Human and murine Cryptococcus neoformans infection selects for common genomic changes in an environmental isolate

A pet cockatoo was the suspected source of Cryptococcus neoformans recovered from the cerebral spinal fluid (CSF) of an immunocompromised patient with cryptococcosis based on the molecular analyses available in 2000. Here we report whole genome sequence analysis of the clinical and cockatoo strains. Both are closely related MATα strains belonging to the VNII lineage, confirming that the human infection likely originated from pet bird exposure. The two strains differ by 61 single nucleotide polymorphisms, including 8 nonsynonymous changes involving 7 genes. To ascertain whether changes in these genes are selected during mammalian infection, we passaged the cockatoo strain in mice. Remarkably, isolates obtained from mouse tissue possess a frame-shift mutation in one of the seven genes altered in the human sample, a gene predicted to encode a SWI-SNF chromatin-remodeling complex protein. Both cockatoo and patient strains as well as mouse passaged isolates obtained from brain tissue had a premature stop codon in a homolog of ZFC3, a predicted single-zinc finger containing protein, which is associated with larger capsules when deleted and appears to have reverted to a full-length protein in the mouse passaged isolates obtained from lung tissue. The patient strain and mouse passaged isolates show variability in the expression of virulence factors, with differences in capsule size, melanization, and rates on non-lytic expulsion from macrophages observed. Our results establish that environmental strains undergo genomic and phenotypic changes during mammalian passage, suggesting that animal virulence can be a mechanism for genetic change and that the genomes of clinical isolates may provide a readout of mutations acquired during infection.


26
Abstract 27 A pet cockatoo was the suspected source of Cryptococcus neoformans recovered from the 28 cerebral spinal fluid (CSF) of an immunocompromised patient with cryptococcosis based 29 on the molecular analyses available in 2000. Here we report whole genome sequence 30 analysis of the clinical and cockatoo strains. Both are closely related MATα strains 31 belonging to the VNII lineage, confirming that the human infection likely originated from 32 pet bird exposure. The two strains differ by 61 single nucleotide polymorphisms, 33 including 8 nonsynonymous changes involving 7 genes. To ascertain whether changes in 34 these genes are selected during mammalian infection, we passaged the cockatoo strain in 35 mice. Remarkably, isolates obtained from mouse tissue possess a frame-shift mutation in 36 one of the seven genes altered in the human sample, a gene predicted to encode a SWI-37 SNF chromatin-remodeling complex protein. Both cockatoo and patient strains as well as 38 mouse passaged isolates obtained from brain tissue had a premature stop codon in a 39 homolog of ZFC3, a predicted single-zinc finger containing protein, which is associated 40 with larger capsules when deleted and appears to have reverted to a full-length protein in 41 the mouse passaged isolates obtained from lung tissue. The patient strain and mouse 42 passaged isolates show variability in the expression of virulence factors, with differences 43 in capsule size, melanization, and rates on non-lytic expulsion from macrophages 44 observed. Our results establish that environmental strains undergo genomic and 45 phenotypic changes during mammalian passage, suggesting that animal virulence can be 46 Introduction 50 51 Cryptococcus neoformans is a human pathogenic fungus that is a major cause of life-52 threatening meningoencephalitis (1). Cryptococcosis is more common in patients with 53 impaired immune systems although occasional disease occurs in individuals with no 54 apparent immune deficits. C. neoformans infection is first thought to be acquired in 55 childhood (2) and is either cleared or can become latent to reactivate if impaired 56 immunity occurs later in life (3). However, disease can also follow exposure to 57 contaminated environmental sources in adults, but the ubiquity of this fungus 58 complicates the identification of point sources. Restriction enzyme polymorphism 59 analysis of patient and environmental samples in New York City revealed that some 60 clinical isolates shared the same restriction patterns and were thus indistinguishable from 61 local infection sources (4), but such analysis lacked the precision to reveal point sources, 62 particularly given that the disease often develops slowly and can be the result of latent, 63

distantly acquired, infection (5). A recent investigation of a cryptococcosis outbreak in a 64
