Abstract
Cryptococcus neoformans most frequently causes disease in immunocompromised patients. However, in Vietnam and east Asia, disease is frequently reported in apparently immunocompetent patients. We have previously shown that almost all such disease is due to a specific lineage of C. neoformans – VNIa-5. However, in HIV-infected patients, infections due to this lineage are not associated with worse outcomes. Here, we demonstrate that the VNIa-5 lineage presents different virulence phenotypes depending on its source. Isolates derived from immunocompetent patients are more virulent than those from HIV-infected patients or the environment. Moreover, the virulence phenotype is plastic – sterile culture filtrate from highly virulent VNIa-5 strains can induce increased virulence in less virulent VNIa-5 isolates, which in turn can then induce increased virulence in their low virulence states. We present evidence that this phenomenon is driven by secreted proteins associated with extra-cellular vesicles.
Introduction
Cryptococcal meningitis is a devastating disease due to infection with encapsulated yeasts of the Cryptococcus genus. The vast majority occurs in HIV-infected patients due to infection with Cryptococcus neoformans, and has a high moratlity rate 1. Disease in HIV-infected patients has been driven by the clonal expansion of a small number of well-defined lineages. These lineages are widely dispersed globally, but a single lineage predominates in most countries2. However, Vietnam is atypical and has two co-dominant circulating lineages (VNIa-5 and VNIa-4), each accounting for approximately 35-40% of cases of meningitis in HIV-infected patients2-4.
In addition to HIV-associated disease, cryptococcal meningitis is also well-described in HIV-uninfected patients in Southeast and East Asia5-9. Such patients account for approximately 20% of cases at our hospital in Vietnam9; the majority of these patients are apparently immunocompetent. Outcomes are similar to those in HIV patients, with 3 month mortality rates in the order of 20-30%9. Contrary to other locations in tropical and sub-tropical areas10,11, most infections in these patients (80%) are due to C. neoformans, rather than C. gattii12. We previously identified an association between the C. neoformans VNIa-5 lineage and disease in these apparently immunocompetent patients, it accounting for approximately 90% of meningitis due to C. neoformans2,3,12. Furthermore, HIV-uninfected patients with cryptococcal meningitis due to other lineages were significantly more likely to have co-morbidities associated with immune compromise12. These data suggest that VNIa-5 isolates have increased pathogenic potential. Of note, we do not observe clustering of isolates within the VNIa-5 lineage according to host immune status, signifying that the entire lineage has the potential to cause disease in immunocompetent people2.
Here, we aimed to understand the pathogenic ability of isolates of the VNIa-5 lineage. Using the Galleria mellonella infection model, we identified differences in virulence between isolates of the VNIa-5 lineage of different ecological backgrounds. The most virulent isolates were from HIV-uninfected patients, which were more virulent than VNIa-5 isolates from HIV-infected patients or the environment. However, we could induce increased virulence in lower virulence VNIa-5 isolates via transfer of sterile culture filtrate derived from highly virulent VNIa-5 isolates from immunocompetent patients. Consequently, these ‘induced’ isolates could sequentially increase the virulence of other low virulence isolates, and their own ‘naïve’ self. This virulence plasticity is mediated by peptide/protein associated with extracellular vesicles and likely underlies the ability of the VNIa-5 lineage to cause disease in immunocompetent people.
Results
Assessing the virulence of C. neoformans VNIa-5
We exploited the Galleria model of infection to compare the relative virulence of 20 VNIa-5 clinical isolates from HIV-uninfected patients with 20 VNIa-4 isolates from HIV-infected patients (Table S1). VNIa-5 infected Galleria had a significantly increased hazard of death compared with VNIa-4 infected Galleria (Hazard Ratio (HR) 1.4, 95% Confidence Interval (95CI) 1.2-1.6, P<0.001) (Figure 1A). After comparing the relative virulence of VNIa-5 isolates according to their source we found that Galleria infection with isolates from immunocompetent patients was associated with a significantly increased hazard of death compared with infection with isolates from HIV-infected patients (HR 2.2. 95CI 1.6 – 3.0, P<0.001), or with isolates from the environment (HR 5.7, 95CI 3.9 – 8.4, P<0.001, Figure 1B).
Hypothesising that the difference in virulence phenotype by ecological background was a function of a previous infection experience we six-fold passaged environmental organisms through Galleria. The virulence phenotype of the environmental isolates remained stable over these multiple passages through Galleria with no change in the hazard of death between the infection with the ‘naïve’ environmental isolate versus infection with that isolate following passage (Figure S1).
