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

Human infection with Cryptococcus causes up to a quarter million AIDS-related deaths annually and is the most common cause of non-viral meningitis in the United States. As an opportunistic fungal pathogen, C. neoformans is distinguished by its ability to adapt to diverse host environments including plants, amoeba 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 a low nutrient-induced gene, an alternative carbon nutrient transporter STL1 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 target-of-rapamycin (TOR)-related multi-gene regulatory mechanism involving the CAC2 subunit of the chromatin assembly complex 1, CAF-1. In this pathway, the TOR-related RNA chaperone, 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. The 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. Author summary The fungus Cryptococcus is a fungal pathogen that kills an estimated quarter of a million individuals yearly and is the most common cause of meningitis in the United States. The fungus is carried in about 10% of the adult population and, after re-activation, causes disease in a wide variety of individuals including HIV-infected as well as immunosuppression either from genetic defects or after immune suppressive treatments due to transplant conditioning, cancer therapy or treatment of autoimmune diseases. The fungus is widely carried in the soil and trees and can infect plants, single cell organisms and even dolphins. However, 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 infect these diverse hosts and identified a nutrient transporter, STL1 to be associated with a marker of poor clinical outcome in a cohort of HIV/AIDS patients. Understanding molecular mechanisms involved in environmental adaptation may help to design better methods of control and treatment of widely dispersed fungal pathogens such as Cryptococcus.


Introduction
Cryptococcus neoformans (Cn) 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 million deaths annually [1,2]. Mortality exceeds 20-50% 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 amoeba [9] and animals such as dolphins [10] and humans [11]. Similarities between mammalian host environments such as the macrophage phagolysosome and those more primitive such as amoeba have been implicated as exerting evolutionary pressure on facultative intracellular pathogens such as Cn [12] and gene expression studies suggest adaptation to nutrient deprivation is important within this environment [13].
Indeed, requirements for the gluconeogenesis enzyme Pck1 [14] and a high affinity glucose transporter [15] 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 [16]. However, the dispensability of isocitrate lyase by Cn during infection suggests that species-specific pathways for starvation tolerance are also important within the mammalian host [17]. In addition, induction of the nutrient-recycling autophagy pathway is particularly important for Cn but less important for other fungal pathogens such as Aspergillus and Candida [18]. For this important stress response, global nutrient regulators such as the target of rapamycin (TOR) [19] play an important link between cryptococcal starvation response, macrophage survival and pathogen fitness [20]. 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 [20]. Similarly, the role of the cAMP nutrient sensing pathway has featured prominently in cryptococcal pathogenesis [21]. In this pathway, adenylyl cyclase is activated by a Ga subunit (Gpa1), resulting in the production of cAMP that binds to regulatory subunits Pkr1 of the PKA complex to release an active form of the catalytic subunit Pka1 which activates downstream proteins [22]. 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 found robust relationships between mammalian infectious outcome and typical virulence factors in vitro, suggesting a need to identify new virulence attributes by novel approaches [23] [24]. Thus, to characterize virulence-associated starvation responses more 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 [25]. These studies demonstrated a high concordance between expression levels by the fungus within infected human cerebrospinal fluid and under nutrient deficiency and identified a large set of 855 upregulated genes. To identify genes within this subgroup whose expression might be most relevant to human infections, isolates from a cohort of HIV/AIDS-infected patients with CM were utilized in a discovery study of attributes related to 10 week mortality, along the lines of a previous hypothesis-driven analysis linking clinical outcome and known virulence attributes [24]. The second most highly differentially-expressed annotated gene (CNAG_01683/CP022331.1) 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 [26]. Furthermore, mouse modeling demonstrated attenuated virulence of an stl1Δ strain, confirming its role in mammalian virulence. STL1 was also identified as a target gene of TOR, regulated by a novel regulatory pathway involving mRNA decay and a multi-component Cac2-Cac1-Ecm15 nuclear shuttle regulatory complex.
These studies thus provide a direct link between TOR signaling starvation response and CAF1mediated regulation of microbial pathogenesis, relevant to human fungal infections.

