, as well as ethanol and tyndallization treatments of samples to select for endospore-forming taxa. A total of 234 taxa (91 previously uncultured) comprising 80 genera and 7 phyla were isolated from mucosal and luminal samples from the ileum, cecum, ascending colon, and spiral colon removed from animals under anesthesia. The diversity of bacteria isolated from the large intestine was less than that detected by next-generation sequence analysis. Long-term enrichments yielded the greatest diversity of recovered bacteria (Shannon’s index [SI] = 4.7). Methods designed to isolate endospore-forming bacteria produced the lowest diversity (SI ≤ 2.7), with tyndallization yielding lower diversity than the ethanol method. However, the isolation frequency of previously uncultured bacteria was highest for ethanol-treated samples (41.9%) and the ichip method (32.5%). The goal of recovering a diverse collection of enteric bacteria was achieved. Importantly, the study findings demonstrate that it is necessary to use a combination of methods in concert to isolate bacteria that are representative of the diversity within the intestines of mammals.

IMPORTANCE This work determined that using a combination of anaerobic isolation methods is necessary to increase the diversity of bacteria recovered from the intestines of monogastric mammals. Direct plating methods have traditionally been used to isolate enteric bacteria, and recent methods (e.g., diffusion methods [i.e., ichip] or differential isolation of endospore-forming bacteria) have been suggested to be superior at increasing diversity, including the recovery of previously uncultured taxa. We showed that long-term enrichment of samples using a variety of media isolated the most diverse and novel bacteria. Application of the ichip method delivered a diversity of bacteria similar to those of enrichment and direct plating methods. Methods that selected for endospore-forming bacteria generated collections that differed in composition from those of other methods with reduced diversity. However, the ethanol treatment frequently isolated novel bacteria. By using a combination of methods in concert, a diverse collection of enteric bacteria was generated for ancillary experimentation.

INTRODUCTION

The intestinal microbiota plays a key role in host health via a multitude of mechanisms (1). One essential function of the microbiota is the metabolism of foods not digested by host-produced enzymes (e.g., complex carbohydrates). For instance, the human genome contains a total of 17 putative digestion-related carbohydrate-active enzymes, compared to >15,000 carbohydrate-active enzymes contained within the enteric microbiota (2). Colonization resistance is another crucial function imparted by microorganisms in the intestine. In this regard, various bacteria competitively occupy habitats and niches required for pathogen invasion, such as within intestinal crypts (3) and the metabolism of carbohydrates (4). To fully elucidate bacterial function, it is critical that bacteria are isolated, characterized, and preserved for experimentation. In this way, candidate organisms can be utilized for the development of innovations in industry and health care.

Culture-independent analyses (next-generation sequencing [NGS]) have greatly advanced our understanding of the diversity of bacteria in natural habitats, including mammalian intestines. However, most research has focused on characterizing the fecal microbiota (5), which is not necessarily representative of the autochthonous bacterial taxa found along the intestinal tract of most mammals (6). Furthermore, the information obtained via traditional NGS analysis is limited, in particular with respect to taxonomic resolution and elucidation of bacterial functional (7–9). This limitation is in part because the resolution of conventional NGS methods is limited to a family/genus level of taxonomic resolution, regardless of the primers selected or bioinformatics programs used (10). Most importantly, bacteria are not recovered, thus eliminating the option to further study and use these organisms for in vivo research and biomedical applications.

Swine are commonly used as a monogastric mammalian model for enteric microbiological studies as they are an important livestock species, and they share many microbiological, anatomical, and associated functional (e.g., immunological) characteristics with human beings compared to other organisms such as mice (11). Pork production accounts for $23.4 billion toward the gross domestic product of the United States, with over a quarter of this value due to the export of pork or pork products (12). Pigs are a particularly good human model as they are of comparable size, possess a similar genome, and have a similar diet to people (13, 14). In addition, their high fecundity and tractability facilitate their use as models to study the mammalian immune system (15). Despite the importance of swine as a livestock species and mammalian model, research investigating the culturable diversity and characterization of the microbiota of pigs is relatively stagnant. The culturable microbiota of pigs was originally evaluated in 1979 (16); however, the application of modern culture-dependent (CD) methods to characterize the enteric microbiota has largely been supplanted by culture-independent evaluations in recent years (17–19). Sequence-based analysis of the intestinal microbiota of pigs identified that species of bacteria within the genera Alloprevotella, Blautia, Clostridium, Lactobacillus, Prevotella, Roseburia, and Ruminococcus and the RC9 gut group comprise their core microbiota (8). However, the global culturability of these organisms in pigs has not been confirmed to date. Importantly, bacterial abundance does not necessarily equate with dominant functions (e.g., niche utilization), and taxa that occur at low densities can occupy critical niches important for host health. As such, it is important that bacteria associated with beneficial niche and habitat utilization in pigs are isolated and preserved for functional assessments.

Numerous isolation methods with many variations in methodology and media have been used to isolate anaerobic bacteria from microbiologically complex substrates (20). These methods mimic the general anoxic state of the distal intestinal tract and primarily involve variations on enrichment (20, 21), direct plating (22), membrane diffusion (e.g., ichip) (23), antibiotic selection (22), and the differential isolation of endospore-forming taxa (24). Of note, the ichip is a series of diffusion chambers that enables the transfer of environmental nutrients into the chamber for improved cultivation of fastidious bacteria (25). In addition, the use of a singular bacterial cell within each diffusion chamber mimics the practice of “culture to extinction” commonly used in ruminant microbiology (26). Fenske et al. (22) evaluated direct plating methods, including the use of antibiotics, and heat treatment of samples to elucidate changes to the diversity of enteric bacteria from feral and industrially raised pigs. However, a comparative examination of different isolation methods is currently lacking.

Using a swine model, we hypothesized that the different isolation techniques applied will recover unique assemblages of bacteria and that the methods can be performed in concert to maximize the estimation of the diversity of enteric bacteria recovered. Furthermore, by employing a combination of methods, a collection of bacteria that is highly representative of the diversity of taxa present in the intestine can be achieved. To test these hypotheses, the objectives of this study were to (i) harvest intestinal segments from adolescent pigs (i.e., ileum, cecum, ascending colon, and spiral colon) implementing methods that prevent the infiltration of ambient oxygen; (ii) isolate bacteria using direct plating, long-term enrichments, a modified ichip method, and strategies that select for endospore-forming taxa; (iii) contrast isolation in carbon dioxide- and nitrogen-predominant atmospheres; (iv) apply bioinformatics methods to compare the efficacies of individual isolation methods/conditions; (v) compare the isolated bacteria to taxa detected using a culture-independent method; and (vi) acquire a comprehensive collection of enteric bacteria that is representative of the diversity of the enteric microbiota for downstream analyses (i.e., as a bioresource for functional analyses).

RESULTS

A diverse collection of enteric bacteria was obtained from the intestines of pigs.

A variety of methods were evaluated throughout this study, including direct plating, enrichments, ichip, and endospore germination methods, including ethanol treatment and tyndallization (Fig. 1). Overall, 1,523 bacterial isolates were recovered and identified, which represented 234 taxa, including 80 genera and 7 phyla (Fig. 2). Phyla represented were Actinobacteria, Bacteroidetes, Firmicutes, Fusobacteria, Lentisphaerae, Proteobacteria, and Synergistetes. The collection was dominated by members of the Firmicutes and Proteobacteria, which represented 70.8% and 16.9% of the taxa isolated, respectively.

FIG 1.