We compared the expression of in vitro phenotypic characteristics associated with virulence in VNIa-5 isolates depending on their source (immunocompetent patients or HIV-infected patients). We compared 15 isolates from immunocompetent patients with 15 isolates from HIV-infected patients and measured growth rates in YPD broth and pooled human cerebrospinal fluid (CSF), capsule size, extracellular urease, phospholipase, and laccase activity. There was no difference in cell diameter, melanin production, urease, or phospholipase production. Isolates from HIV-uninfected patients had moderate but significant increased growth in YPD by 48 hours at 30°C and 37°C compared with isolates from HIV-infected patients (median 4.5×105 CFU/ml (Interqartile Range (IQR) 3.6 X105 - 5.2 ×105) and 4 ×105 CFU/ml (IQR 2.6 ×105 - 5 ×105) for immunocompetent isolates versus 3.3 ×105 CFU/ml (IQR 2.9 ×105 - 4.5 ×105) and 2.8 ×105 CFU/ml (IQR 2.2 ×105 - 3.8 ×105) for HIV derived isolates, P=0.004 and 0.002, respectively). The difference in the ability of isolates derived immunocompetent patients to grow at 37°C compared with 30°C did not reach statistical significance(P=0.058); however, there was statistically significant reduction in growth of isolates from HIV-infected patients at 37°C (P = 0.044) (Figure 2).
In contrast, isolates derived from immunocompetent patients had slower growth in ex vivo CSF compared with isolates derived from HIV-infected patients, with median fungal burdens of 7×103 CFU/ml (IQR4.9 ×103 - 1.0×104) versus 1.9×104 CFU/ml (IQR 6×103 - 2.6×104) after 48 hours, P=0.02 (Figure 2). The isolates derived from immunocompetent patients also elaborated significantly thinner capsules (median thickness 1.3 microns, IQR 0.8 - 1.6) compared with isolates from HIV-infected patients (median 1.5 microns, IQR 1.3 – 1.95), P<0.001).
Low virulence VNIa-5 C. neoformans isolates can be induced to become highly virulent
We hypothesized that the development and/or maintenance of the increased virulence state was due to inter-yeast communication. To test this hypothesis we cultured an environmental VNIa-5 isolate of low virulence in media supplemented with sterilised culture filtrate (sCF) extracted from culture of VNIa-5 isolates with high virulence from immunocompetent patients. Virulence phenotypes were again compared in Galleria. Culture in media supplemented with sCF from all four organisms from immunocompetent patients resulted in increased virulence of the environmental isolates in comparison to their ‘naïve’ state; the hazard of death after induction increased between 2.2 and 2.9-fold (Figure 3). This change in virulence phenotype of the induced environmental isolates was stable over multiple Galleria infection cycles (Figure S2). Conversely, culture in media supplemented with pooled sCF from HIV-associated VNIa-5 isolates had no significant effect on the virulence phenotype of environmental isolates (HR1.4, 95CI 0.8-2.3, P=0.2, ‘induced’ isolate versus naïve isolate; Figure 4).
We repeated the above experiment but cultured an environmental VNIa-5 isolate in media supplemented with sCF from VNIa-4 isolates (necessarily) derived from HIV-infected patients. These conditions did not result in any significant variation in virulence. Similarly, culturing the VNIa-5 environmental isolate in media supplemented with sCF from a previous culture of the same organism, or from the H99 type strain, had no effect on virulence (Figure 4).
Culture filtrate from an induced VNIa-5 isolate will increase virulence in its naïve isogenic self
We next measured the effect of growing a ‘naïve’ low virulence VNIa-5 environmental isolate (LD2) in media supplemented with sCF from its previously ‘induced’ higher virulence self. We cultured the naïve LD2 isolate in culture medium supplemented with sCF from a highly virulent isolate derive from an immunocompetent patient (BMD761) for 48 hours. The cultured yeast cells were harvested by centrifugation, washed, and innoculated onto solid media for single colonies. We termed this isolate ‘iLD2’. iLD2 was again cultured in liquid media for 48 hours and the culture filtrate harvested. The naïve LD2 isolate was cultured in media supplemented with the sCF from iLD2 for 48 hours. We termed the resulting isolate ‘iLD2-induced LD2’. The virulence phenotypes of the isolates from each experiment in the Galleria model – i.e. ‘LD2’, ‘iLD2’, ‘iLD2-induced LD2’ and BMD761 were compared. This experiment demonstrated that, like sCF from BMD761, sCF from iLD2 was itself able to induce an increased state of virulence in its naïve self (nLD2): HR 2.8 (1.7-4.4), P<0.001 (Figure 5).
Virulence induction is associated with known virulence markers
We assessed the effect of culture filtrate induction on known virulence-associated phenotypes in Cryptococcus by comparing the environmental isolated LD2 before (naïve) and after induction (induced). There was no change in thermotolerance or melanization. However, the induced LD2 had superior growth after 48 hours incubation in ex vivo CSF compared with its naïve self (median 5.9 ×104 CFU/ml, IQR 3.6×104 – 7.6×104 vs 3.5 ×104 CFU/ml, IQR 2.3 – 5.7 ×104, P= 0.02, Figure S3a). Similarly, the 48 hour fungal burden in hemolymph of Galleria infected with the induced isolate was significantly higher than in hemolymph when infected with its naïve self (median fungal burden 9.6 ×106 CFU/g body weight (inter-quartile range 6.9 ×106 - 1.4×107) versus 5.3 ×106 CFU/g body weight (IQR 4.0 ×106 - 6.7 ×106, P=0.025), Figure S3b). Induced LD2 isolates eloborated significantly thinner capsules compared with the naïve self both when grown in vitro (median capsule width 4.5 µm, IQR 3.3-6.4 versus 7.2 µm, IQR 4.5 – 9.8, P<0.001, induced versus naïve respectively) and when recovered from larval hemolymph (P=0.03, see Figure S4). Both induced LD2 and naïve LD2 had significantly thinner capsules than the highly virulent immunocompetent associated organism BMD761 (P<0.001 and P=0.03 respectively).