Transcriptional profiling of Cn suggest similarity in adaptive stress response between human CSF and starvation media at 37 o 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 brain in a cohort of solid organ transplant patients [25]. These conditions were utilized to compare transcriptional profiles of a pathogenic Cn strain (serotype A, H99), between nutrient rich and nutrient poor environment and compared the starvation adaptation profiles to that obtained after transfer to pooled human cerebrospinal fluid from patients with cryptococcal meningitis (S1 and S2 Table). A strong correlation was demonstrated between expression level changes in CSF vs. starvation media at 37 o C (Fig 1A; Rsquare = 0.79, slope non-zero p < 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 the transcriptional changes. Of 1,363 genes upregulated in starvation and 1,089 in CSF, 855 were upregulated in both conditions of which 332 were annotated sufficiently to allow GO term assignments ( Fig 1B). Compared to the overall annotated transcriptome (Fig 1C, left panel), transition of Cn to either starvation conditions or CSF resulted in an expansion of gene expression associated with the gene ontology (GO) biological term "carbohydrate metabolic processes" including key enzymes involved in β-oxidation, the glyoxylate cycle and gluconeogenesis (S1 Fig). 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 (S2 Fig). These data thus emphasize similarities in gene expression profiles of Cn between nutrient deprivation and CSF as well as a predominance of genes involved in carbohydrate homeostasis and transport function.

Expression profiling of isolates from a cohort of AIDS-related CM identifies a putative
sugar transporter like gene, STL1 as a candidate biomarker related to clinical outcome.
Further studies were then conducted to determine whether starvation-induced genes may provide candidate biomarkers of clinical outcome that may help to understand human-related pathogenicity of the fungus. Transcriptional profiling under starvation conditions was conducted as described, which previously identified the copper transporter CTR4 as a potential biomarker of brain dissemination in solid organ transplant recipients [27] using a set of serotype A Cn isolates 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 and all 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-weeks mortality of 11/45 (24%), typical of CM in developing countries [3]. Pre-treatment co-variables were significantly different for mental status as measured by Glasgow coma score (p = 0.02), but parameters including CSF initial fungal burden, EFA, CSF opening pressure or WBCs or CD4 count did not differ significantly.
Expression profiles were compared using ANOVA with adjustment, grouping patient isolates associated with mortality at 10 weeks vs. survivors. Ninety-one genes (28 annotated) showed significantly differing expression in patients who died vs. those who lived at least 10-weeks after therapy (S3 Table, adj p < 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] and was selected for further study because of its putative role in alternative carbon metabolism and transport function exemplative of the aggregate transcriptional changes in human CSF. STL1 expression levels by qRT-PCR from the clinical isolates showed a significant correlation with poor early fungicidal activity (EFA, Fig 1D; p = 0.018) [26] as well as a trend with clinical outcome (Fig 1E; p = 0.07); however, the latter was not statistically significant, possibly due to non-microbiological 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 [26,31]. In addition, STL1 expression was demonstrated in a brain autopsy specimen from an HIV/AIDS patient with CM using fluorescent in situ hybridization (66/66 yeast cells with +STL1 signal vs. 9/78 in RNase-treated negative control slides, Fisher's exact test, p < 0.0001; Fig 1F) [20]. Furthermore, qRT-PCR demonstrated that STL1 expression was induced during starvation as well as in the presence of the TOR inhibitor rapamycin (Fig 1G, H). In summary, these data identify a starvation-induced potential biomarker of clinical outcome, STL1, associated with poor microbiological clearance (EFA) during HIV-associated human infections.

