Longitudinal Analysis of the Prevalence, Maintenance, and IgA Response to Species of the Order Bacteroidales in the Human Gut

Abstract

Bacteroidales species are the most abundant Gram-negative bacteria of the human intestinal microbiota. These bacteria evolved to synthesize numerous capsular polysaccharides (PS) that are subject to phase variation. In Bacteroides fragilis, PS synthesis is regulated so that only one of the eight PS biosynthesis loci is transcribed at a time in each bacterium. To determine if the bacteria evolved this unusual property to evade a host IgA response, we directly studied the human fecal ecosystem. We performed a longitudinal analysis of the abundant Bacteroidales species from 15 healthy adults at four intervals over a year. For this study, we used bacterial culture to perform analyses not accurate with DNA-based methods, including quantification of total viable Bacteroidales bacteria, strain maintenance, and IgA responses. Abundant Bacteroidales isolates were identified to the species level using multiplex PCR and 16S rRNA gene sequencing. Arbitrarily primed PCR was used for strain typing. IgA responses to endogenous strains carried over the year were analyzed, and the orientations of the invertible PS locus promoters from the ecosystem were quantified. Subjects consistently harbored from 5 × 10⁸ to 8 × 10¹⁰ Bacteroidales bacteria/g of feces. Within the cohort, 20 different Bacteroidales species were detected at high concentrations. Bacteroides uniformis was the most prevalent; however, abundant Bacteroidales species varied between subjects. Strains could be maintained over the year within the ecosystem at high density. IgA responses were often not induced and did not correlate with the elimination of a strain or major changes in the orientations of the capsular PS locus promoters.

INTRODUCTION

As humans are born, they become rapidly colonized with microbes. Different body sites are colonized by characteristic microbes that differ from those at other sites and often include members that perform functions beneficial to maintaining the health of that particular niche. Interest in these human microbial ecosystems has grown tremendously in the last 5 years owing to their overall importance to human health and disease. This importance is exemplified by the launch in 2007 of the $145 million NIH-funded Human Microbiome Project, designed in part to catalog all of the microbes of the human body and provide complete genome sequences of predominant members of these ecosystems.

At no body site are the concentration and diversity of microbes greater than in the colon. Numerous 16S rRNA gene (11, 29) and metagenomic analyses (13) have analyzed the composition of the human gut microbiota and revealed that this ecosystem harbors up to 1,000 different species. Despite this large number, two phyla of bacteria predominate, the Bacteroidetes and the Firmicutes (reviewed in reference 1). These predominant members have been shown to provide key metabolic functions to their host (15), as well as important developmental (28) and immunologic properties (16, 22). The composition of the intestinal microbiota and the presence or shifts of particular types of bacteria have been correlated with diseases such as colitis, obesity, and colon cancer.

Within the Bacteroidetes phylum, those species that colonize the human gut are largely contained within the order Bacteroidales and collectively are the most abundant Gram-negative bacteria of this ecosystem. Bacteroidales contains more than 30 different human intestinal species, with the most predominant species contained within the genera Bacteroides and Parabacteroides, although Prevotella and Alistipes are also represented.

The genomes of these species reveal unique features that contribute to their success in this dense ecosystem. One conserved feature is the synthesis of a large number of phase-variable capsular polysaccharides (PS). This feature is not present in Bacteroidales species of the oral cavity (9), suggesting specific importance for intestinal survival. The prototype intestinal Bacteroidales strain for PS study is Bacteroides fragilis NCTC 9343, which synthesizes eight different PS. The eight PS biosynthesis loci are each arranged as an operon with a single promoter upstream of the first gene of each region. Seven of these PS biosynthesis locus promoters are flanked by inverted repeat regions, between which the DNA containing the promoter inverts, resulting in phase variation of each PS (17).

There is a second level of regulation of the PS involving the products encoded by the first two genes of each locus. The first, collectively designated the UpxY proteins, are transcriptional antitermination factors necessary to prevent premature transcriptional termination in the 5′ untranslated region of each operon (3). The proteins encoded by the second gene of each locus are termed the UpxZ proteins and function to inhibit the transcriptional antitermination property of subsets of heterologous UpxY proteins (4). The UpxZ proteins have differential inhibitory spectra, establishing a hierarchical mode of regulation. This system has evolved so that if more than one PS locus promoter is simultaneously oriented on, only one will be transcribed.

