The yeast form of the fungus Candida albicans promotes persistence in the gut of gnotobiotic mice

Many microorganisms that cause systemic, life-threatening infections in humans reside as harmless commensals in our digestive tract. Yet little is known about the biology of these microbes in the gut. Here, we visualize the interface between the human commensal and pathogenic fungus Candida albicans and the intestine of mice, a surrogate host. Because the indigenous mouse microbiota restricts C. albicans settlement, we compared the patterns of colonization in the gut of germ free and antibiotic-treated conventionally raised mice. In contrast to the heterogeneous morphologies found in the latter, we establish that in germ free animals the fungus almost uniformly adopts the yeast cell form, a proxy of its commensal state. By screening a collection of C. albicans transcription regulator deletion mutants in gnotobiotic mice, we identify several genes previously unknown to contribute to in vivo fitness. We investigate three of these regulators—ZCF8, ZFU2 and TRY4—and show that indeed they favor the yeast form over other morphologies. Consistent with this finding, we demonstrate that genetically inducing non-yeast cell morphologies is detrimental to the fitness of C. albicans in the gut. Furthermore, the identified regulators promote adherence of the fungus to a surface covered with mucin and to mucus-producing intestinal epithelial cells. In agreement with this result, histology sections indicate that C. albicans dwells in the murine gut in close proximity to the mucus layer. Thus, our findings reveal a set of regulators that endows C. albicans with the ability to endure in the intestine through multiple mechanisms.

Introduction The~100 trillion microorganisms that inhabit the human body comprise members of all three domains of life, i.e. bacteria, archaea and eukaryotes. (Viruses are also prominent components of this assembly). Commonly used approaches to inventory and study the microbiota, such as 16S ribosomal RNA sequencing, are nonetheless designed to target exclusively the bacterial component, inadvertently neglecting the other constituents. Non-bacterial residents of the mammalian gut likely play significant roles in shaping the multiple functions ascribed to the microbiota [1][2][3]. Fungal members of the intestinal community, for example, can affect the immunological responses of the host by dampening or promoting local inflammatory responses [3][4][5][6][7]. Fungi can also shape the composition of the bacterial community in the digestive tract of mammals [4]. Yet little is known about the genetic determinants that endow non-bacterial constituents of the microbiota with the ability to thrive in this niche.
The fungus Candida albicans is a common dweller of the digestive tract and other mucosal surfaces of humans and several warm-blooded animals. On occasion, this microorganism causes either fastidious mucosal disease (e.g. oral thrush) or life-threatening systemic infections. Such disseminated infections typically originate in the intestine of the same patients [8] and a debilitated immune system is the main risk factor associated with serious disease [9]. A growing number of C. albicans infections are also due to the contamination of implanted medical devices such as catheters [10][11][12]. While widely studied, most investigations on C. albicans have focused on traits that are thought to directly contribute to pathogenesis. Hence, the biology of the fungus in the mammalian gut, its more common habitat, remains underexplored.
Mice have been instrumental to uncover how C. albicans causes disease in humans despite the fact that these animals are not the fungus' natural hosts [13]. Adult mice with mature intact microbiota are resistant to intestinal colonization with this fungus [14][15][16]. Accordingly, murine gut colonization in the laboratory is routinely achieved through the oral administration of antibiotics [15,[17][18][19]. This standard mouse model of gut colonization has been employed to identify a handful of C. albicans genes with roles in commensalism [18][19][20][21] as well as mouse commensal bacteria that may prevent C. albicans colonization [14]. These studies have pointed to the importance of metabolic control [22], iron-acquisition regulation [23] and cell-to-cell variability [20] to flourish in this niche. Furthermore, several transcription regulators, i.e. sequence-specific DNA binding proteins, that govern these processes have been uncovered [17][18][19]21]. From the structure of a large transcriptional network built around regulators governing C. albicans gut colonization and/or disseminated infections, it has been postulated that commensalism and pathogenesis traits may be highly interconnected [18,24].
A significant amount of indigenous flora-particularly anaerobes-still persists after administering the antibiotics commonly used to achieve C. albicans intestinal colonization in mice [14]. Hence, a challenge in this experimental setup is to untangle the effects of such undefined microbiota on the biology of the fungus. Germ free mice provide an alternative to circumvent this limitation. Indeed, reports from the early 1970s already demonstrated the feasibility of establishing long-term C. albicans colonization in the gut of various strains of germ free mice [25,26]. Here, by contrasting mice carrying indigenous microbiota to germ free rodents monocolonized with the fungus, we reveal a significant effect of the host's microbial community on the morphology of the fungus. Specifically, we show that germ free animals provide an environment more conducive for C. albicans to adopt the oval-shaped yeast form. Screening a collection of C. albicans deletion mutants in gnotobiotic rodents allowed us to discover a hitherto unknown transcriptional network that promotes the endurance of the yeast form of the fungus in the intestine. Thus, studying the patterns of fungal colonization not only in antibiotic-treated animals but also in germ free mice is likely to reveal novel and relevant aspects of the biology of the various members of the human mycobiota.

Results
Imaging the interface between the fungus Candida albicans and the intestine of germ free or antibiotic-treated conventionally raised mice Murine gut colonization by the fungus C. albicans is typically achieved by the non-selective reduction of the indigenous mouse flora through antibiotic treatment. We [18] and others [15,17,19] have employed these conventionally raised, antibiotic treated animals (hereafter termed 'conventional') to investigate multiple aspects of intestinal colonization by the fungus. Yet the fact that C. albicans does not naturally cohabit with the mouse indigenous microbiota raises questions on whether the flora that remains after antibiotic treatment may confound some of these studies. Thus, as a complementary approach to discover fungal traits required for colonization of the digestive tract, we sought to experiment with gnotobiotic mice. In this setting, C. albicans is administered orally to animals that have been raised in a microbe-free environment (germ free mice), providing a clearly defined study system (Fig 1A). After gavaging an inoculum of *1×10 7 C. albicans cells, the fungus reaches similar loads (in colony forming units per gram of stool) in germ free rodents compared to conventional animals ( Fig  1B). (Comparable numbers were also obtained when plating intestinal and cecum contents.) This observation is in agreement with other reports [14,19].