Scottish hospital revealed how difficult it is to make associations between clinical, and 65 geographically and temporally matched environmental samples (6). 66 In 2000, we reported on the case of an immunosuppressed patient with cryptococcosis 67 who had a pet cockatoo (7). Cryptococcus was also recovered from the cockatoo guano, 68 which was not unexpected as bird guano is a common environmental reservoir for C. 69 neoformans (8,9). Both the patient and bird guano strains were indistinguishable using 70 molecular tools available at the time, resulting in the first example of cryptococcal 71 infection traced from a point source (7). In subsequent years additional cases of C. 72 neoformans infections linked to pet birds were reported (10, 11). Since the original report, 73 genomic sequencing has become routine and there is now a wealth of information 74 available on C. neoformans genomes (12). In this study, we compared the genome 75 sequences of the patient and cockatoo strains and passaged the cockatoo strain in mice to 76 identify genetic and phenotypic changes resulting from mammalian infection. The results 77 validate the earlier conclusion that the clinical and cockatoo strain sequences were closely 78 related, and that the cockatoo was the likely source of infection. We also find evidence of 79 similar genome evolution in mouse-passaged and patient strains. These results implicate 80 molecular weight cut-off (MWCO) centricon filtration unit. The > 3 kDa fraction of EPS 167 was then characterized by NMR. 168 NMR analysis. 1D 1 H NMR data were collected on a Bruker Avance II (600 MHz), 169 equipped with a triple resonance, TCI cryogenic probe, and z-axis pulsed field gradients. 170 Spectra were collected at 60°C, with 128 scans and a free induction decay size of 84336 171 points. Standard Bruker pulse sequences were used to collect the 1D data (p3919gp and 172 zggpw5 colonies were selected from brain and lung CFU plates and used to seed 100 mL YPD 184 cultures, which were allowed to grow for 48 h at 30°C with rotation. Genomic DNA was 185 isolated from each culture following the protocol described in Velegraki et al. (14)  , and variants were called and filtered as described above. A 214 maximum likelihood phylogeny was estimated using segregating SNP sites present in one 215 or more isolates, allowing ambiguity in a maximum of 10% of samples, with RAxML 216 v8.2.12 (21) rapid bootstrapping (GTRCAT substitution model), and visualized with 217 ggtree (R 3.6.0) (22). Aneuploidies were visualized using funpipe (coverage analysis) 218 v0.1.0 (https://github.com/broadinstitute/funpipe), transposon mobilization was 219 assessed through whole genome alignment of the CU and PU assemblies with nucmer v3.1 220 to identify alignment gaps, and copy number variation was assessed using CNVnator v0.3 221 (23). 222 Urease activity assay. C. neoformans strains and isolates were first grown in YPD for 48 223 h at 30˚C. Urea broth comprised of 10 mM KH2PO4, 0.1% Bacto Peptone (Difco), 0.1% D-224 glucose, 0.5% NaCl, 2% urea, and 0.03 mM phenol red, as described by Roberts et al.,225 (24), was inoculated with PBS-washed cells at a density of 1 ´ 10 6 cells/mL for each strain 226 in triplicate. After incubation for 16 h at 30˚C, increased pH of culture media that is 227 indicative of ammonium production due to urease activity was detected by measuring 228 absorbance at 560 nm over a 6 h time course. Absorbance readings that had been 229 corrected by subtraction of media-only background absorbance were plotted against time. 230 Urease activity rates were derived by a simple linear regression model and data were 231 analyzed for statistical significance using an ordinary one-way analysis of variance 232 (ANOVA) with GraphPad Prism 9 software. 233 C. neoformans melanization assay. C. neoformans strains and isolates were first grown 234 in YPD liquid media for 48 h at 30˚C until the cultures were in stationary phase. Cells 235 were washed twice in PBS. 100 µl of washed culture was added to 5 mL Minimal Media 236 with 1 mM L-3,4-dihydroxyphenylalanine (L-DOPA) and grown for 5 d at 30˚C. Cultures 237 were removed and placed into a 6-well plate for imaging. Alternatively, 1 ´ 10 6 PBS-238 washed cells were spotted onto L-DOPA agar in triplicate. Plates were incubated at 30˚C 239 or 37˚C and then photographed after 2, 3, and 6 days using a 12-megapixel camera. Color 240 images were converted to grayscale using Adobe Photoshop, pigmentation intensities 241 were quantified using Image Studio Lite software, and graphed using GraphPad Prism 242 software. Statistical significance was determined with the ordinary one-way ANOVA test, 243 via GraphPad Prism. 