Abstract

Background

   Pleomorphic giant cell adenocarcinoma (PGCA) of the prostate is a rare,
   aggressive variant characterized by multinucleated giant cells,
   sarcomatoid features, and resistance to conventional therapies. Despite
   its recognition in the WHO 2016 guidelines, the molecular drivers and
   clinicopathological correlates of PGCA remain poorly characterized.
   This study presents the first integrative clinicogenomic profiling of
   PGCA, revealing a novel prognostic gene signature with direct
   implications for diagnosis and treatment.

Methods

   We conducted comprehensive clinicopathological and genomic analyses of
   a treatment-refractory PGCA case using histology, immunohistochemistry
   (IHC), whole-exome sequencing (WES), clonal evolution modeling, and
   multicohort validation. IHC assessed key prostate cancer markers (AR,
   AMACR, KLK3, PTEN, NKX3-1, VIM), while WES compared somatic alterations
   in PGCA, adjacent adenocarcinoma, and stromal tissue. Public datasets
   (prostate_dkfz_2018, prad_tcga, prad_mcspc_mskcc_2020) were used for
   external validation.

Results

   PGCA displayed profound pleomorphism, necrosis, and complete loss of
   luminal markers (AR/AMACR/KLK3), along with strong vimentin (VIM)
   expression, consistent with epithelial–mesenchymal transition. WES
   revealed PGCA-specific mutations enriched in cell-cycle and
   inflammatory response pathways, distinct from metabolic alterations in
   the adjacent adenocarcinoma. Clonal evolution analysis showed divergent
   progression from a shared ancestral clone. Importantly, mutations in
   ADAMTS7, CDH1, DRD5, MGAT5, and TP53 emerged as a robust five-gene
   signature predictive of biochemical recurrence, metastasis, and poor
   survival, validated across multiple independent cohorts.

Conclusion

   Our study provides the first molecular roadmap of prostatic PGCA to
   date, establishing a novel five-gene prognostic signature and revealing
   fundamental insights into its pathogenesis through divergent evolution
   from conventional adenocarcinoma. These insights offer new
   opportunities for precise diagnosis, prognostic stratification, and
   targeted therapeutic strategies for this lethal prostate cancer
   variant.

   Keywords: prostate cancer genomics, pleomorphic giant cell
   adenocarcinoma, divergent tumor evolution, cell cycle, apoptosis,
   therapeutic resistance biomarkers

Introduction

   Pleomorphic giant cell adenocarcinoma (PGCA) is a rare and highly
   aggressive malignancy that has been identified in various organs,
   including the lung, pancreas, thyroid, bladder, and prostate
   ([53]1–[54]5). In the context of prostate, PGCA is histologically
   characterized by the presence of bizarre multinucleated giant cells,
   marked nuclear atypia, and sarcomatoid differentiation, setting it
   apart from the conventional prostatic adenocarcinoma ([55]6). With the
   increasing number of similar cases reported in prostate, PGCA was
   classified by the World Health Organization (WHO) 2016 as a variant of
   prostatic adenocarcinoma ([56]7). Clinically, prostatic PGCA manifests
   with non-specific symptoms such as urinary obstruction, hematuria, and
   pelvic pain, yet distinguishes itself through rapid progression, early
   metastasis to bones and visceral organs, and a median survival of less
   than 12 months post-diagnosis ([57]8). In comparison, patients with
   conventional prostatic adenocarcinoma harboring high-grade Gleason
   scores ([58]9, [59]10) exhibit a mortality rate of less than 10% within
   4 years post-diagnosis following standard therapeutic interventions
   ([60]9, [61]10).

   Therapeutic management of prostatic PGCA remains a significant
   challenge due to its intrinsic resistance to conventional treatments
   for prostate cancer (PCa), including androgen deprivation therapy
   (ADT), chemotherapy, and radiation ([62]11). Moreover, previous
   treatments may even be responsible for the clonal evolution of PGCA,
   particularly ADT ([63]6). Compounded by the inefficacy of immunotherapy
   in most advanced PCa cases ([64]12–[65]14), no currently therapeutic
   options can bring progression-free survival (PFS) and overall survival
   (OS) benefits. As an exceptionally rare entity, only approximately 50
   cases have been reported in the medical literatures, accounting for
   <0.1% of all prostatic malignancies ([66]15). Given this rarity,
   prostatic PGCA often presents significant treatment challenges,
   exacerbated by its aggressive nature and poor prognosis, underscoring
   the need for more comprehensive data on this rare pathologic entity.

   Histopathological analyses reveal a high mitotic index, extensive
   necrosis, and heterogeneous cellular architecture, features that
   correlate with its aggressive phenotype ([67]15, [68]16). Limited
   available molecular data have identified a variety of gene alterations
   such as PTEN loss; TP53, BRAC2, and PIK3CA mutation; and TMPRSS2–ERG
   fusion as the genomic signature of PGCA ([69]6, [70]11, [71]17,
   [72]18). However, a comprehensive genomic landscape of PGCA remains
   sparse, with the limited gene alteration information obtained mostly
   from gene panel sequencing. Notably, another critical barrier for
   better understanding the mutational landscape of PGCA besides the
   scarcity is the profound tumor heterogeneity, evidenced by the
   coexisting adenocarcinoma, more differentiated, yet high-grade ([73]8,
   [74]19). Such heterogeneity not only fuels therapeutic resistance but
   also complicates the molecular mechanism underneath ([75]20).

