Molecular type distribution and fluconazole susceptibility of clinical Cryptococcus gattii isolates from South African laboratory-based surveillance, 2005–2013

As is the case globally, Cryptococcus gattii is a less frequent cause of cryptococcosis than Cryptococcus neoformans in South Africa. We performed multilocus sequence typing (MLST) and fluconazole susceptibility testing of 146 isolates randomly selected from 750 South African patients with C. gattii disease identified through enhanced laboratory surveillance, 2005 to 2013. The dominant molecular type was VGIV (101/146, 70%), followed by VGI (40/146, 27%), VGII (3/146, 2%) and VGIII (2/146, 1%). Among the 146 C. gattii isolates, 99 different sequence types (STs) were identified, with ST294 (14/146, 10%) and ST155 (10/146, 7%) being most commonly observed. The fluconazole MIC50 and MIC90 values of 105 (of 146) randomly selected C. gattii isolates were 4 μg/ml and 16 μg/ml, respectively. VGIV isolates had a lower MIC50 value compared to non-VGIV isolates, but these values were within one double-dilution of each other. HIV-seropositive patients had a ten-fold increased adjusted odds of a VGIV infection compared to HIV-seronegative patients, though with small numbers (99/136; 73% vs. 2/10; 20%), the confidence interval (CI) was wide (95% CI: 1.93–55.31, p = 0.006). Whole genome phylogeny of 98 isolates of South Africa’s most prevalent molecular type, VGIV, identified that this molecular type is highly diverse, with two interesting clusters of ten and six closely related isolates being identified, respectively. One of these clusters consisted only of patients from the Mpumalanga Province in South Africa, suggesting a similar environmental source. This study contributed new insights into the global population structure of this important human pathogen.


Introduction
Basidiomycetous fungi within the genus Cryptococcus are the most common cause of meningitis among HIV-seropositive adults in southern Africa [1]. Human disease is mostly caused by two species complexes within the genus, Cryptococcus neoformans and Cryptococcus gattii [2], with C. gattii causing fewer than 1% of cases of cryptococcal disease globally [3]. While more often reported in the literature as causing disease in immunocompetent hosts without underlying medical conditions, C. gattii also causes disease in immunocompromised people [3][4][5].
People infected with C. gattii appear to present with pulmonary cryptococcosis more often than cryptococcal meningitis, with the opposite pattern being observed for disease caused by C. neoformans [4,6]. Among HIV-seropositive adults with advanced immunosuppression, C. gattii disease (meningitis with or without fungaemia) has been reported to be indistinguishable from disease caused by C. neoformans [6,7]. Regardless of the pathogen, flucytosine is recommended in combination with amphotericin B for induction treatment of HIV-associated cryptococcal meningitis, followed by fluconazole monotherapy for the consolidation and maintenance phases of treatment [8].
C. gattii was historically thought to be confined to tropical and subtropical regions and was found less frequently in temperate regions [9]. The environmental range of the two closely-related species overlaps in southern Africa, though C. gattii is a less common pathogen [10]. C. gattii is mainly associated with decomposing plant matter and certain tree species, such as eucalyptus in Australia and California [11,12], firs and oaks in the Pacific Northwest [13] and numerous tropical tree species (e.g. Ficus spp.) [14], with a new niche for C. gattii recently described after its isolation from hyrax faeces in an environmental study carried out in Zambia [12].
Since C. gattii is mainly restricted to sedentary trees, it is not as widely distributed as C. neoformans, which is hypothesized to have a global distribution owing partially to its ability to metabolise pigeon droppings [15]. C. gattii has also demonstrated the ability to adapt to new temperate niches, causing a geographically restricted outbreak in North America [13,[15][16][17]. There are many theories on the unexpected and unexplained emergence of C. gattii in North America. A recently-published hypothesis suggests that C. gattii was transported by ships in contaminated ballast water tanks from South America to North America in 1914 [18]. This organism then established itself in coastal waters. A tsunami in 1964 may then have carried C. gattii into the coastal forests. It adapted subsequently to a new environmental niche (i.e. land/ forest) causing human infections about three decades later [19]. While C. neoformans has been hypothesized to originate from Africa, C. gattii could have originated from South America considering the high genetic diversity of South American C. gattii isolates [10,20].
