Transcriptional Profiling of Patient Isolates Identifies a Novel TOR/Starvation Regulatory Pathway in Cryptococcal Virulence.

Human infection with Cryptococcus causes up to a quarter of a million AIDS-related deaths annually and is the most common cause of nonviral meningitis in the United States. As an opportunistic fungal pathogen, Cryptococcus neoformans is distinguished by its ability to adapt to diverse host environments, including plants, amoebae, and mammals. In the present study, comparative transcriptomics of the fungus within human cerebrospinal fluid identified expression profiles representative of low-nutrient adaptive responses. Transcriptomics of fungal isolates from a cohort of HIV/AIDS patients identified high expression levels of an alternative carbon nutrient transporter gene, STL1, to be associated with poor early fungicidal activity, an important clinical prognostic marker. Mouse modeling and pathway analysis demonstrated a role for STL1 in mammalian pathogenesis and revealed that STL1 expression is regulated by a novel multigene regulatory mechanism involving the CAC2 subunit of the chromatin assembly complex 1, CAF-1. In this pathway, the global regulator of virulence gene VAD1 was found to transcriptionally regulate a cryptococcal homolog of a cytosolic protein, Ecm15, in turn required for nuclear transport of the Cac2 protein. Derepression of STL1 by the CAC2-containing CAF-1 complex was mediated by Cac2 and modulated binding and suppression of the STL1 enhancer element. Derepression of STL1 resulted in enhanced survival and growth of the fungus in the presence of low-nutrient, alternative carbon sources, facilitating virulence in mice. This study underscores the utility of ex vivo expression profiling of fungal clinical isolates and provides fundamental genetic understanding of saprophyte adaption to the human host.IMPORTANCE Cryptococcus is a fungal pathogen that kills an estimated quarter of a million individuals yearly and is the most common cause of nonviral meningitis in the United States. The fungus is carried in about 10% of the adult population and, after reactivation, causes disease in a wide variety of immunosuppressed individuals, including the HIV infected and patients receiving transplant conditioning, cancer therapy, or corticosteroid therapy for autoimmune diseases. The fungus is widely carried in the soil but can also cause infections in plants and mammals. However, the mechanisms for this widespread ability to infect a variety of hosts are poorly understood. The present study identified adaptation to low nutrients as a key property that allows the fungus to inhabit these diverse environments. Further studies identified a nutrient transporter gene, STL1, to be upregulated under low nutrients and to be associated with early fungicidal activity, a marker of poor clinical outcome in a cohort of HIV/AIDS patients. Understanding molecular mechanisms involved in adaptation to the human host may help to design better methods of control and treatment of widely dispersed fungal pathogens such as Cryptococcus.

INTRODUCTION

Cryptococcus neoformans is a major fungal pathogen causing a highly lethal meningoencephalitis (CM), primarily in individuals with impaired host cell immunity, such as those infected with HIV/AIDS, causing a quarter of a million deaths annually (1, 2). Mortality exceeds 20 to 50% despite therapy (3), and recent failed attempts to improve outcomes (4) highlight our profound lack of understanding of the pathophysiology of human infections. While studies in host models such as mice or invertebrates have generated essential insights into fungal virulence, evolutionary differences in host response and pathogen environment suggest a need to model studies using human disease correlates (5).

Peculiar to Cryptococcus is the ability to live for extended periods both within the environment (6) and within the mammalian host in a dormant state (7). It can also cause disease in a wide variety of plants (8), free-living amoebae (9), and humans (10). Similarities between mammalian host environments such as the macrophage phagolysosome and those more primitive organisms such as amoebae have been implicated as exerting evolutionary pressure on facultative intracellular pathogens, such as C. neoformans (11), and gene expression studies suggest adaptation to nutrient deprivation is important within this environment (12). Indeed, requirements for the gluconeogenesis enzyme Pck1 (13) and a high-affinity glucose transporter (14) suggest roles for nutrient homeostasis in cryptococcal virulence. Phagocytosis also stimulates a starvation response in other pathogenic fungi, such as Candida albicans, that induce a shift to fatty acids as a carbon source by upregulating the glyoxylate cycle, requiring the enzyme isocitrate lyase (15). However, the dispensability of isocitrate lyase by C. neoformans during infection suggests that species-specific pathways for starvation tolerance are also important within the mammalian host (16). In addition, induction of the nutrient-recycling autophagy pathway is particularly important for C. neoformans but less important for other fungal pathogens such as Aspergillus and Candida (17). For this important stress response, global nutrient regulators such as the target of rapamycin (TOR) (18) play an important link between cryptococcal starvation response, macrophage survival, and pathogen fitness (19). In addition, downstream of TOR the virulence-associated dead box protein, Vad1, has been shown to act as a global posttranscriptional regulator of TOR-dependent processes, such as autophagy (19). Similarly, the role of the cAMP nutrient-sensing pathway has featured prominently in cryptococcal pathogenesis (20). In this pathway, adenylyl cyclase is activated by a Gα subunit (Gpa1), resulting in the production of cAMP that binds to regulatory subunit Pkr1 of the protein kinase A (PKA) complex to release an active form of the catalytic subunit Pka1, which activates downstream proteins (21). However, little is known regarding regulatory relationships that link these nutrient master regulators and gene targets responsible for mammalian virulence, as well as possible commonalities between the various infective environments, including humans.

Previous studies have not always found robust quantitative relationships between mammalian infectious outcome and classical virulence factors in vitro (22, 23), implying that other mechanisms of virulence remain to be identified. Thus, in the present work, to characterize virulence-associated starvation responses most relevant to human infections, transcriptional profiling of the fungus inoculated into human cerebrospinal fluid (CSF) was first compared to that from a starvation medium previously used to identify a clinical association between expression levels of a CTR4 copper transporter and dissemination to the brain in a cohort of human solid organ transplant patients (24). Comparison of the expression profiles of fungus incubated in human cerebrospinal fluid or under nutrient deficiency identified a large set of 855 genes upregulated under both conditions. To identify genes within this subgroup whose expression might be most relevant to human infections, we compared transcription profiles of isolates from those who died versus those surviving 10 weeks obtained from a previously described cohort of HIV/AIDS-infected patients with CM (23). The second most highly differentially expressed annotated gene (locus tag CNAG_01683 [accession no. CP022331.1]) between patient isolates showed highest homology to a sugar transporter-like gene (STL1) from Saccharomyces cerevisiae implicated previously in glycerol metabolism. STL1 expression levels showed a significant correlation with poor early fungicidal activity (EFA), a clinical prognostic marker of patient outcome (25), and in situ hybridization demonstrated STL1 expression in a brain of an HIV-infected patient with CM. Furthermore, mouse modeling was used to demonstrate attenuated virulence of an stl1Δ strain, confirming its role in mammalian virulence. STL1 was also found to be a target gene of TOR, regulated by a novel regulatory pathway involving mRNA decay and a multicomponent Cac2-Ecm15 nuclear shuttle regulatory complex. These studies thus provide a direct link between STL1, TOR signaling, starvation response, and CAF1-mediated regulation of microbial pathogenesis relevant to human fungal infections.