The induction phenomenon is protein mediated via extracellular vesicles
The induction effect of sCF from hypervirulent BMD761 on the naïve environmental organism LD2 was maintained after storage at -20°C for 28 days. The same observation was made for sCF treated with RNase or DNase. However, boiling sCF, or treating sCF with Proteinase K, abolished the induction effect (Figure 6). We further hypothesized that the induction phenomenon was associated with extracellular vesicles (EVs), which have been implicated in the pathogenic potential of Cryptococcus neoformans and the related organism Cryptococcus gattii13-15. Consequently, we added bovine serum albumin (BSA), which disrupts EV function, to sCF from BMD761 and measured the induction effect on LD213,16. Adding BSA resulted in the loss of the induction effect with no increase in the virulence of the recipient isolate versus its naïve self (HR 0.8; 95CI 0.5, 1.3, P=0.32, Figure S5).
We then generated EV-free supernatant and EV-containing pellet fractions (as shown by electron microscopy) from both BMD761 derived sCF and previously induced LD2 sCF (Figure 8). Incubation of naïve LD2 in media supplemented with supernatants did not result in induction of increased virulence (BMD 761 supernatant: HR 1.2, 95CI 0.8, 1.5; P=0.3, Figure 7B and iLD2 supernatant: HR 1.2, 95CI 0.9, 1.6; P=0.1 see Figure 7D). Conversely, when a naïve LD2 was grown in media supplemented with the EV-containing pellets there was induction of increased virulence (BMD 761 pellet: HR 1.8, 95CI 1.4, 2.4; P<0.001, Figure 7A and iLD2 pellet: HR 2.1, 95CI 1.6, 2.7; P<0.001, see Figure 7C).
Induction is associated with the expression of genes encoding virulence factors and hypothetical proteins
We used RNA-seq to investigate the transcriptional basis of the observed phenotypic change. High quality RNA sequences were generated from four organisms with a confirmed phenotype: an environmental VNIa-5 (low virulence and termed naïve LD2 (nLD2)), a clinical VNIa-5 from an HIV-uninfected patient (BMD761), an environmental VNIa-5 in a state of increased virulence after growth in media supplemented with sCF from clinical VNIa-5 (induced LD2: iLD2) and the naïve environmental VNIa-5 isolate (nLD2) grown in media supplemented with its own sCF from a previous culture of its naïve self (self-induced LD2: siLD2) (Table S2).
The majority of transcriptome replicates clustered by their sample type on a PCA plot of read counts per gene (Figure S6). There was a notable difference in global gene expression between BMD761 and nLD2. The overall transcriptional profile of iLD2 was closer to that of BMD761 after induction. The ‘self-induced’ LD2 also differed in the transcriptional profile from the naïve LD2, was distinct from iLD2, and did not converge with the expression profile of BMD761.
We next performed differential gene expression analysis to identify genes that were significantly (Wald test, Benjamini-Hochberg adjusted P value <0.05, log2 fold change ≥1 and ≤-1) up or down regulated. We filtered 784 differentially expressed genes between BMD761 and naïve LD2; after induction with BMD761 sCF; 244 genes were differentially expressed between BMD761 and induced-LD2 (Figure 9). This contrasts with the large number of differences between these 2 isolates and LD2 in the other states (naïve or self-induced), where we found there were >1500 DEGs (Figure 9).
We aimed to identify upregulated genes in both BMD761 and induced LD2 compared with naïve LD2 that may explain the phenotypic difference. Focussing on genes which were experimentally known to be associated with virulence17, we identified 17 up-regulated virulence genes in both BMD761 and induced LD2 relative to naïve LD2 (Supplementary Table S3). Specfically, we found that QSP1, which encodes a secreted peptide that has previously been implicated in quorum sensing and the upregulation of virulence of C. neoformans, was significantly downregulated in both the highly virulent BMD761 isolate and the induced environmental isolate versus its naïve self (Supplementary Table S3)18.
Discussion
A number of lineages of Cryptococcus neoformans cause meningitis in humans in Vietnam, but the vast majority of cases are due to either VNIa-4 or VNIa-52. We have previously shown that the VNIa-5 lineage is strongly associated with disease in apparently immunocompetent patients2,3. Both VNIa-5 and VNIa-4 frequently affect our (severely immunosuppressed) HIV-infected patients2,3. VNIa-4 rarely causes disease in HIV-uninfected patients, and when it does, the patients are significantly more likely to have underlying diseases associated with immunosuppression9. Despite this difference in ability to infect hosts of different immune competencies, mortality rates are no different in HIV-infected patients whether they are infected with VNIa-4 or VNIa-53. Our work here goes some way towards explaining these observations. We have found that the VNIa-5 lineage displays significant variability in virulence phenotype in the Galleria model of infection, and that this is associated with the particular isolate’s ecological background – isolates from immunocompetent patients are more virulent than those from HIV-infected individuals or those recovered from the environment. The VNIa-5 isolates derived from immunocompetent patients are also significantly more virulent than VNIa-4 isolates, necessarily derived from HIV-infected patients.
