A cDNA library has been constructed in the previous study (Pongpom et al., 2005). Briefly, Talaromyces marneffei ATCC 200051 (CBS119456, F4) human isolated strain was cultured in a brain heart infusion broth (BHIB) at 37°C for 3 days. Total RNA was isolated from the yeast cells with TRIzol^® reagent (Gibco BRL, Gaithersburg, MD, USA). Then the messenger RNA (mRNA) was isolated with an mRNA purification kit (Oligotex; QIAGEN, Germany). SuperScript^®cDNA synthesis system (Gibco BRL) was used in the cDNA library construction in the form of lambda bacteriophage.

The sera were collected from the blood specimens of T. marneffei-infected AIDS patients that were sent for clinical analysis at Maharaj Nakorn Chiang Mai hospital. Five sera were pooled, and an immunoglobulin G was purified from the serum by using the HiTrap Protein G HP column (Amersham Pharmacia Biotech, Uppsala, Sweden). The purified immunoglobulin was incubated with E. coli Y1090 lysate for 2 hours at 37°C to absorb the antibodies directed to the E. coli determinants.

The constructed cDNA library was screened with a purified IgG from a pooled serum. The antibody screening experiment was performed as described previously (Pongpom et al., 2005). The positive phage clones were then isolated and purified by repetition of the antibody screening process until the homogeneous positive signal was generated. Approximately 10⁵ independent plaques from the amplified library were screened with the purified immunoglobulin to obtain immunogenic protein-encoding clones. E. coli Y1090 was infected with the phage library and allowed the plaques to grow at 42°C for 4 h. A 50 mM isopropyl-β-D-thiogalactopyranoside (IPTG)-impregnated nitrocellulose membrane (Hybond C-extra; Amersham Pharmacia Biotech) was overlaid on the surface of the culture plate for 3 h at 37°C. The membrane was lifted and then incubated for 4 h at room temperature in a blocking buffer (5% non-fat dry milk, 0.1% Triton X-100 in Tris buffer saline). Then, the membrane was incubated with the purified IgG (25 μg/ml) and HRP-conjugated goat anti-human IgG (20,000-fold dilution in the blocking buffer) for 1 h at room temperature. The antigen-antibody complex was detected by a chemiluminescent substrate (SuperSignal Substrate West Pico; Pierce, Rockford, IL, USA). The positive phage clones were selected and purified by repeated screening until a homogeneous positive signal was generated. Phage-to-plasmid conversion of positive clones was performed by in vivo excision using E. coli DH10B strain (Gibco BRL) and the insert cDNA were subjected to sequencing by dideoxynucleotide chain termination method.

The DNA sequences of the antigenic protein-encoding clones were deposited in the Genbank database under the following accession numbers: P1; OQ241945, P3; OQ241946, P6; OQ241948, P7; OQ241949, P9; OQ241950, P10; OQ241951, P11; OQ241944, P12; OQ241952, P13; OQ241953, P14; OQ241954, P15; OQ241955, P17; OQ241956, P21; OQ241957, P23; OQ241947, P24; OQ241958, P26; OQ241959, P28; OQ241960.

DNA sequences were submitted to the basic local alignment search tool for nucleotides (BLASTn) to search for sequence similarity against T. marneffei strain ATCC18224, and fungal models Saccharomyces cerevisiae, Schizosaccharomyces pombe, and Neurospora crassa. Amino acid sequences of identified antigenic proteins and other fungal homologs were obtained from BLAST search tool for protein (BLASTp). The Clustal program was employed to perform protein sequence alignment. Functional annotation of identified T. marneffei proteins was investigated using DAVID server (https://david.ncifcrf.gov/tools.jsp). Functional domains were predicted using Prosite (https://prosite.expasy.org/scanprosite/), SMART (http://smart.embl-heidelberg.de/), and Uniprot (https://www.uniprot.org/) tools. If data prediction was available, STRING (Szklarczyk et al., 2021) was used to analyze the protein-protein interaction network (https://string-db.org/cgi/input.pl), and EggNOG server was used for orthology analysis (Hernández-Plaza et al., 2022) (http://eggnog6.embl.de). The presence of secretory signal sequences and protein subcellular localization were predicted using the SignalP and WoLF PSORT programs (Horton et al., 2007; Teufel et al., 2022). The presence of the proteins within T. marneffei extracellular vesicles (EVs) was performed by comparison to available data previously described (Yang et al., 2020). For B-cell linear antigenic epitope prediction, the amino acid sequences were evaluated using the web-based tools Bepipred 2.0, ABCpred, and SVMTriP (Saha and Raghava, 2006; Yao et al., 2012; Jespersen et al., 2017).