As would be expected based on the regulation of the PS, a mutant B. fragilis strain that synthesizes only one PS is able to compete for colonization of the gnotobiotic mouse intestine with wild-type bacteria (8), which is a population of bacteria synthesizing one each of the eight different PS. The question then arises as to why these intestinal Bacteroidales evolved the capacity to synthesize multiple phase-variable PS. As the regulatory system in B. fragilis evolved so that each bacterium transcribes only one PS biosynthesis locus at a time, the mechanism likely serves an evasion-type purpose. Many pathogenic bacteria synthesize surface molecules that are subject to phase variation for the purpose of evading an antibody-mediated response to these often immunogenic surface molecules. Very few studies have analyzed antibody responses to the resident intestinal bacteria in humans. As we are very interested in the biological significance of this process, we set out to determine if intestinal Bacteroidales evolved the capacity to synthesize multiple phase-variable PS for immune evasion. To address this question, we analyzed the bacteria from their natural human intestinal environment to avoid artifactual immunologic responses that may be generated by using animal systems.

The analyses described in this report used a culture-based system to isolate and identify the most abundant Bacteroidales species in at least four fecal samples each from 15 healthy adults over a 12-month period. Unlike other abundant bacteria of the human gut, Bacteroidales bacteria are exquisitely suited for culture-based analysis. Bacteroides and Parabacteroides strains are very aerotolerant, allowing for extended periods in air without loss of viable bacteria, and these species grow well on a selective medium. In performing this longitudinal analysis, we gained valuable data about the human intestinal Bacteroidales population, beyond what can currently be gleaned from DNA-based analyses of the ecosystem. In particular, we directly quantified the viable Bacteroidales species in these ecosystems over time. This culture-based method also allowed us to create a valuable strain bank for follow-up analyses, which was essential for the goals of this study. In addition, despite the enormous amount of data obtained from 16S rRNA gene and metagenomic analyses, they are limited to classification at the genus or species level and do not differentiate strains of the same species, an essential factor in this study.

We were able to address simple yet largely unresolved questions about the abundant Bacteroidales species of the human gut microbiota. These included how many total Bacteroidales bacteria there are within an individual's fecal microbiota, whether certain species are more prevalent than others, whether more than one strain of a species can cocolonize an individual, whether individuals can maintain a strain in their gut microbiota over the year of study, whether an intestinal IgA response to the PS of the resident Bacteroidales species is generated, and if so, whether this immune response correlates with the elimination of that strain from the intestinal microbiota or shifts in the expression of the various PS.

MATERIALS AND METHODS

Stool sample culture.

Fifteen healthy adults were recruited and consented to participate in this study. The study was approved by the Partners Human Research Committee IRB and complied with all relevant federal guidelines and institutional policies. Subjects with any history of gastrointestinal disease or chronic diarrhea were excluded. Participants included 7 females and 8 males 31 to 66 (mean, 42) years old who were of Caucasian, Latino, African American, or Asian descent. Fresh fecal samples were collected for at least four of the following time points: 0, 1, 3, 6, 8/9, and 12 months. The 1-month sample was used for quantification of total Bacteroidales species (Table 1) and was screened only to detect B. fragilis for bft (B. fragilis enterotoxin gene) analysis (Table 2) and was not considered for the species analysis/prevalence studies. Stool samples were self-collected in sterile screw-cap specimen cups, transported at room temperature to the laboratory, and processed within 6 h of collection. Samples were diluted in phosphate-buffered saline (PBS) and plated on brucella laked blood, kanamycin, vancomycin plates (PML Microbiologicals), which are selective for Bacteroidales species from fecal samples grown anaerobically. No Bacteroidales bacteria were selected against, nor was the colony count diminished when the fecal sample was processed 6 h after collection versus immediately (data not shown). All species of Bacteroides, Parabacteroides, and Prevotella that were tested grew on these plates. Alistipes species were excluded from this study, as it was unclear whether these organisms grow under these culture conditions. The plates were placed in an anaerobic chamber at 37°C for 4 days. Colonies arising on the two highest-dilution plates (usually the 10⁻¹⁰, 10⁻⁹, and 10⁻⁸ plates) were counted, and the species were identified by a series of multiplex PCR assays (Fig. 1). Therefore, we were analyzing only the most prevalent species in each sample. Colonies that were not unambiguously identified at the species level were further processed by sequencing the 16S rRNA gene. All colonies on the highest-dilution plate were identified at the species level, and several isolates representing different colony phenotypes (large, small, mucoid, color) on the second-highest-dilution plate were identified. The number of colonies analyzed ranged from 14 to 51 per sample (depending on colony phenotypes), with an average of 29. A minimum of one isolate of each species identified at each time point was frozen at −80°C, with the exception that all identified B. fragilis bacteria were frozen.

