Ethics statement

Animal experiments were performed in accordance with guidance issued by Duke University’s Institutional Animal Care and Use Committee and the U.S. Animal Welfare Act, under approved protocol A148-19-07. Animals were housed in facilities managed by veterinary staff with Duke Lab Animal Research (DLAR) and accredited by the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC).

Strains, laboratory crosses and isolation

All strains were maintained on yeast peptone dextrose (YPD) plates grown at 30°C for 48 hours from -80°C stocks. Overnight cultures for amoeba assays, melanin assays, and RNA isolation were made in liquid YPD at 30°C on a rotor drum.

Amoeba resistance assay

C. neoformans and C. deneoformans strains were grown overnight in 3 mL of liquid YPD on a roller drum before being diluted down to OD₆₀₀ 0.6. 100 μL of diluted culture was spread on solid V8 media petri dishes using glass beads. Plates were grown at 30°C for 60 hours before being removed from the incubator. Acanthamoeba castellanii (ATCC 30234) were grown in ATCC 712 in a 75 mL tissue culture flask. Amoeba were harvested, between passage 5 and 15, from flasks and suspended at a concentration of 10⁶ cells/mL before 50 μL of amoeba culture were pipetted onto the center of the Cryptococcus lawn. Plates were allowed to dry at room temperature on the bench top for 10 minutes and then placed in a 25°C incubator for 12—18 days. Measurements were taken at days 1, 12, and 18. Area of clearance was calculated by subtracting the day 1 measurement from the final day measurement as the day 1 measurement represented the initial spread of the amoeba culture on the plates. Two final time points were used (12 and 18 days) based on amoeba replication rate and activity. Rank order of amoeba affect was conserved between days 12 and 18.

Melanization assay

Melanization was assayed by growing strains on minimal media plates with L-DOPA (7.6 mM L-asparagine monohydrate, 5.6 mM glucose, 10 mM MgSO4, 0.5 mM 3,4-dihydroxy-L-phenylalanine, 0.3 mM thiamine-HCl, and 20 nM biotin) for 72 hours. Plates were then scanned on an Epson Expression XL Flatbed Scanner in reflective mode at 300 dpi. ImageJ was used to calculate greyscale intensity of the colonies. Each sample was measured in triplicate.

Spore dissection

Meiotic progeny were recovered by microdissection of random basidiospores as previously described [41]. Briefly, cells from the two parental strains were each resuspended in sterile water to a density of OD₆₀₀ = 1.0. Equal volumes of cell suspensions were mixed, and 5 μL of the mixture, as well as the two parental strains (serving as negative controls of mating), were spotted onto MS solid medium. The MS plates were incubated in the dark at room temperature (23°C for two weeks, at which time robust hyphae, basidia, and basidiospore chains were produced by the spots from the mixture of the two parental strains. Basidiospores from a large number of basidia in one location along the edge of the mating spot were picked directly from the MS plates using the needle of a dissection microscope each time, then transferred and separated onto YPD solid medium. To reduce the chances of sampling clones from the same basidia, we only separated limited numbers of basidiospores from one location (<5%), and sampled multiple locations, as well as from multiple mating spots.

DNA extraction, library preparation, and sequencing

DNA was extracted with MasterPure Yeast DNA Purification kit and cleaned up with the Zymo Research Genomic Clean and Concentrator kit (following manufacturer’s instructions) followed by quantification with PicoGreen. After quantifying the DNA with PicoGreen, samples were prepped for genomic sequencing using seqWell’s plexWell 96 kit to prepare the libraries. Briefly, samples were individually bar-coded in sets of 96 using randomly inserted transposons, pooled, and then purified. Next, each pooled sample was bar-coded, enriched, and finally size-selected purified. Libraries were sequenced at Duke University’s Sequencing and Genomic Technologies Facility on the NovaSeq 6000 S-Prime with 150 basepair paired end reads. Reads were aligned to the H99 C. neoformans reference genome using BWA. Variant calling was carried out using SAMtools and Freebayes.

Segregants filtering and SNP filtering

Segregants were filtered to remove aneuploidy and clonality as described in detail in [42]. Of the original 384 segregants we were left with 304 after filtering. Variant sites were filtered based on read depth, allelic read depth ratio, quality scores, and minor allele frequency as described in [43]. The total number of bi-allelic variant sites prior to filtering was 59,430; these were reduced down to 46,670 after filtering.

QTL mapping

The 46,670 genetic variants were combined into 4,943 haploblocks, defined by linkage, using methods described previously by Roth et al. [43]. For association testing of amoeba resistance and melanization, a Mann-Whitney U test was used across these 4,943 haploblocks to associate phenotype and genotype, coding the Bt22 and Ftc555-1 genotypes as zero and one respectively. The -log10 (p-value) from the Mann-Whitney U-tests was monitored to identify QTL and 95% confidence intervals were calculated using permutation testing, a thousand times with replacement [44].

Permutation testing

Permutation testing was carried out as described in [45] and [43] for establishing significance thresholds for QTL mapping. A thousand permutations were used for the melanin and amoeba phenotypes. Random assignments of genotype and phenotpe were held constant for every condition tested to preserve autocorrelation between phenotypes. The 95^th percentile of the permuted null distribution were used as the threshold for significance.