The differences in virulence phenotype of the VNIa-5 isolates are not explained by a specific within-lineage substructure - virulent isolates from immunocompetent patients do not cluster within, but are distributed throughout, the VNIa-5 phylogeny2. This suggests that any isolate of the VNIa-5 lineage should have the capacity to express a highly virulent phenotype, and therefore infect immunocompetent patients. Consistent with this, we found that the VNIa-5 virulence phenotype is highly plastic – less virulent isolates, derived from the environment, can be induced into a more virulent state by growing them in culture media that has been supplemented with sCF from highly virulent isolates, derived from immunocompetent patients. While we found HIV derived VNIa-5 isolates were somewhat more virulent that the environment-derived isolates, their sCF lacked this capacity and could not induce increased virulence. The induction effect was also limited to sCF derived from the VNIa-5 lineage – sCF from the VNIa-4 lineage, or from the VNIb H99 type strain, did not have the induction property. Therefore, we hypothesize that the induction effect seen with culture filtrate from immunocompetent patient-derived VNIa-5, and the virulence plasticity of the lineage, are intimately associated with the ability of the lineage to cause disease in the immunocompetent host.
We found the increased virulence of VNIa-5 derived from immunocompetent patients was associated with differences in in vitro virulence phenotypes compared with isolates from HIV-infected patients, although the relationship was not simple. Isolates from immunocompetent patients had faster rates of growth at 30 and 37°C, but had slower growth in ex vivo cerebrospinal fluid and expressed thinner capsules. Induction of the higher virulence state in the environmental isolate LD2 was also associated with increased fungal burdens in Galleria, and expression of thinner capsules both in vitro and in the Galleria model. However, the induced LD2 isolate grew more rapidly in ex vivo CSF compared with its naïve self. It is not clear why the changes in in vitro phenotypes on induction were not entirely consistent with our findings when comparing isolates by source. It may reflect the fact that these measures of virulence are largely validated in relation to the ability to cause disease in humans irrespective of immune background, i.e. including those with HIV infection. Or, it may be that the induction phenomenon we have identified here is only a part of the mechanism that is associated with disease in immunocompetent humans. However, our transcriptional data suggest thet the change in virulence we are measuring in Galleria is real and relevant, because i) it is associated with increased expression of a number of virulence associated genes and ii) we saw a significant and remarkable convergence in gene expression between the highly virulent BMD761 isolate and the induced environmental isolate LD2.
The induction phenomenon we see is robust and repeatable. We do not believe it is an artefact of ‘carry-over’ of virulent organism from the sCF donor isolate, because we used 0.45 micron filters, which should exclude any donor cells, and every culture filtrate was tested for sterility by culture. Furthermore, the induced state is stable following purification by culture on solid media. We can be certain that the convergence of the transcriptomic data from BMD761 and iLD2 is not confounded by contamination because we identified SNPs unique to each isolate within the context of nearly 900 C. neoformans genomes2, and verified the identity of the strains in the RNA-seq data using these markers. Similarly we also do not believe that the effect on virulence in Galleria is due to ‘carryover’ of some agent (eg extra-cellular vesicles) from the donor isolate sCF into Galleria, because induced isolates retain their increased virulence following purification on solid media.
The induction effect appears to include a protein or small peptide component, since the inducing effect of sCF is lost with both boiling and protease treatment. A secreted peptide – QSP1 – has previously been shown to be involved in quorum-sensing and virulence regulation of C. neoformans18. Interestingly, we found that expression of this gene is down-regulated when the low virulence environmental strain is induced into the high virulence state and thus it is unlikely that this gene explains the phenomenon we observed.
Our data also suggest that the induction effect is associated with extra-cellular vesicles – the addition of albumin, which disrupts Cryptococcus vesicles, to sCF inhibited virulence induction16. We used an ultracentrifugation protocol13 to isolate EVs and showed that the induction effect was maintained in the EV containing pellet, and lost in the EV-free supernatant. We confirmed the presence and absence of EVs in the pellet and supernatant respectively with transmission electron microscopy. The populations and compositions of EVs secreted by eukaryotic cells are complex. In addition to proteins they can contain small nucleic acids13. While we found no effect of nuclease treatment on the property of sCF, it is possible that we may have missed an effect because any small nucleic acids may have been protected within vesicles.