C. gattii is divided into six molecular types: VGI, VGII, VGIII, VGIV, VGV and VGVI [21,22], of which five have been proposed as separate species [23], with the molecular type VGV recently identified from six environmental isolates collected from the Zambian Central Miombo woodlands [22]. VGI is mostly found in Australia and Papua New Guinea [24]. VGII is prevalent in Canada and the United States of America (USA), where it has caused a major outbreak [13]. This molecular type is also widely present in Australia [25], Brazil and Colombia [26]. VGIII has been found in the USA [27], Mexico, Brazil and Colombia [26]. The molecular type VGIV has until now been mainly identified from India [28], Colombia, Mexico [29], Botswana, Malawi [30] and Zimbabwe [31]. VGVI has been newly described based on two identical isolates from Mexico [23]. In earlier molecular epidemiology studies including relatively few isolates, most southern African isolates were confirmed as molecular type VGIV [30,32], with this molecular type being mostly associated with disease in HIV-infected individuals [4].
To expand on these earlier findings, we aimed to: (1) determine the molecular diversity of C. gattii in South Africa by genotyping a subset of clinical isolates from enhanced laboratorybased surveillance conducted between 2005 and 2013; (2) describe the fluconazole susceptibility of these isolates, and (3) determine if there was an association between patient clinical characteristics and the infecting molecular type, and the effect of molecular type on in-hospital mortality. In addition, we analysed the genomes of 98 isolates of the most common South African C. gattii molecular type VGIV to place them into a global context.

Ethics statement
Ethics clearance for this study was obtained from the Human Research Ethics Committee (Medical), University of the Witwatersrand with clearance certificate numbers: M160375 and M1809107. All data analysed was anonymized.

Study design and sample selection
We conducted a cross-sectional study nested within laboratory-based surveillance for cryptococcosis in South Africa from 2005 to 2013. We used this time period since routine collection and storage of cryptococcal isolates through surveillance was terminated in 2013. From 1 January 2005 to 30 June 2008, cryptococcal isolates were submitted from all laboratories to the National Institute for Communicable Diseases (NICD). The surveillance methodology changed from 1 July 2008 to 31 December 2013 when only enhanced surveillance sites (ESS) (29 hospitals in 9 provinces), National Health Laboratory Service (NHLS) laboratories in Kwa-Zulu-Natal Province, and pathology laboratories in the private, mining, and military sectors were required to submit isolates. A case was defined as a person diagnosed with cryptococcal disease by any one of the following positive tests during a 30-day period: (1) India ink microscopy on cerebrospinal fluid (CSF) or (2) a positive cryptococcal antigen test on blood or CSF or (3) culture of Cryptococcus from any clinical specimen. Recurrent isolates from the same patients were excluded and only cases with cultured and viable isolates were included [33]. We also excluded isolates that were identified as C. neoformans or other Cryptococcus species. Following these inclusion and exclusion criteria, 750 cases of C. gattii infection in total were available for analysis. HIV infection status was recorded for 387 of these patients at ESS; we excluded all other patients with an unknown HIV infection status. Among these 387 patients, 374 were HIV-seropositive and 13 were HIV-seronegative. One hundred and thirty-six C. gattii isolates were randomly selected from the 374 HIV-seropositive cases using a random-integer generator (https://www.random.org/integers/) and all 13 isolates from HIV-seronegative persons were selected but only ten were viable. In total, 146 unique cases were selected from 387 cases of C. gattii disease between 2005 and 2013 and the corresponding isolates were genotyped (S1 Fig).