Table 1.

Total Bacteroidales CFU in fecal samples from 15 human subjects over 1 year

Subject	Total no. of Bacteroidales CFU/g of fecal sample at:
	0 mo	1 mo	3 mo	6 mo	8/9 mo	12 moa
1	3 × 1010	1 × 1010	3 × 1010		1 × 1010	3 × 1010
2	7 × 109	2 × 109	6 × 109		5 × 108	2 × 109
3	3 × 1010	8 × 109	1 × 1010	1 × 1010		2 × 109, 3 × 109
4	6 × 108	1 × 109	1 × 109	2 × 109		1 × 109
5	5 × 109	7 × 109	6 × 108	8 × 109		2 × 109, 4 × 109
6	1 × 109	9 × 109	1 × 109	3 × 109		3 × 109
7	2 × 109	3 × 109	1 × 109	4 × 109		2 × 1010
8	3 × 1010	1 × 1010	2 × 1010	2 × 1010		2 × 109, 6 × 109
9	1 × 109	1 × 1010	3 × 109	3 × 109		1 × 109, 2 × 109
10	2 × 1010	2 × 1010		2 × 109		3 × 1010, 8 × 1010
11	9 × 109	6 × 109		1 × 1010	7 × 108	2 × 109, 8 × 109
12	2 × 1010	1 × 1010	2 × 1010	2 × 1010		2 × 1010, 2 × 1010
13	5 × 109	2 × 109	1 × 109	5 × 108		1 × 109, 4 × 109
14	2 × 1010	2 × 1010	2 × 1010		2 × 1010	2 × 1010, 2 × 1010
15	2 × 1010	3 × 1010	1 × 1010	2 × 1010		2 × 1010, 3 × 1010

^aTwo different regions of various 12-month fecal samples were analyzed.

Table 2.

Proportions of B. fragilis isolates containing bft

Subject	Proportion of B. fragilis isolates containing bft
	0 mo	1 moa	3 moc	6 mo	9 mo	12 mo	Total
1	0/2	0/7		0/2b		0/11
2			0/1					0/1
3	0/7	0/2	0/5	0/2		0/2	0/18
4	0/18	0/1	0/2	1/6			1/27
5	1/1b		2/2	1/1		5/5	9/9
7	21/21	3/3	3/3	3/3	6/6	36/36
8	0/2		0/1		0/1	0/4
9		0/2					0/2
10	0/1	0/7				0/3b	0/11
11	0/3b					0/3
13		5/5	4/4	1/2		4/4b	14/15
14	17/17	6/6	4/4		5/5	7/7	39/39
15						2/2	2/2

^aThe 1-month collection was screened only for B. fragilis and therefore not used in other tables and figures in this report.

^bSome of these B. fragilis isolates were present on the third-highest-dilution plate and therefore not included in the other analyses of this study.

^cThe B. fragilis isolates from the 3-month sample of subject 1 were not frozen and therefore were not available for bft typing.

Fig. 1.

Multiplex PCR analyses. EtBr-stained agarose gels show the PCR product(s) resulting when each Bacteroidales species was analyzed by multiplex PCR I (top panel), multiplex PCR II (middle panel), and multiplex PCR III (bottom panel). The following type strains were used: B. thetaiotaomicron ATCC 29741, B. vulgatus ATCC 8482, B. fragilis NCTC 9343, B. caccae ATCC 43185, B. ovatus ATCC 8483, P. distasonis ATCC 8503, P. merdae ATCC 43184, B. uniformis ATCC 8452, and B. finegoldii DSM 17565. The B. stercoris, B. eggerthii, B. intestinalis, B. dorei, and P. johnsonii strains were all obtained from this study and confirmed by full 16S rRNA gene sequencing.

Multiplex PCR assays.

