Synergistic interactions between Candida albicans and Enterococcus faecalis promote toxin-dependent host cell damage

Significance

The fungus Candida albicans is a pathobiont colonizing mucosal surfaces of healthy individuals. The interaction with bacteria is generally considered to be antagonistic, with bacteria preventing fungal infection by mediating colonization resistance. In contrast, our study shows that interaction with the bacterium Enterococcus faecalis can result in more severe infections. We identified a bacterial toxin mediating synergistic damage, which explains why synergism was observed for some but not all strains of E. faecalis. Our findings might have clinical implications as C. albicans and E. faecalis occur in the same mucosal niches, benefit from antibiotic treatment, and are coisolated in clinical samples.

Abstract

The fungus Candida albicans and the Gram-positive bacterium Enterococcus faecalis share mucosal niches in the human body. As opportunistic pathogens, both are found to expand population size during dysbiosis, and can cause severe systemic infections in susceptible individuals. Here, we show that the presence of C. albicans results in increased host cell damage by E. faecalis. Furthermore, E. faecalis aggravates oropharyngeal candidiasis in mice. Increased damage is mediated by enterococcal cytolysin, and involves both physical interaction and altered glucose availability. Physical interaction promotes accumulation of bacteria on host cells, facilitating contact of cytolysin with host cells. Glucose depletion by the metabolic activity of the fungus sensitized host cells to cytolysin. This work illustrates how a complex interplay between fungi and bacteria can result in detrimental consequences for the host.

Candida albicans is a fungal pathobiont colonizing different mucosal surfaces of the human body. Colonization is asymptomatic in healthy individuals, but translocation across mucosal barriers in hospitalized and/or immunocompromised patients can lead to severe systemic infection (1, 2). Besides immunosuppression, disturbance of the mucosal microbiome by antibiotic treatment is a major risk factor for disseminated candidiasis (2), suggesting that fungal–bacterial interactions significantly influence the behavior of C. albicans in the host. While most research has focused on antagonistic effects of bacteria on C. albicans virulence or colonization, synergism has been observed for a few bacterial species, including Enterococcus faecalis (3–6).

Enterococci are lactic-acid producing Gram-positive bacteria ubiquitous in nature, and colonizers of animals, including humans (7, 8). Like C. albicans, they are opportunistic pathogens that can cause various infections ranging from urinary tract infections to bacteremia and endocarditis (9, 10). They share various mucosal host niches with C. albicans, like the oral cavity (11), urogenital (12), and the intestinal tract (8). In the gut and the oral cavity, the presence of C. albicans increases the relative abundance of Enterococcus spp. after perturbation of the mucosal microbiota by antibiotics (4, 5, 13, 14) or chemotherapy (15). Furthermore, Candida and enterococci are commonly coisolated from clinical samples, including blood cultures (16–18).

While these observations suggest synergistic interactions between C. albicans and enterococci, with possible clinical consequences, E. faecalis has also been shown to reduce C. albicans virulence by secretion of the bacteriocin EntV. Therefore, it was postulated that E. faecalis promotes a commensal lifestyle of C. albicans by inhibiting fungal filamentation, which in turn would benefit E. faecalis by synergistic interactions (19). Contrary to this hypothesis, however, interaction of E. faecalis and C. albicans on an oral organoid model led to increased tissue erosion and microbial invasion (20). In this model, virulence-associated E. faecalis genes were downregulated, whereas fungal virulence-associated genes were upregulated.

One limitation of the previous studies on C. albicans–E. faecalis interactions is that only few strains were used. Enterococci are known for high genome plasticity, with a small core genome and a large accessory genome contributing to up to 25% of the total genome size (21). Horizontal gene transfer is common in enterococci, involves different types of mobile genetic elements, and has mainly been studied in the context of antimicrobial resistance (22). In addition, the E. faecalis pathogenicity island can be horizontally transferred (23) and contains variations between strains that modulate virulence (24). Virulence of E. faecalis is polyfactorial, with several factors contributing to virulence and host cell damage (25), including capsule, aggregation substance, pili, adhesins, gelatinase, and cytolysin. Cytolysin is a pore-forming toxin consisting of two peptides and causing hemolysis on horse blood agar plates (HBA) and host cell damage (26). Furthermore, it is associated with increased virulence in infection models and patient mortality (27–31). Enrichment of virulence genes, including the cyl operon encoding cytolysin, in strains from clinical infections compared to strains isolated from other sources has been observed in some, but not all studies (32, 33).

To address the possible impact of strain-to-strain variation of E. faecalis on interaction with C. albicans, we screened a larger number of strains in an in vitro intestinal model. We found that the type of bacterial–fungal interaction depends on strain-specific properties of the bacterial partner. Specifically, synergistic host cell damage depended on the E. faecalis cytolysin. We confirmed the in vivo relevance of these findings by using a murine model of oral candidiasis. In this model, cytolysin-producing E. faecalis aggravated invasion and damage, while a cytolysin-negative EntV-producing strain reduced invasion. The presence of C. albicans increased bacterial virulence partly by increasing E. faecalis accumulation and aggregation on host cells. In addition, fungal-mediated nutrient depletion sensitized host cells to pore-forming toxins in vitro. Thus, multiple mechanisms contribute to synergistic damage caused by C. albicans–E. faecalis coinfection.

Results

Host Cell Damage After Coinfection of an In Vitro Intestinal Model with C. albicans and E. faecalis Is Bacterial Strain-Dependent.

Both antagonistic and synergistic interactions between C. albicans and E. faecalis have been reported (5, 15–17, 34, 35). To determine the type of interaction on enterocytes, we used an in vitro cell culture model to quantify host cell damage following either single infection with C. albicans or enterococci, and damage after coinfection. The E. faecalis type strain, the commonly used strains ATCC 29212, OG1RF, and V583, as well as 16 clinical isolates obtained from blood, stool, or urine were analyzed to address possible variations between bacterial strains. While for most of the tested E. faecalis isolates the coinfection damage was comparable to or slightly lower than the sum of damage caused by fungus and bacteria alone, five E. faecalis strains displayed significantly higher coinfection damage under ambient oxygen (21%), reduced oxygen (10%), and low but physiological gut oxygen tension (1% O₂; Fig. 1 A–C) (36). By quantification of bacterial and fungal burden, we could exclude that enhanced host cell damage was due to increased proliferation of microbes during coinfection (SI Appendix, Fig. S1).