Subculture and identification of surveillance isolates
Following primary isolation of Cryptococcus species at diagnostic pathology laboratories, a sweep of the culture plate was inoculated onto Dorset medium (Diagnostic Media Products (DMP), NHLS, Johannesburg, South Africa) and transported to a reference laboratory at the NICD in Johannesburg, where the isolates were identified by phenotypic methods [34] and stored at -70˚C. The patient metadata that accompanied each isolate were captured in a surveillance database. The selected C. gattii isolates were retrieved from -70˚C storage and sub-cultured onto Sabouraud dextrose agar (DMP, NHLS, Johannesburg, South Africa) to check for purity and viability [33]. We phenotypically characterised isolates as previously described [34].

Multilocus sequence typing (MLST) and data analysis
DNA was extracted from single yeast colonies using the Zymo ZR Fungal/Bacterial DNA MiniPrep kit (Zymo Research Corp, USA) following the manufacturer's instructions. The only modification from the manufacturer's instructions was the starting material for DNA extractions, since single yeast colonies grown on Sabouraud agar (DMP, NHLS, Johannesburg, South Africa) at 30˚C after 48h were used. The International Society for Human and Animal Mycology (ISHAM) MLST consensus scheme for C. neoformans and C. gattii containing six housekeeping genes: CAP59, GPD1, LAC1, PLB1, SOD1, URA5 and the intergenic spacer region IGS1 was used as previously published [21] for genotyping. DNA was amplified using a conventional PCR with DreamTaq Hot Start DNA Polymerase (Thermo Fisher Scientific, Waltham, Masschusetts, USA) in an Applied Biosystems 2720 Thermal cycler (Thermo Fisher Scientific, Waltham, Masschusetts, USA) using PCR cycling conditions described previously [21]. Amplicons were visualised on a 1% agarose gel and Sanger sequencing was performed. For each locus, DNA sequences in both forward and reverse directions were obtained and edited using the program Sequencher ver. 5.4.6 (http://www.genecodes.com). Allele types (ATs) and sequence types (STs) were assigned using the online C. gattii MLST database (http://mlst.mycologylab.org). The concatenated DNA sequences for the seven loci (S1 Table) were aligned using the program ClustalW (BioEdit Sequence Alignment Editor). A phylogenetic tree was generated with the program MEGA 5 using the neighbour-joining method with a bootstrap analysis of 100 replicates and the Jukes-Cantor model [35].
Mating type identification was performed by conventional PCR amplification using primer sets and cycling conditions from previous studies [36,37], which are specific to the mating type regions of a and α mating-type cells. Amplicons were visualised on a 1% agarose gel and the mating type was determined using amplification that related to each mating-type's expected amplicon size [36,37].

Fluconazole susceptibility testing
Of the 146 C. gattii isolates, fluconazole susceptibility testing was performed for 105 isolates: 60 randomly selected VGIV (from a total of 101), 40 VGI, three VGII and two VGIII. We determined the fluconazole minimum inhibitory concentration (MIC) values (range: 0.125 μg/ ml to 64 μg/ml) using custom-made broth microdilution (BMD) panels (NICD, Johannesburg) prepared, inoculated and read according to susceptibility testing methods published previously and the Clinical and Laboratory Standards Institute (

Whole genome sequencing and phylogenetic analysis
Whole genome sequencing (WGS) was performed on 98 molecular type VGIV isolates using Illumina MiSeq sequencing technology (Illumina, San Diego, California, USA). Previously extracted DNA samples (extraction procedure described above) were prepared for paired-end sequencing using the NEBNext Ultra II DNA Library Prep Kit for Illumina followed by 2 × 300 bp sequencing on a MiSeq instrument at the Translational Genomics Research Institute (Pathogen and Microbiome Division, Flagstaff, USA). Single nucleotide polymorphisms (SNPs) were detected from raw read data using the publically available Northern Arizona SNP pipeline (NASP) [43]. The pipeline was set to align reads for every sample to the known assembled reference VGIV genome (IND 107, NCBI BioSample number: SAMN01932842) using Burrows-Wheeler Aligner (BWA) [44]. SNP variants were detected using SAMtools [45]. Posterior filtering parameters involved removing positions that had <10x coverage, <90% variant allele calls, and those mapping to duplicated regions in the reference. Downstream filtering by NASP produced the final high-quality or BestSNP alignment that was then used