T. marneffei strain ATCC200051 (CBS119456, F4) was cultured on Sabouraud dextrose agar (SDA) at 25°C for 10-14 days to generate conidia. The conidia were harvested and 1x10⁸ conidia/ml were inoculated in a 50-ml Sabouraud dextrose broth (SDB). Cultures were incubated either at 25°C (mold phase) or 37°C (yeast phase) with continuous shaking at 200 rpm. After 72 hours, cultures were collected by centrifugation at 4°C, 7,000 rpm for 30 min. Total RNA was isolated by using TRIzol^® reagent, treated with DNase I, and converted to cDNA as previously described (Amsri et al., 2021). Quantitative real-time PCR was performed using the SYBR Green qPCR mix (Thunderbird SYBR Green Chemistry, TOYOBO). An actin gene was included as a reference gene. All primers used in this study are listed in Supplemental Table S1 . Calculation of a relative expression was performed using the 2^-(ΔCt) where ΔC_t = C_t actin – C_t target). One-way analysis of variance (ANOVA) with Tukey’s multiple comparisons test was performed to test for a significant difference among pairs, using Prism software (GraphPad, version 7.0). Statistical significance was set to P < 0.05. Graphs were generated using Prism software. Error bars indicate standard deviation.

Heatmap indicating transcript abundance and expression pattern was generated using the HEATMAP hierarchical clustering web tool (https://www.hiv.lanl.gov/content/sequence/HEATMAP/heatmap.html). The log2 fold-change values of relative expression levels (the 2^-(ΔCt) value) were used to construct the heatmap. For visualization purposes, the non-detected expression value of MPLP6 gene in the conidia and mold phases was adjusted to the lowest gene expression value detected in this analysis.

The construction of ΔcpeA and complemented strains were described previously (Pongpom et al., 2013). To generate the Δhsp30 mutant, a similar targeted gene deletion approach was performed. Briefly, the deletion construct, containing the Aspergillus nidulans pyrG selectable marker flanked by 1.5-kb sequences of 5’ upstream and 3’ downstream regions of the hsp30 gene (PMAA_014600), was transformed into the T. marneffei ΔligD, pyrG- strain. The uracil prototroph was selected and the hsp30 gene deletion was confirmed.

The human monocytic cell line THP-1 (ATCC TIB-02) was maintained in RPMI-1640 medium containing 10% (v/v) fetal bovine serum (Gibco; Life Technologies, Carlsbad, CA, USA), 100 U penicillin, and 100 mg/ml streptomycin at 37°C with 5% CO₂. One million cells/well were seeded into a 6-well plate and 100 ng of phorbol myristic acetate (PMA, Sigma, St. Louis, MO, USA) was added to induce macrophage differentiation for 2 days. The macrophage cells were infected with the conidia of ΔcpeA, wild type (ATCC18224) and cpeA complemented strains at a 10:1 ratio of conidia to macrophages. After 2-h incubation, the cells were washed three times with sterile PBS and lysed with a triton X-100. The fungal cells were harvested by centrifugation, diluted 100 times, and spread onto a potato dextrose agar (PDA). The fungus was cultured at 28°C for 2 days and the colonies were counted and calculated in colony-forming units (CFU). The percentage of fungus killed during macrophage infection was calculated using the equation 100-[(CFUx100)/500,000], in which five hundred thousand represents the baseline control number of conidia not infected by macrophage.

One thousand conidia harvested from the T. marneffei wild type (ATCC18224) and the Δhsp30 mutant were inoculated onto the surface of Aspergillus nidulans minimal medium (ANM) agar containing various stressors. The stress-generated conditions included 1 M sorbitol for osmotic stress, 1 mM hydrogen peroxide (H₂O₂) for oxidative stress, and 0.2 M sodium chloride (NaCl) for hyper-salinity stress. The fungal strains were incubated at 37°C for 10 days.