Identification of a VAD1-dependent regulator of STL1, ECM15.
Further studies were conducted to identify TOR-dependent regulatory pathways that could play a role in expression of starvation-associated genes such as STL1. TOR is a particularly important pathway as the TOR inhibitors sirolimus and everolimus, are in widespread clinical use as immunosuppressants [32]. Recently, the cryptococcal RNA chaperone Vad1 was shown to regulate gene expression in a TOR-dependent fashion by recruiting mRNA to a decapping complex protein Dcp2 for decapping, leading to transcript degradation [20]. Previously, we determined the mRNA binding profile of Vad1 in Cn by using RNA immunoprecipitation followed by microarray analysis (RIP-ChIP), and showed that VAD1 binds multiple transcripts, including a gene ECM15 [33]. ECM15 in S. cerevisiae is a non-essential poorly understood protein, proposed to be involved in cell wall competency and was found on a yeast-hybrid screen to bind to a nuclear protein, Cac2 [34,35]. In the current study, ECM15 mRNA binding to Vad1 was confirmed by qRT-PCR after immunoprecipitation of a c-myc-Vad1 fusion protein compared to that of an equivalent precipitation using an untagged strain (Fig 2A). In addition, as shown in Fig 2B, a vad1Δ mutant showed accumulation of ECM15 transcripts compared with wild-type from cells incubated under starvation conditions. Transcriptional profiling of an ecm15Δ strain showed reduced expression of the target gene, STL1 described above (S4 Table), suggesting a regulatory circuit between TOR-dependent regulation and STL1 through VAD1 and ECM15. Phenotypic studies demonstrated a role for ECM15 in expression of known virulence factors of C. neoformans. For example, the anti-phagocytic extracellular capsule was increased in the ecm15Δ mutant cells compared to wild-type under both nutrient replete conditions (Glu +) and nutrient deplete conditions (Glu -) which reverted in the complemented strain ( Fig 2C).
However, expression of the multifunctional laccase enzyme [24] was reduced in the ecm15Δ mutant ( Fig 2D) as was mating ( Fig 2E). These studies thus implicate a role for ECM15 in the regulation of several virulence-associated phenotypes as well as mating of C. neoformans.
Further studies sought to investigate the functional relatedness of the cryptococcal Ecm15 protein to that of S. cerevisiae. The deduced amino acid sequence showed 66.8% identity to its respective homolog of Sc. Susceptibility of a Sc ecm15Δ mutant to the cell wall active agent Calcofluor white previously reported [36] was complemented by the cryptococcal ECM15 gene ( Fig 2F). These results suggest that Cn Ecm15 is a functional homolog of Sc Ecm15p. Extending these results to cell wall phenotypes associated with virulence, ecm15Δ mutants were found to exhibit increased susceptibility to the ionic detergent SDS as well as the anti-fungal cell wall agent caspofungin or the cell membrane disruptor drug fluconazole ( Fig 2G). Interestingly the ecm15Δ mutant also displayed poor growth on nutrient poor agar consisting of asparagine salts (ASN) without glucose consistent with its TOR dependence ( Fig 2G). In addition, the Cn ecm15Δ strain exhibited attenuated virulence in a mouse model (Fig 2H). Interestingly attenuated virulence was also exhibited by an ECM15 overexpressing strain compared to an identical strain containing an equivalent empty vector at the same copy number (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 [37]. Attenuated virulence with deletion or overexpression is a hallmark of the cryptococcal VAD1 gene that shows a similar phenotype [20] and may be due to their close regulatory association. 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.

ECM15 dependent regulation of a chromatin assembly factor, CAC2 involved in extracellular capsule expression and mammalian virulence
Previous whole proteome interaction studies suggested that Sc Ecm15p may interact with Sc Cac2 [34]. 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 sub-telomeric locations and tolerance to ultraviolet irradiation (UV), the latter demonstrated in both Sc and Cn [38,39].
However, its precise role as a potential regulator remains unexplored in eukaryotes. Construction  Table). Further study identified several virulence-related phenotypes shared between CAC2 and ECM15. As shown in Fig 3C and S9 Fig, capsule sizes under nutrient deplete conditions were accentuated in the cac2Δ mutant (a phenocopy of the ecm15Δ mutant), and was suppressed to that of wild-type in the CAC2 complemented strain. Epistatic studies demonstrated that capsule sizes under both nutrient replete and nutrient deplete conditions were restored to WT in ecm15Δ mutant cells overexpressing the CAC2 gene, although overexpression of ECM15 had no effect on the capsular phenotypes of the cac2Δ mutant cells. These results suggest that CAC2 is a suppressor of capsule formation downstream of ECM15. In addition, CAC2 transcription was suppressed in the ecm15Δ mutant strain under starvation conditions, suggesting that ECM15 partially activates CAC2 signaling through a transcriptional process. In addition, demonstration of CAC2 transcript accumulation in the vad1Δ mutant, similar to ECM15 suggests the strong effects VAD1 as a global regulator (Fig 3D and 3E) [14]. Furthermore, CAC2 overexpression resulted in suppression of virulence ( Fig 3F), similar to that of the ECM15 overexpressor strain, suggesting a role of both in the suppression of virulence. However, deletion of CAC2 downstream of ECM15 did not increase mammalian virulence suggesting loss of ECM15-regulated processes as the pathway was traced downward. Taken together, these data suggest an epistatic relationship between ECM15 and CAC2 in the repression of cryptococcal capsule and mammalian virulence.

