Characteristics of cancer mycobiome in patients with acral melanoma

Abstract

Background

Intratumoral fungi have recently been implicated in cancer initiation and progression, with potential as biomarkers for predicting clinical outcomes and treatment response in patients with cancer. However, their role in acral melanoma (AM) has not been previously explored.

Methods

We characterized the mycobiome in AM tumor tissues and adjacent non-tumor tissues. Differences in fungal communities between the two tissues, as well as the prognostic and diagnostic potential of intratumoral fungi, and their associations with the tumor microenvironment and clinicopathologic features, were evaluated through bioinformatics and biostatistical analyses.

Results

Although some intratumoral fungi originated from adjacent tissues, AM tumors exhibited a distinct fungal composition characterized by altered species richness, community structure, and an increased Ascomycota-to-Basidiomycota ratio. Several fungal taxa were identified as potential diagnostic and prognostic biomarkers and showed significant correlations with clinical parameters and immune infiltration. Specifically, the CD68-high samples harbored greater fungal diversity and higher relative abundance of Endocarpon compared with CD68-low samples. Furthermore, fungi–bacteria interactions were characterized by significant negative correlations between their diversity, while positive interspecies interactions dominated the network.

Conclusions

These findings underscore the potential role of the cancer mycobiome in AM and offer new insights into the tumor microenvironment and its implications for cancer prevention and therapy.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Many studies have confirmed the presence of microorganisms within tumors, adding significant complexity to the tumor microenvironment. However, while much research has focused on intratumoral bacteria, studies on intratumoral fungi remain limited.

WHAT THIS STUDY ADDS

• Our study confirms the presence of fungi in acral melanoma (AM) and reveals significant differences in fungal composition between tumor tissues and adjacent non-tumor tissues. The results demonstrate a notable association between intratumoral fungi and patients’ clinicopathological features as well as immune infiltration. Moreover, our findings highlight the potential of certain fungi as prognostic and diagnostic biomarkers for patients with AM.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• Our study provides a new perspective for the complex tumor microenvironment of AM, broadens the understanding of the cancer microbiome, and offers novel insights for cancer diagnosis and intervention.

Introduction

The gut microbiome has been established as a major player in human health and been extensively studied. It has been proved to predict and influence response to treatment in patients with cancer by modulating the immune system.1,3 With advancements in microbial DNA sequencing technologies, increasing attention has been directed toward the cancer microbiome. Evidence has already confirmed the presence of microbes, including bacteria and fungi, within tumors,4,7 adding significant complexity to the tumor microenvironment (TME).

While much of the focus has been on intratumoral bacteria, research on intratumoral fungi remains limited. Although bacterial biomass in tumors is low, the abundance of fungi is even much lower. In primary tumors from The Cancer Genome Atlas database, bacteria accounted for 96% of the microbial population, while fungi comprised only 4%.⁷ In fact, only one fungal cell can be detected per 10,000 tumor cells.⁶ As a result, the study of the cancer mycobiome requires highly sensitive detection methods due to the low fungal burden.

Recent studies have demonstrated the role of fungi in tumor development. For instance, Malassezia, abundant in pancreatic ductal adenocarcinoma, accelerates tumor formation by activating complement C3 through mannose-binding lectin.⁵ Aspergillus sydowii promotes lung cancer progression via IL-1β-mediated expansion and activation of myeloid-derived suppressor cells.⁸ Fungi also contribute to immune tolerance, promoting tumor progression by influencing host innate immunity.⁹ Beyond promoting tumor development, fungi facilitate metastasis by weakening cell adhesion or disrupting epithelial barriers.^{10 11} Furthermore, certain fungi have shown potential as biomarkers for predicting clinical outcomes in patients with cancer.^{6 7 12 13} As a component of the TME, intratumoral fungi play a notable role in tumor promotion, metastasis, and response to treatment.¹⁴ Uncovering the composition of the intratumoral fungal community could enhance our understanding of the TME and reveal novel therapeutic targets for cancer treatment.

To date, no studies have investigated the cancer mycobiome in acral melanoma (AM). In this study, we explored the characteristics of the mycobiome in AM. Our study showed that fungi served as biomarkers for AM prognosis and diagnosis. Our findings suggest that specific fungal species within AM tumors may represent potential targets for cancer prevention and therapy.