for maximum parsimony inference. Only positions present in all genomes with at least 10x depth of coverage and 90% agreement were included. Maximum parsimony trees with 100 bootstrap replicates were constructed using the phangorn library [46] and visualized using MEGA 6 software [47]. We included 47 additional global VGIV whole genome sequences. Nineteen VGIV isolates were obtained from the Australian Medical Mycology Culture Collection (WFCC registration number: WM-1205) and sequenced herein (S2 Table). The remaining VGIV whole genome sequences were obtained from NCBI GenBank; six of which were sequenced in previous studies [17,22,48]. The 47 additional VGIV genomes included 19 isolates from Botswana, 17 isolates from South Africa, four isolates from Colombia, three isolates from India, two isolates from Australia, one isolate from Zambia and the geographical origin of one isolate was unknown. For assessment of specific clusters, trees were drawn for each obtained cluster separately to maximize the high-resolution of comparable positions between samples of each apparent clade. For these specific sub-analyses, draft assemblies of one of the samples in each clade were created de-novo using SPAdes (https://cab.spbu.ru/software/spades/) in careful mode and then used as references for the cluster trees. A cluster was defined as a group of more than two isolates that shared a common ancestor with fewer than 1100 SNPs between isolates. All WGS data were deposited in the NCBI SRA under the BioProject accession number PRJNA804139.

Statistical data analysis
For the 2005-2013 surveillance period, we compared demographic characteristics of patients with C. gattii and C. neoformans disease to determine if there were any differences between these two species complexes. For C. gattii infection, we used logistic regression models to determine associations between patient characteristics and molecular type. Patient case data were collected by nurse surveillance officers at ESS who either interviewed patients or reviewed their medical charts [33]. We defined the molecular type groups as VGIV vs. non-VGIV, since VGIV was the dominant molecular type in South Africa. We included the following variables in the analysis with molecular type as the dependent variable: sex, age, year of diagnosis, geographical region, specimen type, HIV infection status, mental status at diagnosis, CD4+ T-cell (CD4) count at diagnosis, antiretroviral treatment, current antifungal treatment, and tuberculosis treatment. Mental status was categorised as "alert" (a Glasgow Coma Scale [GCS] score of 15) or "not alert" (a GCS score <15 or a patient was recorded to be disorientated, stuporose or comatose at the time of diagnosis) [33]. We divided the geographical regions of South Africa, based on the Koppen-Gieger climate classification zones [49], into mostly temperate (Gauteng, Mpumalanga, KwaZulu-Natal and Western Cape provinces) vs. mostly arid (Northern Cape, Free State, Limpopo, North West and Eastern Cape provinces). In the final multivariable analysis, we only adjusted for factors that were associated with geographical region and were risk factors for infection caused by VGIV isolates; the variables age, sex and HIV infection status all met these criteria. We also separately modelled the effect of molecular type on in-hospital mortality, adjusting for sex, age, mental status at diagnosis, CD4 count at diagnosis, antiretroviral treatment, and current antifungal treatment. We compared the fluconazole MIC 50 values of the VGIV isolates and non-VGIV isolates using a Wilcoxon rank-sum test. We also compared the association between MIC value (�8 μg/ml vs >8 μg/ml) and outcome data of patients infected with C. gattii. Statistical analyses were performed using Stata statistical software ver. 14.1 (StataCorp LP, Texas, USA). In terms of gender differences of cases infected by the two species complexes, 57% (426/748) of patients infected with C. gattii were male while 48% (11,057/23,206) infected with C. neoformans were male (p = 0.0001). More patients diagnosed with either C. neoformans or C. gattii disease were from the mostly temperate provinces of South Africa (C. gattii: 565/750, 75%; C. neoformans: 17,682/23,512, 75%) compared to the mostly arid provinces (C. gattii: 185/750, 25%; C. neoformans: 5,830/23,512, 25%). For both C. gattii and C. neoformans infections, CSF was the most common specimen type from which the pathogen was cultured (C. gattii: 712/750, 95%; C. neoformans: 22,279/23,512, 95%) followed by blood culture (C. gattii: 32/750, 4%; C. neoformans: 1179/23,512, 5%).

