Comparative Genomics of Serial Candida glabrata Isolates and the Rapid Acquisition of Echinocandin Resistance during Therapy

The opportunistic pathogen Candida glabrata shows a concerning increase in drug resistance. Here, we present the analysis of two serial bloodstream isolates, obtained 12 days apart.

ABSTRACT

The opportunistic pathogen Candida glabrata shows a concerning increase in drug resistance. Here, we present the analysis of two serial bloodstream isolates, obtained 12 days apart. Both isolates show pan-azole resistance and echinocandin resistance was acquired during the sampling interval. Genome sequencing identified nine nonsynonymous SNVs between the strains, including a S663P substitution in FKS2 and previously undescribed SNVs in MDE1 and FPR1, offering insight into how C. glabrata acquires drug resistance and adapts to a human host.

INTRODUCTION

Candida glabrata is a commensal of the human microflora but is also a significant cause of both superficial and invasive fungal infections. C. glabrata is the second most common causative agent of invasive candidiasis, accounting for 27% of Candida bloodstream infections in the United States (1). Despite its name, C. glabrata is more closely related to the nonpathogenic model organism Saccharomyces cerevisiae than other Candida family members. C. glabrata is also unique among other Candida species for its haploid genome that forces asexual reproduction. Perhaps as a result, C. glabrata shows high levels of genomic plasticity and chromosomal rearrangements as a means to generate genetic diversity are common (2–4).

C. glabrata possesses MIC₉₀ values five to seven dilution steps higher than C. albicans for azole class antifungals, independent of acquired resistance determinants (5). Beyond that, it can also develop bona fide resistance against azoles, which is often mediated by the upregulation of efflux pumps such as CDR1 and SNQ2 and a corresponding gain-of-function mutation in PDR1 (4). In addition to its reduced azole susceptibility, the echinocandin resistance rate for C. glabrata in the United States more than doubled in the period of 2001 to 2010, going from 4.9 to 12.3% (6). Echinocandin resistance in C. glabrata is frequently the result of mutations in the “hot spot” region of FKS2, one of the catalytic subunits of the β-1,3-glucan synthase target complex (7). Importantly, FKS mutations correlate not only with higher in vitro MICs but also with treatment failure and increased mortality in vivo (6). Here, we present the analysis of two serial C. glabrata bloodstream isolates obtained at 12-day intervals. Both strains show pan-azole resistance and echinocandin resistance evolved during the sampling interval, presumably as a result of the FKS2 S663P mutation identified through whole-genome sequencing in the second isolate. Our results highlight the ability of C. glabrata to rapidly acquire antifungal resistance, underscoring the increasing problem of resistant infections, and provide a global view of the genetic determinants necessary for C. glabrata to survive and persist in the bloodstream.

Patient case report and the development of echinocandin resistance in serial C. glabrata isolates.

Candida glabrata strains NRZ-2016-057 and NRZ-2016-058 are serial isolates from the bloodstream of an AML patient obtained at a 12-day interval. The patient was treated with voriconazole prophylactically from 8 December 2015 to 13 February 2016. During this period, she also received a short course of anidulafungin from 15 to 21 January. On 13 February 2016 the patient was switched from azole to echinocandin therapy and received 70 mg caspofungin on the initial day of treatment and then 50 mg on subsequent days from 14 to 20 February. On 20 February, antifungal therapy was switched to 100 mg of anidulafungin, which was maintained until patient death on 28 February 2016. To help manage the source of infection, the central venous catheter was removed on 14 February 2016 and all other catheters replaced. NRZ-2016-057 was isolated from a blood culture on 13 February 2016 while the patient was on prophylactic voriconazole therapy, and NRZ-2016-058 was obtained on 25 February 2016 from a follow-up blood culture after the patient had been switched to echinocandin therapy (Fig. 1). The relatedness of the isolates was confirmed by multilocus sequence typing (MLST) as described previously (8) and using the allele definitions and MLST profiles described by PubMLST (https://pubmlst.org/cglabrata/) (Table 1). NRZ-2016-057 and NRZ-2016-058 were deposited in and are publicly available via the Jena Microbial Resource Collection under numbers JMRC:NRZ:0308 and JMRC:NRZ:0309, respectively.