The multiplex PCR assays are based on the design of Liu et al. (18) and take advantage of the variable region between the 16S and 23S rRNA genes. All primers used in this study are listed in Table S1 in the supplemental material. Taq Mastermix (NEB) was used for all PCRs, and the conditions were as follows: multiplex PCR I, 2 min at 94°C and then 35 cycles of 94°C for 30 s, 59°C for 30 s, and 68°C for 45 s with a final 60-s extension at 68°C. The same conditions were used for multiplex PCRs II and III except that the annealing temperature was 52°C. PCR products were resolved on 1.4% agarose gels. Larger colonies were first screened using multiplex PCR I, if the species was not identified, multiplex PCR II was performed, and finally multiplex PCR III. Small colonies were first screened using multiplex PCR II, followed by multiplex PCRs I and III. The species Bacteroides caccae, Bacteroides ovatus, Bacteroides fragilis, Bacteroides thetaiotaomicron, Bacteroides vulgatus, and Bacteroides xylanisolvens consistently yielded larger colonies. The species Bacteroides dorei, Bacteroides intestinalis, and Bacteroides uniformis and all Parabacteroides spp. yielded smaller colonies. Bacteroides eggerthii and Bacteroides stercoris yielded colonies of various sizes. Many isolates unresolved by multiplex PCR were identified to the species level by sequencing of a 286-bp region of the 16S rRNA gene (16S rRNA region 1; see Table S1 in the supplemental material). All isolates that were identified as B. vulgatus or B. dorei by multiplex PCR or sequencing of the 286-bp 16S rRNA gene region were retested using primer sets that targeted separate regions of the 16S rRNA gene. Similarly, all strains identified by multiplex PCR as B. ovatus or by 286-bp 16S rRNA gene sequencing as B. ovatus or B. xylanisolvens were rescreened with a new primer set to discriminate these species. For a few isolates, sequencing of an additional portion of the 16S rRNA gene was necessary. All isolates identified as B. stercoris by multiplex PCR were further analyzed by sequencing of the 286-bp 16S rRNA region. In total, we performed full or partial 16S rRNA gene sequencing of 237 isolates.

Arbitrarily primed PCR (AP-PCR).

Chromosomal DNA was isolated from in vitro-grown bacteria. Each AP-PCR was performed using a single 10-bp primer. Each 20-μl PCR mixture consisted of 100 ng of chromosomal DNA, 10 μl of Taq 2× master mix, and 288 pmol of primer. Reaction conditions for all three AP-PCRs were as follows: 2 min at 95°C, followed by 32 cycles of 90°C for 30 s, 28°C for 30 s, and 68°C for 2 min.

PCR screening of B. fragilis isolates for the enterotoxin gene (bft).

PCR was performed using previously described primers that amplify all three isoforms of bft (5) with the following PCR conditions: 95°C for 2 min and then 32 cycles of 95°C for 30 s, 62°C for 30 s, and 68°C for 60 s.

Analysis of IgA, IgG, and IgM responses to endogenous Bacteroidales species.

Aliquots of frozen fecal samples were resuspended in 10 volumes of PBS. For some samples, complete, EDTA-free protease inhibitor cocktail tablets (Roche) were added. All particulate debris was removed by two successive centrifugations at 13,000 × g, and the resulting supernatant was used as the source of soluble Igs. The total IgA in each sample was quantified using the Human IgA ELISA kit (Immunology Consultants Lab, Inc.). To determine if there was IgA specific to Bacteroidales species, whole-cell lysates of in vitro-grown bacteria were electrophoresed on 4 to 12% NuPAGE Bis-Tris gels (Invitrogen). Contents were transferred to polyvinylidene difluoride membranes and probed with the Ig preparation at a final dilution of 1:100. Alkaline phosphatase-conjugated goat anti-human IgA (Invitrogen), IgG (Biosource), or IgM (Sigma) was the secondary antibody used at 1:1,000, and the blots were developed with 5-bromo-4-chloro-3-indolylphosphate (BCIP) substrate (KPL).

Quantitative analysis of the B. fragilis capsular PS promoter orientations in the human ecosystem.