   In the present study, we examined the histology characteristics of a
   prostatic PGCA patient with bone metastasis through H&E staining and
   immunohistochemical profiling. To delineate the mutational landscape of
   PGCA precisely, laser-capture microdissection was employed to isolate
   specific regions, including PGCA, conventional adenocarcinoma, and
   adjacent stromal compartments for whole-exome sequencing (WES).
   Functional enrichment analysis and clonal evolution modeling based on
   somatic mutations were performed to interrogate dysregulated biological
   pathways and the mechanistic origins of PGCA. Gene mutations
   significantly correlated with dismal prognosis in PGCA were identified
   and further verified with the transcriptional profiles of PCa patients
   from public datasets. Our findings revealed the distinct molecular
   signatures of PGCA and coexisting adenocarcinoma, uncovering potential
   driver mutations that may underlie PGCA pathogenesis and confer
   resistance to current therapies.

Materials and methods

Patient characteristics and tissue samples

   One 74-year-old Chinese man with elevated PSA for a month was admitted
   to Shenzhen People’s Hospital in March 2019. Digital rectal examination
   (DRE) showed an obvious intrarectal lump, and then the patient
   underwent needle biopsy. A total of 12 biopsy samples were collected
   from different spots of prostate. After a series of treatments, the
   samples were embedded in paraffin and further sectioned for H&E
   staining. Two qualified pathologists diagnosed this prostatic disorder
   as PGCA according to the morphology alterations. Furthermore, bone scan
   showed potential metastasis in multiple spots. The patient received
   hormone therapy for 3 months with a second elevation of PSA and then
   took chemotherapy additionally. Eventually the patient underwent TURP
   and received the combination of hormone therapy and chemotherapy. The
   resected samples were consistently treated as before and performed H&E
   staining and gene mutation analysis.

Immunohistochemical staining

   Needle biopsy and transurethral resection samples were sectioned for
   immunohistochemical staining to explore the expression characteristics
   of PCa pathology markers. The sections were immersed in 3% H[2]O[2] for
   10 min at room temperature to block endogenous peroxidase and incubated
   in citrate buffer at 95°C for 40 min for antigen retrieval. After being
   blocked with 3% bovine serum albumin for 1 h at room temperature, they
   were incubated by primary antibodies anti-AR (diluted at 1:500, Abcam,
   cat. no. ab133273), AMACR (diluted at 1:100, Abcam, cat. no. ab194396),
   KLK3 (diluted at 1:1,000, Abcam, cat. no. ab76113), NKX3-1 (diluted at
   1:500, Abcam, cat. no. ab196020), PTEN (diluted at 1:2,000, Abcam, cat.
   no. ab267787), and VIM (diluted at 1:500, Abcam, cat. no. ab92547) at
   4°C overnight. HRP-linked secondary antibody (Abcam, cat. no. ab7090)
   incubation was performed subsequently followed by 3,3′-diaminobenzidine
   (DAB) staining. The sections were counterstained with hematoxylin to
   detect nucleus.

Whole-exome sequencing

   To obtain relatively precise genomic information of PGCA, we collected
   different tissue areas mainly consisting of pleomorphic giant cells,
   adenocarcinoma cells, and stromal cells by Applied Biosystems
   ArcturusXT™ LCM. The transurethral resected samples were sectioned and
   put on the special slide used for LCM. After H&E staining, we performed
   LCM to collect PGCA, PCa, and stroma samples from various spots.
   Genomic DNA of these samples were isolated by QIAamp DNA FFPE Tissue
   Kit (QIAGEN, cat. no. 56404) according to the manufacturer’s
   instructions and sent to Genergy Biotechnology Inc., Shanghai, China,
   for whole-exome sequencing (WES).

Data processing and mapping

   Raw sequencing data were processed by fastp (v0.12.6) as the following
   criteria: 1) removing adaptors; 2) removing reads with more than five
   uncertain bases (N); 3) removing reads with more than 40% low quality
   bases (phred quality score ≤20); 4) sliding window trimming (four
   bases). Clean reads were mapped to the reference GRCh37/hg19 genome
   using the BWA mem function. BAM files were sorted by the Sentieon tools
   (v201808), and the duplicate reads were removed using Sambamba
   (v0.6.6).

Gene mutation annotation

   Aligned reads were further processed with GATK (v.4.1.1.0) for base
   quality recalibration. Somatic mutations were determined via MuTect2
   (v4.1.1.0) and annotated by ANNOVAR (v2018-04-16). Those outside the
   target region, supported by <3 reads or covered by <10 reads, were
   disregarded. Low-confidence somatic mutations considered likely
   germline in ESP6500, 1000 Genomes, or ExAC_EAS (global minor allele
   frequency >0.1%) were also dismissed, excluding those with presence in
   the COSMIC database. In addition, synonymous mutations were filtered
   out for further analyses. R package ComplexHeatmap (v1.20.0) was used
   to visualize the mutational landscape.

Clonal evolution analysis

   The evolutionary relationships and temporal order of each area
   collected from PGCA samples during disease progression were evaluated
   by cellular frequency (CCF) of gene mutations using PyClone-VI
   (v0.1.0). Clusters containing at least five mutations were used to
   infer a consensus clonal evolution model using ClonEvol.