Fig. 1.

Strain-dependent synergy during Candida-Enterococcus coinfection of enterocytes results in increased host cell damage. Enterocytes were infected with 10⁶ cells/mL C. albicans for 6 h followed by addition of 10⁵ cells/mL E. faecalis and further incubation for 24 h. Damage was measured by release of host cell lactate dehydrogenase (LDH) into the supernatant, normalized to noninfected controls, and is displayed as percent damage relative to Triton X-100-lysis. Green: monoinfection with C. albicans; dark purple: monoinfection with E. faecalis; orange: sum of single infections; light pink: coinfection. Bars represent mean and SD, n = 3. The sum of LDH release caused by bacterial and fungal monoinfection was compared to the LDH release following coinfection using two-tailed unpaired Student’s t tests. Significant values are indicated by P-values. ns: not significant (P > 0.05). (A) Damage after infection with C. albicans SC5314 and different E. faecalis strains at 21 % O₂,5 % CO₂. (B) Damage at 1 % O_2, 5 % CO₂. (C) Damage at 10 % O_2, 5 % CO₂. (D and E) Damage after infection with E. faecalis B1 and different C. albicans strains (D) or other yeast species (E). WT: Parental strain of the respective deletion mutants (SC5314 for eed1Δ and efg1Δcph1Δ; BWP17 for ece1Δ).

To determine whether the overall higher microbial burden during coinfection led to increased host cell damage, we tested whether the number of bacterial and fungal cells used for infection affects LDH release from host cells. Increasing the bacterial inoculum over a range of 4 log₁₀ led to only a twofold increase in monoinfection damage, and coinfection damage was similar within a range of 10⁴ to 10⁶ bacteria per mL (SI Appendix, Fig. S2A). Similarly, C. albicans-induced damage was relatively stable in monoinfections; however, coinfection damage was concentration-dependent (SI Appendix, Fig. S2B). One possible explanation for the lack of a clear dose-dependent increase of enterocyte damage in monoinfections is that bacterial growth in the system is limited, as suggested by the lack of further increase of bacterial CFU from 24 to 30 h shown in SI Appendix, Fig. S1. For C. albicans, high numbers of fungal filaments might form multiple layers without increased numbers of fungal filaments accessing and damaging enterocytes. This indicated that synergistic host cell damage could not be explained by the higher microbial burden, but suggested that fungal activities were required. Consistent with this, heat-killed C. albicans did not induce synergistic host cell damage during coinfection (SI Appendix, Fig. S2C). Likewise, bacterial viability was required for inducing increased damage (SI Appendix, Fig. S2D), indicating that both microbes contribute to the synergistic virulence phenotype. Furthermore, synergistic damage developed within the last 6 h of the coincubation period, in which bacterial CFU exceeded fungal CFU > 100fold (SI Appendix, Fig. S1) and damage caused by monoinfections remained stable (SI Appendix, Fig. S2E).

C. albicans Virulence Factors Contribute to Damage During Coinfection but Are Not Required for Synergism.

In the experiments presented above, enterocytes were infected with C. albicans first, and bacteria were added after 6 h to prevent a possible negative impact of fast bacterial replication on fungal growth or germ tube induction at early stages of infection. We next tested whether this preincubation was required for synergistic damage and observed that synergism became evident and more pronounced with prolonged preincubation of host cells with C. albicans (SI Appendix, Fig. S2F).

C. albicans rapidly forms germ tubes and hyphae in cell culture conditions, and thereby could contribute to synergistic damage. To test this, we used C. albicans mutants with defects in filamentation (eed1Δ/eed1Δ and efg1Δcph1Δ/efg1Δcph1Δ). As expected, these mutants were strongly reduced in their ability to damage host cells during monoinfection (Fig. 1D). These strains were still able to induce synergistic damage during coinfection. Synergistic damage was also observed with a mutant lacking ECE1 encoding the fungal peptide toxin candidalysin, and the commensal C. albicans strain 101 (Fig. 1D). Overall coinfection damage was lower for the C. albicans mutants and 101 compared to the virulent wildtype strains; this is likely due to the lower or absent fungal-mediated damage in coinfection with these strains. To determine whether the synergistic effect on coinfection damage is limited to C. albicans, we tested additional fungal species: Saccharomyces cerevisiae and Candida glabrata, but not Candida parapsilosis, induced increased coinfection damage (Fig. 1E) despite their lack of cytotoxicity in monoinfections. Taken together, these findings suggested that the increased host cell damage during coinfection was mainly caused by bacterial activities.

Enterococcal Cytolysin Is Necessary for Enhanced Coinfection Damage of Enterocytes.

We hypothesized that synergism with fungal infection is mediated by a distinct virulence factor, or set of factors, found in all synergistic strains, but absent in nonsynergistic isolates. Therefore, we screened the genomes of the E. faecalis isolates used for the presence of virulence-associated genes. Synergistic strains belong to different capsule operon types, but share several virulence-associated genes, e.g. encoding aggregation substance, certain pili, adhesins, and gelatinase (SI Appendix, Table S1). However, this set of genes was also present in nonsynergistic strains, including OG1RF (SI Appendix, Table S2). A notable exception was the cytolysin operon (SI Appendix, Table S3). All synergistic strains displayed β-hemolysis on HBA, and carried the complete cytolysin operon (cyl; SI Appendix, Table S3). The two strains that appear to contain the full cyl operon but displayed no synergism in coinfections, strain B3 and B8, showed no β-hemolysis on HBA at high oxygen, and only B3 displayed weak hemolysis at low oxygen (SI Appendix, Table S3), indicating possible defects in cytolysin regulation or expression.