Histological analyses of C. albicans cells proliferating in mammalian tissues, e.g. on the murine oral or vaginal mucosae or in kidneys, consistently show the filamentous form (also termed hyphae) of the fungus as the predominant morphology. This pattern is expected at sites of infection because filamentation is known to be a key determinant of Candida pathogenesis. The mammalian gut, on the other hand, is an environment where the fungus resides as a commensal and as such-at least a priori-it would not be expected to filament.
To establish what morphology C. albicans adopts in the gut of mice, we collected sections of the colon of germ free and conventional animals colonized with the fungus (Fig 1C). These tissue segments were fixed and processed using a method [27,28] that preserves the mucus layer structure and is compatible with histology and immunofluorescence. We first used Periodic acid-Schiff (PAS) staining because the polysaccharides on the surface of C. albicans are readily stained with this procedure. Furthermore, the intestinal mucus layer and mucus containing cells are also distinguishable with PAS. As shown in Fig 1D, we could easily visualize a large number of oval-shaped cells-reminiscent of the typical yeast morphology of C. albicanslocated in the vicinity of the mucus layer of the intestine of germ free mice monocolonized with the fungus. Similar patterns were observed with an independent C. albicans wild-type isolate (S1 Fig). In contrast to these animals, we observed that C. albicans adopts heterogeneous morphologies (elongated cells, filaments and yeast forms) which are scattered in the colon of conventional rodents (Figs 1D and S2). Quantification of the different cell morphologies in both types of animals indicated that yeast cells are the predominant form in monocolonized mice but not in conventional rodents ( Fig 1E).
To probe whether the yeast-like structures that we observed in gnotobiotic mice indeed correspond to C. albicans, we used immunofluorescence. Colon sections were stained with an anti-Candida specific antibody and DAPI (to stain nuclei in the epithelium). As illustrated in Fig 1F, there was a strong fluorescent signal originating from the anti-Candida antibody in the area of the intestinal lumen adjacent to the mucus layer. The equivalent analysis in conventional animals, on the other hand, could identify only a few C. albicans cells in the area proximal to the mucosa (Figs 1F and S2). From these results we conclude that in mice monocolonized with C. albicans, the fungus populates the region proximal to the intestinal mucus layer and that the yeast form of the fungus is the most prevalent morphology in this niche. By contrast, in conventional animals only a few scattered C. albicans cells are found close to the mucosa and they adopt multiple morphologies. Thus, the experimental setting in which germ free mice are gavaged with the fungus provides a more homogeneous system to study the interplay between the yeast form of C. albicans and the mammalian host in vivo.
Identification of C. albicans regulatory genes that contribute to fitness in the digestive tract of gnotobiotic mice To start dissecting the mechanisms that sustain the proliferation of the fungus in this niche, we screened a collection of C. albicans transcription regulator deletion mutants for their ability to endure in the gut of gnotobiotic mice monocolonized with the fungus. To be able to compare our screening results to other animal models, we focused on the same set of~60 deletion mutants that have previously been probed in a mouse model of disseminated infection and in the antibiotic-promoted murine gut colonization setting [18]. The rationale for screening transcription regulator genes is that transcriptional circuits are central to the regulation of many biological processes; hence, they are likely to be critical components of the gene circuits that mediate gastrointestinal colonization. None of the~60 deletion mutants display growth defect or any other major phenotype under >50 laboratory conditions tested [29]. Thus, this subset of regulatory genes prioritizes the identification of circuits that are likely to be relevant in vivo.
To minimize the number of mice needed for the screen, we assayed the mutants in pools of 15-18 strains each, and quantified their abundance in feces and intestinal contents by qPCR (each strain is signature tagged) for a period of 21 days as described [18] (Fig 2A). In total, the genetic screen in gnotobiotic mice revealed eight C. albicans deletion mutants with reduced fitness in this setting ( Fig 2B) [30][31][32][33] but the fact that they also displayed a phenotype in our study validates the experimental model and screening approach. Another four deletion mutants (zcf8Δ/Δ [orf19.1718], zfu2Δ/Δ [orf19.6781], try4Δ/Δ [orf19.5975] and orf19.5910Δ/Δ) had not been identified before in any other mouse setting. In this report, we investigate further the functions of ZCF8, ZFU2 and TRY4. We excluded ORF19.5910 because it encodes a protein predicted to be part of a histone remodelling complex and as such it likely has pleiotropic effects on gene expression. We corroborated the results of the screen by independently testing the gut colonization phenotype of the zcf8Δ/Δ, zfu2Δ/Δ and try4Δ/Δ strains at different time points after gavage ( Fig 2C). Furthermore, we confirmed that the gut colonization phenotype could be ascribed to the specified loci because adding back the corresponding wild-type copy of the gene to each deletion mutant strain restored, to a large extent, in vivo fitness (Fig 3).