RESULTS

Transcriptional profiling of C. neoformans suggests similarity in adaptive stress responses between human CSF and starvation medium at 37°C.

Previous studies utilized nutrient-deficient media to simulate the environment of the host CSF and successfully correlated expression levels of a CTR4 copper transporter with dissemination to the brain in a cohort of solid organ transplant patients (24). These conditions were utilized to compare transcriptional profiles of a pathogenic C. neoformans strain (H99, serotype A, molecular types VNI and VNII) (26) between nutrient-rich and nutrient-poor environments and to compare the starvation adaptation profiles to that obtained after transfer to pooled human cerebrospinal fluid from patients with cryptococcal meningitis (see Tables S1 and S2 in the supplemental material). Serotype A strains are the predominant strains that cause HIV-associated CM and were selected to reduce potential genetic confounding. A strong correlation was demonstrated between expression-level changes between YPD and either CSF versus starvation medium at 37°C (Fig. 1A) (R2 = 0.79; slope non-zero P value of <0.0001). Transcriptional profiles after transfer to each respective condition demonstrated not only common identities of upregulated genes, but also the comparatively similar magnitudes of transcriptional changes. Of 1,363 genes upregulated in starvation and 1,089 in CSF, 855 were upregulated under both conditions, of which 332 were annotated sufficiently to allow Gene Ontology (GO) term assignments (Fig. 1B). Compared to the overall annotated transcriptome (Fig. 1C, left panel), transition of C. neoformans to either starvation conditions or CSF resulted in an expansion of gene expression associated with the GO biological term “carbohydrate metabolic processes,” including key enzymes involved in β-oxidation, the glyoxylate cycle, and gluconeogenesis (see Fig. S1 in the supplemental material). In contrast, genes in the category “translation” were underrepresented in both media, consistent with the slow growth anticipated under these conditions. Using GO functional terms, transition of fungal cells to either starvation or CSF resulted in differential expression of genes involved in the transport function (see Fig. S2 in the supplemental material). These data thus emphasize similarities in gene expression profiles of C. neoformans between nutrient deprivation and CSF, as well as a predominance of genes involved in carbohydrate homeostasis and transport function.

FIG 1

Expression profiling demonstrates a concordance of cryptococcal genes upregulated in fungus-infected human cerebrospinal fluid and starvation. (A) C. neoformans strain H99 was grown to the mid-log phase in YPD and then subjected to starvation (Starv) in asparagine salts without glucose or sterile cerebrospinal fluid (CSF) at 37°C for 3 h. RNA was recovered and subjected to microarray analysis, and values were expressed as log2 ratios of expression in CSF/YPD. n = 2 for each condition. (B) Venn diagram of genes upregulated >2-fold and adj P < 0.05 after incubation in fungus-infected CSF or under starvation conditions. (C) GO biological process term occurrence among genes upregulated under the indicated conditions. (D) Transcriptional ratio of STL1/ACT1 determined by qRT-PCR of patient isolates compared to EFA (n = 46; regression analysis, n − 2 degrees of freedom [df], P < 0.05). (E) Expression ratios of STL1/ACT1 in survivors at 10 weeks (S [n = 35]) versus those who died (D [n = 11]). (F) Sections of a brain autopsy specimen were obtained from a 42-year-old man with HIV/AIDS who died of severe and diffuse C. neoformans infection and were visualized by bright-field microscopy (BF) or differential interference contrast microscopy (DIC) or stained with Gomori methenamine silver stain (GMS), hematoxylin and eosin (H&E stain), calcofluor white (CFW), or C3-fluorescein-labeled oligonucleotide probes for STL1 transcripts (STL1) by fluorescent in situ hybridization in the presence (+) or absence (−) of RNase A treatment. Arrows point to stained yeast cells. Bar = 10 µm. (G and H) Expression levels of STL1/18S rRNA under mid-log growth in YPD or 3 h of incubation under starvation conditions or in the presence of 5 µM rapamycin (Rap). n = 3 independent experiments. *, P < 0.05, and **, P < 0.01, by Student’s t test. Error bars represent standard deviation.

Metabolic reprogramming of carbohydrate metabolism in C. neoformans in response to incubation in fungus-infected human CSF or starvation. Pathway schemes demonstrate differential expression ratios of C. neoformans after incubation in human fungus-infected CSF or under starvation conditions (starv) relative to mid-log-phase cells grown in YPD of genes expressing enzymes active in β-oxidation, glyoxylate cycle, or gluconeogenesis. Adapted from Derengowski et al. (12). Download FIG S1, PDF file, 0.1 MB.

Expression profiling demonstrates transcriptional expansion of genes related to transporter activity (GO function). Shown is transcriptional profiling of C. neoformans cells as in Fig. 1A displayed by the most common GO function terms. Shown are the percentage of annotated GO terms relative to total count of gene annotations (left upper panel), starvation (right upper panel), CSF alone (right lower panel), or both CSF and starvation (right lower panel). Download FIG S2, PDF file, 0.1 MB.

Genome-wide expression-level changes after transition from YPD to CSF. Download Table S1, PDF file, 0.3 MB.

Genome-wide expression-level changes after transition from YPD to starvation medium. Download Table S2, PDF file, 0.4 MB.

Expression profiling of isolates from a cohort of AIDS-related CM identifies a putative sugar transporter-like gene, STL1, associated with clinical outcome marker EFA.

Further studies were then conducted to determine whether starvation-induced genes may provide associations with clinical outcome that may help to understand human-related pathogenicity of the fungus. Transcriptional profiling under starvation conditions was conducted as described below, which previously identified the copper transporter gene CTR4 as a potential biomarker of brain dissemination in solid organ transplant recipients (27) using a set of serotype A C. neoformans isolates (molecular types VNI and VNII) from a cohort of HIV/AIDS patients described previously (28). A demographic table of the cohort from which isolates were obtained is shown in Table 1. Median age was 36 years, and all patients were treated with amphotericin B-containing regimens. Despite therapy, 5 died by 2 weeks, and an additional 6 died by 10 weeks, generating a 10-week mortality of 11/45 (24%), typical of CM in developing countries (3). Pretreatment covariables were significantly different for mental status, as measured by Glasgow coma score (P = 0.02), but parameters including CSF initial fungal burden, EFA, VN genotype, CSF opening pressure, or white blood cell (WBC) or CD4 cell count did not differ significantly. Expression profiles were compared using analysis of variance (ANOVA) with adjustment, grouping patient isolates associated with mortality at 10 weeks versus survivors. Ninety-one genes (28 annotated) showed significantly different expression in patients who died versus those who lived at least 10 weeks after therapy (see Table S3 in the supplemental material) (adjusted P value [adj P] of <0.02). The first highly expressed gene was the nitroreductase family protein gene (CNAG_02692 [CP022323.1]). The second most highly differentially expressed gene showed highest homology to a sugar transporter-like gene, STL1, from S. cerevisiae previously implicated in glycerol transport (29, 30). STL1 was selected for further study because of its putative role in alternative carbon metabolism and transport function exemplifying the aggregate transcriptional changes in human CSF. STL1 expression levels by quantitative reverse transcription-PCR (qRT-PCR) of the clinical isolates showed a significant correlation with poor early fungicidal activity (EFA [P = 0.018]) (Fig. 1D (25), as well as a trend with clinical outcome (Fig. 1E) (P = 0.07); however, the latter was not statistically significant, possibly due to nonmicrobiological confounders, such as immune responses or delays in medical care. EFA is the rate of fungal clearance in each of the HIV patient’s CSF during therapy and is a microbiological prognostic marker of clinical outcome (25, 31). We did not find a relationship between STL1 expression and molecular genoype (P = 0.73).In addition, STL1 expression was demonstrated in a brain autopsy specimen from an HIV/AIDS patient with CM using fluorescent in situ hybridization (FISH) (66/66 yeast cells with positive STL1 signal versus 9/78 in RNase-treated negative-control slides; Fisher’s exact test, P < 0.0001) (Fig. 1F) (19). Furthermore, qRT-PCR demonstrated that STL1 expression was induced during starvation, as well as in the presence of the TOR inhibitor rapamycin (Fig. 1G and H). TOR is a particularly important pathway as the TOR inhibitors sirolimus and everolimus are in widespread clinical use as immunosuppressants and may play a risk factor in cryptococcosis (32). In summary, these data identify a starvation-induced putative sugar transporter gene, STL1, associated with poor microbiological clearance (EFA) during HIV-associated human infections.