The cyl operon is often located on self-transmissible plasmids (31), and sequencing results indicated the presence of a 64 kb plasmid containing the cyl operon in the synergistic strain B1 (SI Appendix, Fig. S3). To explore the possible role of cytolysin for synergistic damage, we therefore performed plasmid curing of strain B1 and selected hemolytic and nonhemolytic derivative strains for coinfection experiments. The loss of hemolytic activity was associated with the loss of increased coinfection damage (Fig. 2A). Conversely, transmission of hemolytic activity from B1 to the nonsynergistic plasmid-free strain OG1RF by conjugation resulted in increased host cell damage during coinfection (Fig. 2B and SI Appendix, Fig. S2 G and H). This suggested the presence of a conjugative plasmid in B1, and a link between hemolysis and synergistic damage. However, we cannot exclude that some of the hemolytic colonies obtained in the conjugation experiment were spontaneous rifampicin-resistant mutants of the donor strain. Furthermore, other factors encoded on the putative plasmid could have contributed to the phenotype. We therefore analyzed isogenic mutants of E. faecalis OG1X (29, 37–39) harboring plasmids encoding the cytolysin operon and/or the gene encoding for the adhesin aggregation substance (AS) (40), which increases adhesion and internalization of E. faecalis by colonic enterocytes (41). Increased damage of host cells during coinfection was consistently observed for cytolysin positive strains (Fig. 2C), while the presence of AS had no effect on damage. Thus, enterococcal cytolysin is required for the enhanced damage during coinfection, and the presence or absence of the cyl operon explains the differences in coinfection damage observed for the different E. faecalis strains.

Fig. 2.

Synergistic coinfection damage is mediated by the E. faecalis cytolysin. Enterocytes were infected with 10⁶ cells/mL C. albicans SC5314 for 6 h at 1 % O₂ followed by addition of 10⁵ cells/mL E. faecalis and further incubation for 24 h. Damage was measured by release of host cell lactate dehydrogenase (LDH) into the supernatant, normalized to noninfected controls, and is displayed as percent damage relative to Triton X-100-lysis Green: monoinfection with C. albicans; dark purple: monoinfection with E. faecalis; orange: sum of single infections; light pink: coinfection. Bars represent mean and SD, n = 3. The sum of LDH release caused by bacterial and fungal monoinfection was compared to the LDH release following coinfection using two-tailed unpaired Student’s t tests. Significant values are indicated by P-values. ns: not significant (P > 0.05). (A) Loss of hemolytic activity after plasmid curing of E. faecalis B1 coincides with loss of synergistic coinfection damage. Cyl-: nonhemolytic E. faecalis B1 obtained by plasmid curing; Cyl+: E. faecalis B1 isolated obtained from the same plasmid curing approach which retained hemolytic activity. (B) Transfer of hemolytic activity from E. faecalis B1 (donor) to E. faecalis OG1RF (recipient) by conjugation results in synergistic coinfection damage. Cyl+: hemolytic OG1RF derivative; Cyl-: nonhemolytic OG1RF derivative obtained from the same conjugation experiment. (C) Coinfection with OG1X (plasmid free) and OG1X derivatives harboring the following plasmids: pAM714: cytolysin-positive/agglutination substance-positive; pAM944: cytolysin-positive/agglutination substance-negative; pAM9058: cytolysin-negative/agglutination substance-positive) and ATCC 29212 (hemolytic) as control. Synergistic coinfection coincides with a functional cytolysin operon.

The Effect of E. faecalis on the Severity of Murine Oropharyngeal Candidiasis Depends on the Bacterial Cytolysin.

C. albicans and E. faecalis do not only share a niche in the gastrointestinal tract but also in the oral cavity. We therefore quantified host cell damage during mono- and coinfection of oral buccal cells (TR146). For these experiments, we used an MOI of 1 to achieve a similar level of damage with C. albicans alone as observed with enterocytes. Similar to intestinal cells, coinfection damage was increased with cytolysin positive strains of E. faecalis (Fig. 3A). To determine whether these strain-dependent effects can be translated to in vivo infections, we used a murine oropharyngeal candidiasis (OPC) model in which the drinking water was supplemented with different E. faecalis strains after C. albicans infection. Since we observed apoptotic cell death following infection in vitro both for intestinal and epithelial cells (SI Appendix, Fig. S4), apoptosis was included as read out for in vivo invasion in addition to invasion depth; both parameters were significantly correlated (Fig. 3B).

Fig. 3.

Damage of oral cells in vitro and murine oropharyngeal candidiasis in vivo are aggravated by coinfection with cytolysin-producing E. faecalis strains. (A) Human oral buccal cells (TR146) were infected with 10⁵ cells/mL C. albicans SC5314 for 6 h at 1 % O₂ followed by addition of 10⁵ cells/mL E. faecalis and further incubation for 24 h. Damage was measured by release of host cell lactate dehydrogenase (LDH) into the supernatant, normalized to noninfected controls, and is displayed as percent damage relative to Triton X-100-lysis. Green: monoinfection with C. albicans; dark purple: monoinfection with E. faecalis; orange: sum of monoinfections; light pink: coinfection. Bars represent mean and SD, n = 3. The sum of LDH release/apoptosis caused by bacterial and fungal monoinfection was compared to coinfection using two-tailed unpaired Student’s t tests. Significant values are indicated by P-values. ns: not significant (P > 0.05). (B–I) 6 to 7 wk old male corticosteroid-treated C57BL/6 J mice were sublingually inoculated with C. albicans SC5314 for 75 min. 6 h later E. faecalis was added to the drinking water for 24 h (2 × 10⁷ CFUs in 50 mL). (B) Correlation of apoptotic area and invasion depth. Mean values of all experimental groups (C–H) were analyzed by Pearson correlation. (C–E) Coinfection with E. faecalis ATCC 29212 and OG1RF, respectively, n = 5 mice per group. (C) Representative histographs of PAS-stained tongue sections, (Scale bar, 100 µm.) For each animal (n = 5 mice per group), three nonconsecutive sections per tongue (sections from different parts of the tongue) were randomly selected and stained. (D, F, and H) Invasion depth was measured for each lesion visible on histological sections. Three nonconsecutive section per mouse were analyzed. (E, G, and I) Size of apoptotic area was measured for each lesion visible on histological sections. Three nonconsecutive section per mouse were analyzed. (D and E) 5 mice/group. (F and G) Coinfection with different E. faecalis strains. Data from two independent experiments, n = 5 mice/group in total. Pink/purple shades: cytolysin positive; green shades: cytolysin negative. (H and I) Coinfection with E. faecalis OG1X pAM714 (cytolysin positive: Cyl+) and pAM9058 (cytolysin negative: Cyl−), respectively. Data from two independent experiments, n = 6 mice/group in total. (D–I) Data was analyzed by 1-way ANOVA with Tukey’s multiple comparisons test to compare all groups (D–G) or Dunnett’s multiple comparisons test to compare coinfection with C. albicans only (H and I). P-values < 0.05 are indicated in the graphs.

