The pervasive effects of an antibiotic on the human gut microbiota, as revealed by deep 16S rRNA sequencing

Abstract

The human intestinal microbiota is essential to the health of the host and plays a role in nutrition, development, metabolism, pathogen resistance, and regulation of immune responses. Antibiotics may disrupt these coevolved interactions, leading to acute or chronic disease in some individuals. Our understanding of antibiotic-associated disturbance of the microbiota has been limited by the poor sensitivity, inadequate resolution, and significant cost of current research methods. The use of pyrosequencing technology to generate large numbers of 16S rDNA sequence tags circumvents these limitations and has been shown to reveal previously unexplored aspects of the "rare biosphere." We investigated the distal gut bacterial communities of three healthy humans before and after treatment with ciprofloxacin, obtaining more than 7,000 full-length rRNA sequences and over 900,000 pyrosequencing reads from two hypervariable regions of the rRNA gene. A companion paper in PLoS Genetics (see Huse et al., doi: 10.1371/journal.pgen.1000255) shows that the taxonomic information obtained with these methods is concordant. Pyrosequencing of the V6 and V3 variable regions identified 3,300-5,700 taxa that collectively accounted for over 99% of the variable region sequence tags that could be obtained from these samples. Ciprofloxacin treatment influenced the abundance of about a third of the bacterial taxa in the gut, decreasing the taxonomic richness, diversity, and evenness of the community. However, the magnitude of this effect varied among individuals, and some taxa showed interindividual variation in the response to ciprofloxacin. While differences of community composition between individuals were the largest source of variability between samples, we found that two unrelated individuals shared a surprising degree of community similarity. In all three individuals, the taxonomic composition of the community closely resembled its pretreatment state by 4 weeks after the end of treatment, but several taxa failed to recover within 6 months. These pervasive effects of ciprofloxacin on community composition contrast with the reports by participants of normal intestinal function and with prior assumptions of only modest effects of ciprofloxacin on the intestinal microbiota. These observations support the hypothesis of functional redundancy in the human gut microbiota. The rapid return to the pretreatment community composition is indicative of factors promoting community resilience, the nature of which deserves future investigation.

Conflict of interest statement

Competing interests. The authors have declared that no competing interests exist.

Figures

Figure 1. Genetic Distances between Tags and Reference Sequences

The genetic distances between each unique V6 tag (A) and V3 tag (B) and its best hit (or hits) in the appropriate reference database is shown on the vertical axis (see Methods for details); the abundance of the tag is shown on a logarithmic scale on the horizontal axis. Horizontal arrows indicate cumulative totals of tags from most to least abundant. Discrete horizontal rows of symbols near the x-axis correspond to an integer number of nucleotide changes relative to a reference sequence; vertical arrows indicate cumulative tag totals for up to 0, 1, or 2 changes. Abundant tags are likely to be identical to sequences in the reference databases; rare tags display a range of distances.

Figure 2. Rank Abundance Curves

OTU abundance is shown on a logarithmic scale on the vertical axis; OTUs (defined as described in Methods) are listed in rank order along the horizontal axis. The dashed horizontal line at abundance 5 corresponds to an estimated probability of detection of 99%, based on the appearance of a rare OTU in a tag or sequence library according to the Poisson distribution. A comparison of V6, V3, and full-length libraries is shown in (A); the abundance per subject for V6 and V3 refOTUs are shown in (B and C). The maximum V3 refOTU abundance and number of V3 refOTUs are higher for individual A than other participants (C) because three additional samples were analyzed for this subject.

Figure 3. Rarefaction Curves

The vertical axis shows the number of OTUs that would be expected to be found after sampling the number of tags or sequences shown on the horizontal axis. Curvature toward the horizontal indicates the increased sequencing effort required to observe novel OTUs when only rare OTUs remain to be discovered. Text boxes indicate the final observed OTU richness (O), range of estimated OTU richness by the ACE, ICE, Chao 1 and Chao 2 nonparametric estimators (E) [–63], and Good's coverage (C) [64] for each method. The nonparametric estimators indicate that total OTU richness is much higher than currently observed, but are known to underestimate the true richness when the number of observations (i.e., tags or sequences) is small relative to the size of the community [60,63]. Good's coverage, the proportion of tags or sequences found in OTUs containing at least one other member, is an estimate of the probability that a tag or sequence drawn at random from the pool of amplicons will belong to an OTU that has already been detected.