TABLE 1

Baseline clinical and laboratory characteristics

Parametera	Result for patients
	All	Survived	Died
n	45	34	11
Age, yrb	36 (21–62)	37 (21–62)	33 (23–51)
Male, %	53	53	55
Abnormal mental status, %	9	3	27
Fungal burden, log10 CFU/ml	5.9	5.9	5.8
EFA	−0.51	−0.55	−0.37
CSF opening pressure, cm H2Ob	26 (7–76)	26 (10–76)	15 (7–42)
Genotype, % VNI	68	58	42

CFU, colony-forming unit; CSF, cerebrospinal fluid; EFA, early fungicidal activity.

The median is shown with range in parentheses.

Genome-wide expression changes between isolates of patients who died versus survived at 10 weeks. Download Table S3, PDF file, 0.1 MB.

STL1 plays a role in cryptococcal growth on alternative carbon sources and is required for full virulence in mice.

Further characterization demonstrated a role for STL1 in expression of the virulence factors capsule (Fig. 2A) and laccase (Fig. 2B), but STL1 was dispensable for mating (see Fig. S3A in the supplemental material). The induction of capsule under the nutrient-deficient (Glu−) conditions was decreased from a small rim of capsule in the wild type (WT) but markedly increased by overexpression under both YPD (yeast extract-peptone-dextrose) and nutrient-deficient conditions. Because of its putative role in alternative carbon homeostasis and poor fungal clearance and a trend toward worse survival in clinical strains showing increased expression, we tested to see if STL1 overexpression could facilitate growth on alternative carbon sources and play a role in mammalian virulence using a mouse model. Interestingly, STL1 overexpression (Fig. S3B) facilitated increased growth in nutrient limitation medium containing 0.03% of either pyruvate, lactate or acetate—intermediates in gluconeogenesis and previously identified substrates within mammalian brains during CM infections—as well as glycerol, which has been previously described for an STL1 homolog from Candida (Fig. 2C) (29, 33). In addition, STL1 overexpression facilitated increased growth in nutrient limitation medium containing either 0.03% ribose or citrate but not equivalent concentrations of glucose versus an identical strain transformed with empty vector alone in identical copy number (Fig. 2D). These data extend the spectrum of alternative carbon substrates related to fungal STL1. However, STL1 did not facilitate growth on other sugars, including galactose, glucosamine, and rhamnose, suggesting carbon substrate specificity (Fig. S3C). In addition, the stl1Δ strain exhibited moderately reduced virulence in an intravenous mouse brain dissemination model compared to the wild-type strain (Fig. 2E, left panel). However, inoculation of C. neoformans strains overexpressing STL1 resulted in no additional increase in virulence in the highly virulent strain H99, compared with identical strains transformed with empty vector alone (Fig. 2E, right panel). Lack of increased virulence could be because the H99 strain used is already highly virulent (34). Taken together, these data suggest that STL1 is associated with alternative carbon acquisition and plays a role in capsule, laccase activity, and virulence.

FIG 2

STL1 is required for growth on alternative carbon substrates, capsule formation, laccase activity, and mammalian virulence. (A) Indicated strains were incubated on YPD or ASN without glucose (Glu−) agar for 3 days at 30°C and examined by India ink microscopy. Bar = 5 µm. (B) The indicated cells were inoculated on ASN containing 100 mg/ml norepinephrine to assay production of melanin pigment by laccase. (C) The indicated strains were diluted to an A600 of 1.0, and 1:5 serial dilutions (5 μl) were plated on YPD or ASN containing the indicated substrates (0.03%) and incubated at 30°C for 5 days. (D) In the upper panels, the indicated strains were diluted to an A600 of 1.0, and 1:5 serial dilutions (5 μl) were plated on ASN containing the indicated substrates (0.03%) and incubated at 30°C for 7 days. In the lower panels, the indicated strains were cultured with ASN broth containing the indicated substrates (0.03%), and optical density at 600 nm (OD600) was calculated at the indicated time points. (E) Mice were inoculated by tail vein (106 of the indicated cells), and progress was followed until the mice were moribund. n = 10 mice per group. P < 0.0001 (left panel) and P = 0.5 (right panel) by log rank test.

Mating and spotting assay. (A) The indicated MATα strains were coincubated with a MATa mating partner (strain KN99) on nitrogen-limiting mating medium (V8 agar) for 3 weeks at room temperature. The edges of the mating mixtures were photographed (40×). Bar = 500 µm. (B) qRT-PCR of STL1 of indicated strains grown to the mid-log phase in ASN containing 0.03% acetate (without glucose). n = 3 independent experiments. ***, P < 0.001 by Student’s t test. (C) The indicated strains were diluted to an A600 of 1.0, and 1:5 serial dilutions (5 μl) were plated on ASN containing the indicated substrates (0.03%) and incubated at 30°C for 5 to 10 days. Download FIG S3, PDF file, 0.2 MB.

The nuclear regulator CAC2 is a suppressor of STL1 and plays a role in capsule and virulence in mice.

We next sought to investigate mechanisms of starvation-associated STL1 regulation. Previous work had identified ECM15 mRNA as a regulatory target of the starvation/TOR-associated global virulence-associated deadbox (Vad1) protein (35). Although ECM15 is not known to directly regulate gene expression, a whole-proteome interaction study previously suggested that Ecm15 from the yeast Saccharomyces cerevisiae interacts with a second S. cerevisiae protein, ScCac2 (36). Cac2 is a key constituent of the CAF-1 chromatin assembly factor, which assembles histones H3 and H4 and mediates chromatin suppression of genes at subtelomeric locations and tolerance to UV irradiation in both S. cerevisiae and C. neoformans (37, 38). However, its precise role as a potential regulator remains unexplored in eukaryotes. Thus, we sought to determine if a putative VAD1-ECM15-CAC2 regulatory pathway could play a role in the starvation/TOR-dependent regulation of STL1 demonstrated in Fig. 1G and H.