244 Phospholipase activity. Extracellular phospholipase in C. neoformans strains and isolates 245 was tested by the modified method reported by the Chen et al. (25). Egg yolk agar medium 246 was created based on Difco Sabouraud Dextrose Agar media with 8% egg yolk, 1M sodium 247 chloride and 0.05M calcium chloride. After overnight growth in YPD media 3 ul of each 248 strain, with a total of 10,000 cells, were spotted onto egg yolk agar medium. Each strain 249 and isolate was tested on five separate plates and incubated at 30°C. After 72 h and 96 h 250 colonies were photographed and measured with ImageJ software. Activity of 251 phospholipase was analyzed by the ratio of colony diameter to total precipitation 252 diameter, where a ratio equal to 1.0 indicates a lack of phospholipase activity. Statistical 253 significance was determined by an unpaired t-test test, via GraphPad Prism. 254 Heat-ramp and thermal stability analysis. Strains and isolates were maintained at -80°C 255 in glycerol, streaked onto Sabouraud dextrose (SAB) (BD Difco) agar and incubated at 256 30°C for 48 h prior to heat-ramp cell death assays. SAB broth was inoculated from growth 257 on plates to equal densities (OD 0.1) and incubated at 30°C for 18 h in stationary 96-well 258 plates. Isolates were resuspended, diluted 1:5 in fresh SAB broth, and 100 µl was treated 259 with a linear 30°C to 56°C heat-ramp stress over 10 min in a water bath with agitation 260 (Lauda). Untreated and heat-ramp treated strains were immediately spotted (5 µl in SAB) 261 in 5-fold serial dilutions on SAB agar and incubated at 30°C for 48 h to assess viability. 262 CFUs before and after heat-ramp assays were enumerated to calculate relative survival. 263 To determine growth differences at high temperature, untreated strains were spotted on 264 SAB agar and incubated at 37°C for 48 h. Statistical significance was determined with the 265 Anova test, via GraphPad Prism. The relationship of the original strains, previously described (7), is shown in Figure 1, 289 with the original patient strain referred to as patient unpassaged (PU), and the original 290 cockatoo strain referred to as cockatoo unpassaged (CU). Isolates recovered from mice 291 infected with the CU strain are identified as cockatoo passaged brain (CPB) and cockatoo 292 passaged lung (CPL) to denote the tissues from which they were isolated. C. neoformans 293 recovered from patient and cockatoo are referred to as strains to denote different origins 294 while those recovered from mouse tissues are referred to as isolates. Here, we aimed to 295 identify the relatedness of the CU and PU strains, examine genetic changes resulting from 296 mammalian passage of these strains, and interrogate differences across virulence 297 phenotypes arising from passage across hosts and body sites. 298 299 Genomic analysis of patient and cockatoo strains. To determine the relatedness of the PU 300 and CU strains, Illumina reads were aligned to the C. neoformans H99 reference genome 301 (33), and variants were called to identify single nucleotide polymorphisms (SNPs) and 302 insertion/deletion events (indels). Based on a phylogenetic analysis of these samples and 303 a subset of 238 additional whole genome sequences chosen to represent lineages VNIa, 304 VNIb, VNIc, VNBI, VNBII, and VNII, previously described by Desjardins et al. (34), we 305 assigned the PU and CU strains to the globally detected VNII lineage confirming their 306 close relationship (Fig. 2, Supplementary Fig. 1), and determined that both strains 307 possess the MATα mating type. Utilizing ONT read data, we then generated genome 308 assemblies for both PU and CU strains, consisting of complete telomere-to-telomere 309 sequences for each chromosome, polished with Illumina reads. To identify variants, we 310 compared the two assemblies with nucmer (MUMmer). We identified 7 genes with 311 nonsynonymous variants between the two assemblies, all impacting genes with homologs 312 in the C. neoformans var. grubii (H99) genome (Table 1). Of these genes, 2 313 (LQVO5_002184 and LQVO5_000317) had C. neoformans H99 homologs 314 (CNAG_06273 and CNAG_00342) that are repressed during titan cell formation and 315 murine cryptococcal infection, respectively (35,36). For both strains, we were able to 316 generate full-length chromosomal assemblies, consisting of 14 chromosomes, with rRNA 317 content residing on chromosome 4, and chromosome 15 representing the mitochondrion, 318 with equivalent gene sets and high levels of identity between the two assemblies (99.99%), 319 indicating high levels of relatedness (Supplementary Fig. 2a). When these assemblies 320 are compared to the C. neoformans H99 reference assembly, we see sequence identities 321 of 97.7%, for both CU and PU strains to the H99 reference (Supplementary Fig. 2b). the CU strain and observed for signs of illness, but none were apparent. Hence, at day 43 328 the mouse was sacrificed, brain and lung were harvested and homogenized, and the 329 suspensions plated, which yielded 816 and 78 CFU/mg in brain and lungs, respectively. 