Introduction

Recent advances in whole genome sequencing of prokaryotes have led to an increased recognition of the phenomenon of genome diversity within traditionally defined taxa [26]. Genomic diversity is observed among strains of well-characterized pathogens [4, 31] and in organisms isolated from natural microbial communities [23]. Twenty percent to 40% of open reading frames in prokaryotic genomes encode proteins of unknown function, and these uncharacterized genes are over-represented among unique, strain-specific genes [40]. Strain-specific genes are likely related to ecological niche specialization [40], and selection acting on these genes drives evolutionary diversification between strains [22]. It has been proposed that “natural taxa” are comprised of groups of strains or ecotypes that emerge in populations through periodic selection [6]. In gene sequence-based studies, these ecotypes or strains are detected as clusters of sequences where the average distance between ecotypes is greater than the average distances within the cluster. Our understanding of micro-diversity and ecotype dynamics is limited by difficulty in detecting phylogenetically discrete strains in complex samples using common, 16S rRNA sequence-based methods. Also, despite growing sequence-based evidence of sub-species level diversity within natural communities and an increasing appreciation of its potential biological significance, studies of the phenotypic relationships of sub-species groups defined by sequence relationships are rare. Without this information, it is difficult to evaluate the significance of micro-diversity to the function of complex microbial communities.

In culture-independent, 16S rRNA sequence-based microbial ecology studies, polymerase chain reaction (PCR) with universal bacterial primers is conducted on total community DNA and products are sequenced, either via production of clone libraries or direct pyrosequencing. Usually, only a portion of the gene is amplified (approximately 500 bp) to simplify target amplification and sequencing. Taxonomic and phylogenetic assignment of sequences is facilitated by comparison to reference sequence databases, particularly the Ribosomal Database Project [25]. Traditionally, population data is lumped into “operational taxonomic units (OTU)”, which appear as discrete clusters of nearly identical sequences. The reasons for the observed sequence micro-diversity within OTUs identified by 16S rRNA sequence, or any PCR-amplified gene target, include PCR artifacts arising from non-proofreading polymerases [3]. However, fine-scale population structure, including discrete groups of sequences with <1% sequence divergence, is also apparent in 16S rRNA libraries where artifacts have been minimized through technical adjustments [2].

The gene encoding the universal 60 kDa chaperonin (cpn60) has been established as a target for identification of bacterial species and has the advantage of offering superior discrimination between closely related organisms: at the species [17, 21, 28] and sub-species levels [5]. Among protein-coding alternatives to 16S rRNA, cpn60 is the most developed, and a large, curated database of sequence data, cpnDB (http://cpndb.cbr.nrc.ca), is available [18]. Since cpn60 is a protein-coding gene, the codon position of differences between nearly identical sequences generated in microbial ecology studies can be informative. In a study of human vaginal microbial communities where cpn60 was used as a target rather than the 16S rRNA gene [15], the distribution of differences between members of a cluster of sequences with >90% identity to each other was skewed toward synonymous substitutions at the third codon position, rather than the random distribution expected from polymerase error. Similar observations have been made of cpn60 sequence data from microbial populations in animal intestines [10, 16, 19] and an industrial environment [11]. This observation suggests that at least some sequence micro-diversity in microbial populations may be indicative of real strain diversity, rather than technical artifacts. Furthermore, it suggests that cpn60 sequences may be used to characterize sub-species level diversity within natural microbial communities.

The objectives of the current study were to compare sub-species resolution of the cpn60 and 16S rRNA gene targets and to determine if the cpn60 target could be used to detect species and sub-species level changes in a model microbial community: Enterococcus in the feces of pigs from weaning to maturity. We have also used carbon source utilization profiling to investigate the phenotypic diversity among strains within the fecal enterococci and to determine if strains identified based on cpn60 sequence have distinct phenotypes. We chose Enterococcus spp. as a model population since they are easily isolated from the intestinal microbiota of pigs and have a broad repertoire of metabolic capabilities.

