Cryptococcus neoformans Glucuronoxylomannan and Sterylglucoside Are Required for Host Protection in an Animal Vaccination Model.

The number of deaths from cryptococcal meningitis is around 180,000 per year. The disease is the second leading cause of mortality among individuals with AIDS. Antifungal treatment is costly and associated with adverse effects and resistance, evidencing the urgency of development of both therapeutic and prophylactic tools. Here we demonstrate the key roles of polysaccharide- and glycolipid-containing structures in a vaccination model to prevent cryptococcosis.

not altered in mutant cells (see Fig. S1 in the supplemental material), we attributed the differences observed here to the accumulation of SGs.
Deletion of SGL1 in an acapsular strain (⌬cap59) resulted in SG accumulation without affecting cellular morphology. The presumed association between SGs and extracellular GXM led us to delete SGL1 in a mutant lacking a putative regulator of capsular export (CAP59) (30) for further functional analyses. After biolistic DNA delivery, we selected 31 of 280 colonies that were nourseothricin (NAT) resistant and confirmed the correct plasmidial insertion and disruption of the SGL1 gene via Southern blot analysis. Transformant 18 showed the expected band of 3,604 bp ( Fig. 2A and B). To confirm the phenotype that had been associated with the Δsgl1 mutant, we performed TLC and MS analyses ( Fig. 2C and D). The single and double mutants manifested similar levels of SG accumulation, and similar amounts of SGs were observed in the WT and Δcap59 strains. The amounts of SGs seen in mutants lacking the SGL1 gene were increased approximately 3-fold in comparison to WT and Δcap59 cells (P Ͻ 0.005). No significant changes were observed in sphingolipid glucosylceramide and inositol phosphorylceramide (IPC) detection, indicating that the mutant cells lacking SGL1 did not manifest general defects in sphingolipid biosynthesis (data not shown). Morphological analyses by microscopy confirmed a lack of capsular structures in the mutants lacking CAP59 (Fig. 2E). Noticeably, these cells also formed large clumps. On the basis of its   Proliferation analyses demonstrated that the Δcap59/Δsgl1 and Δcap59 strains manifested similar decreased growth rates in DMEM in a 5% CO 2 atmosphere (37°C), particularly at pH 7.4. India ink counterstaining under these conditions revealed Δcap59/Δsgl1 cells with clumped morphology, like strain Δcap59 (Fig. 3A). The analyses performed at pH 7.4 showed the most pronounced growth defects, and further transmission electron microscopy (TEM) analysis revealed that Δcap59 cells had greater numbers of vacuoles than all the other strains (Fig. 3B). The Δsgl1 mutant had both elongated shaping and bigger vacuoles with internal fibrillar-like structures. Additional phenotypic characterization included melanin and urease analyses. At 37°C and 5% CO 2 , all the mutants displayed impaired melanin production compared to WT cells (Fig. 3C). Urease assays revealed similarity among the Δcap59/Δsgl1, Δcap59, and WT strains and a slight increase in enzyme activity in the Δsgl1 mutant (Fig. 3D).
The ⌬cap59/⌬sgl1 mutant was nonpathogenic and was cleared from the lungs 7 days after intranasal injection in a mouse model. Previous studies have shown that both the Δcap59 and Δsgl1 mutants were nonpathogenic in intranasal infection models in mice (7,31). We reproduced these models with the double mutant in the present study. CBA/J mice were susceptible to infection with WT cells, but all animals survived after intranasal inoculation of the mutant strains (Fig. 4A). Fungal loads (CFU counting) were determined 7, 14, 21, and 60 days postinoculation with the mutant strains and 7, 11, 17, and 28 days postinoculation with the WT strain. Only the WT cells were able to reach the brains of the animals (Fig. 4B). Although all mutants were nonpathogenic and unable to reach the brain, the lung fungal burdens differed considerably (Fig. 4C). While CFU levels increased exponentially in mice infected with the WT strain, the Δcap59/Δsgl1 double mutant was cleared from the lungs 7 days after inoculation. The Δcap59 burden was decreased 7 days after inoculation but remained detectable in lung samples throughout the observation. The Δsgl1 mutant was also cleared from the lungs, but only 14 days postinfection. To verify if the double mutant was eliminated from the lungs before day 7 postinfection, we also analyzed CFU levels on days 1, 3, and 5 postinoculation with the Δcap59/Δsgl1 strain. A time-dependent decrease of CFU levels was observed (Fig. 4D).
