Mammalian-evolved strains of the fungal pathogen Cryptococcus neoformans show adaptation via heritable transcript level differences specific to mouse MHC haplotype

Abstract

Background

Organisms adapt to novel environments using changes to genome, gene expression, and protein functions. This study focused on changes that had occurred when a fungal pathogen previously encountered hosts that differed only at the major histocompatibility complex (MHC) region, loci that control immune recognition during the adaptive immune response. To investigate how this fungal pathogen adapted to the host environment, next generation sequencing data were examined from strains of Cryptococcus neoformans (C. neoformans) that had been previously passaged eight times through congenic mice that specifically differed at the MHC locus, H2. Transcript levels and the genomic sequence for each post-adapted fungal strain were examined to identify molecular adaptation strategies via heritable gene expression changes (epigenetic changes) and mutation (DNA changes).

Results

The post-adapted strains displayed repeated changes in transcript levels, as determined by RNA-sequencing. Some of these epigenetically regulated genes (ERGs) only occurred in strains passaged in MHC specific hosts, suggesting possible prior adaptations to specific host MHCs. To our knowledge, this is the first time ERGs have been reported as possible pathogen adaptations to specific host MHC alleles. Additionally, of the total 47 single nucleotide polymorphisms (SNPs) identified, 7 SNPs each were found in 2 or more fungal strains (of the 6 analyzed strains) passaged through different MHC congenic hosts, suggesting much of the fungal mutation-based adaptation was to the mouse host, and not MHC-specific.

Conclusions

These data demonstrated that these passaged C. neoformans strains adapted to the MHC haplotype of a novel mammalian host environment primarily via epigenetic rather than mutation-based mechanisms.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12864-026-13196-7.

Background

Cryptococcus neoformans (C. neoformans) is a global fungal pathogen found in most soil contaminated with bird droppings [1]. Infection with C. neoformans typically presents as either a respiratory infection or meningitis, causing severe illness primarily in the immunocompromised [1], such as in people with AIDS or transplant recipients, where infection is deadly [2, 3]. While the global incidence of cryptococcosis has decreased in recent years, current estimates place it as causing roughly 19% of all AIDS-related deaths [3].

The first line of defense against C. neoformans are innate immune cells, including macrophages in the lung. In addition, other antigen-presenting cells, such as dendritic cells and B cells engulf and phagocytize the invading pathogen and load pathogen-derived peptides onto major histocompatibility complex (MHC) Class II molecules in mice or human leukocyte antigen complex (HLA) in humans [4]. This MHC: peptide complex, which is presented on the cell surface, activates antigen-specific T cells, which activate other immune cells to initiate a response to the infection [5].

Mice have 30 MHC Class I and 9 MHC Class II genes located on a region of chromosome 17 known as the H2 locus; each shows a high degree of genetic variation [6]. The entire H2 locus in mouse inbred strains is designated by a haplotype letter, such as b, k, or q. Because MHC genes activate T cells and the immune response, specific versions of MHC genes (alleles) affect an individual’s susceptibility to infectious and autoimmune diseases [7, 8], and there is strong selection for any pathogen that can escape MHC presentation. In fact, MHC genes are prime targets of pathogen adaptation, as MHC haplotypes drive virulence or escape mutations in a number of viral infections, such as HIV [9], HTLV-1 [10], hepatitis C virus [11], and Friend virus [12]. Additionally, genetic polymorphisms that occur within the MHC are major host factors that influence pathogen adaptation and virulence evolution [13].

While much research has been done on the pathogenesis of C. neoformans, adaptations that occur during infection in a host and how they might relate to virulence are not fully understood. In a prior study published in 2004, congenic mice with differences only in the MHC locus H2 were infected with the lab-adapted strain H99W to investigate MHC-based differences in C. neoformans pathogenicity [14]. Two independent serial passages using liver homogenates were completed for each of five haplotypes: b, d, f, k, and q. This serial passage was repeated for a total of eight times, resulting in ten final strains (two biological replicates per MHC haplotype) [14] (Fig. 1). In that study, mortality rate significantly differed between MHC congenic mouse strains, but there was no evidence supporting MHC-dependent differences in infectious load or mortality when strains from mice of all haplotypes were compared [14]. Given that the initial strain, H99W, is a lab-adapted clinical strain with reduced virulence compared to its parent strain, it was posited that most virulence adaptations were probably adaptations to the host rather than to the specific MHC haplotypes of the mice [14]. Since this first study was completed, advances in next-generation sequencing technology have occurred, allowing whole genomes and transcriptomes to be analyzed. These advances have made it possible to now investigate the molecular adaptation strategies of these evolved strains.

Fig. 1.

Passage of C. neoformans in mice with different major histocompatibility complex (MHC) haplotypes (previously published [14]). This figure depicts the serial passage of each strain through mice of different MHC haplotypes performed in the original study. Two initial infections using C. neoformans strain H99W were performed for each haplotype. When an infected mouse became ill, it was euthanized using CO₂ overdose and the liver homogenized and injected into the next mouse. This method of passage was repeated a total of eight times for each strain, resulting in ten final strains, representing two biological replicates for each MHC haplotype [14]. C. neoformans strains isolated from b, k, and q mice were then analyzed for genomic and transcriptome differences. All animal use complied with federal regulations and the University of Utah’s Institutional Animal Care and Use Committee guidelines (IACUC protocol #02-03005)

Transcript level differences in RNA sequencing experiments include: (1) transient RNA level differences due to the net combination of increased/decreased transcription [15] and mRNA decay [16, 17], (2) heritable RNA level differences due to epigenetic changes (also called epimutations) that result from persistent, self-perpetuating, trans-generational chromatin modifications, such as DNA methylation, histone differences and post-translational modification, and production of noncoding RNAs or induction of stable, positive feedback loops [18–21], and (3) heritable RNA level differences due to mutations to gene regulatory regions or genes involved in epigenetic mechanisms [22]. In typical RNA sequencing experiments where cells exposed to different environments are immediately compared, the primary findings will be in the transient RNA level changes (such as [16], as well as some epigenetic-based differences, although this is not typically acknowledged.

When C. neoformans infects a mouse host, fungal adaptation and response could be due to any combination of transient gene regulation differences (gene regulation), heritable gene expression differences (epigenetics), and/or mutational differences. Once the fungi are removed from the mouse adaptation environment, cells will lose the environment-specific transient response (gene regulation), but retain the heritable changes (epigenetic and mutation-based). In this experimental design, using post-adapted fungi that are far removed from the mouse adaptive environment, only heritable differences, (epigenetic and mutation-based changes to regulatory regions) could be responsible for differences in transcript levels when compared to the parent strain [18, 22].

To test the hypothesis that specific heritable adaptations accounting for the differences in observed virulence occurred in these strains during passaging, we investigated epigenetically regulated gene expression (ERG) and genomic changes in these post-adapted strains using next generation sequencing. These results identified transcript level differences compared to the parent strain that were attributable to the specific MHC adaptation environment or were adaptations to mouse (MHC-independent). None of these heritable changes could be mapped to genomic changes that would directly influence gene expression level, so these transcript level differences must be due to epigenetic alterations. Furthermore, three genes that function in a chromatin-based regulatory complex were identified in one adapted strain as having heritable differences in transcript levels. Finally, genomic changes were minimal, did not include genes known to be involved in epigenetic mechanisms, and were mostly strain-specific with only one mutant gene common across strains adapted to each MHC haplotype. This study advances our understanding of the persistent molecular adaptations of C. neoformans to the host immune environment, providing new molecular targets that may advance current treatments.