Aliquots of frozen fecal samples were resuspended in 10 volumes of PBS. Particulate material was sedimented for 5 min by gravity, and 200 μl of the nonsediment material was used for chromosomal DNA extraction with the Extract-Master fecal DNA extraction kit (Epicentre). For each PS biosynthesis locus promoter, a PCR was performed with primers that flanked the invertible promoter region and the resulting product was digested with a restriction enzyme that cleaves asymmetrically between the inverted repeat elements (32) (see Fig. 5A). Due to differences between the genomic sequences of the strains, neither the PSA promoter region of subject 14 nor the PSF promoter region of subject 7 could be analyzed by this method. To analyze the PSF promoter region of subject 14, a new primer set was designed. All PCR programs had an initial incubation of 2 min at 94°C; 35 cycles of 94°C for 30 s, 59°C for 30 s, and 68°C for 90 s; and then 2 min at 68°C. Fragments were separated by electrophoresis on 2% agarose gels and visualized by ethidium bromide (EtBr) staining. PS promoter orientations were quantified by measuring relative band intensities using ImageJ software (available at http://rsbweb.nih.gov/ij/).

Fig. 5.

Orientations of the B. fragilis PS locus promoters from bacteria in the fecal samples of subjects 7 and 14. (A) Schematic of the PCR digestion method used for quantitative promoter orientation analysis of the PSH locus. The inverted repeats (IR) flanking the invertible DNA region are shown with the promoter (p) indicated. The primers used for PCR and the restriction site are indicated. The sizes of the fragments resulting from digestion of the PCR product when the promoter is in both the on and off orientations are shown. (B and C) EtBr-stained agarose gels demonstrating the fragments resulting from PCR digestion of each invertible PS locus promoter region for 0-, 6-, and 12-month samples from subject 7 or 14. (D and E) Quantification of the percentages of B. fragilis bacteria with each of the PS locus promoters oriented on in the 0-, 6-, and 12-month samples from subjects 7 (D) and 14 (E).

RESULTS

Analysis of Bacteroidales in human stool samples.

Analysis of all 74 fecal samples collected from the 15 subjects demonstrated that the total culturable Bacteroidales concentrations ranged from 5 × 10⁸ to 8 × 10¹⁰ CFU/g of wet stool sample (Table 1). Quantification of the total Bacteroidales species in two separate portions of the same stool sample revealed a slight variation in the total Bacteroidales concentrations, never more than 0.5 log (Table 1, 12-month time points). We used multiplex PCR assays (18) to identify to the species level colonies growing on plates corresponding to the two highest dilutions (usually 10⁻¹⁰, 10⁻⁹, and 10⁻⁸ dilution plates) (Fig. 1). Any colony not unambiguously identified by multiplex PCR was identified to the species level by sequencing a region(s) of the 16S rRNA gene. In total, 20 different Bacteroidales species were identified from 59 samples (1-month samples were not analyzed by multiplex PCR for species identification) (Fig. 2). The only non-Bacteroidales species identified were vancomycin-resistant Clostridium innocuum and Lactobacillus ruminis.

Fig. 2.

Number of stool samples from which a given Bacteroidales species was identified on the two highest-dilution plates from a total of 59 stool samples.

Some Bacteroidales species, such as B. uniformis, B. ovatus, B. thetaiotaomicron, B. vulgatus, B. fragilis, and P. distasonis, were on one of the two highest-dilution plates in more than half of the samples (Fig. 2; see Table S2 in the supplemental material). These species were also those that were detected in the greatest number of subjects, with B. uniformis detected in all 15 subjects and B. ovatus and B. thetaiotaomicron detected in all subjects but one (see Fig. S1 in the supplemental material). B. fragilis, which had been considered to be less abundant than other Bacteroides species of the human gut, was detected at high levels in 31 out of the 59 samples (Fig. 2) and was detected in at least one sample from 11 of the 15 subjects (see Fig. S1 and Table S2 in the supplemental material). Other Bacteroidales species, such as Bacteroides massiliensis, Prevotella copri, Parabacteroides goldsteinii, B. eggerthii, Bacteroides salyersiae, Parabacteroides johnsonii, and Bacteroides nordii, were detected at high concentrations in fewer samples (Fig. 2; see Fig. S1 in the supplemental material). The species detected varied slightly when two different sections of the same stool sample were analyzed (see Table S2 in the supplemental material, 12 months).

Enterotoxigenic B. fragilis carriage.