Acapsular cells were mainly found inside bronchioles and induced a similar immune response when compared to capsular cells. Histologic analysis of the lungs at the beginning of infection (day 3) and at the endpoint (day 14 for the WT and day 60 for the mutants) revealed that all strains were detectable in the lungs in early infection points (Fig. 5A). However, the acapsular mutants, i.e., strains Δcap59/Δsgl1 and Δcap59, were mostly found inside the bronchioles. WT and Δsgl1 strains colonized the lung tissue, and big clusters of immune cells were observed. The presence of neutrophils and macrophages indicated the initiation of a pronounced inflammatory response. We confirmed the infiltration of the neutrophils and recruited macrophages by flow cytometry after the infection with the capsular fungal cells (Fig. 6). The Δsgl1 mutant induced higher number of monocytes and eosinophils (9 and 14 days) and dendritic cells (2, 7 and 14 days) than were seen with the other groups. At 4 days postinfection, the Δcap59/Δsgl1 mutant induced an increase in monocytes and eosinophils compared to the WT and Δcap59 strains. Moreover, the double mutant triggered an immune response similar to that seen with the Δcap59 strain. In the lungs, the levels genomic DNA using the 5=UTR and a gene probe, which are represented by a black bar in panel A. Transformant 18 was named mutant Δcap59/Δsgl1. The double mutant showed an expected band of 3,604 bp and no apparent band when 5=UTR and gene probes were used, respectively. (C) TLC of lipids extracted from Δcap59/Δsgl1, WT, Δcap59, and Δsgl1 cells was performed. The presence of bands representing the size of SGs was observed for both SGL1-deleted mutants, indicating the accumulation of SGs in these mutants. (D) MS analysis of lipids extracted from the fungal cells performed in triplicate showed a 3-fold increase in the levels of SGs in the SGL1-deleted mutants compared to the WT and Δcap59 strains, confirming the TLC result. *, P Ͻ 0.003 (compared to the WT and Δcap59 strains). (E) India ink and transmission electron microscopy (TEM) analysis of the fungal cells grown in YPD for 24 h. Absence of capsule was confirmed for the double mutant and strain Δcap59. Acapsular cells presented with formation of clumps. GlcCer, glucosylceramide.

Sterylglucoside-Related Protection Requires GXM
The double mutant (Δcap59/Δsgl1) displayed characteristics similar to those displayed by the corresponding Δcap59 background strain under host-mimicking conditions. (A) Growth curve of the fungal cells cultivated in DMEM plus 5% CO 2 at 37°C and pH 4.0 (lysosomal milieu; left panel) or pH 7.4 (extracellular milieu; right panel). An impairment of growth was observed by comparisons between the acapsular and capsular strains only at pH 7.4. Data represent means Ϯ standard deviations (SD) of results from an experiment performed in triplicate. #, difference between ⌬cap59/⌬sgl1 strain compared to the other strains (P Ͻ 0.001). $, difference between the WT strain and the acapsular mutants (P Ͻ 0.001). *, difference between the capsular and acapsular mutants (P Ͻ 0.001). (B) India ink and TEM were performed at pH 7.0 since growth impairment was observed. Data show that the Δcap59/Δsgl1 cells did not show any evident changes. However, the Δcap59 strain displayed a higher number of vacuoles and the Δsgl1 strain some cells with elongated shape and big vacuoles with fibrillar-like structures inside. (C) The melanin production activity of all the mutants was impacted compared to the WT strain results. (D) Christensen's urea agar was used for urea formation analysis. The strain Δcap59/Δsgl1 results showed similarity to the Δcap59 mutant results. of proinflammatory cytokines, including interleukin-1␤ (IL-1␤) and granulocyte colonystimulating factor (G-CSF), and of chemokines, including IP10/CXCL10, MIP1-␣/CCL3, MIP1-␤/CCL4, MIP2/CXCL2, and KC/CXCL1, were increased and remained high after the infection with WT cells (Fig. 7). Interestingly, the levels of the same molecules were increased at early points after Δsgl1 infection; however, they subsequently decreased over the duration of the infection. These results indicate that sustaining high levels of these cytokines may stimulate a hyperinflammatory reaction, which could be associated with nonresolution of the infection and, ultimately, with mouse death. Although the host was able to mount an immunological response to the double mutant, the response was insufficient to induce protection, which was found to be more similar to the level induced by the acapsular Δcap59 mutant.
At the endpoints, however, only the WT and Δcap59 cells were found in infected tissues, which supports the CFU data (Fig. 5B). The lungs of animals infected with Δcap59 cells were affected only partially, while the damage associated with the WT infection was more widespread in the lungs. Infections with the Δcap59/Δsgl1 mutant resulted in enlargement and partial destruction of the alveolar space, despite the fact that no fungal cells were found to be present there. Although we did not observe a clear difference in cell infiltration responses in the lungs of mice infected with the various mutants at the late stage of infection, we did see that the infections with the Only WT cells were able to reach the brain. Also of note, the CFU for the WT strain was verified on different days (days 7, 11, 17, and 28) due to the higher susceptibility to inoculation (n ϭ 3 per group in each time point). (C) The CFU counting of the lungs was performed under the same conditions as described for the brain. The Δcap59/Δsgl1 and Δsgl1 cells disappeared from the lungs on days 7 and 14 postinfection, respectively. The Δcap59 cells decreased in number but were maintained in the lungs, whereas the WT cell numbers increased exponentially. (D) Fungal burden of lungs inoculated with the double mutant to assess early time points (since this mutant was cleared from the lungs by day 7 postinfection).
