C. albicans strains and other Candida species used in this study are summarized in Supplementary Table 1. Details about culture conditions and strain construction, including primers used in this study (Supplementary Table 2), can be found in Supplemental Methods. The ploidy of all generated strains was determined via a previously described quantitative flow cytometric method (Pavelka et al., 2010).

Genomic DNA was extracted as described (Rancati et al., 2008). Sequencing libraries were prepared using Illumina sample preparation kits and indexed paired-end (PE) sequences were obtained on an Illumina HiSeq platform. All sequencing data were deposited in NCBI SRA database under accession number SRP056269. Data pre-processing, mapping and variant calling was performed in CLC Genomics Workbench, essentially as described (Liu et al., 2015). Higher level data integration and analysis was performed via customized R scripts. Detailed sequencing protocols and analysis parameters are provided in Supplemental Methods. A summary of the sequencing data and results is reported in Supplementary Tables 3, 4.

Logarithmically-growing C. albicans was washed three times in 50 ml of 1 mM cold phenylmethanesulfonyl fluoride (PMSF). 3 volumes of glass beads were then added to one volume of pellet and vortexed for 10–15 times at 1-min intervals to break the cells. The cell lysate was again washed three times with 50 ml cold PMSF, before addition of 3 volumes of 2% SDS to the pellet and incubation for 10 min at 100°C for two rounds. SDS was removed by washing the pellet in 1 mM PMSF and the pellet was dried overnight at 55°C before re-suspension in 1 mM PMSF to a solution of 0.01 g dry weight / ml and storage at −20°C. Total chitin was extracted and measured as described (Kapteyn et al., 1997). Total mannan were extracted from isolated cell walls by alkali treatment and further precipitated with Fehling's solution as a copper complex as described (Chaffin et al., 1998). Total glucan was extracted as described (Dijkgraaf et al., 1996) and the extracted glucan and mannan components were then measured by the Dubois method (Dubois et al., 1956) using glucose as a standard.

Mouse experiments were conducted according to the rules and guidelines of the Agri-Food and Veterinary Authority (AVA) and the National Advisory Committee for Laboratory Animal Research (NACLAR), Singapore. The experiments were reviewed and approved by the Institutional Review Board of the Biological Resource Center, Singapore (IACUC protocol 140955). Six to ten weeks-old female C57BL/6 wild-type mice were used for all experiments. Systemic virulence was assessed after injection of 5 × 10⁵ CFUs via the tail vein. GI competition experiments were performed essentially as described in Prieto et al. (2014). Briefly, equimolar mixtures of a test strain and a fluorescently tagged (dTomato-expressing) reference strain were intragastrically inoculated in antibiotic-treated mice (2 mg/ml streptomycin + 1500 U/ml penicillin G) and their relative frequency was monitored over time by plating stool homogenates on antibiotics-containing plates and counting fluorescent colonies under a fluorescence stereomicroscope. The relative fitness coefficient of the test strain was then obtained by linear regression according to the following formula: (1)log2[R(t)/R(t0)]=sγR(t−t0), where R(i) represents the ratio between the test strain and the reference strain at time i; s is the selection coefficient; γ_R is the growth rate of the reference strain expressed as cell divisions per hour; t represents the time points in hours and t₀ the initial time point. In vitro fitness competition assays were performed essentially in the same way, except that strain mixtures were serially passaged on a daily basis in YPD medium. More details are provided in the Supplemental Methods.

J774.1 cells were plated at a density of 5 × 10⁵ cells/well in 12-well plates for 24 h. C. albicans strains from overnight cultures were washed twice in PBS and co-cultured with the macrophages at various MOIs, ranging from 1:200 to 1:1600. Under these conditions, >99% of individual C. albicans cells were phagocytosed by the macrophages after 3 h (data not shown). Visible fungal colonies were counted after 24 h, and the percentage of phagosome-escaping cells was determined by linear regression as explained in the Supplemental Methods.

