Dynamic nature of viral and bacterial communities in human faeces

Summary Bacteriophages are a major component of the gut microbiome and are believed to play a role in establishment and stabilization of microbial communities by influencing taxonomic and functional diversity. We show that the activity of lytic and temperate phages can also significantly affect bacterial community structure in a model of extended colonic retention. Intact fresh human feces were incubated anaerobically at 37°C without homogenization and subjected to metagenomic sequencing. We observed subject-specific blooms and collapses of selected bacteriophage and bacterial populations within some individuals. Most notable were striking collapses of Prevotella populations accompanied by increases in specific bacteriophages. In a number of cases, we even observed a shift from one bacterial “enterotype” to another within 48 h. These results confirm that intact feces represents a highly dynamic ecological system and suggests that colonic retention time could have a profound effect on microbiome composition, including a significant impact by bacteriophages.


Highlights
Significant changes occur in human fecal bacteriome and virome during ex vivo incubation

Specific phage and bacterial dynamics are dependent on microbiome context
Collapse of Prevotella population is mirrored by an increase of Prevotella phages Incubation of fecal microbiome can result in changes comparable to a switch of enterotype

INTRODUCTION
Human gut microbiome composition is highly individualized, 1,2 which makes it difficult to define a ''healthy'' gut microbiome and complicates the analysis of compositional changes in microbiota in cohorts associated with different diseases. 3It has also been shown that microbiome individuality can underpin individual-specific responses to diet 4 and medications. 5Inter-individual microbiome variation represents a continuum of states, defined by gradients of concentration of the predominant taxa. 6The concept of ''enterotypes'', or a fixed number of poles (usually three, defined by the dominance of Bacteroides and Prevotella genera, or the Ruminococcaceae family 7 ) to which different intermediate states seem to gravitate, has been used to interpret gut microbiome data.The analysis of the human gut microbiome is further complicated by significant radial and longitudinal variation throughout the anatomy of the gastrointestinal tract 8 and reliance on fecal samples as faithful proxies for the distal gut microbiome.
][11][12] In particular, it has been demonstrated that looser stools with higher moisture content tend to display decreased microbial diversity.For Bacteroides and Ruminococcaceae-dominated enterotypes the Bacteroides:Ruminococcaceae ratio increases with increasing stool water content.It has been suggested that the relatively faster growth rates of Bacteroides can counter their washout at higher gut transit rates.At the same time, firmer stools contain increased quantities of Akkermansia, and the methane-producing archaeon Methanobrevibacter, as well as display a shift from carbohydrate fermentation to protein catabolism in microbial energy generation. 13Interestingly, the enterotype dominated by slow-growing Prevotella is associated with diets rich in plant fiber, 14 shorter transit time, and lower total microbial loads. 12This enterotype does not show any tendency for the enrichment of fast growers, suggesting that a different strategy is used by Prevotella (such as stronger adherence to mucus and epithelia) to avoid washout. 11This concept holds true when tested in vitro (in a three-stage continuous culture system), where increase in transit time resulted in decrease of biomass and diversity in the distal compartments. 15n addition to abiotic properties of the microbial habitat (such as availability and type of nutrients, peristaltic rate, etc.), the role of biotic factors must be taken into account.Bacteriophages (phages) are one of the prominent biotic forces operating in the gut microbiome with potential for exerting control over diversity and composition of the bacteriome.Advances in sequencing technology have paved the way for metagenomic studies that have helped to uncover the extraordinary diversity of largely uncultured and unclassified phage populations in the gut-the virome. 16Phages are as abundant as bacterial cells, reaching 10 11 virus-like particles (VLP) g À1 , 17 and one would predict that such a mass of bacteriophages would exert a strong top-down control on the density and diversity of the bacteriome. 18,19However, ll OPEN ACCESS the majority of these particles belong to viruses that either have been predicted to have a temperate lifestyle, 20 or have been shown to undergo lytic replication in restricted sub-populations of their hosts, both factors that would limit phage-induced mortality in bacterial hosts. 21,22evertheless, it is conceivable that even when microbiota shifts are driven largely by abiotic factors such as changes in colonic retention time (diarrhea or constipation), phages can act as immediate effectors responsible for bacterial mortality, opening the way for replacement of bacterial populations, and changes of overall diversity.The exact mechanisms responsible can range from prophage induction caused by bacterial SOS response, increased susceptibility to virulent phages due to reduced fitness or changes in cell surface structures, or potentially increased accessibility to bacterial prey due to changes in stool consistency.
It is known that feces, despite a high bacterial biomass content and likely near depletion of nutrients, remains a dynamic system with many bacteria, including strict anaerobes, remaining viable, 23,24 metabolically active, 25 and in a state of active DNA replication. 26This opens opportunities for dynamic observations of fecal microbiota, including phage-host interactions, using simple ex vivo models. 27In this study, we simulated extended colonic retention time by anaerobically incubating freshly collected human feces and registering changes in microbial and phage community composition over time.We started from a premise that this extended anaerobic incubation will likely change the composition of fecal microbial communities, in a manner similar to how increased retention time changes the composition of fecal microbiota in the human colon.We also hypothesized that phages (temperate and virulent) might have a direct role in this process, facilitating the demise of certain bacterial populations.Using this simple model we were able to draw parallels with previously observed correlations of microbiota composition and colonic residence time/stool consistency in humans. 11Most importantly, we observed correlations between collapse or overgrowth of certain bacterial populations and either collapse or bloom of phage populations.