Gene functional enrichment analysis

   Mutated genes specific in pleomorphic giant cells and adenocarcinoma
   cells compared with stroma were identified. Functional enrichment
   analysis was performed using Fisher’s exact test as implemented in the
   clusterProfiler package (v3.8.1), with a Bonferroni correction and an
   adjusted p-value of 0.05.

Clinical correlation analysis

   Gene mutations and clinical information of the PCa patients from public
   datasets were downloaded from cBioPortal. The cBioPortal datasets used
   in this analysis include “Prostate Cancer (DKFZ, Cancer Discov 2018),”
   “Prostate Adenocarcinoma (TCGA, PanCancer Atlas),” “Prostate Cancer,
   Metastatic Castration-Sensitive (MSKCC, Cancer Cell 2020),” and
   “Prostate Cancer, PI3K Pathway Alterations” (MSKCC, Nat Med 2021). To
   estimate independent prognostic factors from PGCA mutations, we
   constructed a Cox proportional hazards model with genomic and prognosis
   information from dataset prostate_dkfz_2018. Other public datasets were
   divided into mutant and wild type by candidate gene mutations for PFS
   and OS analysis to verify their clinical relevance to adverse
   prognosis. Furthermore, we used a chi-square (χ2) test to examine the
   correlation of candidate gene mutations with pathology grade (Gleason
   score) and metastatic potentials. DEGs between samples from mutation
   carriers and wild-type PCa patients provided by public datasets were
   identified with Linear Models for Microarray Data (limma) as the
   FDR-adjusted p-value <0.05 and fold change >2. Functional enrichment of
   these DEGs was analyzed as the same method above.

Results

Distinct morphological and immunohistochemical characteristics of PGCA

   A patient with elevated serum prostate-specific antigen (PSA) for a
   whole month underwent DRE and bone emission computed tomography (ECT),
   suggesting a possible prostatic neoplasm with multiple bone metastases
   ([76] Figure 1A ). The morphology observations from needle biopsies
   confirmed the presence of both conventional adenocarcinoma and PGCA
   components ([77] Figure 1A ). The patient was administered with
   systemic therapies including chemotherapy (docetaxel) and
   endocrinotherapy (goserelin and bicalutamide) for 5 months, followed by
   transurethral resection of the prostate (TURP) due to the massive tumor
   burden and the replacement of bicalutamide with abiraterone ([78]
   Figure 1A ). Despite aggressive treatment, the patient exhibited
   disease progression and died within 6 months post-diagnosis,
   underscoring the aggressive nature of PGCA. Needle biopsies and
   transurethral resected samples were both performed H&E staining for
   histological analysis, demonstrating distinct morphological features of
   PGCA. The adenocarcinoma component showed typical glandular structures
   with relatively uniform nuclei and moderate cytoplasm ([79] Figure 1B
   ). In contrast, PGCA exhibited significant cellular pleomorphism,
   abundant multinucleated giant cells and extensive necrosis, showing a
   more aggressive histological phenotype ([80] Figure 1B ). To further
   explore the cellular and molecular basis of PGCA pathogenesis, we
   performed immunohistochemical staining with key PCa markers and found
   that AR and AMACR were highly expressed in the adenocarcinoma
   component, consistent with luminal differentiation, yet no expression
   was found in PGCA, indicative of lineage plasticity (defined as the
   ability of cancer cells to switch lineage identity in response to
   therapeutic pressure or microenvironmental cues, is increasingly
   recognized as a mechanism of treatment resistance and tumor progression
   in prostate cancer) and dedifferentiation ([81] Figure 1C ). As one of
   the downstream genes of AR, KLK3 was also negatively stained in PGCA,
   suggesting that the elevation of PSA was derived from the coexisting
   adenocarcinoma ([82] Figure 1C ). Loss of PTEN and amplification of
   NKX3-1, the representative hallmarks of PCa progression, were found
   both in PGCA and adenocarcinoma components ([83] Figure 1C ).
   Additionally, VIM was highly expressed in the densely fibrotic stroma
   of adenocarcinoma region, not in the tumor cells ([84] Figure 1C ). In
   contrast, the positive signal of VIM staining was significantly
   enriched in the pleomorphic giant cells of the PGCA region, suggesting
   a transition to a more primitive, treatment-resistant phenotype ([85]
   Figure 1C ). These findings indicated that the PGCA component was more
   aggressive and likely to be the fundamental element for therapeutic
   resistance and adverse prognosis.

Figure 1.

   [86]Diagram detailing a PGCA patient's diagnosis and treatment,
   including DRE, bone ECT, and a medication timeline. Microscopic images
   in panel B show needle biopsy and transurethral samples for
   adenocarcinoma and PGCA with magnified views. Panel C displays stained
   tissue sections with markers AR, AMACR, KLK3, NKX3-1, PTEN, and VIM,
   comparing adenocarcinoma and PGCA. Scale bars indicate magnification
   levels.
   [87]Open in a new tab

   Histopathology characterization of PGCA and coexisting adenocarcinoma
   by H&E staining and immunohistochemical evaluation. (A) Schematic
   drawing illustrating diagnostic procedures and therapeutic
   interventions for the PGCA case, including pharmacological management
   and surgical resection. (B) Histopathology examination of biopsy and
   surgical samples from PGCA and adenocarcinoma components. (C)
   Immunohistochemical staining demonstrating the expression
   characteristics of AR, AMACR, KLK3, NKX3-1, PTEN, and VIM in PGCA and
   adenocarcinoma. Positive signals with different antibodies were stained
   in brown. The cell nucleus was stained with hematoxylin and presented
   blue.