Nuclear localization of CAC2 is nutrient-dependent and requires ECM15.
Saccharomyces interactome studies [34], suggested that  Fig 4B). Interestingly, Ecm15 also mis-localized to the cytoplasm in cac2∆ mutant suggesting that an interaction between the two is required for effective nuclear co-localization.
These results suggest that Ecm15 and Cac2 play an interacting role in nuclear targeting, leading to downstream repression of virulence associated phenotypes and genes such as STL1 described in  Because of its putative role in alternative carbon homeostasis and virulence in the clinical isolates, we tested to see if STL1 overexpression could facilitate growth on alternative carbon sources. Interestingly, STL1 overexpression facilitated increased growth in nutrient limitation media 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 the STL1 homolog from Candida ( Fig 5D) [29,41]. In addition, STL1 overexpression facilitated increased growth in nutrient limitation media containing either 0.03% of ribose, or citrate but not equivalent concentrations of glucose vs. an identical strain transformed with empty vector alone in identical copy number (Fig 5E). These data extend the spectrum of alternative carbon substrates related to Stl1. However, STL1 did not facilitate growth on other sugars including galactose, glucosamine, and rhamnose, suggesting carbon substrate specificity (S4 Fig). In addition, the stl1Δ strain exhibited moderately reduced virulence in an intravenous mouse brain dissemination model compared to the wild-type strain ( Fig 5F, left panel). However, inoculation of Cn 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 5F, right panel). Taken together, these data suggest that deletion of cryptococcal CAC2 increases transcript abundance of STL1, a regulator of capsule, sugar transport, laccase activity, and virulence.

Cac2 exhibits TOR-dependent STL1 promoter occupancy
Early studies suggested that orthologs of CAF-1 members such as Cac2 are associated with suppression of genes in the sub-telomeric region [38], which include cryptococcal genes such as  Table); however, the STL1 gene does not reside within this region (location; chromosome 11, CP003830.1: 603,867-606,604). Thus, to identify possible direct STL1 Cac2-specific DNA binding region(s), chromatin immunoprecipitation (ChIP) was performed which demonstrated promoter occupancy within a 100-bp fragment centered at -700 bp from the transcriptional start site of STL1 as well as that of FRE7 in nutrient condition (Fig 6A), otherwise, STL1 Cac2-specific DNA binding efficiency was reduced in both starvation and rapamycin condition (S5 Fig) . To determine the functional significance of the binding, we tested expression levels of a Stl1-green fluorescent protein (GFP) fusion having serial deletions of the STL1 promoter (-1000, -500, -50 from the transcript start site) on basal transcriptional activity in C. neoformans. The plasmid containing an STL1 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 -1,000 to -500 region of STL1 under both nutrient replete (Glu +) and nutrient poor (Glu -), corresponding to the region of Cac2 binding by ChIP (Fig 6B and 6C). However, in the nutrient-rich condition, CAC2-dependent activity in the WT strain was only partially suppressed using the episomal constructs (S6A and S6B Fig) and suggested a requirement for additional chromatin-dependent promoter-binding factor(s) for effective CAC2-dependent chromatin silencing [42]. Thus, linear PCR fragments containing the indicated promoter, STL1 coding region and selection marker were transformed into WT and cac2Δ strains as integrated fragments. As shown in Fig 6B and  Interesting was the heterogeneity in the expression of the cac2Δ strain which may suggest additional post-translational regulatory features.
In summary, these data identify a TOR-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 7).