A striking feature of the induction effect is the stability of the resulting phenotype, and the transmissibility of the phenomenon. One possible explanation is that the effect we are observing is an infection event by a mycovirus. To investigate this possibility we interrogated all unmapped reads from our whole genome and transcriptome sequence data using a metagenomic platform ((http://taxonomer.com, data not shown). We found no evidence of viral infection. A further possibility is that the effect is prion mediated – an EV-associated prion (Sup35p) has been implicated in cell-to-cell communication in Saccharomyces cerevisiae, although the effect is not well-conserved across species19,20. However, the fact that the effect of sCF we observed was heat labile suggests a prion-driven mechanism is unlikely.
On balance, we believe the phenomenon we have described is mediated through EVs associated with one or more peptides/proteins. Small nucleic acids may also be involved. An EV-dependent mechanism is plausible - EV production has been demonstrated for a number of fungal pathogens, including Cryptococcus14,15,21-23. A specific role in the pathogenicity of C.neoformans was first postulated in 200822, and EVs have been shown to contain virulence-associated factors including capsular components and laccase14. In addition to proteins, EVs are also associated with small DNAs and RNAs which could influence virulence13,23-25. More recently, EVs have been shown to mediate the virulence of an outbreak strain of the related species C.gattii13. The EV-mediated effect was dose-dependent, manifested as increased growth, and required both EV-associated protein and RNA, providing strong evidence of EV uptake enabling inter-yeast communication13. The phenomenon differed from our observations in that it was seen in macrophage culture, and onward transmissibility of the phenotype was not reported13.
Where and when the virulence induction phenomenon occurs in relation to human disease is unclear. It may be a rare event that occurs somewhere in the environment, perhaps as a result of interaction with other micro-organisms, or a consequence of animal infection. If this is the case, we would on occasion expect to see such inducing, high virulence isolates in HIV-infected patients. To confirm this we would need to characterise large numbers of isolates from HIV-infected patients. Alternatively, induction might occur in humans themselves, although if so this is probably not driven by innate immunity, since the virulence phenotype was extremely stable in Galleria26-29.
A great challenge in treating HIV-uninfected patients with cryptococcal meningitis is knowing when it is is safe to stop antifungal therapy. Cure may take many months of treatment, and patients may even require life-long antifungal suppressive therapy30. This contrasts with HIV-infected patients, who have a modifiable immune pathology - once antiretroviral therapy has allowed immune recovery antifungal treatment can be stopped safely30. Inter-yeast signalling systems, associated specifically with changes in virulence, are potential therapeutic targets. Disrupting such systems might allow shortened treatment courses or even cure in immunocompetent patients. However, they are only relevant if active in human disease. Currently, it is extremely difficult to obtain sufficient volumes of clinical samples to reliably test hypotheses concerning pathogenesis (generated from laboratory observations) at the site of human disease31. However, the recent development of a novel cerebrospinal fluid filtration technology – neurapheresis – may revolutionise this. Neurapheresis, developed to treat hemorrhagic stroke, can efficiently remove large amounts of pathogen biomass from the cerebrospinal fluid of rabbits experimentally infected with Cryptococcus32-34. Such a system, itself a potential treatment for human cryptococcal disease, could allow confirmation that the phenomenon we describe here is active in human disease.
Conclusion
The virulence phenotype of Cryptococcus neoformans VNIa-5 in the Galleria infection model is associated with the isolate source. Isolates from immunocompetent patients are the most virulent; those from the environment the least. The virulence phenotype is highly plastic and moderated by inter-yeast signalling through the secretion of proteins or peptides associated with extracellular vesicles. Induction of virulence is accompanied by transcriptional re-modelling and convergence of gene expression between the induced and inducer isolates. The induction phenomenon is VNIa-5 specific, and likely key in the lineage’s ability to cause disease in the immunocompetent host. Better understanding this mechanism may reveal novel therapeutic targets.
Funding
Funded by a Wellcome Trust Intermediate Fellowship to JND Grant number: WT097147MA
Acknowledgements
Electron microscopy was performed at the Dunn School EM Facility at the University of Oxford.
Methods
Isolates
Clinical isolates were derived from Vietnamese patients with cryptococcal meningitis who had been enrolled into descriptive and randomised intervention studies performed by our institute in collaboration with the Hospital for Tropical Diseases, Ho Chi Minh City9,35,36. All isolates were from cerebrospinal fluid obtained at the point of diagnosis and/or study entry, and stored on Microbank beads at -80°C prior to revival on Sabouraud dextrose agar. Informed consent was obtained from all participants; all studies received ethical approval through the Oxford Tropical Ethics Committee, The Hospital for Tropical Diseases, and The Ministry of Health of Vietnam. All isolates have been previously sequenced and the data are publically available2. Isolates of the lineage and human immune phenotype of interest (HIV-infected or not) were randomly selected from the isolate database for subsequent experiments. Environmental isolates were obtained via randomised sampling of air, soil and tree bark in urban and rural areas within and around Ho Chi Minh City. All strains used were mating type alpha.