Significant changes in human fecal bacteriome and virome during ex vivo incubation
Fresh fecal samples were collected from eight healthy adult volunteers (subject codes 916, 920, 922, 923, 926, 928, 941, 943).One gram fecal aliquots were held under anaerobic incubation at 37 C to simulate extended colonic retention time.Three aliquots were removed from each sample at each time point: 0 h, 6 h, 24 h, 48 h, and 120 h of incubation (Figure 1A).In order to assess changes in viral and bacterial diversity and taxonomic composition, shotgun sequencing of DNA extracted from the fecal VLP fraction, as well as sequencing of fecal 16S rRNA gene amplicon libraries, were performed. 28Fecal samples collected from six additional subjects were subsequently processed in a similar manner, though without triplicate sampling (Figure 1B).
The assembly of fecal viromes revealed a total of 12,828 viral genomic contigs across all 14 fecal samples, both partial (12,442) and nearly complete (386) that ranged in size from 1 kb to 196 kb (Figure S1A).Apart from a relatively small fraction of eukaryotic dsDNA and ssDNA viruses present in some of the samples, the DNA virome was dominated by phages belonging to the order Crassvirales, other orders and families of the class Caudoviricetes, as well as the order Petitvirales of small tailless phages (family Microviridae, Figure S1).Some of these phage contigs (including 180 out of 386 nearly complete genomes) could be linked to suspected bacterial hosts by means of sequence matches with CRISPR spacer database or sequence similarity to known phages and prophages present in bacterial genomes from RefSeq and HMP reference genome databases.The bacteriome was represented by 15 dominant families of phyla Bacillota, Bacteroidota, Actinobacteriota, and Pseudomonadota, once again in line with typical fecal bacterial composition (Figure 2A). 7,12ased on the results from the incubation of the initial eight fecal samples, we observed a tendency for an increase in viral contig a-diversity over time, for both Shannon and Simpson indices (Pearson r = 0.34 and 0.28 [weak correlation], p = 0.00024 and 0.0025, respectively, Figure S2).A slight tendency (statistically insignificant at this sample size) for a reduction of bacterial Shannon diversity was also observed.The bacterial community structure assessed at the individual operational taxonomic unit (OTU) level demonstrated significant evolution over time in all samples.Measures of b-diversity between samples, such as Bray-Curtis dissimilarity, showed a steady departure over time from the original viral and bacterial community composition (Pearson r = 0.65 for bacterial community [moderate correlation], p = 5.1e-06; r = 0.76 for viral community, p = 1.4e-08; Figure S3).With regard to bacterial composition, this process was more evident in some samples than in others (Figures 2A and S4).It was striking that samples showing the most drastic changes in bacteriome structure over time (individuals 916, 922) were characterized by an initial dominance followed by a precipitous reduction of Prevotellaceae populations during incubation (Figure 2A).This results in dramatic changes, with bacterial structures in some subjects coming to more closely resemble those of other subjects by the end of incubation, rather than their own original bacteriome composition (Figure S4A).While changes in virome composition appear to be equally significant, greater inter-individual variability of phage populations 17 outweighs virome shifts induced by incubation (Figure S4B).Nevertheless, in parallel with the extinction of the dominant Prevotella in samples from individuals 916 and 922, we observed a striking reduction of Petitvirales phages 31 and an increase in unclassified viruses, which results in completely altered bacteriomes and viromes by the end of the incubation (Figure 2B).In the sample from individual 928, the reduction of Prevotella populations was paralleled by a relative increase in Crassvirales phage 32 and a reduction of unclassified viral species (Figure 2B).These observations are consistent with phage proliferation in the fecal samples ex vivo and a resulting depletion of certain bacterial taxa.

