Previous studies have reported the identification of the BSAP that are specific for B. vulgatus (BSAP-3) or B. uniformis (BSAP-2) [12, 13]. For BSAP-3, membrane attack perforin/complex protein sequences (MACPF) were identified in a gene cluster 16 from B. vulgatus/B. dorei that had activity to restrict growth in a species-specific manner of B. vulgatus. Similarly, in B. uniformis, MACPF was also identified that could restrict species-specific replication of B. uniformis which was named BSAP-2. To determine the presence of the gene encoding BSAP-3 or BSAP-2, the individual samples from 3 longitudinal data sets used in this study were used for the gene analysis based on the MACPF cluster 15 and 16 sequence reads using BLAST+ [15] and IGV tools [16].
We first analyzed the human microbiome (HMP) data set that is composed of paired samples taken at separate times (up to 1 year apart) [2, 17]. We selected 30 paired samples that we have previously shown to have related strains of B. vulgatus and B. uniformis over periods between 3 months and 1 year. To satisfy the parameters for WSS analysis to determine relatedness, the sequence reads that had a depth of greater than 3.5 with a minimum sequence coverage greater than 30% were used for further analysis [1]. We screened for the complete genes encoding BSAP-3 for B. vulgatus and BSAP-2 for B. uniformis in each sample pair (Supplementary Table 1). We identified 11 sample pairs out of 30 for the complete gene encoding BSAP-3, and 7 of 28 samples for gene encoding BSAP-2 (2 sample pairs, S64, and S70 had WSS scores below the cutoff value for B. uniformis) (Figure 1). For 30 BSAP-3 and 28 BSAP-2 sample pairs, we found the same BSAP phenotype (either positive or negative) for time points taken at times that were different by 3-6 months.
To further characterize the BSAP negative phenotype, we used the IGV to display the 5’ and 3’ sequences bracketing the BSAP deletion [16] (Supplementary Figures 1A and 1B). We found that 4 of the 19 sample pairs for the BSAP-3 deleted phenotype were identical (S28, S31, S48, and S66) and 10 of the 19 sample pairs (S16, S17, S21, S23, S49, S60, S64, S65, S69, and S70) had a few nucleotide differences bracketing the deletion from the comparison of the two (a and b) time points (Figure 2A and 2B, and Supplemental Figure 1A). Similarly, for BSAP-2 from the HMP, we found 2 of the 20 (S21 and S66) BSAP-2 negative phenotypes had identical 5’ and 3’ sequences and 18 of the 20 sample pairs (S5, S10, S14, S16, S17, S19, S20, S23, S24, S28, S29, S31, S48, S51, S57, S60, S61, and S65) had only a few nucleotide differences 5’ and 3’ sequences bracketing the deletion from the comparison of the a and b time points (Figure 3A and 3B, and Supplemental Figure 1B). Thus, these results establish the existence of a reservoir of these strains for both B. vulgatus and B. unformis in these individuals that is stable over an extended time (3-6 months) that is used in the selection of the dominant fecal strain.
From a further inspection of the BSAP negative phenotype sample sets with IGV, we found a subset of Individuals for BSAP-3 (S5, S14, S18, S62, and S63) and one individual (S49) for BSAP-2 that lacked a complete BSAP gene but did have partial (truncated) sequence fragments in the BSAP gene region. The BSAP-3 in individual S14 was particularly interesting since the B. vulgatus in the two samples were related as determined by WSS with the first sample containing a deleted BSAP-3 while the second, later sample had a partial BSAP-3 gene (Supplementary Figure 1A). Most importantly, we found that the pattern of the partial BSAP genes in these individuals was unique to that sample and thus differed between the two sample times.
To determine the impact of the BSAP gene phenotype following perturbation and recovery of the gut ecosystem, we next analyzed a data set from a previous study that analyzed gut microbes following a single antibiotic treatment [5, 18]. This data set contained the collection of fecal samples from 6 individuals at day 0, day 7 and day 90 that were used as untreated controls (Supplemental Table 2). Analysis of the BSAP genes from these samples found that 3 of the 6 were BSAP-3 positive, while 2 of the 6 were positive for BSAP-2 (Supplementary Table 2 and Supplementary Figures 2A and 2B). The analysis from the control samples for the Raymond et al. study gave similar results with respect to the analysis of the sequences bracketing the BSAP deletions as that found for the analysis of the HMP data set (Supplementary Table 2 and Supplementary Figures 2A and 2B).