We first identified a cryptococcal CAC2 putative protein sequence demonstrating 33% identity to that of the S. cerevisiae. The identity of C. neoformans CAC2 (CnCAC2) was confirmed by restoring UV tolerance to an Sccac2Δ mutant strain (Fig. 3A). To directly evaluate CAC2 in C. neoformans, the cryptococcal cac2Δ strain was generated and demonstrated increased expression of STL1 by qRT-PCR (Fig. 3B). Early studies suggested that orthologs of CAF-1 members such as Cac2 are associated with suppression of genes in the subtelomeric region (37), which include cryptococcal genes such as FRE7 (ferric-chelate reductase: CNAG_00876 [CP022325.1]), CNAG_05333 (XM_012198174.1), and the HpcH/Hpa1 family protein gene (CNAG_06874 [XM_012194223.1]) (see Table S4 in the supplemental material); however, the STL1 gene does not reside within this region (location in chromosome 11 [CP003830.1], positions 603867 to 606604). Thus, to identify a possible direct STL1 Cac2-specific DNA binding region(s), chromatin immunoprecipitation (ChIP) was performed using an anti-green fluorescent protein (anti-GFP) antibody directed against a Cac2-GFP fusion protein. These studies demonstrated Cac2 binding within a 100-bp fragment centered at −700 bp from the transcriptional start site of STL1, as well as that of FRE7 under the nutrient condition (Fig. 3C, top panel). In addition, STL1 Cac2-specific DNA binding efficiency was reduced under both starvation (Fig. 3C, middle panel) and rapamycin (Fig. 3C, bottom panel) conditions. To determine the functional significance of the binding, we tested expression levels of an Stl1-GFP fusion having serial deletions of the STL1 promoter (positions −1,000, −500, and −50 from the transcript start site) on basal transcriptional activity in C. neoformans (see Fig. S4A in the supplemental material). The plasmid containing an STL1 open reading frame (ORF) with various lengths of the STL1 5′-promoter–GFP fusion gene were transformed into wild-type or cac2Δ C. neoformans strains. Empty vector was used as a control. These results indicated the presence of basal transcription originating from the region from −1000 to −500 of STL1 under both nutrient-replete (YPD) and nutrient-poor (Glu−) conditions, corresponding to the region of Cac2 binding by ChIP. However, under the nutrient-rich condition, a subpopulation of WT cells displayed suppression using episomal constructs which replicate autonomously, which could suggest a requirement for chromatin-dependent promoter-binding factor(s) for more effective CAC2-dependent chromatin silencing (39). Thus, linear PCR fragments containing the indicated promoter, STL1 coding region, and selection marker were transformed into WT and cac2Δ strains as integrated fragments, demonstrated comigration with genomic DNA on Southern blot analysis. As shown in Fig. 3D and E, the integrated construct exhibited more effective repressed STL1 expression under nutrient-rich conditions (YPD), which was derepressed under either nutrient-poor conditions (Glu−) or in the cac2Δ mutant strain. Interestingly there was heterogeneity in Stl1-GFP expression in the cac2Δ strain, which may suggest posttranslational modification of the fusion protein.

FIG 3

Cac2 is a repressor of STL1 and capsule and overexpression reduces virulence in mice. (A) The indicated strains were diluted to an A600 of 1.0, and 1:5 serial dilutions (5 μl) were plated on YPD or on YPD with UV irradiation (UV) for 1 min and incubated at 30°C for 3 days. (B) qRT-PCR of STL1 of indicated strains grown to the mid-log phase in YPD and ASN without glucose (Glu−) media. n = 3 independent experiments. *, P < 0.05, and **, P < 0.01, by Student’s t test. (C) Chromatin immunoprecipitation (ChIP) of a Cac2-promoter complex. Nuclear extract from induced cryptococcal cells expressing a Cac2-GFP fusion protein was immunoprecipitated using an anti-GFP antibody (α-GFP Ab) or control IgG (IgG only) and assayed by PCR for the presence of the indicated regions of STL1, FRE7 promoter, or control promoter sequences of ACT1. (D and E) Effects of serial deletion of the STL1 5′-promoter (−1000 from the transcript start site) on basal transcriptional activity in C. neoformans. Linear constructs containing STL1 ORF with indicated regions of the STL1 5′-promoter–GFP fusion gene were transformed into WT or cac2Δ strains of C. neoformans, integration was confirmed by Southern blotting, and cells were grown to the mid-log phase in YPD or ASN salts without glucose (Glu−), and the population was subjected flow cytometry or microscopy (DIC or GFP fluorescence microscopy). Cells transformed with empty vector were used as control. Bar = 5 μm. (F) The indicated strains were inoculated onto the indicated media, recovered, and visualized by India ink microscopy. (G) Indicated cells were inoculated on ASN containing 100 mg/ml norepinephrine to assay production of melanin pigment by laccase. (H) Mice were inoculated by tail vein (106 cells of the indicated strains), and progress was followed until mice were moribund. n = 10 mice per group. P = 0.7 (left panel) and P < 0.0001 (right panel) by log rank test.

Effects of serial deletion of the STL1 5′-promoter (−1000 from the transcript start site) on basal transcriptional activity in C. neoformans. A plasmid containing the STL1 ORF with indicated regions of the STL1 5′-promoter–GFP fusion gene was transformed into the WT strain of C. neoformans and grown to mid-log phase in YPD (Glu+) and ASN salts without glucose (Glu−), and the population was subjected to microscopy (DIC or GFP fluorescence) (A) or flow cytometry (FITC) (B). Empty vector was used as a control. Bar = 5μm. (C) Capsule thickness across mutant strains in C. neoformans. The indicated strains were incubated on YPD or ASN without glucose (Glu−) agar for 3 days at 30°C and examined by India ink microscopy. Bar = 5 µm. Download FIG S4, PDF file, 0.1 MB.

List of genes suppressed by both ECM15 and CAC2 determined by genome-wide expression analysis. Download Table S4, PDF file, 0.1 MB.

Further characterization of CAC2 demonstrated that, similar to the STL1 overexpression strain, the CAC2 deletion strain showed increased capsule expression under nutrient-poor conditions (Fig. 3F), but there was little effect on melanin formation (Fig. 3G), mating (Fig. S3A), or virulence (Fig. 3H, left panel). However, overexpression of CAC2 resulted in markedly reduced virulence in mice (Fig. 3H, right panel), compared to an empty vector control strain, suggestive of a suppression of genes involved in virulence, such as STL1. The cac2Δ strain was thus a phenocopy of the STL1 overexpressor strain, though epistatic experiments demonstrated that STL1 overexpression in the CAC2 overexpressor strain could not restore capsule induction, suggesting that CAC2 may suppress additional genes involved in capsule synthesis besides STL1 (Fig. S4B). However, the data do suggest that CAC2 is a transcriptional repressor of STL1 and is associated with capsule expression and murine virulence.

