Modulation of Post-Antibiotic Bacterial Community Reassembly and Host Response by Candida albicans

The introduction of Candida albicans into cefoperazone-treated mice results in changes in bacterial community reassembly. Our objective was to use high-throughput sequencing to characterize at much greater depth the specific changes in the bacterial microbiome. The colonization of C. albicans significantly altered bacterial community reassembly that was evident at multiple taxonomic levels of resolution. There were marked changes in the levels of Bacteriodetes and Lactobacillaceae. Lachnospiraceae and Ruminococcaceae, the two most abundant bacterial families, did not change in relative proportions after antibiotics, but there were marked genera-level shifts within these two bacterial families. The microbiome shifts occurred in the absence of overt intestinal inflammation. Overall, these experiments demonstrate that the introduction of a single new microbe in numerically inferior numbers into the bacterial microbiome during a broad community disturbance has the potential to significantly alter the subsequent reassembly of the bacterial community as it recovers from that disturbance.

the specific changes in the bacterial community via 454 pyrosequencing of 16S rRNA gene amplicon libraries, similar to our previous studies analyzing intestinal conventionalization dynamics in mice 30 . Since intestinal inflammation can alter the bacterial microbiota, we also sought to determine whether intestinal inflammation was playing a significant role in this process.

Results
Disturbing the indigenous bacterial community by a broad spectrum antibiotic breaks colonization resistance and allows longterm colonization of the ileum and cecum by C. albicans CHN1. C57BL/6 mice were treated with cefoperazone in their drinking water for one week (days 27 to 0), then switched to sterile drinking water and inoculated with C. albicans CHN1 by oral gavage at day 0. The groups of mice were subsequently analyzed for bacterial microbiome changes and C. albicans colonization at days 7 and 21 post-antibiotic treatment (Figs. 1A & 1B). Cefoperazone reduced total bacterial numbers in the intestine by 10,000 fold during the week the mice were drinking the antibiotic water (determined by both culture and total 16S rRNA gene qPCR, data not shown). Upon cessation of the antibiotic, total bacterial numbers returned to pre-antibiotic levels by 72 hours (approx. 10 10 culturable bacteria per g cecal tissue). Neither untreated mice nor those whose microbiome had been disturbed by cefoperazone had detectable C. albicans in their ileum or cecum at days 7 and 21 (Figs. 1A & 1B). When mice were gavaged with C. albicans in the absence of microbiome disturbance, less than half of the mice had detectable C. albicans levels at day 7 and almost none at day 21. In sharp contrast, all of the gavaged mice with a disturbed microbiome had detectable C. albicans colonization at day 7 and the colonization levels did not significantly change through day 21. We have followed mice out through 90 days, and C. albicans levels in C. albicans-gavaged mice remain similar to those seen at day 21 (data not shown). Thus, the indigenous bacterial microbiota exerts strong competitive exclusion against invasion by the yeast C. albicans; however, disturbance of the community by cefoperazone breaks colonization resistance and allows C. albicans to colonize the intestinal microbiome at levels far inferior to those of the bacteria (.10 10 bacteria vs. 10 4 yeast).
After disturbance of the GI microbiota, colonization by C. albicans CHN1 at low numbers can alter the reassembly of the bacterial community. To assess the effects of C. albicans CHN1 colonization on the reassembly of the intestinal bacterial community after broad disturbance, we used 454 pyrosequencing of V3-V5 16S rRNA gene amplicon libraries generated from total cecal mucosal DNA at days 7 and 21 post-disturbance. Sequence data were binned into operational taxonomic units (OTU), based on 97% similarity, as described in the methods.
Two metrics of change in bacterial diversity were calculated for each of the treatment groups at day 7 and day 21: inverse-Simpson values and rarefaction curves (Figs. 1C & 1D). The inoculation of C. albicans into an undisturbed community did not significantly affect diversity. However, there was a significant decrease in the diversity of the bacterial community of the cecum at day 7 post-disturbance compared to untreated mice. The inoculation of C. albicans into a disturbed community further decreased community diversity at day 7 post-antibiotic, although the additional loss of diversity did not score as statistically significant. By day 21, the diversity of both the groups that had suffered a microbiota disturbance had largely recovered to that of the untreated groups, although the diversity of the group that was disturbed and gavaged with C. albicans still trended lower.