Methods

Sample collection

This study included patients newly diagnosed with AM who underwent surgery at Nanjing Drum Tower Hospital between January 2013 and March 2022. Patients with recurrent AM, other malignancies, neoadjuvant treatments prior to surgery, or infectious diseases such as tuberculosis, hepatitis B, or syphilis were excluded. A total of 89 patients were enrolled, from whom clinicopathological data and formalin-fixed paraffin-embedded (FFPE) tissue samples were collected. Additionally, 35 matched adjacent non-tumor tissues and 12 blank paraffin samples were included. Overall survival (OS) was followed for all patients.

Immunohistochemistry staining, multiplex immunofluorescence staining, and fluorescence in situ hybridization staining

Pathologists performed immunohistochemical (IHC) staining on tumor sections to assess molecular markers of T cells (CD3, CD4, CD8, and Foxp3) and macrophages (CD68). Five representative, non-adjacent, non-overlapping areas were randomly selected at 200× magnification to count tumor-infiltrating immune cells. Cell counts were averaged in units of cells per square millimeter. The positive rate of immune cells was calculated as the ratio of immune cells to the total cell count. Two independent pathologists from Drum Tower Hospital, blinded to the sample groupings and clinical data, evaluated the IHC staining and quantified the number of immune cells per high-power field.

For the multiplex immunofluorescence (mIF) analysis, we used a PANO 5-plex IHC kit based on Tyramide Signal Amplification following the manufacturer’s protocol (Panovue, China). AM tumor sections with 4 µm thickness were cut from FFPE tumor specimens for mIF staining. Acidic and alkaline antigen retrieval buffers (Panovue, China) were employed as needed. The antibodies were as follows: anti-1,3-β-glucan (1:50 dilution, Abcam, UK, Cat. no. 233743) Opal650, anti-Aspergillus (1:100, Abcam, UK, Cat. no. 20419) Opal650, anti-CD68 (1:2000 dilution, Proteintech, USA, Cat. no. 66231–2-Ig) Opal570, and anti-SOX10 (1:400 dilution, Proteintech, USA, Cat. no. 66786–1-Ig) Opal520. Nuclei were stained with DAPI (1:100 dilution, Beyotime Biotechnology, China, Cat. no. C10002). For fluorescence in situ hybridization (FISH), tumor and adjacent non-tumor sections were stained using a Cy3-labeled pan-fungal probe D-223 (5'-CCACCCACTTAGAGCTGC-3') and a fluorescein isothiocyanate (FITC)-labeled Aspergillus-specific probe (5'-TGACGGCCCGTTCCAG-3'). Multispectral images for each stained slide were scanned using the PanoVIEW VS200 System (Panovue, China) and acquired using OlyVIA software (Olympus, Japan).

DNA extraction and internal transcribed spacer sequencing

DNA was extracted from these samples, and DNA purity and concentration were tested using a Nanodrop One (Thermo Fisher Scientific, MA, USA). Genomic DNA was then amplified by PCR, according to internal transcribed spacer 2 (ITS2) primers. The quality of the PCR product was subsequently assessed. To monitor and exclude potential contamination, negative controls were included throughout the experimental workflow, comprizing six DNA extraction blanks and six PCR amplification blanks. No contamination was detected in these controls. Using the ALFA-SEQ DNA Library Prep Kit, library construction was performed, and library fragment sizes were assessed using the Qsep400 high-throughput nucleic acid protein analysis system (Hangzhou Houze Biotechnology, Hangzhou, China), and library concentration was measured using Qubit 4.0 (Thermo Fisher Scientific, Waltham, USA). PE250 sequencing of the constructed amplicon libraries was performed based on the Illumina platform (Guangdong Magigene Biotechnology, Guangzhou, China).

Sequencing data processing

Primers were removed from the raw sequencing data using Cutadapt software (https://github.com/marcelm/cutadapt/), facilitating quality control and filtering. Double-ended reads were then merged based on their overlap to generate the original spliced sequences (raw tags). These raw tags were subsequently filtered using fastp V.0.14.1, https://github.com/OpenGene/fastp), an ultra-fast all-in-one FASTQ preprocessor, to produce clean tags.

Species annotation was performed by aligning the representative sequences of each amplicon sequence variant (ASV) against the SILVA database using usearch-sintax/blast. The newly implemented database (ITS/UNITE V.10.0) was used for fungal taxonomic annotation of ASVs. Sequences identified as chloroplasts or mitochondria, as well as those that could not be classified at the kingdom level, were removed. This process yielded the final set of valid tag sequences and a comprehensive ASV table containing the representative sequence taxonomy for each sample.