Nuclear localization of CAC2 is nutrient dependent and requires ECM15.

Saccharomyces interactome studies (36) suggested that ECM15 may regulate CAC2 by a posttranslational interactive mechanism. ECM15 in S. cerevisiae is a nonessential poorly understood protein, proposed to be involved in cell wall competency (36, 40). Interestingly, deletion of ECM15 resulted in at least a 2-fold increased transcription (Fig. 4A) (adj P < 0.05) of 106 genes, while transcription of only 8 genes increased after CAC2 deletion, 4 of which were shared with ECM15. Interestingly, the gene showing highest suppression by both ECM15 and CAC2 was STL1 (Table S4). In addition, examination of protein sequences using cNLS Mapper (41) identified putative nuclear localization sequences in Ecm15 in the region D63 to A91 and Cac2 in the region E778 to V802 (see Fig. S5 in the supplemental material). Thus, to assess for possible regulatory interactions, as well as common nuclear localizations, in Cryptococcus and whether this could change in response to nutrient starvation or rapamycin, both proteins were tagged with GFP. These studies demonstrated colocalization of Cac2-GFP (Pearson correlation coefficient [PCC] = 0.77) and Ecm15-mCherry (PCC = 0.87) with the nuclear dye 4′,6-diamidino-2-phenylindole (DAPI) (Fig. 4B) and with each other (PCC = 0.77) under nutrient-rich conditions (YPD) with cytoplasm localization under starvation or in the present of rapamycin, consistent with a possible role as nuclear suppressors under nutrient-rich conditions and derepression under starvation conditions. Interestingly, rapamycin resulted in cytoplasmic localization of EMC15-mCherry with an additional suppression of visible fluorescence in starvation, which could be due to transcriptional suppression by VAD1, as alluded to above. Epistatic studies were next conducted, which demonstrated that the increased production of capsule in the ecm15Δ mutation could be readily suppressed by CAC2 overexpression, but ECM15 overexpression was not as effective at suppressing the increased capsule phenotype of the cac2Δ mutant (Fig. 4C), suggesting that ECM15 is hypostatic to CAC2. Additional studies demonstrated that deletion of ECM15 resulted in mislocalization of the Cac2-GFP fusion protein under nutrient-replete conditions (WT, CAC2-GFP-DAPI; PCC = 0.77) to the cytoplasm (ecm15Δ, CAC2-GFP-DAPI; PCC = 0.57), similar to that during starvation (WT, CAC2-GFP-DAPI; PCC = 0.53) or in rapamycin-treated cells (WT, CAC2-GFP-DAPI; PCC = 0.24) (Fig. 4D). These results suggest that Ecm15 is required for Cac2 nuclear targeting, leading to downstream repression of virulence-associated phenotypes and genes, including STL1.

FIG 4

Cac2 and Ecm15 shares overlapping expression profiles and undergoes nutrient- and TOR-dependent nuclear localization. (A) Venn diagram of genes showing >2× increased transcription relative to WT and adj P < 0.05 after deletion of ECM15 or CAC2. (B) C. neoformans cells expressing Cac2-GFP or Ecm15-mCherry were incubated in YPD, ASN without glucose (Glu−), or YPD containing rapamycin and observed by fluorescence (DAPI, Cac2, or Ecm15) or differential interference contrast (DIC) microscopy. Bar = 5 µm. (C) Indicated strains were incubated on YPD or ASN without glucose (Glu−), recovered, and observed by India ink microsocopy. (D) Cells expressing Cac2-GFP in wild-type (WT) or ecm15Δ strains were incubated in YPD ASN without glucose (Glu−) or YPD containing rapamycin and observed by fluorescence or by DIC microscopy. Bar = 5 µm.

Prediction of putative nuclear localization signals. Ecm15 (A) and Cac2 (B) putative nuclear localization signals were determined by cNLS Mapper. Download FIG S5, PDF file, 0.1 MB.

ECM15 plays a role in capsule, mating, and cell wall stability and is regulated by VAD1.

Further studies were then conducted to further characterize the cryptococcal ECM15 gene and characterize its relationship with starvation regulatory pathways that could play a role in regulation of genes such as STL1. The deduced amino acid sequence of the cryptococcal Ecm15 showed 66.8% identity to its respective homolog in S. cerevisiae. Susceptibility of an Scecm15Δ mutant to the cell wall active agent calcofluor white as previously reported (42) was complemented by the cryptococcal ECM15 gene (Fig. 5A). These results suggest that CnEcm15 is a functional homolog of ScEcm15p. Phenotypic studies shown in the preceding figure (Fig. 4B) demonstrated a role for ECM15 in extracellular capsule as well as mating, with a subtle effect on melanin formation (Fig. 5B). These studies thus implicate a role for ECM15 in the regulation of several virulence-associated phenotypes as well as mating of C. neoformans.

FIG 5

VAD1 is a repressor of ECM15, a regulator of capsule, laccase activity, and virulence. (A) The indicated strains were diluted to an A600 of 1.0, and 1:5 serial dilutions (5 μl) were plated on YPD medium or YPD containing 50 μg/ml of calcofluor white (CFW) and incubated at 30°C for 3 days. (B) The indicated strains were inoculated on ASN containing 100 mg/ml norepinephrine and observed after 2 days at 37°C (upper panel). Indicated MATα strains were coincubated with a MATa mating partner (strain KN99) on nitrogen-limiting mating medium (V8 agar) for 3 weeks at room temperature. The edges of the mating mixtures were photographed (40×). Bar = 500 µm (lower panel). (C) The indicated strains were diluted to an A600 of 1.0, and 1:5 serial dilutions (5 μl) were plated on YPD, ASN, or YPD containing SDS and incubated at 30°C for 3 to 7 days. (D) Mice were inoculated by tail vein (106 [left] or 103 [right] of the indicated strains), and progress was followed until they were moribund. n = 10 mice per group. P< 0.0001 (left panel) and P < 0.01 (right panel) by log rank test. (E) Lysates from cells expressing a c-myc-tagged Vad1 fusion protein incubated under the indicated conditions were immunoprecipitated followed by RT-PCR/gel electrophoresis using primers for the indicated gene transcripts. (F) Steady-state transcript levels of ECM15 from indicated strains grown to the mid-log phase in YPD or ASN salts without glucose (Glu−). n = 3 independent experiments. *; P < 0.05 by Student’s t test.