FIG 1.

Schematic overview of the patient case report, including antifungal history and isolation from the bloodstream of NRZ-2016-057 and NRZ-2016-058.

TABLE 1.

Antifungal susceptibility of C. glabrata strains NRZ-2016-057 and NRZ-2016-058

Strain	Strain type	MIC (mg/liter)a								Time elapsed (days)
		AMB	POS	VRC	ITC	FLC	ISA	AFG	CAS
NRZ-2016-057	3	0.25	>8	>8	>8	>64	>8	0.03	0.5
NRZ-2016-058	3	0.5	>8	>8	>8	>64	>8	4	>8	12

^aAMB, amphotericin B; POS, posaconazole; VRC, voriconazole; ITC, itraconazole; FLC, fluconazole; ISA, isavuconazole; AFG, anidulafungin; CAS, caspofungin.

The antifungal susceptibility of NRZ-2016-057 and NRZ-2016-058 was determined according to the EUCAST broth dilution method. Briefly, strains were precultured on yeast extract-peptone-dextrose (YPD) plates at 35°C, and then the susceptibility to amphotericin B (European Pharmacopoeia, Strasbourg, France), isavuconazole (Basilea Pharmaceutica International, Ltd., Basel, Switzerland), itraconazole (Janssen-Cilag GmbH, Neuss, Germany), posaconazole (MSD, Rahway, NJ), voriconazole (Pfizer, Inc., Peapack, NJ), fluconazole (Pfizer), anidulafungin (Pfizer), and caspofungin (MSD) was assessed. MICs were determined using a Nepheloskan Ascent type 750 (Labsystems, Helsinki, Finland) after 24 h of incubation at 35°C and are defined as 50% reductions in growth compared to drug-free wells. Two reference strains, Aspergillus fumigatus ATCC 204305 and Candida parapsilosis ATCC 22019 were included as quality control in each set of tests. NRZ-2016-057 showed elevated MICs to all azoles examined but was still susceptible to amphotericin B and echinocandin-class drugs (Table 1). NRZ-2016-058 acquired resistance to anidulafungin and caspofungin during the sampling interval, in addition to its previously obtained azole resistance, but it remained susceptible to amphotericin B.

Whole-genome sequencing of NRZ-2016-057 and NRZ-2016-058.

To determine the genetic changes that accompanied the development of echinocandin resistance, as well as any potential adaptations to the host environment, whole-genome sequencing was performed on NRZ-2016-057 and NRZ-2016-058 using 2x125bp Illumina paired-end sequencing by GATC Biotech (Constance, Germany). Raw genome sequence files (fastq) were uploaded to the NCBI Sequence Read Archive (SRA) and are publicly available under BioProject PRJNA483064.

Raw Illumina reads were processed by Trimmomatic (v0.32) (9) using specific parameters (leading:10 trailing:10 slidingwindow:4:15 minlen:50), ensuring the removal of adapter sequences and reads of low quality. After trimming, the read quality was checked with FastQC (v0.11.5) (10). Reads were aligned to the Candida glabrata reference genome CBS 138 (www.candidagenome.org) applying Novoalign (V3.07.01) in paired-end mode (Novocraft). From 4.4 and 5.5 million paired reads, we obtained 50- and 110-fold coverages for NRZ-2016-057 and NRZ-2016-058, respectively, covering 99% of the genome. Genetic variants, including single nucleotide variants (SNVs) and insertions and deletions (indels), were searched by SAMtools (v1.2; Samtools mpileup –uf) in conjunction with BCFtools using the multiallelic calling model (bcftools call –mv) (11). A minimum of five mapped reads for all reported variants was required. Processing and filtering of vcf files, allowing for the detection of amino acid substitutions in protein sequences, was carried out with the R package VariantAnnotation from Bioconductor (12). Compared to CBS 138, we identified 26,235 synonymous SNVs and 11,471 nonsynonymous SNVs in NRZ-2016-057 (Table 2). This strain also showed 4,168 indels, 171 of which were in coding regions. We detected a 10.8-kb deletion on chromosome A of NRZ-2016-057, containing the uncharacterized open reading frame (ORF) CAGL0A02255g, as well as deletions of similar size on chromosomes H, J, and K, which did not contain any annotated ORFs (Fig. 2). These results indicate that NRZ-2016-057 displays a relatively high degree of genetic diversity compared to CBS 138, an observation in line with previous studies showing high genetic diversity between C. glabrata strains (2, 13). A complete list of genetic changes detected in NRZ-2016-057 compared to CBS 138 can be found in Data Set S1 in the supplemental material.