Analysis of ITS sequencing data

Prior to downstream analyses, the ASV table was decontaminated using the SCRuB algorithm¹⁵ based on the tissue-free paraffin samples, to effectively eliminate potential contaminants. SourceTracker¹⁶ was used to analyze the origin and proportion of cancer fungi. Alpha diversity indices (Richness, Chao1, Shannon diversity, Simpson diversity, abundance-based coverage estimation (ACE), Phylogenetic diversity) were calculated using the R package “vegan” based on the ASV table (https://cran.r-project.org/web/packages/vegan/index.html). Beta diversity was assessed through non-metric multidimensional scaling (NMDS) using the vegan package based on Bray-Curtis distance. Microbial differences between groups were conducted using the linear discriminant analysis (LDA) effect size (LEfSe) algorithm¹⁷ with an LDA score threshold of three, implemented in the R package “microbiomeMarker” and “microeco” package¹⁸ (http://huttenhower.sph.harvard.edu/lefse/). Benjamini-Hochberg method was performed to correct the p value (false discovery rate, FDR <0.05).

Bacterial–fungal interactions with relative abundances each of the top 60% were analyzed using the “Hmisc” package and those correlations (Spearman’s correlation coefficients: r >0.6 or < −0.6, p<0.05) were visualized using “Gephi”. Spearman correlation analysis was used to analyze the relationship between fungal and bacterial diversity indices in paired samples.

Survival analyses were conducted using the Kaplan-Meier method and compared via log-rank tests. Variable selection for multivariable Cox regression was performed using Least Absolute Shrinkage and Selection Operator (LASSO) regression. The diagnostic performance of cancer fungi was analyzed using receiver operating characteristic (ROC) curves and the area under the curves (AUC).

Given the non-normal distribution of microbial data, the Mann-Whitney U test (Wilcoxon rank-sum test) was primarily used for between-group comparisons. For datasets meeting normality assumptions, Student’s t-test was applied. Statistical significance was defined as a two-sided p value ≤0.05.

Results

ITS sequencing reveals distinct microbial profiles in AM tissues and adjacent non-tumor tissues

Of the 136 samples initially collected, 21 failed to generate sequencing results, leaving 115 samples with valid data (online supplemental figure S1). The rarefaction curve plateaued, indicating that the sequencing depth was sufficient for all samples (online supplemental figure S2). SourceTracker analysis identified the origin of fungi in AM tumor tissues (figure 1A), showing that adjacent non-cancerous tissue was a significant source, contributing 61.9%±0.61% of the fungal population. A considerable portion of the fungal population (38.0%±0.6%) was of unknown origin, with minimal contamination from laboratory sources (0.08%±0.09%), demonstrating the effectiveness of the decontamination process (online supplemental table S1).

Figure 1. Comparative analysis of fungal profiles in AM tumors and adjacent non-tumor tissues. (A) SourceTracker analysis of intratumoral fungi, showing the origins and proportions of fungal communities in 85 patients with AM. (B) Venn diagram of shared and unique ASVs between tumor and adjacent tissues. (C–E) Taxonomic composition of fungal communities at the phylum, family, and species levels. (F–H) Boxplots of alpha diversity indices (Richness, Simpson, and Shannon) comparing tumor and non-tumor tissues. (I) Comparison of the Ascomycota/Basidiomycota ratio between groups. (J) NMDS analysis based on Bray-Curtis distance. (K) LEfSe evolutionary branch diagram. Circles from the inside out indicate phylogenetic levels from phylum to species. (L) LEfSe analysis identifying taxa with significantly different abundances (LDA score >3). Taxonomic prefixes: c, class; o, order; f, family; g, genus; s, species. Italic font indicates genus/species taxonomic ranks. AM, acral melanoma; ASV, amplicon sequence variant; LDA, linear discriminant analysis; LEfSe, linear discriminant analysis effect size; NMDS, non-metric multidimensional scaling.