Mutational landscape of PGCA and its aberrant enrichment cell cycle and
apoptosis

   To precisely elucidate the mutational profiles of prostatic PGCA, we
   performed WES on microdissected samples, including PGCA, coexisting
   adenocarcinoma and stroma ([88] Figure 2A ). Genes mutated specifically
   in the PGCA component and significantly associated with dismal
   prognosis were identified and further validated with transcriptional
   profiles ([89] Figure 2A ). Mutations including single-nucleotide
   variants (SNVs) and insertions/deletions (indels) across genomic
   regions showed that PGCA exhibited a higher proportion of mutations in
   gene expression-related regions and fewer mutations in intergenic
   regions, demonstrating a significantly elevated mutational burden in
   PGCA ([90] Figure 2B ). The mutational signature analysis showed a
   notable prevalence of C > T base substitution in both PGCA and
   adenocarcinoma components, suggesting potential DNA repair deficiencies
   or treatment resistance mechanisms ([91] Figure 2C ). This C>T
   transition pattern is a hallmark of spontaneous 5-methylcytosine
   deamination, often linked to aging, oxidative stress, or
   treatment-induced DNA damage ([92]21). Its prominence suggests
   potential defects in DNA repair mechanisms or therapeutic
   pressure-driven mutagenesis that may contribute to the aggressive
   behavior of PGCA. With the background the mutations in stroma, we found
   a striking predominance of Signature87 (SBS87, thiopurine chemotherapy)
   in PGCA, suggesting a potential association with resistance to
   DNA-damaging therapies ([93] Figure 2D ). Furthermore, indel signatures
   ID5, correlated with the age of cancer diagnosis, was significantly
   enriched in adenocarcinoma and also reported may accumulate in normal
   cells ([94] Figure 2D ). In contrast, the indel signatures involved in
   DNA mismatch repair deficiency, including ID3, ID4, ID7, and ID16, were
   exclusive to PGCA, further underscoring the genomic instability ([95]
   Figure 2D ).

Figure 2.

   [96]Diagram illustrating the research process and results related to
   Pleomorphic Giant Cell Adenocarcinoma of the Prostate (PGCA) and
   adenocarcinoma. Panel A shows the workflow from sample collection to
   transcript validation. Panel B displays pie charts of mutation types:
   SNV and Indel for both adenocarcinoma and PGCA, with a legend for
   different genomic regions. Panel C contains bar charts for mutation
   counts, comparing signatures across adenocarcinoma and PGCA. Panel D
   shows pie charts visualizing the distribution of different mutational
   signatures (SBS and ID) in adenocarcinoma and PGCA.
   [97]Open in a new tab

   Mutational signatures of PGCA and adenocarcinoma revealed by WES. (A)
   Experimental design of genomic analysis and transcriptional validation.
   LCM-isolated PGCA, adenocarcinoma, and stroma components from TRUP to
   perform WES. Prognostic gene mutations identified by PFS and OS
   analyses and further validated by transcriptional profiles from public
   datasets. (B) Proportional distribution of SNVs and indels across
   genomic regions. (C) Bar plot showing the mutation signatures of PGCA
   and adenocarcinoma. (D) Pie charts illustrating compositional
   differences in mutation signatures.

   Particularly, somatic mutations in ANKRD1, FASN, and TP53 were
   identified in both PGCA and adenocarcinoma components, with these
   alterations demonstrating significant associations with adverse
   prognosis in PCa ([98] Figure 3A ). The mutated genes in PGCA were
   predominantly enriched in cell cycle dysregulation, such as negative
   regulation of DNA biosynthetic process and cell-cycle G2/M phase
   transition, explaining the presence of bizarre multinucleated giant
   cells in PGCA by mitotic failure ([99] Figures 3B, C ). Furthermore,
   they were also involved in inflammatory factor response with potential
   correlation with apoptosis and immunosuppression, including
   TP53-related DNA damage response, response to inflammatory factors like
   TGFβ and TNF, and negative regulation of the TGFβ receptor signaling
   pathway, collectively contributing to enhanced DNA damage tolerance,
   immunosuppressive microenvironment, and therapeutic resistance ([100]
   Figures 3B, C ). In contrast, adenocarcinoma exhibited mutational
   enrichment in TP53-regulated apoptosis, such as positive regulation of
   DNA damage response, signal transduction by the p53 (tumor suppressor
   protein coded by TP53 gene) class mediator, and positive regulation of
   the signal transduction by p53 class mediator, suggesting a survival
   advantage under therapeutic pressure ([101] Figures 3B, C ).
   Furthermore, adenocarcinoma showed marked aberrations in central carbon
   metabolism and fatty acid biosynthesis, both of which are
   well-established contributors to malignant progression. Central carbon
   metabolism, particularly through enhanced glycolysis (Warburg effect),
   provides rapidly proliferating tumor cells with energy and biosynthetic
   precursors. Aberrant fatty acid biosynthesis, often driven by
   upregulation or mutation of FASN, promotes membrane biosynthesis,
   oncogenic signaling, and lipid-mediated protection from oxidative
   stress, supporting tumor growth and resistance to apoptosis ([102]
   Figures 3B, C ) ([103]22–[104]24). Taken together, the mutational
   landscape of PGCA was defined by defects in cell-cycle regulation and
   immune response pathways, whereas the coexisting adenocarcinoma
   displayed abnormalities in metabolic signaling and p53-mediated
   apoptosis. These molecular distinctions underline the divergent
   biological behaviors and clinical outcomes of PGCA and conventional
   prostate adenocarcinoma.