Resistance to mCRAMP was determined on logarithmically growing cells re-suspended in sterile PBS and incubated at 37°C for 1–2 h. Cell pellets were then re-suspended in a 2-fold serial dilution of mCRAMP 1–39 (Synpeptide) and incubated at 37°C for 2 h, followed by a PBS wash, live/dead staining with 30 μM propidium iodide for 15 min and flow cytometric analysis. Half-maximal effective concentrations (EC₅₀) were determined by non-linear regression in GraphPad Prism. The half-maximal inhibitory concentrations (IC₅₀) of acetic and lactic acid on fungal growth were determined in vitro via turbidimetric assays as described previously (Cottier et al., 2015a) with some minor modifications (refer to Supplemental Methods). Oxidative stress resistance was determined on PBS-washed overnight cultures incubated in PBS at 37°C for 2 h. Cells were then treated with a range of tert-Butyl hydroperoxide (tBOOH) concentrations, incubated at 37°C for 1 h and subjected to live/dead analysis as described above. The NO sensitivity assay was performed following previous protocol (Chiranand et al., 2008) with some minor modifications (refer to Supplemental Methods).

Overnight cultures of freshly revitalized C. albicans strains and Candida spp. were normalized to a final OD₆₀₀ of 2 and washed once with PBS. A Freedom EVO 150 liquid-handling robot (Tecan) performed serial dilutions of the cultures and spotted 3 μl of each serial dilution onto YPD agar omnitrays representing one of a wide range of growth conditions listed in Supplementary Table 5. Each condition was tested a minimum of three independent times and growth data was acquired by a desktop scanner and analyzed using a custom R script for automated spot detection and intensity measurements followed by non-linear curve fitting across the range of serial dilutions and determination of relative growth scores as explained in Supplemental Methods.

Logarithmically-growing Candida cells were stained with anti-β-1,3-glucan monoclonal antibody (mouse IgG; Biosupplies) followed by anti-mouse IgG Alexa Fluor 488 (Molecular Probes) staining, as described previously (Wartenberg et al., 2014) with some modifications (refer to Supplemental Methods). The level of β-1,3-glucan exposure on the cell surface was then quantified on a MACSquant VYB flow cytometer (Miltenyi).

Dectin-1, Dectin-2, TLR2, and TLR4 signaling measurements were performed using a NF-kB reporter cell line (HEK-Blue™ hDectin-1b, HEK-Blue™ mDectin-2 Cells, HEK-Blue™ mTLR2 cells, HEK-Blue™ mTLR4 cells Invivogen) and its parental cell line (HEK-Blue™-null1-v or HEK-Blue™ null2 cells) as the negative control. Briefly, 1 × 10⁵ cells/well were stimulated with 1 × 10⁶ formaldehyde-fixed Candida cells/well and incubated in 96-well flat bottom plates for 6-20 h. Finally, NF-kB activity was quantified with the QUANTI-Blue (Invivogen) reagent accordingly to the manufacturer's instructions.

All data were analyzed using Excel (Microsoft) and Prism (Graphpad) software. Data is reported as mean ± SD. To compare wild-type C. albicans, isogenic C. albicans cell wall mutants and Candida spp. strains fitter than or as fit as wild-type C. albicans and isogenic C. albicans cell wall mutants and Candida spp. strains less fit than wild-type C. albicans, non-parametric Mann-Whitney test was used. For all other comparisons, statistical significance was calculated using two-tailed unpaired t-tests with Welch's correction. In all analyses, p ≤ 0.05 were regarded as statistically significant.

To assess the contribution of each cell wall component to GI tract fitness, we first generated an isogenic set of C. albicans mutants in the prototroph and otherwise wild-type reference strain SC5314, each deleted in one or more genes illustrated in Figure 1A. In particular, we generated homozygous mutants in CHS3, PMR1, MNT1, OCH1, and MNN4 using the SAT-Flipper technique, which allows inducible removal of the deletion cassette, leaving no ectopic selection marker in the genome (Sasse and Morschhauser, 2012). Since GSC1 is an essential gene, we successfully generated the corresponding heterozygous but not the homozygous mutant; on the other hand, because MNT1 was shown to be partially redundant with MNT2 (Munro et al., 2005), we also generated a double homozygous strain carrying both gene deletions in homozygosity. Using quantitative staining of cellular DNA content followed by flow cytometry, we confirmed all strains in the collection to be diploid (Figure 1B). Whole-genome sequencing further indicated absence of either whole-chromosome or segmental aneuploidy and presence of only a few point mutations, most of which were heterozygous and either non-coding or synonymous (Figure 1C, Supplementary Table 4). The mnt1/mnt1;mnt2/mnt2 strain was the only strain containing a non-synonymous mutation in homozygosity and was therefore reconstructed. The reconstructed mnt1/mnt1;mnt2/mnt2 strain was then re-sequenced and found not to carry this or other homozygous non-synonymous mutation (Supplementary Table 4). Experiments critical for this study have been repeated in parallel using both mnt1/mnt1;mnt2/mnt2 strains and found not to lead to significantly different outcomes (Supplementary Figure 1). This collection of C. albicans cell wall mutants therefore represents a carefully constructed isogenic set of strains that is well suited for quantitative phenotypic analyses without potential confounders due to undocumented genetic or karyotypic differences. Moreover, since these strains were generated without using any auxotrophic marker, they are expected to show no intrinsic deficiency in their ability to grow in nutrient-limited environments, e.g., during colonization or infection of certain host tissues, and should therefore represent an invaluable resource to study the role of the C. albicans cell wall during interaction with the host.