Sterylglucoside-Related Protection Requires GXM ® SGL1-deleted mutants induced an increase in the number of lymphocytes and alveolar macrophages compared to the results seen with the WT-infected mouse (9 and/or 14 days). There were noticeable differences in the cytokine levels between the mice infected with the WT strain and those infected with the mutant strains. In addition to corroborating the CFU data, these histological and immunological findings suggest that  2,4,7,9,14, and 30 postinfection. At the early stages, there was an increase in the numbers of neutrophils and recruited macrophages on day 4 postinfection in mice infected with capsular mutants. Although the Δcap59 mutant induced an increase of neutrophil numbers at day 2, a marked decrease was observed afterwards. The Δsgl1 mutant induced an even higher level of cell infiltration than did the WT at day 4. This mutant also triggered higher numbers of monocytes and eosinophils (at 9 and 14 days) and dendritic cells (at 2, 7, and 14 days) than were seen with the other groups. At 4 days postinfection, the Δcap59/Δsgl1 mutant induced an increase in the levels of monocytes and eosinophils compared to the WT and Δcap59 strains. At later stages of infection, no clear difference was observed among the mutants. However, all the mutants, including the SGL1-deleted mutants, induced an increase in the number of lymphocytes and alveolar macrophages compared to WT-infected mice, respectively (9 and/or 14 days). Furthermore, the double mutant triggered an immune response similar to that triggered by the Δcap59 strain. the lung immune response to the Δcap59/Δsgl1 mutant was different from the immune response to the Δsgl1 mutant. The Δsgl1 mutant-induced cytokine and chemokine responses led to recruitment of leukocytes, mainly neutrophils and macrophages, with a subsequent decrease in the levels of inflammatory stimuli over the course of the infection, which may be the basis of the process by which this mutant is eliminated from the lung compared to WT.
In addition to the lung homogenate, we also analyzed levels of cytokine production from blood serum (Fig. S2A) and bronchoalveolar lavage (BAL) fluid ( Fig. S3A) for the different infections. The production of cytokines in the BAL fluid was equivalent to the lung tissue results. However, we did not observe any explicit differences in the cytokine profiles in the serum of WT and Δsgl1 mutant-treated mice as seen in the lung homogenate, although we did see greater immune response stimulation in animals infected with the WT strain than in those infected with the mutants. The quantitative assessment of immune cells for the blood revealed an increase in the levels of lymphocytes of Δcap59/Δsgl1 mutant-injected mice after infection resolution (day 10) (Fig. S4A). Collectively, these data suggest that the double mutant was able to induce an immune response, although to a lesser degree than the Δsgl1 mutant.

GXM is required for SG-related protection in a vaccination murine model.
Our group has shown that exposure of mice to the Δsgl1 mutant triggers a protective immune response after challenge with a WT strain in both normal and immunocompromised (CD4 ϩ cell-depleted) mice (7). GXM is known to induce antibody-mediated immunity against C. neoformans (11,19,27). Due to the connections between SGs and GXM described in this study, we tested the Δcap59/Δsgl1 mutant in a vaccination model consisting of intranasal infection of CBA/J animals with the live attenuated Δcap59/ Δsgl1 strain followed by a rechallenge with the WT strain 30 days later and assessed survival for an additional 60 days (Fig. 8). As previously shown, the Δsgl1 mutant induced 100% protection in animals (*, P Ͻ 0.0001). However, vaccination with Δcap59/ Δsgl1 double mutant resulted in significantly (80%) higher mortality (*, P ϭ 0.0014 [versus phosphate-buffered saline {PBS}]; *, P ϭ 0.0021 [versus ⌬sgl1]). As expected, PBS-treated animals succumbed to the infection after rechallenge with WT cells. Only one animal survived vaccination with the Δcap59 mutant, which was similar to the results seen with animals immunized with the Δcap59/Δsgl1 strain (P ϭ 0.1504). To get a better understanding of the induced protection, we examined the immunological response through flow cytometry (Fig. 9) and cytokine analysis (Fig. 10) after the rechallenge with WT in the vaccination model. We did observe an increase in the levels of dendritic cells and CD4 ϩ T cells (day 8) in mice vaccinated with the Δsgl1 strain but not in those vaccinated with the Δcap59 and the Δcap59/Δsgl1 strains. We also observed the same profile of increased cytokine levels during the early stages of infection, which returned to basal levels as the infection progressed for Δsgl1 mutanttreated lungs compared to the other groups. These cytokines included CXCL1 (KC), IL-6, CCL3 (MIP1-␣), CCL4 (MIP1-␤), and CXCL2 (MIP2). We saw not only a lower number of lymphocyte cells in the blood samples of Δsgl1 mutant-pretreated mice (Fig. S4B) but also noticeable differences in the cytokine profiles, mainly at late stages, of mice that were pretreated with this strain versus the other groups (Fig. S2B). In the BAL fluid analysis, however, we did not find any similarity in the levels of cytokines to those determined in the lung homogenate analysis. We saw an increase of cytokine levels only at late stages in PBS-pretreated animals, although only Δsgl1 mutant-vaccinated    to analyze if the vesicles could be used as a vaccine instead of the live attenuated mutant cells. Since previous results suggested that CD4 ϩ T cell-mediated immunity from the protective effects was not required, we used a model of infection with Galleria mellonella, which lacks T cell-mediated immunity. This model also requires smaller amounts of EVs than the mouse model. Vesicles were isolated from WT, Δcap59/Δsgl1, Δcap59, and Δsgl1 cells and used to treat G. mellonella before challenging the animals with WT cells. Vesicle amounts were normalized to the amounts of glucosylceramide (GlcCer) and sterols, which are typical EV components (33). Vesicle fractions for G. mellonella stimulation contained 2 M sterol, which corresponded to 0.22 pmol of GlcCer (Table 2). G. mellonella survival analysis (Fig. 11) revealed that although all  Sterylglucoside-Related Protection Requires GXM ® animals died at the end of the experiment, those that received Δsgl1 EVs were 100% protected until day 3 (*, P ϭ 0.001) and this vaccination resulted in delayed lethality compared to PBS (*, P ϭ 0.0006). Interestingly, animals receiving WT EVs died sooner than the ones that received only PBS (*, P ϭ 0.0062). Vaccination with Δcap59/Δsgl1 and Δcap59 EVs resulted in no protection compared to PBS vaccination (P Ͼ 0.01).