Next, the bacterial community structure of the experimental groups was assessed using canonical correspondence analysis (CCA). This is a method of direct gradient analysis commonly used in ecology 37,38 where the community data matrix is subjected to a weighted linear regression on the constraining variables, and a correspondence analysis performed on the fitted values. This both focuses the data on the question of interest and creates a testable model for assessment of significant effects. In our study, the variation was constrained to treatment group and time (Fig. 1E). The communities in the untreated (undisturbed) mice shifted very little over the course of the experiments (day 7 vs. day 21). The addition of C. albicans to the undisturbed state had very little effect on the bacterial communities at day 7 or day 21. In contrast, disturbance of the bacterial community by cefoperazone treatment resulted in a significant shift (p , 0.005) away from the untreated state, even seven days after the cessation of antibiotics. The introduction of C. albicans into this disturbed bacterial community resulted in a community structure that was not only significantly different from that of untreated mice, but was also significantly different from mice treated with antibiotics alone. Between days 7 and 21 post-antibiotic, the bacterial intestinal communities in both of the disturbed groups changed to become more similar to that in the undisturbed group (i.e., reassembly). However, the bacterial microbiota of mice treated with antibiotics (disturbed) and of antibiotic-disturbed with C. albicans gavage remained significantly different from each other.
Colonization of the ileum and cecum by C. albicans elicits a host response but does not induce intestinal inflammation. Mice were examined at days 7 and 21 post-antibiotic for intestinal inflammation (histology) and at day 21 for changes in intestinal and mesenteric lymph node gene expression by high-throughput qPCR. When C. albicans was introduced into an undisturbed microbiota (Fig. 2B), very few genes showed any significant change in expression, while disturbance of the microbiota caused a number of genes to be significantly altered in expression (Fig. 2C). The addition of C. albicans to a disturbed microbiota (Fig. 2D) resulted in changes in gene expression not seen in the other groups. The effects of these treatments on the expression of immune modulatory molecules in the mesenteric lymph nodes followed a similar trend ( Fig. 2A). However, despite the active differential regulation of genes, there was no evidence of inflammation in histological sections of the small intestine (defined by absence of leukocyte infiltrates, edema or changes in epithelium structure) and no histological evidence of hyphal invasion (Fig. 3). We have reported previously that C. albicans colonization does not induce cecal inflammation 25,29 .
Introduction of C. albicans into a disturbed bacterial community can alter the diversity and membership of the community. We next characterized the bacterial lineages that shifted during disturbance and recovery, by using RDP-based taxonomic assignment of the bacterial OTUs (97% identity), as described in the methods. At day 7, significant phylum-level shifts were evident in both the antibioticdisturbed community and the community of antibiotic-disturbed with C. albicans gavage (Fig. 4A). There were significant decreases in both the Bacteroidetes and the Synergistetes. The dominant phylum was the Firmicutes, which remained unchanged no matter the treatment. Analysis at a deeper taxonomic level, revealed that there were a number of marked changes in bacterial community membership at the family-level (Figs. 4B-4E). Both the Rikenellaceae and the Porphyromonadaceae, among the top 5 most abundant families in the undisturbed treatment group, were all but non-existent in either of the disturbed treatment groups. In contrast, bacteria from less abundant families (in untreated mice) bloomed in the mice with the disturbed microbiota.