Population genomics

Raw Illumina sequencing reads generated by Desjardins et al. [46] were downloaded from the NIH Sequence Read Archive (BioProject ID PRJNA382844). For each of the 387 BioSamples (corresponding strains of interest) associated with the BioProject, we created a reference-based genome assembly based on aligning paired-end sequence data to the genome of the C. neoformans reference strain H99 (FungiDB R53 [47]). In cases where there were multiple sequencing runs for a given BioSample, we used the sequencing run containing the largest number of paired-end reads. To create reference-based genome assemblies we aligned reads to the H99 reference genome using BWA (v0.7.17-r1188; [48]), called variants using FreeBayes (v1.3.5; [49]), and generated strain-specific consensus assemblies by instantiating the called variants onto the reference genome. The read alignment, variant calling, and consensus assembly were carried out using the Snippy (https://github.com/tseemann/snippy) pipeline tool.

Following construction of consensus assemblies, genome feature annotation was “lifted over” from the H99 reference genome to each strain-specific genome using the software tool Liftoff (v1.6.3; [50]). The polish option of Liftoff was employed to re-align exons in cases where the lift-over procedure resulted in start/stop codon loss or introduced an in- frame stop codon. Based on the polished lift-over annotation, the AGAT GTF/GFF Toolkit software (https://github.com/NBISweden/AGAT) was used to predict protein sequences for all annotated genes in each strain-specific assembly using the agat_sp_extract_sequences.pl script. Where multiple protein isoforms are annotated in the reference genome, we generated predictions for each isoform.

Candidate BZP4 loss-of-function alleles were identified as those cases where the predicted length of the amino acid sequence of Bzp4 is < 90% of the modal protein length estimated from the entire set of strains.

To identify candidate regulatory alleles upstream of BZP4, the SAMtools coverage program was employed to summarize the coverage and read depth in the 1 kb region upstream of BZP4. Strains where the proportion of covered bases was less than 80% and there was reduced read-depth relative to the surrounding genomic region were classified as candidate regulatory deletions. These candidate regions were subsequently confirmed by manual inspection of read alignments.

Population genetic sequence diversity statistics such as π and Tajima’s D were estimated for BZP4 as well as all predicted protein coding genes on chromosome 8 and all 178 transcription factors identified by Jung et al. [51]. These estimates were calculated from multiple-sequence alignments generated from the reference based assemblies described above. Alignment were trimmed using ClipKIT [52], and the statistics were calculated using the population genetic statistical functions implemented in DendroPy [53].

RNA isolation and sequencing

12 segregants and duplicates of the parental strains were used for the analyses comprising 16 individual samples. Samples were grown overnight in liquid YPD on a rollerdrum and were added to V8 petri dishes. V8 cultures were grown at 30°C for 60 hours. Amoeba cultures were collected and suspended at a concentration of 1 × 10⁶ cells/mL. 450 μL of amoeba culture was added to the center of the Cryptococcus lawn with slight agitation to aid in the spread of the culture. Plates were allowed to dry on the benchtop for 30 minutes before being incubated at 25°C for 48 hours. A consistent area of 30 cm² was cut from the plates and then scraped to collect cells. Collected cells were resuspended in 1 mL of PBS and were placed in dry ice for 10 minutes. Samples were then lyophilized for 12—18 hours. Whole RNA was extracted using the RNAeasy Plant Mini Kit (Qiagen 74904).

Control samples did not have amoeba added to them, but were handled in a similar fashion in every other aspect of the protocol.

Libraries were prepared and sequenced by the Duke sequencing core using the Illumina NextSeq 500 High Output Kit producing 150-base pair paired end reads.

RNAseq analysis

Reads were aligned using the CNA3 of H99 C. neoformans var. grubii (accession GCA_000149245.3). from the Ensemble Fungi database. Reads were aligned using Kallisto [54].

Analysis of RNA sequences was performed using Deseq2 in R. Briefly, transcript abundance was normalized using a built in median of ratios method. Samples were normalized based on condition (amobea or control). GO term analysis was performed using the FungiDB GeneByLocusTag tool. Correlation of gene expression with BZP4 was performed using the normalized count output from Deseq2. The z-score for each gene was calculated before a Pearson correlation between BZP4 and all other genes was established. Genes with a correlation of r > |0.5| were further subset by differential expression based on the BZP4 allele.

Sequence motif analysis

DNA sequence motif analysis of the upstream 1 kb regions of genes whose expression correlates with BZP4 were carried out using the program XSTREME [55], part of the MEME Suite [56]. A control set of sequences was generated by randomly selecting the upstream regulatory sequences of 500 genes. An E-value of 10⁻⁸ was used as a cutoff to identify enriched sequence motifs. BZP4-correlated regulatory sequences were compared to three independent sets of control sequences, and we focused on motifs that were below the E-value cutoff in each of the three comparisons.

Tissue culture

The J774A.1 macrophage cell line was cultured in T-75 flasks [Fisher Scientific] in Dulbecco’s Modified Eagle medium, low glucose (DMEM) [Sigma-Aldrich], supplemented with 10% live fetal bovine serum (FBS) [Sigma-Aldrich], 2mM L-glutamine [Sigma-Aldrich], and 1% Penicillin and Streptomycin solution [Sigma-Aldrich] at 37°C and 5% CO₂.