330 Three individual colonies were selected from the brain (CPB1-3) and lung (CPL1-3) CFU 331 plates. These isolates were sequenced and studied for phenotypic characteristics. 332 333 Genomic analysis of mouse-passaged isolates. To identify variants arising from the 334 passage of the CU strain in mice, we aligned Illumina data generated for all mouse evolved 335 isolates to our CU genome assembly. Samples aligned with an average coverage of 775X 336 across the CU reference. We found a frameshift variant in one of the seven genes altered 337 in the patient strain, LQVO5_000317 (Table 1). This frameshift in LQVO5_000317 was 338 present in all mouse evolved isolates from both brain and lung tissue (CPB1-3, CPL 1-3) 339 ( Table 2). The commonality of this variant across brain and lung isolates suggests the 340 mutation was acquired at a common site prior to dissemination. This gene is a homolog 341 of CNAG_00342 in the C. neoformans H99 (VNI) genome, which is predicted via 342 eukaryotic orthologous groupings to function as a SWI-SNF chromatin-remodeling 343 complex protein (KOG2510), and is down-regulated in a murine lung infection model 344 (36), consistent with loss-of-function mutations in mammals.
A second variant impacting only the mouse passaged isolates collected from lung tissue 347 resulted in the loss of a premature stop codon in LQVO5_004463, a truncated homolog 348 of CNAG_05940, a predicted Zinc-finger domain protein (ZFC3). A premature stop 349 codon truncating the ZFC3 homolog (LQVO5_004463) is present in both the CU and PU 350 strains, as well as the CPB isolates, but appears to have reverted to wild type in the CPL 351 isolates. When LQVO5_004463 is extended through the loss of this premature stop 352 codon, the resulting gene encodes a full-length protein comparable to CNAG_05940 in 353 sequence and structure (Supplementary Fig. 3). Zfc3 has a single predicted zinc finger 354 and a striking number of serine and threonine residues in the LQVO5_004463 homolog, 355 CNAG_05940, totaling 21.9% of all residues present. The predicted structure of this 356 protein includes long intrinsically disordered regions that may transition to structured 357 regions upon binding to a substrate (Supplementary Fig. 3c). Interestingly, deletion 358 of CNAG_05940 in C. neoformans results in strains with significantly increased capsule 359 content (37), and this gene is thought to be a target of the virulence implicated 360 transcription factors Gat201 and Liv3 (38). The presence of variants impacting shared 361 genes in both patient and cockatoo-derived mouse isolates is consistent with the 362 hypothesis that the patient's C. neoformans infection was also derived from the pet 363 cockatoo. 364

365
To assess the frequency of loss-of-function mutations in the SWI-SNF and ZFC3 366 homologs among clinical and environmental Cryptococcus isolates, we looked for loss-367 of-function variants in 387 published isolates from both patient and environmental 368 sources (34). We found 3 clinical samples (from the lineages VNI, VNII, and VNB) with 369 frameshift variants in CNAG_00342 (SWI-SNF) and 24 samples with frameshift variants 370 present in CNAG_05940 (ZFC3) ( Table 2). VNI and VNB isolates from both clinical and 371 environmental sources are impacted by frameshift variants in CNAG_05940. To identify 372 large-scale genomic variation in these CU, PU, CPB and CPL isolates, we looked for 373 evidence of aneuploidy and copy number variation (CNV) based on sequence coverage, 374 however, we saw no evidence of either aneuploidy or significant CNVs arising in response 375 to human, bird, or murine passage. 376 CNAG_00342 and CNAG_05940 expression in published datasets. To further probe the 378 role of these genes altered in the patient and mouse isolates, we analyzed eight publicly 379 deposited RNA-seq datasets of C. neoformans strains H99 and KN99α (Table 3) Both genes were mostly upregulated in H99 strains and downregulated in KN99α rather 386 than between conditions with the caveat that many of the H99 strains were exposed to 387 ambient CO2 levels while both KN99α datasets were exposed to 5% CO2, indicating strain 388 and condition specific changes that highlight variability in expression. 389

390