Extending these results to cell wall phenotypes associated with virulence, as suggested by studies in S. cerevisiae, the cryptococcal ecm15Δ mutants were found to exhibit increased susceptibility to the ionic detergent SDS (Fig. 5C). Interestingly the ecm15Δ mutant also displayed poor growth on nutrient-poor agar consisting of asparagine salts (ASN) without glucose (Fig. 5C), similar to that of the stl1Δ mutant. In addition, the Cnecm15Δ strain exhibited attenuated virulence in a mouse model (Fig. 5D, left panel). Interestingly attenuated virulence was also exhibited by an ECM15-overexpressing strain (Fig. 5D, right panel). ECM15 was overexpressed using an ACT1 constitutive promoter and compared with strains expressing empty vector alone in equivalent copy number, as previously described (43). As alluded to above, the cryptococcal RNA chaperone Vad1 regulates gene expression in a starvation-dependent fashion by recruiting mRNA to a decapping complex protein, Dcp2, leading to transcript degradation (19), and previous studies demonstrated that VAD1 binds ECM15 mRNA (35). In the present study, ECM15 mRNA binding to Vad1 was confirmed by qRT-PCR after radioimmunoprecipitation of a c-myc-Vad1 fusion protein (RIP-qRT-PCR) compared to that of an equivalent precipitation using an untagged strain (Fig. 5E). In addition, a vad1Δ mutant showed accumulation of ECM15 transcripts compared with the wild type from cells incubated under starvation conditions (Fig. 5F), suggesting that VAD1 is a repressor of ECM15. Taken together, these data suggest that deletion of cryptococcal VAD1 increases transcript abundance of ECM15, a regulator of capsule, mating, laccase, cell wall stability, starvation tolerance, and mammalian virulence.

In summary, these data identify a starvation-dependent epistatic regulatory pathway involving VAD1, ECM15, and CAC2 controlling growth on alternative carbon sources, expression of the virulence factors capsule and laccase, and mammalian virulence (Fig. 6).

FIG 6

Scheme of starvation regulation of STL1 related to capsule formation and mammalian virulence.

DISCUSSION

Pathogen survival and virulence entail selection for traits necessary within the infective niche. For opportunist pathogens such as C. neoformans, evolutionary pressures exerted during residence in nutrient-poor soil or free-living amoebae (9), or in the presence of diphenolic toxins within plants (8), have likely shaped expression patterns necessary for effective mammalian infection. After infection, C. neoformans is also thought to reside in the human host for a prolonged time as a latent, growth-arrested organism (7). Thus, nutrient limitation could provide a common condition that shapes adaptation and pathogen optimization within diverse environments and was the subject of the present study.

In this study, expression patterns of C. neoformans within human cerebrospinal fluid from actively infected patients was highly correlated with that induced under starvation conditions at 37°C in both identity and levels of transcription. In addition, of the 222 genes previously reported to be upregulated after incubation in the macrophage cell line J774.1 (12), 147 were also found to be upregulated in CSF in the present studies, and of the 323 genes upregulated in free-living amoeba reported in the same paper, 164 were also upregulated in CSF and similar numbers in starvation, suggesting metabolic response commonality between these infective niches. Brain infections with C. neoformans are typically associated with low CSF glucose levels (3), and the pooled patient CSF used in the present studies reflected this. Interestingly, several studies suggest that the macrophage phagolysosome is also a nutrient-deprived environment (4, 43), implicating starvation response pathways in cryptococcal virulence in a number of human infective niches. Specifically, CSF incubation resulted in reduced expression of genes related to protein translation typical of growth limitations from reduced nutrients (44). In contrast, exposure to infected human CSF resulted in increased gene expression of processes related to sugar transport and alternative substrate utilization exemplified by genes involved in β-oxidation, the glyoxylate cycle, and gluconeogenesis. This metabolic reprogramming toward more efficient utilization of alternative substrates is shared by other organisms, such as Candida albicans (45). However, these expression patterns differed from that reported from RNA isolated directly from patient CSF of two CM patients, which more closely resembled that from nutrient-replete media and could have reflected an unusually vigorous growth condition or the higher levels of CSF glucose concentrations typical of some HIV/AIDS patients (46). However, such disparate results caution that expression of diverse strains from individuals with varied clinical presentations may be required to understand heterogeneity within the human infective niche.

Thus, the present studies utilized multiple patient isolates from a cohort of previously described HIV/AIDS patients presenting with cryptococcal meningoencephalitis and treated uniformly with amphotericin-based regimens to study gene expression under the same nutrient limitation conditions described above. These studies identified an STL1 sugar transporter whose expression levels were negatively associated with low rates of clearance from the CSF and showed a trend toward increased 10-week mortality. STL1 is named as a sugar transporter gene but has also been implicated in glycerol acquisition (29, 47), as was found for C. neoformans in the present studies with additional roles in lactate, pyruvate, and acetate metabolism. Interestingly, glycerol is elevated during cell membrane degradation and brain injury (48), and acetate has been demonstrated in large amounts by magnetic imaging spectroscopy during human cryptococcal infections (49, 50), and thus, STL1 may play a role in acquisition of these metabolites during infection. Early fungicidal activity (EFA) is a research measurement of CSF microbiological clearance in patients and is an important clinical prognostic marker of mortality. It reflects both microbiological virulence and drug responses, as well as host aspects of the infection (51). Notable in the present studies was reduced expression of the virulence factor laccase in the stl1Δ mutant. High laccase expression has previously been shown to be correlated with low EFA and worse patient survival, and higher expression could have contributed to microbiological retention in the CSF (23). While elevated STL1 expression tended to suggest poor survival, large variability in expression levels between wild-type C. neoformans strains could have impacted the significance of the relationship. Such variability was previously demonstrated by other cryptococcal biomarkers, such as CTR4, which demonstrated over 100-fold difference among clinical isolates from a cohort of solid organ transport patients (24). There also exist a number of potential nonmicrobiological factors that affect mortality. For example, initial patient mental status was more frequently altered in those who died and is an important risk factor for death (3), as was found in our cohort. Other previously reported clinical risk factors such as age and opening pressure on lumbar puncture did not differ between groups in our study. The relative immune response of the host could also be an important contributor to death in primary infections with C. neoformans. However, previous studies have shown only minor associations between mortality and markers such as CD4 count, peripheral white blood cell count, or CSF white cell count (3). The present cohort did not include patients with cryptococcal immune reconstitution syndrome (cIRIS), where immune responses may be more predominant (52), or other inflammatory syndromes associated with non-HIV-infected individuals (53). In addition, increased expression alone does not necessarily identify genes important to virulence, exemplified by the virulence factor laccase, which is downregulated at elevated temperatures (54), or the highly upregulated PTP1 sugar transporter, which also demonstrated no such role in virulence (12). Other upregulated genes such as PCK1 have had variable roles in virulence, with mutants attenuated in mice (13) but not in rabbits (14). All of these considerations led us to attempt to provide additional validation of the human cohort fungal expression studies by utilizing a model with genetically defined C. neoformans isolates. These studies did show differences in virulence between wild-type and stl1Δ strains, further implicating STL1 in fungal virulence. The intravenous model was tailored to the clinical cohort that all had brain infections and focuses on pathogen-related outcomes of murine brain infection, rather than a pulmonary (intratracheal or posterior pharyngeal inoculation) model, in which survival is more related to altered lung pathology (55). However, additional cohorts reflecting different clinical scenarios may be required to fully understand the complexity of the cryptococcus-human host interaction.