Materials and methods

Strain selection

DNA and RNA sequences for six strains of C. neoformans were analyzed and compared to determine if different adaptation strategies were used in response to hosts with different MHC haplotypes. C. neoformans strain H99W (the non-passaged control strain) was previously passaged in B10 congenic mice with differences only in the MHC locus H2. Each mouse had the MHC haplotypes b, d, f, k, or q, and two independent serial passages were performed for each haplotype (Fig. 1) [14]. Of the ten resultant strains from the original experiment, only those passaged through the b, k, and q haplotypes were used for current analyses. The strains passaged through the d haplotype mice were excluded because one passage was performed three times rather than the eight times performed for the other strains. The strains passaged through the f haplotype mice were excluded because RNA sequencing data was unavailable for one of the two passaged strains. After isolation from mouse hosts, passaged C. neoformans strains were frozen at-80 °C. Strains were grown from frozen culture twice; once to streak onto yeast peptone dextrose plates to send to UCSF and once to grow strains up for DNA or RNA isolation.

Genomic DNA sequencing

For genomic DNA libraries, cultures were inoculated from frozen stocks of C. neoformans containing 4–5 distinct colonies grown from a single mouse liver homogenate that may or may not have evolved independently in the mouse [14]. Each strain (two biological replicates per MHC haplotype) was grown in 50 mL of yeast extract, peptone, adenine, dextrose at 30 °C for 1–2 days and DNA was isolated using the standard CTAB protocol for DNA isolation described in [23]. Next, the DNA was sheared using an M220 Covaris sonicator with microTUBE AFA (SKU: 520045) and the following parameters: duty factor: 20%, peak/displayed power 50 W, cycles/burst: 200, duration: 65 s, and temperature: 20 °C. The NEBNext^® Ultra™ II DNA Library Prep with Sample Purification Beads kit (cat. no. E7103S) was used to generate Illumina libraries, using the maximum recommended input (1 µg) to minimize PCR cycles. Libraries were sequenced on an Illumina HiSeq 4000 with paired-end reads of 100 bp for each strain, including the parent H99W control strain. Each library consisted of roughly 10–13 million sequences, with output files in fastq format.

RNA sequencing

For RNA libraries, single colonies of each mouse-passaged or the H99W parent control strain were grown from frozen stocks in triplicate in yeast nitrogen base at 30 °C for 13–15 h to OD₆₀₀ = 1.0 to obtain enough cells for RNA extraction, as the passaged strains grow slower than the wild-type H99W parent strain in this medium. Total RNA was isolated following the standard TRIzol protocol described in [23]. The NEBNext^® Poly(A) mRNA Magnetic Isolation Module (cat. no. E7490S) was used to enrich mRNA from 1 µg of total RNA. Libraries were generated using the NEBNext^® Ultra™ II Directional RNA Library Prep Kit for Illumina^® (cat. no. E7760S). Libraries were sequenced on an Illumina NovaSeq SP200 with paired-end reads of 100 bp for each strain. Each library consisted of roughly 13–16 million sequences, with output files in fastq format.

RNA sequencing pipeline

The RNA sequencing pipeline was used to identify changes in transcript levels and to predict the functional effects on cell physiology (Fig. S1). The quality of the RNA sequencing files was checked using FastQC ver. 0.11.5 [24]. STAR ver. 2.5.3a [25] was then used to align reads to the genome and produce gene expression counts that were then used to check for outliers in the replicates using the edgeR package (version 4.6.1) [26] and an R script based on the tutorial by [27] and uploaded to GitHub at https://github.com/rebeccaseipelt/Cn_MHC_Evolution. CuffDiff2 ver. 2.2.1 [28] was used to identify genes in the passaged strains with statistically different (q ≤ 0.05) and biologically relevant (fold difference ≥ 2) transcript level differences as compared to the parent strain. The lists of genes whose transcript levels were different (epigenetic regulated genes; ERGs) for all passaged strains and pairs of strains were then compared using a Venn diagram generator to identify common and unique ERGs. The database for annotation, visualization and integrated discovery (DAVID) [29] was used to evaluate which biological pathways were preferentially affected in each strain, strain pair, and overall. DAVID predicted one gene, CNAG_04090, to be an ortholog of Candida albicans Sko1, which C. neoformans is believed to lack. Therefore, protein domain analyses for translated CNAG_04090 and Candida albicans Sko1 (XP_019330633.1) were performed using the conserved domain database search [30].

Genome variant pipeline

The genome variant pipeline was used to identify mutations that occurred in the passaged strains and to determine what functional effect these mutations may have caused (Fig. S2). The quality of the DNA sequencing files was ensured using FastQC ver. 0.11.5 [24]. The reads were of high quality with average quality scores of 30/40 or higher and none needed to be trimmed. Bowtie2 ver. 2.2.4 [31] was used to align the reads to the reference genome (GCA_000149245.3 at Ensembl Fungi [32]) and output a .sam file. Samtools ver. 1.7 [33] was then used to convert the files from .sam to .bam, sort, index, and finally convert the files into .bcf files for variant calling. BCFtools ver. 1.10 [34] was used to call and locate the variants and variant positions. SnpEff ver. 4.3.1 [35] was used to map variant locations to genome features, such as genes, exons, coding, or regulatory changes, as well as the level of functional effect to the encoded protein. High and moderate effects were predicted to change the encoded protein, with high effects altering protein size (nonsense mutations, deletions, or insertions), moderate effects were usually missense mutations, and low effects were synonymous/silent mutations that did not affect the protein coding region. Since fungi have little space between genes, SNPEff often identified multiple genes as potentially affected for each variant. Custom Python code, available on GitHub at https://github.com/rebeccaseipelt/Cn_MHC_Evolution, was used to remove low quality variant calls, exclude variants found in the parent strain, and compile a list of all variant effects and the individual genes affected. Since strains were passaged independently it was expected that genes, but not specific variant locations, would be common, so all variant genes rather than specific locations were the focus. The resultant gene lists were then compared using a Venn diagram generator (https://bioinformatics.psb.ugent.be/webtools/Venn/) for all passaged strains to identify common and unique genes per strain and per haplotype, as well as to identify common mutational strategies. Since so few genes were identified, their functions were investigated using Fungi Ensembl [32]. To identify larger genomic changes, sequencing reads were filtered for “split” reads or truncated reads and then visualized using a genome browser to manually identify and confirm specific, larger structural changes.

Statistics

For the RNA sequencing pipeline, CuffDiff default statistical calculations for identifying statistical significance were utilized, including the q-value, which is a corrected p-value for multiple comparisons [28]. For the genome variant pipeline, variant calls with a quality score cutoff of 80 or greater, which represents a 1 in 10⁸ chance of error, were used [34]. Pathway enrichment analysis was performed using DAVID [29]. DAVID utilizes a modified one-tailed Fisher Exact p-value, referred to as the EASE score. The EASE score is calculated using one less gene than the total count and is thus more conservative than a traditional Fisher Exact p-value. Statistical significance was determined using DAVID’s default EASE threshold of 0.1 [29].

Results

Molecular mechanisms for fungal adaptation to mice with specific MHC haplotypes could include transient gene expression changes, heritable changes in transcript levels (epigenetic), and/or genomic changes (mutations). To more fully characterize persistent, heritable molecular adaptation strategies, both heritable transcript level differences and DNA changes of strains that had been previously passaged through B10 congenic mice with different MHC haplotypes were examined using next generation sequencing technologies. Two bioinformatics pipelines were utilized: the RNA sequencing pipeline (Fig. S1) compared transcript levels of strains previously passaged through different MHC-congenic hosts, and the genome variant pipeline (Fig. S2) compared DNA-based adaptations of the same strains.