Association between patients' clinical characteristics and molecular type
After adjusting for sex, age and HIV infection status, patients diagnosed in the mostly temperate provinces of South Africa had a 43% reduced odds of a VGIV infection than patients from PLOS NEGLECTED TROPICAL DISEASES mostly arid provinces, though the 95% CI spanned 1 (0.21-1.58) and the p-value was large (0.28) (S3 and S4 Tables). We also found that HIV-seropositive patients had a ten times increased odds of a VGIV infection compared to those who were HIV-seronegative in the same adjusted analysis, though with small case numbers, the CI was very wide (95% CI: 1.93-55.31, p = 0.006) (S3 and S4 Tables).

Fluconazole susceptibility testing
The average fluconazole MIC 50 and MIC 90 values for all 105 tested C. gattii isolates (60 VGIV, 40 VGI, 3 VGII and 2 VGIII) were 4 μg/ml and 16 μg/ml, respectively, with a geometric mean of 4.53 μg/ml (Table 1). There was a statistical difference in the MIC 50 values between the VGIV isolates and non-VGIV isolates (p = 0.001), though these differed by only one doubledilution (4 μg/ml for VGIV isolates vs. 8 μg/ml for non-VGIV isolates) ( Table 1). Most of the VGI, VGII and VGIV isolates had a MIC value of �8 μg/ml, which is considered to be wildtype according to published ECVs [41,42]. However, the two VGIII isolates, six VGIV isolates and six VGI isolates were considered to be non-wild-type [41,42]. On unadjusted analysis, patients with a cryptococcal isolate that had a fluconazole MIC value of greater than 8 μg/ml were more than one and a half times more likely to die, though this estimate also had a 95% CI spanning 1 and a large p-value (95% CI: 0.47-5.64; p = 0.43).

Whole genome sequencing of 98 C. gattii VGIV isolates
In order to place our South African VGIV isolates into a global context, we compared the WGS data between South African VGIV isolates and global VGIV isolates. Fig 2 shows the WGS SNP analysis of 98 VGIV isolates from this study, the VGIV reference strain IND 107 and 47 global VGIV isolates. There were 123,891 SNPs called. The overall minimum, median and maximum number of SNPs among all 146 isolates was 0; 13,309 and 21,804, respectively. Kmer-based typing of these genomes in-silico using a custom kraken database indicated that none of our isolates belonged genetically to the new molecular type VGV identified recently in Zambia [22]. Scattered throughout the tree, there were ten clusters overall (see labelled clusters in Fig 2 and Table 2). Seven clusters included South African VGIV isolates only. Cluster 1 consisted of four isolates. The median number of SNPs was 198 between three isolates (924, 1428 and 274). The fourth isolate (4102) was separated by 850 SNPs from the other three isolates in this cluster. Isolate numbers 924 and 1428 were from patients living in the same town, in the Mpumalanga Province, diagnosed at the same facility just over two months apart in 2011. Isolates 3096, 786 and 3470 were in cluster 2 and the median number of SNPs was 404. All three patients from whom these isolates originated lived in the Gauteng Province and two patients were diagnosed in the same facility less than two months apart in 2013. There were two interesting clusters of closely related isolates observed (clusters 3 and 4, Fig 2 and Table 2). Cluster  Table 2). There were fewer than 700 SNPs between isolate number 1005 and two South African samples sequenced from the Australian Medical Mycology Culture Collection in cluster 6. In cluster 7, there were two isolates from our study (numbers 3943 and 841) collected in 2008 and 2011 respectively, obtained from patients living in Gauteng. This cluster also included South African samples (WM0420 and WM779) sequenced from the Australian Medical Mycology Culture Collection separated from each other by less than 20 SNPs but separated from isolate number 841 and isolate number 3943 by fewer than 1100 SNPs. There were two clusters that included South African VGIV isolates from our study and some of the VGIV isolates from other