Furthermore, there were notable changes at day 7 within the Lachnospiraceae (Fig. 6A) and Ruminococcaceae (Fig. 6B). These two bacterial families were (1) the most abundant in all the experimental groups and (2) did not appear to change in relative proportions after antibiotics. At the genus-level, the disturbance of the microbiota resulted in a marked shift in the dominant genera within www.nature.com/scientificreports SCIENTIFIC REPORTS | 3 : 2191 | DOI: 10.1038/srep02191 Figure 1 | The effect of C. albicans colonization on bacterial community structure. C57BL/6 mice were treated with cefoperazone in their drinking water for one week (days -7 to 0), switched to sterile drinking water and then inoculated with C. albicans CHN1 by oral gavage at day 0. The groups of mice were then analyzed for bacterial microbiome changes and C. albicans colonization at days 7 and 21 (n 5 8-12 mice per treatment per time point). Panels A and B demonstrate the growth of C. albicans from the ileum and cecum at day 7 and 21 (L.O.D. of 10 1.7 ). The inverse-Simpson diversity index was utilized to quantify the diversity of the cecal communities at day 7 and day 21 (panel C). Significance was tested using 1-way ANOVA (p , 0.05). Rarefaction analysis was also employed to view what idealized collector's curve based on the data would look like at day 7 and day 21 (panel D). The mean curve of 8-10 rarefaction curves is displayed as a bold line and the dotted lines represent the standard error. Panel E is a canonical correspondence analysis (CCA) plot of both day 7 and 21 normalized microbial composition data where the data has been constrained by treatment and time. To better visualize the group clusters, each data point is connected to its group centroid.
www.nature.com/scientificreports SCIENTIFIC REPORTS | 3 : 2191 | DOI: 10.1038/srep02191 these two familes (Fig. 6). In undisturbed communities, Lachnobacterium, Marvinbryantia and Syntrophococcus constituted approximately 80% of the Lacnospiraceae. However, after antibiotic treatment, they accounted for ,50% of the population at day 7. There were marked blooms of the rare Lachnospiraceae genera Coprococcus, Anaerosporobacter, Dorea and Robinsoniella in antibiotic-only mice. In contrast, C. albicans colonization of the disturbed community was associated with the rapid recovery of Lachnobacterium and Syntrophococcus to undisturbed levels, outgrowth of Roseburia, and Robinsoniella, and reduction in Coprococcus and Dorea levels (Fig. 6). In the Ruminococcaceae, disturbance had very little effect on the dominant genera within the community; however, the presence of C. albicans in the disturbed community resulted in almost a complete elimination of Oscillibacter from the community.
A similar analysis was next carried out at the phylum, family and genus-level at day 21 post-disturbance. The bacterial community structure of the antibiotic-treated mice still had not recovered to a pre-antibiotic state (Fig. 5). Most notably, the levels of the Bacteroidetes phylum was 100 fold lower level than in undisturbed levels (Fig. 5A). However, the presence of low levels of C. albicans within the disturbed community was associated with higher Bacteroidetes levels than in antibiotic only treated mice, similar to untreated mice (Fig. 5A). The blooms in low abundance bacteria that were seen at day 7 were no longer evident at day 21. Relative levels of six of the eleven most dominant families were still markedly depressed in antibiotic-only mice at day 21. In contrast, the presence of C. albicans during the recovery process was associated with ten of the eleven most dominant bacterial families being found at relative levels similar to that found in untreated (undisturbed) mice. The one notable  Table S1. Fold changes are depicted relative to the undisturbed state. Panels B-D depict genes in which a statistically significant difference occurred are displayed for each of the treatment groups (B) C. albicans alone, (C) disturbed microbiota, or (D) disturbed microbiota colonized with C. albicans. Significant differences in gene expression were determined using SAM (p , 0.05, false discovery rate set at zero (q 5 0.00)) are displayed for each of the treatment groups) as described in Methods.
www.nature.com/scientificreports SCIENTIFIC REPORTS | 3 : 2191 | DOI: 10.1038/srep02191 exception was the very low levels of the Lactobacillaceae (denoted by arrows, Figs. 5D & 5E) in both the groups that had undergone an antibiotic-disturbance. In contrast to day 7, there were no longer genus-level differences in the Lachnospiraceae and Ruminococcaceae at day 21 (Fig. 7), largely indicating that the Lachnospiraceae populations had finally recovered fully from cefoperazone treatment.