C. albicans filamentation negatively associates with gut colonization
The C. albicans genes ZCF8, ZFU2 and TRY4 lack clear orthologs in the model yeast Saccharomyces cerevisiae [34]; therefore, it was not possible to use orthology to infer any biological process that might be associated with the phenotype reported here. To determine how ZCF8, ZFU2 and TRY4 may promote gut colonization, we first assayed the proliferation of the deletion mutants through spot assays under a variety of in vitro conditions that had not been tested before. It should be noted that, as described above, Homann et al. [29] had already conducted a large phenotypic screening of these mutants but found no overt phenotype. We observed slight growth impairment in the three deletion mutants when ethanol was used as carbon source in minimal synthetic medium (YNB without amino acids) (S3 Fig) and noticed that they formed wrinkles ( Fig 4A) when spotted on rich media (YPD). Wrinkling in C. albicans is a well-established proxy for filamentation; indeed, we confirmed by microscopy that the wrinkling sectors contained a mixture of yeast and filamenting cells. Strikingly, colony wrinkling in the mutants happened even when the agar plates were incubated at 30˚C, well below the body temperature (37˚C) that normally induces filamentation in C. albicans. We confirmed that the colony wrinkling phenotype could be ascribed to the specified loci because adding back the corresponding wild-type copy of the gene to each deletion mutant strain restored, at least in

Identification of C. albicans transcription regulators that govern gut colonization in gnotobiotic mice. (A) Schematic
representation of the screening approach to identify genetic determinants of in vivo fitness in the fungus C. albicans. Each one of four pools consisting of 15-18 signature-tagged strains was gavaged in four germ-free mice. The abundance of each strain in the inoculum and after recovery from fecal pellets was determined by qPCR with oligos complementary to the signature tags. (B) Log 2 (recovered/ inoculum) values for each deletion mutant at one or 21 days post gavage (each column represents one mouse). Color intensity indicates reduction (blue) or accumulation (red). The order in which the mutants are displayed reflect hierarchical clustering. The eight mutants that cluster at the top display a consistent pattern of reduction in abundance in all mice at day 21. Four of these deletion mutants (orf19.4941, orf19.2315, orf19.921 and orf19.4722) had previously been identified in a screening conducted in the antibioticpromoted gut colonization murine model. The deletion mutants try4Δ/Δ [orf19.5975], zfu2Δ/Δ [orf19.6781], zcf8Δ/Δ [orf19.1718], and orf19.5910Δ/Δ cluster together with the previous group but had not been identified before in any other mouse setting. (C) Validation of the gut fitness defect phenotype of three newly identified genes. Germ free animals were gavaged with 1:1 mixtures of the wild-type reference strain and try4Δ/Δ, zfu2Δ/Δ or zcf8Δ/Δ. The abundance of each strain in the inoculum (I) and after recovery from fecal pellets (R) at day 7, 15 and 21 was determined by qPCR (strains were barcoded). The deletion mutant strains were progressively depleted from the fecal pellets relative to the wild-type reference strain (P values are indicated on top of each comparison). Each circle represents the measurement from one mouse. Statistical analysis by the Mann-Whitney test. NS, not significant.
https://doi.org/10.1371/journal.ppat.1006699.g002 Adding back a wild-type copy of the respective genes to the deletion mutant strains restores in vivo fitness. Germ free animals were gavaged with 1:1 mixtures of each deletion mutant and the corresponding gene add-back strain in three or four mice. The abundance of each strain in the inoculum (I) and after recovery from fecal pellets (R) was estimated by qPCR at day 1, 9 and 21 post gavage (strains were barcoded). In contrast to the deletion mutants which were depleted over time from the fecal pellets, the add-back strains maintained their relative abundance (P values are indicated on top of each comparison). Each circle represents the measurement from one mouse. Statistical analysis by two-tailed unpaired t-test assuming unequal variance. NS, not significant.
https://doi.org/10.1371/journal.ppat.1006699.g003 part, the smooth colony phenotype (S4 Fig). Since a major feature of the C. albicans cells residing in the colon of gnotobiotic rodents is their yeast morphology (Fig 1D and 1E), these findings raised the possibility that an increased tendency to filament may account for, at least in part, the gut colonization defect of the zcf8, zfu2 and try4 mutants. More generally, this notion implies that filamentation in itself is a detrimental factor for intestinal colonization.
To independently test whether there was a link between filamentation and impaired gut colonization, we constructed a strain that ectopically expressed UME6, a well-studied regulator that promotes filamentation in C. albicans [35][36][37]. In this strain, UME6 transcription was driven by the strong, constitutive promoter TDH3 [38]. UME6 overexpression has been shown to be sufficient to drive filamentation in this organism even under non-filamenting conditions [35]. When gavaged in germ-free animals, this strain was severely outcompeted by its wildtype parental strain (Fig 4B) indicating that filamentation in C. albicans is indeed detrimental for gut colonization.

Transcriptional network associated with the regulators ZCF8, ZFU2 and TRY4
Our results thus far suggested that the C. albicans genes ZCF8, ZFU2 and TRY4-which encode proteins containing putative sequence-specific DNA binding domains (zinc cluster or zinc fingers)-negatively regulate filamentation under certain conditions. Since we demonstrate that inducing filamentation in this fungus is detrimental to gut colonization, the connection between the regulators and morphology could explain their fitness defect in gnotobiotic mice. To dissect how ZCF8, ZFU2 and TRY4 can influence morphology, and whether other cellular functions may also account for their colonization phenotype, we next sought to uncover the entire gene repertoire under control of the three regulators. For this, we conducted transcriptome analyses (RNA sequencing) comparing the wild-type reference strain to each one of the three deletion mutants. To try to mimic at least some of the in vivo conditions that we observed in the intestine (yeast morphology, association to a surface and body temperature), RNA was prepared from cells growing on a semisolid surface (Todd-Hewitt agar) at 37˚C. We determined that C. albicans cells remain in the yeast form in this culture medium under the described experimental conditions (>99% of cells were in the yeast form and devoid of germ tubes upon inspection under the microscope). Clearly, these conditions do not fully recapitulate the environment found in the mammalian gut; hence, this caveat needs to be taken into account in the interpretation of the data.