DISCUSSION
In this work, we showed that SGs and GXM are functionally related in the Δsgl1 mutant. To test whether the accumulation of SGs was itself the only factor that would induce protection in a mouse model, we successfully generated the double mutant Δcap59/Δsgl1. The phenotypic characteristics of the Δcap59/Δsgl1 mutant were generally similar to those of the corresponding Δcap59 background strain. The Δcap59/Δsgl1 mutant was rapidly cleared from the lungs. This response from the host was faster than that seen in the Δsgl1 mutant-infected mouse, whereas the Δcap59 mutant infection persisted in the lungs throughout the window of observation. Confirming the CFU data, lung histology showed that the SGL1-deleted mutants were cleared from the lungs at late stages of the infection, and the immunological characterization showed that the assembled inflammatory responses were clearly different between these mutants. All mutants were nonpathogenic, and, when used as live attenuated vaccines, the Δcap59/ Δsgl1 mutant showed a loss of protective properties compared to the Δsgl1 mutant, which indicates that SGs alone are not sufficient for protection. In fact, GXM seems to play an essential role in protection because when EVs, which notoriously contain GXM (33), were used as vaccine formulations in G. mellonella, we found that EVs enriched in SGs, but lacking GXM, were not able to protect the host in a manner similar to that seen with EVs enriched in both SGs and GXM.
SGs and GXM can modulate the host's immune system (12,23). Studies with plant SGs have also suggested a particular role of these immunostimulatory molecules in the balance of Th1 and Th2 immune responses (12,(14)(15)(16)(17). Studies in mammalian cells have reported that glycosylation of cholesterol stimulated by Helicobacter pylori may impair phagocytosis, T cell activation, and, ultimately, bacterial clearance (34). However, little has been elucidated about the roles of fungal SGs. Besides being associated with Around thirty Galleria mellonella larvae were injected through the proleg with 10 l of PBS or an EV suspension of the WT, Δcap59, Δsgl1, or Δcap59/Δsgl1 strain containing 2 M sterol. After 24 h, animals received an injection of 2 ϫ 10 4 WT cells, and host survival was monitored. G. mellonella that had changed to pupae during the experiment were eliminated from the analysis. Animals that received WT EVs died sooner than the ones that received PBS treatment only (*, P ϭ 0.0062). Although all animals died in the end of the experiment, Δsgl1 mutant EV pretreatment guaranteed 100% protection compared to PBS until day 3 (*, P ϭ 0.001) and delayed lethality until the end (*, P ϭ 0.0006). When treated with Δcap59/Δsgl1 and Δcap59 EVs, animals showed levels of survival similar to those seen with PBS treatment (P Ͼ 0.01; some animals died between the EV treatment and WT challenge, and the animals that became pupae were excluded from the analysis).
Colombo et al.
