A multidimensional assessment of in-host fitness costs of drug resistance in the opportunistic fungal pathogen Candida glabrata

The global rise of antimicrobial resistance poses a serious threat to public health. Because drug-resistant (DR) pathogens typically carry mutations in genes involved in critical cellular functions, they may be less fit under drug-free conditions than their susceptible counterparts. As such, the limited use of antimicrobial drugs has been proposed as a practical strategy to diminish the prevalence of DR strains. However, in many cases the fitness of DR pathogens under host conditions is unknown. Candida (Nakaseomyces) glabrata is a prevalent opportunistic fungal pathogen notable for its high rate of fluconazole resistance (FLZR), echinocandin resistance (ECR), and multidrug resistance (MDR) relative to other Candida pathogens. Nonetheless, the fitness of C. glabrata MDR isolates is poorly characterized, and studies of FLZR isolate fitness have produced contradictory findings. Two important host niches for C. glabrata are macrophages, in which it can survive and proliferate, and the gut. Herein, by employing a comprehensive collection of clinical and isogenic C. glabrata isolates, we show that FLZR C. glabrata isolates are less fit inside macrophages than susceptible isolates and that this fitness cost is reversed by acquiring ECR mutations in FKS1/2 genes. Interestingly, dual-RNAseq revealed that macrophages infected with DR isolates mount an inflammatory response whereas the intracellular DR cells downregulate processes required for in-host adaptation. Consistently, DR isolates were outcompeted by their susceptible counterparts in the context of gut colonization and in the kidneys of systemically infected mice, whereas they showed comparable fitness in the spleen. Collectively, our study shows that macrophage-rich organs, such as the spleen, favor the retention of DR isolates, potentially reducing the utility of limited antifungal use to decrease the burden of DR C. glabrata in the context of candidemia. Author summary The rise of multidrug resistant (MDR) strains of fungal pathogens, notably Candida glabrata, poses a significant clinical challenge because of the limited number of antifungal drugs available for use. Thus, it is vital to minimize the prevalence of drug resistance in the clinic. Because in some bacterial and fungal species drug resistance is accompanied by a fitness cost, implementation of limited antibiotic or antifungal drug use in the clinic has been suggested as a practical way to favor the spread of susceptible isolates. However, it is not clear whether this strategy can work for MDR C. glabrata, as its fitness costs have not been systematically examined, particularly in the context of the host. Herein, we show that MDR C. glabrata isolates can replicate within macrophages as well as susceptible isolates, and this result was consistent with gene expression changes in the infected macrophages. In animal models, MDR strains were unfit in the context of the gastrointestinal tract and kidney, but their fitness in the spleen was comparable to that of susceptible strains. Accordingly, the potential of limited antifungal use to reduce the prevalence of MDR strains of C. glabrata strongly depends on the host reservoir of infection.


Introduction
Antimicrobial resistance (AMR) is a leading cause of death worldwide and is regarded as one of the most 67 pressing current medical challenges (1). Because antimicrobial drugs typically target cellular processes 68 critical for growth and virulence, mutations that cause drug resistance may attenuate critical enzymes 69 causing diminished growth rates, lower virulence, or reduced transmission. Therefore, in the absence of 70 drug pressure, these mutants may have lower fitness than drug-sensitive strains, resulting in gradual 71 eradication of resistant mutants and dominance of susceptible counterparts. Accordingly, restriction in the 72 use of antimicrobial drugs has been proposed as a practical strategy to decrease AMR rate in clinical 73 settings (2, 3). Thus far, this strategy has led to contradictory findings, with some clinical centers reporting 74 a significant reduction in the rate of AMR (4-7) but others reporting otherwise (8-10). Thus,  Candida (Nakaseomyces) glabrata is a major human fungal pathogen and the second leading cause of 82 candidemia in many countries (11). C. glabrata rapidly develops resistance during antifungal treatment, 83 with numerous studies reporting alarming increases in the prevalence of fluconazole-, echinocandin-, and 84 multidrug-resistance (FLZR, ECR, and MDR, respectively) (12-17). Fluconazole exerts fungistatic 85 activity in Candida and targets Erg11, one of the critical proteins involved in ergosterol biosynthesis, 86 whereas echinocandins (caspofungin, micafungin, and anidulafungin) are fungicidal and act by inhibiting 87 the catalytic subunits of β-1,3-glucan synthase, Fks1 and Fks2. Mechanisms underpinning FLZR in C. 88 glabrata mainly involve gain-of-function (GOF) mutations in the transcription factor Pdr1, which results 89 in overexpression of efflux pumps, whereas ECR is mainly associated with mutations in two hotspot 90 regions (HS1 and HS2) of Fks1 and Fks2 (18). 91 Interestingly, studies have shown that FLZR C. glabrata isolates are more virulent in the context of 92 systemic infections (19), less effectively phagocytosed by macrophages, and more strongly adherent to 93 epithelial cells compared to susceptible isolates (19,20). Additionally, susceptible wild-type and 94 laboratory-generated FLZR Cg isolates induced comparable virulence when tested in Galleria larvae (21). 95 On the other hand, FLZR C. glabrata isolates harboring gain of function (GOF) Pdr1 mutations were 96 found to be more susceptible to oxidative stress associated with innate immune cells, and that incubation 97 in H2O2 resulted in acquisition of secondary suppressor mutations that inactivated Pdr1 activity and 98 restored oxidative stress resistance (22). Accordingly, laboratory generated FLZR C. glabrata isolates 99 were found to be both susceptible to H2O2 and more effectively killed by neutrophils (23). Similarly, 100 FLZR C. lusitaniae isolates harboring GOF mutations in Mrr1 are more susceptible to H2O2 and therefore 101 secondary suppressor mutations arise during the course of infection to enhance fitness by reversing the susceptibility testing (AFST) and sequencing of PDR1 and FKS1 and FKS2 HS regions were performed 147 for all clinical and CBS138-derived isolates (Supplementary Tables 1 and 2). 148 Variation in fitness patterns of drug-resistant C. glabrata strains during in-vitro stress 149 First, we examined the in vitro fitness of all our C. glabrata strains (both clinical and isogenic isolates, n= stress (ER stress) (tunicamycin 5µg/ml), and osmotic stress (0.5M NaCl). YPD broth containing standard glucose concentration (2%) and set at pH7 was used as the stress-free control. To measure growth rates, 156 OD600 of overnight cultures was adjusted at 0.2 in either stress-free or stress-containing medium and 157 changes in OD600 were monitored kinetically over 15 hours.