Materials and Methods

Swine Feces Collection

Rectal fecal samples were collected from pigs in a biosecure, high health production facility. Collections were made at 3, 9, and 15 weeks of age. Fourteen pigs were included in the study: two to four pigs from each of four litters, and a mixture of male and female piglets were sampled from each litter. Piglets at 3 weeks were suckling. Diet composition at 9 weeks consisted of 55% wheat, 26% soymeal, and 15% barley. At 15 weeks, the diet consisted of 30% corn, 23% barley, 14% soymeal, 8% canola meal, and 23% wheat.

Bacterial Culture

Fecal samples were diluted in peptone water and plated on mEnterococcus agar (Difco) using a spiral plater to facilitate colony counting. Plates were incubated at 35°C for 48 h. The average number of colonies from at least three plates was calculated for each sample to determine colony-forming units per gram. Isolated colonies were picked into 150 μL of Luria-Bertani broth with 12% (v/v) glycerol and incubated overnight at 35°C prior to freezing at −80°C for storage.

Type strains of Enterococcus faecalis (ATCC 19433), Enterococcus hirae (ATCC 8043), Enterococcus faecium (ATCC 19434), Enterococcus avium (ATCC 14025), and Enterococcus mundtii (ATCC 43186) were obtained from the American Type Culture Collection.

PCR and Sequencing

All PCR reactions were performed in 96-well plates using 1 μL of glycerol stock of each bacterial isolate as template. The region of the 16S rRNA gene corresponding to nucleotides 11-536 of the Escherichia coli 16S rRNA gene (encompassing variable regions V1, V2, and V3) was amplified with universal primers F1 (5′-GAG TTT GAT CCT GGC TCA G-3′) and R2 (5′-GWA TTA CCG CGG CKG CTG-3′) [9]. PCR amplifications were performed in 50 μL reactions consisting of 5 μL of a 10× PCR reaction buffer, 2 mM MgCl2, 0.2 mM each of a dNTP mixture, 0.4 μM of each primer, and 1 U of HP Taq DNA polymerase (UBI, Calgary, AB, Canada) using an Eppendorf Mastercycler EP. The amplification program was 3 min at 95°C followed by 40 cycles of 95°C for 30 s, 55°C for 30 s, 72°C for 30 s, and a final extension of 10 min at 72°C.

For amplification of the cpn60 universal target region corresponding to nucleotides 274-828 of the E. coli cpn60 sequence, primers JH0064 (5′-GAT ATY GCW GGW GAY GGW ACA ACD AC-3′) and JH0065 (5′-CGR CGR TCR CCR AAS CCS GGH GCT-3′) were used. JH0064 and JH0065 were designed based on universal cpn60 primers H279 (5′-GAI III GCI GGI GAY GGI ACI ACI AC-3′) and H280 (5′-YKI YKI TCI CCR AAI CCI GGI GCY TT-3′) and modified to match the consensus sequences of E. hirae, E. faecalis, Enterococcus cecorum, and E. faecium, which were previously reported to be present in pig feces. These primers were used instead of universal cpn60 primers H279 and H280 for this Enterococcus-focused study to avoid the expense of inosine-containing primers. PCR reaction components were identical to those described for 16S rRNA. The amplification program for JH0064/JH0065 was 3 min at 95°C followed by 40 cycles of 95°C for 30 s, 62°C for 60 s, 72°C for 60 s, and a final extension of 10 min at 72°C.

PCR products were purified using the PCR Microarray Purification kit (Arrayit Corporation, Sunnyvale, CA, USA). Purified products were sequenced directly with the amplification primers. Raw sequence data for individual PCR products was manually edited and assembled using the Staden Package with the default settings for quality cutoff in the Pregap4 module [37]. Other sequence manipulation was accomplished using EMBOSS [34]. Phylogenetic trees were created using PHYLIP [12] and viewed with NJplot [32].

Species identification of sequences was made by comparison of isolate sequences to the Greengenes database (16S rRNA sequences; http://greengenes.lbl.gov) or cpnDB (cpn60 sequences; http://cpndb.cbr.nrc.ca) [18].