® protective effects in a pulmonary cryptococcosis model and with fungal clearance (7), levels of SGs of Pichia pastoris were found to be increased under stress conditions, including heat shock and exposure to ethanol (35). SGs also participate in autophagymediated degradation of peroxisomes (36). Accumulation of SGs was observed in a C. neoformans mutant lacking the APT1 gene, which encodes a putative flippase responsible for the disposition of lipids in the membranes, in an asymmetric way (37). This mutant is hypopathogenic and was associated with reduced GXM production, defective vesicle trafficking, and impaired lipid biosynthesis (37,38). The lower pathogenicity of the Δapt1 mutant associated with increased SG content agrees with our results. Our data also show that accumulation of SGs was evident in cellular and extracellular fractions, including vesicles. Differences between the Δsgl1 and Δcap59/Δsgl1 strains in colonizing and in inducing an immune response in our pulmonary cryptococcosis model were also observed. Immune defenses in the lungs at early stages of infection were stimulated in waves, with the aim of maintaining a balance between sufficient protection and minimal damage effects (39,40). In fact, SGs seem to stimulate proliferation of eosinophils and lymphocyte cells as reported previously (14,17), but such an increase in cytokine levels (14,16,17) was not clearly observed in our model. On the other hand, mice injected with capsular cells showed induction of chemokines. These chemokines are involved in leukocyte recruitment and polarization of Th1/Th2 immune responses to C. neoformans infection, evidencing its central role in cell-mediated immunity (41). However, sustained levels of these cytokines, as observed solely in the WT-infected group, can lead to immunopathology and be associated with the nonresolution of infection and animal death. These results suggest a main role of GXM under conditions of SG accumulation in the early stages of infection. We also suggest that SGs are adjuvants for the protective effects of GXM. The majority of adjuvants activate antigen-presenting cells, mostly dendritic cells, leading to subsequent differentiation of antigen-specific T cells (42,43). Interestingly, glycolipids have been studied as potential vaccine adjuvants by stimulating the invariant natural killer T (iNKT) cells (42). Several microbial glycolipids are recognized by these cells, which can trigger secretion of Th1 and Th2 cytokines simultaneously, inducing the transactivation of dendritic cells, macrophages, natural killer cells, and B and T cells (42). The effects of a synthetic glycolipid in stimulating iNKT cells revealed an increase in the level of gamma interferon (IFN-␥) and a decrease of in the fungal burden in animals infected with WT C. neoformans cells (44), which suggests that SGs may have been contributing to the clearance of the mutants through this effect.
In our vaccination model, mutant cells lacking CAP59 were unable to induce protection. These results were different from those obtained by Specht and colleagues (5), who observed partial protection in animals pretreated with the Δcap59 strain. This discrepancy might be explained by the usage of different mouse strains (C57BL/6J mice instead of the CBA/J mice used in our study) and the administration route of Δcap59 (subcutaneous instead of intranasal in our study).
The mechanism by which GXM is involved in the anticryptococcal protection needs to be further investigated. Early studies have reported the main role of CD4 ϩ T cells and Th1 responses in the protection against cryptococcosis in animal models (45)(46)(47)(48). We did observe an increase of the CD4 ϩ T cell population in the lungs of mice pretreated with the Δsgl1 mutant in the vaccination model, which indicates that CD4 ϩ T cell are stimulated by GXM-SGs to orchestrate a protective immune response. However, our previous data did show that, under conditions of CD4 ϩ T cell depletion, the protection induced by GXM in the presence of the SG accumulation may go beyond the Th1 response, since protection was also previously observed under conditions of CD4 ϩ T cell deficiency (7). Interestingly, as observed with the Δsgl1 mutant, a C. neoformans strain genetically modified to produce IFN-␥ (H99␥) induced protection in CD4 ϩ T cell-depleted mice (49). Indeed, CD8 ϩ T cells participated in protection in a cryptococcosis model in a manner independent of CD4 ϩ T cells but dependent on IFN-␥ (50). It is known that CD8 ϩ tissue-resident memory T cells (CD8 ϩ T RM ) provide local protection against secondary infection (51,52), and such cells could be involved in the protective effect observed in our study. Different studies have shown that microbial polysaccharides can activate classical and memory CD8 ϩ T cells (53,54); thus, it is possible that C. neoformans GXM could activate lung memory CD8 ϩ T cells, particularly when GXM is associated with SGs. In addition, GXM can modulate innate responses through interactions with Toll-like receptors (TLR) 2 and 4 and with CD14, CD18, and Fc␥RIIB receptors (23,55,56). These responses result in the modulation of neutrophils, macrophages, monocytes, and dendritic cells (57,58). In fact, we did observe massive infiltration of neutrophils and macrophages in the lungs of mice infected with the capsular mutant. The Δsgl1 strain induced even greater infiltration than was seen with systems where mice were infected with the WT strain. In association with SGs, GXM bioactivity and binding to immune receptors can be modified, as previously demonstrated after association of the polysaccharide with chitin oligomers (59).
Although GXM-based vaccines seem to protect through antibody-mediated immunity (11,19,27), passive transfer of immune serum from animals injected with the Δsgl1 strain to naive mice did not induce protection against C. neoformans infection (data not shown), suggesting that the protective effect induced by the Δsgl1 strain may not be antibody mediated. However, more studies are necessary to address this issue and to rule out the effect of this mechanism of protection in our model. For instance, analysis of the protective effects observed here in infection models using B cell knockout mice will be crucial to reveal the role of antibodies in SG/GXM-mediated immunological protection against cryptococcosis.
EVs containing SGs and GXM obtained from the Δsgl1 mutant delayed lethal infections with C. neoformans in a G. mellonella model. Although the delayed mortality was clear in comparisons between the different protective treatments, all animals ultimately died after challenge with the WT strain. The loss of the protective effect of the administration of EVs in this system may have been due to the lack of adaptive immunity in G. mellonella (60) and/or to insufficient administration of EVs. In fact, only a single injection of EVs was performed. In the mouse vaccination model, the administration of live attenuated Δsgl1 cells may continuously produce and release EVs in the airways. The lung fungal burden of Δsgl1 cells started to decrease after day 3 (7), suggesting that there is enough time for the production and secretion of EVs containing GXM and enriched in SGs in the airways during this period. Additional studies are under way to verify the protective effect of these enriched EVs on the innate immune cells. In this sense, mice models of EV-mediated protection are likely more informative than the invertebrate model used in our study to reveal the mechanisms of immunological protection. However, a major limitation of murine models of EV-based vaccination is related to the lack of appropriate protocols for obtaining large amounts of EVs. This limitation will be likely overcome in a near future, as a consequence of the recent development of a facilitated protocol of EV isolation from fungal cultures (61).