countries. The minimum, median and maximum number of SNPs were 0, 6 and 478 SNPs, respectively, between four Colombian isolates [50], a publically available sequence from a South African animal isolate and isolate number 3994 (cluster 8, Fig 2 and Table 2). The patient infected with isolate number 3994 was from the Gauteng Province and was diagnosed in September 2008. The minimum, median and maximum number of SNPs were 0, 318 and 908 SNPs, respectively, between isolate number 547 (patient from Gauteng Province) and isolate number 3882 (patient from Mpumalanga Province) and fifteen Botswanan samples that were isolated from CSF in 2000 and 2012 (cluster 9, Fig 2 and Table 2). There was one cluster that contained no South African VGIV isolates (cluster 10, Fig  2). This cluster included two Indian isolates (isolate numbers: WM2363 and WM2364), an isolate with unknown origin (WM14321) as well as the reference VGIV isolate (IND107). There were less than five SNPs between isolates WM2363, WM2364 and WM14321 with at most 50 SNPs between these three isolates and the reference.
We then analysed the Gauteng/Limpopo cluster and Mpumalanga cluster separately using the draft assembly of one of our isolates (isolate numbers 918 and 976, respectively) as a reference in NASP. There were 1600 SNPs called within the Gauteng/Limpopo cluster of ten isolates (Fig 3A and 3C). The minimum, median and maximum number of SNPs were 52, 426 and 574 SNPs, respectively. Six patients from whom these isolates originated were male and four patients were female. Seven patients were from the Limpopo Province; four of which were diagnosed at the same facility; and three were from the Gauteng Province. The cases did not cluster in time:  Table). The map in Fig 3C highlights the clustered cases in yellow from the Gauteng, Limpopo and Mpumalanga Provinces. The ten patients in the Gauteng/Limpopo cluster were all located in towns within a 355 km radius. The Gauteng patients were from three towns located within a 70 km radius. The Limpopo patients were from two towns also located within a 70 km radius. The Mpumalanga cluster of six isolates from our study and two from a previously published South African study [48] contained 1649 SNPs among all genomes (Fig 3B and 3C). The minimum, median and maximum number of SNPs were 0, 598 and 699 SNPs, respectively. The six isolates from our study originated from five male patients and one female patient, with five patients diagnosed at the same facility. The cases did not cluster in time: one patient each was  Table). The two previously published sequences were from isolates cultured from another South African patient with recurrent cryptococcosis between 2005 and 2009 [48]. Two isolates, number 70 and number 2153, within this cluster were distanced by zero SNPs between them. These patients lived in two towns that were about 30 kilometres apart, but both were diagnosed at the same facility two years apart (2008 and 2010) (S8 Table). As mentioned before, the map in Fig 3C highlights the clustered cases from the Mpumalanga, Gauteng and Limpopo Provinces shown in yellow. The patients from the Mpumalanga Province with isolates that were closely related were all located in four towns in close proximity (within a radius of 51 km).

Discussion
Laboratory-based surveillance for cryptococcosis, a common HIV-associated opportunistic infection in South Africa, provided a platform to determine the molecular epidemiology of C. gattii. Only 3% of viable isolates were identified as belonging to the C. gattii species complex with the remainder identified as belonging to the C. neoformans species complex. Previous clinical studies from neighbouring southern African countries have reported a much higher proportion of C. gattii infections, such as 13% in Botswana and Malawi [30] or 17% in Zimbabwe [31]. However, these studies included small case numbers, where cryptococcal isolates were obtained from patients admitted to a single facility or cases were not randomly selected from a defined study population.