First, the transcriptomes of previously adapted C. neoformans strains were investigated. Since the pathogen’s gene expression was measured long after mouse passage, changes to transcript levels must be due to either regulatory genomic or epigenetic changes. The RNA sequencing pipeline (Fig. S1) was used to compare heritable transcript level differences between individual strains and strain-pairs. To confirm that each strain pair had similar overall gene expression and to identify outliers, gene expression was clustered and visualized as a principal component analysis plot. Having confirmed appropriate clustering of replicates, transcript levels for each gene were compared (evolved strain to the parent strain, H99W) to identify genes with statistically and biologically significant differences in transcript levels. First, genes having a fold difference less than 2 or a significance value (q-value) less than 0.05 were removed from each list. These gene lists were then compared between the strain-pairs and across all strains.

Forty-two epigenetic regulated genes (ERGs) common to all passaged strains were identified, 40 of which had higher transcript levels in all adapted strains compared to the parent, suggesting that increased heritable differences in transcript levels were important in the adaptation of the common C. neoformans lab strain to a mouse host (Fig. 2A; Table 1). Of the 42 ERGs identified, 19 genes had annotated functions, whereas the remaining 23 encoded hypothetical proteins. Of those with annotated functions, many have predicted transporter functions (CNAG_05914, CNAG_00539, CNAG_04617, and CNAG_00796), while others have predicted metabolic functions. Two genes are predicted to play a role in gluconeogenesis, CNAG_04217 and CNAG_06374, which encode phosphoenolpyruvate carboxykinase (EC 4.1.1.49) and malate dehydrogenase (EC 1.1.1.40), respectively. These genes are involved in the carbon fixation pathway, which was preferentially altered (EASE score 0.0487, Fisher Exact p-value 0.00137) (Fig. 3A). Another predicted metabolism gene CNAG_06623, which encodes inositol oxygenase (EC 1.13.99.1), plays a role in ascorbate and aldarate metabolism. Additionally, another group of genes are predicted to be related to protein processing in the endoplasmic reticulum (ER), CNAG_01750, CNAG_06150, CNAG_04200, and CNAG_04449, which encode hsp72-like protein, hsp90-like protein, a C2 domain protein, and an alpha-1,6-mannosyltransferase, respectively. CNAG_01750 and CNAG_06150 encode heat shock proteins and are involved in protein processing [36, 37]. CNAG_04200 encodes a protein that contains a C2 domain that is predicted to tether the ER to the plasma membrane to facilitate lipid transfer [38]. The last, CNAG_04449, encodes an alpha-1,6-mannosyltransferase and is predicted to be involved with the extension of oligosaccharides in the Golgi apparatus [39, 40]. While the ER pathway itself was not identified (EASE score 0.177, Fisher Exact p-value 0.0196), these genes associated with the ER suggest that alterations to this pathway may also be common to all strain-pairs (Fig. 3B). The higher net transcript levels of these genes in all of the mouse-passaged strains suggests that after adaptation in mice, C. neoformans had adapted to nutrient deprivation by heritably upregulating gluconeogenesis, activated the unfolded protein response to mitigate the increased temperature of the host [41], activated ascorbate and aldarate metabolism to defend against oxidative stress [42], and remodeled the cell wall to mask itself from the immune response [40]. Thus, heritable changes in transcript levels for genes involved in multiple pathways had occurred in the passaged strains, suggesting a diverse set of adaptations to the host environment.

Fig. 2.

Common, shared, and unique ERGs among strain-pairs (A), strain-partners (B-D), and metabolism ERGs (E). The quantity of shared genes between noted strain-pairs (A, B, F) or strain partners (C, D) are displayed as numbers within each Venn diagram. A Epigenetic regulated genes (ERGs) were identified using CuffDiff [28] and the overlap between strain partners was identified. These common ERGs were then compared by host major histocompatibility complex (MHC) type (B, K, Q) with 42 common ERGs identified among all strains. B ERGs in each independently evolved strain from host MHC type B were identified using CuffDiff and the overlap identified (204 genes). C ERGs in each independently evolved strain from host MHC type K were identified using CuffDiff and the overlap identified (110 genes). D ERGs in each independently evolved strain from host MHC type Q were identified using CuffDiff and the overlap identified (140 genes). E Metabolism ERGs were identified using CuffDiff. Genes common to both strain partners were annotated to specific pathways using the database for annotation, visualization and integrated discovery (DAVID) [29]. Genes annotated to metabolism were compared across strain-pairs with three common metabolism ERGs identified among all strains. This figure was generated using a Venn diagram generator at bioinformatics.psb.ugent.be/webtools/Venn/

Table 1.

The 42 ERGs common to mouse-adapted strains

Gene	Function	Level	Domain Functions (hypothetical proteins only)	Average fold difference
CNAG_02701	Hypothetical protein	Higher	BAG domain	146.5
CNAG_02691	Hypothetical protein	Higher	BAG domain	33.19
CNAG_07493	Hypothetical protein	Higher	Polar residues, disordered region	28.04
CNAG_04027	Hypothetical protein	Higher	Polar residues, disordered region	13.48
CNAG_04217	Phosphoenolpyruvate carboxykinase (ATP)	Higher		9.752
CNAG_04103	Hypothetical protein	Higher	Transmembrane ion channel	8.992
CNAG_05914	MFS transporter, SP family, general alpha glucoside: H symporter	Higher		8.93
CNAG_04139	Hypothetical protein	Higher	Polar residues, disordered region	8.86
CNAG_01750	Hsp72-like protein	Higher		8.27
CNAG_05874	Hypothetical protein	Higher	Polar residues, disordered region	7.022
CNAG_04185	Hypothetical protein	Higher	Polar residues, disordered region	6.922
CNAG_00539	Membrane transporter	Higher		6.787
CNAG_00306	Hypothetical protein	Higher	Metal ion binding	6.5
CNAG_05842	Cytochrome P450	Higher		5.63
CNAG_06374	Malate dehydrogenase (oxaloacetate-decarboxylating)(NADP)	Higher		5.315
CNAG_02990	Nuclear protein	Higher		5.21
CNAG_05873	Hypothetical protein	Higher	Polar residues, disordered region	4.878
CNAG_03333	Cytoplasmic protein	Higher		4.782
CNAG_04409	Hypothetical protein	Higher	Polar residues, disordered region	4.683
CNAG_06623	Inositol oxygenase	Higher		4.365
CNAG_03347	ATP-dependent Clp protease ATP-binding subunit ClpB	Higher		4.075
CNAG_05654	Hypothetical protein	Higher	Transmembrane (pH response)	3.927
CNAG_04617	OPT family small oligopeptide transporter	Higher		3.852
CNAG_03231	Hypothetical protein	Lower	Signal	3.678
CNAG_04630	Hypothetical protein	Higher	bZIP domains	3.632
CNAG_02553	Short-chain dehydrogenase	Higher		3.503
CNAG_06150	Hsp90-like protein	Higher		3.49
CNAG_03059	Hypothetical protein	Higher	Metal binding (zinc ion binding/ zinc fingers)	3.48
CNAG_04756	Hypothetical protein	Higher	Rnase	3.465
CNAG_02877	Hypothetical protein	Higher	Zinc-fingers	3.437
CNAG_00474	Hypothetical protein	Higher	Polar residues, disordered region	3.287
CNAG_00337	Hypothetical protein	Higher	Polar residues, disordered region	2.98
CNAG_04200	C2 domain-containing protein	Higher		2.88
CNAG_06845	STE/STE11/CDC15 protein kinase	Lower		2.842
CNAG_00796	ATP-binding cassette, subfamily B (MDR/TAP), member 1	Higher		2.71
CNAG_05872	Endopeptidase	Higher		2.682
CNAG_00156	Hypothetical protein	Higher	Zinc-Fingers	2.635
CNAG_02674	Hypothetical protein	Higher	Peptidase/ hydrolase/thioredoxin	2.635
CNAG_06244	Hypothetical protein	Higher	Polar residues, disordered region	2.635
CNAG_02314	Hypothetical protein	Higher	Transmembrane	2.472
CNAG_04202	Cytosolic Fe-S cluster assembly factor NAR1	Higher		2.463
CNAG_04449	Alpha-1,6-mannosyltransferase	Lower		2.39

Fig. 3.