Curiously, the Δsgl1 EV mutants are much larger than the EVs from WT or reconstituted strain (data not shown). This was confirmed in vitro when we synthesized an ethanol dilution of small unilamellar vesicles from binary mixtures of purified SGs and GlcCer extracted from C. neoformans according to previously established methods (62). We noted that addition of SGs systematically increased their size and reduced their diffusion coefficient, measured using the Stokes-Einstein equation (63) (see Fig. S5 in the supplemental material). This increase in vesicle size and the observed decrease in diffusion may reduce the speed of vesicle movement and thus increase the duration of interactions between vesicles and host cells. Given the notable immunomodulatory properties of GXM (23,55) and SGs (32,64), the increased interaction time might lead to a stronger host immune response, resulting in the elimination of the primary infection and in the stimulation of host protection against a secondary infection.
EVs from Candida albicans produced protective effects in a G. mellonella model (65). In our study, however, the presence of C. neoformans WT EVs seemed to exacerbate the infection instead of controlling it. In fact, there is experimental evidence that pretreatment with cryptococcal EVs may enhance disease progression in mice (66). Besides, it has been recently reported that the presence of GXM in cryptococcal EVs, but not the presence of GXM alone, augments virulence of C. gattii in vitro (67). These functional differences between C. albicans and C. neoformans EVs may be explained by distinct vesicle compositions, including the presence of species-specific proteins and complex carbohydrates, such as GXM (65,68). Notably, our results showing that EVs enriched in SGs and GXM contributed to disease control suggest that these antigens can also be used in immunotherapeutic approaches against cryptococcosis. However, this hypothesis requires validation in models of antigen administration after challenging mice with C. neoformans.
In conclusion, our study identified GXM as a key factor in induction of protection by a live attenuated C. neoformans vaccine under conditions of SG accumulation. Our results also suggest that the host immune protection most likely initiates in the lung and is perhaps mediated by macrophages, neutrophils, dendritic cells, and CD8 ϩ T RM but not by CD4 ϩ T cells. Further studies are necessary to identify the specific host cellular component(s) responsible for protection. Finally, our studies highlighted the potential use of fungal EVs enriched in GXM and SGs as a potential formulation for fungal vaccines. This cell-free formulation would be particularly attractive in treatment of immunocompromised patients at risk of developing cryptococcosis.

MATERIALS AND METHODS
Supplemental materials and methods. Supplemental materials and methods are described in Text S1 in the supplemental material.
Strains. The strains used were C. neoformans strain var. grubii H99 wild type (WT); mutant Δcap59, an acapsular strain kindly provided by NIH by June Kwon Chung; and mutant Δsgl1, a strain that accumulates sterylglucosides (SGs) and that was developed by our group in Maurizio Del Poeta's laboratory (7). Growth conditions are detailed in each experimental description. (10 6  (ii) Exo-GXM purification, quantification, and thin-layer chromatography (TLC). WT or Δsgl1 cells were grown in yeast extract-peptone-dextrose (YPD) overnight and transferred to MM for 48 h. Purification of secreted GXM was done by ultrafiltration as previously described (70). The amount of purified GXM from each strain was measured by phenol-sulfuric acid methodology in 96-well plates as described previously by Masuko and collaborators (71). To characterize the lipids in the purified GXM, 6 mg/ml of GXM from WT and Δsgl1 cells was lyophilized and lipid extraction was then performed (see details under "Lipid extraction, mass spectrometry (MS), and TLC" below). The samples were resuspended in 30 l of CHCl 3 /MeOH (2:1 ratio) and analyzed by TLC, developed with chloroform/methanol/water (65:25:4 [vol/vol/vol]), and stained with iodine and resorcinol.

Imaging of capsular GXM and purification and analysis of exo-GXM. (i) Scanning electron microscopy (SEM). Fungal cells
(iii) Dynamic light scattering (DLS). The GXM samples obtained from the different strains were diluted 100-fold in filtered Milli-Q water. The diameter and polydispersity of the obtained polysaccharide suspension were measured by the use of a DynaPro-99 DLS instrument (Wyatt Technology Corporation, Santa Barbara, CA) (72).