Discussion
These experiments demonstrate that the introduction of a single new microbe (the yeast C. albicans CHN1) in numerically inferior numbers into the bacterial microbiome during a broad community disturbance can significantly alter the subsequent reassembly of the bacterial community as it recovers from that disturbance. This occurs in the absence of overt intestinal inflammation (leukocyte infiltrates, edema or changes in epithelium structure) but is associated with changes in host gene expression in the intestinal epithelium. We have previously demonstrated that the recovery of the microbiota in the presence of C. albicans leads to a state of systemic hyperresponsiveness 24,25 . The effects of the antibiotics and yeast were localized to the GI tract because (1) cefoperazone is poorly absorbed when delivered in the drinking water, (2) we have been able to demonstrate that oral treatment with cefoperazone is unable to clear a cefoperazonesensitive bacteria from the lungs of mice (data not shown), and (3) yeast are not introduced into the lungs following oral gavage 24,25 . We propose that our current studies provide additional insight into potential pathways by which colonization with a numerically inferior organism after antibiotics can have a disproportionate effect on the immune system without inducing intestinal inflammation: by affecting the reassembly of the bacterial microbiome.
Our study demonstrates that C. albicans can colonize the mouse cecum and ileum without causing overt intestinal inflammation but the presence of C. albicans is associated with changes in bacterial community reassembly. Although C. albicans, the introduced microbe, is not a normal member of the mouse microbiota, it is known to first colonize humans within a few weeks after birth 39 . In mice, C. albicans is excluded by the normal microbiota; however, colonization can take place if the GI microbiota is disturbed.
Interestingly, while C. albicans causes a mild gastritis in the stomach, the fungi colonizes other parts of the lower GI tract with no evidence of local inflammation 31 . A diverse and abundant fungal microbiome, as detected by molecular techniques has been reported previously, using mice from both a specific pathogen-free (SPF) source and from a colony with a restricted bacterial flora (RF mice) 40 . In a recently published study with Dectin-1 knockout and wildtype mice, high levels of indigenous fungi were reported in wildtype laboratory mice, especially C. tropicalis and accounted for 65% of the fungi detected by ITS1-2 amplicon pyrosequencing, although no culture data was presented for this yeast that can be readily cultured on Sabouraud Dextrose Agar and identified on Chromagar 41 . In our laboratory, as we have reported here, we have observed that SPF mice routinely lack culturable fungi. Treatment with a broad spectrum antibiotic, such as cefoperazone, can result in the culture of fungi from intestinal samples on SDA, such as C. tropicalis 25,29,31 , but after cessation of cefoperazone, these yeast are rarely detected in the mice. The systemic hyperresponsiveness of Candida-colonized mice to challenge with an aeroallergen 24,25 indicates that intestinal C. albicans colonization can alter the immune response of the host. The recent report demonstrating a role for Candida species in driving colitis in Dectin-1 knockout and wildtype mice 41 further underscores the potential of the fungal microbiome to alter host immunity and inflammation.
The current work also highlights that members of the Porphyromonadaceae, Rikenellaceae, Lachnospiraceae and Lactobacillaceae bacterial families are some of the most susceptible to perturbations in reassembly following cefoperazone treatment (1/2 C. albicans). While it is unknown how the mouse microbiota prevents C. albicans colonization, previous studies have demonstrated the ability of C. albicans to compete with bacteria (particularly amongst members of the Firmicutes) 42 . Interactions between C. albicans and other phyla, such as the Bacteroidetes, are not well investigated, although it has been suggested that Lactobacillus species are critical 15,43 . To induce a disturbance in the microbial community and enable C. albicans colonization, cefoperazone was used. This is a poorly absorbed thirdgeneration cephalosporin with excellent activity against anaerobic bacteria 44 . It can facilitate gastric colonization by C. albicans 31 .  