Methods used to isolate anaerobic bacteria from the intestines of pigs. (A) Intestinal segments were ligated and removed from live animals and transferred to an anoxic atmosphere. Liberation of mucosal and luminal contents washingwas accomplished by washing in reduced Columbia broth, followed by homogenization. (B) Aliquots were exposed to ethanol (EtOH) or heat and processed by enrichment and direct plating, along with samples not exposed to ethanol or heat. (C) Long-term enrichments (12 weeks) in basal media amended with blood, xylan, or mucus at 37°C in anoxic atmospheres. (D) Streak plating of the enrichment broths onto basal media containing blood, xylan, or mucus. (E) Dilution of the bacterial suspension within the homogenate to elimination (determined by light microscopy using a Petroff-Hausser chamber). (F) Addition of the appropriate dilution into individual cells of the ichip apparatus, followed by adherence of the top nonpermeable membrane to the apparatus. (G) Placement of the ichip apparatus into fresh rumen fluid. The rumen fluid served as a source of nutrients, vitamins, and cofactors, which were able to diffuse into the medium within the ichip chamber via 0.02-μm-diameter pores in the bottom membrane (i.e., but preventing the passage of bacteria). (H) After 12 weeks, the medium in individual ichip chambers was streaked onto Columbia blood agar. (I) Streak plating of the homogenate onto a variety of agar media (see the text and the supplemental material for a description of the media used). Cultures were maintained for 7 days at 37°C in anoxic atmospheres. Nitrogen-predominant (85% N₂, 10% CO₂, and 5% H₂) and carbon dioxide-predominant (i.e., 90% CO₂ and 10% H₂) atmospheres were used, with the exception of the ethanol and tyndallization methods, which were conducted in the nitrogen-predominant atmosphere. Individual bacterial colonies on agar media were streaked for purity and identified by sequencing the 16S rRNA gene, with comparison to bacterial sequences in the Ribosomal Database Project. Isolates were accessioned into the Intestinal Bacterial Collection (IBaC) at the Lethbridge Research and Development Centre, which included storage of bacteria over liquid nitrogen and lyophilization.

FIG 2.

Phylogram of the bacterial species isolated from the ileum, cecum, ascending colon, and spiral colon of pigs. The bacteria were identified as potentially new species (green) (>95 and ≤97% similarity), genera (blue) (>92 and ≤95% similarity), and families (red) (≤92% similarity) compared to 16S rRNA gene sequences in the Ribosomal Database Project (35).

Atmosphere had a minimal influence on the diversity of bacteria isolated.

All isolation methods were conducted within N₂- and CO₂-predominant atmospheres (referred to as N₂ and CO₂ atmospheres hereafter), with the exception of endospore-forming bacterial selection methods (i.e., the tyndallization and ethanol exposure methods), which were conducted only in the N₂ atmosphere to be consistent with that of Browne et al. (24). No difference (P ≥ 0.992) in the diversity of recovered bacteria was observed between methods conducted in the N₂ and CO₂ atmospheres. The UPGMA (unweighted pair group method with arithmetic mean) values of Euclidian distances were compared using the pvclust package in R (27, 28), and similarities were observed among isolation methods regardless of the atmosphere used, with the exception of direct plating (Fig. 3). Bacterial taxa recovered using the ichip, enrichment, and direct plating methods clustered into discrete groups regardless of the isolation atmosphere (P ≥ 0.950). However, bacterial taxa isolated by direct plating onto Columbia blood agar (CBA) or Dehority’s agar containing 5% porcine mucus or 5% xylan within the CO₂ atmosphere clustered with taxa isolated by enrichment (P = 1.000). In contrast, these collections formed distinct clusters if isolated within the N₂ atmosphere (P = 0.890). Given the similarity in the isolation of bacteria in the two atmospheres, data were combined across atmospheres and are presented as such hereafter.

FIG 3.

Dendrograms of Euclidean distances generated through the UPGMA function of the vegan package of R (28). The pvclust package in R was used to generate values of approximately unbiased (AU) (top number) P values (in percentages) and bootstrap probability (BP) (bottom number) values (nboot = 1,000) as measures of certainty for clusters. Encircled letters indicate collections of bacteria where the AU values were ≥95%, indicating that these clusters are strongly supported by the data. Bacterial communities were generated from bacteria isolated from within N₂ or CO₂ atmospheres. Bacteria were isolated using direct plating (DP), enrichment (Enrich), or a modified ichip (Ichip) method; after a 3-h exposure to 70% ethanol (EtOH); or after a 30-min exposure to 100°C (tyndallization [Tyn]). Media used were Bacteroides bile esculin agar (BactBileEsc), Bacteroides bile esculin agar with 100 μg ml⁻¹ gentamicin sulfate (BBEGent), Columbia medium with 10% sheep’s blood (Blood), modified Dehority’s medium with 5% porcine mucus III (Mucus), modified Dehority’s medium with 5% xylan (Xylan), or Columbia agar supplemented with 10% sheep’s blood and 0.1% sodium taurocholate (Blood&Bile). The bracket denoted with A shows clustering of bacterial taxa isolated by enrichment, the bracket denoted with B shows clustering of bacteria isolated using methods that select for endospore-forming taxa, and the bracket denoted with C shows clustering of bacterial taxa isolated using the ichip. Data are combined across intestinal locations.

The isolation strategies used influenced the diversity and composition of bacteria recovered.

Shannon’s index (SI) values differed (χ² = 72.2; P > 0.010) among isolation methods and media, as determined by the Kruskal-Wallis rank sum test in R (29). However, a post hoc test (30) showed no difference (Fig. 4) (P ≥ 0.160) among the methods, supported further by a q test (q = 1) (31). Comparison of bacterial diversity by isolation strategy (i.e., direct plating, enrichment, ichip, tyndallization, and ethanol treatment) across media revealed differences (P < 0.001) among the five strategies employed. This was attributed to the low diversity of bacteria recovered by the methods designed to isolate endospore-forming taxa. The diversity of bacteria isolated by long-term enrichment, direct plating, and ichip methods did not differ (P ≥ 0.989), yielding SI values of ≤2.5, ≤2.2, and ≤1.5, respectively (see Table S1 in the supplemental material). The diversities of bacteria isolated using the two methods that select for endospore-forming taxa (i.e., ethanol treatment and tyndallization) did not differ (P ≥ 0.786), yielding SIs of ≤0.8 and ≤1.3, respectively. The diversity of bacteria isolated using the tyndallization method was lower than those for the enrichment (P = 0.005), direct plating (P = 0.033), and ichip (P = 0.070) methods. A trend toward a lower SI was observed for the ethanol treatment method relative to enrichment method (P = 0.149).

FIG 4.

Distributions of Shannon’s diversity indices of bacteria isolated from pigs using different isolation methods. Bacteria were isolated in this study using direct plating (DP) methods, enrichments (Enrich), or a modified ichip (Ichip) method; after a 3-h exposure to 70% ethanol (EtOH); or after a 30-min exposure to 100°C (Tyn). Media used were Bacteroides bile esculin agar (BactBileEsc), Bacteroides bile esculin agar with 100 μg ml⁻¹ gentamicin sulfate (BBEGent), Columbia medium with 10% sheep’s blood (Blood), modified Dehority’s agar with 5% porcine mucus III (Mucus), modified Dehority’s medium with 5% xylan (Xylan), or Columbia agar with 10% sheep’s blood and 0.1% sodium taurocholate (Blood&Bile). The center lines in the box plot represent the median value, the size of the box plot represents the distribution within a confidence of 95%, and the vertical lines and dots associated with the box plot represent the total distribution of the data. The box plots were generated using the geom_boxplot function of the ggplot package in R, and Shannon’s diversity was determined using the vegan package of R (28). Data are combined across intestinal locations.

Bacterial taxa recovered using individual isolation strategies differed.