Figure 3.

   [105]Heatmaps and scatter plots displaying gene alterations and
   pathways in adenocarcinoma. Panel A features a heatmap of gene
   alterations with a legend for mutation types. Panel B shows scatter
   plots for enrichment in pathways like TGFβ signaling. Panel C presents
   plots for cancer types and apoptosis pathways, showing gene counts,
   RichFactor, and significance levels.
   [106]Open in a new tab

   Mutational profiles and functional enrichment analysis. (A) Mutational
   landscape of PGCA and adenocarcinoma compared with stromal tissue. Rows
   represented mutated genes and sorted according to the frequency of
   mutations. Columns represent different legions of PGCA and
   adenocarcinoma. Different colors represent different types of
   mutations. (B) GO terms and (C) KEGG pathways enriched in PGCA and
   adenocarcinoma mutations.

Clonal evolution of heterogeneous PGCA and coexisting adenocarcinoma

   Given that the adenocarcinoma components survived the aggressive
   treatments like PGCA, and their mutational enrichment in TP53-regulated
   apoptosis, we suspected if these malignant cells could be the precursor
   cells or daughter cells of PGCA. Therefore, we conducted a
   comprehensive analysis of clonal evolution in PGCA and the coexisting
   adenocarcinoma ([107] Figure 4A ). The clonalities of mutations
   represented by the variant allele frequency (VAF) distributions were
   divided into a total of three clusters ([108] Figure 4B ). Both of the
   PGCA and adenocarcinoma components contained cluster 1, making it the
   founding clone ([109] Figures 4C, D ). The adenocarcinoma contained
   cluster 2 uniquely, and the cluster 3 were specifically found in PGCA,
   indicating distinct clonal hierarchies between PGCA and coexisting
   adenocarcinoma ([110] Figures 4C, D ). The temporal evolution
   trajectory of clonal dynamics revealed a linear evolution pattern of
   both components that the ancestor clone 1 developed into two branches
   across time, with the end of clone 2 as the adenocarcinoma and clone 3
   as the PGCA ([111] Figures 4E, F ). These findings revealed that PGCA
   and adenocarcinoma had the same origin and developed into two different
   forms of tumor cells during the malignancy progression.

Figure 4.

   [112]Flowchart illustrating a clonal evolution analysis process: A)
   Three-step process involving variant read counts, clone identification
   via clustering, and clonal evolution visualization. B) Two graphs
   showing variant allele frequency (VAF) for adenocarcinoma and PGCA
   across three time points. C) Two density plots for adenocarcinoma and
   PGCA showing evolution over time. D) Cluster diagrams illustrating
   adenocarcinoma and PGCA clonal populations. E) Clonal network diagram
   highlighting relationships between clones and ancestral nodes. F)
   Phylogenetic tree showing clonal relationships. Symbols indicate sample
   founding clones, zero cellular fractions, and ancestors of founding
   clones.
   [113]Open in a new tab

   Clonal evolution patterns in PGCA and adenocarcinoma components. (A)
   Clonal evolution inference pipeline. (B) Boxplot comparing variant
   allele frequency (VAF) of each cluster. (C) Bell plot showing that the
   mutation signatures changed over the pseudo-tumor progression. (D)
   Spheres of cells presenting clonal subpopulations. (E, F) Node-based
   and branch-based trees depicting the clonal relationships of PGCA and
   adenocarcinoma.

Identification of PGCA mutations associated with adverse prognosis

   To explore the potential clinical relevance of mutations identified in
   PGCA, we first analyzed the somatic mutations detected in PGCA via
   cBioWES and focused on those with likely roles in tumor progression.
   Given the limited number of PGCA cases and associated clinical
   follow-up, we employed the prostate_dkfz_2018 dataset—comprising a
   larger prostate cancer cohort with comprehensive genomic and prognostic
   data—to assess the prognostic value of PGCA-derived mutations. Using a
   Cox proportional hazards model, we identified five genes—ADAMTS7, CDH1,
   DRD5, MGAT5, and TP53—whose mutations were significantly associated
   with biochemical recurrence ([114] Figure 5A ). These five genes were
   initially identified as somatically mutated in the PGCA component and
   subsequently validated for prognostic relevance using the
   prostate_dkfz_2018 dataset due to its larger sample size and clinical
   annotation. Functional annotation of these five genes revealed diverse
   roles in tumor progression. ADAMTS7 is involved in extracellular matrix
   remodeling and promotes metastasis ([115]25). CDH1, encoding
   E-cadherin, regulates epithelial adhesion, and its loss facilitates
   epithelial–mesenchymal transition (EMT) ([116]26, [117]27). DRD5, a
   dopamine receptor, is implicated in immune modulation and cell
   signaling ([118]28). MGAT5 contributes to aberrant glycosylation
   promoting invasion and immune evasion ([119]29). TP53 is a master
   regulator of DNA damage response and apoptosis, and its mutation
   disrupts genome stability and enables therapeutic resistance ([120]30).
   The PFS analysis was performed to examine the correlation of candidate
   gene mutations and clinical recurrence, exhibiting significantly
   reduced recurrence-free rates in mutation carriers ([121] Figure 5B ).
   These findings were also validated in an alternative PCa cohort from
   public dataset prad_tcga, where mutation carriers exhibited
   significantly shorter recurrence-free survival ([122] Figure 5C ).
   Furthermore, we analyzed the OS rates of patients harboring mutations
   in the identified genes with the prad_mcspc_mskcc_2020 and
   prad_pik3r1_msk_2021 cohorts and found a significantly shorter OS
   compared with wild-type counterparts ([123] Figure 5D ). Together,
   these findings explicitly emphasized the prognostic value of identified
   mutation signature in PGCA.