Carbon Source Utilization Profiling

Carbon source utilization profiles were generated using the BIOLOG GP2 microplate (Biolog Incorporated, Hayward, CA, USA) according to the manufacturer’s instructions. Prior to analysis, strains were resurrected from −80°C glycerol stocks on trypticase soy agar plates containing 5% defibrinated sheep blood and incubated for 48 h at 35°C. Results were recorded at 18 h post-inoculation. Statistical analysis of carbon source utilization patterns was accomplished using Statistical Package for the Social Sciences (SPSS) software (SPSS Inc., Chicago, USA).

Results

Enterococcus Population Size Decreases with Age

Table 1 summarizes the total Enterococcus colony counts for all pigs over the three time periods. Enormous variation in counts was observed in the weaning pigs (3 weeks of age), ranging from 0 in pig 61 (no growth observed on mEnterococcus agar plates) to >108 cfu/g in pigs 58, 64, and 82. Observations at the later times were more consistent. By 15 weeks, total Enterococcus counts had dropped to 103 to 104 cfu/g in all pigs for which data was collected, except pig 64 where no growth was observed. All isolates examined were determined to be Enterococcus spp.

Table 1 Total Enterococcus colony counts in pig feces
Full size table

Identification of Enterococcus spp. by 16S rRNA and cpn60 Sequencing

To compare the utilities of the 16S rRNA and cpn60 targets for Enterococcus species identification and strain level resolution, we amplified and sequenced both targets from each of 169 isolates representing all three time points, from pig 81. BLASTclust was used to identify 100% identical sequences within the collection. Using this criterion, the 16S rRNA sequences reduced to 23 different sequences, and the cpn60 data set contained 43 different sequences. Three species (E. hirae, E. faecalis, and E. faecium) were identified based on cpn60 sequence identity to type strains. Resolution of the E. hirae and E. faecium strains by 16S rRNA sequence was unclear since the type strain sequences for these species are 99% identical to each other over the 510-bp region examined. The cpn60 sequences for these two species are only 85% identical to each other. Interspecific and intraspecific distances were larger for the cpn60 sequences, resulting in clear resolution of species and the identification of two distinct subgroups with E. hirae (Fig. 1). Based on these initial results from pig 81, all further characterization of strains was done only with the cpn60 target.

Figure 1
figure 1

Phylogenetic trees based on partial 16S rRNA (left) and cpn60 (right) sequences for isolates from pig81 at 3, 9, and 15 weeks of age. Sequences from type strains of Enterococcus hirae, Enterococcus faecalis, and Enterococcus faecium are included for reference. Labels indicate the number of isolates with identical sequences at each point in the tree. Label colors indicate the age of pig from which the sequenced isolate was obtained. The tree is a consensus of 100 neighbor-joined trees, and bootstrap values are indicated at the major branch points

Full size image

A total of 694 isolates from pigs 57 (13 isolates), 67 (45 isolates), 68 (185 isolates), 71 (9 isolates), 78 (156 isolates), 81 (168 isolates), and 88 (118 isolates) were subjected to cpn60 PCR and sequencing. These pigs were selected since isolates were available at each time point. In most cases, all isolates cultured from a given pig at each time were collected for analysis. In cases were more than 96 isolates were available, only the first 96 randomly chosen isolates were examined. Although five species were identified (E. faecium, E. hirae, E. faecalis, E. mundtii, and E. avium), all but three isolates were E. faecalis or E. hirae. Figure 2 is a phylogenetic tree of the unique cpn60 sequences from each pig at each time (198 sequences representing 694 isolates). Representative sequences have been deposited in Genbank and assigned accession numbers GQ405204–GQ405360. Sub-species divisions with solid bootstrap support are evident within E. faecalis and E. hirae, and the sequence identities between these intraspecific groups are 96–98%. All sub-species groups contain sequences from multiple pigs. A closer examination of the E. hirae species cluster (107 sequences) shows that of the 66 variable positions in an alignment of these sequences (552 bp), the ratio of codon positions (first:second:third) is 19:13:34, indicating a skew toward the third position. Within hirae-1 and hirae-2 subgroups, codon position of differences was 18:9:15 and 10:6:7, respectively.