B. fragilis is unique among the intestinal Bacteroides species in that some strains contain bft, which encodes an enterotoxin (27). Those B. fragilis strains containing bft, termed ETBF, have been correlated with acute and persistent diarrheal disease (27), flare-ups of inflammatory bowel disease (2, 25), and colorectal cancer (31, 34). One of the advantages of the culture-based approach to species identification is that the isolates were frozen for follow-up analyses, such as bft typing. For these analyses, we included B. fragilis isolates from the 1-month samples and on the third-highest-dilution plate, yielding a total of 178 B. fragilis isolates from 13 of the15 subjects. Using a previously described PCR screening method that detects all three isotypes of bft, we found that the genomes of B. fragilis strains from six subjects contained bft (Table 2). For four of these six subjects, all isolated B. fragilis strains were ETBF, and in three subjects, intestinal carriage of ETBF was detected over the entire year of study. For two subjects (4 and 13), both toxigenic and nontoxigenic strains were detected in the same sample (Table 2 and Fig. 3 D).

Fig. 3.

Strain typing by AP-PCR. (A) AP-PCR amplicons resulting from four longitudinally collected B. fragilis isolates from subjects 7 and 5 and B. fragilis type strain NCTC 9343. The top panel is AP-PCR1, and the bottom panel shows the products from the same isolates with AP-PCR3. (B) AP-PCR amplicons resulting from four longitudinally collected B. fragilis isolates from subjects 14 and 3 and B. fragilis NCTC 9343. The top panel is AP-PCR1, and the bottom panel shows the products from the same isolates with AP-PCR3. (C) AP-PCR amplicons resulting from four longitudinally collected B. uniformis isolates from subjects 14 and 3 and B. uniformis type strain ATCC 8452. The top panel is AP-PCR2, and the bottom panel shows the products from the same isolates with AP-PCR3. (D) AP-PCR products amplified from six B. fragilis isolates from the 6-month stool sample of subject 4. The top panel is AP-PCR1, the middle panel AP-PCR3, and the bottom panel shows amplification of bft.

Strain typing.

One of the main purposes of this study was to determine if synthesis of multiple phase-variable PS by the intestinal Bacteroidales species evolved to allow the organism to evade a host-mediated antibody response. As there is great intraspecies genetic variability in the capsular PS biosynthesis loci (6, 7, 10, 24), different strains of the same species synthesize antigenically and immunologically distinct PS. Therefore, it is necessary to strain type the isolates in order to analyze local host antibody responses to a particular strain over time. We selected isolates of species that were carried at high density at all time points over the year of study and focused primarily on subjects consistently harboring B. fragilis. Strain typing of selected isolates was performed using AP-PCR assays (33). We found that the combination of two AP-PCRs was able to discriminate different strains of the same species. Using these AP-PCRs, we found that subjects 3, 5, 7, and 14 each harbored a consistent B. fragilis strain over the year of study which differed from the B. fragilis strains harbored by the other subjects (Fig. 3A and B). In addition, subjects 3 and 14 each harbored a consistent B. uniformis strain over the year of study (Fig. 3C). This was not true of all subjects, as the results demonstrated that some individuals harbored different strains over time or were colonized by two strains of the same species simultaneously. For example, we confirmed that the one bft-positive isolate from the 6-month sample from subject 4 produced an AP-PCR profile different from that of the bft-negative isolates that predominated in this individual (Fig. 3D). These data demonstrate that an individual can simultaneously harbor two distinct strains of the same Bacteroidales species at high concentrations.

Local antibody responses to abundant Bacteroidales species.

We sought to determine if humans generate a fecal IgA response to their endogenous Bacteroidales species. A comprehensive study has shown that intestinal IgA obtained from fecal samples has a very high correlation with specific IgA obtained from intestinal lavage (14). Therefore, the soluble fraction of the fecal samples was used as the source of antibody for these analyses. For some samples, protease inhibitor was added to the fecal samples, but no differences in specific or total IgA was detected with or without protease inhibitor. The total IgA levels in these soluble fractions ranged from 62 to 1,137 μg/g of stool sample (Fig. 4). We first analyzed if there was an IgA response in the 3-month fecal sample of subject 1 to each of the species that were isolated from the initial (month 0) sample of this subject. No IgA response to the high-molecular-weight capsular PS of these strains was detected (Fig. 4A). The only IgA response detected in these samples was to a P. distasonis molecule that migrated just above the 98-kDa marker and may be an S-layer protein previously described in this species (12). We next analyzed the antibody response to strains carried at high density over the year of study from subjects 3, 5, 7, and 14. The B. uniformis and/or B. fragilis isolates from the 0-, 3-, 6-, and 12-month samples from these subjects were grown in culture, and cell lysates were processed for Western immunoblotting. The blots were probed with the soluble fraction of the 12-month stool sample of the corresponding subject at a final dilution of 1:100. Subjects 3 and 5 did not have detectable IgA to the resident B. uniformis and/or B. fragilis strains they carried at high density over the year of study (data shown for subject 3, Fig. 4B). Interestingly, there was a slight IgA reactivity to the PS of B. fragilis type strain NCTC9343 in subject 3, indicating that this person was likely colonized at one point with a strain that synthesized at least one PS in common with this strain (Fig. 4B). In subjects 7 and 14, there was an IgA response to the heterogeneously sized high-molecular-weight PS of the B. fragilis strains colonizing these individuals (Fig. 4C and D) but no IgA response to the B. uniformis strain of subject 14 (negative data not shown). We analyzed if there was an IgM or IgG at 12 months in subjects 5 and 7 to their resident B. uniformis and/or B. fragilis strains and found no detectable responses. Next, we analyzed if the IgA response to the B. fragilis PS of subjects 7 and 14 were present at the initial collection point and found that these immune responses had persisted in these subjects for the year of study (Fig. 4E, results for subject 7 shown).