In total, 3,268 unique ASVs were identified in tumor tissues and 1,020 in adjacent non-cancerous tissues, with 465 ASVs shared between both tissue types (figure 1B). At the phylum level, Ascomycota and Basidiomycota were the dominant phyla in both tumor and adjacent tissues (figure 1C). At the family level, Aspergillaceae was the most abundant family in both tissue types, with Cordycipitaceae and Cladosporiaceae notably enriched in tumor tissues (figure 1D). At the species level, Aspergillus clavatophorus was predominant in both tissues, while Simplicillus lamellicola (S. lamellicola), Aspergillus glabripes (A. glabripes), Cladosporium cladosporioides, Aspergillus reticulatus, Aspergillus halophilicus (A. halophilicus), Aspergillus leucocarpus, and Trichophyton rubrum (T. rubrum) were enriched in tumor tissues (figure 1E). Stacked bar plots at the phylum, family, and species levels suggested broadly similar fungal compositions between tumor and adjacent tissues, whereas heatmaps of the top 20 most abundant fungal genera revealed subtle compositional differences (online supplemental figure S3).

Alpha diversity analysis showed a significant difference in Simpson diversity (p=0.0136), but not in other diversity indices (figure 1F–H, online supplemental table S2). However, the ratio of Ascomycota to Basidiomycota (A/B ratio) was significantly higher in AM tumor tissues compared with adjacent tissues (Mann-Whitney U test, p=0.045, figure 1I, online supplemental table S3), suggesting dysregulation of the fungal community in tumor tissues.

Beta diversity analysis using NMDS demonstrated a significant difference in species composition between tumor and adjacent non-tumor tissues (p=0.001), with tumor tissues showing higher microbial community similarity (figure 1J). LEfSe analysis, corrected using the Benjamini-Hochberg method, identified multiple taxa significantly enriched in AM tumors, including but not limited to the genera‌ Aspergillus, Simplicillium, Trichophyton, and the species S. lamellicola, Aspergillus glabripes (A. glabripes), A. halophilicus, and T. rubrum (figure 1L). Wilcoxon test supported these findings (online supplemental figure S4).

These findings were further validated by FISH and mIF, which confirmed the presence and enrichment of Aspergillus in AM tumor tissues (figure 2A–C).

Figure 2. Visualization of fungal presence in AM tumor and adjacent tissues. (A) FISH detection of 28S rRNA (D223 probe, Cy3, red), Aspergillus (FITC, green), and DAPI (blue) in tumor and adjacent tissue sections. (B) Fluorescence signal quantification of D223 probe and Aspergillus probe in tumor tissue sections and adjacent paracancerous tissue sections. Both tissue types were randomly selected from 85 patients for staining, with 5 tumor samples and 4 adjacent paracancerous samples analyzed. (C) Immunofluorescence staining for Aspergillus (red), SOX10 (green), and DAPI (blue) in AM tumor tissue and paracancerous tissue sections. AM, acral melanoma; DAPI, 4′,6-diamidino-2-phenylindole;FISH, fluorescence in situ hybridization; FITC,fluorescein isothiocyanate. * p ＜0.05.

Potential of intratumoral fungi as biomarkers of AM

To investigate the prognostic value of intratumoral fungi, we first examined the relationship between microbial diversity and clinical outcomes in patients with AM. Patients were divided into two groups based on OS time, and alpha and beta diversity indices were assessed. However, no significant differences were observed between patients with long and short OS (onlinesupplemental tables S2 S4). While A/B ratios significantly predicted OS (figure 3D, online supplemental table S3). Clinicopathologic characteristics of the cohort are presented in table 1.

Figure 3. Identification of prognostically relevant fungi in AM tumors. (A–D) Kaplan-Meier curves for fungal genera/species (Endocarpon, Irpex lacteus, Fusarium sambucinum) and the A/B ratio. (E) LASSO coefficient profiles of 20 fungal taxa associated with prognosis. (F) Cross-validation plot for optimal lambda selection. Three variables were retained for multivariable Cox modeling. A/B ratio, Ascomycota to Basidiomycota ratio; AM, acral melanoma; LASSO, Least Absolute Shrinkage and Selection Operator.

Table 1. Clinicopathological information of patients with AM enrolled in ITS sequencing.

Characteristics	N (%)
Age
>65	43 (50.6)
≤65	42 (49.4)
Sex
Female	44 (51.8)
Male	41 (48.2)
Stage
0	3 (3.5)
I	18 (21.2)
II	38 (44.7)
III	25 (29.4)
Unknown*	1 (1.2)
Breslow thickness
Tis	3 (3.5)
≤1.0 mm	11 (12.9)
1.01–2.0 mm	16 (18.8)
2.01–4.0 mm	20 (23.5)
>4.0 mm	35 (41.2)
Primary site
Hand	11 (12.9)
Foot	74 (87.1)
Ulceration
With	48 (56.5)
Without	37 (43.5)
RLN metastasis
With	21 (24.7)
Without	63 (74.1)
Unknown*	1 (1.2)
LSAT
Yes	19 (22.4)
No	66 (77.6)
Survival
Alive	54 (63.5)
Dead	31 (36.5)

^*One patient with T4 lesion refused sentinel lymph node biopsy.