The present studies also identified an important molecular pathway in starvation response and its relation to human infections by examining STL1 as a target gene of the starvation/TOR stress pathway. TOR is an important mediator of the starvation/stress response, as well as cryptococcal survival (56), and inhibitors of TOR such as sirolimus and everolimus are in widespread use in patient populations at risk for CM, including transplant recipients (57). More recent work has shown that many TOR-dependent starvation processes are regulated via mRNA stability in yeast (58), and an important TOR-dependent regulator of mRNA stability, VAD1, is also a major virulence determinant in C. neoformans (19). The present studies identified a role for TOR/VAD1-dependent regulation of a novel CAC2/ECM15 regulatory pair of nuclear factors that demonstrated starvation-dependent regulation of STL1. In C. neoformans, a deletion mutant of CAC2 was previously found to exhibit normal growth characteristics and a slight susceptibility to UV irradiation (59). In S. cerevisiae, deletion of CAF-1 subunits such as CAC2 results in increased UV sensitivity and defects in subtelomeric gene silencing (37), suggesting a role in stress response, but the present studies are the first describing the role of Cac2 as a dynamic regulator of nutrient levels. As expected of an STL1 repressor, CAC2 overexpression resulted in reduced virulence in mice. The virulence phenotype was more evident in the CAC2 overexpressor than the STL1 knockout mutant and suggests a combined effect on virulence of multiple CAC2-regulated genes, some of which may not have reached significance in the patient cohort. Another ECM15/CAC2-dependent gene, CNAG_00876 (CP022325.1), has been previously identified as a ferric-chelate reductase gene (FRE7), regulating the important iron acquisition pathway in C. neoformans (60). The FRE7 gene was also upregulated in the patient cohort, although this did not approach significance and was not further characterized. The STL1 regulatory pathway (Fig. 6) has been simplified into a linear chain for study purposes but is likely much more complex due to the presence of other known virulence pathways related to alternative carbon acquisition/glycerol acquisition and metabolism, including HOG1, calcineurin, PI3K, PLC1, and PKA1 (19, 61 – 66). Repression by CAC2 may be chromatin dependent as promoter studies demonstrated greater CAC2-dependent suppression in glucose when the STL1 promoter constructs were integrated. Indeed, previous studies suggested that the Cac2-Caf1 complex binds preferentially to modified histones that have not been reported in plasmid DNA (67) and may thus be required as part of the chromatin repressor complex (68). The related factor Cac1 has recently been shown by single-particle electron microscopy to act as a histone binding platform, linking Cac2 within the Caf1-histone assembly complex implicated in maintenance of molecular architecture (69); however, the present studies extend the role of Cac2 within the Caf1 complex from merely a housekeeping function to that of a dynamic regulator of TOR-mediated nutrient stress response. Future studies may provide more detailed structural mechanistic insight for this complex regulation in eukaryotes and highlight the utility of C. neoformans as a model eukaryote. Nevertheless, the present data demonstrate that the VAD1/ECM15/CAC2 regulatory pathway is integral to the connection of the starvation/TOR virulence response through the sugar transporter gene STL1 (Fig. 6).

MATERIALS AND METHODS

Ethics statement.

Written informed consent was obtained, and the study was approved by the Research Ethics Committee of the University of Cape Town, the Medicines Control Council of South Africa, and the London-Surrey Borders Research Ethics Committee on behalf of St. George’s University of London and by an institutional review board (IRB)-approved protocol from the National Institute of Allergy and Infectious Diseases. All experimental procedures were conducted under a protocol approved by the Institutional Animal Care and Use Committee of the Intramural Research Program of the NIAID, NIH (protocol no. LCIM12E). All experimental studies were approved by the relevant NIAID Animal Care and Use Committee, as per the Guide for the Care and Use of Laboratory Animals (70) from the National Research Council of the National Academies, Washington, DC.

Study subjects.

Subjects providing isolates for expression analysis were control participants in a randomized trial of adjunctive gamma interferon (IFN-γ) in HIV-infected patients and a randomized trial examining alternative amphotericin B combinations, both described previously (71). We attempted to standardize the antifungal drug factor impact on EFA by only selecting for inclusion patients from a single clinical trial site treated using similar study protocols with amphotericin B-based induction regimens (i.e., no fluconazole and no adjunctive IFN-γ, both of which we know lead to significantly slower/faster clearance). CSF from 2 pooled donors was provided from stored specimens under an observational protocol previously described (53). As shown in Table 1, the 45 patients had their age, CD4 count, etc., recorded. All strains were previously serotyped to be serotype A and were genotyped by multilocus sequence typing (MLST) (72).

Strains and media.

Experiments were conducted in a genetic background of C. neoformans WT strain serotype A H99 (MATα [ATCC 208821]), which was the kind gift of J. Perfect. A complete list of strains used in this study is described in Table S5A in the supplemental material. Escherichia coli DH10B (Invitrogen) was the host strain for recovery and amplification of plasmids. The fungal strains were grown in YPD (2% glucose, 1% yeast extract, 2% Bacto-peptone) or YPD agar (YPD and 2% agar). Asparagine minimum selective medium (ASN) for transformant selection and incubation of YPD-grown mid-log cells for transcriptional studies or for detection of laccase production were described previously (73). V8 juice medium was used for mating assays as described previously (74).

Strains, primers, and probes used in this study. (A) Strains used in this study. (B) Primers used in this study. (C) FISH probes used in this study. Download Table S5, PDF file, 0.1 MB.

Microarray experiments.

Isolates from patients in Table 1 or the H99 lab strain were grown to mid-log phase and then transferred to asparagine medium without glucose (1 g/liter asparagine, 10 mM sodium phosphate [pH 7.4], and 0.25 g/liter MgSO4) or human CSF pooled from 4 patients with cryptococcal meningitis (having glucose concentrations from 10 to 25 mg/dl [0.55 to 1.4 mmol/liter]) and incubated for 3 h at 37°C, and RNA was recovered as described previously (24). Serotype A H99-based microarrays (hybridization probe sequences described previously in reference 27) with two unique probes (3 replicate features per probe) for each of 6,969 transcripts (1 per locus) were used for expression profiling. Two-color cohybridizations were performed with a common reference sample in the Cy5 channel on each array. Agilent Feature Extraction software (v.11.5.1.1, protocol GE2_107_Sep09) estimated the median pixel intensity of each feature, which was then log2 transformed. Replicate RNA samples from CSF or starvation buffer conditions were generated in two independent experiments. The reference sample was grown in YPD and used both for loess normalization and as a comparison condition in ANOVA. Each feature signal was normalized by loess (locally weighted scatterplot smoothing) and averaged per locus (3 replicates of two probes for 1 transcript per locus). A mixed-effects ANOVA model (fixed effect of growth medium, random effect of array identification) was computed, and expression difference estimates for each gene were calculated for CSF or starvation medium versus YPD. For the patient isolates (Table 1), the reference pool was created from 12 of the samples. Probe signals were summarized as the median of the 3 replicate features, then loess-normalized ratios against the cognate reference signal were averaged for the two probes per locus. A mixed-effects ANOVA model (fixed effect of survival or mortality time, random effect of study group, two levels) was computed, and expression difference estimates for each gene were calculated for week 2 mortality versus survived, week 10 mortality versus survived, or both week 2 and week 10 mortality versus survived. For both experiments, the false-discovery rate (FDR) was estimated from raw ANOVA P values to compensate for multiple testing of 6,969 genes. SAS and JMP/Genomics software (SAS, Cary, NC) was used for statistical analysis.