Figure 5.

   [124]A composite figure displaying various data visualizations related
   to gene mutations and prostate cancer: A) Table with gene IDs, logrank
   test values, p-values, and hazard ratios represented by a forest plot.
   B) Kaplan-Meier graph showing recurrence-free rates over time for
   mutation vs. wildtype, with significant p-value. C) Kaplan-Meier graph
   comparing recurrence-free rates based on Gleason scores, showing
   statistically significant difference. D) Kaplan-Meier graph comparing
   survival rates over time, exhibiting a significant difference. E)
   Stacked bar chart illustrating prostate cancer patient proportion by
   mutation status and Gleason score across multiple studies. F) Bar chart
   showing mutation frequencies in various studies, categorized by Gleason
   scores. G) Stacked bar chart of prostate cancer patient proportions
   based on mutation status and metastasis presence. H) Bar chart
   highlighting mutation frequencies across different studies, categorized
   by primary versus metastatic status. Each graph includes color codes
   and legends for differentiation.
   [125]Open in a new tab

   Clinical correlation of PGCA mutations in PCa cohorts. (A) Cox
   regression analysis identifying mutations associated with biochemical
   recurrence (prostate_dkfz_2018 dataset). (B, C) Kaplan–Meier curves
   predicting the recurrence-free rates of PCa patients with signature
   mutations (prostate_dkfz_2018 and prad_tcga dataset). (D) OS analysis
   with PCa patients based on the signature mutations
   (prad_mcspc_mskcc_2020 and prad_pik3r1_msk_2021 dataset). (E) Pathology
   grades (Gleason score ≥8 and ≤7) distribution stratified by mutation
   status. (F) Mutation prevalence across pathology grades (Gleason score
   ≥8 and ≤7). (G) Metastasis incidence by mutation status. (H) Mutation
   frequency comparison between primary and metastatic tumors. * p < 0.05,
   ** p < 0.01, *** p < 0.001.

   Enormous evidence has demonstrated that prognostic factors generally
   participated in malignant progression. We therefore evaluated the
   potential associations of candidate gene mutations with pathology
   grades (Gleason score) and metastatic status. Stratified analysis by
   Gleason score revealed that mutation carriers were significantly
   enriched in high-grade tumors (Gleason score ≥ 8) across multiple
   cohorts, suggesting a positive correlation of the identified gene
   mutations and pathology grades ([126] Figures 5E, F ). Additionally,
   the mutation carriers exhibited notably higher frequencies of
   metastasis, and the gene mutations were mostly enriched in metastatic
   samples, suggesting a potential role for these mutations in driving
   metastatic progression ([127] Figures 5G, H ). These clinical
   correlation analyses with multiple cohorts revealed that mutations in
   ADAMTS7, CDH1, DRD5, MGAT5, and TP53 significantly correlated with and
   potentially responsible for the malignant progression and therapeutic
   resistance.

Transcriptional validation of candidate gene mutations in malignant
progression

   To delineate the molecular landscape of PCa driven by PGCA prognostic
   mutations, we examined the differentially expressed genes (DEGs)
   between mutation carriers and non-carriers from several cohorts,
   including prad_su2c_2019, prad_tcga, and prostate_dkfz_2018 ([128]
   Figure 6A ). These datasets were specifically selected because they
   represent extensive, conventional PCa cohorts with rich genomic and
   clinical annotations, enabling robust comparative validation of
   PGCA-derived mutations in a broader oncological context. Although PGCA
   is a rare and aggressive variant, evaluating its candidate mutations
   within these larger, conventional PCa cohorts allowed us to investigate
   the wider relevance and prognostic potential of these alterations
   beyond the rare subtype. Patients with candidate gene mutations highly
   expressed genes related to cell-cycle arrest and epithelial–mesenchymal
   transition (EMT), such as negative regulation of mitotic cell-cycle
   phase transition, negative regulation of mitotic sister chromatid
   separation, positive regulation of EMT, and extracellular matrix
   organization ([129] Figure 6B ). These processes were consistent with
   the typical PGCA features including the presence of multinucleated
   cells and enhanced metastatic potentials. Furthermore, we found that
   the mutation carriers showed high expression levels of genes enriched
   in IL-4 production and negative regulation of the immune system
   process, suggesting an immunosuppressive potential for tumor growth
   ([130] Figure 6B ). On the other hand, KEGG signaling pathway
   enrichment analysis showed that mutation carriers exhibited aberrant
   activation of p53 and PI3K-Akt signaling pathway, both critical for
   survival and therapeutic resistance, and dysregulation of biological
   processes including DNA replication, cell cycle, and homologous
   recombination, collectively contributing to genomic instability ([131]
   Figure 6C ). Concurrently, integration of multicohort transcriptomic
   data revealed that PGCA candidate mutations drive transcriptional
   reprogramming mainly related to cell-cycle dysregulation, metastasis,
   and anti-apoptosis.