The evaluation of Euclidian distances (28) revealed that the different isolation strategies recovered dissimilar bacterial taxa (Fig. S1). The application of pvclust (27) confirmed this observation (Fig. 5). Bacterial taxa isolated by long-term enrichment and direct plating methods clustered together (P = 0.870), whereas bacteria recovered using the ichip formed a unique clade (P < 0.050). Bacteria recovered by tyndallization and the ethanol treatment clustered together regardless of the isolation method employed posttreatment (P = 1.000). An examination of the 25 most frequently isolated taxa revealed that Bacillus spp. were most commonly isolated using the tyndallization and ethanol treatment methods (Fig. 6). In contrast, Lactobacillus spp. were most abundantly isolated using the ethanol method. The use of long-term enrichments, direct plating, and the ichip method resulted in the isolation of diverse genera not within the top 25 taxa recovered (i.e., “other bacteria”). Direct plating methods were commonly associated with the isolation of Escherichia/Shigella spp. However, the inclusion of gentamicin into CBA for direct plating assisted in the isolation of Bacteroides spp. A comparison of the specific recoveries of bacterial phyla revealed that the enrichment methods recovered bacteria from all seven phyla (Fig. 7). Direct plating, the ichip method, and endospore selection methods recovered bacteria from six, four, and three phyla, respectively.

FIG 5.

Dendrograms of Euclidean distances generated through the UPGMA function of the vegan package of R (28). The pvclust package in R was used to generate values of approximately unbiased (AU) (top number) P values (in percentages) and bootstrap probability (BP) (bottom number) values (nboot = 1,000) as measures of certainty for clusters. Encircled letters indicate collections of bacteria where the AU values were ≥95%, indicating that these clusters are strongly supported by the data. Bacteria were isolated using direct plating (DP), enrichment (Enrich), or a modified ichip (Ichip) method; after a 3-h exposure to 70% ethanol (EtOH); or after a 30 min exposure to 100°C (Tyn). Media used were Bacteroides bile esculin agar (BactBileEsc), Bacteroides bile esculin agar with 100 μg ml⁻¹ gentamicin sulfate (BBEGent), Columbia medium with 10% sheep’s blood (Blood), modified Dehority’s medium with 5% porcine mucus III (Mucus), modified Dehority’s medium with 5% xylan (Xylan), or Columbia agar with 10% sheep’s blood and 0.1% sodium taurocholate (Blood&Bile). The bracket denoted with A shows clustering of bacterial taxa isolated by enrichment, and the bracket denoted with B shows clustering of bacteria isolated using methods that select for endospore-forming taxa. Data are combined across intestinal locations.

FIG 6.

Heat map of Euclidean distances and relative abundances of the 25 most abundantly isolated bacterial genera as well as “other” bacteria. Bacteria were isolated using direct plating (DP), enrichment (Enrich), or a modified ichip (Ichip) method; after a 3-h exposure to 70% ethanol (EtOH); or after a 30-min exposure to 100°C (Tyn). Media used were Bacteroides bile esculin agar (BactBileEsc), Bacteroides bile esculin agar with 100 μg ml⁻¹ gentamicin sulfate (BBEGent), Columbia medium with 10% sheep’s blood (Blood), modified Dehority’s medium with 5% porcine mucus III (Mucus), modified Dehority’s medium with 5% xylan (Xylan), or Columbia agar supplemented with 10% sheep’s blood and 0.1% sodium taurocholate (Blood&Bile). The heat map and dendrogram were generated using the heatmap and Hclust functions in the vegan package of R (28). Bacteria were isolated from the ileum, cecum, ascending colon, and spiral colon. Data are combined across intestinal locations.

FIG 7.

Cladograms showing the distribution of bacterial taxa isolated by an individual isolation method (red circles) relative to bacteria isolated using other methods (gold circles). Bacteria were isolated using direct plating (A), enrichment (B), a modified ichip method (C), after a 3-h exposure to 70% ethanol (D), and after a 30-min exposure to 100°C (E). Taxonomic levels in the cladograms range from superkingdom (in the center) to species (at the perimeter). Cladograms were generated using the online LEfSe tool available at https://huttenhower.sph.harvard.edu/galaxy.

To further evaluate the taxon selection by isolation strategy, the linear discriminant analysis effect size (LEfSe) algorithm was applied at a class level of resolution (32). LEfSe determined that the direct plating method was the most successful for the isolation of Erysipelotrichia and Gammaproteobacteria (Fig. S2). The enrichment methods (enriched on blood, mucus, or xylan) were associated with an increased abundance of Clostridia, whereas the ichip method was associated with increased isolation of Actinobacteria. It is noteworthy that of the methods applied to select for endospore-forming taxa, bacteria within the class Bacilli were commonly recovered using both ethanol treatment and tyndallization (primarily Bacillus species), whereas Lactobacillus spp. were commonly isolated following ethanol treatment.

Isolation methods yielded differing proportions of novel species, genera, and families.

Bacteria potentially representing new species, genera, and families were determined by evaluating their similarities to 16S rRNA sequences contained in the Ribosomal Database Project (RDP); specifically, new species were defined as those being >95 and ≤97% similar, new genera were defined as being >92 and ≤95% similar, and new families were defined as being ≤92% similar to sequences contained in the RDP. The ichip method was found to yield the most bacterial taxa belonging to novel families (24.1%) (Fig. 8). Long-term enrichment yielded large numbers of novel bacterial species, and enrichments containing blood, xylan, and mucus recovered 39.9, 31.1, and 26.8% novel species, respectively. Direct plating on CBA (DP-blood) was the most effective direct plating medium for isolating novel taxa, and 24.5% of the recovered isolates were identified as novel species. Strategies designed to select taxa that form endospores yielded the smallest numbers of novel species, with ≤21.8% and ≤7.2% novel taxa isolated with the ethanol treatment and tyndallization methods, respectively. Novel species were isolated at similar frequencies throughout the intestinal tract. However, when comparing the presence of novel species, genera, and families, the ileum tended to yield fewer members of novel families than locations in the large intestine (8.0% versus >13% frequency of isolating novel taxa) (Fig. S3).

FIG 8.

Frequency of novel bacterial taxa (percent) recovered using the evaluated isolation methods and media. Novel taxa were delineated at >95 and ≤97% similarity (species), >92 and ≤95% similarity (genus), and ≤92% similarity (family and higher) relative to 16S rRNA gene sequences within the Ribosomal Database Project (35). Bacteria were isolated using direct plating (DP), enrichment (Enrich), a modified ichip (Ichip) method, after a 3-h exposure to 70% ethanol (EtOH), or after a 30-min exposure to 100°C (Tyn). Media used were Bacteroides bile esculin agar (BactBileEsc), Bacteroides bile esculin agar with 100 μg ml⁻¹ gentamicin sulfate (BBEGent), Columbia medium with 10% sheep’s blood (Blood), modified Dehority’s medium with 5% porcine mucus III (Mucus), modified Dehority’s medium with 5% xylan (Xylan), or Columbia agar with 10% sheep’s blood and 0.1% sodium taurocholate (Blood&Bile). Bacteria were isolated from the ileum, cecum, ascending colon, and spiral colon. Data are combined across intestinal locations.

Specific methods favored the isolation of novel bacteria.

Unique novel bacteria isolated during this study were found most abundantly from enrichment methods (Fig. S4). The ethanol spore germination method and ichip methods were responsible for the isolation of 13 novel bacterial species, while direct plating was responsible for 12 novel bacteria. Overlapping novel bacterial species were observed between the enrichment and direct plating methods as well as between the enrichment and ichip methods. In both cases, six novel bacterial species were found to overlap between methods. The method associated with the isolation of the fewest novel bacterial species was tyndallization, which was associated with the isolation of seven bacterial species in total.

Most novel bacteria that were isolated were recovered from specific intestinal locations. Enrichment methods isolated 13, 11, and 11 novel species from the ileum, cecum, and spiral colon, respectively (Fig. S5). In contrast, the most novel bacteria isolated by direct plating were from the spiral colon (n = 9), cecum (n = 5), and ileum (n = 3). Although many bacteria were observed to be location specific, others were recovered from multiple locations. For example, 22 novel species, recovered using enrichment, were observed across locations. Although the ileum was a large repository of novel species, the majority of these bacteria were isolated using the enrichment, direct plating, and ichip methods; comparatively few novel taxa were recovered from the ileum using the tyndallization and ethanol treatment methods.

The diversity of isolated bacteria did not differ along the intestinal tract.