Figure 2
figure 2

Phylogenetic trees based on alignment of 198 unique cpn60 gene from 694 Enterococcus isolates. Representative sequences from each pig at each time are included. Type strains of Enterococcus faecalis, Enterococcus hirae, Enterococcus faecium, Enterococcus mundtii, and Enterococcus avium are included for reference. The tree is a consensus of 100 neighbor-joined trees, and bootstrap values are indicated at the major branch points. Sub-species groups are also indicated. Colored dots indicate the age of pig from which the sequences were obtained

Full size image

Temporal Shifts in Sub-species Groups Within the Genus Enterococcus

Figures 2 and 3 describe the observed temporal changes in fecal Enterococcus populations. At 3 weeks, the isolate collection was dominated by E. faecalis which accounted for over 98% of characterized isolates. Two isolates were classified as hirae-3 and one as E. faecium. At 9 weeks, the vast majority of isolates were identified as E. hirae (93% hirae-1 and 1% hirae-3). By 15 weeks, E. hirae was still the dominant cultured species, but there was a pronounced shift to subgroup hirae-2, with only one hirae-1 isolate identified from pig71. No subgroup hirae-3 isolates were detected at 15 weeks. E. faecalis subgroup faecalis-2 was observed at all times in all pigs, while faecalis-1 was only detected in suckling pigs 68 and 88 (3 weeks).

Figure 3
figure 3

Proportional representation of the various species and strains in isolates from fecal samples at 3, 9, and 15 weeks of age. The total number of isolates analyzed at each age is indicated below the x-axis. Sub-species group names correspond to Fig. 2

Full size image

Carbon Source Utilization Profiles of Enterococcus Sub-species Groups

To determine if the sub-species groups identified based on cpn60 sequence and phylogeny were also distinguishable based on phenotype, we determined carbon source utilization profiles from the 198 isolates representing the sequences shown in Fig. 2 as well as type strains of E. faecalis (ATCC 19433), E. hirae (ATCC 8043), E. faecium (ATCC 19434), E. avium (ATCC 14025), and E. mundtii (ATCC 43186), using the BIOLOG GP2 system. Ninety-five sole carbon sources are evaluated in each plate. None of the isolates had profiles identical to their corresponding type strains.

Among all E. hirae isolates, 34 of the carbon sources were utilized by at least one isolate (Table 2, Supplementary Fig. 1a). For 14 of these carbon sources, a significant difference in the proportion of isolates in each subgroup utilizing the carbon source was observed, based on binary logistic regression analysis with p < 0.05 considered significant. Hirae-2 was collectively able to utilize 30 of the 34 carbon sources with 18 of these being utilized by at least 50% of the isolates. Although subgroup hirae-1 was collectively able to utilize 31 of the 34 carbon sources, only six were utilized by more than 50% of the isolates, indicative of the greater degree of variability within hirae-1 compared to hirae-2. Individual hirae-2 isolates utilized an average of 16.5 carbon sources (range from 3–27) compared to an average of 8.6 for hirae-1 isolates (range 1–24). The most differentiating carbon sources for hirae-1 and hirae-2 were α-D-lactose, D-ribose, and uridine, which were positively associated with hirae-2. Subgroup hirae-3 isolates were collectively only able to utilize 16 carbon sources, although only five isolates were examined.

Table 2 Carbon source utilization profile of Enterococcus hirae strains
Full size table

E. faecalis isolates were collectively able to utilize 50 of the carbon sources in the GP2 plate (Table 3, Supplementary Fig. 1b). Subgroups faecalis-1 and faecalis-2 could be distinguished on the basis of eight carbon sources whose utilization was positively associated with faecalis-2 (based on chi-square or Fisher’s exact test, p < 0.05). Faecalis-3 had only two members, providing inadequate information for meaningful comparison with the other groups. In subgroup faecalis-2, 29 of the 50 carbon sources were utilized by at least 50% of the isolates and of those, 24 were utilized by 90–100% of isolates. This is consistent with the more homogenous nature of carbon source utilization profiles observed for E. faecalis subgroups compared to E. hirae. Within faecalis-2 were isolates from all three ages of pig, but isolates from the age categories or individual pigs were not distinguishable by carbon source utilization profile (Supplementary Fig. 1b).