In vitro phenotyping
Temperature – dependent growth
Growth at high temperature and in ex vivo human CSF were tested as previously described with modifications for quantitative assessment37. Cryptococcus spp were cultured on Sabouraud agar for 2 days at 30°C. Inocula were made from single colonies and adjusted to 108 cells/ml, then serially 10-fold diluted and spotted in duplicate onto YPD agar in 5µl aliquots. Incubation was at 30°C or 37°C for 48 hours. After 48 hours, the number of colony forming units (CFU) was counted and expressed as CFU/ml.
Capsule production
Capusle thickness was assessed through growth on Dulbecco Modified Eagle Medium (DMEM) [supplemented with 4.5g/L glucose, L-glutamine, sodium pyruvate], NaHCO3 250mM, NaMOPS 1M, Neomycin 200mg/ml, Cefotaxime 100 mg/ml38. Plates were incubated at 37°C in 5 % CO2 for 5 days. A suspension from a single colony was made in India ink visualized at 100X magnification using light microscopy (CX41, Olympus, Japan). Images were captured using a built-in DP71 camera (Olympus, Japan) and visualised using ImageJ (https://imagej.nih.gov/ij/index.html). Capsular thickness was calculated by subtracting yeast cell body diameter from the whole cell diameter. Ten individual microscopic yeast cells were assessed for each isolate.
Laccase activity
Laccase activity was assessed by adding an inoculum of 2×104 cells into 96 microwell plates containing L-DOPA medium (0.1% glucose anhydrous, 0.1% L-asparagine, 0.3% KH2PO4, 0.025% MgSO4•7H2O, and 0.01% L-DOPA, pH 5.56) with incubation for 16 hours at 37°C followed by 48 hours at 25°C with shaking at 800 rpm to induce melanin production. After incubation, the supernatants were harvested by centrifugation (4,000 g for 5 minutes) and the amount of pigment produced determined spectrophotometrically at a wavelength of 475-nm. Absorabance was converted into laccase units (U) by a factor of 0.001 OD = 1 U39. A Two biological replicates were performed for each isolate.
Urease and phospholipase production
Urease production was confirmed by spotting 10μl of a 106cells/mL cell suspension onto Christensen’s agar with incubation at 30°C. Cryptococcus neoformans H99 type strain and Candida parapsilosis ATCC 22019 were used as positive and negative controls. Cultures were observed daily for the characteristic pink discoloration.
Extracellular phospholipase activity was confirmed on egg yolk medium as previously described, with minor modifications again using a 5 µl aliquot of C. neoformans yeast suspension (106 cells/ml) with incubation at 30°C for 72 hours40. The diameter of the precipitation zone (D) formed around a colony in relation to the diameter of the respective colony itself (d) was recorded for 5 selected colonies of each isolate (N=1450 total). The D/d ratio for each isolate was calculated. H99 was included for reference in each experimental batch. The final result for each isolate was expressed as the ratio D/d ratio. Type strain H99 was used as a quality control.
Survival in ex vivo cerebrospinal fluid
CSF was prepared by pooling and filtering (Millipore 0.45 micron membranes, Merck) CSF samples taken prior to antifungal therapy from HIV-infected patients participating in clinical trials. CSF was stored at -80°C prior to use. An inoculum of 106 cells/mL of each isolate of interest was prepared using a Cellometer X2 (Nexcelom) cell counter. 10 uL of inoculum was added to 90 uL of CSF and PBS followed by incubation at 30°C for 3 days. After 72 hours the resulting culture was serially tenfold diluted, spotted onto Sabouraud plates and incubated at 30°C for 3 days for quantification. Tested trains were inoculated alongside H99 and the ΔENA1 negative control (which lacks a cation-ATPase-transporter resulting in decreased viability in human CSF and macrophages, strain provided by the Perfect Lab, Duke University)37. Two biological replicates were performed per isolate.
Galleria infection model
Late instar isogenic wild type Galleria mellonella larvae (30 days from oviposition of the adult moth) were sourced from U U Animal, (Ho Chi Minh city, Vietnam). All larvae were kept at 16°C in bran in the dark. Larvae for experimentation were selected according to the following criteria: weight 250-300 mg, healthy colour (beige). Yeasts were revived from frozen stock (Microbank Beads) on SDA plates for 72 hours. Single colonies were picked and cultured in YPD broth for 24 to 48 hours in a shaking incubator (SI-300, Jeio Tech), at 150 rpm at 30°C until the concentration of cells was at least 108 cells/mL. The pellet was collected with centrifuge at 8000 rpm in 1 minute and subsequently washed twice with PBS. The cell suspension was adjusted to a density of 108 cells/mL using a Cellometer X2 (Nexcelom Bioscience, USA). Ten μL of inoculum (106 cells/larva) were used for injection per larva. All inocula were relabeled by an independent person such that the operator/assessor was blind to the nature of the inoculum used in each batch of larvae (i.e. the survival experiments were blinded). Larvae were inoculated through injection using a sterile Hamilton syringe into the most caudal left pro-leg. 15-20 larvae were infected per cryptococcal isolate. Every survival experiment was internally contemporaneously controlled using larvae of the same batch, a negative control (PBS) (physical injury and sterility, data not show in survival plots), and uninjected larvae and the 2 or more experimental arms of interest. Infected larvae were incubated at 37°C for ten days and checked daily for mortality using physical stimulation with forceps.