There were no differences (χ² = 1.12; P = 0.773) in the diversity of bacteria isolated from the ileum, cecum, ascending colon, and spiral colon (Fig. S6 and Table S2). This observation was confirmed by evaluating q values (q ≥ 0.99) (31).

Bacterial diversity determined by isolation and next-generation sequence analysis differed in the large intestine.

NGS analysis indicated that the α-diversity of communities in the ileum was reduced (P < 0.001) relative to the cecum and spiral colon of pigs (Fig. 9). Furthermore, the β-diversity of bacterial communities in the ileum differed (P < 0.002) from that of communities in the cecum and spiral colon (Fig. S7). Although there was no difference in the α-diversity of bacteria in the ileum as determined by culture detection and NGS analysis (P > 0.970), the diversity of isolated bacteria was low (P < 0.001) in the cecum and spiral colon of pigs relative to the diversity in this location determined by NGS analysis (Fig. 9).

FIG 9.

Shannon’s diversity index of bacteria from/in the intestinal tract of piglets (i.e., ileum, cecum, and spiral colon) as determined by culture detection (CD) and next-generation sequence analysis (NGS). Bacteria were isolated using direct plating, 12-week enrichment, a modified ichip method, ethanol treatment, and tyndallization.

Recovered bacterial taxa differed among intestinal locations.

A comparison of approximately unbiased (AU) P values for UPGMA clustering using the pvclust function in R (27) revealed that bacterial taxa recovered from the cecum, ascending colon, and spiral colon of pigs were similar to each other (P = 0.990) (Fig. S8), but bacteria isolated from the ileum clustered separately from those recovered from the large intestine. In this regard, Escherichia/Shigella spp. were frequently isolated from the ileum, whereas Clostridium sensu stricto, Eubacterium, and Bacteroides spp. were more commonly recovered from the large intestine (Fig. S9). Ancillary analysis of bacterial class abundance using LEfSe (32) revealed that bacteria belonging to Bacilli, Fusobacteriia, and Gammaproteobacteria were more frequently isolated from the ileum (Fig. S10). Members of the Actinobacteria, Bacteroidia, Erysipelotrichia, and Deltaproteobacteria were more commonly recovered from the cecum, while members of the Clostridia, Negativicutes, and Synergistia were most frequently isolated from the ascending colon.

Isolated bacterial taxa overlapped with taxa detected by NGS analysis.

In total, the ileum yielded bacteria belonging to 15 classes (Fig. 10A), which was a considerably smaller number of classes than that observed in the large intestine. Bacteria belonging to a total of 24 and 27 classes were detected in the cecum and spiral colon, respectively (Fig. 10B and C). In the ileum, four classes were detected by NGS analysis but not by isolation, and bacteria belonging to three classes were recovered only by isolation. In the cecum and spiral colon, 10 and 11 bacterial classes were detected by both isolation and NGS analysis, respectively, whereas 12 classes from the cecum and 14 classes from the spiral colon were detected by NGS analysis only. Bacterial belonging to two classes were detected in the cecum and spiral colon only by isolation. Bacteria in the classes Betaproteobacteria, Deltaproteobacteria, and Synergistia (isolated solely from the ileum) were detected only by isolation, whereas bacteria in the classes Alphaproteobacteria, Campylobacteria, Chlamydiae, Coriobacteriia, Deferribacteres, Fibrobacteria, Melainabacteria, Methanobacteria, Mollicutes, Oxyphotobacteria, Spirochaetia, and Thermoplasmata were detected only by NGS analysis. Regardless of the location, bacteria belonging to different classes were most commonly isolated using 12-week enrichments, direct plating, and, to a lesser extent, the modified ichip method.

FIG 10.

Bacterial classes isolated/detected from the intestinal tract of piglets by culture detection (CD) and next-generation sequence analysis (NGS). Culture-dependent methods used were direct plating (DP), 12-week enrichment (Enrich), a modified ichip (Ichip) method, ethanol exposure (EtOH), and tyndallization (Tyn). (A) Ileum; (B) cecum; (C) spiral colon. Black dots indicate positive isolation/detection. The plots were generated using the UpSetR package in R (87) and show the number of bacterial classes by method.

DISCUSSION

The intestinal microbiota of mammals is comprised of diverse bacteria, and the comprehensive isolation of these taxa is required to fully ascertain their function (33). Pioneers in anaerobic microbiology developed and used a variety of methods leading to the creation of manuals such as the VPI anaerobic manual (34). However, a comparative analysis of various methods including novel strategies is lacking. We hypothesized that a combination of isolation methods used in concert is necessary to comprehensively recover bacteria from the intestines of swine as a model mammal. We applied strict anaerobic conditions and compared a variety of isolation strategies, including direct plating, long-term enrichments, and an ichip method modified to recover enteric bacteria, along with methods designed to isolate endospore-forming bacteria. Study findings supported the study hypothesis, indicating that individual techniques select for different bacteria. In total, 234 taxa comprising 80 genera and 7 phyla were isolated from the intestinal tract of pigs, with 24% of the bacteria representing taxa that were previously undescribed compared to taxa within the RDP (35).

The gaseous atmosphere within the intestinal lumen varies in composition spatially. While published data are lacking for pigs, the gaseous atmosphere in the proximal gastrointestinal tract (GIT) of human beings is comprised of ≤10% O₂, 20 to 90% N₂, 10 to 30% CO₂, ≤50% H₂, and ≤10% CH₄ (36). As digesta move through the GIT, the residual O₂ that is present within the lumen of the large intestine is rapidly utilized by facultative anaerobes (37), resulting in an ecosystem that is devoid of O₂ (38). Many enteric bacteria in their vegetative states are exceptionally sensitive to O₂ exposure (24). Although substrates such as digesta can remain in a reduced state for a period of time following removal from the intestine (39), exposure to O₂ in an ambient atmosphere can result in the death of oxygen-sensitive bacteria, thereby reducing the diversity of bacteria recovered (24). For this reason, we employed methods to preclude any exposure of bacteria to O₂. Specifically, we ligated intestinal segments in living animals under general anesthesia, excised the segments, placed the segments in an atmosphere devoid of O₂, transferred the segments to an anaerobic chamber within 60 min of removal of the segments, and processed the samples within a CO₂ or N₂ atmosphere. In contrast to our approach, other researchers such as Fenske et al. (22) collected luminal contents by squeezing digesta into tubes containing 40% anaerobic glycerol followed by flash-freezing of the samples. Although they successfully isolated a diverse assemblage of bacteria, the opportunity for O₂ infiltration coupled with freezing of samples could result in the death of bacterial cells (22), particularly of fastidious bacterial taxa present at low cell densities. The care that we extended in precluding exposure to O₂ may have contributed to the high diversity of bacteria that we isolated from the intestine of pigs.

Traditionally, the isolation of anaerobic bacteria from the distal intestine of animals has relied on the use of atmospheres that are either CO₂ or N₂ predominant. For example, the original isolation of enteric bacteria from pigs was conducted in an atmosphere comprised of 80% N₂, 10% H₂, and 10% CO₂ (16). The impacts of atmosphere on bacterial selection as a function of the isolation method employed have not been extensively examined, and we hypothesized that using multiple anoxic atmospheres for direct plating and enrichment of bacteria would substantially increase the diversity of bacteria recovered. Furthermore, we predicted that the isolation of fastidious bacteria would be favored in the CO₂ atmosphere, as a high-CO₂-content atmosphere occurs in the intestine (36). However, the compositions of bacteria recovered between the CO₂ and N₂ atmospheres using enrichment methods were largely similar. This may be attributed to the production of CO₂ during static bacterial growth, which would have generated a predominantly CO₂ atmosphere within the enclosed enrichment broths and the ichip (36, 40). High-CO₂ atmospheres have been shown to differentially impact the growth of specific bacteria. For example, Streptococcus faecalis exhibits improved growth in a CO₂ atmosphere, whereas the growth of Bacillus cereus is inhibited (41). In addition, the total absence of CO₂ is detrimental to the growth of many bacteria such as Bacteroides fragilis (42). Consistent with the possibility that CO₂ generation during enrichment negated differences between the two atmospheres, we observed that bacterial taxa isolated on agar media in petri dishes differed between the N₂ and CO₂ atmospheres, which has been observed previously (43). Genera that we isolated more frequently using direct plating within the CO₂ atmosphere included members of the Clostridium sensu stricto and Clostridium XIVa groups as well as Eubacterium, Bacteroides, and Christensenella spp. Thus, our findings support the use of a single atmosphere for conducting isolations using the ichip and enrichment methods (i.e., the isolation method and not the atmosphere was associated with the composition of bacteria isolated); however, isolations using direct plating should employ both atmospheres.