Specific phage and bacterial dynamics are dependent on microbiome context
In order to provide a more detailed view of the dynamics of both the phage and bacterial populations and to identify drivers of microbiome change during incubation, analysis of Spearman rank correlation of fractional abundance of bacterial OTUs and viral genomic contigs against time of incubation was performed on a per subject basis.Each individual fecal sample was characterized by a specific pattern of bacterial OTUs and viral contigs which showed strong positive or negative correlation with the time of incubation (p < 0.05 with Benjamini-Hochberg correction; Figure 3A; Table S3).When correlated viral and bacterial species were grouped to bacterial family and viral order levels, a pattern emerges across individual samples, where the majority of families/orders demonstrate mixed behavior within and across individual fecal samples, with some being either consistently negatively correlated with time (bacterial families Prevotellaceae, Rikenellaceae, phage order Petitvirales) or positively correlated (most of the Crassvirales and other class Caudoviricetes phages, bacterial families Acidominococcaceae, Actinomycetaceae, Coriobacteriaceae, Lactobacillaceae, most of the Bifidobacteriaceae and Peptostreptococcaceae, Figure 3A).We then focused our analysis on a subset of these time-correlated taxa comprised of possible phage-host pairs (inferred at bacterial genus level) present in the same fecal sample.We observed cases when the normalized relative abundance of bacterial OTUs declined in parallel with increase in relative abundance of the corresponding phage (e.g., Bacteroides in individuals 922, 926, and 943; Coprococcus and Odoribacter in subject 926), and vice versa (e.g., Bacteroides in individuals 923 and 941; Blautia in subject 928; Figure 3B).Such variable behavior of bacterial strains belonging to the same genera in different contexts can be explained, among other factors, by differences in subject-specific virome composition (both temperate and virulent phages) and differential sensitivity to virulent phage infection or prophage induction in subject-specific bacterial strains.
Taken together, our observations of dramatic changes in both bacterial and viral composition over time in at least some fecal samples incubated anaerobically (Figures 2 and S4), an increase in viral a-diversity (Figure S2A), correlation of specific bacteriome and virome members with time of incubation, and opposite trends in relative abundance of bacteria and their predicted phage, are all consistent with a direct role for phages in bacterial mortality and the concomitant compositional changes of fecal bacteriome.More specifically, observed changes in bacteriome composition can be underpinned by collapses of some bacterial populations caused by virulent phage predation, induction of prophages caused by starvation and other stresses, and overgrowth of other populations, unaffected by phage/prophage attacks.However, other causative factors, such as differential response of microbiota to the inevitable aerobic exposure of fecal samples between sample collection and the beginning of anaerobic incubation, as well as incubation itself, cannot be ruled out.

Collapse of Prevotella population is mirrored by an increase of Prevotella phages
Due to the compositional nature of metagenomic data, it is not obvious how the observed changes in relative abundance would translate into changes in absolute abundance of the corresponding bacterial and viral taxa.To clarify that, we proceeded to collect six additional samples from healthy subjects (including one donor from the original eight) to validate our original observations using plate counts, qPCR, and metagenomics.Collected samples were incubated anaerobically in a similar manner, but without triplicates, for either 0-6-24-48-96 h (subjects 925, 927, and 931) or 0-24-72-144 h (subjects 916-1, 921, and 924; Figure 1B).
This new set of fecal samples displayed dynamics similar to our earlier observations, with dramatic collapses of Prevotella populations and changes of virome composition in Prevotella-rich samples (Figures 4A and 4B).To further investigate this phenomenon we focused our attention on a group of complete circular viral genomic contigs with proven links to the Prevotella genus.These included a 95.6 kb virulent crAsslike phage genome present in sample 931 (exact 34 nt match to a CRISPR spacer GGGTCAGTAGCCATAAGAGTAAATGCAACATCATCAG in Prevotella copri strain Indica; Figure 4C), and a 6.3 kb virulent Microviridae (order Petitvirales) genome present in sample 916-1 (94% and 97% identity matches with CRISPR spacers GACGTATCAGCAGGAGCAGTTTGTTCAGGGGTT and TAATGGAACTATTCTTTATCCTCAATGGG ATGAT in Prevotella copri strain Indica).Relative quantification of these contigs using qPCR in total community DNA as well as in the VLPassociated fraction of fecal DNA revealed discordant dynamics (Figure 4D).While Prevotella levels plummeted in both samples 916-1 and 931, a 95.6 kb crAss-like phage genome, demonstrated a rapid increase that was especially pronounced in the VLP DNA fraction.The 6.3 kb Microviridae phage genome sharply increased in the total DNA fraction but unexpectedly showed a slight decrease in the VLP fraction.These results further support the notion that drastic microbiota changes seen in Prevotella-rich samples 916-1 and 931 can be attributed to phage activity as one of the contributing factors.