Table 3 Carbon source utilization profile differences between Enterococcus faecalis strains
Full size table

Discussion

There is ample evidence that diversity in microbial communities (at various taxonomic levels) is a significant parameter related to community function. Comparative studies where a community is examined over time or conditions are still relatively rare, but there are examples of changes in diversity that are either the cause or effect of biological changes. Metal contamination resulted in at least 99.9% reduction in diversity in a pristine soil, although the total number of bacteria remained approximately the same [13]; diversity in colonic microbial populations in inflammatory bowel disease patients is 30–50% of the diversity observed in healthy patients [29]; and diet composition and antimicrobial use affect microbial community sequence diversity in the intestines of pigs and chickens [10, 16]. Recently, based on genome sequencing of cultured prokaryotes, it has become evident that biologically significant genomic diversity exists within traditionally defined species, and this diversity may also be an important factor in microbial community dynamics. Questions regarding genomic diversity in natural microbial communities have been more difficult to address since rather than comparisons of two or a few fully sequenced genomes, a complex community of hundreds to thousands of different taxa must be considered. Although micro-diversity is discernable even in some carefully constructed 16S rRNA sequence libraries that include nearly full-length gene sequences (approximately 1.5 kb) [2], we sought to demonstrate that this fine-scale microbial community architecture is more clearly resolved using cpn60 universal target sequences.

The development of pig intestinal microbiota from birth to the “climax community” [35] of the adult offers an interesting model system. Dramatic changes in the structure and function of the microbiota occur, with the most dramatic changes associated with weaning when piglets are abruptly converted from a milk diet to solid food. Characterization of microbial succession in pigs to date has involved methods that do not detect changes occurring below the level of traditionally defined species [20, 38].

Our observation of a decrease in the number of culturable enterococci in feces from weaning to 15 weeks is consistent with previous observations of the progression from a neonatal intestinal microbiota specialized for the digestion of a milk diet to the more complex and diverse microbiota required for the digestion of complex diet ingredients. In pigs, as in other animals including humans, there is a succession from coliforms and streptococci, including Enterococcus, to the obligate anaerobes that dominate the climax community [19, 24]. The high degree of variability we observed in the suckling pig colony counts (from undetectable to >108 cfu/g) may reflect that our sampling represented only a “snapshot” at a single time of a complex and inherently unstable environment. However, by 15 weeks, the range of total culturable Enterococcus colony counts had decreased (3.33 × 102 to 3.47 × 104 cfu/g) suggesting that the microbiota of suckling piglets is more variable between pigs than that of adult animals.

One approach to describing genomic diversity is comparative genomics based on whole genome sequencing. Studies of Enterococcus spp., particularly E. faecalis, have demonstrated the genomic plasticity of these organisms and that they are replete with strain-specific genes [1, 27, 30]. A subset of 1,722 genes has been proposed for the E. faecalis core genome, with at least another 1,000 ORFs identified as being variably associated with individual strains characterized to date [36]. Strain-specific genes include putative virulence factors, antimicrobial resistance factors, as well as those involved in homeostasis, nutrient uptake, and metabolism. Selection acting on genes related to niche specialization is thought to be a major driver of strain diversification [22]. Our understanding of the range of genomic diversity in this genus is expanding rapidly with approximately 50 Enterococcus genome projects in progress (at the time of writing) and expansion of focus from pathogens to include commensal Enterococcus strains. However, current logistical constraints limit comparative genomic studies to selected isolates, rather than in-depth studies of natural microbial populations, which are more readily studied using a combination of targeted sequencing coupled with phenotypic characterization.