Functional annotation of genes with transcript level differences that are common to all strain-pairs. A Carbon fixation. The green boxes represent annotated proteins in the carbon fixation pathway for C. neoformans, while the white boxes represent unidentified proteins. Each protein is identified by its enzyme commission identifier, EC. Enzymes whose functions are predicted to be affected are noted by red outlines. B Protein processing in the endoplasmic reticulum. The green boxes represent annotated proteins in the protein processing pathway for C. neoformans, while the white boxes represent unidentified proteins. Proteins whose functions are predicted to be affected are noted by red outlines. Both panels of this figure were generated using DAVID [29] and adapted from the Kyoto Encyclopedia of Genes and Genomes (KEGG) [43–45]

In this analysis, one gene, CNAG_02701, had high net transcript levels in all passaged strains when compared to the parent strain (Table S1) and encodes a hypothetical protein containing Bcl-2-associated athanogen (BAG) domains. Proteins containing these domains regulate chaperones [46]. A recent study on a novel antifungal treatment on C. neoformans has provided further evidence of this gene’s potential role, as it, as well as several heat shock proteins, are upregulated in response to a fungicidal treatment affecting the integrity of the cell wall and membrane [47].

Next, ERGs specific to the mouse host’s MHC haplotype were investigated. Strains passaged in b haplotype mice (B1 and B2) shared a total of 204 ERGs (Fig. 2B). The B strain-pair had unique transcript level differences in three genes that interact with the Polycomb Repressive Complex (PRC) (EASE score 0.0253, Fisher Exact p-value 0.0019): CNAG_01013 (higher level), CNAG_01098 (lower level), and CNAG_05696 (higher level). These encode COMPASS component SWD3 (WDR5), pre-rRNA-processing protein IPI1, and ubiquitin-conjugating enzyme E2 D (UBE2D), respectively. WDR5 is a member of the COMPASS complex, which regulates H3K4 methylation [48]. IPI1 (Interacting Protein of Ipi3) is part of a complex (often involving Rix1 and Ipi3) involved in the processing of precursor ribosomal RNA and is involved in 60 S ribosome maturation [49]. Ubiquitin-conjugating enzyme E2 binds activated ubiquitin acquired from ubiquitin-conjugating enzyme E1 [50]. These MHC-specific differences in ERGs suggest that in b haplotype mice, C. neoformans is using WDR5-mediated methylation mechanisms to enable the expression of specific virulence factors [51] while simultaneously using IPI1 reduction to slow translation and growth and E2-mediated ubiquitination to suppress non-essential processes and rapidly recycle proteins for survival in the host [52]. Strain B1 had a further 85 ERGs identified (Fig. 2B and Table S1) and strain B2 had a further 216 ERGs identified (Fig. 2B and Table S1).

For strains passaged in k haplotype mice (K1 and K2), a total of 110 ERGs were shared (Fig. 2C). The K strain-pair had unique transcript level differences for the mitogen-activated protein kinase (MAPK) signaling pathway (EASE score 0.0823, Fisher Exact p-value 0.0133) (Fig. 4, red stars). There were three ERGs present in the pathway: CNAG_04090 (increased level), CNAG_05968 (decreased level), and CNAG_05155 (increased level). These encode hypothetical proteins predicted to be an ATF family transcription factor (Sko1), a cell division control protein 42 (Cdc42), and a protein-tyrosine-phosphatase (Ptp 2,3), respectively. Cdc42 is a Rho family GTPase [53]. Ptp2,3 catalyzes a reaction involving a protein with a phosphotyrosine group and water, resulting in removal of the phosphate group [54]. These MHC-specific differences in ERGs suggest that in k haplotype mice, C. neoformans has adapted to osmotic stress with higher Sko1 transcript levels [55], has suppressed growth with lower Cdc42 transcript levels, possibly to conserve energy or evade immune detection [56], and has prevented hyper-activation of the HOG pathway, which can be lethal if left unregulated [57], with higher Ptp2.3 transcript levels. In addition to the common ERGs, strain K1 had a further 99 DEGs identified, whereas strain K2 had 367 unique ERGs identified (Fig. 2C and Table S1).

Fig. 4.

Functional annotation of genes with transcript level differences common to the K strain-pair in the mitogen-activated protein kinase (MAPK) signaling pathway. The green boxes represent annotated proteins in the MAPK signaling pathway for C. neoformans, while the white boxes represent unidentified proteins. Each protein is identified by its enzyme commission identifier, EC. Enzymes and proteins whose functions are predicted to be affected are noted by red stars. This figure was generated using the database for annotation, visualization and integrated discovery (DAVID) [29] and adapted from the Kyoto Encyclopedia of Genes and Genomes (KEGG) [43–45]

Finally, strains passaged in q haplotype mice (Q1 and Q2) shared a total of 140 ERGs (Fig. 2D). The Q strain-pair exhibited unique transcript level differences compared to the parent strain to three different pathways: ascorbate and aldarate metabolism (EASE score 0.0141, Fisher Exact p-value 0.000786) (Fig. 5A, red stars), cysteine and methionine metabolism (EASE score 0.0867, Fisher Exact p-value 0.0138) (Fig. 5B, red stars), and pyruvate metabolism (EASE score 0.0981, Fisher Exact p-value 0.0168) (Fig. 6, red stars).

Fig. 5.

Functional annotation of genes with transcript level differences that are unique and common to the Q strain-pair. A Ascorbate and aldarate metabolism. B Cysteine and methionine metabolism. The green boxes represent annotated proteins for C. neoformans, while the white boxes represent unidentified proteins. Each protein is identified by its enzyme commission identifier, EC. Enzymes whose functions are predicted to be affected are noted by red stars. Both panels of this figure were generated using the database for annotation, visualization and integrated discovery (DAVID) [29] and adapted from the Kyoto Encyclopedia of Genes and Genomes (KEGG) [43–45]

Fig. 6.

Functional annotation of genes with transcript level differences common to the Q strain-pair in pyruvate metabolism. The green boxes represent annotated proteins in the pyruvate metabolism pathway for C. neoformans, while the white boxes represent unidentified proteins. Each protein is identified by its enzyme commission identifier, EC. Enzymes whose functions are predicted to be affected are noted by red stars. Reactions catalyzed by these enzymes are represented by red lines. This figure was generated using the database for annotation, visualization and integrated discovery (DAVID) [29] and adapted from the Kyoto Encyclopedia of Genes and Genomes (KEGG) [43–45]

In the ascorbate and aldarate metabolism pathway (Fig. 5A), there were three ERGs present: CNAG_01078 (higher level), CNAG_04753 (higher level), and CNAG_06623 (higher level). These encode aldehyde dehydrogenase (EC 1.2.1.3), gluconolactonase (EC 3.1.1.17), and inositol oxygenase (EC 1.13.99.1), respectively. Aldehyde dehydrogenase catalyzes the reaction of an aldehyde, NAD+, and water to produce a carboxylate, NADH, and hydrogen [58]. Gluconolactonase catalyzes a reaction between D-glucono-1,5-lactone and water to produce D-gluconate [59]. Inositol oxygenase catalyzes a reaction between myo-inositol and oxygen to produce D-glucuronate and water [60]. The higher transcript levels of these MHC-specific ERGs after adaptation in q haplotype mice suggest that C. neoformans has adapted to scavenging inositol [61] to synthesize capsule or generate energy (via the pentose phosphate pathway) to produce reducing power (NADPH) and detoxifying enzymes (ALDH [62]), enabling better survival in the oxidative environment in the mouse host.