Cefoperazone has also been shown to be able to disturb the mouse cecal microbiota for greater than six weeks after washout if the mice are housed in isolation 1 . This delay in reassembly is in contrast to the observation that a cecal microbiota community can completely reassemble within seven days following transfer from a conventional to a germ-free mouse, despite a complete destruction of community structure during the first few days after transfer 30 . In addition, cefoperazone-treated mice will completely and quickly reassemble their cecal microbiota if allowed to recover in the presence of an untreated mouse 1 . While currently of unknown significance, the association between C. albicans colonization and altered Lachnospiraceae composition is intriguing in light of the report that Lachnospiraceae communities can be important in colonization resistance against enteric bacterial pathogens 45 . We have observed that the intestinal microbiota shift caused by the introduction of C. albicans into cefoperazone-treated mice is not associated with mucosal inflammation, but it is also not a neutral event for the host. The pattern of genes down-regulated in mice with a disturbed microbiota and C. albicans colonization is consistent with an environment where the host is less responsive to microbial signals. Down-regulated genes of interest included Ffar2 (Gpr43) 46-48 , Nod1 49,50 , Gpr35 51,52 , and the TLRs 26,53 . These are all well-characterized as environmental sensors that play important roles in inflammation. At the same time, there is a decrease in the expression of various anti-microbial peptides and tight-junction proteins that might render the host less capable of maintaining microbial homeostasis [54][55][56] . abundance graphs constructed for the phylum and family-level taxonomies. Panel A is the phylum-level classification for the day 21 undisturbed (red), undisturbed plus C. albicans (orange), disturbed (green) or disturbed plus C. albicans (blue) groups. Panels B-E are the family-level taxonomic distributions. X-axis ordering of the rank abundance graphs is based on the relative distribution in the undisturbed community. The arrows in panels D and E highlight the Lactobacillaceae, which remain significantly reduced at day 21 following cefoperazone treatment. Data are mean 6 SEM; n 5 8-10 mice per group from 3 separate experiments. Thus, it is possible that the communication between the host and the GI microbiota has been disrupted but not to the breaking point to create intestinal inflammation. It also remains possible that, despite a lack of intestinal inflammation, the mild inflammation in the stomach may be influencing gene expression in the cecum, although it is unclear how this signaling would take place. Another possibility is that formation of discrete foci of yeast invasion and inflammation is occurring in the intestine but is undetectable at the level of resolution used in this study. Finally, active suppression of host responsiveness by C. albicans or other members of the microbiota remains a possibility because it has been previously demonstrated that C. albicans can produce prostaglandins and related oxylipins [57][58][59] . In summary,  we have provided a proof-of-concept study that low-level colonization and/or blooms of yeast within the microbiome following antibiotic therapy can potentially have unintended consequences on bacterial microbiota reassembly and host mucosal immunity.

Methods
Animals and housing. Female C57BL/6 mice (Jackson Laboratories, Indianapolis, IN) were housed 5 mice to a cage under specific-pathogen-free conditions in enclosed filter-top cages. The experiments (at days 7 and 21 post-treatment) were performed on 3 separate occasions. Food and sterile water were given ad libitum. Food remained constant throughout the experiment to minimize the effect of diet on the microbiota. Mice were maintained on grates by the Unit for Laboratory Animal Medicine (ULAM) at the University of Michigan (Ann Arbor, MI) to prevent coprophagy. All protocols were approved by the University of Michigan University Committee on Use and Care of Animals (UCUCA).
Antibiotic treatment. Cefoperazone (0.5 mg/ml) (Sigma-Aldrich, St. Louis, MO) was administered orally to mice ad libitum in drinking water. Antibiotic treatment was continued for 7 days (days 27 to day 0) prior to C. albicans colonization. After 7 days, antibiotic containing drinking water was replaced with sterile water.
C. albicans GI inoculation. C. albicans strain CHN1 (a human clinical isolate) was grown as previously described 31 , diluted to 2 3 10 8 CFU/ml in sterile nonpyrogenic saline, and 10 7 CFU in 50 ml gavaged using a 24-gauge feeding needle attached to a 1ml syringe mounted on a Stepper repetitive pipette (Tridak, Brookfield, CT) to deliver an equivalent amount of inoculum to each mouse. The inoculums were plated on Sabouraud dextrose agar (SDA) to verify the number of colony forming units (CFU) delivered.