Heat tolerance. To determine if resistance to cell death may be related to virulence in 391 these strains, we compared the mouse passaged isolates to the patient strain in a cell death 392 assay that has been previously demonstrated to induce gene-dependent cell death in S. 393 cerevisiae, and more recently in C. neoformans (39,40). To assess cell death 394 susceptibility, a transient sublethal heat-ramp (not heat shock) was applied to all isolates 395 and survival was determined by CFUs when plated at 30°C. Interestingly, all mouse, 396 cockatoo, and patient strains were death-resistant when compared to the lab strain, 397 KN99α (Fig. 3A). Among the isolates tested, the three mouse-passaged brain isolates 398 (CPB1-3) were significantly more death-resistant than the lung isolates (CPL1-3), and the 399 patient (PU) strain (p=0.0324 and p = 0.002, test=ANOVA) (Fig. 3B). However, cell 400 death-resistance does not appear to reflect a gain of heat tolerance at body temperature 401 as all isolates grew indistinguishably when untreated samples were plated on SAB agar 402 and incubated at 37°C (Fig. 3C). Although the lab strain KN99α may be slightly more 403 robust than the CU, PU, CPB, and CPL isolates at both 30°C and 37°C, no differences in 404 CFU number or size among the isolates was observed, suggesting that heat-ramp cell 405 death resistance is not directly correlated with the ability to grow at high temperature. 406 407 Growth in vitro and virulence factor expression. We analyzed both the CU and PU strains 408 and the CPB and CPL isolates for growth in vitro (Fig. 4) and expression of four phenotypes known to be associated with virulence factors: capsule, melanin, urease, and 410 phospholipase (Figs. 5, 6). All four isolates grew well in culture with the CPB1 isolate 411 recovered from brain tissue growing faster than the others (Fig. 4). All isolates expressed 412 each of these virulence factors but there were subtle differences observed. The original 413 cockatoo strain (CU), patient strain (PU), and mouse passaged CU derivative isolates 414 varied little with regards to the capsule, except that CPL1 cells had significantly smaller 415 capsules than the other isolates (p=1x10 -4 , 4x10 -8 , 3.8x10 -6 , CPL1 vs CU, PU, CPB1, 416 respectively; test= paired t-test) (Fig. 5), consistent with a functional ZFC3. Since a prior 417 analysis of sequential isolates recovered from individual patients showed changes in 418 polysaccharide structure (41) which provides a spectral signature for the repeating triad of glucuronoxylomannan (4.8-423 5.4ppm) (42) showed the same peak-set for CU and PU strains and CPB and CPL isolates 424 (4.91, 4.97, 4.99, 5.10, 5.12, 5.15. 5.16, 5.20) (Fig. 5C), implying conservation of EPS 425 structure in these strains and isolates. Isolates from the brain of the CU-infected mouse, 426 CPB1-3, had faster rates and higher total percentages of melanization when compared to 427 the other isolates, including the parental CU strain, the PU strain, and the CPL1-3 isolates 428 from the lungs of the mouse (Fig. 6A-C). The CPB1 isolate was significantly more 429 melanized than CU at 37˚C when grown on L-DOPA-agar (p=0.0365, 0.0009, 0.0002; 2, 430 3, 6 d; test = ANOVA). For phospholipase, the size of the precipitation zone varied across 431 isolates and increased with the time of incubation. After 72h of incubation at 30˚C, the 432 isolate recovered from the mouse brain (CPB1) presented significantly lower (p = 0.033, 433 test=t-test) phospholipase activity than the original CU strain. After 96h the 434 phospholipase activity of CPB1 was significantly lower (p = 0.0049, test=t-test) than in 435 all other strains and isolates (Fig. 6D). In contrast, the expression of urease was 436 comparable among the four isolates (Fig. 6E). 437 438 Virulence in Galleria mellonella. To identify strain differences in virulence, we tested the 439 CU, PU, CPB, and CPL isolates in the invertebrate model G. mellonella (Fig. 6F). This 440 model has been used previously to compare virulence of Cryptococcus isolates, which roughly correlates to virulence in mammalian models (1,2). We found no statistical 442 differences between the virulence of the CU and PU strains. There was enhanced virulence 443 of the CPB1 brain isolate compared to its parental CU strain, and reduced virulence of the 444 CPL1 lung isolate. While all strains and isolates tested were virulent, these results show 445 subtle changes in virulence for an insect host. 446

447
