Genotypic diversity and unrecognized antifungal resistance among populations of Candida glabrata from positive blood cultures

The longstanding paradigm is that most bloodstream infections (BSIs) are caused by a single organism. We performed whole genome sequencing of five-to-ten strains from blood culture (BC) bottles in each of ten patients with Candida glabrata BSI. We demonstrated that BCs contained mixed populations of clonal but genetically diverse strains. Genetically distinct strains from two patients exhibited phenotypes that were potentially important during BSIs, including differences in susceptibility to antifungal agents and phagocytosis. In both patients, the clinical microbiology lab recovered a fluconazole-susceptible index strain, but we identified mixed fluconazole-susceptible and –resistant populations. Diversity in drug susceptibility was likely clinically relevant, as fluconazole-resistant strains were subsequently recovered by the clinical laboratory during persistent or relapsing infections. In one patient, unrecognized respiration-deficient small colony variants were fluconazole-resistant and significantly attenuated for virulence during murine candidiasis. Our data suggest a new population-based paradigm of C. glabrata genotypic and phenotypic diversity during BSIs.


17
The longstanding paradigm is that most bloodstream infections (BSIs) are caused by a 18 single organism. We performed whole genome sequencing of five-to-ten strains from blood 19 culture (BC) bottles in each of ten patients with Candida glabrata BSI. We demonstrated that BCs 20 contained mixed populations of clonal but genetically diverse strains. Genetically distinct strains 21 from two patients exhibited phenotypes that were potentially important during BSIs, including   and whole or partial chromosomal aneuploidies 7