Overall, the RNA-seq experiment identified 395 protein-coding transcripts whose expression was dependent on one or more of the three regulators evaluated (P adj < 0.01; Fig 5A, S1 Table; 560 targets at P adj < 0.05; 670 targets at P adj < 0.1). While each individual regulator, ZCF8, ZFU2 and TRY4, is connected to expression changes in 262, 192 and 211 transcripts, respectively, the large amount of shared targets (89 transcripts in common for all the three proteins, P = 6.09×10 −216 ; 38 additional transcripts shared only by ZCF8 and ZFU2, P < 8.09×10 −17 ; 37 shared only by ZCF8 and TRY4, P < 1.33×10 −14 ; and 15 in common only between ZFU2 and TRY4, P < 8.77×10 −4 ) is a strong indicator of a significant degree of overlap in the functions governed by the three proteins ( Fig 5B). Even more remarkable is the finding that, in the set of shared targets, the direction of the change in transcript levels, i.e. activation or repression, is consistent in all but 2 of the 179 shared genes (Fig 5B): This means that if the expression of a gene is influenced by all three regulators, it is subject to positive regulation by all three or to negative regulation by all three as well. The same applies to transcripts that are shared by two regulators. All these features of the network are robust to threshold changes (see Materials and Methods and S1 Table). From this global analysis we conclude that ZCF8, ZFU2 and TRY4 distribution of all analyzed transcripts (*6100 genes; each dot represents one transcript) when comparing the indicated deletion mutant to the wildtype reference strain. Plotted is the log 2 fold change in expression (Y-axis) as a function of the mean of the normalized read counts (X-axis) for each transcript. Colored triangles correspond to transcripts whose expression is activated (i.e. positive regulation) (red) or repressed (i.e. negative regulation) (blue) by each regulator at P adj < 0.01. (B) Network view of the full set of genes whose expression is influenced by the identified C. albicans transcription regulators. The nodes of the network (in shades of green) are transcription regulators identified in our genetic screen. Large circles are composed of small colored ellipses. Each one of the latter represents an individual target gene. Positive (i.e. activation) and negative (i.e. repression) regulation is indicated by red and blue colors, respectively. The two black small ellipses depict the two genes that show discordant regulation (i.e. they are positively controlled by one regulator and negatively regulated by another). The shown network is completely based on RNA-seq data generated in this study. govern, to a significant extent, overlapping cellular functions and that they may act in concert to promote gut colonization.
Can a plausible link between the three regulators and filamentation be inferred from the RNA-seq data? Among the top genes whose expression is promoted by ZCF8, ZFU2 and TRY4 (based on the magnitude of expression differences between reference strain and deletion mutants), we found the putative regulator FGR17 (Table 1) which has been reported to influence filamentation in C. albicans although with conflicting observations [29,39]. We established that, under conditions that are likely to be relevant in the gut (37˚C and growing on a semisolid surface), FGR17 indeed contributes to preclude filamentation because the fgr17 deletion mutant strain displayed higher levels of colony wrinkling (S5 Fig) phenocoping the zcf8,  zfu2 and try4 deletion mutants (Fig 4A). As expected, the wrinkling areas contained a mixture of yeast and filamenting cells when observed under the microscope. Furthermore, when grown in embedded soft agar, the fgr17 deletion mutant developed longer and denser projections compared to the wild-type reference strain (S5 Fig). Thus, our transcriptome data revealed FGR17, a negative regulator of filamentation, as a prominent target of regulation shared by ZCF8, ZFU2 and TRY4.
Ectopic expression of WOR3, another driver of C. albicans morphology conversion, reduces fitness in the gut of gnotobiotic mice Another top target of regulation common to ZCF8, ZFU2 and TRY4 (based on the magnitude of expression differences between reference strain and deletion mutants) was WOR3 (Table 1). WOR3 is a gene that when overproduced induces formation of elongated opaque cells [40], a heritable morphology that C. albicans can adopt under certain environmental conditions [41,42] and that is distinct from the oval-shaped yeast form. Our RNA-seq data showed that the regulators ZCF8, ZFU2 and TRY4 negatively control WOR3 transcript levels (Fig 6A) raising the possibility that high levels of WOR3 may be detrimental to gut colonization. To test this hypothesis, we constructed a strain that ectopically expressed WOR3 (by placing the coding region of this gene under control of the strong, constitutive promoter TDH3). WOR3 overexpression has been shown to be sufficient to drive in mass conversion to the opaque morphology [40]. When gavaged in germ-free animals, this strain was outcompeted by its wildtype parental strain ( Fig 6B). Thus, just as filamentation, WOR3-driven opaque cell formation also impairs the fitness of the fungus in the gut of gnotobiotic mice. These results imply that keeping WOR3 transcript levels low (Fig 6A) is another mechanism whereby the three regulators (ZCF8, ZFU2 and TRY4) may foster gut colonization. In fact, these findings further support our initial observation (Fig 1) that the oval-shaped yeast form of the fungus is the most prevalent cell morphology in this niche.
While additional mechanisms of morphology control cannot be ruled out, our results indicate that by promoting transcription of FGR17 and repressing WOR3, three of the gut colonization regulators that we identified in our screen can favor the yeast form over filament and opaque cell formation, respectively.