Phagocytosis assay

To measure the phagocytosis of various Cryptococcus segregants by macrophages, J774A.1 cells were seeded at a density of 1x10⁵ cells per well of a 24-well plate [Greiner Bio-One], then incubated overnight at 37°C and 5% CO₂. At the same time, an overnight culture of C. neoformans parental strains or segregants was set up by picking a fungal colony from YPD agar plates (50g/L YPD broth powder [Sigma-Aldrich], 2% Agar [MP Biomedical]) and resuspending it in 3mL liquid YPD broth (50 g/L YPD broth powder [Sigma-Aldrich]). The culture was then incubated at 25°C overnight under constant rotation (200 rpm).

On the day of the assay, macrophages were activated using 150ng/mL phorbol 12-myristate 13-acetate (PMA) [Sigma-Aldrich] for 1 hour at 37°C. PMA stimulation was performed in serum-free media to eliminate the contribution of complement proteins during phagocytosis. To prepare C. neoformans for infection, overnight C. neoformans cultures were washed two times in 1X PBS, counted using a hemacytometer, and fungi was incubated with macrophages at a multiplicity of infection (MOI) of 10:1. The infection was allowed to take place for 2h at 37°C and 5% CO₂. After 2 h infection, as much extracellular Cryptococcus as possible was washed off using 1X PBS.

Fluorescent microscopy imaging

The number of phagocytosed fungi was quantified from images from a fluorescent microscope. To distinguish between phagocytosed and extracellular C. neoformans, wells were treated with 10 μg/mL calcofluor white (CFW) [Sigma-Aldrich] for 10 mins at 37°C. The wells were washed again with PBS to remove residual CFW. Fluorescent microscopy images were acquired at 20X magnification using the Nikon Eclipse Ti inverted microscope [Nikon] fitted with the QICAM Fast 1394 camera [Hamamatsu]. Images were analysed using the Fiji image processing software [ImageJ]. To quantify the number of phagocytosed Cryptococcus from the resulting images, the total number of ingested C. neoformans was counted in 200 macrophages, then the values were applied to the following equation: ((number of phagocytosed C. neoformans/number of macrophages) × 100).

Intracellular proliferation rate assay and time-lapse imaging

To investigate the intracellular proliferation rate (IPR) of Cryptococcus strains within macrophages, infected macrophages were captured at a regular interval over an extended period. Time-lapse imaging was performed by running the phagocytosis assay as usual, then after washing off extracellular Cryptococcus with 1X PBS, serum-free culture media was added back into the wells before imaging. Images were captured using the Nikon Eclipse Ti microscope at 20X magnification. Images were acquired every 5 minutes for 18 hours at 37°C and 5% CO₂.

The resulting video was analysed using Fiji [ImageJ] and IPR was determined by quantifying the total number of internalised fungi in 200 macrophages at the ‘first frame’ (time point 0 (T0)) and ‘last frame’ (T10). The resulting values were used in the following equation: ((number of phagocytosed C. neoformans/number of macrophages) × 100). Next, the number of phagocytosed fungi at T10 was divided by the number of phagocytosed fungi at T0 to give the IPR (IPR = T10/T0).

Mouse infections

Cryptococcus strains for inoculation were grown overnight in 5 ml of YPD broth at 30°C in a roller drum. Cells were pelleted by centrifugation and washed twice with sterile PBS. The cell pellet was resuspended in PBS, diluted, and counted by hemocytometer. The final inoculum was adjusted to a cell density of 4 x 10⁶ CFU/ml. Test groups consisting of five male and five female A/J mice aged 4–5 weeks were purchased from Jackson Labs (stock #000646) and infected via intranasal instillation. Mice were anesthetized using isoflurane administered with a calibrated vaporizer. 25 μl of the prepared inoculum was pipetted into the nares one drop at a time until the full volume containing 10⁵ CFU was inhaled. Mice were observed until fully recovered from anesthesia. Following infection, mice were monitored daily for symptoms of disease progression including weight loss, labored breathing, lack of grooming, social isolation, and any signs of pain or distress. Mice were euthanized upon reaching humane endpoints according to guidelines set forth by Duke University’s Animal Care and Use Program. Survival curves were plotted using GraphPad Prism version 8 and analyzed using log-rank (Mantel-Cox) statistical test.

Comparison of amoeba resistance in diverse C. neoformans Strains

We developed a plate-based assay to quantify Cryptococcus resistance to predation by the amoeba, Acanthamoeba castellanii (Fig 1A; [57]). Briefly, an established lawn of Cryptococcus cells was inoculated with a drop of amoeba. After a defined period of time, the cleared (consumed) portion of the lawn is quantified and used as a measure of resistance. The larger the clearance area, the less resistant the cells are to amoeba consumption.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Amoeba resistance varies between strains of C. neoformans.