AM, acral melanoma; ITS, internal transcribed spacer; LSAT, lack of suggested adjuvant treatment; RLN, regional lymph node metastasis.

We next performed log-rank tests at the genus and species levels. Several fungal taxa were significantly associated with OS (p≤0.05), including Endocarpon, Irpex lacteus (I. lacteus), and Fusarium sambucinum (F. sambucinum) (figure 3A–C), with additional taxa listed in online supplemental table S5. Notably, patients with low relative abundance of Endocarpon or high relative abundance of taxa such as Apiotrichum, Blastobotrys, Cyberlindnera, Irpex, Mrakia, Pronectria, Pichia, Sordaria, Toxicocladosporium, Trichomerium, Apiotrichum scarabaeorum, Blastobotrys persicus, F. sambucinum, I. lacteus, Pronectria loweniae, and Sordaria fimicola exhibited poorer OS (online supplemental table S5). While Naganishia, Phialemonium, and A. glabripes showed a trend toward association with OS, these did not reach statistical significance.

To reduce model overfitting, we applied LASSO regression to these 20 taxa, identifying three key variables for inclusion in multivariable COX regression (figure 3E,F). After adjusting for age, clinical stage, and Breslow thickness (online supplemental table S6), multivariable Cox regression identified Endocarpon (HR 0.16, 95% CI: 0.05 to 0.46, p<0.001), I. lacteus (HR 4.75, 95% CI: 1.70 to 13.26, p=0.003), and A/B ratio (HR 4.57, 95% CI: 1.27 to 16.50, p=0.020) as independent prognostic factors for OS (table 2).

Table 2. Multivariable Cox analysis of intratumoral fungi associated with overall survival.

Factor	HR (95% CI)	P value
Age (>65 vs ≤65)	6.08 (2.21 to 16.74)	<0.001
Stage (III vs 0–II)	5.42 (2.16 to 13.58)	<0.001
Breslow thickness (>2 vs ≤2)	6.95 (1.89 to 25.52)	0.003
Endocarpon (high vs low)	0.16 (0.05 to 0.46)	<0.001
Naganishia (high vs low)	0.75 (0.24 to 2.37)	0.628
Irpex lacteus (high vs low)	4.75 (1.70 to 13.26)	0.003
A/B ratio (high vs low)	4.57 (1.27 to 16.50)	0.020

AB ratio, Ascomycota to Basidiomycota ratio; ASV, amplicon sequence variant.

In addition, ROC curve analysis was performed to evaluate the diagnostic potential of A/B ratio and fungi that were significantly enriched in tumor tissue. Fungi with AUC >0.8 and p<0.05 were shown in figure 4. The results indicated that the orders Onygenales (AUC=0.841, 95% CI 0.751 to 0.931, p<0.001) and Hypocreales (AUC=0.813, 95% CI 0.714 to 0.912, p<0.001), the family Arthrodermataceae (AUC=0.854, 95% CI 0.764 to 0.944, p<0.001), the genera Trichophyton (AUC=0.854, 95% CI 0.764 to 0.944, p<0.001) and Simplicillium (AUC=0.906, 95% CI 0.833 to 0.98, p<0.001), and the species T. rubrum (AUC=0.862, 95% CI 0.779 to 0.945, p<0.001), A. halophilicus (AUC=0.9, 95% CI 0.842 to 0.959, p<0.001), A. glabripes (AUC=0.857, 95% CI 0.778 to 0.937, p<0.001) and S. lamellicola (AUC=0.906, 95% CI 0.833 to 0.98, p<0.001) demonstrated good diagnostic validity (figure 4). The A/B ratio also exhibited moderate diagnostic value (AUC=0.63, 95% CI 0.514 to 0.747, p<0.001, online supplemental figure S5).

Figure 4. Diagnostic potential of intratumoral fungi in melanoma. ROC curves of fungi with AUC >0.8 at various taxonomic levels: Order—Onygenales, Hypocreales; Family—Arthrodermataceae; Genus—Trichophyton, Simplicillium; Species—Trichophyton rubrum, Aspergillus halophilicus, Aspergillus glabripes, Simplicillium lamellicola. AUC, area under the curve; ROC, receiver operating characteristic.