ZCF8, ZFU2 and TRY4 promote C. albicans adherence We established above that the yeast morphology is a key attribute of C. albicans in the gastrointestinal tract of animals monocolonized with the fungus. Accordingly, the regulators ZCF8, ZFU2 and TRY4 control the expression of determinants predicted to favor the yeast form. However, our RNA-seq data clearly indicates that the expression of many other genes are influenced by ZCF8, ZFU2 and TRY4. Thus, we next sought to explore whether additional functions could contribute to the colonization phenotype. Gene Ontology term searches retrieved 'cell wall' and 'cell periphery' (P = 0.015 if considering the set of 395 transcripts that showed changes at the most stringent threshold; P = 4.47×10 −3 if considering the set of 558 targets; P = 8.4×10 −4 if considering the set of 670 targets included at the lower stringency threshold) as well as 'response to oxidative stress' (P = 2.1×10 −4 if considering the set of 395 transcripts that showed changes at the most stringent threshold; P = 1.2×10 −4 if considering the set of 558 targets; P = 5×10 −4 if considering the set of 670 targets included at the lower stringency threshold) as cellular components, functions or processes overrepresented in the full RNA-seq dataset. Almost a third (n = 119) of the genes included in the network, however, are currently annotated as having 'unknown function' [43].
The zcf8, zfu2 and try4 mutants had previously been identified in an in vitro genetic screening for cell-substrate adherence (using the silicone poly-dimethyl siloxane as substrate) [44]. A fourth gene identified in that screen, TRY6, is also embedded in the transcriptional circuit that we mapped here (Fig 5). Consistent with such a role, we found that several members of the agglutinin-like sequence (ALS) family of adhesins (ALS2, ALS4, ALS5, and ALS9) [45] were targets of regulation of ZCF8, ZFU2 and/or TRY4 (Fig 5B and S1 Table).
Likewise, as described in the previous paragraph, 'cell wall' and 'cell surface' are GO terms enriched in the network (Fig 5). Thus, we considered the possibility that, in addition to promoting adherence to silicone, they could also contribute to the attachment to surfaces relevant in the context of the gastrointestinal tract.
To explore this idea, we first probed whether the C. albicans reference strain and the deletion mutants could adhere to surfaces coated with mucin, the main component of the intestinal mucus. We devised an in vitro assay (Fig 7A and Materials and Methods) in which yeast cells were added on top of an agar layer containing mucin or PBS as control. As shown in Fig  7B, the addition of mucin promotes the adherence of C. albicans to the agar layer. The three deletion mutant strains, however, displayed impaired adherence in the presence of mucin. Collectively, these findings indicate that ZCF8, ZFU2 and TRY4 are needed for the fungus to adhere to surfaces such as those coated with silicone [44] or mucin. To further evaluate the role of the identified regulators in promoting adherence to a relevant surface, we probed the interactions between C. albicans and intestinal epithelial cells that produce mucus. Using the HT29-MTX-E12 cell line, which is derived from human colorectal adenocarcinoma cells (see Materials and Methods), we found that wild-type C. albicans forms cell clusters attached to the surface of the monolayer (Fig 7C). The formation of these clusters is fully dependent on the presence of epithelial cells because no clusters were found in the absence of the HT29-MTX monolayer. Furthermore, the C. albicans cells were strongly attached to the epithelial cells because the clusters persisted despite stringent washing. The C. albicans zcf8 deletion mutant strain showed a statistically significant (P < 0.05) reduction in the formation of cell clusters on top of the monolayer (Fig 7D) whereas the try4 and zfu2 mutants displayed a trend towards decrease (albeit their P values did not reach statistical significance at P < 0.05). These results demonstrate that at least one of the identified regulators, ZCF8, can also foster the interaction between the fungus and mucus derived from human intestinal cells.
Taken together, our findings indicate that in addition to influencing morphology, the regulators ZCF8, ZFU2 and TRY4 can also promote adherence to surfaces that are relevant in the intestine of the host.

Discussion
Fungi are prominent components of the human microbiota [3,46], yet the mechanisms that these eukaryotic microorganisms employ to flourish as commensals in the mammalian gut remain largely unexplored. In this report, we have delineated the interface between C. albicans and the murine intestinal epithelium and established that the oval-shaped yeast morphology is a key determinant of C. albicans' ability to persist in the gastrointestinal tract of gnotobiotic mice monocolonized with the fungus. Several lines of evidence support this conclusion: First, the fungus almost uniformly adopts the oval-shaped yeast cell morphology in the gut of monocolonized mice (Fig 1); second, the ectopic expression of UME6 or WOR3, two regulators that can drive the formation of filaments or opaque cells, respectively, significantly impairs the fitness of the fungus in this niche (Figs 4B and 6B); and third, deletion of any of three regulatory genes (ZCF8, ZFU2 and TRY4) that negatively control filamentation and the expression of WOR3 reduces fitness in the murine gut (Figs 2C, 3, 4A and 6A). While different C. albicans isolates may vary greatly in their ability to colonize mucosal surfaces, we show that at least two different strains (SC5314 and WO-1) remain mostly in the yeast form in the colon of monocolonized rodents. Therefore, the milieu in the intestine of germ free mice is conducive to such Candida's morphology.
C. albicans, as most fungi, can adopt diverse cell morphologies. The mammalian body temperature (37˚C) as well as the presence of serum are robust inducers of one of these morphologies, the filamentous form. Accordingly, this is the form typically found in histology sections of mammalian tissues infected with C. albicans (oral and vaginal mucosae or internal organs such as kidneys). The mammalian gut, on the other hand, is a niche where the fungus resides, for the most part, as a harmless commensal. Our observation that C. albicans remains in the yeast form in monocolonized animals is in agreement with, and supports, such principle. The yeast form of the fungus is, therefore, stable at body temperature, implying that conditions in the gut environment or host-derived factors override the temperature-and/or serum-induced filamentation. An unusual morphology, termed the GUT phenotype [17,47], has been postulated as a specialized form of C. albicans adapted to the digestive tract. While we cannot rule out the existence of this morphology in our experimental setting (because, for example, such specialized cells could be restricted to specific compartments in the digestive tract), our findings show that the more familiar form of the fungus, the oval-shaped yeast cells, are abundant and stable in this niche (Fig 1). Thus, our findings suggest that no additional morphologies need to be invoked to explain the commensal state of the fungus in the digestive tract.