A. A schematic overview of the amoeba resistance assay. A lawn of Cryptococcus is grown on V8 media for 60 hours. A. castellanii are added to the center of the lawn. A. castellanii consume the cells across the plate for a period of 12 or 18 days (based on amoeba activity). Cryptococcus strains able to resist amoeba have smaller areas of consumption when imaged. B. A. castellanii in co-culture with C. neoformans expressing RFP on solid V8 media. White arrows indicate examples of C. neoformans—amoeba interaction. C. Boxplots representing amoeba resistance phenotypes for a diverse set of C. neoformans strains after 18 days of amoeba co-culture. The x-axis displays the strains assayed. The y-axis represents the area the amoeba consumed. Smaller clearance areas indicate greater resistance to amoeba.

https://doi.org/10.1371/journal.ppat.1011763.g001

This amoeba resistance assay was applied to a diverse set of C. neoformans strains that represent major sub-lineages within this species (Table 1; [25, 58–61]). This assay revealed extensive variation in amoeba resistance between strain backgrounds (Fig 1C). Notably, there is no simple relationship between the site of collection and amoeba resistance; clinical strains isolated from patient samples exhibited both resistant and sensitive amoeba resistance phenotypes.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Genetically diverse C. neoformans strains surveyed for amoeba resistance.

https://doi.org/10.1371/journal.ppat.1011763.t001

Mapping populations and genome sequencing

From our collection of genetically diverse strains, strains of opposite mating type (MAT a-MATα) were identified that differed in their resistance to amoeba. We carried out pairwise mating tests to identify strain pairs with sporulation and germination efficiency suitable for establishing a large genetic mapping population. C. neoformans strains Bt22 and Ftc555-1 were chosen for further analysis based on spore viability and differences in their amoeba resistance. The low-resistance strain Bt22 (MAT a) is a clinical isolate while the high-resistance strain Ftc555-1 (MATα) is an environmental isolate collected from a mopane tree; both strains were collected in Botswana [61]. By manual spore dissection, we isolated 384 progeny from a cross between these two strains. The genomes of these progeny were then sequenced on the Illumina NovaSeq 6000 platform to an average depth of ~15×. Based on the resulting sequence data, the progeny were filtered based on criteria including sequencing depth, read quality, elevated ploidy, and clonality. After filtering, the final mapping population was composed of 304 recombinant progeny. 46,670 variable sites were identified between the parental strains that were collapsed into 4,943 haploblocks.

Cross-species amoeba resistance QTL

The F₁ segregants generated from the Bt22 × Ftc555-1 cross exhibited a diverse response to amoeba predation (mean of predation area 23.46 cm² and SD 14.99 cm²). There is a substantial amount of transgressive segregation—16.9% of the segregants exhibited resistance higher than Ftc555-1, and 38.4% of segregants exhibited lower resistance than Bt22 (Fig 2B). Substantial transgressive segregation suggests that epistatic interactions between parental alleles contribute to both increased and decreased resistance beyond the parental phenotypes. Segregant genotypes and phenotypes were combined to carry out QTL mapping based on a marker regression approach [43]. This QTL analysis revealed that genetic variation for amoeba resistance in the mapping population is dominated by a single, major effect locus on chromosome 8 (Fig 3A). Segregants with Bt22 haplotypes at the chromosome 8 QTL peak exhibited significantly larger zones of amoeba clearance than those offspring with the Ftc555-1 haplotype (Fig 3B). The QTL on chromosome 8 explains an astonishing 62% of variation in amoeba resistance.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Phenotypic variation in amoeba resistance in mapping populations derived from both C. neoformans and C. deneoformans crosses.

A. Representative images of plates from amoeba resistance assays. On the plates, the area consumed by amoeba are highlighted in yellow. Parental strains are shown in the middle panels, and transgressive segregants are on the left and right. B. A histogram displaying amoeba resistance of segregants in the C. neoformans cross. The x-axis represents the amoeba clearance area. Phenotypes of the two parental strains are indicated in orange (Ftc555-1) and blue (Bt22). C. A histogram of the C. deneoformans F₁ progeny amoeba phenotypes. Phenotypes of the two parental strains are indicated in orange (431) and blue (XL280).

https://doi.org/10.1371/journal.ppat.1011763.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Amoeba resistance QTL for C. neoformans and C. deneoformans.

A. Manhattan plot representing the association between genotype and amoeba resistance in the C. neoformans mapping populations. The dotted line indicates the significance threshold by permutation. B. Distributions of segregant phenotypes associated with the QTL peak on chromosome 8 for C. neoformans. The x-axis represents allelic state at the QTL peak. This peak explains 62% of the amoeba resistance variation (R²). C. Manhattan plot for amoeba resistance in the C. deneoformans mapping population. This peak explains 62% of the amoeba resistance variation (R²). D. Segregant phenotypes by chromosome 7 genotype for C. deneoformans. E. A magnified view of the 95% confidence interval of the C. neoformans QTL for amoeba resistance. Barred lines at the top are the 95% confidence intervals for the C. deneoformans and C. neoformans amoeba resistance QTL. F. Gene diagrams for the region around the BZP4 gene for Ftc555-1 and Bt22. UTRs are shown in yellow.

