Copy number loss of KDM5D may be a predictive biomarker for ATR inhibitor treatment in male patients with pulmonary squamous cell carcinoma

Abstract A limited number of patients with lung squamous cell carcinoma (SCC) benefit clinically from molecular targeted drugs because of a lack of targetable driver alterations. We aimed to understand the prevalence and clinical significance of lysine‐specific demethylase 5D (KDM5D) copy number loss in SCC and explore its potential as a predictive biomarker for ataxia‐telangiectasia and Rad3‐related (ATR) inhibitor treatment. We evaluated KDM5D copy number loss in 173 surgically resected SCCs from male patients using fluorescence in situ hybridization. KDM5D copy number loss was detected in 75 of the 173 patients (43%). Genome‐wide expression profiles of the transcription start sites (TSSs) were obtained from 17 SCCs, for which the cap analysis of gene expression assay was performed, revealing that upregulated genes in tumors with the KDM5D copy number loss are associated with ‘cell cycle’, whereas downregulated genes in tumors with KDM5D copy number loss were associated with ‘immune response’. Clinicopathologically, SCCs with KDM5D copy number loss were associated with late pathological stage (p = 0.0085) and high stromal content (p = 0.0254). Multiplexed fluorescent immunohistochemistry showed that the number of tumor‐infiltrating CD8+/T‐bet+ T cells was lower in SCCs with KDM5D copy number loss than in wild‐type tumors. In conclusion, approximately 40% of the male patients with SCC exhibited KDM5D copy number loss. Tumors in patients who show this distinct phenotype can be ‘cold tumors’, which are characterized by the paucity of tumor T‐cell infiltration and usually do not respond to immunotherapy. Thus, they may be candidates for trials with ATR inhibitors.

In cancers that lack targetable molecular alterations, targeting the well-defined phenotypic hallmarks could have therapeutic benefits [8].One such hallmark is genomic instability, which promotes genetic diversity and thereby drives the acquisition of multiple hallmark capabilities.DNA damage resulting from unabated replication (i.e.DNA replication stress) is a major source of genomic instability in cancers and precancerous lesions [9][10][11].Cellular response to DNA damage is mainly regulated by the ataxia-telangiectasia mutated-checkpoint kinase 2 and ataxia-telangiectasia and Rad3-related (ATR)-checkpoint kinase 1 pathways [12].Pharmacological inhibition of ATR exacerbates cell death and enhances antitumor activity of inhibitors of DNA topoisomerase I, a nuclear enzyme that suppresses genomic instability [13,14].Indeed, prostate adenocarcinoma cells with DNA replication stress induced by the loss of expression of male-specific histone-demethylase lysine demethylase 5D (KDM5D) are sensitive to an ATR inhibitor [15].This observation indicates that the loss of expression of KDM5D could serve as a biomarker to predict the efficacy of ATR inhibition, which may help select patients eligible for this new targeted therapy [16].KDM5D, located on the Y chromosome, is capable of demethylating the active transcriptional markers H3K4me2 and H3K4me3 [17].Despite extensive genome-wide analysis using next-generation sequencing, copy number alteration of genes on the Y chromosome has not been well studied.
Although ATR inhibitors have not yet been approved by the US Food and Drug Administration for lung cancer treatment, a phase II trial of an ATR inhibitor (berzosertib) in combination with a highly selective topoisomerase I inhibitor (topotecan) in patients with relapsed small cell lung cancer (SCLC), all of whom had evidence of disease progression after conventional treatment including chemotherapy or immunotherapy before participation, showed prolonged survival [18].Furthermore, preliminary signs of clinical efficacy of berzosertib in combination with gemcitabine or another ATR inhibitor, ceralasertib, in combination with carboplatin were observed in a small number of patients with advanced tumors including NSCLC, suggesting an early proof of concept for this drug class; however, these trials suggested that further clinical trials may be best undertaken in a patient population selected for potential biomarkers [19][20][21].
Here, we analyzed the ramifications of KDM5D copy number loss in SCC using molecular and clinicopathologic approaches.Histologically, we focused on cellular findings correlated with worse clinical outcomes in SCC, including nuclear size and tumor budding [22], and stromal findings to clarify clinicopathological characteristics of patients with KDM5D copy number loss.We aimed to investigate whether such loss is a suitable predictive biomarker for ATR inhibitor treatment in male patients with SCC in further clinical trials.