Lipid extraction, mass spectrometry (MS), and TLC. Lipid extraction was performed before every TLC and MS analysis in this study using the protocols of Singh and colleagues (73). For determination of SG levels in cell pellets, GXM, and EVs, we used different volumes (10 ml, 1 ml, and 50 l, respectively) from a culture of 1 liter. We dried the samples and extract the lipids for MS analysis. For mutant characterization, we cultivated the strains in yeast nitrogen base (YNB) broth for 24 h at 30°C with shaking and a pellet with 5 ϫ 10 8 cells was used for lipid extraction. The extraction followed three major steps: Mandala's method (74), followed by the methodology of Bligh and Dyer (75) and then base hydrolysis (73). The samples obtained were resuspended in chloroform/methanol (2:1) for the TLC, and purified SGs and GlcCer were used as references. For MS analysis, a standard lipid mix from Avanti Polar Lipids (Alabaster, AL) was prepared and added to the cells for calibration purposes prior to the first step of lipid extraction. Ergosteryl-3-␤-glucosides (SGs) and various sphingolipids and IPC were analyzed as described previously (37,76). The samples were resuspended in 150 l of ammonium formate (1 mM)-0.2% formic acid-methanol. The separation was processed through a Thermo Accela high-performance liquid chromatography (HPLC) system (San Jose, CA). The HPLC system was attached to the heated electrospray ionization (HESI) source of a Thermo TSQ Quantum Ultra triple quadrupole mass spectrometer (San Jose, CA). Data were analyzed on Thermo Xcalibur 2.2 Quan Browser software and normalized to inorganic phosphate content (7,76,77).
Sterylglucoside-Related Protection Requires GXM ® Purification of EV and quantitative analysis. EV isolation was performed for the WT, Δcap59, Δsgl1, and Δcap59/Δsgl1 strains following the protocol of Rodrigues and collaborators (78). After being maintained for 48 h in YNB, the culture was centrifuged for 15 min at 4,000, 10,000, and 15,000 ϫ g to separate the cell debris. The remaining supernatant was concentrated using an Amicon system and a membrane of 100 kDa. Next, the concentrated supernatant was ultracentrifuged at 100,000 ϫ g to isolate EVs from the supernatant. Quantification of sterols within the vesicles was performed using an Amplex red kit (Thermo Fisher Scientific), and, for more-specific quantification of the SGs (ergosteryl-3-␤glucosides), MS analysis was used. GlcCer was also analyzed using this technique.
Deletion of the SGL1 gene in the ⌬cap59 mutant. In order to delete the SGL1 gene (CNAG_05607) in the Δcap59 mutant, which is hygromycin resistant, we used the same disruption cassette, psgl1 (pCR-5UTR-NAT1-3UTR-TOPO), and a protocol similar to that used in the previously reported work involving the Δsgl1 mutant (7). Briefly, we used a biolistic DNA delivery method (79) to introduce the plasmid psgl1 into the Δcap59 mutant. Transformants from different colonies were cultivated in YPD and NAT-containing YPD plates (100 g/ml) to check stability and to select the colonies of transformants. The genomic DNA of these selected colonies was then isolated for Southern blot analysis. We used EcoRV and KpnI restriction enzymes to digest the DNA and two different probes to check the correct insertion of the fragment. We identified a band in the Southern blot using a 5= untranslated region (5=UTR) probe of 4,910 bp for the fungal cells that had the SGL1 gene intact and without any alteration in the gene sequence and a band of 3,604 bp for the fungal cells with a deletion of SGL1 and an insertion of NAT cassette in the correct position. For the gene probe, the cells that had the SGL1 gene disrupted did not present any visual band.
Double mutant characterization. (i) Growth curve. Fungal cells were grown in YPD overnight at 30°C with shaking. They were washed with PBS, and 10 5 cells/ml were added into 10 ml of Dulbecco's modified eagle medium (DMEM) at pH 4 or pH 7.4 at 37°C and 5% CO 2 . Samples of 100 l were taken after 24, 48, or 72 h and plated in YPD plates for CFU verification.
(ii) Morphology and capsule analysis. The cells were growth in YPD overnight at 30°C with shaking or in DMEM media at pH 4 or pH 7.4 at 37°C and in the presence of 5% CO 2 . The capsule was analyzed through counterstaining of the cells with India ink and examination under ϫ100 magnification. In addition, transmission electron microscopy (TEM) was performed to further characterize the capsule and cell structure. The protocol for the agarose embedding was similar to that described by Munshi and colleagues (76), and pictures were taken at magnifications of ϫ23,000 and ϫ49,000. A Leica EM UC7 ultramicrotome was used for sectioning. The sections were placed on 300-mesh copper grids and stained with uranyl acetate and lead citrate.
(iii) Characterization of virulence factors. Melanin and urease assays were performed at 37°C with 5% CO 2 . Plates for melanin testing were prepared with MM agar with addition of 1 mM L-3,4dihydroxyphenylalanine (L-DOPA). Fungal cells were spotted at different dilutions (10 7 , 10 6 , 10 5 , 10 4 , and 10 3 ) and visually analyzed after 96 h of cultivation. The production of melanin was determined by the dark color of the cells (80). Plates for urease test were prepared with Christensen's urea agar, and 5 ϫ 10 4 cells were spotted in the center of the plate. Cells were cultivated for 120 h and assessed by difference represented by the pink color of the colony and the size of the pink halo (81,82).
Mice. CBA/J mice (Envigo, Indianapolis, IN) (3 to 4 weeks of age) were used in this study. The animals had access to food and water ad libitum. All animal procedures were approved by the Stony Brook University Institutional Animal Care and Use Committee (protocol no. 341588) and followed the guidelines of American Veterinary Medical Association.