In the cysteine and methionine metabolism pathway, there were three ERGs present in the pathway: CNAG_00637 (higher level), CNAG_06300 (higher level), and CNAG_01252 (higher level). These encode cystathionine beta-synthase (EC 4.2.1.22), a glutamate-cysteine ligase catalytic subunit (EC 6.3.2.2), and thiosulfate/3-mercaptopyruvate sulfurtransferase (EC 2.8.1.2), respectively (Fig. 5B). Cystathionine beta-synthase catalyzes the reaction of L-serine and L-homocysteine to produce L-cystathionine and water [63]. Glutamate-cysteine ligase catalyzes the reaction between ATP, L-glutamate, and L-cysteine to produce ADP, phosphate, and gamma-L-glutamyl-L-cysteine [64]. Thiosulfate/3-mercaptopyruvate sulfurtransferase catalyzes the reaction of 2-oxo-3-sulfanylpropanoate and a reduced thioredoxin to produce pyruvate, hydrogen sulfide, and an oxidized thioredoxin [65]. The higher transcript levels of these MHC-specific ERGs in C. neoformans post-adaptation in q haplotype mice suggest that the fungus has adapted to produce more glutathione, which enables defense against oxidative stress [66], and more thiosulfate/3-mercaptopyruvate sulfurtransferase to counter copper toxicity in the host [67].

In the pyruvate metabolism pathway, there were three ERGs present in the pathway (Fig. 6): CNAG_01078 (higher level), CNAG_06374 (higher level), and CNAG_04217 (higher level). These encode aldehyde dehydrogenase (EC 1.2.1.3), malate dehydrogenase (EC 1.1.1.40), and phosphoenolpyruvate carboxykinase (EC 4.1.1.49), respectively. Aldehyde dehydrogenase catalyzes the reaction of an aldehyde, NAD+, and water to produce a carboxylate, NADH, and hydrogen [58]. Malate dehydrogenase catalyzes a reaction between (S)-malate, NADP+, and oxaloacetate to produce pyruvate, carbon dioxide, and NADPH [68]. Phosphoenolpyruvate carboxykinase catalyzes a reaction between ATP and oxaloacetate to produce ADP, phosphoenolpyruvate, and carbon dioxide [69]. CNAG_06374 and CNAG_04217 also had higher transcript levels in both the B and K strains as well, and were involved in a variety of pathways, including gluconeogenesis and carbon fixation, as mentioned above. The higher transcript levels for these genes suggest that after mouse adaptation, C. neoformans has adapted to increase gluconeogenesis [70] and reducing power and has switched to using alternate carbon sources [71], bypassing normal glycolysis. In addition to the common ERGs, strain Q1 had another 347 ERGs identified (Fig. 2D and Table S1) and strain Q2 had 27 additional ERGs identified (Fig. 2D and Table S1).

In examining pathways common to strain pairs, strain-pairs passaged through mice with b and q haplotypes shared nine ERGs related to metabolism (B: EASE score 0.088, Fisher Exact p-value 0.062; Q: EASE score 0.0043, Fisher Exact p-value 0.00243)(Fig. 2E; Table 2). Of these nine genes, three also had differences in transcript levels in the k haplotype strain-pair. Interestingly, all nine genes had higher transcript levels across all strains in which their level was different. The coordinate adaptation of these nine enzymes suggests four major survival strategies of C. neoformans in MHC congenic mice: more gluconeogenesis to make its own sugar [72], utilization of inositol to synthesize capsule [61], scavenging of nitrogen, and optimization of the interconversion of glucose-6-phosphate anomers to drive energy production [73].

Table 2.

Nine metabolism-related epigenetic regulated genes common to B and Q strain-pairs

Gene ID1	Function2	Pathway3	Level4	Also in K?5
CNAG_06374	Malate dehydrogenase	Pyruvate metabolism	Higher	Yes
CNAG_01078	Aldehyde dehydrogenase	Glycolysis	Higher
CNAG_04307	Urate oxidase	Purine metabolism	Higher
CNAG_05329	Myo-inositol 2-dehydrogenase	Inositol phosphate metabolism	Higher
CNAG_02365	Amidase	Arginine and proline metabolism	Higher
CNAG_04217	Phosphoenolpyruvate carboxykinase	Glycolysis	Higher	Yes
CNAG_06666	Starch phosphorylase	Starch and sucrose metabolism	Higher
CNAG_07559	Glucose-6 phosphate 1-epimerase	Glycolysis	Higher
CNAG_06623	Inositol oxygenase	Ascorbate and aldarate metabolism	Higher	Yes

¹ Official gene symbol

² Gene function

³ Gene pathway

⁴ Fold difference direction

⁵ Found in K strain pair

Of these ERGs, three were predicted components of the glycolysis pathway (Fig. 7): CNAG_01078 (higher level), CNAG_07559 (higher level), and CNAG_04217 (higher level). These encode aldehyde dehydrogenase (EC 1.2.1.3), glucose-6-phosphate 1-epimerase (EC 5.1.3.15), and phosphoenolpyruvate carboxykinase (EC 4.1.1.49), respectively. Aldehyde dehydrogenase catalyzes the reaction of an aldehyde, NAD+, and water to produce a carboxylate, NADH, and hydrogen [58]. Glucose-6-phosphate 1-epimerase converts alpha-D-glucose 6-phosphate into beta-D-glucose 6-phosphate [74]. Phosphoenolpyruvate carboxykinase catalyzes the reaction between ATP and oxaloacetate to produce ADP, phosphoenolpyruvate, and carbon dioxide [69]. The epigenetic adaptation of these glycolysis genes with higher transcript levels reinforces the fact that energy production is required by C. neoformans to successfully infect a host.

Fig. 7.

Functional annotation of genes with different transcript levels common to the B and Q strain-pairs in glycolysis. The green boxes represent annotated proteins in the glycolysis pathway for C. neoformans, while the white boxes represent unidentified proteins. Each protein is identified by its enzyme commission identifier, EC. Enzymes whose functions are predicted to be affected are noted by red stars. Reactions catalyzed by these enzymes are represented by red lines. This figure was generated using the database for annotation, visualization and integrated discovery (DAVID) [29] and adapted from the Kyoto Encyclopedia of Genes and Genomes (KEGG) [43–45]

To determine whether these heritable gene expression differences were due to DNA-based changes in regulatory regions and to identify any DNA-based adaptation strategies, genomic variants in each adapted strain were investigated. After variants shared with the parent strain and variant genes with no or low evidence were removed, genomic changes across all strains were investigated. Very few genomic variants not already found in the parent strain were identified, with some potential heterogeneity, particularly when coverage was low (see Table 3 asterisks). These potential variants were included so as not to eliminate any potential variation. While there were no whole chromosome duplications, there were rearrangements, including partial duplications and large deletions, particularly among sub-telomere regions, although none were consistent for both members of a strain pair. Of the 47 single nucleotide polymorphisms identified, each variant was located within its gene or regulatory region and its effect on encoded protein was predicted using SNPEff. Several genes were often identified for a particular variant location by SNPEff due to the tightly packed C. neoformans genome where transcription units may overlap, including coding and regulatory regions. When we compared the variant locations to the epigenetic regulated gene lists (ERGs), we found no evidence of changes in the regulatory or coding region of any ERG in any strain. These results indicated that epigenetic mechanisms, rather than regulatory mutations, were responsible for the differences in transcript levels observed for the strains.