Culture Filtrate preparation
All isolates were grown in 7.5 ml YPD broth at 30°C with shaking for 48 hours followed by centrifugation at 8000 rpm for 1 minute. The resulting supernatant was filtered using a Millipore membrane filter 0.45 μm (Merck), and checked for sterility by plating onto Sabouraud’s agar and into BHI broth with incubation at 30°C for seven days.
Induction experiments
2.5 mls of the culture filtrate of interest was added to 5mls of sterile YPD broth (total volume 7.5mls). A single purified colony of the isolate of interest was added to the resulting culture medium and incubated for 48 hours with shaking (150rpm) at 30°C. Cells were separated by centrifugation (1 minute at 8000 rpm), washed twice with PBS and pellet spread on Sabouraud agar plate for single colony purification. Single colony was then inoculated into YPD broth, and incubated with shaking for 48 hours for inoculum preparation.
Confirmation of virulence phenotype stability in Galleria
Yeast was recovered from Galleria hemolymph, cultured for purity on SDA, recultured in YPD and reinoculated into Galleria as described 6 fold. Similarly purified culture derived from hemolymph was repeatedly subcultured on solid media, frozen on Microbank beads, revived, and re-inoculated into the Galleria model.
Quantification of fungal burden in larvae
Larvae infected with the strains of interest (5 dead larvae/biological replicate) were collected at the point of death and gently squeezed to obtain 1 gram of hemolymph and fat body using a sterile knife. Glass beads (3mm) and 1 mL of sterile water were added for homogenization at 30Hz for 10 minutes (TissueLyser II, Qiagen). Homogenates were diluted with PBS and inoculated onto SAB plates and incubated for 3 days at 30°C. The number of colonies were counted to calculate the CFU/gram/larvae.
Capsule size measurement from Galleria
Following infection with 106 cells, larvae were sacrificed 48 hours post-infection and hemolymph extracted as above. Hemolymph was stained with India ink to visualise capsule using light microscopy (at 1000X magnification) with an Olympus DP71 digital camera (Olympus, Japan) and the proprietary Image-J software.
Treatment of culture filtrate with freezing, heat, protease, and nuclease
To check the stability of culture filtrate exposed to prolonged frozen storage, the culture filtrate of BMD761 was kept at -20 °C for 33 days. The filtrate was brought to room temperature prior to induction experiments. To denature protein by heat, filtrate of BMD761 was boiled at 97 °C for 2 minutes. To denature protein chemically, filtrate was incubated with Proteinase K (Sigma-Aldrich, UK) at a concentration of 100 μg/mL in the presence of 30 mM Tris.HCl (pH=8) and 5 mM CaCl2 at 37 °C for 1.5 hours. To degrade DNA, 28 μL DNAse I (Ambion, ThermoFisher, UK) was added to 700 μL DNAse I buffer and 7 mL filtrate for 1 hour at 37°C. To remove RNA, 2.5 mL filtrate was incubated with 50 μL RNAse cocktail of RNAse A and RNAse T1 (Ambion, ThermoFisher Scientific, UK) for 1 hour at 37°C.
Confirmation of digestion of protein and nucleic acid in filtrate
Protein digestion was confirmed by running treated and untreated culture filtrate on precast polyacrylamide gel 4-20 % (Mini-PROTEAN® TGX Stain-Free™ Precast Gels, BIO-RAD) at 200V for 40 mins. Precision Plus Protein Dual Xtra Standards (BIO-RAD) were loaded alongside samples, including a proteinase K treated ladder.
Samples treated with nuclease were confirmed by loading in agarose gel 1.5 % followed by electrophoresis for 60 mins at 100V, with 100bp DNA ladder.
Extracellular vesicle preparation
Isolates for experimentation were grown to stationary phase in 500ml of YPD broth at 30°C with shaking for 48 hours prior to transfer to 200mL centrifugation vessels. These were spun at 15,000g for 10 minutes at 4°C. Supernatant was decanted and centrifugation repeated for a further 10 minutes, then filtered through 0.45um filter membranes. The filtrate was then concentrated using Amicon 100kDa ultrafiltration columns with centrifugation at 5000g for 15 minutes at 4°C. The EV containing retentate, was then ultracentrifuged at 150,000g at 4°C for 1 hour to provide EV containing pellet and EV-free supernatant.
Electron Microscopy
Pellet and supernatant resulting from ultracentrifugation were fixed in 2% paraformaldehyde (PFA) in 0.1M sodium phosphate buffer at room temperature for 30 minutes. Samples were stored at 4 °C for up to 3 days before negative staining for Transmission Electron Microscopy (TEM). For negative staining, 10 µl of sample was applied to a freshly glow discharged carbon filmed 300 mesh Cu grid for 2 mins, blotted and stained with 2% uranyl acetate for 10 sec, then blotted and air dried. Grids were imaged in a Tecnai 12 TEM (Thermo Fisher) operated at 120 kV using a Gatan OneView camera.