https://doi.org/10.1371/journal.ppat.1011763.g003

To determine if there are similarities in the genetic architecture of amoeba resistance between closely related pathogenic species of Cryptococcus, we carried out a similar analysis using a mapping population derived from a C. deneoformans cross described in [43]. This cross, between strains XL280a and 431α, consists of 90 recombinant progeny. XL280a and 431α have only modest differences in amoeba resistance, but similar to the findings in C. neoformans, the C. deneoformans offspring exhibited a high degree of transgressive segregation for this trait. 15.4% of offspring displayed negative transgressive segregation (segregants with lower amoeba resistance than XL280a), 54.8% positive transgressive segregation (segregants with lower amoeba resistance than 431α), and 29.8% non-transgressive (Fig 2A and 2C). QTL analysis of this population identified a significant peak on chromosome 7 that explains 23% of the variance of amoeba resistance (Fig 3C). Progeny with the 431α allele on chromosome 7 have a higher average resistance to amoeba (Fig 3D). By examination of the genes under the QTL peak on chromosome 7, this region was found to be orthologous to the C. neoformans QTL peak on chromosome 8. These two regions share 82% nucleotide sequence identity and conserved synteny (Fig 3E), suggesting there are conserved genes required for amoeba resistance within the Cryptococcus species complex.

Identification of an amoeba resistance gene

To identify candidate causal variants for the QTL region on chromosome 8 in the C. neoformans cross, we analyzed the predicted effect of nucleotide sequence differences on annotated features in this region (S1 Table). 25 genes were within the identified 64 kb region that comprises the 95% confidence interval for this QTL. Ten of these genes have an annotated function or have homology to annotated genes in other fungal species. Across all of the genes in the QTL region, 31 synonymous mutations, 29 non-synonymous mutations, and two indels were identified. The two indels both result in nonsense mutations but the predicted genes in which they occur have no characterized function and no known gene deletion phenotype. Characterization of genetic variation in non-coding regions in the QTL region led to the identification of a large sequence difference between Bt22 and Ftc555-1, a 1789 bp deletion occurs in the Bt22 background in the intergenic region between the BZP4 and HSP78 genes. The Bt22 variant truncates 100 bp of the annotated BZP4 5’ UTR and 1689 bp further upstream (Fig 3F).

With C. neoformans phylogenetic data and short-read sequence data from [46], a strain was identified, Bt45, that is nearly genetically identical to Bt22 (~200 SNP differences) but does not share the deletion upstream of BZP4 (Fig 4A). Like Bt22, Bt45 is a clinical isolate collected in Botswana from a patient with HIV/AIDS in the early 2000s [60, 62]. Bt45 was found to exhibit significantly greater amoeba resistance than Bt22 (pairwise t-test, p < 0.0005), though not to the level observed for Ftc555-1 (Fig 4B). This comparison between these two nearly genetically identical strains is analogous to an “allele exchange” experiment and provides strong evidence that the non-coding variant identified upstream of BZP4 is the likely causal variant underlying the Chromosome 8 QTL.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Disruption of BZP4 reduces amoeba resistance.

A. Models of the genomic regions surrounding BZP4 in Bt22 and Bt45. B. Amoeba resistance assay for two independent gene deletions of BZP4 in the Ftc555-1 background and a closely related strain of Bt22, Bt45. Bars are colored for comparison between closely related strains. The x-axis represents the strain tested. The y-axis represents the area amoeba consumed.

https://doi.org/10.1371/journal.ppat.1011763.g004

To provide further evidence of BZP4’s contribution to amoeba resistance, CRISPR-Cas9 editing was utilized to delete BZP4 in the Ftc555-1 background [63, 64]. Two independent bzp4Δ mutants were isolated and their amoeba resistance phenotypes were assessed. Both mutants exhibited a significant reduction in amoeba resistance (pairwise t-test, p < 0.0005) and melanization (Fig 4B and S3 Fig).

In sum, multiple lines of evidence suggest that the 1789 bp deletion we identified upstream of BZP4 is the causal variant for the large effect amoeba resistance QTL we identified on chromosome 8. For the sake of conciseness, in the text that follows the two allelic states at this locus are referred to as BZP4^B (Bt22 allele) and BZP4^F (Ftc555-1 allele).

BZP4 is a pleiotropic QTG for amoeba resistance and melanization

BZP4 is a transcription factor that has been shown to play a role in regulation of the melanin synthesis pathway under nutrient deprivation conditions [65]. Variation at the BZP4 locus was previously identified in a genome-wide association study (GWAS) for melanization [46]. In that study, bzp4 loss-of-function mutations, found exclusively in clinical isolates, were shown to correlate with reduced melanization. A later study found that decreased expression of BZP4 is correlated with decreased melanization in VNI clinical isolates [66].

Based on the role of BZP4 in the regulation of melanin synthesis, we reasoned that the BZP4 variant identified in the amoeba resistance mapping might also result in differences in the ability to produce melanin. While neither of the parent strains in the cross lacks melanin, Bt22 exhibits less melanin pigmentation than Ftc555-1 when grown under the same inducting conditions (Fig 5A). The C. neoformans mapping population was assayed for the ability to produce melanin when grown on L-DOPA plates. Segregants in the cross ranged from completely white (devoid of melanin) to a deep ebony color accompanied by melanin leaking into the surrounding media (Fig 5A). Across the segragants, 23.45% of the segregants displayed positive transgressive segregation (more melanized than Ftc555-1) while 24.43% displayed negative transgressive segregation (less melanized than Bt22). When the joint distribution of amoeba resistance and melaninization phenotypes among the offspring was assessed, a positive but non-linear relationship was observed (Fig 5D).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Melanization and amoeba resistance share the same QTL.