Table 3.

Overlap of confirmed variant genes from B, K, and Q strains

Strains	Genes with Variants Predicted to Have High, Moderate, and Splicing SnpEff Effects	Genes with Variants Predicted to Have Low and Modifier SnpEff Effects
B1 K2 Q1 Q2	CNAG_03121	n/a
B1 Q2	n/a	CNAG_03123
K1 Q2	CNAG_05838 (* for K1 and confirmed for Q2)	CNAG_05840* CNAG_12685* CNAG_05837*
K2 Q1	n/a	CNAG_03120
B1	CNAG_00604* CNAG_03723	CNAG_03721 CNAG_00605* CNAG_00601* CNAG_00602* CNAG_00603* CNAG_03724
B2	CNAG_02745	CNAG_02747 CNAG_05140* CNAG_12370* CNAG_05138*
K1	CNAG_00545*	CNAG_12179* CNAG_12180* CNAG_03896* CNAG_12183* CNAG_00544* CNAG_03897*
K2	CNAG_02594	CNAG_13098* CNAG_06266* CNAG_07922* CNAG_13097* CNAG_07919* CNAG_12303 CNAG_12304
Q1	CNAG_04433	CNAG_04434 CNAG_00513 CNAG_00509 CNAG_00512
Q2	CNAG_00292	CNAG_01485 CNAG_01486 CNAG_05842 CNAG_01487 CNAG_12042 CNAG_00294

* Variant call in this gene may be due to strain population difference

To further study any genomic-based adaptation strategies in these data, we next focused broadly on variant genes because strain evolution by independent passage would most likely involve common gene-level changes rather than common exact base mutations. In addition, variant locations were often annotated to several genes due to the small intergenic space and overlapping transcription units in this species. So as not to exclude any potentially important genes, all variant genes were investigated and categorized by likely effect. When variant genes originating from different strains were compared, all but two variant genes were unique to a single strain (Table 3). Of these unique variant genes, functions were widely varied, not redundant or in similar biological pathways. Of the two variant genes that were not unique, one variant gene (CNAG_05838) was shared by a K-adapted strain (K1) and a Q-adapted strain (Q2). Only one gene was common to a strain pair (adapted in the q haplotype mice). However, this one variant gene was shared among four of the strains (CNAG_03121), representing at least one strain from each haplotype (Table 3). This gene would be the most likely to represent adaptation to a mouse host, since it was shared across all haplotypes, if not across all strains. This gene was affected by missense mutations in strains B1, Q1, and Q2, and predicted to have a moderate level of effect. In strain K2, it contained a frameshift mutation, which was predicted to have a high level of effect (Table 4).

Table 4.

Haplotype involvement of confirmed significant effect variant genes from B, K, and Q strains

	Strains	Gene	Function 1	B1 Mutation 2	B2 Mutation 2	K1 Mutation 2	K2 Mutation 2	Q1 Mutation 2	Q2 Mutation 2
3	B1; K2; Q1; Q2	CNAG_03121	Hypothetical protein; pfam domain: Mediator complex subunit Med13, C-terminal	Missense variant; Moderate effect; chr8:118736	na	na	Frameshift variant; High effect; chr8:117808	Missense variant; Moderate effect; chr8:115989	Missense variant; Moderate effect; chr8:120014
2	K1 & Q2	CNAG_05838	GTPase activating protein	na	na	Splice donor variant; High effect; chr7:1009627 *	na	na	Stop gain variant; High effect; chr7:1011535
1	Q1	CNAG_04433	CMGC/DYRK/DYRK2 protein kinase	na	na	na	na	Frameshift variant; High effect; Chr9:858090	na
	Q2	CNAG_00292	Hypothetical protein: Pfam domain: Zinc finger, ZPR1-type	na	na	na	na	na	Missense variant; Moderate effect; Chr1:753753
	K1	CNAG_00545	Cohesin loading factor subunit SCC2	na	na	Disruptive inframe insertion; Moderate effect; Chr1:390308 *	na	na	na
	B1	CNAG_00604	Hypothetical protein; Pfam domain: Zinc finger, RING-type, eukaryotic	Missense variant; Moderate effect; Chr1:1551946 *	na	na	na	na	na
	K2	CNAG_02594	Hypothetical protein; no Pfam domains detected	na	na	na	Missense variant; Moderate effect; Chr3:1272619	na	na
	B1	CNAG_03723	Nuclear protein	Missense variant; Moderate effect; Chr2:626711	na	na	na	na	na
	B2	CNAG_02745	Hypothetical protein; Pfam domain: Ribonuclease III domain	na	Missense variant; Moderate effect; Chr3:856395	na	na	na	na

¹ Gene function or predicted Pfam domain as noted in Fungi Ensembl

² Variant type and variant effect predicted by SNPeff; Variant call location (chromosome: base); High effect variants are noted with bold font. Moderate effect variants are noted with italics font.

* Variant call at this location may be due to strain population difference

CNAG_03121 is a hypothetical gene predicted to encode mediator complex subunit Med13. Despite it being listed as a hypothetical gene, one study identifies it as an ortholog to the med13 gene of S. cerevisiae and shows that it affects gene expression [75]. However, this complex and gene have not been studied in detail in C. neoformans, though they have been shown to play a role in virulence in C. albicans [76].

With the role of Med13 in gene regulation, we wanted to determine if variation in this gene might induce the heritable gene expression differences we observed, so we compared the list of ERGs from this study to those of the Med13 knockout strain ([75]; data not shown). Several genes were common, but the overlap was inconsistent across the four strains carrying a mutation in CNAG_03121. Additionally, some of the genes from the knockout study also showed transcript level differences common to strains that did not carry a mutation in CNAG_03121. Therefore, it is hard to say with certainty whether these changes in expression are due to the mutation of this gene.

Taken together, these data suggest that all MHC-specific adaptations that could be responsible for the virulence differences of these post-adapted strains appear to be either due to transient gene regulation, which was not measured here, or were epigenetic in origin since there were no mutations that were MHC-specific or in regulatory regions of the haplotype-specific genes that showed different transcript levels compared to the parent strain.

Discussion

To subvert the immune system, pathogens employ many different strategies, including a variety of differential expression mechanisms, including changes in transcript levels that are heritable and/or transient, as well as genomic mutations. This study examined the adaptation strategies employed by the less virulent H99W strain of C. neoformans when serially passaged through B10 congenic mice with differences only in MHC haplotype.

The H99W strain was obtained after laboratory passage from the original H99 strain (isolated by John Perfect, Duke University). It has reduced melanin production, impaired mating, and attenuated virulence compared to the parent H99O strain [77], making it a good choice for serial passage experiments, as serial passage generally increases microbial virulence [78]. The H99W strain differs from the more virulent parent strain by nine SNPs and nine indels [77], none of which were changed after passaging through MHC-congenic mice, suggesting the adaptations that occurred in a mouse host are different than those occurring during laboratory passage.

Here, we utilized a two-prong approach, first to identify heritable gene expression differences and second, to identify any shared mutations that could either be haplotype-specific, or cause heritable gene expression differences, including MHC-specific adaptations and adaptations to the mouse host.