Enrichment isolation is a frequently used method to isolate bacteria from microbiologically complex samples. A major limitation of enrichment isolation is that fast-growing taxa and subtypes within taxa can be differentially favored, which can result in a reduced diversity of bacteria recovered. In the current study, we used long-term enrichments, and we observed that this strategy recovered the most diverse collection of bacteria. This result was unexpected as more modern methods, such as the ichip or those that select for endospore-forming bacteria, were anticipated to outperform traditional isolation methods such as enrichment (24, 44). A possible explanation for the high diversity of bacteria that we recovered using enrichment is the extended duration of the enrichment step that we employed (12 weeks) coupled with the diversity of media used. It is noteworthy that the 12-week enrichment period that we used is substantially longer than the ≤6-week period typically reported in the literature (45). Previous research using an 8-week enrichment period for cattle feces recovered a similarly high diversity of bacteria (21). However, the 8-week enrichment period employed by Ziemer (21) recovered novel bacteria at a frequency rate of 98%, which was higher than the 39% rate observed in the current study; this may be attributed to her utilization of continuous culture compared to static enrichment (21). The media used for enrichments in the current study were supplemented with blood, mucus, or xylan. These amendments were selected for individual reasons, including (i) their role as an antinutritive component of the pig diet that has been shown to be effective at isolating novel bacteria during enrichments (i.e., xylan) (21, 46), (ii) their routine use for isolating bacteria (i.e., blood) (47, 48), and (iii) because they provide limiting glycans within the intestinal environment (i.e., mucus) (49). The high diversity of bacteria that we recovered from the intestine of pigs using long-term enrichments validates the effectiveness of this strategy and emphasizes the importance of extending the duration of the enrichment period.

Many bacterial taxa within the Firmicutes form endospores, and these taxa are difficult to isolate due to poor endospore germination coupled with the predominance of non-endospore-producing bacteria, which obscure colonies emanating from endospores. The use of methods that select for endospore-forming bacteria has been promoted as a way to improve the isolation of novel taxa (24). We applied the ethanol treatment method described by Browne et al. (24) as well as tyndallization. The tyndallization method was originally developed as a method to kill endospores in foods (50). The method works by killing vegetative cells and stimulating the germination of endospores by heat exposure at atmospheric pressure. We observed that both of the methods evaluated, regardless of the postisolation strategy applied, isolated similar bacterial taxa and overwhelmingly isolated bacteria known to produce endospores. Most notably, Bacillus licheniformis was frequently isolated and represented ∼50% of the bacteria recovered. We observed that the diversity of bacteria isolated following ethanol treatment was higher than that found following tyndallization, suggesting that the heat treatment did not appreciably stimulate endospore germination. It is noteworthy that Lactobacillus johnsonii, a non-endospore-forming bacterium, was commonly isolated using the ethanol method but was absent from tyndallized samples. Tolerance to ethanol has been observed within Lactobacillus spp. (51), which may explain the isolation of L. johnsonii using the ethanol method. Although we did not isolate Lactobacillus spp. from heat-treated samples, Fenske et al. (22) frequently isolated lactobacilli (e.g., Lactobacillus reuteri) from Tamworth pigs after exposure of digesta samples to heat followed by direct plating. In contrast to the report of Browne et al. (24), the diversity of bacteria that we recovered using endospore selections was comparatively low. Fenske et al. (22) similarly observed that the application of the heat treatment method resulted in a change in the community of bacteria isolated compared to other traditional direct plating strategies. In the current study, bacterial isolates were selected based on colony morphology and were subsequently identified by sequencing the 16S rRNA gene of the isolated bacteria. This, coupled with the overabundance of dominant isolates, may have reduced the diversity of bacteria that we observed for the ethanol exposure and tyndallization methods. In this regard, the utilization of methods that allow the screening of a larger number of isolates, such as matrix-assisted laser desorption ionization–time of flight (MALDI-TOF) instrumentation, may identify novel endospore-forming taxa that are present at low frequencies (52). Although the diversity of bacteria isolated by endospore-selective methods was low relative to the other methods examined in the current study, novel taxa were recovered at rates of 21.8% and 7.1% for the ethanol and tyndallization methods, respectively. This illustrates the value of using methods that select for endospore-forming taxa, which may be enhanced with the use of characterization strategies such as the MALDI-TOF method (52, 53). Importantly, our findings show that relying exclusively on methods that select for endospore-forming taxa is insufficient to isolate bacteria that are representative of the taxa present in the intestine of pigs and likely other animals.

The ichip method has been shown to efficiently isolate novel taxa from soil and other environments, including from the oral cavity of human beings (54, 55). A salient advantage of the ichip method is that slow-growing and nutritionally fastidious bacterial taxa are allowed to proliferate in the enrichment medium independent of other faster-growing or nutritionally competitive bacteria, thereby facilitating their isolation. We adapted the ichip method for the isolation of enteric bacteria by modifying the enrichment material for their isolation. The use of rumen fluid has not been previously used with the ichip, and this substrate is a source of rare-occurring vitamins, cofactors, and other nutrients that are required for the growth of fastidious bacteria (45, 56). Fresh rumen fluid can be easily harvested in relatively large quantities from cannulated cattle, and nutrients are abundant in the liquid fraction (i.e., compared to pig feces or digesta obtained from the large intestine). Furthermore, digesta from the distal intestine of pigs (e.g., cecum, ascending colon, and spiral colon) would be expected to contain fewer nutrients and growth factors, and feces would be expected to contain even fewer due to their assimilation by microorganisms and the host. Using the ichip method, relatively diverse bacterial taxa were isolated from the intestines of pigs, but the bacteria isolated represented only four of the seven phyla recovered in the study. Within these four phyla, however, 54 bacterial species were previously uncultured relative to taxa within the RDP (35). Using the ichip method, 26% of recovered bacterial isolates from the intestines of pigs were previously uncultured taxa. Although the value of the ichip method for isolating novel bacteria from a variety of substrates has been previously demonstrated (e.g., from soil and the human oral cavity) (44, 54, 55), it has not previously been applied to isolate bacteria from intestines. Although a number of novel taxa were isolated using the modified ichip in the current study, the method was substantially more labor-intensive than other methods such as long-term enrichment. Moreover, used on its own, the ichip method did not reveal the diversity of bacteria that was present in the intestine of pigs. Thus, the method should be used in combination with other isolation strategies.