Associations between intratumoral fungi and clinicopathologic features

Our analysis of fungal diversity and its association with clinicopathological characteristics revealed no significant differences in either alpha or beta diversity across different patient groups (onlinesupplemental tables S2 S4). However, the relative abundance of individual fungal taxa was associated with specific clinicopathological features.

Using the Mann-Whitney U test, we observed that fungal profiles varied by tumor location. Patients with AM on different primary sites exhibited differences in the A/B ratio and in the abundance of Sordariomycetes, Cordycipitaceae, and A. halophilicus (p=0.008, 0.031, 0.047, and 0.044, respectively; onlinesupplemental figure S6ad and table S7). Notably, higher abundances of these taxa were found in tumors located on the feet, while higher A/B ratios were observed in hand lesions. Age also influenced fungal abundance, as older patients exhibited a higher relative abundance of A. glabripes (p=0.005, onlinesupplemental figure S6e table S7). Moreover, A. halophilicus was more abundant in patients with ≥2 mm Breslow thickness and in those with ulceration (p=0.024, p=0.03, respectively, onlinesupplemental S6f, g table S7).

Associations between fungal biomarkers and immune infiltration

Tumor microorganisms are crucial components of the TME. To assess their possible role in immune modulation within the TME, we investigated the relationship between intratumoral fungal diversity and the infiltration of immune cells, including CD3, CD4, CD8, Foxp3, and CD68-positive cells. The analysis was stratified by the median positive rate of immune cell populations.

No significant differences in alpha diversity were observed in the CD4 and Foxp3 groups (online supplemental table S2). However, in the CD68 group, significant differences in alpha diversity were identified. Based on indices such as Richness, Chao1, and Phylogenetic diversity, the high CD68 group exhibited significantly higher fungal diversity than the low CD68 group (figure 5A–C, online supplemental table S2). Similarly, higher Simpson diversity was observed in the high CD3 and high CD8 groups (online supplemental table S2). In contrast, no significant differences in beta diversity were noted across any immune cell groups (online supplemental table S4).

Figure 5. Associations between tumor-associated macrophages in the TME and intratumoral fungi. (A–C) Correlation of macrophage infiltration with fungal diversity indices (Richness, Chao1, Phylogenetic Diversity). (D) Immunofluorescence staining for SOX10 (green), CD68 (yellow), and β-glucan (red); arrows indicate CD68 colocalization with β-glucan. (E) FISH analysis of SOX10 (green), CD68 (yellow), and Aspergillus (red); White arrows indicate Aspergillus signals. FISH, fluorescence in situ hybridization; TME, tumor microenvironment.

We further analyzed individual fungal taxa in relation to immune cell infiltration (onlinesupplemental figure S7 table S8). The results indicated that the A/B ratio was higher in both the low CD3 and low Foxp3 groups (p=0.014, p=0.011, online supplemental figure S7a, c), suggesting a potential association between this ratio and reduced T-cell or regulatory T-cell presence, while the genus Naganishia was more abundant in the high CD3 group (p=0.024, online supplemental figure S7b). In the CD4 group, negative associations were found in abundances of Cordycipitaceae, Simplicillium, and S. lamellicola (p=0.021, p=0.036, p=0.036, online supplemental figure S7d–f). Additionally, the high CD68 group displayed significantly higher abundances of Sordariomycetes, F. sambucinum, Endocarpon compared with the low CD68 group (p=0.038, p=0.032, p=0.024, online supplemental figure S7g–i), reinforcing the link between fungal presence and macrophage infiltration.

Given that macrophages are the predominant immune cell type in the TME of AM, as previously reported,¹⁹ and that our findings further confirm significant associations between macrophage infiltration and microbial diversity, as well as specific fungal taxa, we pursued mIF staining to localize fungi and macrophages within the AM microenvironment. The results revealed punctate fungal signals in the perinuclear region and extracellular space of tumor cells, with β-glucans mainly colocalizing with CD68 on the surface of macrophages (figure 5D). These indicated that fungi were recognized and phagocytosed by macrophages. In contrast, mIF staining of Aspergillus, a tumor-enriched genus, revealed its localization primarily to tumor cells and extracellular regions, with minimal colocalization with macrophages (figure 5E).