We observed 99 STs among our clinical C. gattii isolates, indicating a high genetic diversity, with VGIV being the dominant molecular type in South Africa. This confirms the results of studies undertaken in Zimbabwe, Botswana and Malawi [30,31]. In an earlier South African study on isolates from episodes of recurrent cryptococcosis, there were eight C. gattii isolates recovered from four patients and four isolates each had the molecular type VGI and VGIV, respectively [34]. In our larger study, we were able to describe the molecular type distribution of C. gattii in South African patients. Twenty-seven per cent of the clinical isolates in our study had the VGI molecular type; this molecular type has been described in a previous study as being dominant among Kenyan clinical and environmental isolates [51]. In a study of 350 clinical, veterinary and environmental C. gattii isolates from Africa, Asia, Europe, North America and South America, VGII was the dominant molecular type followed by VGI [52]. There were lower percentages of VGIII and VGIV isolates. For the twenty-eight African isolates (countries unspecified), accounting for 8% of the total number of isolates included in this global study, VGI was the dominant molecular type, followed by VGIV. The authors concluded that VGI and VGII are globally distributed whereas VGIII and VGIV are minor molecular types that seem to be geographically restricted to specific locations [52]. However, in this specific study, a larger number of isolates was obtained from Europe, North America and South America, and therefore this collection was not representative of the global molecular type distribution.
The fluconazole MIC 50 values, MIC 90 values and ranges overlapped for isolates with all four molecular types in our single-centre study. Our study therefore does not provide sufficient evidence for any meaningful difference in fluconazole susceptibility in relation to molecular type. Compared to the fluconazole MIC data for C. neoformans isolates from a previous South African study using the same surveillance platform, the MIC 50 and MIC 90 values for C. gattii differed by two fold higher dilutions [33]. A previous study showed that there was a higher level of heteroresistance to fluconazole in C. gattii than in C. neoformans [53]. These C. gattii isolates were also more resistant to xenobiotics than C. neoformans which could be due to the ecological niche C. gattii inhabits since tree hollows may harbour various ubiquitous saprophytic microorganisms that produce xenobiotics [53]. All C. gattii isolates showed heteroresistance in this one study; therefore, this is an intrinsic property of C. gattii that the authors speculate reflect an acquisition from its ecological niche but this needs to be studied further [53].
We found incidentally that HIV-seropositive patients had an increased odds of a VGIV infection compared to those HIV-seronegative, after adjusting for age, sex and climatic region.
With a small sample size, our estimate of this effect is likely to be imprecise. However, previous studies in southern Africa have also shown that the VGIV molecular type is a major cause of C. gattii meningitis in HIV-seropositive patients [30,31]. In general, VGIV and VGIII have been reported to cause infection in HIV-seropositive patients whereas VGI and VGII are reported to cause infection mainly in immunocompetent hosts [4,5,54,55]. We are uncertain of what could explain this difference but Harris and colleagues previously speculated that the C. gattii molecular type, sequence type, fungal population genetics and environmental distribution "probably interact to mediate infection in patients with varying degrees of immune competence" [56]. Based on the point estimate, patients infected with the VGIV molecular type were less likely to die in hospital compared to patients infected with the non-VGIV molecular types, although we may have found this difference entirely by chance with a large p-value. Beale et al. observed the opposite for C. neoformans whereby the VNB genotype was associated with a worse survival outcome [57]. Recent phenotypic variation results suggest that the VGIV molecular type may be a less virulent pathogen [58]. These authors observed that the VGIV molecular type is similar to other molecular types in terms of capsule production and cell size. Both VGIV and VGIII molecular types were more sensitive to temperature compared to VGI and VGII that are considered most virulent [58]. VGIV also produces irregular cells, a variant phenotype not shown to be related to patient death when compared to other variant phenotypes, such as giant cells, micro cells and shed extracellular capsule produced by Cryptococcus species [58]. There is also evidence that virulence is not specifically associated with a certain molecular type [59,60]. In one of these studies, Galleria mellonella larvae were infected with the four C. gattii molecular types and highly virulent isolates were observed from all molecular types, suggesting that virulence may be related to strain-specific characteristics regardless of the molecular type [59].