Application of hemolytic E. faecalis alone, in the absence of C. albicans infection, did not cause any histological alterations (Fig. 3C). The consequences of E. faecalis on OPC were strain-dependent: The cytolysin-negative strain OG1RF, which was previously shown to reduce C. albicans virulence in a Caenorhabditis elegans model by production of EntV (34, 35), led to a significant reduction of the depth of fungal invasion (Fig. 3D) and areas of apoptosis (Fig. 3E). A minor, but statistically significant damage-reducing effect of EntV was also observed in vitro (SI Appendix, Fig. S5). In contrast, the hemolytic strain ATCC 29212 led to more extensive lesions (Fig. 3 D and E and SI Appendix, Fig. S6A). Increased OPC severity correlating with the presence of the cyl operon was also observed in other E. faecalis strains (Fig. 3 F and G). Importantly, comparison of two isogenic E. faecalis OG1X derivatives showed enhanced tissue invasion and apoptosis only for the cytolysin-positive strain (Fig. 3 H and I and SI Appendix, Fig. S6B).

Plating of homogenized tongue tissue from uninfected and infected mice confirmed the absence of endogenous C. albicans strains (SI Appendix, Fig. S7 A and B). Quantification of fungal burden showed a significant but minor increase in C. albicans CFU in coinfection with 2/10 strains (SI Appendix, Fig. S7); both E. faecalis strains were cytolysin positive. However, no increased fungal burden was observed with the other four hemolytic E. faecalis strains tested, three of which led to significantly increased tissue damage. Thus, increased fungal burden might contribute to the augmented severity of OPC for some but not all strains. The effect of C. albicans on E. faecalis CFU was variable; a significant positive effect was observed for E. faecalis ATCC 29212 and OG1RF (SI Appendix, Fig. S7A), but not for E. faecalis OG1X (SI Appendix, Fig. S7B). Importantly, within experiments bacterial burden did not differ between cytolysin-positive and -negative strains in either the presence or absence of C. albicans (SI Appendix, Fig. S7 A–C).

Physical Interaction with Fungi Leads to Accumulation of Bacteria on Host Cells.

While soluble factors contributed to increased damage of coinfections, physical contact could also be involved. To test this, we separated microbes from host cells and/or each other by using cell culture inserts. Separation of both bacteria and fungi from enterocytes completely abolished host cell damage (Fig. 4A, orange bar), consistent with the requirement of adhesion and invasion for damage caused by C. albicans (42, 43), and the lack of solubility of E. faecalis cytolysin in most liquid media (44, 45). The presence of E. faecalis in the inserts did not affect host cell damage when C. albicans was in contact with enterocytes (Fig. 4A, green/turquoise bars). In contrast, a significant increase of damage by E. faecalis was observed if C. albicans was applied to the insert (Fig. 4A, purple bars), confirming the role of soluble factors in enterocyte damage. However, the damage in this setting, as well as the sum of the damage caused by either microbe on enterocytes when physically separated from the other, was significantly lower than that observed if both microbes were in physical contact with each other and the host cells (Fig. 4A, pink bar), indicating a significant contribution of contact-mediated mechanisms. Similarly, the presence of S. cerevisiae in the transwell insert significantly increased enterococcal host cell damage, but damage was lower than that caused by coinfection (Fig. 4 A, Right). Independent of physical contact, the presence of C. parapsilosis did not affect enterocyte damage (Fig. 4 A, Right). One possibility for the higher damage caused by coinfection is that adhesion of bacteria to fungal elements, as observed by electron microscopy (Fig. 4B), increases the contact of bacteria with host cells. To test this, we carefully removed the supernatant after infection and plated the supernatant and lysed host cell layer separately. While the majority of bacteria were found in the supernatant of monoinfections, significantly more bacteria were found on the host cell layer than in the supernatant after coinfection with C. albicans, reflecting a significant increase of bacteria on enterocytes (Fig. 4C). To test whether this effect is specific for C. albicans, we also analyzed S. cerevisiae, which showed synergistic coinfection damage, and C. parapsilosis, which did not. C. parapsilosis had no effect on bacterial distribution (Fig. 4C), significantly more bacteria were isolated from supernatant than enterocytes. This is consistent with the failure of this fungus to induce synergistic coinfection damage. S. cerevisiae displayed intermediate results, bacterial numbers on enterocytes were increased (Fig. 4C), but the increase was not statistically significant. This could be due to the significantly weaker adhesion of S. cerevisiae to enterocytes compared to C. albicans (Fig. 4D). This difference was not only observed for enterocytes but also with the plastic surface of cell culture plates (Fig. 4E). Adhesion of S. cerevisiae and C. parapsilosis was comparable (Fig. 4 D and E), suggesting differences in the interaction with E. faecalis might mediate the differences in bacterial location during coinfection observed for these species (Fig. 4C).

Fig. 4.