Discussion
Pathogen survival and virulence entails selection for traits necessary within the infective niche.
For opportunist pathogens such as C. neoformans, evolutionary pressure exerted during environmental residence by resistance in free-living amoeba [9] or in the presence of diphenolic toxins within plants [8], have likely shaped the expression patterns necessary for effective mammalian infection. In the example of C. neoformans, the organism is also thought to reside in the host in many cases as a latent organism, which undergoes reactivation after immunosuppression by diseases such as HIV [7]. Thus, expression patterns of environmental C.
neoformans strains may also undergo optimization through microevolution after exposure to the mammalian infective niche which has been demonstrated in mice [37] and may influence clinical outcome in humans [25]. These combined pressures have resulted in an organism highly adapted for survival and growth under nutrient limitation conditions [43]. The molecular patterns that shape adaptation and optimization within these different environments was the subject of the present study.
In these studies, expression patterns of C. neoformans within human cerebrospinal fluid from actively infected patients was highly correlated with that induced under starvation conditions at 37 o C both in 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 [13], 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 in the present studies, 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 reflected this, ranging from 10-25 mg/dL (0.55-1.4 mmol/l) in the present studies. Interestingly, several studies suggest that the macrophage phagolysosome is also a nutrient-deprived environment and such stress results in pathogen adaptation for survival [4,37]. These commonalities reflective of starvation response thus implicate starvation response pathways in cryptococcal virulence in humans.
However, these expression patterns differed from that reported from RNA isolated directly from patient CSF of two CM patients that more closely resembled that from nutrient replete media and could have reflected an unusually vigorous growth condition or higher levels of CSF glucose concentrations typical of some HIV/AIDS patients [44]. However, such results caution that expression within a number of different environments may be required to understand heterogeneity within the human infective niche. In the present studies, CSF incubation resulted in reduced expression of genes related to protein translation typical of growth limitations from reduced nutrients [45]. Conversely, 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 towards more efficient utilization of alternative substrates and low glucose is shared by other organisms such as Candida albicans that also modifies its metabolism to assimilate these alternative substrates [46]. However, increased expression alone does not necessarily identify genes important to virulence, exemplified by the virulence factor laccase that is down regulated at elevated temperatures [47]. Disappointingly, genes demonstrating increased expression within pathogenic niches have shown variable roles in mammalian virulence. For example, the highly upregulated pyruvate kinase PYK1 and hexose kinase I and II (HXK1/HXK2) involved in alternative substrate demonstrate attenuated virulence in a mouse model [15], whereas the highly upregulated PTP1 sugar transporter demonstrated no such role [13]. Other upregulated genes such as PCK1 have had variable roles in virulence, with mutants attenuated in mice [14], but not in rabbits [15]. Such discordant results demonstrate the limitations of exclusive reliance on model host studies of virulence.
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. An important benefit of human-associated studies is that they may reflect better the specific environment of the human host vs. those derived from mouse studies alone, which may have limitations [48]. For example, mice strains such as C57BL/6J may elicit a significant neutrophilic and even eosinophilic response in lungs after infection with clinical fungal isolates, which is quite different from the histiocytic response of humans including giant cell formation, depending on the relative cellular immunity of the infected patient [49]. Drawbacks are that it is difficult to control for heterogeneity among unrelated fungal organisms and their hosts. The present studies identified a potential role for an STL1 sugar transporter required for survival and growth under low nutrient conditions and was also prioritized for further study because it was representative of an alternative substrate and transporter gene population showing an expansion in expression after transition to infected human CSF. STL1 is named as a sugar transporter but has been better characterized as having a role in glycerol acquisition [29,50] which was also found in the present studies in addition to roles in lactate, pyruvate and acetate metabolism. Interestingly, glycerol is elevated during cell membrane degradation and brain injury [51] and acetate has been demonstrated in large amounts by magnetic imaging spectroscopy during human cryptococcal infections [52,53] and thus STL1 may play a role in acquisition of these metabolites during infection. Expression levels of STL1 showed a trend towards and association with 10-weeks mortality in HIV-related CM and were also associated with an important prognostic marker of microbiological clearance during human infections, EFA. EFA is a research measure of CSF microbiological clearance in patients which reflects both microbiological and drug responses as well as host aspects of the infection that affects clearance and demonstrates reduced clearance in groups who die [54]. In addition to its role as a prognostic marker EFA has been secondarily proposed as a treatment surrogate, related to antifungal activities that distinguished amphotericin b vs. standard fluconazole therapy [26] but has not proven to be robust surrogate for the second role in multiple randomized-controlled studies of amphotericin-based regimens, according to Institute of Medicine guidelines [31].
Indeed, we found that EFA in the present cohort of patients treated with amphotericin-related regimens was broadly similar. Notable was the reduced expression of the virulence-factor laccase in the stl1∆ mutant which has previously been shown to be correlated with EFA and patient survival and could have contributed to microbiological retention in the CSF [24]. However, we were not able to demonstrate a statistically significant correlation of STL1 expression with patient survival with the numbers of available patient isolates which may be due to the large variability in expression levels between wild-type strains. 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 [25]. However, it would be premature to speculate as to the role of STL1 as a clinical biomarker and is likely co-related to other microbiological markers. There also exists a number of potential non-microbiological 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. Another important contributor to death in primary infections with C. neoformans is the relative immune response of the host although previous studies have shown only minor relationships of markers such as CD4 count, peripheral white blood count or CSF white count [3]. The present cohort did not include patients with cryptococcal immune reconstitution syndrome (cIRIS) where immune responses may be more predominant [55] or other inflammatory syndromes associated with non-HIV infected individuals [56]. All of these considerations led us to attempt to provide additional validation of the human cohort fungal expression studies by utilizing a mouse model which showed a modest difference in virulence between wild-type and stl1Δ strains, further suggesting a role for STL1 in mammalian infections. 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 where survival is more related to altered lung pathology [57]. However, additional cohorts reflecting differing clinical scenarios may be required to fully understand the complexity of the cryptococcal-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 TOR stress pathway. TOR is an important mediator of the starvation/stress response, is important for cryptococcal survival [58] and inhibitors of TOR such as sirolimus and everolimus are in widespread use in patient populations at risk for CM including transplant recipients [59]. More recent work has shown that many TOR-dependent starvation processes are regulated via mRNA stability in yeast [60] and an important TOR-dependent regulator of mRNA stability, VAD1 is also a major virulence determinant in C. neoformans [20]. 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 deletant mutant of CAC2 was previously found to exhibit normal growth characteristics and a slight susceptibility to UV irradiation [61]. In S. cerevisiae, deletion of CAF-1 subunits such as CAC2 results in increased UV sensitivity and silencing defects of sub-telomeric genes [38], suggesting a role in stress response, but the present studies are the first describing the role of Cac2 as a dynamic regulator, per se. Interestingly, only CAC2 overexpression resulted in reduced virulence in mice, but this would be expected as it is an STL1 repressor. 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 that 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 (FRE7), regulating the important iron acquisition pathway in C. neoformans [62]. The FRE7 gene was also upregulated in the patient cohort, though did not approach significance and was not further characterized. This regulatory pathway 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 [20,[63][64][65][66][67][68][69]. Repression by CAC2 may be chromatin dependent as promotor studies demonstrated greater CAC2-dependent suppression in glucose when the 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 [70] and may thus be required as part of the chromatin repressor complex [71]. 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 [72]; however, the present studies extend the role of Cac2 within the Caf1 complex from merely a house-keeping function to that of a dynamic regulator of TOR-mediated nutrient stress response. Future studies may further help to provide more detailed structural mechanistic insight for this complex regulation in eukaryotes and highlights the utility of C. neoformans as a model eukaryote. Nevertheless, the present data demonstrates that the VAD1/ECM15/CAC2 regulatory pathway is integral to the connection of the starvation/TOR virulence response through the sugar transporter STL1 (Fig 7). Study Subjects. Subjects providing isolates for expression analysis were control participants in a randomized trial of adjunctive IFN-γ in HIV-infected patients, and a randomized trial examining alternative amphotericin B combinations, both described previously [73]. 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 AmB-based induction regimens (ie no fluconazole and no adjunctive IFN gamma 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 [56]. As shown in Table 1, the 45 patients had an age, CD4, etc. All strains were previously serotyped to be serotype A [74].

