Abstract

   Background: Tertiary lymphoid structures (TLSs), organized clusters of
   immune cells within non-lymphoid tissues, significantly influence tumor
   progression and therapeutic response. However, their prognostic
   relevance and underlying regulatory mechanisms in clear cell renal cell
   carcinoma (ccRCC) remain insufficiently characterized.

   Methods: We integrated transcriptomic and clinical data from 928 ccRCC
   patients to construct a TLS-related prognostic RiskScore using machine
   learning algorithms. TLS maturation heterogeneity was characterized via
   immunohistochemistry and multiplex immunofluorescence analyses. The
   functional role of interferon regulatory factor 4 (IRF4), a key
   regulator within the TLS gene network, was investigated using in vitro
   assays. Single-cell RNA sequencing (scRNA-seq) and spatial
   transcriptomics were employed to dissect the involvement of IRF4 in TLS
   formation and maturation.

   Results: The derived TLS-associated signature RiskScore, comprising
   CCL22, LOXL1, LIPA, ADAM8, and SAA1, effectively stratified patients
   into distinct prognostic groups and showed robust associations with
   clinical parameters, tumor microenvironment (TME) features, and
   predicted immunotherapy responses. Functional assays demonstrated that
   IRF4 significantly enhanced the malignant phenotype of 786-O and 769-P
   ccRCC cells. Clinically, elevated IRF4 expression independently
   predicted worse patient outcomes, characterized by a predominance of
   immature TLS phenotypes, reduced TLS density, and diminished CD8⁺ T
   cell infiltration. Mechanistically, scRNA-seq analyses revealed that
   active IRF4 signaling was predominantly confined to immature B cell
   states and was inversely associated with TLS maturation trajectories.
   Spatial transcriptomics further confirmed IRF4 enrichment within TLS
   regions, notably spatially segregated from high endothelial venules
   (HEVs) and mature TLS compartments.

   Conclusion: In conclusion, this study establishes a robust TLS-related
   prognostic signature for ccRCC and elucidates the mechanistic role of
   IRF4 in promoting TLS immaturity and immune dysfunction. By potentially
   recruiting immature B cells while impairing their maturation, IRF4
   contributes to an ineffective anti-tumor immune landscape, offering a
   promising target for therapeutic intervention.

   Keywords: tertiary lymphoid structures (TLSs), clear cell renal cell
   carcinoma (ccRCC), tumor microenvironment (TME), interferon regulatory
   factor 4 (IRF4), single-cell RNA sequencing (scRNA-seq), spatial
   transcriptomics analysis

Introduction

   Renal cell carcinoma (RCC) represents a significant cause of cancer
   mortality worldwide [55]^1^-[56]^3. In 2022, it was estimated that
   there were over 7.37×10^4 novel cases and 2.40×10^4 fatalities due to
   RCC [57]^4. RCC prevalence in China has increased over the past thirty
   years, and the rate of new diagnoses per 100,000 individuals has nearly
   tripled, rising from 1.16 in 1990 to 3.21 in 2019[58]^5. Clear cell
   renal cell carcinoma (ccRCC), the most frequent histological subtype,
   accounts for approximately 70% of all RCC cases [59]^6. Genetic
   alterations in key driver genes such as VHL, SETD2, PBRM1, and BAP1 are
   characteristic of ccRCC, contributing to genomic instability and
   impaired DNA repair mechanisms, thereby potentially promoting ccRCC
   tumorigenesis [60]^7.