Incubation of fecal microbiomes can result in changes comparable to a switch of enterotypes
The scale of changes in bacterial communities during incubation prompted us to fit these changes into the context of the ''enterotype'' concept.To do that, we combined the 16S rRNA sequencing data generated here with data from a year-long longitudinal observation (monthly sampling) of fecal bacteriomes in ten individuals, 17 as well as longitudinally collected samples (three during a year for each subject, n = 38) from a healthy control cohort of a comparative IBD microbiome/virome study 33 (Figure 5A).Since all samples were processed in the same lab using the same protocol, and some donor subjects were enrolled in all three studies, we deemed these three datasets directly comparable.K-means clustering of Bray-Curtis dissimilarities between all samples readily separated them into three partially overlapping clusters, or ''enterotypes'' 1-3 (Figure 5B).A number of incubated fecal samples (916, 922, and 931), as well as longitudinally sampled subjects 916 and 922, migrated from one enterotype to another during the course of the studies (Figure 5A).We then used Simpson diversity index to measure variance of the enterotype label over time within each subject (or each incubated fecal sample).Confirming our observations of instability of Prevotella-dominated samples, subjects from all three studies predominantly assigned to enterotype 3 (family Prevotellaceae-enriched in our designation) showed significantly greater instability (switching between enterotypes, Mann-Whitney test p values 0.024-0.047,Figures 5C and   A B 5D), compared to subjects predominantly assigned to enterotypes 1-2 (enriched in members of the family Bacteroidaceae, and also either Ruminococcaceae or Lachnospiraceae, Figures 5E-5G).

DISCUSSION
In this study, we utilized a simple model of anaerobic incubation of unprocessed fecal material at 37 C as a proxy for extended colonic retention time.We observed a steady departure from the original composition of both bacteriome and virome, the magnitude of which was dependent on the original microbiome composition.Fecal samples dominated by Prevotella (enterotype 3 in this study and enterotype II according to Arumugam et al. 7 ) were especially prone to change under these conditions, showing a dramatic reduction of Prevotella read counts over time.
Our in vitro results complement previous human studies that had demonstrated colonic retention/transit time and linked stool consistency as strong predictors of microbiota composition, with looser stools associated with the Prevotella enterotype and firmer stools associated with the Bacteroides/Ruminococcaceae enterotypes. 11,15Importantly, the absolute abundance of Prevotella was reported as relatively well conserved across populations, making high relative abundance of Prevotella a feature typically found in samples with low total microbial loads. 12In addition to differences seen in large population cohorts, previous experiment with exposure of a single healthy human fecal microbiota sample to an in vitro simulated longer colonic transit time was sufficient to recapitulate some of the microbiota changes seen in elderly and slow transit time patients. 15The precise mechanisms responsible for these changes were not fully understood at the time, but our results suggest that the action of bacteriophages, either via prophage induction or attacks of virulent (lytic) phages, may have a significant contribution to depletion of some and overgrowth of other bacterial populations.Many of the changes in bacterial communities seen in this study as a result of anaerobic fecal incubation, although highly divergent and subject-specific, occurred in parallel with changes in the phage communities.An example of that was the increase in Crassvirales and decrease in Petitvirales (family Microviridae) phages linked to Prevotella.The Crassvirales comprise a diverse and largely uncultured group of bacteriophages that were previously linked, through a variety of approaches, to bacterial hosts in the phylum Bacteroidota, mainly genus Bacteroides (evidence ranging from isolation in culture, single-cell bacterial genomics, and bioinformatics approaches 21,34,35 ).No Crassvirales phages targeting Prevotella have been isolated in culture so far; however, an enrichment of a related phage in the presence of Prevotella in vitro has been reported. 36An increase in overall virome diversity and a tendency toward a decrease in bacterial diversity (insignificant at this sample size) were general features of all samples.
Our findings of rapid parallel changes in bacterial and phage communities in the incubated feces may indicate that bacteria exposed to prolonged residence in a purportedly nutrient-depleted environment might undergo a shift in energy state sufficient to launch prophage induction or to significantly increase sensitivity to virulent phages.These events can potentially cause a dramatic change in microbiota composition, including a shift from enterotype II (Prevotella) toward enterotype III (Bacillota) in some samples.Using mice colonized with simple artificial communities, it has been demonstrated that waves of spontaneous prophage induction can lead to dramatic bacterial mortality and change in community structure. 37Such events can indeed be triggered by global stress and bacterial SOS response, resulting from increase in transit time, nutrient starvation, and microaeration. 38,39mitations of this study Include: (i) The time gap between voiding and processing of fecal samples and their aerobic exposure.This could have resulted in microbiota composition-dependent oxidative stress and increased mortality in certain taxonomic groups.At the same time, previous studies have shown that raw fecal samples retain good bacterial viability when stored at ambient conditions for up to 24 h. 25 (ii) The use of 16S rRNA amplicon profiling, as opposed to shotgun metagenomics, to study the composition of bacterial microbiota.This limited our taxonomic resolution to genus level and prevented us from establishing phage-host connections in a more direct fashion (CRISPR spacer matches, prophages in metagenomically assembled genomes etc).(iii) Relatively low number of samples used, which, due to heterogeneity of microbiome compositions could have hindered some common trends in phage-host dynamics that would be more visible in larger donor populations.(iv) The simplified nature of the model and its inability to take into account such covariates of increased colonic residence as nutrient fluxes (including as a result of phage-induced lysis, as well as between the human host and bacteria), changes in gas composition, pH, water absorption etc.
Nevertheless, our data suggests that integrating the analysis of both bacterial and phage communities would provide a more complete understanding of the composition of the bacteriome in feces or in the distal gut, particularly following long colonic residence times.The absence of certain taxa, such as Prevotella, could be an accurate reflection of the community structure or could be a result of phage-induced lysis.If the latter was the case, analyzing the bacteriome in isolation would provide a misleading picture of the true bacterial composition.However, the presence of high levels of Prevotella-specific phage would confirm that Prevotella had been present in significant amounts since every phage is a biomarker of the recent presence of the host bacterium.In order to extend the usefulness of this observation we need to establish more phage-host relationships within the gut microbiome, an effort that would certainly provide better insights into bacterial structure in the microbiome.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