. 49
The longstanding paradigm is that most BSIs, including candidemia, are caused 50 by a population of genetically identical strains derived from a single organism that passes 51 through a bottleneck ("single organism" or "independent action" hypothesis). 11,12 Standard 52 microbiology laboratory protocol in processing positive clinical cultures is to test single 53 microbial strains from morphologically distinct colonies. Increasingly, WGS data 54 demonstrate that colonization or chronic infections by various bacteria may be caused by 55 a population of strains in which genetic diversity emerges during long-term host 56 (S-DD). Four patients died within 30 days of C. glabrata BSI diagnosis (patients D, K, L, 80 P). 81 82 WGS of C. glabrata strains. We streaked 10 µL aliquots from the initial positive blood 83 culture bottle from each of the 10 patients onto SDA plates, which were incubated at 35ºC for 84 48 hours. For 9 patients, colonies were indistinguishable by morphotypes. We isolated strains 85 from 4 to 9 from randomly selected colonies. In the tenth patient (patient J), colonies were 86 indistinguishable by morphotypes at 48 and 72 hours of incubation. At 84 hours, however, 87 pinpoint colonies were evident, admixed with a greater number of larger colonies [ Figure 1]. 88 For the remainder of the manuscript, we refer the larger and pinpoint colonies as normal 89 (NCV) and small colony variants (SCV), respectively. We isolated 4 NCV and 5 SCV strains 90 for further study. The index J strain grew as a NCV. For each patient, the index strain was 91 labeled as strain #1, and other independently isolated strains were labeled as #2-10 (#2-5 for 92 patient I). In patient J, strains #1-5 and #6-10 were NCV and SCV, respectively, on SDA 93 plates. All BSI strains underwent next generation sequencing (Illumina NextSeq). Index 94 strains also underwent sequencing by Oxford Nanopore (ONT) using MinION showed that PDR1 and CDR1 expression were significantly higher in L4 than in other L 175 strains [ Figure 4]. To show that G346C affects azole MICs, we first deleted PDR1 in 176 strains L4 and BG2, then engineered either a PDR1 G346C mutation or wild-type PDR1 177 in these backgrounds. Azole MICs were higher against the engineered PDR1 G346C 178 strains (256 and 64 µg/mL against L4 and BG2, respectively) than against respective 179 strains with wild-type PDR1 (32 and 16 µg/mL, respectively) [Supplemental Table 3]. L8 180 did not carry a PDR1 mutation, but CDR1 expression was increased by >12-fold 181 compared with the index strain L1. 182 Two of 10 strains from the initial positive blood culture were fluconazole-resistant. 183 Blood cultures ~48 hours later, after institution of fluconazole, were fluconazole-resistant. 184 Resistant strains harbored the PDR1 G346C mutation described above. 185 Patient J. Patient J was treated with caspofungin for C. glabrata BSI, after which 186 he was placed on voriconazole prophylaxis. Forty-seven days after BSI, he developed 187 intra-abdominal infection due to fluconazole-resistant C. glabrata SCV (strain J11). 188 Strains J1-J4 were NCV in size and morphology on YPD and RPMI agar plates, 189 and in growth rates in YPD broth at 37ºC. Strains J6-J10 and J11 were consistent with 190 SCVs (i.e., respiration-deficient Candida petite mutants by a constellation of phenotypes, 191 including defective growth in YPD liquid medium, inability to grow on YP-glycerol medium, 192 and deep-violet color on agar containing eosin Y) [ Figure 5A and 5B]. 19 Strain J5, which 193 was identified as forming NCV at 48 hours on SDA agar, was intermediate to clones J6-194 J10 and J1-J4 in growth rate in liquid YPD medium. By flow cytometry, cells of strain J5 195 were comparable in size to those of strains J6-J10 following growth in liquid YPD, and 196 significantly smaller than those of strains J1-J4 [ Figure 5C]. J5 also resembled SCV 197 strains J6-J10 in inability to grow on YP-glycerol medium, and in deep-violet color on 198 eosin Y and trypan blue indicator plate. Therefore, on balance, strain J5 was most 199 consistent with a SCV. 200 We used GENOME-STRip to estimate mitochondrial genome copy numbers (mtN) 201 of J strains from mapped short reads. Estimated mtN ranged from 1 to 35. Most SCV 202 strains had ≤18 copies (J9, n=1; J5, n=8; J10, n=13; J7 and J8, n=18). In contrast, 203 estimated mtN for strains J1-4 and J6 were 30 and 35, respectively. The average mtN 204 for strains from other patients was 28 (range: 24-31 copies). We further estimated mtN 205 using quantitative PCR of NCV strain J1 and SCV strains J9 and J11. Estimated mtN was 206 significantly higher for J1 than for J9 and J11 (mean 18.4±2.9 vs 0.4±0.1 and 0.3±0.1, 207 respectively; p=0.001, ANOVA). 208 J1 had significantly greater mitochondrial staining with rhodamine-123, as evident 209 by FACS sorting, than did J9 [ Figure 5D]. Electron microscopic images of J1 and J9 210 corroborated striking differences in numbers of mitochondria per-cell [ Figure 5E]. To 211 assess respiration in greater detail, we exposed strains J1 and J9 to various mitochondrial 212 electron transport chain inhibitors. As expected, oxygen consumption by strain J1 was 213 significantly reduced upon exposure to antimycin A, sodium cyanide or salicylhydroxamic 214 acid (SHAM) [ Figure 5F]. In contrast, oxygen consumption by strain J9 was already 215 diminished in the absence of electron transport inhibition, and it was not reduced upon 216 exposure to antimycin A, sodium cyanide or SHAM. 217 In previous reports, respiration-deficient C. albicans petite mutants induced in mice 218 during in vivo passage experiments or resulting from disruption of Mip1 DNA polymerase 219 were relatively resistant to phagocytosis. 20 We assayed phagocytosis and killing of 220 strains J1, J5 and J9 by freshly harvested human neutrophils. SCV strains J5 and J9 were 221 more resistant than J1 to neutrophil phagocytosis ( Studies have shown that dysfunctional mitochondria activate PDR1 and, in turn, 228 CDR1. 21 As expected from these reports, PDR1 and CDR1 were up-regulated by 229 14.6±4.0-fold and 257±11.8-fold, respectively, in strain J9 compared with strain J1 (RT-230 PCR). Along these lines, fluconazole and voriconazole MICs were higher against strains 231 J6-J10 (≥ 64 µg/mL and 16 µg/mL, respectively) than they were against J1-J4 (4-8 µg/mL 232 and 0.5 µg/mL, respectively). Fluconazole and voriconazole MICs against J5 (32 µg/mL 233 and 1 µg/mL, respectively) were intermediate to those against the other strains. 234 Finally, we compared strains J1 and J9 for virulence during hematogenously 235 disseminated infections of mice. None of the mice died at day 21 following lateral tail vein 236 injections of 10 7 CFU of either strain. J9 caused significantly lower tissues burdens than 237 J1 in kidneys and spleens at 1, 3 and 7 days, and in livers at 7 days [ Figure 6] 238