   ccRCC is characterized by significant heterogeneity, leading to
   variable patient prognoses and highlighting the critical need for
   improved biomarkers beyond traditional clinicopathological features. In
   recent years, considerable research effort has focused on developing
   molecular signatures derived from high-throughput data to better
   stratify patients and predict outcomes. These efforts have explored
   various biological signatures of ccRCC. For instance, researchers have
   investigated signatures linked to specific cell death pathways
   [61]^8^,[62]^9, cancer stemness and non-coding RNAs [63]^10, tumor
   microenvironment (TME) and immune components [64]^11^,[65]^12. This
   landscape of recently proposed biomarkers underscores the diverse
   strategies being employed and sets the stage for introducing novel
   signatures, like the TLS-based one in this study, which may offer
   unique insights into the tumor's immune landscape and patient
   prognosis. Despite identifying numerous potential genomic biomarkers,
   the translation of these findings into routinely used tissue or
   blood-based molecular biomarkers for guiding clinical decision-making
   in ccRCC remains a significant challenge [66]^13.

   TME is an intricate and dynamic system composed of diverse cell types,
   including endothelial cells, fibroblasts, and immune cells [67]^14. The
   TME plays a critical role in tumor initiation, progression, and
   metastasis, and modulates therapeutic response [68]^15^,[69]^16, being
   actively shaped by components such as cancer-associated fibroblasts
   [70]^17. Furthermore, systemic host factors, such as the gut
   microbiota, and intrinsic microenvironmental conditions, such as
   hypoxia, are increasingly recognized as modulators of the TME
   [71]^18^,[72]^19. Recent research indicates that tumor-associated
   immunity is evident in tertiary lymphoid structures (TLSs) [73]^20.
   TLSs form in nonlymphoid tissues under pathological conditions,
   including cancer, rather than during physiological states
   [74]^21^,[75]^22. Structurally, TLSs are featured by a central cluster
   of CD20^+ B cells surrounded by CD3^+ T cells, similar to the lymphoid
   follicles in secondary lymphoid organs (SLOs) [76]^23. The T cell
   region within TLSs consists of CD4^+ T follicular helper (Tfh) cells,
   CD8^+ cytotoxic T lymphocytes, CD4^+ T helper 1 (TH1) cells, and
   regulatory T cells (Tregs) [77]^24^,[78]^25. Unlike SLOs, TLSs
   typically lack a distinct capsule, potentially facilitating their
   cellular constituents to interact directly with adjacent tissue
   [79]^26.

   TLSs are implicated in initiating and enhancing adaptive immune
   responses [80]^27^,[81]^28. T cells within TLSs interact with mature
   dendritic cells (DCs) and B cells, inducing T cell differentiation and
   the development of germinal centers (GCs) [82]^21. After B cells
   achieve complete maturation within TLSs, they become plasma cells that
   secrete high-affinity IgG and IgA, potentially increasing the tumor's
   response to immunotherapy [83]^29. TLSs are increasingly recognized as
   crucial organizing centers for anti-tumor immunity, influencing
   prognosis and therapeutic response across numerous malignancies
   [84]^30. Accumulating evidence across diverse malignancies, such as
   melanoma [85]^31, ovarian cancer [86]^32, non-small cell lung cancer
   [87]^33, bladder cancer [88]^34, and pancreatic cancer [89]^35,
   consistently links the presence, density, and particularly the
   maturation state of TLSs with patient prognosis and efficacy of immune
   checkpoint inhibitor (ICI) therapies. Indeed, understanding the broader
   TME's impact is crucial, as studies highlight its pivotal role in
   regulating intercellular communication, shaping treatment responses,
   and driving resistance to chemoradiation, targeted therapy, and
   immunotherapy [90]^36. However, the presence of TLSs in various tumors
   can differ significantly, which may lead to variable outcomes [91]^37.
   Factors influencing TLS maturation, such as local metabolic constraints
   [92]^38, and the specific cellular interactions within them, including
   T-cell activation levels [93]^35, may contribute to this heterogeneity.
   In ccRCC, the different localization and maturity of TLSs can lead to
   different prognoses, with tumor-proximal TLSs being more mature and
   exhibiting better prognoses [94]^39. Therefore, prognostic models for
   ccRCC incorporating TLS molecular markers hold potential clinical
   significance.