Infection, fungal burden, and histology. All the infections were performed intranasally. Animals were anesthetized with 60 l xylazine/ketamine solution intraperitoneally (95 mg of ketamine and 5 mg of xylazine per kilogram of animal body weight). Fungal cells, including WT, Δcap59, Δsgl1, and Δcap59/Δsgl1 cells, were grown overnight in YPD, counted, and resuspended in PBS. Inoculation of 7 ϫ 10 5 cells per animal in a volume of 20 l was performed through nasal inhalation. Animals were checked daily, and those that had more than 20% weight loss and appeared moribund or in pain were euthanized using CO 2 inhalation followed by cervical dislocation. For survival curve analysis, 8 animals per group were monitored during 60 days after infection. After this period, the lungs and brains of the animals that had survived (2 per group) were used for histology analysis. Hematoxylin and eosin (H&E) and periodic acid-Schiff (PAS) staining was performed. Histological analyses of the lung and brain of animals injected with the strains were also performed after 3 days of infection. Because the WT-injected mice were getting sick and dying at around day 14, we also performed histology in this time point for this strain. For fungal burden studies, 3 animals per strain were used at each time point. The homogenization of lung and brain was performed with 10 ml of PBS using a Stomacher 80 blender (Seward, United Kingdom). The samples were plated on YPD plates for CFU analysis after 48 h. Vaccination in murine model. For vaccination studies, we infected 10 mice with PBS or with Δcap59, Δsgl1, and Δcap59/Δsgl1 cells intranasally. We used the same techniques and numbers of cells used for survival curve analysis. After 30 days, animals were rechallenged with 7 ϫ 10 5 cells of the WT strain. Survival of mice was monitored for 60 days after the rechallenge.
Lung immune cell populations and cytokine analyses. The detection of immune cell populations and cytokines was performed after pulmonary infection model and after WT rechallenge in the vaccination model. The procedures used for the two models were the same as described previously, except for the inoculation of 5 ϫ 10 5 fungal cells and the time points. Separate groups of 3 mice per treatment each day were used for the two analyses. For the last time point (30 days), we did not have the WT-treated group since those mice had succumbed to infection previously. In the vaccination model, at 30 days after the WT challenge, only the Δsgl1-pretreated mice were alive. The immune cell populations were analyzed by flow cytometry. Lungs were harvested into sterile PBS, minced in fluorescenceactivated cell sorter (FACS) buffer with added collagenase and DNase I, homogenized through a 70-m-pore-size filter, lysed in Ack lysis buffer, washed with PBS, and filtered again, leaving a single-cell suspension. The cell suspensions were subjected to Fc blocking and then stained with a fluorescencelabeled antibody cocktail. The antibody cocktail included viability stain AF700, CD103-allophycocyanin (CD103-APC), major histocompatibility complex class II-fluorescein isothiocyanate (MHCII-FITC), LY6Cperidinin chlorophyll protein-Cy5.5 (LY6C-PerCPCy5.5), Ly6G-phycoerythrin (Ly6G-PE), Siglec F-PE Texas Red, CD11c-PE Cy7, F4/80-BV421, CD11b-BV510, CD4-BV605, CD19-BV650, CD45-BV711, and CD8-BV785. The stained cells were resuspended in FACS buffer, run on a BD LSR II flow cytometer (BD Biosciences), and analyzed using Flowjo version 10. The detection of cytokines in the lung tissue was analyzed using a Milliplex Mouse Cytokine/Chemokine Magnetic Bead Premixed 25 Plex kit (Millipore Sigma), following the 2-h incubation protocol and using PBS as the matrix for the samples. The samples were immediately run on a Bioplex 200 system running Bioplex manager version 4.1.1. Before homogenization of the lungs, BAL fluid was collected for a separate analysis (see the figures in the supplemental material). Cytokine measurements that fell either above or below the detection limits were excluded to reduce bias; therefore, the graphical illustrations of all relevant cytokines portray values within the detection limits of the assay.
Vaccination study with the EVs in an invertebrate model. For vaccination study with the EVs, G. mellonella was used in its final instar larval stage (60). Thirty larvae were injected with 10 l of PBS or an EV suspension of WT, Δcap59, Δsgl1, or Δcap59/Δsgl1 containing 2 M sterol (based on WT quantification). Injection into the hemocoel through the prolegs was performed using a Hamilton syringe. The animals were kept in a 37°C incubator in the dark. After 24 h, larvae were injected with 10 l of PBS containing 2 ϫ 10 4 cells of the WT strain. The larvae were monitored daily for survival.
Statistical analysis. Statistical analysis was performed using GraphPad Prism software version 7.0. Data are presented as means Ϯ standard errors of the means. One-way analysis of variance (ANOVA) or a two-way repeated-measure ANOVA was performed followed by a Tukey's post hoc test for multiple comparisons. For survival curve determinations, statistical analysis was performed using the log rank (Mantel-Cox) test for survival analyses, and data represent the mean percentages of survival.