In previous studies, we identified seventeen T. marneffei proteins that exhibited a reaction with antibodies present in the sera of talaromycosis patients (Pongpom et al., 2005). Briefly, an antibody screening experiment was conducted on 10,000 plaque-forming units of the cDNA library to obtain the antigenic protein-encoding genes. E. coli Y1090 strain was infected with the bacteriophage lambda library and protein expression was induced. The immunoreaction was performed using purified IgG from sera infected with T. marneffei as a probe to detect the antigenic proteins. Chemiluminescent detected distinct positive signals that could be repeated in secondary and tertiary screenings ( Figure 1 ) (Pongpom, 2004). Usually, the purified clones were obtained after the second round of immunoreaction. Seventeen positive plaques were purified, and the cDNA inserts were sequenced. Three antigenic protein-encoding genes, cpeA, MPLP6, and hsp30 had been previously characterized. They encoded for the catalase-peroxidase (CpeA), MP1-like protein 6 (Mplp6p), and Hsp30 respectively, and they were proven to contain immunogenic properties (Pongpom et al., 2005; Vanittanakom et al., 2009; Pongpom and Vanittanakom, 2011; Pongpom et al., 2013). The remaining fourteen clones in the antigenic protein dataset was subjected to bioinformatic analyses in this current study. The complete sequencing data were deposited at GenBank and are publicly available ( Supplemental Data S1 ; See materials and methods).

The DNA sequences of 14 clones with antigenic properties were subjected to gene identification using the BLAST nucleotide search against the T. marneffei strain ATCC 18224. Since the functions of these genes were unknown in T. marneffei, we searched for experimentally verified homologous proteins in model yeasts (Saccharomyces cerevisiae and Shizosaccharomyces pombe) and a model filamentous fungus (Neurospora crassa) using BLAST and Uniprot. Results were summarized in Table 1 and Supplemental Tables S2–S5 . The following genes were discovered from our original screening result: glutathione peroxidase, NADH-ubiquinone oxidoreductase, 60S ribosomal protein, and stearic desaturase. The others were novel genes whose characteristics have never been previously described ( Table 1 ).

To explore protein functional activities, we utilized ScanProsite and SMART (Simple Modular Architecture Research Tool) tools for functional domain and critical region searches (Schultz et al., 1998; de Castro et al., 2006). Putative protein localization was predicted using Uniprot program. If subcellular compartment prediction for T. marneffei proteins was not annotated in Uniprot, we used the protein homologs for the search instead. The web-based tool String was used to search for protein-protein interaction networks and functional enrichment. Lastly, we used the WoLF PSORT and SignalP to identify amino acid sequences for cellular sorting signals.

The antigenic proteins were predicted to localize in five major compartments ( Supplemental Table S2 ). First, the protein P14 was predicted to localize in the nucleus, due to its C2H2-zinc finger domain that is commonly found in the transcription factor. Second, several clones were predicted to be cytoplasmic proteins, including fructose-1,6-biphosphatase (Fbp1), ribosomal protein RPL20A, glutathione peroxidase (Gpx1), and thiamine synthase Nmt1. The glutathione peroxidase (Gpx1) was predicted to be located both in the cytosol and the mitochondria. Third, organelles-localizing proteins included the endoplasmic reticulum (ER) membrane proteins Pho88 and stearic acid desaturase SdeA, a vesicular membrane protein Mon1, and a putative mitochondrial inner membrane protein ubiquinone oxidoreductase Nuo21.3. Fourth, the Mplp6p and P26 hypothetical proteins contained signal peptides, suggesting that they might be secretory proteins that localize to the cell wall and excrete extracellularly. Thus, we postulated that Mplp6p and P26 were the cell wall or cell membrane-associated proteins. Lastly, several proteins were predicted to localize in the cell membrane since they possessed the detectable transmembrane portion, or the functional domain associated with the membrane. These membrane-localized proteins were Ham13, Mon1, Fus1, Pho88, Marvel and SdeA.

Epitope is the specific region of the antigen where the antibody binds. We sought to determine the peptide regions of identified antigens that were potentially recognized by patient antibodies. Therefore, we screened these proteins for putative B-cell epitopes using the bioinformatics analysis approach. As prediction is preferable when overlapped results are obtained from the use of more than one software program (Faria et al., 2011), we mapped putative epitope regions using three different bioinformatic tools (Bepipred, ABCpred, and SVMTriP). All epitope regions obtained from each program and overlapped sequences are presented in Supplemental Data S2 . All antigenic proteins contained the overlapped epitope regions. The top three antigenic proteins that contained the highest number of predicted epitope peptides are as follows: P26 (Epitope = 95.7%), NADH-ubiquinone oxidoreductase Nuo21.3 (Epitope = 81.8%), and Nbr1 (Epitope = 73.7%) ( Figure 2 ). These proteins containing high percentages of B-cell epitopes served as candidate antigens for further investigation.