The direct plating method is commonly used to isolate bacteria from the intestines of animals and has the primary advantage of tailoring the isolation strategy with specific carbohydrates or antibiotics (22). In addition, bacteria can be isolated using minimal media to isolate carbohydrate-specific bacteria, such as xyloglucan (57). However, carbohydrate-specific bacteria can also be determined after isolation using a complex medium (58). In addition, the use of blood for the isolation of bacteria has been shown to be both efficient and cost-effective (59). It is noteworthy that we used Columbia broth (CB)/Columbia agar (CA) as a basal medium based on our experience that it supports the growth of most anaerobic taxa in our collection, and it has been successfully used to isolate anaerobic enteric bacteria previously (60). We observed that the composition of the bacteria recovered by direct plating conducted in the CO₂ atmosphere was similar to that recovered by long-term enrichment; however, the diversity of bacteria isolated by direct plating was reduced. As discussed above, the direct plating method was observed to be more affected by the isolation atmosphere than any other method, and the composition of bacteria isolated by direct plating within the CO₂ atmosphere was similar to that obtained by long-term enrichment. In an attempt to optimize the isolation of slower-growing bacteria, we maintained cultures for 7 days. We also utilized a variety of growth factors in media. In this regard, direct plating onto media containing 100 μg ml⁻¹ gentamicin sulfate altered the composition of bacteria collected, largely because of the selection of Bacteroides spp. This is expected as gentamicin at 200 μg ml⁻¹ is often used for this purpose (61). We chose to use a lower concentration of gentamicin in an attempt to isolate members of the Firmicutes phylum, such as Clostridium and Megasphaera spp., which were observed previously in chickens (60). This concentration is the suggested amount contained in Bacteroides bile esculin agar, which has been used successfully for isolations from pigs (61, 62). Although the diversity of bacteria recovered by direct plating in the current study was found to be lower than that recovered by the enrichment or ichip methods, the use of a variety of media resulted in the isolation of bacteria representing six of the seven phyla recovered in the study. Our findings show that direct plating recovered unique taxa, but if the goal is to isolate bacteria that represent the diversity of taxa present in the intestine, then relying on direct plating alone is inadequate.

Characterization of bacteria using NGS analysis has shown that the structure of communities differs between the distal small intestine and large intestine of mammals (63, 64). We observed that the diversity of bacteria determined by NGS analysis was higher in the cecum and spiral colon than that determined by CD methods, but there was no difference between the two strategies for the ileum. Cummings et al. (65) compared standard isolation methods using direct plating techniques to NGS-based detection in bronchoalveolar lavage fluids. They observed that NGS more frequently revealed anaerobic fastidious bacteria than direct plating. However, more specific characterizations could be ascertained by direct plating in samples with low bacterial densities. When comparing the overlapping taxa observed in our study, similar results were observed as NGS identified four classes of bacteria in the ileum that were not observed by CD methods. This contrasted with the more diverse communities of bacteria in the spiral colon, where 14 bacterial classes were observed only by NGS analysis.

Using CD methods, the composition of the isolated bacterial taxa differed in the ileum in comparison to the cecum, ascending colon, and spiral colon of pigs. This variation in bacterial structure is due in part to the higher concentrations of O₂ in the small intestine (38). We observed that the culturable microbiota of the ileum of swine was predominated by bacteria within the Proteobacteria, which is consistent with NGS results (66). The growth of Proteobacteria in the small intestine is favored by their ability to utilize O₂ as a terminal electron acceptor (i.e., many taxa are facultative anaerobes), coupled with their relative tolerance to acidic conditions (67, 68). We also isolated Firmicutes bacteria from the ileum. Bacteria belonging to the Firmicutes are more prevalent in the ileum than the jejunum, possibly as a result of the introduction of bacteria into the ileum from the cecum via the ileal-cecal junction (69). At present, it is unclear if Firmicutes taxa are autochthonous in the ileum and if they have an important functional role in the distal small intestine of pigs. Although the merits of NGS approaches are recognized, the comprehensive isolation of bacteria from the GIT is crucial to elucidate their ecological role in the intestine, including the characterization of niches that individual taxa, and subtypes within taxa, occupy.

The establishment of anaerobic bacterial collections requires considerable microbiological expertise and a specialized infrastructure to isolate, characterize, and store diverse taxa. Moreover, the maintenance of culture collections requires a significant expenditure in operational funds. Over the past decade, emphasis has focused on the characterization of enteric communities using NGS as opposed to CD analysis (70). It is noteworthy that the structures of the intestinal microbiota of pigs infected with Salmonella enterica serovar Typhimurium examined by NGS and CD evaluations did not correspond and that bacteria possibly conferring colonization resistance were not detected by NGS (71). More recently, research has been directed toward utilizing advanced methods to elucidate function (e.g., transcriptomics and metabolomics). Although these methods have provided key information on community function, access to living anaerobic bacteria allows complementary research to further ascertain function. For example, genetic variation among strains in relation to function and empirical evaluation using gnotobiotic animal models can provide information that cannot be obtained using metagenomics methods. However, the use of metagenomic assembled genomes is closing this gap as it enables the “binning” of specific nucleotide segments to generate and construct genomes in silico, enabling the identification of specific genes associated with bacterial taxa (72). Crucial to empirical functional assessments is access to bacteria that represent the diversity of autochthonous taxa that exist in the intestine, including from specific locations/habitats. This requires the concerted application of culture-dependent methodologies. The findings of the current study show that diverse bacteria can be successfully recovered from the intestines of pigs (i.e., as a monogastric mammalian model). However, reliance on a single or limited number of isolation methods is insufficient, and it is necessary to utilize a number of strategies in concert. In this regard, we applied strict anaerobic microbiological methods and used a number of isolation strategies, including direct plating, enrichment, ichip, and ethanol treatment and tyndallization, to differentially recover endospore-forming taxa. Using these methods, we recovered in excess of 200 species of bacteria belonging to 7 phyla, including many previously undescribed bacteria. There are 500 to 1,000 bacterial species present in the GIT of human beings (73–75). Comparisons between humans and pigs have revealed similar diversity and richness between animals, with 96% shared pathways between the microbiota of these animals, indicating a conserved structure and function of the adapted microbiota of humans and pigs (11). Although a relatively limited number of taxa are thought to contribute disproportionately to the total numbers of bacterial cells present in the intestine, bacterial abundance is often not correlated with function (76, 77). This emphasizes the importance of continuing efforts to isolate, characterize, and store bacterial taxa present in the GIT as a bioresource for functional analyses. It is noteworthy that the cost of completing the isolations, identifications, and subsequent storage of bacteria in the current study is in line with the cost of completing advanced molecular-based approaches (i.e., metagenomic assembled genomes). Even with the rapid advancement of culture-independent technologies and the expected decrease in the cost of conducting these analyses, the acquisition of actual bacteria is necessary to fully elucidate function toward innovation achievement (i.e., in conjunction with “omic” approaches). In conclusion, the comprehensive isolation and acquisition of bacteria from the GIT of mammals must target multiple locations and incorporate a variety of isolation approaches beyond short-term direct plating or enrichment isolation culture.

MATERIALS AND METHODS

Ethics approvals.

Approval to utilize pigs in the study was obtained from the Agriculture and Agri-Food Canada (AAFC) Lethbridge Research and Development Centre (LeRDC) Animal Care Committee (ACC) before the commencement of the experiment (animal use protocol number 1512). In addition, approvals to collect blood from sheep for use in microbiological media (animal use protocol number 1922) and to collect rumen fluid from cannulated cattle (animal use protocol number 1909) were also obtained from the LeRDC ACC in advance.

Animals and husbandry.

Twenty-four castrated large white Landrace cross pigs (6 weeks of age) were used in the experiment. The pigs were not exposed to antibiotics at any point, and the dam was not administered antibiotics for at least 1 year before parturition. Pigs were transferred to the Livestock Containment Unit and allowed to acclimatize for 1 week within the facility before the commencement of the experiment. Animals were provided a minipellet ration diet that was free of antibiotics (Proform Pig Starter 2; Hi-Pro Feeds, Okotoks, AB, Canada). Feed was provided daily, and pigs were permitted to eat and drink ad libitum. Straw was used for bedding, and toys were provided for environmental enrichment.

Intestinal sample collection and processing.