The distinct gut environments found in germ free and in conventional animals appear to influence the morphology and distribution C. albicans cells in the colon: In germ free animals, fungal cells were found in significant numbers in the vicinity of the mucus layer whereas only few C. albicans cells were visible in the equivalent region in conventional animals (Fig 1D and  1F). Because the fungal load in stool samples is comparable between both types of animals ( Fig  1B), such patterns are unlikely due to differences in the number of colonizing cells. Instead, this could reflect intrinsic differences in accessibility to the intestinal mucus layer. For example, the mucus in the colon of conventional animals is considered impenetrable whereas in germ free mice it can be penetrated rather easily [48]. Alternatively, but not mutually exclusive, the under-development of the immune system in germ free mice may reduce the spatial segregation between the intestinal mucosa and microbes which is maintained by secreted antimicrobials such as RegIIIγ [49]. Further studies would be needed to establish whether these or other mechanisms underlie the documented differences in C. albicans localization in the murine colon between both types of animals.
What is the basis for the preference of the yeast form over hyphae in the gut of gnotobiotic rodents? One possibility is that the filamentous form induces, and is more susceptible to, gut immune effectors (e.g. antimicrobial peptides). Our finding that the levels of the neutrophil stimulating chemokine G-CSF (Granulocyte-colony stimulating factor) are higher in intestinal tissue carrying the filamentous form of the fungus (S6 Fig) supports this notion. An alternative scenario, not mutually exclusive with the previous idea, is that both morphologies differ in their resistance/ susceptibility to the gut environment (e.g. metabolites that change in abundance depending on the microbial composition). It is plausible that the pressure to maintain C. albicans in the yeast form in monocolonized mice arises from the need to down-regulate any filament-associated virulence factors (i.e. that would cause damage to host cells) and avoid eliciting a strong anti-fungal or pro-inflammatory host response. On the other hand, in the presence of competing and/or antagonistic microbes (e.g. in conventional animals), the fungus may have to employ heterogeneous morphologies to be able to better compete and persist in such complex ecosystem.
By employing gnotobiotic mice as the experimental system to carry out a genetic screen, we identified previously undescribed genes that contribute to fitness in this niche. The three new regulators that we uncover here (ZCF8, ZFU2 and TRY4) have a role in gut colonization (by contrast, none of them showed any phenotype in similar screenings conducted in the standard mouse model of disseminated candidiasis [18,50]). We infer at least three complementary mechanisms whereby the identified regulators may promote fitness in the gastrointestinal tract: First, they favor the yeast over the filamentous morphology (Figs 4 and S4); second, they keep the transcript levels of the regulator WOR3 low, thus preventing WOR3-induced opaque cell formation (Fig 6); and third, they contribute to adherence to both mucin-containing surfaces and mucus-producing intestinal epithelial cells (Fig 7). The transcriptome profiling of the three deletion mutants (Fig 5) clearly indicates that there are many other transcripts under control of ZCF8, ZFU2 and TRY4. These targets of regulation portray a variety of cellular processes. Therefore, additional functions may contribute as well to the mouse colonization phenotype of the identified regulators. For example, another GO term enriched in the network, 'response to oxidative stress' may also be relevant because oxidative stress responses reportedly play important roles in colonization of the digestive tract [51].
A potential link between the three regulators, adherence and gut colonization is inferred by our in vitro assays on two relevant substrates (intestinal mucin and intestinal epithelial cells). Intestinal bacteria such as Lactobacillus rhamnosus and Streptococcus gallolyticus are known to harbour genes encoding proteins that mediate binding to mucus [52,53]. To our knowledge, however, this has not been reported for fungi. The bacterial machinery that promotes adhesion to intestinal mucus has been shown to play a key role in enabling colonization of the mouse colon [53]. This is in agreement with our finding that the zcf8, zfu2 and try4 mutants display, on the one hand, reduced ability to bind to a surface containing mucin and, on the other hand, impaired gut colonization. Indeed, histology sections of the colon show that the fungus lies in clusters of cells in close proximity to the intestinal mucus layer (Fig 1D and 1F). Since a variety of known and predicted cell surface proteins (e.g. the agglutinin-like sequence adhesins ALS2, ALS4, ALS5, and ALS9) are embedded in the network that we map here, we speculate that the adherence phenotype that we observe is the result of the misregulation of not a single but multiple of these loci. Since intestinal mucins have been shown to suppress virulence traits in this fungus [54] and such traits include adhesins, it is plausible that mucin itself influences adherence. However, the mucin-responsive adhesins (e.g. ALS1 and ALS3) [54] are associated mostly with hyphae formation. By contrast, the adhesins that are targets of regulation of ZCF8, ZFU2 and TRY4 (e.g. ALS2 and ALS4) are unaffected by mucin [54].