Interactions with macrophages. The interaction of C. neoformans with macrophages is 448 unusual in that ingestion results in transient intracellular residence, which may be 449 followed by non-lytic exocytosis whereby the fungal cell exits the phagocytic cell without 450 lysing the latter (43, 44). Given that this process involves a complex choreography of 451 cellular events that must occur in synchrony we considered it a sensitive indicator of C. 452 neoformans-macrophage interaction, and examined its frequency for the CU, PU, CPL, 453 and CPB isolates (Fig. 7). The results show that the CPB1 isolate manifested a 454 significantly higher frequency of non-lytic exocytosis relative to the other strains and 455 isolates (p=0.013, test of equal proportions with Bonferroni correction). 456 457 Interactions with amoeba. Since amoebae are predators of C. neoformans and amoeba-458 fungal interactions have been proposed to select for traits that function during 459 mammalian virulence, we evaluated whether passage through human and mice affected 460 the interaction with A. castellanii. In an assay favoring amoeba predation through the 461 presence of divalent cations (45), we observed that C. neoformans strains recovered from 462 the patient (PU) and cockatoo (CU), as well as mouse-passaged isolates (CPL and CPB) 463 were equally susceptible to predation (Fig. S4). All fungal strains experienced minimal 464 killing by A. castellanii at 24 h, and modest CFU rebounds by 48 h, but differences 465 between the CU and PU strains or CPL and CPB isolates were not significant (Fig. S4A). 466 This pattern of initial CFU decrease and subsequent rebound was not seen for the same 467 strains or isolates incubated in only DPBS (Fig. S4B). To ascertain whether we could reproduce some of the changes observed in the patient 503 strain relative to the cockatoo strain we passaged the latter in mice. In all mouse passaged 504 isolates we observed one frameshift variant that impacts the same gene (LQVO5_000317, homolog of the H99 gene CNAG_00342) observed in the patient strain. A second variant 506 impacting only the mouse passaged isolates collected from lung tissue results in the loss 507 of a premature stop codon for LQVO5_004463, a homolog of CNAG_05940 (ZFC3). 508 ZFC3 is a transcription factor that is repressed during titan cell induction in addition to 509 being a target of the transcription factors Gat201 and Liv3 (38). Changes in length of the 510 protein encoded by LQVO5_000317 are likely to impact expression of this gene's targets. 511 Analysis of the expression of CNAG_00342 and CNAG_05940 in publicly available 512 databases involving conditions related to animal passage and infection revealed variation 513 in the expression of these genes, however, it is unclear whether the condition or genetic 514 background is responsible for this variation. Hence, it may be that the genomic changes 515 observed in the cockatoo strain upon mouse or human passage represent selection of this 516 particular genotype during infection, which may be specific to this genetic background. 517 Variation of gene expression across strains, correlating with genetic groupings and 518 lineage, have been observed for clinical and environmental isolates (50), and isolates 519 derived directly from human CSF (51). Consequently, we caution against generalization 520 from the results to other cryptococcal strains until there is greater sampling of changes 521 associated with virulence; given the plasticity of the C. neoformans genome, there may be 522 many solutions to the problem of survival in the ecologic niche defined by these hosts. 523

524
The occurrence of numerous C. neoformans genetic changes in the form of SNPs, 525 deletions, and insertions suggests that fungal replication in mammalian hosts may select 526 for specific changes. Although this is the first genomic study of a C. neoformans strain 527 before and after human passage, other studies have reported genetic changes during 528 infection. Analysis of serial isolates from patients shows chromosome rearrangements, 529 ploidy alterations, SNPs, insertions, and deletions (52). One mechanism of mutagenesis 530 is transposon mobilization during infection (53), however, we saw no evidence of 531 transposon mobilization upon mammalian passage. C. neoformans replicates within 532 macrophages in vivo (54), and internalized fungal cells are exposed to oxygen-and 533 nitrogen-derived radicals that can be mutagenic (55). A recent study of C. albicans during 534 in vitro and in vivo passage suggested a higher mutation rate in vivo (56). The fact that 535 mammalian infection may select for specific genetic changes in C. neoformans suggests 536 that those microbes capable of prolonged residence in hostile environments such as human hosts can acquire genetic changes, such that the capacity for virulence could 538 provide a shortcut to greater genetic variation. 