Fungal–bacterial contact leads to accumulation of bacteria at the host cell layer. (A) Mono- and coinfection of enterocytes with C. albicans SC5314 and E. faecalis ATCC 29212 (Left) or E. faecalis ATCC 29212 and S. cerevisiae and C. parapsilosis, respectively (Right) using transwell inserts with 0.4 µm pores size for spatial separation. Enterocytes were infected with 10⁶ fungal cells/mL for 6 h, followed by addition of 10⁵ cells/mL E. faecalis and further incubation for 24 h. Damage was measured by release of host cell lactate dehydrogenase (LDH) into the supernatant, normalized to noninfected controls, and is displayed as percent damage relative to Triton X-100-lysis. The scheme below depicts the location of fungal/bacterial cells in the well or insert, respectively. Data are shown as mean with SD, n = 3. Statistical significance was analyzed for C. albicans by two-tailored unpaired Student’s t test with P-values indicated in the graph (ns = not significant; P > 0.05). For S. cerevisiae and C. parapsilosis, damage caused by coinfection was compared to E. faecalis on enterocytes with fungi in the insert by 1-way ANOVA followed by Dunnett’s multiple comparisons test; significant differences are indicated by the P-value in the graph. For S. cerevisiae, coinfection damage and damage by E. faecalis on enterocytes with fungi in the insert was analyzed by the ratio paired t test due to variations of coinfection damage between experiments. (B) Scanning electron micrographs, representative images from two independent experiments. The images were colored manually; turquoise: fungi, fuchsia: bacteria. Without host cells: E. faecalis and C. albicans coculture in KBM for 24 h. Coinfection: Enterocyte coinfection with E. faecalis ATCC 29212 and C. albicans SC5314 (24 h). (C) Ratio of bacterial cells recovered from the supernatant or the host cell layer after monoinfection with E. faecalis ATCC 29212 or coinfection with C. albicans SC5314, S. cerevisiae, or C. parapsilosis. Data are shown as mean with SD, n = 3, and was analyzed by comparing coinfection data for supernatant and cell layer, respectively, to monoinfection by 1-way ANOVA followed by Dunnett’s multiple comparisons test. Statistical significance (P > 0.05) is indicated by p-values in the graph. (D and E) Ratio of fungal cells recovered from the supernatant and the host cell layer (D) or plastic surface (E) after 10⁶ fungal cells/mL were incubated for 30 h. Data are shown as mean with SD, n = 3, and was analyzed by 1-way ANOVA followed by Dunnett’s multiple comparisons test comparing C. albicans to S. cerevisiae and C. parapsilosis, respectively. P-values are indicated in the graph.

Candida-Induced Glucose Starvation Sensitizes Enterocytes to Bacterial Damage and Pore-Forming Toxin.

To test whether soluble factors produced by or changes in media composition due to the metabolic activity of C. albicans affect bacterial virulence, we generated Candida-conditioned medium (CCM) by growing fungi either alone (CCM-) or on enterocytes (CCM+) in KBM medium for 30 h. These media were then used for infection of enterocytes with E. faecalis. Both media led to significantly increased host cell damage during E. faecalis monoinfection (Fig. 5A), which was more pronounced than the effect of C. albicans in inserts during transwell experiments (Fig. 4A). Therefore, we hypothesized that CCM either contained specific molecules produced by C. albicans that cause damage or that nutrient depletion due to the metabolic activity of the fungus affected bacterial virulence. Glucose in naïve KBM (9 mM) was completely depleted after 12 h incubation of C. albicans, irrespective of the strain used (SI Appendix, Fig. S8A). Of note, in the transwell assay, higher glucose concentrations were observed in the well with E. faecalis compared to the insert with C. albicans (SI Appendix, Fig. S8B). Thus, the stronger effect of CCM compared to transwell experiments could be due to differences in glucose depletion. Furthermore, while naïve KBM contains no lactate, a considerable amount (13 mM) was observed after incubation of enterocytes in medium for the duration of infection experiments. Addition of C. albicans resulted in three-fold lower levels of lactate, likely due to its utilization by the fungus. To determine the effect of lactate on host cell damage caused by E. faecalis, CCM- was spiked with lactate and used for infection of enterocytes with E. faecalis. Lactate addition had no significant effect on E. faecalis mediated damage (SI Appendix, Fig. S8C). In contrast, glucose supplementation reduced bacterial-induced damage to the level observed with fresh media (Fig. 5A). While glucose deprivation led to higher LDH release during bacterial infection, transcription of cylL_L and cylL_S, encoding the two cytolysin subunits, was not altered (Fig. 5B). Similarly, no clear changes in cylL_L and cylL_S were observed for coincubation with C. albicans (SI Appendix, Fig. S9). This suggested that cytolysin expression is not regulated by glucose. However, while both C. albicans and S. cerevisiae led to glucose depletion, C. parapsilosis did not (SI Appendix, Fig. S8D), suggesting a link between glucose and synergistic damage independent of cyl expression. Hypothesizing that glucose affects host cell susceptibility to the toxin, we used melittin, as a second small pore forming toxin found in bee venom (46). Melittin-induced host cell damage was increased in the absence of glucose (Fig. 5C), suggesting that glucose-depletion by C. albicans leads to enhanced effects of bacterial cytolysin by sensitizing enterocytes. Addition of melittin to enterocyte cultures infected with C. albicans, S. cerevisiae, or C. parapsilosis resulted in synergistic damage with C. albicans but not C. parapsilosis, similar to the observations with hemolytic E. faecalis (Fig. 5D). In contrast to coinfection, no synergism was observed with S. cerevisiae. Furthermore, conditioned media of both, C. parapsilosis and S. cerevisiae, increased damage by E. faecalis to the extent observed for C. albicans (SI Appendix, Fig. S8E). Thus, even though glucose depletion affects host cell susceptibility to melittin and bacterial damage, it does not fully explain the synergism during coinfection and the effects of conditioned media.

Fig. 5.

Glucose depletion does not lead to upregulation of cytolysin secretion but sensitizes host cells to damage. (A) C. albicans SC5314 conditioned media (CCM) were prepared by incubating C. albicans SC5314 for 30 h (1% O₂, 5% CO₂) in KBM with (CCM+) or without enterocytes (CCM−). The medium was sterile filtered and used for infection of enterocytes with E. faecalis ATCC 29212. Conditioned media were used directly or supplemented with 9 mM glucose (+gluc). Data (n = 3) was analyzed by 1-way ANOVA with Tukey’s multiple comparisons test to compare all groups; P-values < 0.05 of comparisons within the same type of CCM (±) are indicated in the graphs. Damage was measured by release of host cell lactate dehydrogenase (LDH) into the supernatant, normalized to noninfected controls, and is displayed as percent damage relative to Triton X-100-lysis. (B) Expression of the large (cylLL) and small (cylLS) cytolysin subunits normalized to the housekeeping gene gyrA in E. faecalis ATCC 29212 cultured in CCM and controls for 18 h. Data (n = 3) was analyzed for each gene by 1-way ANOVA with Tukey’s multiple comparisons test to compare all groups. (C) Melittin-induced enterocyte damage (determined as described for (A)) in CCM+ supplemented with or without glucose (PBS as control). n = 3, data are shown as mean with SD, groups were compared with two-tailored unpaired Student’s t test. (D) Effect of addition of melittin on enterocyte damage during infection with C. albicans SC5314, S. cerevisiae, or C. parapsilosis (n = 3 independent experiments). The sum of damage (determined as described for (A)) caused by the fungus and melittin separately (orange) was compared to the damage during coincubation (pink) for each fungal species with two-tailored unpaired Student’s t test.