Figure 1 .
Figure 1.Overview of the experimental design Human fecal samples were collected and subjected to anaerobic incubation at 37 C.(A) Fecal samples from eight subjects were divided into three 1 g aliquots and incubated for 0-6-24-48-120 h, followed by shotgun metaviromics and 16S rRNA amplicon sequencing.(B) Samples from a second group of six subjects were processed similarly but without triplicates and subjected to plating and qPCR assays in addition to sequencing.

Figure 2 .
Figure 2. Changes in bacteriome and virome composition during anaerobic incubation of eight human fecal samples (A) Relative abundance of bacterial families reaching at least 5% of read counts in any of the samples.(B) Relative abundance of viral contigs (black outlines) reaching at least 0.1% of read counts in any of the samples.Bars are colored by viral order.All samples were taken as biological triplicates (duplicates in some cases due to failed DNA extraction/sequencing).

Figure 3 .
Figure 3. Bacterial OTUs and phage genomic contigs follow different patterns of correlation of their relative abundance with time of incubation (A) Bacterial OTUs (grouped by family) and viral contigs (grouped by viral order) showing positive or negative Spearman rank correlation (p < 0.05) with time within individual fecal samples.(B) Normalized (Z score) relative abundance of predicted phage-host pairs (viral contigs linked to bacterial host OTUs via CRISPR spacer matches and inference from closely related phage genomes in IMG/VR3 database.See STAR Methods) showing positive or negative correlation with time of incubation (filled circles are median values, vertical bars show range between minimum and maximum value across three biological replicates).

Figure 4 .Figure 5 .
Figure 4. Phages linked to Prevotella expand as populations of their hosts collapse in six additionally collected fecal samples (A) Relative abundance of bacterial families reaching at least 5% of read counts in any of the samples.(B) Relative abundance of viral contigs (black outlines) reaching at least 0.1% of read counts in any of the samples.Bars are colored by viral order.(C) Circular maps of phage genomic contigs Node_1 (a Crassvirales order phage), Node_83 (a Petitvirales order phage).(D) qPCR relative quantification of complete circular phage genomic contigs Node_1, Node_83 and genus Prevotella 16S rRNA gene in two fecal samples showing rapid decrease of Prevotella populations (filled circles are median values, vertical bars show range between minimum and maximum value across three technical replicates).

TABLE d
RESOURCE AVAILABILITY B Lead contact B Materials availability B Data and code availability d EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS B Collection and incubation of fecal samples d METHOD DETAILS B Extraction and shotgun sequencing of VLP-associated DNA B Bacterial microbiota profiling B Bioinformatic analysis of VLP metagenomic data B Bacterial cultures B qPCR d QUANTIFICATION AND STATISTICAL ANALYSIS SUPPLEMENTAL INFORMATION Supplemental information can be found online at https://doi.org/10.1016/j.isci.2023.108778.