TABLE 2.

Whole-genome comparison of CBS 138 and NRZ-2016-057

Parameter	Synonymous SNVs	Nonsynonymous SNVs	Indels	Indels in CDS
Total no.	26,235	11,471	4,168	171
Percentage vs CBS 138	0.21	0.09	0.03	0.003

FIG 2.

Read depth of NRZ-2016-057 across each chromosome when aligned to the CBS 138 genome. Highlighted regions indicate where spikes in read depth colocalize with the genomic location of putative or confirmed adhesin genes, suggesting increases in copy number.

C. glabrata contains a large number of adhesins and the ability of the organism to adhere to host cells and other structures is a key virulence determinant for the organism (reviewed in reference 14) and, in general, the number of adhesins correlates with pathogenicity within the Nakaseomyces clade (15). Interestingly, we observed large spikes in read depth that predominantly corresponded to the genomic location of annotated adhesin genes in CBS 138, suggesting increases in the copy number of adhesin genes (Fig. 2). To confirm spikes in read coverage were not PCR duplicates, duplicates were marked and removed using Picard (v1.134; remove_duplicates=true assume_sorted=true).

Genomic comparison between NRZ-2016-057 and NRZ-2016-058.

Genomic analysis of NRZ-2016-058 compared to NRZ-2016-057 with reference to CBS 138 identified two synonymous and nine nonsynonymous SNVs between the putative parent and offspring strain (Table 3). No indels were observed in NRZ-2016-058 that were not also present in the parent strain. In agreement with the development of phenotypic echinocandin resistance, we identified a T1987C (S663P) substitution in the FKS2 coding sequence of NRZ-2016-058, a mutation previously shown to be sufficient to induce echinocandin resistance in C. glabrata (16). However, we also detected a nonsynonymous SNV in MDE1, a gene that is downregulated in response to caspofungin in S. cerevisiae and results in enhanced resistance to caspofungin upon its deletion in this organism, suggesting that there could potentially be additional mechanisms contributing to the echinocandin resistance observed in NRZ-2016-058 (17, 18). Another previously undescribed nonsynonymous SNV between the two isolates is in FPR1, a TOR signaling member whose expression is upregulated in biofilm versus planktonic growth (19). The S. cerevisiae ortholog of FPR1 binds to rapamycin and FK506, both of which show antifungal activity. We also identified nonsynonymous SNVs in genes associated with metabolism and nutrient acquisition and the core stress response, including CAGL0M09999g and HXT4/6/7. The exact function of the genomic changes observed between NRZ-2016-057 and NRZ-2016-058 remains undefined; however, they presumably and collectively play a role in adapting the fungus to the host environment and the unique environmental stresses and antifungal therapy encountered there. It is also possible that these genetic changes conferred early or small increases in echinocandin resistance before the FKS2 S633P mutation was acquired.

TABLE 3.