Fig. 4.

Western immunoblot analysis of local human antibody responses to abundant endogenous Bacteroidales strains. (A) IgA reactivity in the 3-month sample from subject 1 to the Bacteroidales strains isolated from this subject at the initial collection point. (B to D) IgA reactivity of the 12-month sample from subjects 3, 7, and 14 to the endogenous B. fragilis strain that each had carried over the year of study. (E) IgA reactivity from the initial stool sample from subject 7 to the same strains of panel C.

B. fragilis capsular PS promoter orientations in subjects 7 and 14.

Subjects 7 and 14 both mounted an IgA response to their endogenous B. fragilis strain, yet these strains persisted in these ecosystems in the presence of this immune response. If the synthesis of multiple phase-variable PS evolved to allow these bacteria to persist in the presence of an IgA response, we would expect to observe shifts in the PS synthesized by these bacteria over time. Different strains of B. fragilis synthesize structurally and immunologically distinct PS. Therefore, it is not feasible to directly analyze the PS itself because we lack knowledge of the PS types of each strain and immunological reagents for their analysis. The PS promoter regions and the first two regulatory genes of each region are relatively conserved between strains. This allows for the analysis of the relative orientations of the seven invertible PS biosynthesis locus promoters directly from fecal samples. By combining the data on the orientations of these promoters with our knowledge of the inhibitory spectra of the UpxZ proteins encoded by each locus, we can predict PS expression in vivo. DNA was isolated from the 0-, 6-, and 12-month fecal samples of subjects 7 and 14. For each promoter region, a PCR digestion assay was performed (32) which quantified the percentages of bacteria from a mixed population with an invertible DNA region in each orientation. For each region, a PCR was performed with primers that anneal outside the invertible region so that there was no bias in the PCR for either the on or the off orientation (schematic shown in Fig. 5 A). The product is digested with an enzyme that cuts asymmetrically between the invertible DNA region. When a promoter is in the on orientation for transcription of the downstream PS biosynthesis operon, the two resulting PCR fragments are different in size from the two fragments that result when this promoter is in the off orientation. These PCR fragments are resolved on an agarose gel and EtBr stained, and densitometry is used to quantify the bands corresponding to the on and off orientations. These values correlate with the percentage of the B. fragilis population with these PS promoters oriented in each direction. In subject 7, there are some changes in the percentage of bacteria with the PSA, PSE, and PSH locus promoters oriented on when comparing the initial collection to the 6-month sample (Fig. 5B and D). Most notable is that in the initial sample, the PSE promoter was on in 75% of the population, in only 37% in the 6-month sample, and in 24% of the population in the 12-month sample. This was the greatest shift observed among any of the promoters analyzed. In particular, there were very few changes in the population between the 6- and 12-month samples. Similarly, the B. fragilis population of subject 14 was extremely stable with regard to PS locus promoter orientations over the year of study (Fig. 5C and E). In nearly all of the B. fragilis bacteria of this ecosystem, the PSG locus promoter was consistently oriented on. Our in vitro studies have shown that the UpgZ protein, encoded by the PSG region, has a very wide spectrum of inhibition, preventing the transcription of all PS biosynthesis loci except PSG and PSH (4). Due to the fact that nearly all of the B. fragilis isolates from subject 14 had the PSH biosynthesis locus promoter oriented off at all time points, we predict that nearly all of the B. fragilis bacteria of subject 14 consistently and exclusively expressed PSG over the year of study.