Tissue isolation. Following euthanasia, the GI tract was sterilely removed and the cecum separated from the colon and small intestine. A 1 cm 2 tissue strip from the cecal tip was removed, rinsed in sterile 13 PBS and divided for DNA and RNA isolation. The remaining portion of the cecum was cut open along the greater curvature and rinsed in sterile 13 PBS. Thin slices along the outer portion were removed and fixed overnight in 10% Neutral Buffered Formalin for histological analysis. The remaining tissue was homogenized in sterile water using a Tissuemiser (Fisher Scientific). Aliquots of the homogenates were plated in 10-fold serial dilutions on Sabouraud dextrose agar (Difco). For the small intestine, the distal half of the organ was separated, cut open longitudinally and rinsed in sterile 13 PBS. A 1 cm 2 tissue strip was taken from the middle and divided for DNA and RNA isolation. The distal portion of the remaining tissue was prepared for histological analysis using the "Swiss-roll" technique whereby the tissue is rolled and fixed overnight in 10% Neutral Buffered Formalin, which more readily allows visualization of the entire length of the intestinal sample. The proximal portion of the remaining tissue was homogenized in sterile water using a Tissuemiser (Fisher Scientific). Aliquots of the homogenates were plated in 10-fold serial dilutions on Sabouraud dextrose agar (Difco) to determine levels of C. albicans colonization.
DNA extraction. Genomic DNA was extracted from tissue homogenized in 360 ml ATL buffer (Qiagen DNeasy Blood & Tissue kit) using a modified protocol previously demonstrated to isolate bacterial DNA 31 .
454-prosequencing. The V3-V5 region of the bacterial 16S rRNA gene was targeted using barcoded primer sets corresponding to 357F and 929R. These primers sets were originally developed by the Broad Institute. The PCR conditions were 95uC for 2 min., followed by 30 cycles of (95uC 20 sec, 50uC 30 sec, 72uC 5 min), holding at 4uC. Quality control and sequencing was carried out at the University of Michigan, using the Roche 454 GS-FLX Titanium platform.
16S rRNA gene data analysis. Sequence data were processed and analyzed using the software mothur v. 1.21 32 according to the protocol provided (http:// www.mothur.org). The operational taxonomic units (OTUs) were binned at 97% similarity (to create shared community file and the phylotyped, genus-level grouping, file). For the purposes of this study, we focused on OTUs that were present at greater than 0.5% of the sample population. These files, along with the file containing the taxonomic information for the OTUs, were imported and further analyzed in R. The R-package vegan 33 was used for diversity analyses and ordinations, as well as a custom R script for sorting classification results into tables (ClassifierSorter v.2). Classification of OTUs was carried out using the implementation of the RDP Classifier and RDP taxonomy training set 7 34,35 contained in mothur. The number of sequences in each sample group can be found in Supplemental Table S2.
Quantification of mRNA levels using high-throughput quantitative PCR. RNA was isolated from each tissue using TRIzol (Ambion) and the RNeasy Mini Kit (Qiagen). RNA concentration was determined using a Nanodrop instrument (Thermo Scientific). Complementary DNA (cDNA) was generated using the RT 2 First Strand Kit (Qiagen). Expression levels of genes under study were determined using Mouse RT 2 Profiler PCR assays, custom-made to contain replicate sets of 48 primer pairs (Qiagen). Each well of the replicate sets was loaded with 2.5 ng of cDNA reaction product. Each card was run on a LightCycler 480 Real-Time PCR system (Roche). The relative RNA expression levels were inferred from the C t values. The geometric mean of ß-actin and GAPDH levels was used as the internal control for the generation of DCt values. The heatmap was generated in R using the package gplots 36 .
Statistics. A standard one-way ANOVA with a Tukey post-test was used to determine whether statistical differences existed in C. albicans CFU counts and in diversity metrics (p , 0.05 considered significant). The R implementation of the Statistical Analysis of Microarrays (SAM) (R-package samr) was used to determine statistical differences in gene expression with significance (p , 0.05) being set for q 5 0.00 (qvalue of a gene equals the minimum estimated false discovery rate at which it appears significant). Differences in microbial community centroids were determined using a permutational MANOVA (function adonis() in the R-package vegan). The significance of constraints in constrained ordinations was tested using an ANOVAlike permutation test (function anova.cca in R-package vegan). All statistical analyses were performed in R or Prism v5.