Preparation of inducing broth using ultracentrifugation product
Inducing broth consisted of 50uL ultracentrifuged product (either pellet or supernatant fraction) added to fresh sterile YPD to achieve a final volume of 7.5 mL. A 100uL volume was taken for culture as a test of sterility. Naive strains (a single colony each) were then inoculated into inducing broth and incubated for 48 hours with shaking at 30°C. Yeast were harvested in the normal manner via centrifugation 8000 rpm and washed twice with PBS prior to subculture on solid agar. Inoculum for larva infection (108 cells/mL) were prepared in PBS using a Nexelom cell counter; 10 uL (106 cells) were injected into the larvae. Inoculum densities were checked through serial dilution and culture.
RNAseq
Three replicates of each strain were cultured in YPD broth for 96 hours at 30°C to determine log phase for RNA extraction. An inoculum of 106 cells was grown in 7.5 mL of YPD broth by agitating (200 rpm). At interval time point post-inoculation (0, 6, 20, 24, 30, 44, 54, 68, 72 and 96 hours), an aliquot of 50 uL was taken for quantification by plating in Sabouraud agar plates. These plates were exposed to incubation at 30°C for 48 hours. Growth curves were plotted using ggplot package and R version 3.4.0 software (R Foundation for Statistical Computing, Vienna, Austria). Growth rates were determined using the formula growth rate=(log10Nt-log10No)*2.303/t-to, where N indicates cell concentration at a particular time point; and t indicates the time point).
For transcriptional experiments, single colonies of each isolate of interest were revived from beads (Microbank, UK) stored at -80°C. Revived isolates (approximately 106 cells, counted using a Cellometer) were grown in 7.5 ml of YPD broth by agitating for 18 hours at 30°C to reach mid-log phase. RNA was harvested using the RiboPure™-Yeast RNA Isolation Kit (Ambion, USA) according to the manufacturer’s instructions. Extracted RNA samples were subjected to quality control using an Agilent 2100 Bioanalyzer (Agilent, USA). The integrity of total RNA was estimated using the RIN (RNA Integrity Number), which ranges from 1 to 10, with 1 being the most degraded. Samples with RIN greater than 7 qualified for RNA-Seq. 6 biological replicates were prepared of every experimental condition. RNA was eluted in isopropanol and shipped at room temperature to Macrogen (Seoul, Korea) for library preparation and sequencing.
Paired-end RNA-Seq libraries were constructed using the TruSeq stranded mRNA preparation kit (Illumina, USA). Resulting cDNAs were ligated with sequencing adaptors and sequenced using the Illumina HiSeq 2000 platform, generating ∼11.2 million reads of 150 bp per sample, resulting in an estimated 177-fold coverage.
RNA sequencing analysis
Raw reads were checked for quality using FastQC (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Adapters were removed using scythe (https://github.com/vsbuffalo/scythe). The resulting reads were then further checked based on Phred score and read length in order to trim low quality regions using Trimmomatic (http://www.usadellab.org/cms/?page=trimmomatic). Sequenced reads were aligned with the C. neoformans var. grubii H99 reference genome using HISAT2 (https://ccb.jhu.edu/software/hisat2/index.shtml). SAM (Sequences Alignment Map) files, output of read alignment, were converted to BAM (Binary Alignment Map) files using SAMtools (http://www.htslib.org). Finally featureCounts was used to count reads assigned to genomic features (genes, exons, chromosome locations) on the reference genome41. The output table of raw counts per gene for each sample was imported for gene expression analysis using DESeq42. Library size differences were normalized internally43.
Metagenomic analysis
Taxonomer, a metagenomics-based pathogen detection tool (available at http://taxonomer.com), was used to profile unmapped mRNA expression reads and unmapped DNA from (previous) genome sequencing from the Cryptococcus isolates. The web-based interface was used to visualize the summary of taxonomic composition and read abundance.
Statistical data analysis
All statistical analyses were done using R software R version 3.4.244. Kaplan-Meyer survival curves were plotted using ggplot2 and survminer package45,46. The Log-rank test was used to compare differences in survival. Hazard Ratios were estimated using the Cox proportional-hazards model. The Wilcoxon rank sum test was used to compare fungal burdens between isolates. Capsule size was compared using the Kruskal-Wallis test with P values adjusted for multiple comparisons with the Benjamini-Hochberg method.
Supporting Information
Supplementary Figures
Footnotes
Trieu Phan Hai, trieuph{at}oucru.org, Thanh Lam Tuan, thanhlt.311088{at}gmail.com, Duong Van Anh, anhdv{at}oucru.org, Trinh Nguyen Mai, maitrinh473{at}gmail.com, Lan Nguyen Phu Huong, bshuonglan{at}gmail.com, Guy E Thwaites, gthwaites{at}oucru.org, Errin Johnson, errin.johnson{at}path.ox.ac.uk, Nguyen Van Vinh Chau, chaunvv{at}oucru.org, Stephen Baker, sgb47{at}medschl.cam.ac.uk, Philip M Ashton, pashton{at}oucru.org, Jeremy N Day, jday{at}oucru.org