158
Although there was significant individual variation in growth rates between strains, the following trends 159 were observed. In stress-free YPD, no subset of clinical isolates showed a significant fitness cost, but 160 CBS138-derived, isogenic ECR and susceptible isogenic isolates showed a slightly higher growth rate 161 than FLZR and MDR isolates ( Figures 1A and 1B showed the highest IR rates, whereas the MDR isolates were the least fit inside the macrophages ( Figure   179 1C). For clinical isolates, the IR rates of the MDR isolates were similar to susceptible and ECR isolates, 180 whereas the FLZR isolates had the lowest IR rates ( Figure 1D). This difference between IR rates of clinical  Consistent with a previous study (20), isogenic CBS138-derived FLZR isolates had the lowest 186 phagocytosis rate ( Figure 1E), whereas phagocytosis rates were similar for all clinical isolates regardless 187 of susceptibility profile ( Figure 1F). Of note, the lower IR rate of isogenic FLZR isolates was not simply 188 due to their lower phagocytosis rate because, first, clinical FLZR isolates had phagocytosis rates similar 189 to other isolates and, second, all isolates had very similar CFUs at 3 hpi. Altogether, these observations  vs. R663G (Figure 2A), were introduced into a CBS138-derived FLZR isolate carrying PDR1 G1079E (see 204 methods section). S629P and S663P are the most prevalent and R631G and R665G are the least prevalent 205 mutations found among clinical ECR isolates (14,16,28).

206
The CRISPR-Cas9-generated MDR isolates, the parental FLZR strain, and CBS138 (susceptible wild-207 type control) were used to infect THP1 macrophages, and IR rates were determined at 3, 6, 24, and 48 hpi. showing a higher IR rate compared to their counterparts in FKS1-HS1. Interestingly, mutations with a 212 higher clinical prevalence also had a significantly higher IR rates than the less prevalent mutations at the 213 same locus (i.e., S629P and S663P vs. R631G and R665G) ( Figure 2B). Of all four MDR mutants, MDR-214 FKS2 S663P had the highest IR rate, comparable to that of the susceptible strain, CBS138. growth defect as their parental FLZR petite isolates ( Figure 2C), consistent with the fact that the petite phenotype mainly stems from defective mitochondria and cannot be restored by introducing echinocandin 222 resistance. 223 Overall, these observations suggest that the intracellular growth defect of FLZR isolates carrying GOF 224 mutations in PDR1, but not mitochondrial dysfunction (petite), can be rescued by fks mutations conferring 225 ECR, reinforcing our previous observations that isogenic and clinical MDR strains have higher IR rates 226 than FLZR strains.   249 To better understand the host response toward FLZR, MDR, and ECR isolates and vice versa, we 250 performed a dual-RNAseq analysis to assess pathogen and host transcriptomes. We selected the pan-251 susceptible CBS138 and the FLZR isolates, as well as the MDR-FKS2 S663P and MDR-FKS2 R631G isolates 252 since these fks mutations had the highest and lowest impact on the intra-macrophage fitness of the parental 253 FLZR isolates, respectively. THP1 macrophages were infected with these isolates and RNA samples were 254 isolated at 3 and 24 hpi and analyzed by RNAseq.