Sequence-based detection of sub-species level diversity in a population requires the use of a gene target with sufficient resolving power. Cpn60 sequences have previously been demonstrated to be useful for identification of phenotypically similar Enterococcus spp. of clinical importance [14]. A direct comparison of partial 16S rRNA and cpn60 sequences from isolates from pig 81 demonstrates the relative resolution of these two targets (Fig. 1). Pairwise interspecies sequence identities are lower for cpn60, allowing clear discrimination of the species included (85% pairwise identities between E. faecalis, E. faecium, and E. hirae). The amplified region of 16S rRNA (including the first three variable regions) was insufficient to discriminate between E. faecium and E. hirae. In fact, a comparison of published full-length 16S rRNA sequences for the type strains of these species (Genbank Accessions AJ276355 and AJ276356) shows that they are 99% identical over the entire length of the gene. Using cpn60 sequences, we were also able to identify two sub-species groups within E. hirae. Average pairwise sequence identities between these two groups were 96–98%, but they were nonetheless distinguishable with a high bootstrap value. These divisions, initially recognized in pig 81, were supported by additional isolates from all other pigs (Fig. 2). We also detected two sub-species groups within E. faecalis with similar inter-group sequence similarities.

The species represented in the culture collection were likely influenced by the isolation methods we employed. For example, while we did not isolate any E. cecorum, a species previously reported in pigs [7], there have been reports that this species does not grow well on common selective media for enterococci [8]. However, regardless of the culture method biases, isolation conditions were consistent at each time, making it possible to compare isolates selected. Dramatic shifts in the Enterococcus population profile were observed over the three time points, and these changes occurred simultaneously among all pigs examined (Figs. 2 and 3).

The fact that Enterococcus population changes occurred simultaneously indicates that the cause of these changes is likely something in common between all individual animals, probably diet. Fecal collection times corresponded to three different feed regimens: sow milk at 3 weeks and two different grower diets at 9 and 15 weeks. Between 9 and 15 weeks, dietary changes included a 50% decrease in wheat content and the introduction of corn and canola meal. The shift from hirae-1 to hirae-2 between 9 and 15 weeks may be the result of diet-based selection. The hirae-2 isolates were generally capable of utilizing more of the sole carbon sources in the BIOLOG GP2 assay, perhaps indicating that they are more metabolically capable and were positively selected for with a more complex diet. This would be analogous to the changes in Lactobacillus metabolic profiles observed in the small intestine of weaning pigs where carbon source utilization pattern changes mirrored changes in diet composition [33]. The coordinated shifts in Enterococcus populations within the animals in this study is in interesting contrast to descriptions of microbial succession in human intestinal microbiota where much more individual-to-individual variation occurs, likely due to the individual environments and diets of the subjects [23, 39].

Analysis of the carbon source utilization profiles of representative isolates from all sub-species groups defined by cpn60-based phylogeny demonstrates that these groups also have distinguishable phenotypes and represent what have been described as ecotypes: groups of closely related strains occupying the same ecological niche [6]. This is best illustrated by the comparison of hirae-1 and hirae-2. The utilization patterns for 14 of the carbon sources evaluated showed significant differences between the E. hirae sub-species groups making it possible to classify isolates as hirae-1 or hirae-2 based on carbon source utilization parameters. For E. faecalis, faecalis-1 isolates were only found at 3 weeks in pigs 68 and 88. These isolates were distinguishable from faecalis-2 isolates based on cpn60 sequence and eight carbon sources whose utilization was positively association with faecalis-2 (Table 3, Supplemental Fig. 1a). Unlike the hirae-1 to hirae-2 shift, E. faecalis strains belonging to the faecalis-2 group were isolated throughout the study and consistent with their definition based on cpn60 sequence; very little variation in carbon source utilization pattern was observed within this group (Supplemental Fig. 1b). These observations suggest that while ecological conditions in the intestine selected for hirae-2 strains between 9 and 15 weeks, faecalis-2 strains occupy a niche unaffected by the diet change in this time frame.

Taken together, our results demonstrate that it is possible to use cpn60 sequences to identify phenotypically distinct strains or ecotypes within an intestinal microbial community. Lack of recognition of micro-diversity in many microbial ecology studies is attributable in part to the size of sequence collections produced. If a relatively small number of sequences from a single gene are generated from a population, it is unlikely that phylogenetically discrete sub-species groups will be identified. However, with the advent of next-generation sequencing technologies, much deeper data sets can be produced, increasing the likelihood of identifying these groups, assuming that the target sequence chosen is sufficiently informative. If micro-diversity is a significant parameter in microbial population dynamics, then exploitation of the cpn60 universal target, as demonstrated in this study, is a useful approach for its elucidation.