Discussion

C. albicans and E. faecalis coexist on mucosal surfaces in the human body as part of the normal microbiota. Furthermore, both thrive under dysbiosis, and the presence of C. albicans correlates with higher colonization levels of E. faecalis (4, 5, 14, 15), suggesting synergistic interactions. However, antagonistic effects on virulence mediated by the E. faecalis bacteriocin EntV have also been described (34, 35). This raises the question whether this antagonistic interaction neutralizes the increased risk of infection usually associated with elevated fungal and bacterial colonization. We chose to address this by using an in vitro model that allowed us to assess the consequences of coinfection on enterocyte damage using a collection of reference and laboratory strains, as well as clinical isolates.

This approach revealed that the consequences of coinfection were highly strain-dependent. The majority of E. faecalis strains, including the commonly used laboratory strain OG1RF, led to a neutral outcome, in which the damage after coinfection was similar to the sum of damage inflicted by each microbe alone. A distinct set of strains, however, displayed significantly increased host cell damage after coinfection with C. albicans. One feature of these strains was hemolytic growth on horse blood agar plates, which is typically caused by the E. faecalis cytolysin (47). We confirmed the presence of the cytolysin operon in these strains, with the exception of strain ATCC 19433. This strain furthermore caused hemolysis and detectable enterocyte damage only at high oxygen concentrations, whereas coinfection synergy of the other tested strains was more pronounced at low oxygen. As cytolysin expression has been reported to be higher at low oxygen (48), this together with the absence of a detectable cytolysin operon strongly suggests that ATCC 19433 carries another factor responsible for the observed hemolysis. Two strains, B3 and B8, showed no or only weak hemolytic activity on HBA and no synergistic damage in vitro despite presence of the complete operon in the genome. While not tested in this study, a possible explanation is a lack of cytolysin operon expression in these strains. Importantly, gain or loss of hemolytic activity by plasmid curing, plasmid transfer, and using isogenic mutants correlated with synergistic coinfection damage in different genetic strain backgrounds. This indicates that increased damage caused by coinfections can be attributed to enterococcal cytolysin.

Increased toxin production by bacteria in the presence of C. albicans was demonstrated for Staphylococcus aureus (49), but could not be confirmed in this study for E. faecalis on the transcriptional level. Attempts to detect and quantify the small cytolysin peptides in in vitro experiments by unbiased proteomics were unsuccessful. Thus, while we cannot exclude that coinfection with C. albicans results in increased cytolysin production by E. faecalis, it is likely not the main or only mechanism involved.

An alternative explanation for the increased enterocyte damage upon coinfection is increased local toxin concentration on enterocytes. The importance of localized toxin secretion has recently been demonstrated for the fungal toxin candidalysin that is released in fungal invasion pockets (50, 51). Consistent with this hypothesis, physical contact was necessary for full synergistic damage. Of note, cytolysin expression and the cytolytic activity of the produced toxin molecules is affected by the presence of target cells: In the absence of target cells, the toxin subunits CylL_S and CylL_L form a stable complex lacking cytolytic activity. CylL_L has a higher affinity to lipid membranes than CylL_S, thereby leading to an excess of free CylL_S in the presence of target cells. Free CylL_S exceeding a threshold can bind to the membrane protein CylR1, which modulates the repressive function of CylR1 and CylR2, resulting in expression of the operon (26, 52, 53). As a consequence, cytolysin-positive E. faecalis strains can cause hemolysis on blood agar plates, but fail to lyse erythrocytes in liquid media (54). In addition, quorum-sensing autoinduction depending on cell density drives cytolysin expression and adherence at the infection site promotes the toxic effects of cytolysin (52). The presence of C. albicans resulted in a higher number of bacterial cells in association with enterocytes, and thereby possibly enhanced cytolytic activity.

Binding to fungal cells has also been reported for S. aureus and oral streptococci (55, 56). While S. aureus binds preferably to Als3 expressed on hyphae (56), synergistic damage by E. faecalis also occurred with filament-deficient C. albicans strains and other yeast species, suggesting a different binding mechanism. The synergistic coinfection damage observed with nonfilamentous fungi furthermore implies that pronounced fungal invasion and creation of invasion pockets are not necessary for increased cytolysin-mediated damage, although they did contribute to overall host cell damage. S. cerevisiae, which led to increased damage in E. faecalis coinfection, also tended to increase the number of bacteria on host cells. It should be noted that binding of S. cerevisiae to host cells and the plastic surface was weaker than that of C. albicans. This possibly resulted in the loss of bacteria during washing steps, and consequently underestimation of the number of bacteria in contact with enterocytes during coinfection with S. cerevisiae. Furthermore, coinfection with C. parapsilosis had no effect on bacterial association with the host cell layer and did not result in synergistic damage. As the adhesion of C. parapsilosis and S. cerevisiae to host cells and the plastic surface was comparable, different factors might mediate binding to surfaces and interactions with E. faecalis.

Taken together, our results suggest that physical interactions between fungi and bacteria, resulting in an increased concentration of bacteria on host cells enhance cytolysin-mediated damage. This would be consistent with the enhancing effects that target cells have on cytolysin regulation (26, 52, 53), hemolysis on solid but not in liquid media (54), the short active half-life of cytolysin in solution in the absence of nearby membrane targets (57), and the lack of damage enhancement if bacteria were located on the insert during transwell experiments.