WGS and phenotypic characterization of C. glabrata strain J11. Strain J11 239
was recovered by the clinical microbiology lab from a culture of an intra-abdominal 240 abscess 47 days after the index BSI. J11 was a SCV that demonstrated slow growth in 241 YPD media, inability to grow in YP glycerol, small size on SDA plates, purple colonies on 242 eosin plates, and resistance to fluconazole, voriconazole and posaconazole. J11 243 clustered with other J strains on nuclear genome phylogenetic tree [ Figure 7A]. 244 Therefore, J11 was highly related to J strains recovered from the initial blood culture. On 245 mitochondrial phylogenetic tree, strain J11 clustered with SCV strains, and was closest 246 to strain J8 [ Figure 7B]. Most studies of Candida diversity in patients have characterized longitudinal, rather 281 than contemporaneous strains. WGSs of C. glabrata or C. albicans strains from serial 282 oral, vaginal, blood, stool and respiratory cultures revealed within-patient differences in 283 SNPs, indels, gene CNVs, aneuploidies, and loss of heterozygosity (LOH, for C. 284 albicans) 5,7,[22][23][24][25][26] . Longitudinal emergence of antifungal resistance, often associated with 285 appearance of resistance-conferring mutations, is well-recognized among C. 286 glabrata. 26,27 In contrast, there are few studies of Candida diversity from a single site at 287 a given timepoint. In a Candida auris outbreak investigation, probabilistic analysis of 288 WGS data from 6-12 pooled colonies suggested mixed colonization or disease in ~25% 289 of patients 28 ; it is unclear if heterogenous populations were identified from blood 290 cultures. In another study, WGS of strains from 3 independent C. albicans colonies from 291 oral cultures of healthy volunteers clearly demonstrated that each strain was unique, 292 mostly due to SNPs and short-range LOH 29 . Finally, in a multi-center study, 5.6% and 293 8.1% of positive Candida cultures from blood and any clinical sites, respectively, had 294 antifungal polyresistance, defined as heterogenous susceptibility testing results among 5 295 independent colonies 30 . Such mixed populations were detected in 15.3% of C. glabrata-296 positive clinical samples, a frequency in keeping with our description of unrecognized 297 azole-resistant strains in 2 of 10 patients. Contemporaneous strains in the multicenter 298 study did not differ by multilocus sequence type, but WGS was not performed. 299 C. glabrata and most pathogenic Candida spp. are GI tract commensals. 300 Genotypically and phenotypically diverse bacterial populations are increasingly 301 recognized during colonization and chronic infections of non-sterile sites, including GI 302 tract, lungs, and skin [13][14][15] . Aside from our recent demonstration that positive blood cultures 303 of patients with carbapenem-resistant K. pneumoniae BSIs were comprised of genetically 304 variant, clonal strains that differed in antibiotic susceptibility and virulence 16 , there are 305 scant data on microbial diversity during acute infections of putatively sterile 306 sites. 31 Numbers of nucleotide differences (i.e., SNPs, indels) between 307 contemporaneous C. glabrata BSI strains here (pairwise average per-patient: ~3,500) 308 were broadly comparable to those previously reported among contemporaneous C. 309 albicans from oral cultures (~500-5,100 SNPs) 29 . Mutations in our strains were 310 predominantly synonymous and found in non-coding regions. Similar findings in a 311 previous study of C. albicans passed in vitro and in vivo were felt to reflect purifying 312 selection that limited accumulation of mutations in protein-coding sequences 3 . More 313 recently, however, investigators were surprised to find enrichment of non-synonymous, 314 coding sequence mutations among serial C. glabrata BSI strains 26 . Reasons for 315 discrepancies between studies are unclear, and merit further exploration. 316 In keeping with previous data for C. glabrata, we found that genes encoding 317 adhesins and other cell wall proteins were over-represented as sites of non-synonymous 318 because they were not visualized on blood or SDA agar plates until ≥84 hours after sub-345 culture. Therefore, SCV prevalence may be under-estimated unless culture plates are 346 assiduously monitored for several days after Candida growth is detected. Our finding has 347 clinical significance since patient J subsequently had a relapsing infection due to a 348 fluconazole-resistant SCV strain (J11) that clustered phylogenetically with SCVs from the 349 initial blood culture. 350 Mitochondria possess their own genetic material that evolves independently from 351 the nuclear genome. In certain C. glabrata strains, the mitochondrial genome is hyper-352 diverse compared to the nuclear genome 26 . On the whole, we found that mitochondrial 353 genome phylogeny was less sensitive than WGS phylogeny in identifying within-patient 354 strain diversity. Nevertheless, data for J strains indicate that mitochondrial genome 355 analysis may be a useful complement to WGS phylogeny for C. glabrata with apparent 356 respiration deficiencies, SCV phenotypes, growth defects or decreased antifungal 357 susceptibility. It is unclear what precipitated mitochondrial dysfunction and emergence of 358 SCVs among J strains. Azole exposure has been linked to mitochondrial damage 37-359 41 , and patient J received 28 days of fluconazole prophylaxis prior to BSI. In fact, the 360 relatively few clinical C. glabrata SCV strains reported to date were recovered from azole-361 experienced patients. 38,42 Alternatively, mitochondrial damage could have stemmed from 362 oxidative stress from host phagocytes. 20 Regardless of mechanism, dysfunctional 363 mitochondria activate compensatory responses that confer adaptive advantages with 364 cross-resistance to both azole and phagocytic killing 20,21,35 . Compared to NCV strain J1, 365 SCV strain J9 was relatively resistant to neutrophil phagocytosis and killing, but it was 366 significantly attenuated for virulence in mice with hematogenously disseminated 367 candidiasis. Azole resistance in SCV strains was associated with growth defects under 368 various conditions in absence of drug exposure, which likely offset potential advantages 369 afforded by evasion of phagocytosis. 370 Previous studies of virulence of C. glabrata SCVs have yielded conflicting 371 results. In one report, an SCV generated by ethidium bromide treatment exhibited 372 reduced virulence during murine disseminated candidiasis. 43 In another study, however, 373 a fluconazole-resistant oropharyngeal C. glabrata SCV was more virulent than an 374 antecedent fluconazole-susceptible strain 38 . We previously reported that a C. 375 albicans SCV that emerged after hematogenous passage through mouse organs caused 376 lower mortality and acute tissue burdens during disseminated candidiasis than its parent 377 strain, but it persisted within tissue for a prolonged period 19 . Results are difficult to 378 compare between studies since strains were not isogenic and they were created through 379 different methods. Moreover, virulence is a complex and multi-factorial phenomenon, 380 which can vary based on site of infection and conditions within a given host. For example, 381 SCV strain J9 was less virulent than NCV J1 following intravenous inoculation in our 382 mouse model, but nevertheless SCVs were able to persist in patient J after resolution of 383 candidemia and re-emerge to cause relapsing infection. The concept of virulence is 384 particularly complex for an opportunistic pathogen like C. glabrata that has significant 385 redundancy in virulence determinants, lacks a dominant virulence factor, and often 386 causes diseases in patients with immunodeficiencies and other host defense 387