A. Manhattan plot representing the association between genotype and melanization in the C. neoformans plot. The y-axis represents the strength of the association between genotype and light reflected (the degree of melanization). The x-axis represents the genomic location of the haploblocks used in the associations. The dotted line represents a significance threshold determined by a permutation test. Representative images of parents and segregants on L-DOPA media are also included. Each colony is from the randomized plates employed for QTL mapping. Images are brightened 30% to better display the difference in pigmentation. B. Segregant phenotypes at the maximum significance value of the QTL on chromosome 8. The x-axis represents the segregant allele at the maximum significance of the QTL. The y-axis represents the light reflected off of the colony when melanized. Parental strains are indicated in orange (Ftc555-1) and blue (Bt22). C. A magnified view of the chromosome 8 QTL peaks for amoeba resistance and melanization illustrating the QTL overlap. D. Plot comparing amoeba resistance and melanization phenotypes. The x-axis represents melanization and the y-axis represents amoeba resistance. Each dot represents a single segregant. Segregants are colored by their allele at the chromosome 8 QTL. E. Relationship between amoeba resistance and melanization in natural isolates. The linear regression line, and the coefficient of determination, are indicated for the linear model fit to the data with the strain H99 (outlier, top left) excluded.

https://doi.org/10.1371/journal.ppat.1011763.g005

QTL mapping based on the melanization phenotypes identified a major peak on chromosome 8 nearly identical in location to the QTL for amoeba resistance (Fig 5C). This QTL explains a remarkable 50.2% of the phenotypic variation for melanization (Fig 5B). Based on the similarity of the QTL for amoeba resistance and melanization (Fig 5C), as well as the previously demonstrated role of BZP4 in the regulation of melanin synthesis, we propose that the non-coding deletion upstream of BZP4 has pleiotropic effects on both of these traits.

In contrast to the findings in C. neoformans, the chromosome 7 QTL for amoeba resistance in the C. deneoformans cross does not appear to have a pleiotropic effect on melanization. Instead, a nonsense mutation in the gene RIC8 is primarily responsible for variation in melanization for this cross as described in an earlier study from our research groups [43].

To test whether the relationship observed between amoeba resistance and melanization in our C. neoformans mapping population holds more broadly, we again employed the genetically diverse collection of C. neoformans strains described above. This collection was supplemented with additional strains that have predicted BZP4 loss-of-function mutations, and amoeba resistance and melanization for each isolate was measured. All strains, including those with predicted BZP4 loss-of-function mutations, are capable of producing melanin given sufficient incubation time (S3 Fig). However, large differences in the rate of melanization between strains were observed and images taken at two days of growth as a measure of the variation were used. When comparing amoeba resistance and melanization across the diverse strain set, the reference strain H99 is notable as an outlier in terms of the bivariate relationship among these traits (S4 Fig). With H99 excluded, there is a strong linear relationship between amoeba resistance and melanization (R² = 0.58) (Fig 5E). A notable trend among strains with predicted BZP4 loss-of-function mutations is that those that melanize more readily (Bt103 and Bt102) are more resistant to amoeba than those that melanize slowly (Bt22 and Bt75) (Fig 5E).

Gene expression differences associated with BZP4 allelic variation

Because the BZP4 allele identified involves a deletion of a large upstream non-coding region, we hypothesized that the phenotypic effects of this allele are mediated by a reduction in the expression of the BZP4 gene, with consequent effects on the downstream targets of this transcription factor. To test this hypothesis, gene expression was profiled with RNAseq. Using six offspring with each BZP4 genotype (6 for Bt22 and 6 for Ftc555-1; 12 strains in total) transcriptional responses when grown on V8 medium were compared with or without the addition of amoeba.

BZP4 was significantly differentially expressed between strains with the BZP4^B and BZP4^F alleles, in both amoeba and non-amoeba conditions (Fig 6A and 6B). BZP4^B strains exhibited an average 1.83-Log₂ fold decrease in expression relative to BZP4^F strains when co-cultured with amoeba and a 2.06-Log₂ fold decrease when amoeba were absent (Fig 6A). No other gene within the chromosome 8 QTL showed statistically significant differences in expression. While there are differences in BZP4 expression between genotypes, no significant change in the expression of BZP4 was observed between control and amoeba conditions (Fig 6A). This suggests that the effect of the BZP4 allelic differences identified is not specific to the amoeba challenge conditions of our assay.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. BZP4 expression differs significantly between genotypes.