539 540 Our observation that human and mouse passage was associated with the emergence of 541 genetically different variants has important implications for the understanding of C. 542 neoformans genomics, virulence, and pathogenesis. Furthermore, the finding that 543 passage of an environmental strain of C. neoformans through humans and mice resulted 544 in genetic changes suggests that clinical strains may have been modified by residence in 545 human hosts, where they are exposed to higher constant temperature and must survive 546 attack by the immune system. However, these changes did not confer increased resistance 547 to amoeba, in contrast with previous findings that the passage of environmental isolates 548 in amoebae results in both genetic changes and the emergence of pleiotropic phenotypes 549 (57). Since Entamoeba spp. and slime molds can occur in bird feces (58, 59), the CU strain 550 may already be maximally resistant to amoeba predation. Comparison of genomic 551 changes in mammalian and amoeba passaged Cryptococcus revealed no obvious 552 commonalities, consistent with the notion that even whilst these hosts provide similar 553 challenges to C. neoformans, such as surviving phagocytosis and phagosome residence, 554 they constitute different selective environments. Overall, these results indicate that the 555 cryptococcal genome is highly malleable such that genetic changes can accumulate 556 rapidly. 557 558 Comparison of clinical and environmental isolates for genomic differences such as the 559 ones found in this study revealed multiple instances of frameshift variants in both 560 CNAG_00342 and CNAG_05940 in clinical and environmental isolates spanning 561 lineages VNI, VNII, and VNB (34). How these genetic changes affect virulence and 562 pathogenesis is a question for future studies. In this regard, a comparison of the virulence 563 of 10 clinical and 11 environmental C. neoformans isolates in mice revealed that 7 clinical 564 isolates and only 1 environmental isolate caused lethal infection (60). Considering these 565 results in the light of our findings suggest that the clinical strains in that study were 566 perhaps more virulent because of genetic changes that occurred or were selected for by 567 human passage. 568 The comparison of C. neoformans characteristics associated with virulence among 570 cockatoo, patient, and mouse passaged isolates in this study revealed subtle phenotypic 571 changes. There were differences in average capsule size between isolates recovered from 572 mouse lung and brain, but NMR analysis of the major polysaccharide component revealed 573 no major changes. The finding of organ related differences in capsule size is consistent 574 with prior reports (61). However, the finding that the GXM structure was unchanged 575 contrasted with the prior observation that serial clinical isolates from patients with 576 persistent infection manifested changes in the polysaccharide structure (41), suggesting 577 that for the CU strain studied here the polysaccharide structure was more stable. The 578 major difference in the polysaccharide structures involved the extent of acetylation, with 579 the CPB isolate having more than others. No major differences were observed in urease 580 expression, but the mouse-passaged brain isolates manifested faster melanization. 581 Despite melanin variability across the CPL isolates, we did not see genetic variation within 582 genes known to be involved in melanization in these isolates. Similarly, mouse-passaged 583 isolates and patient strains grew better at higher temperatures than the cockatoo strain 584 possibly reflecting a period of adaptation to thermal mammalian conditions. In general, 585 there were no major changes in phenotypes associated with virulence during mammalian 586 passage consistent with the notion that these attributes exist primarily for environmental 587 survival and only accidentally confer upon C. neoformans the capacity for mammalian 588 virulence. Nevertheless, we do note that the mouse passaged isolates recovered from 589 brain tissue manifested faster growth, higher melanization, increased rates of non-lytic 590 escape, and killed moth larvae faster, consistent with a relative gain in virulence during 591 animal passage. These inter-strain and -isolate phenotypic differences highlight the 592 tremendous variation apparent in closely related C. neoformans strains, a phenomenon 593 that contributes to virulence (62) and is also apparent in pleiotropic variants generated 594 by amoeba predation (57) and phenotypic switching (63). 595

596