   Interferon regulatory factor 4 (IRF4) serves as one of the genetic
   markers for identifying TLSs in cancers, and it fulfills a complex role
   in immune regulation [95]^23. Increased IRF4 levels enhance the
   differentiation of CD4^+ CD25^low effector T (Teff) cells and reduce T
   follicular helper (Tfh) cell numbers [96]^40. Additionally, IRF4
   directs effector regulatory T (Treg) cell differentiation and promotes
   immune suppression [97]^41. Upregulated IRF4^+ Treg cells within tumors
   are significantly linked to early tumor recurrence as well as worse
   disease-free survival (DFS) and overall survival (OS) [98]^42. Although
   IRF4 fosters proliferation and continuous differentiation of CD8^+ T
   cells [99]^43, persistently high IRF4 can lead to CD8^+ T cell
   exhaustion [100]^44. Specifically, recent studies confirm that IRF4
   upregulation following T cell activation impedes human CD8 T cell
   effector function, while promoting cell proliferation and PD-1
   expression, contributing to an exhausted phenotype in
   tumor-infiltrating lymphocytes (TILs) rather than activation alone
   [101]^45. Moreover, IRF4 induces M2-type macrophages [102]^46,
   contributing to an immunosuppressive TME. This process can be
   influenced by upstream signaling [103]^47 and by post-translational
   modifications [104]^48.

   Previous studies have elucidated IRF4's involvement across various
   tumor types. Anaplastic large cell lymphoma shows a dependency on IRF4
   signaling, with MYC as a key target of IRF4[105]^49. Furthermore,
   mutations affecting IRF4-DNA binding can upregulate genes specific to
   human lymphomas [106]^50. IRF4 mRNA escalates from normal tissues to
   oral submucous fibrosis and oral squamous cell carcinoma (OSCC),
   paralleling immune infiltration [107]^51. Moreover, IRF4 acts as an
   oncogenic factor in human non-small cell lung cancer, partly through
   triggering the Notch-Akt pathway [108]^52. In multiple myeloma (MM),
   IRF4 is a critical transcription factor whose activity and cellular
   growth effects can be deregulated by the loss of the inhibitory protein
   BCL7A, which normally limits its DNA binding activity [109]^53. Due to
   its specific overexpression and role in mediating MM progression and
   survival, IRF4 is actively being pursued as a therapeutic target, with
   novel direct small molecule inhibitors that bind its DNA-binding domain
   being designed and synthesized [110]^54 and antisense
   oligonucleotide-based approaches being explored to silence its
   expression [111]^55. Nevertheless, the precise functions of IRF4 in
   ccRCC are yet to be elucidated.

   This study employed an integrative approach to elucidate the complex
   role of TLSs in ccRCC. We developed and validated a robust TLS-related
   signature capable of predicting patient prognosis, TME characteristics,
   and immunotherapy response. Then, since IRF4 is a key TLS-associated
   gene with unclear functions in ccRCC, we investigated its specific
   contribution to tumor progression and TLS maturation. By combining in
   vitro functional assays with single-cell and spatial transcriptomics
   analyses, we specifically sought to determine how IRF4 expression
   influences B cell dynamics within the TME, contributes to TLS
   heterogeneity, and ultimately impacts ccRCC progression, thereby
   identifying potential mechanisms driving immune evasion and revealing
   novel therapeutic vulnerabilities.

Materials and Methods

Raw Data Collection and Standardization Process

   This study compiled transcriptomic and clinical data for 763 ccRCC
   patients from several online databases, including the Cancer Genome
   Atlas (TCGA) database, clinical proteomic tumor analysis consortium
   (CPTAC) database, European Molecular Biology Laboratory (EMBL)
   database, and International Cancer Genome Consortium (ICGC) database,
   using datasets TCGA-KIRC, CPTAC-3, E-MTAB-3267, and RECA-EU. To ensure
   comparability and standardization, RNA sequencing data initially
   presented as fragments per kilobase of transcript per million mapped
   reads (FPKM) were converted to transcripts per million (TPM), followed
   by log2(TPM+1) transformation. Additionally, we mitigated the batch
   effect using the “ComBat” algorithm of the “sva” package. The study
   also incorporated genetic alteration data from the TCGA database,
   covering somatic mutations, copy number variations (CNVs), and tumor
   mutational burden (TMB). All patient consent and ethical approvals were
   appropriately secured in the original studies. Baseline clinical data
   for the patients were presented in [112]Table S1.