While physical interaction played a prominent role for coinfection damage in vitro, culture-conditioned media also had an effect on enterococcal host cell damage. This implied that consumption of nutrients or release of small molecules by C. albicans could affect bacterial virulence. This might also explain why synergistic damage by C. albicans only occurred if enterocytes were infected with C. albicans first, and why the synergistic effect was more pronounced with increasing preincubation time and fungal cell numbers: Ongoing fungal metabolic activity would lead to increased nutrient depletion or conversion over time. Production of small molecules, including secreted metabolites, might be more pronounced in the absence of bacterial nutrient competition. Here, we focused on glucose consumption by C. albicans as an activity that would lead to depletion of nutrients available to the bacteria. While glucose depletion increased host cell damage by E. faecalis, this was not attributable to increased cytolysin production. Furthermore, the use of KBM medium with a high buffering capacity prevented media acidification during the experiments. Therefore, we considered effects of glucose on host cells. Because recombinant cytolysin is not available, we used melittin as a model pore-forming toxin, and observed that damage by melittin was reduced by glucose supplementation. A protective effect of luminal glucose was also observed for LPS-, and Giardia-induced enterocyte apoptosis in vitro (58, 59), and ischemia/reperfusion-induced enterocyte damage and barrier dysfunction in vivo (60). While the mechanisms by which glucose reduces toxin-mediated damage remain unclear, the negative effects of fungal glucose depletion have been demonstrated for C. albicans and macrophages. Infected macrophages increase glycolysis and rely on glucose for survival. C. albicans rapid glucose uptake depletes glucose available for the macrophage, thereby causing macrophage death (61). This illustrates that nutrient depletion by microbes can affect host cell susceptibility. Glucose might also contribute to osmotic balance. Interestingly, though, the addition of melittin to enterocytes infected with C. albicans, S. cerevisiae, or C. parapsilosis resulted in synergistic damage with C. albicans only, whereas coinfection of E. faecalis also resulted in synergistic damage with S. cerevisiae. This indicates differences between the toxins, and that results obtained with melittin cannot be directly translated to enterococcal cytolysin. Furthermore, infection of enterocytes with E. faecalis in conditioned media from all three fungal species resulted in increased host cell damage. This cannot be attributed to glucose depletion, and for C. parapsilosis is in stark contrast to the lack of synergistic damage in coinfection experiments. It seems possible that various soluble factors influence host cell susceptibility either positively or negatively, and that the ultimate outcome depends on the quality and quantity of different factors.

Cytolysin-dependent synergistic damage in vitro was not restricted to enterocytes but also occurred with oral epithelial cells. Since intestinal colonization of mice with neither C. albicans nor E. faecalis results in detectable epithelial damage (19, 62, 63), we decided to use a murine oropharyngeal candidiasis (OPC) model in which mucosal damage can be more easily measured to test whether cytolysin-dependent synergistic damage occurs in vivo. Furthermore, physical interaction between C. albicans and E. faecalis has been observed in mixed biofilms on the surface of tongues during murine OPC in immunocompromised mice (15). The results confirmed increased host damage by coinfection with C. albicans and cytolysin-producing E. faecalis. However, the magnitude of synergistic interactions observed in vitro did not always match the extent of synergism in vivo. Given that several virulence factors contribute to E. faecalis virulence, it is likely that the genetic background of the bacteria affects the relative contribution of cytolysin to coinfection damage in OPC. However, the comparison of the isogenic OG1X strains that differ only in cytolysin clearly demonstrate the impact of cytolysin. Additionally, the cytolysin-negative strain OG1RF led to reduced fungal invasion depth. This likely reflects EntV-mediated reduction of filamentation and virulence, which was described in other infection models (34, 35), but was not evident in in vitro enterocyte infection. While it appears likely that the physical interactions between C. albicans and E. faecalis in vitro also affect in vivo coinfection, the role of metabolism and soluble factors might differ. Regarding glucose availability, for example, it should be noted that normal blood glucose levels (70 to 100 mg/dL) are only slightly lower than the amount of glucose in cell culture media, and remain relatively constant over time, providing continuous supply for tissues. However, changes in local perfusion due to inflammation might create microenvironments in which glucose concentrations are lower independent of microbial activities, or increased blood flow might compensate for increased glucose consumption. It therefore remains unclear if and to which extent C. albicans’ metabolic activity affects glucose concentrations in vivo.

In summary, this study demonstrates that the outcome of C. albicans–E. faecalis coinfection depends on strain-specific bacterial properties, and we identified cytolysin as the factor aggravating host damage in vitro and in vivo. Therefore, species identification alone might not be sufficient to predict the risk of infections in patients with concurrent bacterial and fungal overgrowth. Furthermore, both physical interactions between the microbes and metabolic activity altering the environment and affecting host cell resilience contribute to increased damage during coinfection.

Materials and Methods

Detailed descriptions of the methods and materials are provided in the SI Appendix.

Strains and Culture Conditions.

All strains used in this study are listed in SI Appendix, Tabled S4 and S5. Fungal overnight cultures were grown in yeast extract peptone dextrose (YPD; 2% glucose, 2% peptone, and 1% yeast extract). Bacterial overnight cultures were grown in liquid Todd-Hewitt-Broth medium (THB). Heat inactivation was performed for 1 h at 65 °C in a water bath.

In Vitro Infection Models.

For enterocyte infection a 7:3 mixture of the two human enterocyte cell lines C2BBe1 and mucus producing HT29-MTX-E12 cells differentiated for 14 d were used. TR146 cells (human oral buccal cells) were grown 48 h. Infection experiments were performed with KBM (Gold Keratinocyte Medium) as medium due to its high buffering capacity. The infectious dose was 1 × 10⁶ fungal cells per mL (MOI = 1) unless stated otherwise. After 6 h (or as indicated) 1 × 10⁵/mL enterococci were added. After an additional 24 h of incubation, samples were analyzed. Due to the fast growth of the bacteria, bacterial CFU exceeded fungal CFU at the end of the experiment (SI Appendix, Fig. 1S). Transwell experiments were performed using 0.4 µm inserts.

Quantification of Host Cell Damage, Microbial Burden, and Glucose and Lactate.

Host cell damage was quantified by measuring lactate dehydrogenase (LDH) release into the supernatant using the Cytotoxicity Detection Kit (Roche). Data were normalized to Triton X-lysed controls. Apoptosis was quantified using the FAM-DEVD-OPH in vitro Apoptosis Detection Reagent (ICT-6356) according to the manufacturer’s protocol.