We note with interest that the mouse passaged isolates recovered from brain tissue (CPB) 597 were more resistant to thermal stress. Since the mouse passaged isolates recovered from 598 lung tissue (CPL) did not show this phenotype we cannot attribute this to simple thermal 599 adaptation to mammalian temperatures. Furthermore, the comparison of genetic 600 variants between CPB and CPL isolates did not reveal genetic changes that are known to confer increased thermal stability. Consequently, the most likely explanation for 602 increased thermal tolerance in CPB isolates is epigenetic change, possibly associated with 603 altered metabolic states, which may allow for greater survival during rapid heating. Brain 604 and lung environments are expected to differ in catecholamine concentrations (64) and 605 inflammatory responses (65-67). Although a mechanistic investigation of this 606 phenomenon is beyond the scope of this paper, we note that if this phenomenon occurs 607 in nature, it could provide environmental fungi capable of mammalian infection with a 608 mechanism for rapid thermal adaptation that could increase their fitness during climate 609 change. This in turn raises the specter that fungi capable of mammalian infection could 610 increase in prevalence with warming climates. 611

612
In summary, genomic analysis of cockatoo, human, and mouse passaged isolates strongly 613 supports our earlier inference that human infection resulted from exposure to a pet 614 cockatoo. In this study, the comparison of cryptococcal genomes of the incident bird and 615 patient strains with mouse-passaged isolates revealed the occurrence of common genetic 616 changes during mammalian passage. Previously we showed that passage of C. 617 neoformans in mice promotes the appearance of new electrophoretic karyotypes (69). 618 Similarly, in the ascomycete C. albicans, infection was associated with larger changes in 619 heterozygosity during murine passage (56) and human infection (70) than occur in vitro. 620 The capacity for virulence in pathogenic microbial species is not without cost as evident 621 by genome reduction and host specialization (71). The finding that mammalian infection 622 promotes genomic changes in both C. neoformans and C. albicans suggests that the 623 capacity for virulence can provide a mechanism for more rapid evolutionary change 624 through selection and adaptation in the mammalian host, which brings new parameters 625 for consideration when evaluating the cost-benefit equation for mammalian virulence in 626 pathogenic fungi. 627   two temperatures expressed as a ratio of linear phase slopes indicates a significant growth 675 advantage at 37˚C for the mouse-passaged brain isolate compared to the other strains and 676 isolates. (C) Both the patient strain and mouse-derived isolates show a significant 677 decrease in the length of lag time at 37˚C compared to 30˚C. Statistical significance was 678 determined using an ordinary one-way ANOVA (ns = not significant, *p < 0.05, **p < 679 0.01, ***p < 0.001, ****p < 0.0001). after growth at 30˚C for 5 days. There are no major differences in melanization between 695 the PU strain or CPL isolates, which generally show melanization consistent with the 696 parental strain (CU). (B-C) Scatter plot graphs (upper panel) and representative images 697 (lower panels) of pigment production expressed as a percentage of the H99 reference 698 strain for the cockatoo (CU) and human (PU) strains, and mouse-passaged brain (CPB) 699 and lung (CPB) isolates grown for the indicated amount of time on L-DOPA-agar at 700 either 30˚C (B) or 37˚C (C). Compared to the original cockatoo strain, only the mouse-701 derived brain isolate shows a significant increase in pigmentation at 37˚C. (D) C. 702 neoformans were inoculated onto egg yolk agar and incubated at 30C. After 72h and 703 96h of incubation phospholipase production was analyzed by measuring the ratio of 704 colony diameter to precipitate + colony diameter on the plate, where a ratio value equal 705 to 1.0 indicates a lack of phospholipase activity. (E) Time course of urease activity for 706 the indicated strains of C. neoformans grown at 30˚C in urea broth. Increased pH of 707 culture media that results from the conversion of urea to ammonium was quantified by 708 measuring the absorbance of cell culture media at 560 nm relative to a cell-free control. 709 Urease activity rates were not statistically different between strains. (F) There is no 710 statistical difference in virulence between the CU and PU strains in the G. mellonella 711 model system. However, the CPB1 isolate from the mouse had significantly enhanced 712 virulence when compared to the parental CU strain. Log-rank Mantel-Cox test was 713 performed using GraphPad PRISM and corrected for multiple comparisons using the 714 Bonferroni method. 715 716 Figure 7. Non-Lytic escape frequency of each strain during BMDM infection.

718
CPB was the only isolate with a significantly increased frequency of non-lytic escape 719 compared to the original CU strain. * Indicates P < 0.05 via a test of equal proportions 720 with Bonferroni correction. 721 722