255
Transcriptomic responses of C. glabrata isolates to macrophage internalization 256 To get a general overview of the transcriptomics profiles of all studied C. glabrata samples we performed   plot of all studied C. glabrata samples across studied conditions. The plot is based on vst-transformed read count data generated by DESeq2. Labels on the data points correspond to drug susceptibility profiles of each strain: S -Susceptible, FR -Fluconazole Resistant, MDR -Multidrug Resistant. Percentages on PC1 and PC2 axes indicate the total amount of variance described by each axis. B. GO term enrichment analysis (category "Biological Process") of up-regulated genes at a given comparison of C. glabrata strains shown on the X axis. C. GO term enrichment analysis (category "Biological Process") of down-regulated genes at a given comparison of C. glabrata strains shown on the X axis. For (B) and (C) the numbers underneath the comparisons correspond to the "counts" of clusterProfiler (i.e. total number of genes assigned to GO categories). GeneRatio corresponds to the ratio between the number of input genes assigned to a given GO category and "counts". Only significant (padj<0.05) enrichments are shown. Adjustment of pvalues is done by Benjamini-Hochberg procedure.
Transcriptomic responses of macrophages to infection by the different C. glabrata strains 288 We first performed a PCA analysis to obtain an overall view on the transcriptomes of macrophages  marginally increased the defective IR rate of the FLZR strain. In sum, we observed that the more similar 313 the IR rate, the more similar the macrophage transcriptomic responses are.

314
A similar pattern was observed for both up-and down-regulated genes ( Figures 5C and 5D, respectively).

318
In the same comparison, macrophages infected with FLZR and MDR-FKS1 R631G strains predominantly 319 downregulated processes related to cell division, such as chromosome segregation, nuclear division, 320 mitotic cell-cycle checkpoint, among others. Percentages on PC1 and PC2 axes indicate the total amount of variance described by each axis. B. GO term enrichment analysis (category "Biological Process") of up-regulated genes of macrophages infected with C. glabrata strains (as depicted on X axis) compared to uninfected macrophages. C. GO term enrichment categories (when available) of up-regulated genes of macrophages infected with different C. glabrata strains (see the X axis for specific comparisons). D. GO term enrichment categories (when available) of down-regulated genes of macrophages infected with different C. glabrata strains (see the X axis for specific comparisons). For (B), (C) and (D) the numbers underneath the comparisons correspond to the "counts" of clusterProfiler (i.e., total number of genes assigned to GO categories). GeneRatio corresponds to the ratio between the number of input genes assigned to a given GO category and "counts". Only significant (padj<0.05) enrichments are shown. Adjustment of p-values is done by Benjamini-Hochberg procedure. E. Transcripts related to "classically" activated macrophages show significantly enriched expression in the macrophages infected with FLZR and MDR-FKS1 R631G strains. Plots depict enrichment of the "classically" activated macrophage transcriptional module (61) for macrophages infected with the indicated strains at 24hpi. Normalized enrichment score (NES) and adjusted P values are shown in the inset. transcriptional module to be significantly enriched in the FLZR-and MDR-FKS1 R631G -infected 328 macrophages ( Figure 5E). 329 Collectively, major differences observed at 24 hpi are associated with GO terms related to chemotaxis, 330 lipid metabolism, cytokine production, among others, while down-regulated pathways are mainly related 331 to cell division. included GFP-expressing and non-fluorescent isolates. 343 We previously showed that GFP-expressing CBS138 are slightly less fit in the GI tract than their non-344 fluorescent counterparts (https://www.biorxiv.org/content/10.1101/2023.06.15.545195v1). Nonetheless, 345 our GI-tract colonization revealed that a GFP-expressing susceptible isolate readily outcompeted non-346 fluorescent FLZR ( Figure 6A) and MDR-FKS2 S663P isolates ( Figure 6B). Surprisingly, the GI-tract 347 competition also showed that a non-fluorescent FLZR isolate readily outcompeted a GFP-expressing 348 MDR-FKS2 S663P , which had previously shown high fitness inside macrophages ( Figure 6C). Altogether, 349 these results indicated that susceptible isolates readily outcompeted both FLZR and MDR isolates and that 350 FLZR isolates may be more fit than the MDR isolates in the context of the GI tract.  Our in vitro experiments indicated that MDR and susceptible C. glabrata isolates tolerated stresses 390 characteristic of the macrophage/phagosome environment better than FLZR isolates. Consistent with these 391 in vitro data, MDR and susceptible isolates had a significantly higher intra-macrophage growth rate 392 compared to FLZR isolates. Interestingly, MDR isolates carrying FKS1 S629P and FKS2 S663P , which are the 393 most common clinically relevant mutations, had a higher IR rate compared with clinically less prevalent 394 mutations, such as FKS R631G and FKS2 R665G (16, 28). This potentially explains the high prevalence of 395 FKS1 S629P and FKS2 S663P among clinical isolates. Of note, the lower IR rate of FLZR isolates is not simply 396 due to a lower phagocytosis rate, because the number of intracellular C. glabrata cells were similar across 397 all phenotypes at 3 hpi and because phagocytosis rates were similar among these clinical isolates.