Consensus Clustering Analysis

   Our study utilized a consensus clustering approach to divide ccRCC
   patients into separate subgroups according to TLS-related genes and
   differentially expressed genes (DEGs), respectively. We executed
   consensus hierarchical clustering and confirmed the ideal cluster
   quantity and the distribution of patients using the
   “ConsensusClusterPlus” package in R, performing 1000 repetitions for
   credibility assurance [113]^56.

Correlation of Clinical Traits and Immune Landscape with Molecular Profiles

   Several clinical variables were taken into consideration, including
   patient age, gender, tumor grade, and Stage. Utilizing the “survival”
   and “survminer” packages in R software, we performed the Kaplan-Meier
   (K-M) analysis to compare prognostic outcomes [114]^57. Then, we
   estimated the prevalence of immune cells within samples by implementing
   the CIBERSORT algorithm [115]^58, and we pinpointed the immune cells
   infiltrating proportion through the single-sample gene set enrichment
   analysis (ssGSEA) algorithm [116]^59. Additionally, we calculated the
   ESTIMATE score for each ccRCC specimen using the ESTIMATE algorithm
   [117]^60, and compared immune checkpoints (ICPs) expression across
   subgroups.

Functional and Pathway Enrichment Analysis

   We performed Gene Ontology (GO) and Kyoto Encyclopedia of Genes and
   Genomes (KEGG) pathway enrichment analyses with the “clusterProfiler”
   package [118]^61. Additionally, we utilized the KEGG gene set
   (c2.cp.kegg.v7.5.1) to execute gene set variation analysis (GSVA) to
   underscore functional discrepancies among clusters, based on the
   adjusted P value < 0.001 and the threshold of |log2-fold change (FC)| >
   0.2[119]^62.

Distinctive Expression and Development of the TLS-Related Predictive
Signature

   The “limma” package in R software was utilized to pinpoint DEGs among
   different TLS-gene-related clusters, setting the selection threshold at
   |log2-FC| > 1.25 and the adjusted P value < 0.001[120]^63. We then
   devised a prognostic scoring system (RiskScore). To recognize
   prognosis-related genes, univariate Cox regression analysis was
   performed, considering DEGs expression levels and survival data.
   Subsequently, we used the least absolute shrinkage and selection
   operator (LASSO) technique and multivariate Cox regression analysis to
   create an optimal predictive model. RiskScore was calculated as
   follows: RiskScore = h0(t) × exp (expression of CCL22 × corresponding
   coefficient + expression of LOXL1 × corresponding coefficient +
   expression of LIPA × corresponding coefficient + expression of ADAM8 ×
   corresponding coefficient + expression of SAA1 × corresponding
   coefficient). Patients were then categorized into low or high
   RiskGroups based on the median RiskScore.

Clinical Relevance of the Predictive Signature and Establishment of a
Prognostic Outcome Prediction Nomogram

   We compared clinical outcomes between the two RiskGroups by employing
   K-M analysis. We confirmed the predictive accuracy of the RiskScore
   system using a receiver operating characteristic (ROC) curve, created
   with the “survival ROC” package in R. Moreover, the prognostic
   relevance of the signature was investigated by categorizing ccRCC
   patients according to their clinical features.