Fungi–bacteria interactions

We previously investigated the bacterial component of AM (unpublished). Available 16S rRNA sequencing data were obtained from 89 tumor tissues and 12 matched paracancerous tissues, with microbial contaminants removed using the SCRuB algorithm (online supplemental figure S7). To further elucidate the complexity of the intratumoral microbial microenvironment, we explored associations between fungal and bacterial communities.

Our analysis of six diversity indices revealed significant negative correlations between fungal and bacterial diversity, specifically in ACE, Chao1, Phylogenetic diversity, and Richness indices (p<0.05, figure 6A–F). This finding indicates that higher fungal diversity may be associated with lower bacterial diversity within AM tumor tissues.

Figure 6. The associations between fungi and bacteria. (A–F) Correlation analysis of fungal and bacterial diversity indices (ACE, Chao1, Phylogenetic diversity, Richness, Shannon diversity, Simpson diversity) in AM tumor tissues. (G) Bacteria–fungi interaction network diagram of adjacent non-tumor tissues. (H) Bacteria–fungi interaction network diagram of AM tumor tissues. Each node in the network represents an ASV, with node colors indicating the kingdom affiliation of the ASVs. The lines connecting the nodes (edges) signify significant interactions between ASVs, defined by a correlation coefficient greater than 0.6 or less than −0.6 (p<0.05). Blue lines denote positive interactions, while red lines represent negative interactions. ACE, abundance-based coverage estimation; AM, acral melanoma; ASV, amplicon sequence variant.

Network analysis (figure 6G,H) of ASVs accounting for the top 60% of relative abundance, including both bacteria and fungi, showed that the ecological network of AM tumor tissues (comprizing 155 nodes and 970 edges) was less modular compared with paracancerous tissues (comprizing 462 nodes and 3,581 edges). This difference may suggest lower system stability and resilience in the TME. In AM tumor tissues, positive interactions dominated the network, accounting for 99.9% of the interactions, while negative interactions constituted only 0.1%, and were identified as bacteria–fungi interactions (figure 6H, online supplemental table S9). Similarly, in paracancerous tissues, positive interactions occupied an absolute predominance (99.6%) of the network, and negative interactions accounted for 0.4% (figure 6G, online supplemental table S10). As for the intrakingdom interactions within AM tumor tissues, bacterial (n=712 edges) interactions were more prevalent compared with fungal interactions (n=250 edges). Furthermore, there were substantial cross-kingdom (fungi–bacteria) microbial interactions (0.8%), although interactions were more common within kingdoms (99.2%).

Taken together, these findings highlight ecological interactions between bacteria and fungi in AM, suggesting that the dynamics of these microbial communities may play a critical role in the development and progression of the disease.

Discussion

The current study provides a detailed characterization of the cancer mycobiomes in AM tumors. Our findings revealed that not all AM tumor samples tested positive for fungal signals. Our analysis identified a state of fungal community dysregulation in AM tumors and revealed significant differences in fungal composition from adjacent tissues. These insights not only contribute to our understanding of the microbial landscape in AM but also highlight the potential for certain fungi to serve as biomarkers for prognosis and diagnosis in patients with AM. Additionally, our findings indicate complex interactions between fungi and bacteria within the TME.

Consistent with previous findings, AM tumor fungi were partially derived from adjacent non-tumor tissue,²⁰ but distinct from it. In addition to significant differences in alpha diversity and beta diversity, we also observed significant variation in A/B ratio between AM tissues compared with adjacent tissues. Previous reports indicated that melanoma tissues typically exhibit the lowest A/B ratio among various solid tumors.⁷ However, our findings revealed that AM tissues possessed a significantly higher A/B ratio than their adjacent non-tumor tissues.

Currently, most studies have focused on the bacterial microbiota and its impact on human health and disease, with limited research on fungi and their associations with human tumors. In this study, through LASSO and multivariable Cox analysis, we identified Endocarpon, I. lacteus, and A/B ratio as independent predictors for OS in patients with AM. I. lacteus, a basidiomycete fungus, has been reported to produce purified polysaccharides with antimembranous glomerulonephritis activity.²¹ However, its role as an intratumoral fungus remains unclear. In this study, high relative abundance of I. lacteus was associated with shorter OS. Conversely, Endocarpon, a globally distributed lichen-forming genus,²² was identified as a positive prognostic factor. Notably, extracts derived from Endocarpon pusillum have demonstrated anti-gastric cancer activity in vitro,²³ suggesting potential therapeutic relevance.