Changes in gene expression patterns can reflect a change of biological activity or vice versa. To better understand the function of identified genes, we assessed the gene expression profiles of all antigenic encoding genes. Since dimorphism is an essentially biological process of T. marneffei, required for the virulence of this fungal pathogen. We reasoned that the identification of phase-specific proteins would be beneficial for developing detection tools in both clinical and research settings. Our antigenic protein screening experiment was performed with T. marneffei growing in only the yeast phase, and therefore we wondered if the expression levels of these genes would change with T. marneffei in different morphologies. To investigate the gene expression profiles of identified genes, conidia from the T. marneffei were either directly harvested or inoculated into Sabouraud’s dextrose broth and grown for 3 days at 25°C or 37°C to induce either the mold or yeast phase, respectively. RNA was collected from these samples, and quantitative real-time PCR was performed to assess transcript levels. A heat map was generated using log2 values of gene expression data (2^-ΔCt), so that we could compare both (i) gene expression fold-change and (ii) transcript abundance across genes and phases ( Figure 3 ; Supplemental Table S6 ). The color scheme assigned green to genes with the highest abundance and red to genes with the lowest expression. In addition, hierarchical analysis was performed to group genes with similar expression profiles.

As shown in Figure 3 , the genes that exhibited yeast-phase-specific expression were MPLP6 and Fus1. As a control, the MPLP6 transcript was not detected in either mold or conidia, consistent with previous studies (Pongpom and Vanittanakom, 2011). The Fus1 transcript was significantly upregulated by 9-fold in the yeast phase and expressed at very low levels in the mold and conidial phases ( Figures 3 , 4B ). Furthermore, the Nmt1 gene was differentially upregulated in the yeast phase, exhibiting almost a 90-fold increase in the yeast sample ( Figures 3 , 5A ). Notably, the Nmt1 transcript in the yeast phase was one of the highest abundant transcripts. We did not detect any genes that were exclusively expressed in the mold phase. However, RPL20A ( Figures 3 , 5B ) and P26 ( Figures 3 , 6B ) genes showed the highest fold change increase in the mold form, upregulating by 23- and 123-fold, respectively. We noted that the RPL20A and P26 genes were partially upregulated in the yeast phase ( Figures 3 , 5B , 6B ). Although none of the identified genes were specifically expressed in conidia, we found that Marvel ( Figures 3 , 7B ) and hsp30 ( Figures 3 , 5C ) genes were highly expressed in conidia. These two genes were also genes that exhibited the highest fold change in each phase. The Marvel gene was upregulated by 900-fold and 152-fold, and the hsp30 gene was upregulated by 400-fold and 96-fold in the conidia and yeast forms, respectively. Lastly, our gene expression analysis revealed that RPL20A and Nuo21.3 showed the highest transcript abundance among identified genes ( Figure 3 ).

For those reported genes, our current gene expression profile was consistent with previously published data. For example, expression levels of hsp30 ( Figure 5C ), Nuo21.3 ( Figure 5D ), and fbp1 ( Figure 5E ) genes showed high induction in the conidia and yeast phases (Pongpom and Vanittanakom, 2016). Only the gpx1 gene showed a different expression profile compared to previous reports, which could be attributed to variations in media.

The membrane trafficking system is crucial for maintaining proper cargo distribution and compartmentalization within cells, as well as for facilitating autophagy, a cellular degradation process (Yang and Rosenwald, 2014). In autophagy, a double-membrane structure is formed and then fused with the lysosome or vacuole for hydrolysis and recycling of building blocks. We noticed that the majority of identified antigenic proteins (53%) were predicted to localize in the cell membrane, membrane-bound organelles, and extracellular compartment ( Figure 8A ; Supplemental Table S2 ). Also, some proteins had an overlapping role in autophagy ( Table 1 ). To better understand the function of the antigenic proteins, Gene Ontology (GO) enrichment and putative biological functions of all identified genes were investigated using the DAVID server. Strikingly, there were many antigenic proteins that showed enrichment in membrane-related functions ( Figure 8B ; Table 1 ; Supplemental Tables S3 , S5 ). Thus, we hypothesized that these proteins function in the membrane trafficking pathway and play a role in antigen presentation during T. marneffei infection inside the macrophages. The relevance of antigenic proteins or their homologous proteins to the membrane trafficking system was highlighted, as described below.