In our WGS analysis, we observed a total of ten closely-related clusters and we zoomed in on two clusters specifically which we called the Gauteng/Limpopo cluster and Mpumalanga cluster consisting of ten and six isolates, respectively. The median number of SNPs for each of these two clusters were 426 and 598, respectively. A previous study described nine VGI isolates isolated from patients living in the south-eastern states of the USA that clustered together with 41,000 SNPs and 4558 SNPs between any two isolates [61]. Firacative and colleagues were able to define two major clusters of VGIII isolates in their study, separating serotype B from serotype C isolates. There were 88,337 SNPs for one cluster and 79,945 SNPs for the second cluster [60]. Three clusters were observed amongst the VGII outbreak isolates in the Pacific North West region of North America, made up of VGIIa, VGIIb and VGIIc, with 107, 132 and 137 SNPs within each of those subtypes, respectively, indicating that these VGII populations in that region were clonal with a low genetic diversity [17,61]. Our results suggest that, like VGI and VGIII, the VGIV molecular type also has a high genetic diversity. We observed that one of the isolates from our study and a previously isolated South African sample were related to four previously isolated environmental Colombian VGIV isolates with <480 SNPs [50]. This suggests some level of relatedness between VGIV isolates from different geographical regions or a similar environmental exposure, although we do not have the travel history of the patients to confirm this. The possible relatedness between Colombian and South African isolates could be due to genetic exchange between C. gattii molecular types before the continental drift [20,62]. According to one hypothesis, before the ancient supercontinent Pangea split into the present continents, C. neoformans and C. gattii shared a common Pangean ancestor somewhere in between South America and Africa since these two continents were next to each other on Pangea [20]. This continental drift dispersal hypothesis has further been applied to the molecular types within the C. gattii species complex [62].
Our surveillance data also consisted of the patients' residential addresses and healthcare facility at the time of clinical diagnosis. However, this does not take into account the long latency period of cryptococcosis in some cases and relocation of residence between provinces. The Gauteng, Limpopo, North West and Mpumalanga Provinces are located in the northern parts of South Africa. These provinces are inland. However, climatic differences between these provinces do exist. The Gauteng Province is elevated and temperate, whereas parts of the Mpumalanga Province are low-lying, subtropical to tropical and humid but also dry in some areas. The North West Province and the Limpopo Province are hot and dry. There was a cluster of six isolates from patients living in the Lowveld region of Mpumalanga Province in South Africa, suggesting a common environmental source. Based on our findings, we can speculate that the VGIV molecular type may be widely distributed in this subtropical environment. The reason for its widespread environmental distribution could be due to this genotype's high fitness in the environment. This has been observed with triazole-resistant Aspergillus fumigatus with the TR/L98H substitution [63] as well as in C. gattii in a recent study done in Zambia [22]. In the Zambian study, C. gattii was mostly associated with the Zambian Central Miombo woodlands that is dry and tropical [12]. The genomes of six of these isolates were recently sequenced and found to represent a new molecular type, VGV [22]. This molecular type was found in hyrax-associated environments making this a new environmental niche for C. gattii. We did not find any VGV isolates in our collection of clinical C. gattii isolates. As Farrer and colleagues conclude, it is possible that this molecular type is strictly environmental and yet to cause clinical infection [22]. Further studies of clinical and environmental C. gattii isolates from southern Africa should be performed to gain a better understanding of this pathogenic species complex [10].
In conclusion, our study showed that South African clinical C. gattii isolates, collected through enhanced laboratory-based surveillance from 2005-2013, were highly diverse, the VGIV molecular type dominated, and these isolates also mostly had low fluconazole MIC values. HIV-seropositive patients were more likely to be infected with VGIV isolates. We observed two interesting clusters of closely related VGIV isolates, with one cluster consisting of patients from the Mpumalanga Province in South Africa, suggesting a similar environmental source. Our large study provides a broader baseline for more extensive WGS studies in southern Africa, to enable us to gain a more complete picture of C. gattii, the lesser known cause of HIV-associated cryptococcal meningitis in this part of the world.