To study MHC-specific, as well as MHC-independent changes in transcript levels in the serially passaged strains, transcriptome level analyses were conducted. This analysis identified 42 genes whose transcript levels differed from the parent strain and were shared by all strain-pairs, 19 of which have an annotated function (Table 1). In total, there were three metabolic pathway genes impacted in all six passaged strains, with two having a role in carbon metabolism and one having a role in ascorbate and aldarate metabolism. There were also four common ERGs that had a role in protein processing in the ER, and another four common ERGs having transporter functions. Of these four transporter genes, three are Major Facilitator Superfamily (MFS) transporters, which transport a variety of molecules, including both drugs and metabolites [79]. The last is an OPT family oligopeptide transporter, which is involved in the transport of glutathione and metal chelates [80]. These data suggest that enhancement of these pathways are important, MHC-independent adaptations to growth in a mouse host and include functional effects of upregulating gluconeogenesis and the unfolded protein response, as well as defense against oxidative stress.

Interestingly, all passaged strains had significantly higher transcript levels for the gene CNAG_02701, which encodes a hypothetical protein containing BAG domains, which are known to have a role in the heat shock response [46]. This gene is upregulated in a recent study of C. neoformans, along with several known heat shock proteins, in response to an antifungal treatment [47]. These data suggest, not surprisingly, that the pathogen experiences and adapts to increased stress when infecting a mouse host.

When biological pathways were examined for this gene list, shared changes to carbon fixation and protein processing in the ER were identified (Fig. 3; Table 1). Additionally, when strain-pairs passaged through the same MHC haplotype mice were compared, often an altered pathway was identifiable as enriched. Since the strains passaged through mice of a particular haplotype were performed independently, this implies that these unique changes may be important adaptations to MHC haplotypes, rather than general adaptations to the mouse host. To our knowledge, this is the first time ERGs have been reported as possible pathogen adaptations to specific host MHC alleles.

The ERGs shared by both B strains interact with the PRC, which plays a large role in gene repression or activation. All three genes identified are traditionally associated with PRC1 and have conserved roles in regulation of virulence factors, repression of target genes, and ribosome maturation, including in fungi such as C. neoformans that lack a traditional PRC1 complex. Thus, C. neoformans’ adaptation to b MHC haplotype mice led to different expression of genes responsible for regulation of virulence factors, which may have contributed to the increased virulence of these strains.

When unique transcript level differences shared between strains K1 and K2 were examined, a MAPK signaling pathway appeared to be affected (Fig. 4). MAPK signaling pathways are important to C. neoformans during infection and serve many functions from stress response to reproduction [81]. Within the MAPK signaling pathways, there were three genes that exhibited different gene expression. All three genes play a role in the HOG pathway, which is important to C. neoformans virulence [82]. In S. cerevisiae, Hog1 is phosphorylated in response to osmotic stress; however, in C. neoformans, the opposite has been reported [81]. Interestingly, the two genes with higher transcript levels, Ptp2,3 and Sko1 encode proteins that repress the stress genes activated by the HOG pathway [83, 84]. Conversely, the gene with lower transcript levels, Cdc42, is crucial to the activation of the HOG pathway [85]. Together, these data suggest that in K1 and K2 passaged strains, the HOG pathway may be diminished due to some difference in osmotic stress when adapting to infection of k MHC haplotype mice, which may affect the virulence of these strains.

Finally, there were three affected pathways specific to strains passaged through haplotype q mice (strains Q1 and Q2): ascorbate and aldarate metabolism (Fig. 5A), cysteine and methionine metabolism (Fig. 5B), and pyruvate metabolism (Fig. 6). Since the genes identified in these pathways had higher transcript levels in both Q strains, these data suggest that these differences were MHC-specific and resulted in increased gluconeogenesis, increased defense against oxidative stress and copper toxicity, increased reducing power, and increased utilization of alternate carbon sources, which likely contributed to the significantly increased virulence of the Q strains.

When shared genes among the groups were investigated, 18 metabolism genes exhibited different transcript levels in the Q strain-pair, 9 of which were shared with the B strain-pair, and 3 were shared between all passaged strains (Fig. 2E; Table 2). A few of these genes were involved in pathways only shown to be preferentially altered in the Q strain-pair, while some others were involved in different metabolic pathways without enough ERGs to be considered preferentially altered. However, there were three overlapping metabolism-related genes between the B and Q strain-pairs that were part of the glycolysis pathway, and one of these was also different in the K strain-pair (Fig. 7). There were also two genes with higher transcript levels that were common to all strain-pairs that were part of the carbon fixation pathway, as mentioned above (Fig. 3; Table 1). These common differences in gene expression in metabolic pathways, which included increased energy production and capsule synthesis, indicate that these were MHC-independent adaptations to the mouse host.

In addition to shared metabolic changes, there were also two genes with higher transcript levels that encode heat shock proteins and were shared by all strain-pairs. These heat shock proteins are involved in the stress response, as well as regulation of unfolded proteins in the ER [36, 37]. Members of the heat shock protein 70 family are upregulated in C. neoformans during infection [86]. Additionally, heat shock protein 90 is important for the virulence of C. neoformans; blocking the function of this protein results in increased drug susceptibility and decreased virulence, as this protein is essential and when inhibited, compromises the permeability of the cell membrane [87]. These shared concordant differences in gene expression also likely indicate an adaptation to a mouse host.

Consistent with this study, several previous studies have shown that gene expression in C. neoformans is significantly altered during infection, with significant changes between early and late-stage infections [88, 89]. Additionally, some studies showed significant differences in gene expression between different clinical lineages of C. neoformans, even when grown in congenic mice and isolated from the same location [89]. Furthermore, epigenetic mechanisms appear to be a significant source of drug resistance adaptation in fungi [18, 90]. While previous studies have shown that genomic mutations may help drive microevolution in C. neoformans [91], these data suggest that changes to gene expression, whether by gene regulation or by epigenetic mechanisms, is one of the primary ways that C. neoformans reacts and adapts to novel host environments, with further alterations occurring over time. It also suggests that differences in the host, potentially in MHC haplotype, may lead to heritable differences in expression.

To determine whether these differences in gene expression could be due to regulatory mutations or epigenetic changes, and also to identify any MHC-specific genomic changes, variant gene analysis was performed for each of the passaged strains. After filtering out mutations shared with H99W, as well as those with low or no evidence, there were no variant genes shared across all passaged strains and only one variant gene shared between at least one strain of each strain-pair. In the B1 strain, CNAG_03723, is a chromatin remodeling protein that shows increased expression under higher temperatures [92] and oxidative stress conditions [93], indicating its potential involvement in stress response pathways. In the B2 strain, CNAG_02745, is a DICER-like protein that has enhanced expression levels in capsule-inducing conditions, suggesting a possible involvement in virulence [94]. Since these two genes were not mutated in both B strains, they are likely not MHC-specific. In the K1 strain, CNAG_00545, a cohesion loading factor subunit SCC2, is significantly upregulated in hypoxic conditions [95], which aligns with the observed DEG expression changes in the HOG pathway in the K strains. In the K2 strain, the hypothetical protein CNAG_02594, shows higher expression in the ccr4 mutant strain at 37 °C compared to the wild-type strain under the same condition, indicating a potential role in the transcriptional response to host temperature stress [92]. Since these two genes were not mutated in both K strains, they are likely not MHC-specific. Additionally, a few of the genes identified in the Q strains were knocked out and tested for virulence in mice in Liu et al. [96]). Liu et al. devised a signature-tagged mutagenesis (STM) score for each of the knockout strains they tested. Knockout strains with negative STM scores < − 2.5 had reduced virulence whereas knockout strains with positive STM scores > 2.5 had increased virulence. CNAG_05838, which was identified as a high effect variant in the Q2 strain, and is a GTPase activating protein had a STM score of 5.28 whereas CNAG_04433, which was identified as a high effect variant in the Q1 strain, and is a CMGC/DYRK/DYRK2 protein kinase had a STM score of 2.92. The positive STM scores of these genes suggest that when these genes are deleted, the strains are hypervirulent in mice. Thus, these genomic mutations may have also contributed to the increased virulence of the Q strains. However, since these genes were not mutated in both Q strains, these genes may not be MHC-specific.