A. Points representing the difference in transcript abundance between conditions in transcript per million (TPM) in segregants with each parental allele under the chromosome 8 QTL. The y-axis represents the total transcript counts. Boxplots are colored by parental allele. Significance is determined by ANOVA. “**” p < 0.005; “***” p < 0.0005. B. A Venn Diagram displaying the number of genes with increased and decreased expression for amoeba and control conditions based on the parental allele. Genes with decreased expression are colored in orange and those with increased expression are colored in blue. C. A hierarchically clustered heatmap of genes that strongly correlate with BZP4 expression and are differentially expressed based on the BZP4 allele in amoeba treated samples. Rows represent genes. Columns represent individual strains (parents and segregants). The colors of each cell represents z-scores. Colored rows at the bottom represent strains by their BZP4 genotype and their amoeba sensitivity (area of clearance, measured in cm²).

https://doi.org/10.1371/journal.ppat.1011763.g006

Given that BZP4 was identified as a candidate QTG for melanization, and the transcription factor it encodes has been previously implicated in the regulation of melanin synthesis genes, we predicted that such genes would also exhibit differences in expression as a function of BZP4 genotype. Contrary to this prediction, we found that no major melanin synthesis genes were significantly differentially expressed between genotypes (when filtered for biological significance by fold change), when measured on the V8 growth media used in the amoeba experiments (S2 Table).

Investigating genome wide expression differences, 587 genes exhibited a greater than 2-fold difference between genotypes in either amoeba and control conditions. Of the 587 differentially expressed genes, 130 are shared between conditions, 254 are specific to the control conditions, and 203 are amoeba specific (Fig 6B). Using GO term analysis for the 90 shared genes that have increased expression when BZP4 expression is reduced, we find transmembrane transporter activity, oxidoreductase activity, and transition metal ion binding to increase in activity with reduced BZP4 expression. Interestingly, there is a paucity of GO predictions for the 40 shared genes that are decreased in expression with reduced BZP4 expression, as many of the genes are hypothetical or poorly characterized. The number of genes that have an inverse transcriptional relationship with BZP4 expression indicates a role in gene repression. GO terms specific to amoeba conditions further include transmembrane transporter activity and oxidoreductase activity.

To identify genes potentially regulated by Bzp4 itself, we further focused on genes whose expression is strongly (anti-)correlated with BZP4 (Pearson correlation, |R| ≥ 0.5) and that show differential expression between strains with the BZP4^B and BZP4^F alleles (log₂-fold change > 1 and p_adj ≤ 0.05). This subsetting identified 36 genes that are positively correlated with BZP4 and 62 genes that are negatively correlated (Fig 6C and S3 Table). Hierarchical clustering of these genes, illustrated in Fig 6C, emphasizes the highly distinct expression patterns that these genes exhibit between strains with high and low amoeba resistance. The GO profile of this gene set is highly similar to that of the broader set of differentially expressed genes.

To identify potential regulatory motifs in the promoter regions of genes with similar expression patterns to BZP4, we used the motif analysis tool XSTREME [55, 56] to analyze 1 kb regions upstream of the 98 BZP4-correlated genes (positive and negatively correlated genes analyzed separately). Using this approach we identified the enriched sequence motif CACAKGCWA (K = T/G, W = T/A; S1 Fig) which is found in 25% of the genes positively correlated with BZP4, including BZP4 itself, compared to a background rate of <1% across sets of 500 random genes. This enriched motif bears similarity to the general E-box motif CANNTG typically bound by basic Helix-Loop-Helix (bHLH) transcription factors [67] and a particularly close match to the PBE-box motif, CACATG, which is bound by phytochrome interacting bHLH transcription factors [68]. No consistent sequence motif was identified among the negatively correlated genes.

Population genetics of BZP4

Providing a more detailed understanding of allelic variation at BZP4 in C. neoformans, we reanalyzed BZP4 sequence variation based on 387 sequenced C. neoformans strains, originally described in Desjardins et al. [46].

Desjardins et al. identified four strains (Bt75, Bt102, Bt103, Bt147) within the VNBI and VNBII lineages with BZP4 loss-of-function (LOF) alleles. Our bioinformatics analysis similarly predicted LOF alleles for these four strains and identified three additional VNI strains with likely LOF alleles—Bt3, Bt107, and Bt156. All seven of these strains with predicted BZP4 LOF alleles were isolated in clinical settings. Based on the phylogenetic relationships of these strains, and the specific locations of the stop-gain mutations, it is likely that each of these alleles arose independently.

To explore genetic variation at the BZP4 locus more broadly, we calculated two widely used measures of nucleotide sequence variation: a) π, the average per-base number of variable sites between pairs of individuals in a population; and b) Tajima’s D, a summary statistic that can be used to identify genomic regions with an excess of rare alleles. Negative values of Tajima’s D, which indicates an excess of rare alleles, can be caused by selective sweeps but can also result from demographic phenomena such as population bottlenecks. We compared BZP4 estimates of π and Tajima’s D to estimates of the same parameters for 177 other transcription factors [51] and to all other annotated genes on chromosome 8 on a per lineage basis (S4 Table). In the VNI lineage, rare BZP4 alleles are somewhat elevated (Tajima’s D = -1.97) compared to other genes on chromosome 8 and somewhat elevated compared to other transcription factors (S4 Table). There is no evidence for an overabundance of rare BZP4 alleles in the VNBI or VNBII lineages compared to the chromosome 8 average or relative to other transcription factors.