impairments. 388
A strength of our study design is that C. glabrata were collected as blood cultures 389 were being processed according to standard clinical microbiology lab practices. We 390 believe that genotypic and phenotypic diversity emerged at sites of colonization such as 391 the GI tract prior to bloodstream inoculation. While mutations occurring during blood 392 culture incubation may have also contributed to within-patient differences among strains, 393 these are unlikely to account for the extent of diversity we observed in each of our 10 394 patients. Indeed, incubation of index strain J1 in a sterile blood culture bottle resulted in 395 significantly fewer nucleotide differences among recovered strains than were observed 396 among J1-J10 (p<0.0001), and it did not lead to emergence of SCVs. In a WGS analyses 397 of C. glabrata from longitudinal cultures of various body sites over 0-90 days, investigators 398 previously concluded that inter-strain genetic variation was primarily due to standing, pre-399 existing diversity within the population rather than to accumulation of new mutations 7    Gene deletions/duplications were analyzed using two strategies. The first 486 employed GENOME-Strip 54 and was based on coverage depth in addition to information 487 from split reads, paired reads, and the assembly of reads. In the second strategy, we 488 employed a Nextflow pipeline that performs de novo assembly of long reads from the 489 representative of each patient's strains using FLYE 55 . Short reads were mapped to the 490 representative strain's assembly using BWA, and Pilon 56 was used for error correction 491 and to polish and produce the final assembly. Assembled genomes within each patient 492 were aligned using progressiveMauve 57

. The alignments were visually inspected in 493
Mauve to detect gene deletions/duplications. To estimate mitochondrial copy number for 494 each strain, we used GSUTILS scripts from GENOME-STRIP, which can estimate a 495 gene/genomic region copy number. All programs were run using the University of 496 Pittsburgh Center for Research Computing resources. 497

Reverse transcription-quantitative PCR (qRT-PCR). RNA was harvested from yeast 498 cells in exponential growth in Yeast Extract-Peptone-Dextrose (YPD) broth using the 499
RiboPure RNA purification kit for yeast (Invitrogen), and then treated with DNaseI. cDNA 500 was synthesized using Verso cDNA synthesis kit (Fisher Scientific). Genomic DNA 501 contamination was checked by PCR with primers flanking the intron of C. glabrata ACT1. 502 Primers used are summarized in Supplemental Table 1 43,58,59 . Quantitative PCR with 503 SYBR Green qPCR Master Mix (Fisher Scientific) was performed with 1:10 diluted cDNA. 504 Target gene expression was calculated using the ΔΔCT method, with normalization to the 505 housekeeping genes ACT1 and Cg18S. All experiments were done in independent 506 biological triplicates and are shown as mean with standard deviation (SD) for each time 507