ZCF8, ZFU2 and TRY4 have been shown to mediate the adherence of C. albicans cells to medical devices such as catheters [44]. The attachment to these devices is the first step in the formation of biofilms, a common problem in the clinic [11,12]. Imaging of C. albicans biofilms formed in vitro and in vivo show that the basal layer of the biofilm, i.e. the layer that is in direct contact with the device, is made out mostly of cells in the yeast form [55]. This observation further supports the role played by all three regulators specifically in the adherence of the yeast cell morphology (as opposed to adherence by the filamentous form which is mediated by other factors [56]). Thus, the circuitry directed by ZCF8, ZFU2 and TRY4 may be employed by C. albicans in strikingly different contexts (intestine and medical devices), although the underlying function is essentially the same in either setting. These findings illustrate how the same gene repertoire that sustains C. albicans colonization in the mammalian gut as commensal, can be diverted to drive attachment to other surfaces and fuel bloodstream infections.
Of the eight transcription regulators that we identified in our genetic screen in gnotobiotic mice monocolonized with C. albicans, four displayed a similar phenotype in a gut colonization model that employed antibiotic-treated conventionally raised rodents [18]. The fact that these regulatory genes contribute to the fitness of C. albicans in the gut independently of the presence or absence of microbiota suggests that their function is aimed primarily at coping with demands imposed by the host (as opposed to demands that could originate in competing microbes). The role of TYE7, RTG1 and RTG3 in regulating nutrient acquisition and metabolism [30,31,33] may fall in this category. Consistent with this notion, the RTG1 and RTG3 genes also contribute to Candida fitness in bloodstream infections [18,50], i.e. in an environment that is devoid of other microbes. On the other hand, the remaining hits that we identified in the screen displayed a fitness defect only in gnotobiotic mice monocolonized with the fungus. The spatial distribution of C. albicans in the colon of gnotobiotic mice differs significantly from that of conventionally raised rodents colonized with help of antibiotic treatment (Fig 1). The different behaviour of the mutant strains between both settings may thus reflect an effect of the residual microbiota that remains after antibiotic treatment or, alternatively, differences in the murine response to the presence/absence of microbiota. Untangling these effects would require further studies, although it is unclear whether the indigenous microbiota found in conventional mice is relevant to the study of C. albicans commensalism because the fungus does not naturally cohabit with the murine microbiome.
In sum, we have shown that the use of gnotobiotic mice monocolonized with C. albicans can reveal new aspects of the biology of this human gut commensal. The fact that in this experimental setting the fungal cells adopt a homogeneous yeast morphology-as opposed to the morphologically heterogenous population found in antibiotic-treated mice-indicates that such system is particularly suited to explore the biology of this form of C. albicans in the mammalian gut. Furthermore, since the intestinal microbiota can significantly influence the proliferation of the fungus, gnotobiotic mice will become instrumental in dissecting the in vivo interplay with relevant members of the gut community. As exemplified by the gene regulators uncovered here, a more systematic study of Candida's commensal lifestyle shall reveal the extent to which fungal traits that allow C. albicans to thrive in the gut are co-opted to initiate or sustain invasive proliferation and cause disease.

Materials and methods
Strains and DNA oligos used in this study are listed in S2 and S3 Tables, respectively. All C. albicans strains were derivatives from the wild-type reference strain SN152 [57] unless indicated otherwise. Gene deletions were constructed as described [57]. Overexpression strains were constructed by placing a copy of the TDH3 promoter upstream of the coding sequences of interest as described previously [38].
Gastrointestinal tract colonization C. albicans strains were grown overnight at 30˚C in YPD liquid medium, washed twice with sterile 1×PBS, and counted in a hemocytometer. Conventional NMRI mice were treated with antibiotics and gavaged as described [18]. Germ free female NMRI mice (20-22 g) were gavaged with *1×10 7 C. albicans cells (in a 0.1 or 0.2 mL volume). The animals were maintained in cages within a sterile isolator throughout the experiment. Colonization was monitored by collecting and plating fecal pellets (produced at the time of collection) at various time points after inoculation. Cecum contents were also collected and plated at day 21 post gavage, when the experiment was terminated. The screening procedure and analysis was carried out exactly as described [18,50]. The number of animals per group (n) was four to five, as this number has been shown to provide a reasonable balance between statistical power and the high cost of experimenting with germ free mice [14]. All experiments with germ free animals were conducted in the Core Facility for Germ-Free Research of the Karolinska Institute (Stockholm, Sweden).

Tissue collection, histology and immunohistochemistry
Complete colons were removed from the animals at the end of the experiment (day 21 post gavage), excised and cut in *1 cm pieces. Tissues were fixed and processed as described [28] with minor modifications. Briefly, samples were fixed using methacarn solution (60% methanol, 30% chloroform, 10% acetic acid) for 5 days. Successive washes in methanol 35 min and ethanol 30 min were repeated until there was no sign of acid remaining in the samples. Paraffin embedding was conducted on a paraffin embedding machine (Leica EG 1160) using a standard program consisting of washing steps in increasing ethanol concentrations, several xylol baths and 3 hours in melted paraffin at 62˚C. Paraffin blocks were stored at room temperature.
For histology analysis, paraffin blocks were cut in 2-4 μm sections. After deparaffinization in xylol and rehydration in decreasing ethanol concentrations, sections were incubated in Periodic acid followed by Schiff´s reagent (standard PAS staining). Hematoxylin was used to stain nuclei. Sections were dehydrated using increasing ethanol concentrations and xylol and embedded in Entellan. For immunohistochemistry, sections were processed as described [27]. Briefly, after deparaffinization and rehydration, C. albicans cells were visualized using a rabbit anti-C. albicans antibody (Abcam) diluted 1:500 in blocking buffer at 4˚C for 12-16 h in a humid chamber. After successive washing steps in PBS + 0.5% Tween-20 (PBST), the secondary antibody (Alexa Fluor 488 goat anti-rabbit IgG, Life Technologies) was applied at a 1:500 dilution in blocking buffer with 10 μg/ml DAPI (Sigma) and incubated for 2 hours at RT in a humid chamber. Slides were washed subsequently with PBST and mounted in Mowiol (Sigma).