Strains and Media.
Experiments were conducted in a genetic background of C. neoformans WT strain serotype A H99 (MAT α, ATCC 208821) and was the kind gift of J. Perfect. A complete list of strains used in this study is described in S5 Table. Escherichia coli DH10B (Invitrogen) was the host strain for recovery and amplification of plasmids. The fungal strains were grown in YPD medium (2% glucose, 1% yeast extract, 2% Bacto-peptone) or YPD agar medium (YPD and 2% agar). Asparagine minimum selective medium (ASN) for transformant selection and for detection of laccase production was previously described [75]. V8 juice medium were used for mating assays as described [76]. Table 1 or the H99 lab strain were grown to midlog phase and then transferred to asparagine medium without glucose (asparagine, 1g/L, 10 mM sodium phosphate, pH 7.4 and 0.25 g/l MgSO4) or human CSF pooled from 4 patients with cryptococcal meningitis and incubated for 3 h at 37°C and RNA recovered as described previously [25]. Serotype A H99-based microarrays (hybridization probe sequences described previously in [27]) with two unique probes (3 replicate features per probe) for each of 6,969 transcripts (one per locus) were used for expression profiling. Two-color co-hybridizations 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 and averaged per locus (3 replicates of two probes for 1 transcript per locus). A mixed-effects ANOVA model (fixed effect of growth media, random effect of array ID) was computed, and expression difference estimates for each gene were calculated for CSF or starvation media vs. YPD. For the patient isolates (Table 1) Overexpression, disruption and complementation of ECM15, CAC2, and STL1 in C.