point. 508
For mitochondrial copy number determination, we used quantitative PCR, with 509 COX1 as target for mitochondrial gene and ACT1 for control (Supplemental Table 1  were made in distilled water, which were then cultured on YPD agar plates for colony 568 count enumeration. Percentage of adherence was calculated by dividing colony forming 569

units (CFUs) of adherent cells over CFU of input cells and multiplied times 100. 570
Phagocytosis and killing assays. Phagocytosis and killing assays were 571 performed as previously described with slight modification. 19 Briefly, fresh 572 polymorphonuclear cells (PMNs) were resuspended in RPMI 1640. Opsonized C. 573 glabrata cells (50% normal human serum at 37 0 C for 30 minutes) were incubated with 574   Table 2. Within-patient differences in gene copy number variants. Data were analyzed using Nextflow pipeline. Black and gray boxes represent presence and absence of specific genes, respectively. Deletions of 12 genes conferred withinpatient differences between strains in 4 patients; 8 of these genes encoded adhesins. We did not observe any within-patient gene duplication.

Gene Clones
Gene description Putative adhesin-like cell wall protein (adhesin cluster I)

EPA1
Putative adhesin-like cell wall protein (adhesin cluster I)

EPA5
Putative adhesin-like cell wall protein (adhesin cluster I)

3A.
Phylogenomic tree for all strains, estimated using Maximum Likelihood with RaxML. The tree represents mitochondrial genome nucleotide alignment of the 94 strains, based on variant calling data. We used 100 iteration bootstrapping, and only values higher than 50 are shown. Strains J1-J5 carried fewer mitochondrial nucleotide differences than did J6-J10, and they clustered in a clade with K, P and ST strains, in keeping with nuclear phylogeny. Strains J6-J10 (red asterisks) showed much higher divergence than did J1-J5, and fall outside of the clade that include strains J1-J5.  5C. Cell sizes of J strains, as measured by FACs. Cells (1X10 6 CFU/ml) were grown in 4 ml of YPD broth at 30 o C for 24 hours. Strains J1-J4 and J5-J10 demonstrate larger and smaller cell diameters, respectively.

5D. Flow cytometric analysis of rhodamine 123-stained J1 (upper panel) and J9
(lower panel) strains. Cells (1X10 6 CFU/ml) were grown in 4 ml of YPD broth overnight at 30 o C, then treated with 1 mM sodium azide before rhodamine 123 staining. Unstained cells (a and b) are presented as controls. Stained cells for J1 and J9 cells, before and after treatment with sodium azide, are presented in c and d, and e and f, respectively. Xaxis denotes rhodamine-123 fluorescence intensity. The histogram bins are normalized to peak, and the Y-axis denotes percentage of maximal value. Under routine growth conditions, J1 cells exhibited greater fluorescence than did control J9 cells, consistent with higher mitochondrial activity. Intensity of staining of J1 was reduced by roughly 50% following inhibition of electron transport with 1 mM sodium azide. In contrast, sodium azide did not reduce staining of respiratory-deficient strain J9.

5E. Transmission electron micrographs of strains J1 and J9.
Representative images shown here highlight reduced numbers and aberrant morphology of mitochondria in small colony variant strain J9, compared to normal colony variant J1. Several normal appearing mitochondria are denoted by white asterisks (*). * * * * 5F. Evaluation of respiratory status of strains J1 and J9 using electron transport inhibitors. Cells grown in synthetic completed medium for 48 hours at 30 o C were untreated or exposed to sodium cyanide 100mM, salicylhydroxamic acid (SHAM) 100 mM or antimycin 1 µM. Oxygen consumption was measured as phosphorescent probe signal intensities (y-axis) versus time (x-axis). In absence of drug exposure, oxygen consumption by strain J1 was greater than that by strain J9. Each of the electron transport inhibitors resulted in significant reduction in J1 oxygen consumption, without impacting consumption by respiratory-deficient strain J9.  7A. Radial phylogenetic tree of the nuclear genomes of all strains belonging to ST3 from patients J, K, P and ST. Note that J11 clustered with the 10 BSI strains from patient J (J1-J10). Other strains clustered by patient.
7B. Phylogenetic tree of mitochondrial genome of the 11 J strains. Note the NCV strains (bottom 5 strains, J1-J5) cluster with each other and the SCV strains (top 6 strains, J6-J10 and the J11) cluster with each other. J11 is closest to J8 by mitochondrial genome.