   Integrating the RiskScore with clinical parameters, we formulated a
   nomogram for forecasting the 1-year, 3-year, and 5-year overall
   survival (OS) rates of ccRCC patients, utilizing the “rms”, “regplot”,
   and “survival” packages in R. To determine the accuracy and robustness
   of the nomogram, we conducted calibration curves, time-dependent ROC
   curves, and decision curve analysis (DCA).

Forecasting the Treatment Response to ICI Therapy

   To forecast the response to immune checkpoint inhibitor (ICI) therapy
   in ccRCC patients, we utilized the tumor immune dysfunction and
   exclusion (TIDE) analysis. TIDE assesses two principal immune escape
   mechanisms: impaired T cell functionality and the inhibition of T cell
   infiltration with low cytotoxic T lymphocyte (CTL) counts [121]^64. Low
   TIDE score patients are anticipated to have a favorable response to
   immunotherapy. Furthermore, we corroborated the response to
   immunotherapy of the TLS-related signature by analyzing the David Liu
   cohort, which involved 121 patients with metastatic melanoma who
   received anti-PD-1 blocking agent treatment (nivolumab or
   pembrolizumab) [122]^65. We also validated the result with CheckMate
   009 and 025 (CheckMate cohort), and 136 advanced ccRCC patients treated
   with nivolumab (anti-PD-1) [123]^66.

Specimen Collection and H&E Staining

   In our study, we included a cohort of 60 patients diagnosed with ccRCC
   who were treated at the Department of Urology, Fudan University
   Shanghai Cancer Center (FUSCC, Shanghai, China). We also included
   clinical data for 105 ccRCC cases from the Department of Urology,
   FUSCC, between 2010 and 2023, to examine the relationship between IRF4
   expression and prognosis. These patients underwent surgical procedures
   during which tissue samples were collected for further analysis. The
   study design and the collection and use of these tissue samples were
   carried out with strict adherence to the Declaration of Helsinki. The
   study was approved by the FUSCC Ethics Committee, and all patients
   provided informed permission.

   To assess the extent of lymphocytic infiltration, histopathological
   examinations were conducted. Hematoxylin and eosin (H&E) staining was
   conducted on slides of both ccRCC and adjacent normal tissues following
   reported protocols [124]^67, and they were carefully reviewed by two
   experienced pathologists independently.

TLS Maturation Analysis with Multiplex Immunofluorescence (mIF) Staining
Assays

   To further investigate the tumor microenvironment [125]^68^-[126]^70,
   we employed independent IHC staining on adjacent slides and multiplex
   immunofluorescence staining assay, conducted in collaboration with
   Shanghai KR Pharmtech, Inc., Ltd. (Shanghai, China). Specifically, we
   used the 7-color multiplex immunofluorescence kit (KR Pharmtech, Inc.,
   Ltd. (Shanghai, China) to analyze the maturation of TLSs. For the
   staining process, we selected a panel of specific antibodies, namely
   CD8 (Abcam, ab178089, 1:100), CD20 (Dako, L26, IR604), CD21 (Abcam,
   ab227662, 1:100), CD23 (Abcam, ab315289,1:100), CK (Abcam, ab7753,
   1:100), and PD-L1 (CST, E1L3N, 13 684S, 1:400) [127]^71^,[128]^72.

   The tissue slides obtained were first treated with Antibody Diluent
   Block buffer. Following this, the slides were incubated with the
   primary antibody for 40-60 minutes. Then, the slides were treated with
   a polymer horseradish peroxidase (HRP)-conjugated secondary antibody
   for 10 minutes. Next, the Fluorophore Working Solution was applied to
   the slides for 10 minutes, visualizing the complexes. The nuclei within
   the tissue slides were stained with 4',6-diamidino-2-phenylindole
   (DAPI). Finally, whole slide scans were conducted at 20× magnification
   using the KR-HT5 system (KR Pharmtech, Inc., Ltd. (Shanghai, China)).
   The acquired images were then analyzed using inForm 2.4.0 software.