Figure 4. Relative Abundance of Taxa across Samples

The heatmaps show the relative abundance per sample of V3 refOTUs with total abundance of at least ten, after normalizing to an equal number of tags per sample (see Methods). Color intensity for all panels is directly proportional to the logarithm of normalized OTU abundance from 0 to 300; the 473 highest refOTU abundance values per sample, ranging from 300 to 19,964, are shown at maximum color intensity. Each column in the heatmaps represents one sample, arranged chronologically from left to right within individuals; blue bars above columns indicate Cp-associated samples. Each row represents a V3 refOTU, arranged alphabetically in a taxonomic hierarchy from phylum to genus, and by decreasing abundance within a genus. (A) All 1,450 V3 refOTUs with total normalized abundance of at least ten. The staggered bars on the left indicate phyla (Ac = Actinobacteria, Ba = Bacteroidetes, Fi = Firmicutes, Pr = Proteobacteria, Ve = Verrucomicrobia). Numbers on the right represent prominent genera or higher taxonomic ranks when refOTUs could not be classified to the genus level (see Methods for details): 1. Bifidobacteria, 2. Bacteroides, 3. Parabacteroides, 4. Alistipes, 5. Oscillospira, 6. Dialister, 7. Clostridium, 8. Dorea, 9. Faecalibacterium, 10. Subdoligranulum, 11. Clostridiaceae (family), 12. Eubacterium, 13. Anaerostipes, 14. Coprococcus and Lachnospira (2 thin bands at this position), 15. Roseburia, 16. Ruminococcus, 17. Lachnospiraceae (family), 18. Clostridiales (order), 19. Firmicutes (phylum), 20. Sutterella, 21. Akkermansia. (B–D) The normalized relative abundance in detail (same data as (A)) for V3 refOTUs in the three most abundant genera, Bacteroides, Faecalibacterium, and Roseburia.

Figure 5. Diversity Statistics

(A) Observed taxon richness (number of V3 refOTUs) per sample; Cp-associated samples have significantly fewer OTUs than pre- and post-Cp samples for individuals A and B (p < 0.005) but not individual C (p = 0.129). (B) Shannon diversity index; Cp-associated samples are significantly less diverse than other samples for all individuals (p < 0.001). (C) Shannon equitability index; OTU abundance in Cp-associated samples is significantly less evenly distributed than OTU abundance in other samples for all individuals (p < 0.001 for A and B, p < 0.05 for C). Formulas for diversity and evenness are given in Methods; significance is assessed as the probability that the Cp-associated value is drawn from the lower tail of a normal distribution with mean and variance as calculated from the other samples.

Figure 6. PCA of Relative Taxon Abundance

Principal component analysis (PCA) of log2-transformed normalized abundance for 1,450 V3 refOTUs with normalized total abundance of at least ten (same data as in Figure 4A). Cp-associated samples are shown with open symbols, other samples with filled symbols. (A) PCA axis 1 (accounting for 35.3% of intersample variation) separates the samples of individual B from those of individuals A and C; PCA axis 2 (18.2% of intersample variation) separates Cp-associated samples from others. The separation of the Cp-associated samples from others is greatest for individual A, intermediate for B, and least for C, consistent with the heatmaps (Figure 4) and diversity statistics (Figure 5). (B) PCA axis 3 (9.7% of intersample variation) captures temporal variability in non-Cp samples of individual A, and to a lesser extent, individual C. Three samples collected from individual A in the week prior to Cp treatment cluster together at one extreme of the range of axis 3 values. (C) PCA axis 2 versus PCA axis 3. (D) Scree plot showing the proportion of variance explained by the first 5 PCA axes; the dashed line at 0.059 indicates the amount of variance expected for each axis with random, uncorrelated data.

Figure 7. Abundant Taxa in the genus Bacteroides

The relative abundance per subject of Bacteroides V3 refOTUs, expressed as the percentage of all V3 tags from that subject. The 19 most abundant Bacteroides refOTUs range from the first to the 91st most abundant taxa overall. Eighteen of these 19 Bacteroides refOTUs differ significantly in abundance between subjects (p < 0.05, FDR = 1.9%, NS marks the exception), but the abundance of the Bacteroides genus as a whole does not (p = 0.075, FDR 15%). Numbers in the legend indicate the rank order normalized abundance of refOTUs within Bacteroides, and correspond to the designation of the OTU according to the convention of this report (i.e., “1” represents V3_Bacteroides_refOTU_1.) The dominant tag in 15 of the 19 most abundant Bacteroides refOTUs has a perfect match to reference sequences from the cultivated species named in the legend.

Figure 8. Individualized Cp Responses

Normalized abundance is shown on the vertical axis for selected V3 refOTUs (scales differ between panels); samples are ordered chronologically along the horizontal axis. Cp-associated samples are shown with open symbols. (A–H) Eight taxa among the 90 most abundant V3 refOTUs that show contrasting responses to Cp among individuals. Temporal variability is also evident at times not associated with Cp use.