We next adopted a previously proposed in vivo competition assay aimed at comparing the ability of different C. albicans strains to colonize the mouse GI tract (Prieto et al., 2014) and extended it into a quantitative in vivo competitive fitness assay. Briefly, the original method entailed orally gavaging antibiotic-treated mice with equimolar mixtures of a test strain and a fluorescently tagged reference strain, followed by counting fluorescent and non-fluorescent colonies in plated serial dilutions of stool samples collected at specific time points from the double-colonized mice. As expected, competition of wild-type SC5314 cells against the fluorescently tagged version of the SC5314 strain led to no significant change in the relative proportion of fluorescent colonies in the mouse stool over time, indicating that presence of the fluorescence cassette was associated with no or minimal fitness cost in the mouse GI tract (Figure 2A). In contrast, our collection of isogenic cell wall mutants displayed a variety of behaviors in this assay, ranging from undetectable to significant decrease in relative frequency depending on the genotype (Figures 2B–E). In particular, while the gsc1/GSC1 and the mnn4/mnn4 mutants were essentially undistinguishable from the wild-type strain, the mnt1/mnt1 strain showed a consistent and progressive drop in relative frequency during the competition assay, reaching ~20% after ~60 h. The mnt1/mnt1; mnt2/mnt2 mutant, however, was completely outcompeted by the fluorescent strain in <40 h, suggesting that MNT1 and MNT2 are not completely redundant with each other for GI tract fitness. Moreover, while chs3/chs3 and pmr1/pmr1 mutants were outcompeted by the reference strain in a similar amount of time as the mnt1/mnt1;mnt2/mnt2 strain, the och1/och1 strain was outcompeted in ~10 h.

We then developed a regression method that allows us to rigorously quantitate the relative competitive fitness of each test strain using a time series of relative frequencies of fluorescent colonies as an input (see Supplemental Methods for details). By this analysis, the gsc1/GSC1 and mnn4/mnn4 strains were at least as fit as wild-type C. albicans, while all other cell wall mutants in this study displayed a wide range of fitness disadvantages (Figure 2F). In an attempt to correlate GI tract fitness with cell wall composition, we then quantified the total amounts of chitin, β-glucans and mannans in all strains but no significant association was found. In particular, among the strains with significantly reduced GI tract fitness, mnt/mnt1;MNT2/MNT2 and mnt1/mnt1;mnt2/mnt2 showed no significant alterations in crude cell wall composition; on the other hand, chs3/chs3 had significantly reduced chitin and β-glucan and significantly increased mannan levels and och1/och1 had significantly increased chitin and significantly decreased mannan levels. Overall, while genetic alterations in cell wall biosynthetic pathways clearly affect the ability of C. albicans to colonize the mouse GI tract, differential fitness in this environment cannot be attributed to relative compositional differences in the three main polysaccharide layers.

In order to extend the generality of our findings beyond C. albicans, we then repeated the in vivo competition assays in distantly related Candida species (Figures 3A–E). Interestingly, while C. krusei and C. parapsilosis were significantly less fit, C. glabrata, C. tropicalis, and C. dubliniensis were significantly fitter than C. albicans in the mouse GI tract (Figure 3F). In accordance with the cell wall mutants' data, compositional analysis of the major cell wall polysaccharides did not yield any significant association with GI tract fitness. In particular, whereas both C. glabrata and C. dubliniensis were significantly fitter than C. albicans in the mouse GI environment, the former had significantly reduced and the latter significantly increased chitin levels in comparison to C. albicans. Moreover, whereas both C. tropicalis and C. krusei showed significantly decreased β-glucan levels, the former was significantly fitter and the latter significantly less fit then C. albicans. Hence, despite the existence of a large variation in both GI tract fitness and cell wall composition within the Candida genus, these two traits appear to be largely independent from each other.