398
Although the mechanisms underpinning differential intracellular fitness warrant further investigation, 399 recent studies have shown that the FLZR phenotype in C. glabrata (28) and C. lusitaniae (24) renders 400 such isolates more susceptible to oxidative stresses and that continuous exposure to oxidative stress-  Intriguingly, the spleen appeared to be more permissive for retention and persistence of drug-resistant 421 isolates, suggesting that macrophage-rich organs may serve as a viable reservoir for the emergence,  It should be noted that clinical drug-resistant isolates often acquire secondary mutations, and perhaps 446 epigenetic modifications, to compensate for the fitness cost associated with antifungal resistance (2, 3).

447
Unlike clinical strains, the isogenic drug-resistant C. glabrata isolates tested in this study were generated 448 in the absence of host-relevant constraints. As such, future studies using drug-resistant isolates obtained 449 from various mouse organs and sequential isolates collected from humans will shed additional light on  and plates incubated at 37°C in 5% CO2 for 3 hrs. Three hours pi, all the wells were extensively washed 483 with PBS and fresh RPMI was added. Of note, the MOI of 5/1 was used for the competition assays.

484
To calculate the IR rate, macrophages were lysed with one ml of ice-cold water at 3, 6, 24, and 48 hpi 485 and plated on YPD agar plates. IR rate was calculated by dividing the intracellular CFU over the CFU of 486 the initial inoculum and data were presented as percentage. The RPMI collected at 3hrs was plated on 487 YPD plates using which the phagocytosis rate was determined. Phagocytosis rate was defined as the 488 supernatant CFU over the initial inoculum CFU and the values were subtracted from 100. 489 mutations in HS1-Fks1, S629P and R631G, and 2 mutations in HS1-Fks2, S663P and R665G. The codons 492 underlying these mutations were adopted from previous studies (14,47). To introduce each mutation, we 493 used two overlapping ultramer primers in which the codon the desired mutation was introduced. Of note, 494 the PAM site was also mutated using redundant codon sequences to prevent CRISPR-Cas9 cut of the 495 desired amplicons. For each mutation, we carried out two PCR using forward primer and reverse ultramer 496 primer as well as the forward ultramer primer and the reverse primer (Supplementary Table 4). 497 Subsequently, these two PCR products were fused together using short forward and reverse primers, which 498 were sequenced following PCR product purification. Each fused PCR product should carry two mutations, 499 a non-synonymous mutation conferring ECR and a silent mutation in PAM site.

500
Competent C. glabrata cells were prepared using Frozen-EZ Yeast Transformation Kit (Zymo Research) 501 and transformation followed an electroporation-based protocol described previously (49) and gRNAs 502 listed in Supplementary Table 4. The transformants were incubated in fresh RPMI for 2 hours, followed 503 by spreading them on YPD plates containing 0.125µg/ml of micafungin and incubated for one week in 504 37°C. Positive colonies were subjected to PCR using diagnostic primers (Supplementary Table 4) and 505 subjected to sequencing. ECR colonies should contain the two previously described mutations.

506
Macrophage damage assay 507 To measure the extent of damage incurred by C. glabrata isolates to macrophages, we measured the level 508 of lactate dehydrogenase using a commercial kit (Sigma) (50). Briefly, macrophages infected with the 509 MOI of 5/1 were extensively washed with PBS 3 hpi, followed by adding fresh RPMI and incubation in 510 CO2 incubator at 37°C for another 21hrs. After 24 hours, supernatant samples were collected and LDH 511 was determined as described previously (50). The OD value of each replicate was subtracted from that of 512 the background control (uninfected macrophages) and the corrected value was divided by that of high 513 control (uninfected macrophages treated with 0.25% of Triton X-100 for 3 minutes). The corrected 514 normalized values were presented as percentage.  Macrophages infected with the MOI of 5/1 were extensively washed 3 hpi and fresh RPMI was added to 525 wells to be further incubated at 37°C. After extensive PBS wash at each step, macrophages were subjected 526 to a manual RNA extraction protocol described elsewhere. The RNA samples were treated with RNase free-DNase and further purified using RNeasy kit (QIAGEN) per manufacturer's instruction.   Table. To assess an enrichment of transcripts of the "classically" activated macrophages,