Unique nonsynonymous SNVs observed in NRZ-2016-057 and NRZ-2016-058 compared to CBS 138^a

Locus	Gene	Nucleotide positionb	Nucleotide change	Amino acid change	GO biological process annotation(s)	GO molecular function annotation(s)
CBS 138 vs NRZ-2016-057
CAGL0G02541g		ChrG: 234003	G→A	S→N	ND	ND
CAGL0L13002g		ChrL:1391142	T→C	F→L	Mismatch repair, mitochondrial DNA repair	ATP-binding, DNA-dependent ATPase activity, dinucleotide insertion or deletion binding, guanine/thymine misrepair binding
CBS 138 vs NRZ-2016-058
CAGL0A02233g	HXT4/6/7	ChrA:236338	C→A	A →S	Transmembrane transport	Glucose transmembrane transporter activity
CAGL0G02519g		ChrG:229079	G→A	Q→stop	Protein kinase C-activating G-protein coupled receptor signaling pathway	Diacylglycerol kinase activity
CAGL0I04774g		ChrI:428298	G→A	A→V	Cellular calcium ion homeostasis	Enzyme regulator activity
CAGL0K04037g	FKS2	ChrK:375361	T→C	S→P	Cellular response to drug, 1,3-β-d-glucan biosynthesis pathway, ascospore wall assembly, pathogenesis	1,3-β-d-glucan synthase activity
CAGL0K09724g	FPR1	ChrK:952002	C→T	Q→stop	TOR signaling, cellular response to drug, chromatin reorganization, conjugation with cellular fusion, protein folding, regulation of homoserine biosynthetic process.	Macrolide binding, peptidyl-prolyl cis-trans isomerase activity
CAGL0M07876g	MDE1	ChrM:785919	T→C	I→V	l-Methionine salvage from methylthioadenosine	Metal ion binding, methylthioribulose 1-phosphate dehydratase activity
CAGL0M09999g		ChrM:992392	T→C	L→P	Negative regulation of Ras protein signal transduction, negative regulation of cAMP biosynthetic process, positive regulation of GTPase activity, regulation of adenylate cyclase activity	GTPase activator activity

^aLists of nonsynonymous SNVs observed in NRZ-2016-057 and NRZ-2016-058 compared to CBS 138 strain were compiled, and unique variants for each strain were identified by overlap analysis. ND, none determined.

^bNucleotide position from the CBS 138 genome.

Interestingly, aside from the FKS2 mutation, none of the changes observed between NRZ-2016-057 and NRZ-2016-058 overlapped those seen in another study following the genomics changes that accompanied the development of echinocandin resistance in a different set of C. glabrata isolates (16), highlighting the diverse mechanisms by which this organism can adapt to the selective pressures applied by the host and antifungal therapy. Moreover, despite the fact that the S633P mutation in FKS2 comes at a fitness cost to the organism in the absence of echinocandins (16, 20), we did not observe any significant growth differences between CBS 138, NRZ-2016-057, and NRZ-2016-058 in RPMI or YPD medium (data not shown), suggesting that one or more of the unique SNVs identified in NRZ-2016-058 could serve as a compensatory change for the reduced fitness resulting from the FKS2 mutation, potentially by upregulating FKS gene expression to compensate for the reduced catalytic potential of the mutated 1,3-β-d-glucan synthase complex, as has been shown for other C. glabrata isolates containing FKS hot spot mutations (20).

Finally, though NRZ-2016-057 and NRZ-2016-058 were resistant to all azoles tested, they do not possess any of the previously described mutations associated with azole resistance in C. glabrata, which are most commonly medicated via gain of function mutations in PDR1 (21). However, we did identify a novel R761M polymorphism in this gene not previously described among the sequences present in the GenBank nr collection, in addition to V91I, L98S, and D243N polymorphisms which have been previously described in both azole-susceptible and azole-resistant isolates (22). Finally, neither strain showed any genetic changes in the mismatch repair complex gene MSH2 that has also been shown to drive multidrug resistance in C. glabrata (23). The not-yet-deciphered mechanism by which azole resistance is mediated in NRZ-2016-057 and NRZ-2016-058 is an intriguing question warranting further investigation.