Samples were collected from live pigs under general anesthesia. Pigs were initially sedated with ketamine (Vetalar; Fort Dodge, IA) and xylazine (Xylamax; Bimeda, Cambridge, ON, Canada) at doses of 22 mg kg⁻¹ and 2.2 mg kg⁻¹ of body weight, respectively. Animals were placed in dorsal recumbency on a v-trough surgical table and intubated, and general anesthesia was established with isoflurane (Abbott Laboratories, North Chicago, IL) at 1,500 ml min⁻¹ O₂. The abdomen was disinfected with 2% chlorhexidine and 4% isopropyl alcohol (Stanhexidine; Omega Laboratories Ltd., Montreal, QC, Canada), 70% ethanol, and 10% povidone-iodine (Prepodyne; West Penetone Inc., Ville D’Anjou, QC, Canada). A longitudinal ventral laparotomy was established. Segments of the intestine (∼10 cm long) were collected from the ileum, cecum, ascending colon, and spiral colon. To ensure the integrity of the intestinal segment (e.g., to prevent the infiltration of air into the segment) and to minimize the release of digesta, double ligatures were established at the two ends of the segment, and the segment was then excised from the intestine by cutting between the two ligatures. Care was taken to ligate mesenteric blood vessels immediately prior to intestinal segment removal to ensure the maintenance of blood flow to adjacent intestinal tissue. The segments were placed in a 3.5-liter anaerobic jar (catalog number HP0011A; Oxoid, Nepean, ON, Canada) within 3 to 5 min of removal from the animal, the ambient atmosphere in the jar was removed by a vacuum and replaced with N₂ gas, and the segments were transported to the laboratory for processing. Before placement in an anaerobic chamber, samples were placed on ice. The time from sample removal from the animals to its placement in the anaerobic chamber was less than 20 min.

Sample processing.

Ligated intestinal samples were transferred into an anaerobic chamber (Forma 1025; Forma Scientific Inc., Marietta, OH) containing a predominantly N₂ atmosphere (i.e., 85% N₂, 10% CO₂, and 5% H₂). Once in the chamber, ligatures were aseptically removed, and the intestinal segment was incised to expose the mucosal surface and luminal contents. A 1-cm² sample of the intestinal wall was removed and transferred into 5 ml of reduced Columbia broth (CB) (HiMedia Laboratories LLC, West Chester, PA) contained in a 50-ml Cellstar polypropylene tube (Greiner Bio-One International, Kremsmünster, Austria). The sample was gently inverted 30 times to separate the digesta from the mucosa. The intestinal segment was then transferred to a new 50-ml tube containing reduced CB, and both tubes were vortexed (high setting for 30 s). The resultant suspension was used to isolate bacteria by various methods. The tubes containing samples were then placed into an anaerobic jar (Oxoid), the lid was sealed, and the jar with tubes was transferred into a separate Forma 1025 anaerobic chamber with a predominantly CO₂ atmosphere (i.e., 90% CO₂ and 10% H₂). Once in the CO₂ atmosphere chamber, the enteric samples were processed to obtain luminal or mucosal samples for isolation of bacteria as done previously (60). Samples were processed for isolation as outlined above (Fig. 1).

Bacteriological media.

All media used during this study were reduced before use. Media used included a basal Dehority’s medium (78) supplemented with xylan or porcine mucus type III (see Table S3 in the supplemental material), Columbia broth supplemented with 10% blood, and Bacteroides bile esculin agar with and without 100 μg liter⁻¹ gentamicin (Table S4). The liquid medium prepared without the addition of cysteine was autoclaved for 5 min. Once autoclaved, warm media were transferred into an N₂ atmosphere chamber and vigorously agitated to displace oxygen. When the medium cooled, it was decanted into a bottle containing the desired cysteine content (1 g liter⁻¹) to remove any residual oxygen, and the bottle was sealed with a stopper, removed from the chamber, and autoclaved for 30 min at 121°C at 100 kPa. Media prepared for use in the CO₂ atmosphere were prepared in the same fashion; however, 40 ml liter⁻¹ of 8% sodium carbonate (Sigma-Aldrich Canada, Oakville, ON, Canada) was added before autoclaving to prevent acidification.

(i) Enrichment broths.

Broths for enrichment were prepared in a fashion similar to that for liquid media, with the exception that after the first 5-min autoclave cycle, 10 ml of the medium was transferred into 15-ml glass tubes (Fisher Scientific, Ottawa, ON, Canada). Tubes were sealed with a screw cap, and oxidization was detected by the addition of resazurin (final concentration, 1 μg ml⁻¹) (Table S5). For enrichment cultures containing blood, 10% whole sheep blood was added to liquid media containing resazurin, dispensed into glass tubes, and autoclaved; after autoclaving, the curdling of blood allowed the visualization of resazurin oxidation.

(ii) Agar media.

Media containing 1.5% agar (BD Difco, Fisher Scientific) and 1 μg ml⁻¹ resazurin (Sigma-Aldrich Canada) were autoclaved for 30 min at 121°C at 100 kPa (Table S6). Molten media were dispensed into petri dishes, immediately transferred directly into the N₂ and CO₂ atmosphere chambers, and maintained in the chambers for 10 days before use.

Isolation strategy.

A number of isolation methods were used. These included (i) direct plating, (ii) long-term enrichment, (iii) the use of an ichip adapted for enteric isolations, and (iv) the selection of endospore-forming taxa via ethanol treatment and tyndallization, followed by direct plating and enrichment (Fig. 1). In all instances, isolations were conducted under strict anoxic conditions (CO₂ and/or N₂ atmospheres) at 37°C.

(i) Direct plating.

The bacterial suspension from intestines in CB was streaked (10 μl) onto various agar media (Table S7) within the N₂ or CO₂ atmosphere chambers, and cultures were incubated for 7 days.

(ii) Enrichment broths.

To generate enrichment cultures, 20 μl of the bacterial suspension from intestines in CB was added to 10 ml of enrichment media (Table S8). Cultures were maintained in the N₂ or CO₂ atmosphere chambers for 12 weeks, and enriched bacteria were isolated by direct plating.

(iii) ichip.

A modified version of the ichip method described previously (44) was used (Table S9 and Fig. S11 to S13). Instead of a three-dimensional printed manifold, we generated a diffusion chamber from a pipette tip holder as done previously (44). We also used a membrane containing 0.02-μm-diameter pores (Sterlitech, Kent, WA), which was sealed to the apparatus bottom using Silicone II sealant (General Electric Company, Fairfield, CT). ichips were then autoclaved in a sealable container and placed in the N₂ and CO₂ atmosphere anaerobic chambers for 24 h before use. Phosphate-buffered saline (PBS) (0.01 M sodium phosphate [pH 7.4]) amended with 0.5% agarose was reduced by autoclaving for 5 min at 121°C and then quickly transferred into an anaerobic chamber, where it was vigorously mixed, and l-cysteine HCl (1 g liter⁻¹) was added to assist in its reduction. Once cooled, the liquid was transferred into a clean bottle, sealed with a rubber septum cap, and autoclaved. A 50-μl aliquot of the intestinal sample diluted in reduced PBS with agarose was stained with 50 μl of a solution of 0.4% trypan blue (Sigma-Aldrich Canada). Stained bacterial cells in PBS were enumerated visually using a Petroff-Hausser chamber (VWR International, Radnor, PA) at a ×100 magnification under bright field and oil emersion using a Zeiss Axioskop 3 microscope (Carl Zeiss Canada Ltd., Toronto, ON, Canada). Within the anaerobic chamber, the appropriate volume of the intestinal sample was added to reduced PBS to obtain a target density of 1 bacterium in 200 μl. The diluted sample (200 μl) was then dispensed into individual chambers of the ichip apparatus. The top of the chamber was then securely sealed with a Microseal ‘B’ PCR plate sealing film (Bio-Rad Laboratories, Hercules, CA). The apparatus was placed into fresh rumen fluid obtained from fistulated cattle, ensuring that the porous membrane was submerged; the rumen fluid was used as a source of nutrients, including rare nutrients, vitamins, and cofactors (44, 79). ichips were maintained for 12 weeks, and the rumen fluid was replaced at 2-week intervals. After the incubation period, the ichip was removed from the rumen fluid, residual fluid was removed, the ichip was surface sanitized with 70% ethanol, and the top membrane was carefully removed. A 10-μl subsample was taken from each chamber and streaked onto reduced Columbia agar (CA) (HiMedia Laboratories) amended with 10% whole sheep blood (CBA).