Gerstein et al., 2019 [97] have used logistic regression and principal component analysis of whole genome sequences of 38 clinical isolates to identify 40 genes that are associated with human survival, immunologic response, or clinical parameters. Of those 40 genes, CNAG_05662, which was upregulated 17-fold in the Q1 strain and encodes ITR4, a protein involved in transport or sensing of 5-carbon and 6-carbon sugar alcohols, was associated with human survival. When they tested a knockout of this gene in mice, it was associated with decreased virulence, suggesting this gene is involved in virulence in a mouse model. Thus, the upregulation of this gene may have also contributed to the virulence of the Q1 strain.

These data suggest that there were no MHC-specific mutations. These data contrast with those of a similar study using passaged lines of Friend virus in different MHC-congenic mice where phenotypic adaptation to specific host MHC haplotypes was observed [12]. The fact that MHC-specific mutations were not readily evident after passage with C. neoformans may be due to the smaller genome of Friend virus and possibly an increased mutation rate in the virus compared to that of C. neoformans.

The gene that was shared between at least one strain of each strain-pair, was CNAG_03121 and showed a missense variant in B1, Q1, and Q2, and a frameshift variant in K2 (Table 3). Previous RNA-sequencing analyses suggest that this gene is upregulated under thermal [92] and oxidative stress [93] conditions and in rim101 [98] and msl1 mutants [99] (mutant strains that are involved in regulatory pathways). Since its expression fluctuates across stress and environmental contexts, it may have a role in fungal survival in the host.

This hypothetical gene is predicted to encode the kinase domain subunit MED13 of the mediator complex, which plays a role in gene regulation in other fungal pathogens, such as C. albicans [100]. In C. albicans, the mediator complex plays a role in the regulation of genes with a variety of functions, including many virulence genes, such as those related to yeast-hyphal transition, formation of filaments, and production of adhesins; med13 homozygous-knockouts possess defects in these areas [76]. A study in C. neoformans identifies CNAG_03121 as an ortholog to S. cerevisiae’s med13 and further shows that deletion of this gene affects the expression of certain genes, specifically those involved in the glyoxylate pathway [75]. Upon comparing the list of ERGs from this study to those of the knockout strain, there were several that overlap; however the mutation status of CNAG_03121 did not accurately predict expression of these genes in our data. Therefore, it is hard to say with certainty how these differences in expression relate to the mutation of this gene. However, these data suggest that a change in this subunit and, therefore, in this complex, could be advantageous in a mouse host, potentially having a variety of effects on growth and virulence, regardless of MHChaplotype.

Other studies have described microevolution of C. neoformans in various hosts and its correlation with virulence. Ormerod et al., 2013 [101] have found that differences in carbon utilization affect virulence of serial relapse isolates from humans, which aligns with the differences in carbon metabolism pathways that we identified in the mouse-passaged strains. Additionally, Chen et al., 2017 [102] have identified a number of genes that are involved in stress responses or may be important for fitness and survival within the human host, including CNAG_00156, a transcription factor activated by calcineurin that regulates cell wall integrity [103]. This gene was upregulated in all of the mouse-passaged strains (Table 1) suggesting that this gene is important in infection of both a mouse and a human host.

There are a few limitations to this study. Since this study utilized gene expression of yeast strains not immediately isolated from the mouse host, we only identified persistent, self-perpetuating gene expression differences. While there are likely additional transient differences, dynamic gene regulation aspects to adaptation would not be detectable in these data. In addition, the cultures used for genomic DNA isolation represented a mixture of strains isolated from mice rather than a single, defined strain. This may explain the variability observed in some of the variant calls and it is possible that some true, relevant variants may have been dismissed due to a lack of evidence at the locus. Furthermore, while it is possible that additional epigenetic or mutation changes occurred post-adaptation, growing the strains in vitro for such a small amount of time compared to the time spent in the host [strains were passaged in MHC congenic mice for 102–250 days (b ~ 200–250 days; k ~ 194–250 days; q ~ 103–177 days], and the fact that the parent and adapted strains were grown in the same media for this short time, suggests that any additional epigenetic or mutation changes are likely to be negligible, or at least identical in all or most strains. Future studies utilizing newer, more robust sequencing technologies that yield longer reads and higher read depth may enable detection of structural and rare variants.

Conclusions

Taken together, this study provides a more gene-centered and mechanism-encompassing analysis to the adaptation strategies utilized by this fungal pathogen when adapted to a novel host environment, particularly related to metabolism and carbon fixation. CNAG_03121 was the only variant gene predicted to have a medium or high level of effect shared across at least one strain passaged through mice of each haplotype and thus is likely MHC-independent and related to mouse adaptation. The original study performed on these strains showed phenotypic differences in virulence and concluded that most adaptations were likely to be adaptations to mice rather than adaptations to a particular MHC haplotype [14]. While this new molecular-level investigation of these strains supports this conclusion, the current transcriptomics data also identified a number of heritable, MHC-specific ERGs, which is the first time ERGs have been reported as possible pathogen adaptations to specific host MHC alleles. We hypothesize that these differences may provide some advantage to the yeast in mice of a particular MHC haplotype, whether they are related to host immune evasion, stress response, and/or metabolism. Thus, these MHC-specific ERGs should be investigated in greater detail, particularly when there are better annotations of the exact functions of the many hypothetical genes in the genome of C. neoformans.

Authors' contributions

M.J.N. conducted the genomic and RNA sequencing of the passaged strains. C.B.A., E.E.M., and R.L.S-T. conducted genomic and transcriptomic analyses of the sequencing data. H.D.M. and W.K.P. provided funding. W.K.P. supervised E.E.M. while she passaged C. neoformans in the different strains of MHC-congenic mice to generate the passaged strains. H.D.M. supervised M.J.N. while he sequenced the passaged strains. C.B.A. wrote the first draft of the manuscript and all authors edited and reviewed the manuscript. E.E.M. and R.L.S-T. directed and supervised the omics analyses.

Funding

At the time the strains were created, WKP was awarded an R01 from NIGMS (R01GM039578); At the time of sequencing, HDM was awarded an R01 from NIGMS (R01AI100272). CBA was awarded a research grant from the Department of Biology, MTSU. The funders did not play a role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Data availability

The genomic and transcriptomic data generated in this study have been submitted to the NCBI BioProject database (https://www.ncbi.nlm.nih.gov/bioproject/) under accession number PRJNA1284378: https://www.ncbi.nlm.nih.gov/sra/?term=PRJNA1284378. Identifiers and hyperlinks at NCBI for run, biosample, and experiment, as well as sample descriptions are noted in Supplemental Table 2. The custom Python and R script were uploaded to GitHub at https://github.com/rebeccaseipelt/Cn_MHC_Evolution.

Declarations

Ethics approval and consent to participate

No live animals or human participants were involved in the current study.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.