Figure 6.

   [132]Three sections labeled A, B, and C. Section A contains three
   volcano plots with gene expression data from three studies:
   prad_su2c_2019, prad_tcga, and prostate_dkfz_2018. Genes with
   significant fold changes are highlighted. Section B shows enrichment
   dot plots with different biological processes listed for each study.
   P-values and Rich Factors are displayed along the axes. Section C
   presents pathway analysis results, with various pathways and their
   significance levels marked on dot plots. Each plot compares gene
   expressions, biological processes, or pathways relevant to prostate
   cancer research.
   [133]Open in a new tab

   Transcriptomic alterations associated with signature mutations. (A)
   Volcano plot showing the DEGs between mutation-positive and
   mutation-negative tumors. (B) Biological processes and (C) signaling
   pathways of upregulated genes in PCa patients with signature mutations.

Discussion

   Emerging evidence indicated pleomorphic giant cells with polyploidy in
   tumors associated with drug resistance and adverse survival ([134]31).
   In this study, a prostatic PGCA patient administered with chemotherapy
   and endocrinotherapy alongside TURP only lived for 6 months
   post-diagnosis. Histopathology examinations showed a distinct phenotype
   characterized by cellular pleomorphism, multinucleated giant cells, and
   necrosis. Cellular pleomorphism of PGCA indicated significant
   malignancy of tumor cells, given that conventional adenocarcinoma, even
   with a high Gleason score, typically consists of cells with relatively
   uniform nuclei ([135]32). Most multinucleated giant cells, induced by
   cell-cycle arrest, cell–cell fusion, or therapeutic stress, were
   identified around necrotic areas probably caused by the hypoxic
   microenvironment ([136]33, [137]34). Furthermore, abundant studies have
   shown that the pleomorphic giant cells exhibited stem cell-like
   features with a similar proliferation cycle to embryonic development
   and dominated tumor metastasis ([138]35–[139]38). The unique
   immunohistochemical profiles examined from our study showed the absence
   of AR, AMACR, and KLK3 expression, coupled with high VIM expression in
   pleomorphic giant cells. Loss of AR and the downstream gene expression
   indicated insensitivity to androgen, suggesting resistance to
   endocrinotherapy, which was reported before in individual cases
   ([140]8). A large number of studies showed that an elevated expression
   of VIM suggested a loss of epithelial differentiation and acquisition
   of mesenchymal traits, potentially linked to EMT and metastatic
   behaviors ([141]39, [142]40). These findings demonstrated the
   potentially critical roles of pleomorphic giant cells in multiple drug
   resistance, attributed to the embryonic morphology, loss of AR, and
   overexpression of VIM.

   The mutational landscape of PGCA components, examined by WES with
   samples obtained from LMD, showed significant enrichment in cell-cycle
   dysregulation and growth factor response with potential correlation
   with apoptosis. Among the mutated genes, TP53 and CDC25B have been
   verified capable of preventing G2/M-phase cell-cycle arrest, suggesting
   that the multinucleated giant cells were induced by cell-cycle arrest
   in PGCA ([143]41–[144]43). Early studies indicated that TGFβ and TNF
   were critical factors responsible for inflammatory microenvironment
   construction in malignant tumors and were also involved in a variety of
   cellular processes, including proliferation, motility, and apoptosis
   ([145]44, [146]45). Mutations in genes associated with these signaling
   pathways could potentially cause apoptotic effects on tumor cells and
   promoted survival from aggressive treatments and the immune system in
   PGCA. In comparison, the coexisting adenocarcinoma also possessed TP53
   mutation, among other gene mutations associated with TP53-mediated
   apoptotic dysfunction and metabolic aberrations. These findings
   demonstrated that both PGCA and coexisting adenocarcinoma possessed
   TP53 mutation and presented tolerance to chemotherapy and
   endocrinotherapy. Given that the pleomorphic giant cells have been
   reported capable of generating daughter cells with even more aggressive
   phenotypes through asymmetric division ([147]46, [148]47), we analyzed
   the development trajectory of these two malignant components to examine
   if they had a directive evolutionary relationship. The clonal evolution
   analysis showed that they shared the same clonal origin yet diverged
   during malignant progression, acquiring distinct genomic signatures.
   Such lineage bifurcation explained the clinical aggressiveness of
   prostatic PGCA and its resistance to conventional therapies, also
   reminding us the aggressive malignancy of coexisting adenocarcinoma.