DISCUSSION

This study reveals novel findings about the abundant Bacteroidales species present in human feces, many of which have not been obtained by DNA-based analyses. The data revealed that the total Bacteroidales concentration within these 15 subjects was between 5 × 10⁸ and 8 × 10¹⁰ CFU/g of wet feces. An individual can simultaneously harbor many different Bacteroidales species at high concentrations; in one individual, 10 species were identified at high concentrations in a single sample. Within the cohort of 15 healthy adults, we identified a total of 20 different Bacteroidales species at high concentrations. Although there was not a predominant Bacteroidales species present in the ecosystems of all individuals, particular species such as B. uniformis, B. ovatus, B. thetaiotaomicron, and B. vulgatus were more frequently identified at high concentrations within these human samples. While DNA-based analyses may be limited to the identification of organisms at the genus or species level, our analyses allowed classification at the strain level. Our results indicate that an individual can harbor more than one strain of the same species at one time and can maintain the same strain in his or her ecosystem for at least a year.

Previous studies reported recovery rates of B. fragilis of 6 to 70% (27), whereas B. fragilis was cultured from 87% of our healthy cohort, including those on the third-highest-dilution plate. Several reports have also documented the prevalence of ETBF in human samples (2, 27, 35), as well as the longitudinal carriage of ETBF in children with diarrhea (23, 26). In our study, 6 of the 13 subjects with B. fragilis harbored ETBF and 4 of these individuals carried these strains at high density throughout the year. Since we only analyzed strains present at high concentrations, it is possible that the ETBF carriage rate is even greater. Previously, ETBF carriage in human stool samples from healthy subjects has been reported at between 2 and 30% (2, 27, 35), whereas ETBF carriage was 40% within our 15 subjects. As ETBF has been correlated with persistent diarrhea in children, colitis, and the development of colon cancer, the long-term carriage of ETBF deserves further study.

Very few studies of the human gut microbiota have analyzed local IgA responses to predominant members, and no studies have analyzed specific IgA responses to abundant Bacteroidales species in the normal adult human gut. In this study, we did not detect IgA to several strains harbored at high density for the year of study. For those strains to which an IgA response was mounted, there was no correlation with either elimination of that strain or major alterations in the orientations of the invertible PS locus promoters. Thus, IgA responses to these bacteria likely do not influence the composition of the intestinal microbiota in the normal adult human gut. Rather, studies suggest that IgA helps maintain the luminal compartmentalization of intestinal bacteria by reducing the densities of the surface-associated bacteria and restricting the penetration of the epithelial layer by these bacteria (19, 21, 30). A study with mice showed that some commensals penetrate the epithelial cell layer and survive within dendritic cells, triggering the production of IgA (20). It is possible that the B. fragilis strains of subjects 7 and 14 interacted closely with the mucosal surface and were sampled by dendritic cells. The IgA response to these bacteria might then limit this interaction with the host mucosal surface.

The data strongly suggest that the synthesis of multiple phase-variable PS by Bacteroides species did not evolve to facilitate evasion of a host-mediated immune response. In subject 14, very few changes in the orientations of the PS locus promoters of the endogenous B. fragilis population were detected, despite the long-term IgA response to these molecules. In these relatively stable ecosystems, there is little variation in the orientations of these promoters over time. There are, however, substantial differences in the orientations of various B. fragilis PS locus promoters when organisms of the two ecosystems analyzed are compared. The PSG promoter was oriented off in nearly all of the B. fragilis population of subject 7, while nearly all of the B. fragilis bacteria of subject 14 had this promoter oriented on. A previous study with mice suggests that different host gut microbiotas contribute to differences in B. fragilis PS promoter orientations (32). Demonstration of such a correlation awaits further study.

It is possible that these species evolved to synthesize multiple phase-variable PS to evade other factors such as phage, products synthesized by other members of the microbiota, or predatory organisms. In this scenario, substantial shifts in the orientations of these promoters and thus PS expression would occur following introduction of the specific PS-targeting factor. Analysis of human gut microbiotas in which the B. fragilis capsular PS locus promoters demonstrate substantial changes over time may reveal factors contributing to the evolution of this biological process.