Overexpression, disruption, and complementation of ECM15, CAC2, and STL1 in C. neoformans.

Standard methods were used for overexpression, disruption, and complementation of the ECM15, CAC2, and STL1 genes in strain H99 (MATα), as described previously using two PCR-amplified fragments and a 1.3-kb PCR fragment of the URA5 gene previously described to effect a deletion within the target coding regions and was complemented using a 1-kb of up- and downstream genomic fragment of the target genes (13, 75, 76). Complementation in all cases retained the original deletion construct. The primers used in this study are listed in Table S5B.

Ecm15-mCherry and Cac2-GFP fusion proteins.

The cryptococcal shuttle vector pORA-YP142 (77) was used to express a fusion between the Ecm15 protein and a synthetic mCherry protein (Cneo-mCherry), utilizing C. neoformans codon usage produced by standard methods (75). The plasmid was digested with NdeI and PstI, and a PCR-amplified fragment of the H99 ECM15 gene-containing promoter region was digested with NdeI and PstI and ligated into compatible sites to produce YP148. The plasmids were recovered, the sequences were verified, and the plasmids were linearized with SceI and transformed into C. neoformans H99 MATα FOA cells by electroporation using standard methods (78). All cell preparations grown on nonselective medium were assayed at the end of each experiment by simultaneous inoculation of selective and nonselective plates to verify >90% retention of plasmid. Colocalization was observed using a Leica DMI 6000B microscope with a Hamamatsu camera using LAS AF6000 v.2.1.2 software (Leica). Pearson’s correlation coefficients of colocalization were calculated on a minimum of 20 cells using Leica LASX software. Predicted nuclear localization sequences were identified in the Ecm15 and Cac2 proteins using cNLS Mapper (41).

qRT-PCR experiments.

C. neoformans H99 strains were grown on YPD or ASN without glucose. Real-time PCR was performed using primer sets as described in Table S5B. Reverse transcription was performed on DNase-treated RNA using the iScript kit (Bio-Rad Laboratories) according to the manufacturer's protocol. PCRs were set up using iQ SYBR green supermix (Bio-Rad Laboratories), according to the manufacturer's protocol. qRT-PCR was performed using a Bio-Rad iCycler (MyiQ2).

FISH.

Sections of a brain autopsy specimen were obtained from a 42-year-old male who died of severe and diffuse C. neoformans infection as reported previously (13). Fluorescent in situ hybridization (FISH) was performed as previously described (19). Briefly, cells were washed once with 1× phosphate-buffered saline (PBS) and fixed for 4 h with 4% wt/vol paraformaldehyde in PBS at 4°C. Probes were labeled at the STL1 ORF with C3-fluorescein (LGC Bioresearch [Table S5C]). For the negative control, samples were treated with RNase A (50 µg/ml) for 1 h at 37°C, prior to the hybridization step. Fixed cells were hybridized in 20 μl of hybridization buffer (0.9 M NaCl, 0.01% wt/vol SDS, 20 mM Tris-HCl [pH 7.2], 20% formamide) with 5 ng of C3 fluorescein-labeled probe and incubated at 46°C for 16 h. After incubation, cells were pelleted by centrifugation and resuspended in 1 ml of prewarmed washing buffer (20 mM Tris-HCl [pH 8.0], 0.01% wt/vol SDS, 5 mM EDTA, 225 mM NaCl) for 30 min at 46°C. The slides were then mounted in ProLong Gold antifade reagent (Invitrogen) and observed using a Leica DMI 6000B microscope with a Hamamatsu camera and LAS AF6000 v.2.1.2 software (Leica). Cells were scored as positive by a blind observer and analyzed by Fisher’s exact test.

Measurement of capsular size and virulence studies.

To induce capsule, yeast cells were grown on ASN at 30°C for 4 days. Capsule was measured by microscopy after the fungal cells were suspended in India ink (79) and laccase by melanin production on norepinephrine agar. Virulence studies were conducted according to a previously described intravenous mouse meningoencephalitis model (80) using 10 CBA/J mice for each C. neoformans strain. All experimental procedures were conducted under a protocol approved by the Institutional Animal Care and Use Committee of the Intramural Research Program of the NIAID, NIH.

ChIP assay.

The ChIP assay was adapted and modified from a previously described protocol (81) for C. neoformans. PCR detection of the STL1 gene was performed using a primer set described in Table S5B. Input DNA was used as a positive control, consisting of unprecipitated genomic DNA as a loading control and to show intact function of each primer set and PCR.

Promoter deletion studies.

The 5′-truncated promoter sequences were obtained by PCR using a single reverse primer and different forward primers (positions −1000, −500, and −20 from the transcript start site) carrying BglII and NdeI restriction sites (Table S5B). The amplified fragments were inserted upstream of the GFP gene coding region of the plasmid, producing a series of STL1 promoter-STL1-GFP vectors. After verification by sequence analysis, the confirmed constructs or linear amplified fragments containing the indicated upstream region, STL1 reading frame, and GFP marker and terminator as indicated were transformed into WT or cac2Δ strains of C. neoformans. Integrated constructs were confirmed by uncut Southern blots (81). These strains were subjected to microscopy (differential interference contrast [DIC] or GFP fluorescence) or flow cytometry (fluorescein isothiocyanate [FITC]) to determine the promoter activity of STL1 under glucose or starvation conditions.

Statistics.

Errors were expressed as standard error of the mean (SEM). Fluorescence-positive cells (Fig. 1) were scored in a blind fashion and analyzed by Fisher’s exact test.

Calculations involving statistics available from the ANOVA output (JMP/Genomics version 8.0) include mean square error (MSE), error degrees of freedom (DDF), model degrees of freedom (NDF), and the proportion of the variance accounted for by the model (R2). From these we derive the following: SSE (sum of squares error) = MSE × DDF, SSM (sum of squares model) = SSE × R2/(1 − R2), MSM (mean square model) = SSM/NDF, and F ratio = MSM/MSE.

Raw P values are retrieved from the cumulative probability distribution of the F ratio using the parameters F ratio, NDF, and DDF (e.g., with the SAS function PROBF), where n (total samples) = NDF + DDF + 1 and C (number of groups) = NDF + 1.

A hypothetical increase of n samples per group by iteration i (steps) leads to an increase in DDF, DDF(ni) = n + 2 × n × i – C, which propagates to recalculations of SSE, SSM, F ratio, and P value for i = 1,2,3… iterations.

Data availability.

Data from the microarray experiments have been deposited in the National Library of Medicine Gene Expression Ontology (GEO) database.