There are several identified proteins that contain functional domains or regions related to membrane trafficking functions. However, the homologs of these proteins have not been experimentally characterized in any fungal model species, thus the actual functions have not been validated. First, the MARVEL-domain containing protein (TmMarvel) was found in our screening result. MARVEL (MAL and related proteins for vesicle formation and membrane link) is a conserved domain that is comprised of four transmembrane helices (Sánchez-Pulido et al., 2002). The function of this domain is to mediate protein-lipid interaction, which is required for the localization or formation of diverse specialized membrane subdomains (Yaffe et al., 2012; Rubio-Ramos et al., 2021). The TmMarvel-encoding gene (PMAA_006490) gives two different transcripts, the first one encoding the 173-amino acid protein (TmMarvelA) with four transmembrane helices ( Figure 7A ), and the other encoding 125-amino acid protein (TmMarvelB) with only three transmembrane helices (data not shown). As mentioned in the previous section, gene expression profiling revealed that the expression of the TmMarvel encoding gene was drastically upregulated in the conidia and yeast phases, increasing by 900-fold and 152-fold, respectively ( Figure 7B ). Collectively, TmMarvel was likely upregulated in conidia and yeast to mediate specific membrane apposition events.

Second, the antigenic clone P26 contained the signal peptide, which is the short N-terminal amino acid sequence that targets proteins to the secretory pathway in eukaryotes. As shown in Figure 6A , the P26 protein contains the signal peptide with good predictive value (signal probability, 0.9998; SignalP 6.0) (Teufel et al., 2022). The protein localization was predicted using WoLF PSORT (Horton et al., 2007), and the result indicated that P26 protein contained the ER membrane retention signals (XXRR-like motif), RSIL, in the N-terminus ( Figure 6A ), and it might localize extracellularly (predictive values; 55.6%: extracellular, including cell wall, 22.2%: mitochondrial, 11.1%: nuclear, and 11.1%: cytoplasmic). Gene expression analysis revealed that P26 encoding gene was highly upregulated in both the mold and yeast phases. Its expression was elevated by 123- and 68-fold in the mold and yeast phases, respectively ( Figure 6B ). This result suggests that the P26 protein likely enters the secretory pathway, necessary for both mold and yeast morphology.

To provide proof-of-concept that information from our antigenic-protein profiling could lead to an understanding of the interplay between the host and pathogen interaction, we sought to characterize the functions of these antigenic proteins under physiological conditions. First, we turned our attention to the catalase-peroxidase encoding gene, cpeA. The immunogenic property of the CpeA protein was confirmed, and the ΔcpeA mutant was previously constructed (Pongpom et al., 2005; Pongpom et al., 2013). The ΔcpeA mutant showed dramatic growth reduction when the cells were exposed to hydrogen peroxide, indicating its role in the oxidative stress response (Pongpom et al., 2013). However, the involvement of the CpeA protein in the interaction with host cells during fungal invasion and infection remained unknown. In this study, the human cell line THP-1, differentiated into macrophage-like cells by PMA treatment, was infected with T. marneffei conidia at MOI 10 and assessed the percentage of fungal strains killed by the macrophages. As shown in Figure 13A , the null mutant ΔcpeA was 15.2% more susceptible to macrophage killing than the wild type and complemented strains, suggesting that this protein could be involved in fungal survival during macrophage engulfment.

Second, we focused on the hsp30 gene as it showed the highest levels of gene upregulation during the pathogenic yeast phase ( Figures 3 , 5C ) and the Hsp30 protein has previously been validated to be immunogenic (Vanittanakom et al., 2009). The hsp30 gene was subjected to targeted gene deletion, and phenotypic analysis was performed in a null deletion mutant (Δhsp30). We examined the role of hsp30 during conidiation, germination, and yeast-to-mold phase transition. We found that the Δhsp30 mutant exhibited a normal phenotype. Also, the Δhsp30 mutant was not sensitive to the anti-fungal drug caspofungin (data not shown). Interestingly, the Δhsp30 mutant displayed growth sensitivity to various stressors, and this phenotype was detected only under 37°C-induced yeast morphology ( Figure 13B ). At 37°C, the Δhsp30 mutant was sensitive to NaCl (0.2 M), sorbitol (1 M), and hydrogen peroxide (1 mM), suggesting the role of hsp30 in high salt stress, osmotic stress, and oxidative stress, respectively. Our results demonstrated that hsp30 functions prominently under elevated temperature, consistent with its highly upregulated levels during the yeast phase where the cells encounter an increased temperature of 37°C.

Together, our phenotypic characterization of the null mutants indicated that both immunogenic proteins, CpeA and Hsp30, contribute to T. marneffei stress adaptation under the yeast physiological conditions.