(iv) Isolation of endospore-forming bacteria—ethanol.

The basic method of Browne et al. (24) was used, with some modifications. In this regard, intestinal samples were stored at −80°C until processing. In addition, ethanol was filtered before placement into the anaerobic chambers, the YCFA medium used by Browne et al. (24) was replaced with CBA medium, cultures were incubated for 7 days (i.e., instead of 3 days), and a 12-week enrichment was used to isolate bacteria. Frozen samples were thawed in the anaerobic chamber containing an N₂ atmosphere, placed in 1 ml of reduced CB, and vortexed (high setting for 30 s); 0.5 ml of the suspension was transferred to a new sterile vial; and 0.5 ml of reduced 70% ethanol was added to the suspension (1:1 ratio). The ethanol suspensions were incubated for 3 h. Bacteria were recovered by direct plating and enrichment. For isolation using direct plating, 10-μl subsamples were streaked onto CBA containing 0.1% sodium taurocholate (Sigma-Aldrich Canada) and maintained in an N₂ atmosphere anaerobic chamber. In addition, 20-μl subsamples were added to 10 ml of reduced enrichment broth in Hungate tubes. The enrichment media used were Dehority’s medium amended with 0.5% xylan from beechwood (Sigma-Aldrich Canada) or porcine mucus III (Sigma-Aldrich Canada) as well as CB with 5% sheep blood. After the enrichment period in the N₂ atmosphere, 10 μl from each culture was streaked onto CBA and maintained in the N₂ atmosphere as described above.

(v) Isolation of endospore-forming bacteria—tyndallization.

Tyndallization of samples was conducted in concert with the ethanol treatment method. Intestinal samples were placed in 1.0-ml glass vials (Fisher Scientific) within the N₂ atmosphere anaerobic chamber. The vials were sealed to maintain samples in a reduced state, removed from the anaerobic chambers, and placed at 100°C for 30 min. Vials were then transferred back to the anaerobic chamber, and bacteria were isolated by direct plating and enrichment as described above for the ethanol treatment.

Recovery and storage of bacteria.

Based on colony morphology, representative colonies were selected, streaked for purity on CBA, and maintained in the atmosphere in which they were isolated. Colonies were then streaked for biomass on CBA. After ≥7 days, biomass was removed from the surface of the medium and transferred to tubes containing 1.5 ml of reduced CB with 40% glycerol. The tubes were then sealed and removed from the anaerobic chambers, snap-frozen on dry ice, and transferred to −80°C for medium-term storage (Table S10). In addition, biomass was placed into tubes and stored at −80°C for the subsequent extraction of genomic DNA.

Identification of bacteria.

Biomass for the extraction of genomic DNA was thawed at room temperature, after which 300 μl of lysis buffer (plant lysis buffer 102; AutoGen Inc., Holliston, MA) supplemented with 10 mg ml⁻¹ lysozyme (Thermo Scientific Inc., Waltham, MA) was added to the biomass for cell lysis. Cells were lysed at 37°C overnight and then placed at 50°C for 1 h. Genomic DNA was then extracted using an AutoGen 740 instrument (AutoGen Inc.) according to the manufacturer’s recommendations. The 16S rRNA gene was amplified using the 27F and 1492R primers as previously described (80) and was partially sequenced by Eurofins Genomics (Toronto, ON, Canada) using the 27F primers as previously described (35, 60). Sequence chromatograms were visualized, assessed for quality, and trimmed using v5.3.9 Geneious software (Geneious, San Diego, CA). Trimming was conducted to ensure that chromatogram peaks were of good signal intensity and generally formed individual peaks. Trimmed sequences in fasta format were queried against reference sequences within the RDP using the sequence match program (SeqMatch [http://rdp.cme.msu.edu/seqmatch/seqmatch_intro.jsp]) (35) with the following settings selected: strain = both (type and non-type), source = isolates, size = both (>1,200 and <1,200), quality = good, taxonomy = nomenclature, and K nearest neighbors (KNN) matches = 1. Sequences are available in Zenodo (https://doi.org/10.5281/zenodo.3728174). Representative isolates were accessioned in the IBaC at AAFC LeRDC and appropriately stored.

Next-generation sequence characterization of bacterial communities.

DNA extracted from digesta was processed with an Illumina protocol for creating 16S rRNA sequencing libraries. Extracted genomic DNA was amplified with Illumina indexed adaptor primers (V4 Schloss primers) (81). The PCR mixture contained 5 μl of PCR buffer, 1 μl of 10 mM deoxynucleoside triphosphates (dNTPs), 1 μl of 25 mM MgCl₂, 2.5 of μl of each primer, 0.25 μl of Hot Start Taq DNA polymerase (Qiagen Inc., Toronto, Ontario, Canada), 32.8 μl of molecular-grade water, and 5 μl of bacterial community DNA. Amplicons were purified with a QIAquick PCR purification kit (Qiagen Inc.) according to the manufacturer’s recommendations. The effectiveness of the cleanup was checked by agarose gel electrophoresis, followed by quantification with a Qubit instrument (Fisher Scientific). Indexed DNA libraries were normalized to 1.5 ng μl⁻¹ and pooled. A PhiX control (10%) was run with the normalized DNA library, and both were denatured and diluted to 8 pM prior to loading into the MiSeq reagent kit, V2 (500 cycles) (Illumina, San Diego, CA). An average of 5,673 16S rRNA gene amplicon reads were obtained per sample, with 94% of the reads passing the Q₃₀ score. QIIME2 (82) was used to classify bacterial reads for digesta communities. This analysis was done using the Tourmaline Snakemake reproducible workflow to automate QIIME2 (version 2019.7) analyses (https://github.com/ropolomx/tourmaline). Briefly, raw reads were denoised with DADA2 (83), and representative amplicon sequence variants (ASVs) were generated. A phylogenetic tree of ASV sequences was generated, and the taxonomy of each ASV was identified by using a naive Bayes classifier pretrained with the reference SILVA 132 database (silva-132-99-515-806-nb-classifier.qza). Alpha diversity metrics, including the number of taxa observed, the Shannon index of diversity, and the inverse Simpson index, were calculated. The phyloseq package (version 1.28.0) of R, version 3.6.1, was used to evaluate beta diversity with a principal-coordinate analysis (PCoA) of the calculated unweighted UniFrac distances, generating an ordination plot. Differential abundance between intestinal locations was detected with analysis of composition of microbiomes (ANCOM) in QIIME2 (84).

Data analyses.

Bacterial taxa were collated with intestinal location, intestinal sample type (i.e., mucosa associated or digesta), isolation atmosphere, isolation method, and isolation medium. Bacterial diversity was determined by Shannon’s index of diversity, and community composition was determined using the vegan and pvclust packages in R (28, 29). Analyses were conducted similarly to that done by Moote et al. (60). Detection of the differential abundance of taxa isolated was done via LEfSe methods (32). The analyses were done using the LEfSe tool at http://huttenhower.sph.harvard.edu/galaxy/ (32). Dendrograms were generated from the 16S rRNA gene sequences using the MEGA-X software package (85), multiple-sequence alignments were made using the embedded MUSCLE tool, and dendrograms were generated using UPGMA linkage with a bootstrap value of 500. The Newick alignment file was then transferred to FigTree, version 1.4.4, to generate the dendrogram (86). Experimental statistics were evaluated using the vegan package in R (28), and statistical inferences were determined using the Kruskal-Wallis rank sum test in R (29) and the Nemenyi test with χ² approximations (30) to identify significant changes to Shannon’s indices. False discovery rates were determined via the q value program in R (31). Overlapping taxa were evaluated using the UpSetR program in R (87) to determine the presence/absence of overlapping taxa between methods and intestinal locations.

Data availability.

The raw sequencing reads were submitted to the Sequence Read Archive of the NCBI under BioProject accession number PRJNA612572.