Transcriptome analysis
The C. albicans reference strain and deletion mutants were grown on Todd-Hewitt agar at 37˚C for 24 h. Cell collection, RNA purification, library preparation and sequencing were carried out as we recently described [58]. Two biological replicates of each strain were analyzed.
Quality control, mapping and differential gene expression analysis. We obtained between 27-42 million reads per sample. Low quality reads were eliminated and adapter sequences trimmed using Trimmomatic [59] (v0.36) with default settings for single end reads. Over 95% of reads in every sample passed the quality control and were aligned to the C. albicans reference genome Build 21 using STAR [60] (v2.5.2b) with default parameters. From the resulting BAM files, raw read counts were extracted with HTSeq [61] (v0.6.1) using the respective annotation. Read counts were loaded into R (v3.3.2) and analyzed with the DESeq2 [62] package (v1.14.1). With our depth of sequencing, we detected a significant number of reads in >6100 annotated ORFs (99%). Differential gene expression analysis was performed using the default DESeq2 workflow. Adjusted P-values of 0.1, 0.05 and 0.01, which were obtained through the Benjamini-Hochberg-method [63], were taken as cut-offs for significance. The resulting networks and lists of genes at different cut-offs are shown in Fig 4 and S1 Table. To verify the validity of the analysis, we determined the P-value distribution across the samples. For the ZCF8 dataset, an overestimation of the null model variance was detected and corrected using fdrtool [64]. As an additional measure to reduce the rate of false positives in the corrected dataset, we applied a minimum fold-change threshold (log 2 values of |0.33|). This ad hoc measure does not change the overall structure of the inferred network. Cytoscape [65] (v3.4) was employed to visualize the network and generate the displayed graphs.

Adherence assays
Six-well plates containing solidified 1% agar were overlaid with 50% mucin (Mucin from porcine stomach Type III, Sigma) or 1×PBS. Plates were incubated at 4˚C overnight; excess mucin and 1×PBS were then removed from the wells. *1×10 3 C. albicans cells from an overnight culture in YPD were added to each well. Plates were incubated 8 hours at 37˚C under anaerobic conditions and the wells washed twice with 1×PBS to remove non-adherent cells. YPD agar was then overlaid and the plates incubated at 30˚C until colonies could be visualized and quantified.
Adherence to the mucus-producing human epithelial colorectal adenocarcinoma cell line HT29-MTX-E12 (Sigma-Aldrich, Germany) was evaluated by fluorescence microscopy. Briefly, 5×10 5 HT29-MTX-E12 cells were seeded on transwell membranes (Corning Transwell 0.4 μm pore polycarbonate membrane; 12 mm inserts purchased from Sigma-Aldrich, Germany) in DMEM medium supplemented with 10% FBS, 5% MEM amino acids and 5% L-Glutamine and allowed to grow for 21 days (medium was exchanged twice per week). One hour prior to infection with 5×10 6 C. albicans cells, the medium was changed to DMEM with 5% MEM amino acids and 5% L-Glutamine (i.e. without FBS). Fungal cells were incubated on the HT29-MTX-E12 monolayer for 1 h at 37˚C and 5% CO 2 . After co-incubation, unattached cells were removed by stringent washing (three times) with PBS and fixed in methacarn solution (60% methanol, 30% chloroform and 10% acetic acid) for 24 h. For histology analyses (shown in Fig 7C), membranes were washed twice with methanol for 35 min and once with ethanol for 30 min, dehydrated and embedded in paraffin. Sections were processed and PAS stained as described above for intestinal tissue. For adherence quantification, C. albicans cells were stained with a primary anti-C. albicans antibody (Life technologies) and a secondary Alexa Fluor 594 Goat anti-Rabbit antibody (Invitrogen). HT29-MTX-E12 nuclei were stained with Hoechst. Photographs from 25-30 fields per transwell (each strain was assayed in 3 transwells) were evaluated with ImageJ (v1.50i) to estimate the area of the monolayer covered by C. albicans cell clusters.
Colony morphology and embedded agar assays C. albicans strains were grown overnight in YPD liquid medium at 30˚C. To evaluate overall colony morphology, two methodologies were used. In the first method, fungal cells were spotted on YPD agar, plates were incubated at 30˚C and photographs were taken after 48 h. In a second method, fungal cells were plated on YPD agar at a density of 20-30 colony-forming units per plate; plates were incubated at 37˚C and photographs of single colonies were taken after 96 h.
To evaluate the morphology of colonies embedded in agar, cells were placed on top of a thin layer of soft (1%) YPD agar at a density of *100 colony-forming units per plate. Before solidification, another layer of soft agar was overlaid on top. Plates were left to dry for a few hours and then incubated at 37˚C. Photographs of embedded colonies were taken after 48 h.

Statistical analyses
The statistical analysis of the genetic screen was conducted as described [24,50]. The significance of the overlaps between the differentially expressed gene datasets was estimated using the hypergeometric test. To calculate the significance of the multi-set intersection between TRY4, ZFU2 und ZCF8 we used the SuperExactTest package (v0.9.9.2) for R. The t-test (twotailed) for unpaired samples assuming unequal variance was used to compare the adherence of multiple strains to the mucin-containing agar (Fig 7B). The Mann-Whitney U test was employed to compare the adherence to the mucus-producing cell line HT29-MTX-E12 ( Fig  7D).

Data availability
The RNA-seq data generated in this study has been submitted to the Gene Expression Omnibus (GEO) database (Accession number GSE96547).