   Images were classified into TLS maturation stages as follows: early TLS
   (E-TLS) depicted dense lymphocytic aggregates of mixed CD8^+T and B
   cells (CD8 and CD20 positive) without follicular DCs (FDCs) and GC
   (CD21 and CD23 signals negative); primary follicle-like TLS (PFL-TLS)
   showed dense lymphocytic aggregates with FDCs and absence of a GC (CD21
   positive but without CD23 signals); secondary follicle-like TLS
   (SFL-TLS) were identified by both FDCs and a GC (CD21 and CD23
   positive)[129]^39. TLS-positive samples in the study were categorized
   as follows: SFL-TLS positive, containing at least one instance of
   SFL-TLS; PFL-TLS positive, with at least one PFL-TLS occurrence but no
   SFL-TLS; E-TLS positive, exhibiting neither PFL-TLS nor SFL-TLS
   [130]^39^,[131]^73.

Immunohistochemistry and Scoring Assays

   After identifying at least one instance of TLSs in H&E-stained slides,
   adjacent slides were subjected to immunohistochemical (IHC) staining,
   performed as previously described [132]^67, and IRF4 (ab315394, Abcam)
   was identified.

   In order to quantify the intensity of IRF4 staining, an
   immunohistochemistry scoring system was employed, named IHCscore. The
   staining number score was determined based on the percentage of cells
   showing positive staining: a score of 0 was assigned for negative
   staining, 1 for 1-25% staining, 2 for 26-50% staining, 3 for 51-75%
   staining, and 4 for 76-100% staining. The staining color scored 0 for
   negative, 1 for weak, 2 for medium, and 3 for strong staining. The
   final IHCscore was calculated by multiplying the staining number score
   by the color score [133]^74. Based on the median IHCscore, the samples
   were further categorized into high and low groups.

Cell Culture and siRNA Transfection

   The human ccRCC cell lines 786-O and 769-P were acquired from the Type
   Culture Collection Cell Bank, Chinese Academy of Sciences. These cells
   were nurtured in RPMI 1640 medium, supplemented with 10% fetal bovine
   serum and 1% penicillin/streptomycin. Cells were maintained in a 37°C
   incubator with 5% CO[2]. Cells were seeded in 10-cm dishes at 50%
   confluency one day before transfection. Transfection commenced when
   cell density approached 70%. Hieff Trans^® Liposomal Transfection
   Reagent (Yeasen, Shanghai, China) was used to prepare complexes with
   the negative control and IRF4 siRNA, according to the protocol from the
   manufacturer. After a 15-minute room temperature incubation, these
   complexes were introduced to the cells. We harvested the cells at 72
   hours post-transfection for later analysis.

Western Blotting Assay

   Cells were lysed using Cell lysis buffer for Western and IP (P0013,
   Beyotime) supplemented with protease inhibitors. Protein concentrations
   were determined using the BCA assay (P0009, Beyotime). Equal amounts of
   protein were separated by 10% SDS-PAGE and transferred onto PVDF
   membranes. Membranes were blocked with 5% non-fat milk in TBST for 1
   hour at room temperature. Membranes were then incubated overnight at
   4°C with primary antibodies diluted in blocking buffer. The following
   primary antibodies were used: IRF4 (11247-2-AP, Proteintech), CXCL13
   (10927-1-AP, Proteintech), BCL6 (21187-1-AP, Proteintech), PD-L1
   (28076-1-AP, Proteintech), and GAPDH (60004-1-Ig, Proteintech). After
   washing with TBST, membranes were incubated with horseradish peroxidase
   (HRP)-conjugated secondary antibodies for 1 hour at room temperature.
   The following secondary antibodies were used: Goat anti-Rabbit IgG
   (H+L) (SA00001-2, Proteintech) and Goat anti-Mouse IgG (H+L)
   (SA00001-1, Proteintech). Bands were visualized using an enhanced
   chemiluminescence (ECL) detection kit (36222ES, Yeason) and imaged
   using ChemiDoc Imaging System (Bio-Rad).