We next selected 17 individual sample sets from the Raymond et al. study that were given a treatment of cephalexin with samples collected one day after treatment and 82 days after (a total of 90 days after pre sample). We had previously analyzed this data set using our strain tracking analysis to show recovery of the pre-antibiotic strain at the day 90 samples for most individuals. Analysis of BSAP genes from the pre sample demonstrated that 3 of the 17 were BSAP-3 and 4 out of 17 were BSAP-2 positive (Figure 4). Although for most of the sample sets the BSAP phenotype did not change following antibiotics, we did find for BSAP-3 one individual (P15) was partial BSAP-3 at day 0 and following the antibiotics at day 7, but by day 90 was BSAP-3 positive with no strain change (Figure 4A, Supplementary Figures 3A, and Supplementary Table 2). In another individual (P17), we found a new strain on day 7 that was partial BSAP-3, but at day 90 the strain was related to the pre sample and was also BSAP-3 positive. Finally, one individual of note, P4, had a deleted BSAP-3 gene at days 0 and 7, with a B. vulgatus sub strain at day 90 with a complete BSAP-3 gene. This individual was the only example where we found a BSAP phenotype change from negative to positive without a strain change (Supplementary Figure 3A). The correlation of the dominant fecal strain in these two individuals with the presence of the BSAP positive phenotype supports that the BSAP-3 positive phenotype restricted the B. vulgatus with the BSAP-3 negative phenotype in the recovery of the GIT ecosystem following the cephalexin perturbation.
To further characterize the impact of the antibiotic on the BSAP phenotype negative samples, we next analyzed the sequence changes using IGV [16]. For the BSAP-3 phenotype negative samples, we found 8 samples in which the strains were related at Day 0 and Day 90 had identical or few nucleotide differences in the 5 and 3 regions bracketing the BSAP-3 deletion (P2, P3, P9, P13, P18, P19, P20, and P21) demonstrating the recovery of the pre antibiotic sub strain (Figure 4A, and Supplementary Figure 3A). For the BSAP-2 negative phenotype samples, 10 samples that had BSAP-2 phenotype negative for the Day 0 and Day 90 samples all had identical or few nucleotide differences in the 5’ and 3’ regions bracketing the deleted BSAP gene (P2, P9, P10, P11, P13, P15, P17, P18, P20, P21) (Figure 4B and Supplementary Figure 3B). Collectively, these results are consistent with the resiliency of the strain reservoir in the capacity to recover following perturbation of the GIT.
To investigate the impact of a greater perturbation on the strain reservoir, we characterized the strain stability of individuals that had been treated with a suppressive antibiotics cocktail consisting of 3 antibiotics (meropenem, gentamicin, and vancomycin) [5, 19]. In this longitudinal study, we found that in some instances the treatment resulted in a strain change from the pre antibiotic strain. Overall, 2 of 10 individuals showed BSAP-3 positive and 3 of 10 showed BSAP-2 positive (Figure 5, Supplementary Figure 4, and Supplementary Table 3). We found several examples where there was a change in an individual from BSAP negative to the BSAP positive phenotype. Individual P1 had a strain change early after antibiotics that were BSAP-3 negative, but the same samples at later times (42 and 180 days) were the same strain as the pre-strain and were both also BSAP-3 positive (Figure 5A). Another individual (P4) that was a partial BSAP gene on day 0 had a strain change with the new strains at day 42 and 180 that were both BSAP-3 positive (Figure 5A). Similarly, in this individual, for B. uniformis there was also a new strain at days 42 and 180 that were both BSAP-2 positive (Figure 5A and 5B). Finally, for P11 BSAP-2 where there was no change in the strain of B. uniformis but the BSAP-2 phenotype changed from partial to positive (Figure 5B).
IGV analysis of the BSAP-3 negative samples revealed a complex pattern, most probably due to the strain change in individuals that occurred following the 3 antibiotics. For BSAP-3, we found that only 1 of 10 samples (P1) had positive BSAP-3 at days 0, 42, and 180 (Figure 5A). We did find two individuals, P11 and P12 where the B. vulgatus strain at Day 0 was related to the strain at Day 180. In the case of P11, the 5 and 3 regions at Day 0 and Day 180 fecal dominant sub strains were different, whereas the 5 and 3 sequences from P12 were the same at Day 0 and 180 (Figure 5A and Supplementary Figure 4A). For B uniformis, individuals P3 and P9 had related strains at Day 0 and 180 and both the 5 and 3 sequences were the same (Figure 5B and Supplementary Figure 4B).