Fungal burden was quantified by plating serial dilutions on YPD agar containing 50 µg/mL gentamicin and 100 µg/mL doxycycline. For bacteria, samples were plated on LB agar, when necessary supplemented with 50 µg/mL nystatin to suppress fungal growth. Glucose (mmol/l) was measured using a Contour blood sugar meter (Ascensia Diabetes Care, Germany), lactate was quantified by the Institute of Clinical Chemistry and Laboratory Diagnostics, Jena University Hospital, Jena, Germany, using routine methods. Additional details are provided in the SI Appendix.

Generation of C. albicans Conditioned Media.

Enterocytes or empty wells containing KBM only were inoculated as described above with 1 × 10⁶ cells/mL C. albicans in 30 mL and incubated for 30 h at 37 °C with 5% CO2 and 1% O₂. The supernatant was removed, filtered through a 2 µm Filtropur sterile filter (Sarstedt), and stored at −80 °C. Conditioned medium was used either directly or spiked with glucose (Carl Roth) or L-(+)-lactate (Sigma) to a final concentration of 9 mM or 13 mM, respectively.

E. faecalis Plasmid Curing and Conjugation.

Plasmid curing was performed as described by McHugh and Swartz (64), details are provided in the SI Appendix. To facilitate plasmid transfer by conjugation, filter mating was performed as described by Haug et al. (65) with strain B1 as donor and OG1RF as recipient. Details are provided in the SI Appendix.

Bacterial DNA Isolation and Whole Genome Sequencing.

Genomic DNA was isolated using the Qiagen Genomic-tip 100/G or DNeasy Blood & Tissue Kit according to the manual for gram positive bacteria.

DNA was quantified, fragmented, and used to prepare sequencing libraries using the TruSeq Nano kit (Illumina)Sequencing libraries were sequenced on Illumina Nextseq creating 300 bp paired-end reads. Sequencing coverage was 100-fold. Sequencing results were compared to reference sequences for the cytolysin operon (KEGG entry for the Enterococcus faecalis ATCC 29212 (T03320) p1 plasmid (genes DR75_2951 - DR75_2958) and the sequences of other known Enterococcus virulence factors retrieved from the Virulence Factors of Pathogenic Bacteria database (http://www.mgc.ac.cn/VFs/main.htm), to determine their presence in the strains.

Additional sequencing (BioSample accessions SAMN50298714, SAMN50298715) was conducted using the Nanopore sequencing technology (ONT). Bacterial DNA was isolated using The ZymoBIOMICS DNA Miniprep Kit for long read sequencing. 1 µg size-selected DNA was used with the SQK-LSK114 kit (Oxford Nanopore Technologies, Oxford, UK) to prepare the sequencing library. Reads of the samples were basecalled using Dorado (v7.3.11) (Oxford Nanopore Technologies) using the high super accuracy basecalling model. The basecalled reads were assembled and polished using Flye (v2.9.3), Medaka (v1.11.3), and Racon (v1.4.20).

PCR of Enterococcal Genomic DNA.

PCR with Phusion High-Fidelity DNA-Polymerase (ThermoFisher) was performed as stated in the manual with the following settings. 35 cycles, denaturation 10 s, annealing at 55 °C 30 s, elongation 45 s, final extension 10 min. Genomic DNA was diluted 1:100 and used as 1 µL. Final reaction volume 20 µL. All primers used to amplify cytolysin genes are shown in SI Appendix, Table S4.

Quantitative Real-Time PCR.

Details on RNA isolation and generation of cDNA are provided in the SI Appendix, Supplementary Methods. qRT-PCR was conducted in biological and technical triplicates with the Brilliant II SYBR Green QPCR Master Mix (#600828, Agilent) in the Stratagene Mx3005 device (Agilent). 1:10 dilutions of each cDNA sample were used. The primer efficiency was determined with genomic DNA of the E. faecalis strain ATCC 29212 in 1:10 dilutions from 100 µg/mL to 0.1 µg/mL. The respective gene expression was calculated with the Δct method relative to the housekeeping gene gyrA. All primers are listed in SI Appendix, Table S4.

Mouse Model of Oropharyngeal Candidiasis.

A detailed description of the animal experiments is provided in the SI Appendix, Supplementary Methods. Briefly, 6 to 7 wk old male C57BL/6 J mice (Jackson laboratories, USA) were randomly assigned to the infection groups. Oral C. albicans infection was induced as described previously (66–68). 6 h post C. albicans infection E. faecalis was added to the drinking water for 24 h (2 × 10⁷ CFUs in 50 mL).

Mice to be infected with E. faecalis OG1X + pAM714 and OG1X + pAM9058 were started on antibiotics (gentamicin 15 μg/mL; erythromycin 50 μg/mL) in the drinking water at day -2, and antibiotic supplementation was continued throughout the experiment. Erythromycin was used to ensure plasmid maintenance, gentamicin to suppress endogenous bacteria (OG1X is gentamicin resistant).

The mice were killed 2 d after C. albicans infection, and C. albicans and E. faecalis were quantified by plating serial dilutions or organ homogenates.

For histology and immunohistochemistry, tongues were cut in half, fixed in zinc buffered formalin, and embedded in paraffin. The depth of individual fungal lesions relative to the surface of the tongue (69) were determined using PROGRES GRYPHAX® software version 1.1.8.153 (Jenoptik). Apoptotic cells were stained by immunohistochemistry using the ApopTag in situ apoptosis detection kit (EMD Millipore) as described previously (70), and apoptotic areas were quantified using PROGRES GRYPHAX^® software (Jenoptik). Samples were blinded before evaluation.

Ethics Statement.

All animal work was approved by the Institutional Animal Care and Use Committee (IACUC) of the Lundquist Institute at Harbor-UCLA Medical Center (protocol number 30927-01R).

Statistical Analysis.

Statistical analyses were performed using Graphpad Prism version 9.2.0. The type of statistical analysis used is indicated for each dataset in the figure legend.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

Enterococcus genome sequencing data have been deposited in Zenodo (71, 72). Other study data are included in the article and/or SI Appendix.