In addition to survival analysis, certain fungal taxa demonstrated diagnostic efficacy through ROC curve analysis. These findings support the clinical utility of these fungi as potential biomarkers and therapeutic targets, highlighting the need for further exploration into the role of the mycobiome in cancer biology and treatment strategies.

We also found significant correlations between specific fungi and various clinicopathological features, including age, primary site of the tumor, ulceration status, and Breslow thickness. Notably, A. halophilicus was more abundant in patients with thick tumors or with ulceration, suggesting a possible link between this fungal species and tumor aggressiveness. The relevance and causality of these findings, however, remain to be established through further research.

Tumor-associated microbes have been reported to influence the tumor immune microenvironment in various ways, and a concept of tumor microbe microenvironment was initially described by Ma et al.¹⁴ Previous studies have analyzed the association of tumor microbes with immune infiltrative characteristics.^{24 25} Our study is the first to investigate the association between the mycobiome and the tumor immune microenvironment in AM. We observed that macrophages in the TME were linked to both fungal diversity and specific fungal abundance. IF staining provided evidence of fungal presence in macrophages, cancer cells, and extracellular regions, aligning with previous observations.⁷ β-glucans, known as major components of microbial cell walls, can bind to macrophage surface receptors, enhancing macrophage phagocytic activity and modulating anti-melanoma immunity.²⁶ These results suggest that fungal colonization may be influenced by microenvironmental factors and indicate a potential regulatory role of fungi in the AM microenvironment. However, the role of intratumoral fungi-derived β-glucans requires further validation in future studies. Among the identified fungi, Aspergillus was significantly enriched in AM tumor tissues, as demonstrated by ITS sequencing and FISH. While Aspergillus species have been implicated in tumor progression,^{8 27} emerging evidence also suggests potential anticancer effects of Aspergillus and its metabolites.28,30 These dual roles underscore the need to further explore Aspergillus in the context of melanoma pathogenesis, treatment response, and therapeutic targeting.

Microorganisms obtain nutrients through mechanisms of competition or cooperation. In our study, the negative relationship observed between multiple diversity indices of bacteria and fungi in AM tumor tissue implies that, collectively, these two kingdoms may be engaged in competitive interactions. However, we also observed frequent positive interactions between specific species, particularly within kingdoms. This suggests that species with certain genetic relationships might not primarily compete from an individual perspective. The formation of modules through close interactions contributes to community stability. Additionally, our network analysis confirmed clear cross-kingdom interactions between tumor fungi and bacteria. In these interkingdom relationships, positive interactions were much more prevalent than negative interactions, a trend also reported in previous studies.^{7 31} The cross-kingdom interactions between fungi and bacteria emphasize the importance of evaluating the tumor microbe microenvironment as an integrated system.

Our study has several limitations. First, while we detected fungal DNA, it remains unclear whether cancer-associated fungi are viable and capable of surviving within the TME. Second, although we took precautions to minimize contamination, such as using tissue-free paraffin and negative controls, the possibility of false-positive results cannot be entirely ruled out. Third, the lack of multisite sampling, aside from adjacent non-tumor tissues, limits our ability to draw definitive conclusions regarding the origins of tumor-associated fungi. Additionally, the lack of an independent external validation cohort constrains the generalizability of our findings.

Despite these limitations, our study broadens the understanding of the cancer microbiome. As the first analysis of cancer mycobiomes in AM tissues, this research provides valuable insights that could inform future studies and advance cancer diagnostics and therapeutics. Overall, our findings emphasize the significance of the microbiota in the TME and its potential as biomarkers for AM.

Conclusion

Our study demonstrates that AM tumor tissues possess distinct microbial compositions compared with adjacent non-tumor tissues, with a portion of tumor fungi deriving from these adjacent areas. By examining the relationships between tumor-associated fungi and clinicopathological features, immune infiltration characteristics within the TME, and clinical outcomes in patients with AM, we underscore the significance of the cancer microbiota as an integral component of the TME, with potential as both prognostic and diagnostic biomarkers. Additionally, this research provides the first insights into bacterial–fungal interactions in AM, contributing to a deeper understanding of tumor-associated microorganisms. In summary, our findings illuminate the characteristics of the cancer mycobiome in AM, paving the way for future studies and clinical applications.

Data availability statement

Data are available upon reasonable request.