Cellular Functional Experiments In Vitro

   The proliferation capability of cell lines was detected by colony
   formation assay and CCK-8 assay, respectively. For colony formation
   assay, we seeded 786-O or 769-P into 6-well plates with 1000 cells per
   well. After 14 days of culture in complete medium, the plates were
   gently washed twice with PBS, fixed with 4% paraformaldehyde for 20
   minutes at room temperature, and then stained with 0.1% Crystal Violet
   solution for 20 minutes at room temperature. Following staining, plates
   were carefully washed with water to remove excess stain and allowed to
   air dry before being photographed on a luminous board. The colonies
   were quantified if they contained more than 50 cells.

   CCK-8 assay was conducted in 96-well plates, with each well containing
   2000 cells in 100 µl of complete culture medium. Measurements were
   taken on day 1, 2, 3, 4, and 5 after incubation, respectively, by CCK-8
   assay kit (Beyotime). After adding 10 µl CCK-8 reagent to each well,
   cell culture proceeded for an additional 2 hours shielded from light.
   Subsequently, absorbance was measured at the wavelength of 450 nm using
   Microplate Spectrophotometer (BioTek Instruments Inc.).

   Transwell assay was performed to assess cell invasion. Transwell
   chambers (Corning) coated with Matrigel (Corning) were loaded with 200
   µl FBS-free culture medium containing 20,000 cells. Then, they were
   placed in 24-well plates, which contained 800 µl complete culture
   medium per well. After culturing for 24 hours, non-invaded cells
   remaining on the upper surface of the Transwell membrane were carefully
   removed using a cotton swab. The chambers were then washed with PBS,
   and the invaded cells on the lower surface were fixed with 4%
   paraformaldehyde for 20 minutes at room temperature. After fixation,
   the membranes were stained with 0.1% Crystal Violet solution for 20
   minutes. Excess stain was removed by washing with PBS, and the
   membranes were allowed to air dry before being carefully excised,
   mounted on slides, and photographed.

   Wound healing assay was used to evaluate cell migration. Cells were
   seeded into 6-well plates and allowed to grow until confluent, and a
   scratch was then made to create a wound. After 24 hours of incubation
   in FBS-free culture medium, the wound gap area was photographed and
   measured.

Flow Cytometry Assay

   Flow cytometry was performed to examine the apoptosis rate of cells.
   Cells were labeled with Annexin V-PE and 7-AAD (Multi Sciences)
   following the manufacturer's instructions. Then, detection was carried
   out using the LSRFortessa flow cytometer (BD Biosciences), and results
   were analyzed with CytExpert software.

Single-Cell RNA Sequencing and Data Preprocessing

   In this study, we collected scRNA-seq data from 19 ccRCC samples
   sourced from [134]GSE207493. The single-cell data from the 10x platform
   were preprocessed following the standard protocol outlined in the R
   package Seurat [135]^75. Doublets were eliminated using
   DoubletFinder[136]^76, and low-quality cells were filtered out based on
   specific criteria: cells with fewer than 200 or more than 6000 detected
   genes, or those with a mitochondrial gene ratio exceeding 20%.
   Additionally, low-expression genes, defined as those expressed in fewer
   than 5 cells, were also excluded. Cell cycle status was assessed using
   the CellCycleScoring() function, and potential interference from cell
   cycle effects was mitigated through the regression algorithm
   implemented in ScaleData(). Following data normalization and
   standardization, Harmony was employed to correct for batch effects, and
   the top 2000 highly variable genes were selected. Dimensionality
   reduction was executed using Principal Component Analysis (PCA), with
   the first 15 principal components chosen for t-SNE visualization. Cell
   clustering was performed using the FindClusters() function, with a
   resolution parameter set to 0.5. Finally, cell subpopulations were
   manually annotated based on classical marker genes and references from