We next sought to identify factors underlying the variation in mouse GI tract fitness across all our C. albicans cell wall mutants and the other Candida species, employing a systematic profiling of a large number of quantitative phenotypic parameters (Supplementary Figure 2). Consistent with above findings, no significant association was found between competitive fitness in the mouse GI tract (Supplementary Figure 3A) and total chitin, mannan, β-glucan, β-1,6-glucan or β-1,3-glucan levels (Supplementary Figures 3B–F). Moreover, no association was seen between in vivo and in vitro fitness (Figure 4A), or between in vivo fitness and ability to grow under a variety of physiological or stressful environmental conditions (Supplementary Figures 3G–R), ruling out simple explanations based on overall fitness effects in some of the tested strains. Secretion of the mouse cathelicidin-related antimicrobial peptide (mCRAMP) in the GI lumen was recently shown to play an important role in controlling C. albicans colonization in the mouse GI tract (Fan et al., 2015); however no significant correlation between mouse GI tract fitness and mCRAMP resistance was observed (Figure 4B). Microbiota-derived weak organic acids, such as acetic and lactic acid, were shown to be fungistatic and proposed to contribute to colonization resistance against C. albicans in the mouse GI tract (Cottier et al., 2015b), however Candida strains and species displaying the highest competitive fitness in the mouse GI tract did not show a significant increase in acetic acid or lactic acid resistance (Figures 4C,D). An important feature of gut commensal microbes is their ability to survive in the presence of bile (Strati et al., 2016), but GI fitness was uncorrelated to resistance to ox bile or bile salts (Figures 4E,F). A trade-off between gut commensalism and systemic virulence was recently proposed (Pande et al., 2013), but no association was found between GI tract fitness and ability to kill mice during systemic infections, as all tested cell wall mutants and Candida species were significantly less virulent that wild-type C. albicans (Figure 4G). Moreover, with the only exception of the och1 mutant, the reduced systemic virulence exhibited by the C. albicans cell wall mutants did not associate with the ability to survive in phagosome-like conditions (low pH, high ROS, high NO) (Figures 4H–J). Indeed, none of the cell wall mutants displayed obvious hyphal morphogenesis defects (Supplementary Figure 4) and their ability to escape from macrophages after phagocytosis associated neither with systemic virulence nor with GI tract fitness (Figure 4K).

Because GI tract fitness was not associated with any plausible microbe-intrinsic factor or phenotype that we tested so far, we reasoned that the ability of fungal cells to colonize the gut environment might be better explained by factors related to host-microbe interactions. We therefore screened our library of C. albicans cell wall mutants and Candida species for their ability to trigger signal transduction events downstream of some of the main PRRs known to be important for fungal cell recognition. To this end, we employed commercially available human cell lines expressing individual PRRs and an inducible NF-κB reporter. While TLR4 signaling was essentially undetectable in response to any tested strain (data not shown), large variation in the ability to trigger TLR2- and Dectin-2-mediated signaling was observed between the Candida strains and species; this however did not associate with differences in GI tract fitness (Figures 5A,B). Interestingly, Candida strains and species displayed a wide range of abilities to signal through the Dectin-1 receptor, and this was significantly associated with their competitive fitness in the mouse GI tract (Figure 5C). This prompted us to reevaluate our initial observation about the lack of association between GI fitness and cell wall composition, and to investigate a potential contribution not of the total amount but of the level of exposure of β-glucan (the primary PAMP for Dectin-1) on the surface of the fungal cells. Consistent with predictions, β-glucan surface exposure was significantly associated with mouse GI tract fitness across the entire range of C. albicans cell wall mutants and Candida species (Figure 5D). Overall these data indicate that the fungal cell wall may play an important role in the colonization of the mammalian GI tract, and that the level of β-glucan exposure on the cell surface correlates with a loss in competitive fitness of fungi in this host environment.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.