   Our clinical correlation analysis with multiple public cohorts
   identified a total of five genes, namely, ADAMTS7, CDH1, DRD5, MGAT5,
   and TP53, whose mutations were significantly associated with malignant
   progression and adverse prognosis in PCa. TP53 mutations,
   well-documented in various malignancies associated with enhanced
   genomic instability and therapeutic resistance, were identified in both
   PGCA and adenocarcinoma ([149]48, [150]49). The remaining gene
   mutations were all PGCA specific, among which ADAMTS7 mutations were
   found associated with oncogenesis and dismal prognosis in a variety of
   cancers ([151]50–[152]52). ADAMTS7 is a cartilage oligomeric matrix
   protein-cleaving enzyme involved in extracellular matrix remodeling,
   whose family members are also linked to tumor progression by affecting
   the interplay between the malignant cells and local microenvironment
   ([153]53). Notably, tumor-suppressor CDH1 (E-cadherin) mutations were
   frequently inactivated in metastatic tumors with immunosuppressive
   microenvironment ([154]54–[155]58), likely contributing to immune
   evasion and migrative phenotype of pleomorphic giant cells in this
   case. DRD5 is a G protein-coupled receptor with differential expression
   and mutational profiles in various cancers. In 12 tumor types including
   bladder cancer, breast cancer, esophageal squamous cell carcinoma, and
   head and neck cancer, hypermethylation in the DRD5 promoter region
   leads to gene expression silencing, whereas in non-small cell lung
   cancer, DRD5 expression was elevated in response to docetaxel ([156]59,
   [157]60). Activation of DRD5 by the selective agonist [158]SKF83959
   upregulates reactive oxygen species (ROS) levels in tumor cells,
   thereby activating the mitochondrial apoptosis pathway while inhibiting
   the mTOR pathway to induce autophagic cell death ([159]28). However,
   there were also reports indicating that mutations or suppression of
   DRD5 could induce an immunocompromised microenvironment and promote
   cell survival from therapeutic drugs ([160]28, [161]61, [162]62). MGAT5
   encoding a glycosyltransferase was found overexpressed in a variety of
   malignancies, unlike our findings, whose loss function was noted for
   tumor suppression and inhibition of metastasis ([163]63–[164]66),
   implicating different roles for PGCA pathogenesis. In ovarian cancer,
   upregulation of MGAT5 could also exacerbate resistance to anti-PD-L1
   immunotherapy by catalyzing branched N-glycans, which enhanced its
   binding with PD-1 on CD8^+ T cells ([165]67). Furthermore,
   transcriptional profiling confirmed that mutations in these genes drive
   transcriptional reprogramming linked to cell-cycle dysregulation,
   antiapoptosis, immunosuppression, and metastasis, reinforcing their
   functional relevance to the malignant maintenance of PGCA phenotypes.

   This study is based on a single well-characterized case, which may
   limit broad generalization. However, given the extreme rarity of
   prostatic PGCA, our comprehensive multiomics and pathological profiling
   offers valuable insights and establishes a foundation for future
   studies in larger cohorts.

Conclusion

   This study provides the first integrative characterization of prostatic
   pleomorphic giant cell adenocarcinoma (PGCA), revealing its distinct
   histological, immunohistochemical, and genomic features. We identified
   key mutations—particularly in ADAMTS7, CDH1, DRD5, MGAT5, and
   TP53—associated with malignant progression and poor prognosis. PGCA
   showed transcriptional reprogramming related to cell-cycle
   dysregulation, immune evasion, and therapeutic resistance. Our findings
   uncover novel molecular insights into PGCA pathogenesis and suggest
   potential biomarkers and therapeutic targets for this rare and
   aggressive prostate cancer subtype.

Funding Statement

   The author(s) declare financial support was received for the research
   and/or publication of this article. The work was supported by the
   National Natural Science Foundation of China (No. 82303339); Shenzhen
   Science and Technology Program (No. JCYJ20240813101919026, No.
   JCYJ20220530152405011); Shenzhen Third People’s Hospital Research Fund
   (Nos. 24250G1016, 24250G1017, 25270G1017, and 22250G1047); and Shenzhen
   High-level Hospital Construction Fund (No. 24252G1002).

Data availability statement

   The WES data of this study is available in China National Center for
   Bioinformation / National Genomics Data Center under the accession code
   HRA010637. The VCF files containing mutational information has been
   deposited in National Genomics Data Center
   ([166]https://ngdc.cncb.ac.cn/) under the accession number of
   subPRO062622.

Ethics statement

   The studies involving humans were approved by the Ethics Committee of
   Shenzhen People’s Hospital, under approval number LL-KY-2021696. The
   studies were conducted in accordance with the local legislation and
   institutional requirements. The participant provided their written
   informed consent to participate in this study. Written informed consent
   was obtained from the individual(s) for the publication of any
   potentially identifiable images or data included in this article.

Author contributions

   XM: Conceptualization, Funding acquisition, Writing – original draft,
   Writing – review & editing. KC: Data curation, Investigation, Writing –
   review & editing. JZ: Formal analysis, Methodology, Writing – review &
   editing. LL: Funding acquisition, Resources, Writing – review &
   editing. JL: Funding acquisition, Resources, Writing – review &
   editing. KH: Validation, Writing – review & editing. HZ: Validation,
   Writing – review & editing. DL: Methodology, Writing – review &
   editing. JG: Visualization, Writing – review & editing. CF:
   Supervision, Writing – review & editing. XZ: Funding acquisition,
   Writing – review & editing. WL: Funding acquisition, Writing – review &
   editing. LC: Software, Writing – review & editing. LY: Supervision,
   Validation, Writing – review & editing. XM: Supervision, Writing –
   review & editing. ZC: Funding acquisition, Project administration,
   Supervision, Writing – review & editing.

Conflict of interest

   The authors declare that the research was conducted in the absence of
   any commercial or financial relationships that could be construed as a
   potential conflict of interest.

Generative AI statement

   The author(s) declare that no Generative AI was used in the creation of
   this manuscript.

Publisher’s note

   All claims expressed in this article are solely those of the authors
   and do not necessarily represent those of their affiliated
   organizations, or those of the publisher, the editors and the
   reviewers. Any product that may be evaluated in this article, or claim
   that may be made by its manufacturer, is not guaranteed or endorsed by
   the publisher.

References