Microarray experiments. Isolates from patients in
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 down stream genomic fragment of the target genes [14,77,78]. Complementation in all cases retained the original deletion construct. Primers used in this study listed in S6 Table. Ecm15-mCherry and Cac2-GFP fusion proteins The cryptococcal shuttle vector pORA-YP142 [79] was used to express a fusion between the Ecm15 protein and a synthetic mCherry protein (Cneo-mCherry), utilizing C. neoformans codon usage produced using standard methods [77]. 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 [80]. All cells preparations grown on nonselective media are assayed at the end of each experiment by simultaneous inoculation of selective and non-selective plates to verify >90% retention of plasmid. Co-localization was observed using a Leica DMI 6000B microscope with a Hamamatsu camera using LAS AF6000 ver 2.1.2 software (Leica). Predicted nuclear localization sequences were identified in the Ecm15 and Cac2 protein using cNLS-Mapper using a cut-off score of 2 [40].
qRT-PCR experiments C. neoformans strains H99 were grown on YPD or ASN without glucose media. Real-time PCR was performed using a primer sets as described in S6 Table. 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).

Fluorescent in situ Hybridization (FISH).
Sections of a brain autopsy specimen were obtained from a 42 year old male who died of severe and diffuse C. neoformans infection and was reported previously [14]. FISH was performed as describe [20]. Briefly, cells were washed once with 1× PBS and fixed for 4 h with 4% w/v paraformaldehyde in PBS at 4°C. Probes were labelled at the STL1 ORF with the C3-fluorocein (S7 Table; LGC Bioresearch). For negative control, samples were treated with RNase A (50 µg/mL) for 1 h at 37°C, prior to the hybridization step.

Measurement of capsular size and virulence studies.
To induce capsule, yeast cells were grown on ASN media in a 30°C for 4 days. Capsule was measured by microscopy after the fungal cells were suspended in India ink [81], and laccase by melanin production on norepinephrine agar. Virulence studies were conducted according to a previously described intravenous mouse meningoencephalitis model [82] using 10 CBA/J mice for each C. Chromatin Immunoprecipitation (ChIP) assay The ChIP assay was adapted and modified from a previously described protocol [83] for C. neoformans. PCR detection of the STL1 was performed using a primer sets as described in S5 Table. '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 reaction.
Promoter deletion studies 5'-truncated promoter sequences were obtained by PCR using a single reverse primer and different forward primers (-1000, -500, -20 bp from the transcript start site) carrying BglII and NdeI restriction sites (S5 Table). The amplified fragments were inserted in the upstream of the GFP gene coding region of the plasmid, producing a series of STL1pro-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 [83]. These strains           Table. FISH probes used in this study.