For the VNBI lineage, there are sufficient numbers of environmental and clinical strains (74 and 112 strains respectively) to compare Tajima’s D by isolation source. Estimates of Tajima’s D for BZP4 are lower in clinical strains (Tajima’s D = -1.47) compared to environmental strains (Tajima’s D = -0.66). However, the difference in Tajima’s D between these two groups is not unusual relative to other genes on chromosome 8 or compared to other transcription factors.

Epistatic QTLs for amoeba resistance and melanization

While the pleiotropic chromosome 8 QTL we identified in the C. neoformans cross explains a large portion of variation for both amoeba resistance and melanization, both traits show continuous rather than bimodal distributions and there is a large degree of transgressive segregation. These observations suggested that there are likely additional alleles, perhaps interacting epistatically with the major effect allele on chromosome 8, that contribute to phenotypic differences in both of these traits. To test for epistatic interactions, our mapping population was subdivided based on genotype at the chromosome 8 QTL peak, and we re-ran the QTL mapping procedure for each subpopulation (S2 Fig). For amoeba resistance, a single epistatic QTL was found on chromosome 5, exclusive to the segregants with the Ftc555-1 allele at the chromosome 8 QTL (S2(A) Fig). This epistatic chromosome 5 QTL explains 19% of the variation within that subgroup and it increases the overall variance explained for amoeba resistance to 64%. We identified two epistatic QTL for melanization, one in the segregants that have the Bt22 chromosome 8 QTL allele and the second in those that have the Ftc555-1 allele (S2(B) Fig). The epistatic QTL in the Bt22 background is found on chromosome 1 and explains 26% of the variation within that subgroup. The epistatic allele in the Ftc555-1 background occurs on chromosome 7 and accounts for a more modest 7.8% of variance. With these epistatic interactions included, the variance in melanization explained by all of the QTL identified increases to 56.4%. Evidence of epistasis for amoeba resistance and melanization highlights the importance of strain background and the impact of individual allelic differences on traits of interest.

Comparing amoeba resistance and virulence

The accidental pathogen hypothesis is based on the similarities between amoeba and macrophage interactions with Cryptococcus. Both amoeba and macrophages employ similar methods of detecting, phagocytosing, and degrading fungal cells but the question remains: does survival when challenged with one phagocyte relate to success with the other? To answer this question, we measured the intracellular proliferation rate (IPR) in macrophages of progeny from the Bt22 × Ftc555-1 cross. F₁ progeny were chosen to represent opposite extremes in terms of their amoeba resistance phenotypes. These low and high resistance strains were assayed alongside the parental strains. J774A.1 murine macrophages were infected with C. neoformans cells and the internal proliferation rate of yeast cells was measured using time lapse microscopy as described in the Methods. All F₁ progeny assayed, regardless of BZP4 genotypes and amoeba resistance phenotypes, showed similar macrophage internal proliferation rates (Fig 7A). Phagocytic index, another measure of yeast-macrophage interactions, also showed no association with BZP4 allelic variation or amoeba resistance (S5 Fig). Thus, in contrast to the predictions of the accidental pathogen hypothesis, we do not observe a relationship between amoeba resistance alleles and survival in macrophages.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. Amoeba resistance and measures of virulence are uncorrelated.

A. Barplots representing the internal proliferation rate of parental strains and segregants. Boxplots are colored by the chromosome 8 allele. Orange boxes have the Ftc555-1 allele and blue have the Bt22 allele, darker colors indicate segregants. Dots represent individual measurements. Strains are oriented in rank order of amoeba resistance (highest to lowest resistance). Significance determined by ANOVA (p = 0.29 F = 1.28). B. Survival curves for animals infected with the parental strains and a group of F₁ segregants. H99 is in black, Ftc555-1 is in orange, and Bt22 is in blue. Segregant curves are colored by parental allele under the chromosome 8 QTL, dark blue represents the Bt22 allele and darker orange represents the Ftc555-1 allele.

https://doi.org/10.1371/journal.ppat.1011763.g007

To explore the relationship between amoeba resistance and the ability to cause disease in animal models, an equal number of 4–5 week old male and female A/J mice were intranasally infected with F₁ progeny from the C. neoformans cross, the parental strains, and the reference strain H99. Survival was monitored for a period of 179 days, with animals sacrificed based on disease progression symptoms (Fig 7B). Though this analysis involve only a modest number of strains, we observed no relationship between BZP4 genotype and virulence in mice. The virulence of the parental strains, Bt22 and Ftc555-1, is the opposite of their amoeba resistance phenotypes. Bt22, which exhibits low resistance to amoeba predation, has modest virulence with a time to 50% lethality (LT50) of ~92 days. This is in stark contrast to Ftc555-1, which is highly resistant to amoeba, but is completely avirulent in the mouse model of infection employed. The reference strain, H99, is strongly virulent in mice (LT50 ~21 days) but has very low amoeba resistance. Furthermore, we detected no association between LT50 estimated from murine survival curves and the BZP4 genotype of a small number of segregants (Fig 7B).

Finding a lack of correlation between amoeba resistance and virulence for strains from our mapping population, the analysis was broadened to include an additional nine, genotypically diverse C. neoformans strains. Using the same intranasal murine infection model described above, we found that virulence was highly variable among strains, but again found no correlation between amoeba resistance and LT50 measures of mouse survival (S5 Fig) [69].