Study population
We screened the archives of the Department of Human Pathology, Juntendo University School of Medicine, Tokyo, Japan to identify patients who underwent complete lung tumor resection between January 2008 and January 2014.We obtained data on the following clinicopathological parameters: age, sex, smoking status, tumor size, lymphovascular invasion, lymph node and distant metastases, resection type, and mutation status of EGFR and KRAS.All tumors were resected at the Department of General Thoracic Surgery of Juntendo University Hospital.Pathological diagnoses were based on the 2021 WHO classification [23].To clarify clinicopathological features of SCC with KDM5D copy number loss, we excluded female patients in the present study because KDM5D is located on the Y chromosome so, theoretically, no female patients with SCC express KDM5D.The archives contained data from 194 male patients with SCCs.All patients were followed-up via regular physical and blood examinations, with mandatory radiography, computed tomography, or magnetic resonance imaging.Additionally, 65 male patients with SCLC who underwent surgical resection at Juntendo University Hospital between 2008 and 2020 were assigned to the control group to confirm the association of copy number loss of KDM5D, which was detected using fluorescence in situ hybridization (FISH) and droplet digital PCR (ddPCR).In both groups, all tissues were fixed in 10% buffered formalin and embedded in FFPE after routine processing, and these tumors were assembled into tissue microarrays (TMAs), using 1.5-2.0-mmcores sampled from one or two different representative areas of each FFPE tissue block.The study was performed in accordance with the Declaration of Helsinki.The patients were not required to provide consent for participation in the study, but we offered them an opt-out option, which allows patients to express a choice to object to their confidential patient information being used in this research.Individuals cannot be identified from the data presented herein.This study was approved by the

Histopathologic examination
We evaluated tumor budding, nuclear size, stromal content, and spread through alveolar spaces using representative slides of SCCs.Digital assessment approach for nuclear size in SCC For measuring nuclear area, TMA images which were archived as high-resolution whole-slide images were analyzed in QuPath (Version 0.4.3), an opensource software for digital image analysis [24].We analyzed the nuclear area of tumor cells, with a minimum of 100 cells, detecting and labeling with the QuPath application 'cell detection' and 'detection measurement' in each TMA core.We used QuPath's polygon tool for annotation of areas of interest after detecting TMA cores.After visually comparing different settings of the parameter for optimal detection of tumor cells, we used the default or proper settings for each case (supplementary material, Figure S1A).Overlapping nuclei that could not be measured were omitted.We also analyzed nuclear area of lymphocytes by selecting five TMA cores and averaging them as a control (supplementary material, Figure S1B).

FISH
FISH analysis of the TMAs was performed using a two-color KDM5D/X chromosome (CenX) probe mix, which was provided by the Molecular Cytogenesis Core at Memorial Sloan Kettering Cancer Center (New York, USA).From the 194 male patients with SCCs, TMAs of 173 tumors were available for FISH analysis.The probe mix consisted of a bacterial artificial chromosome clone containing full-length KDM5D (Yq11) (RP11-188C1 and RP11-204P21; labeled with red dUTP) and a centromeric repeat plasmid specific to CenX (pSV2x5; labeled with green dUTP).Probe labeling, tissue processing, hybridization, posthybridization washing, and fluorescence detection were performed according to standard laboratory procedures [15].Images were obtained with an all-in-one fluorescence microscope (BZ-X800; KEYENCE, Osaka, Japan) equipped with a Plan Apochromat Â40 objective (NA0.95,BZ PA40; KEYENCE).Red fluorescence was detected using a TexasRed filter (ex: 560/40 nm, em: 630/75 nm, dichroic: 585 nm, OP-87765; KEYENCE) and green fluorescence was detected using a GFP filter (ex: 470/40 nm, em: 525/50 nm, dichroic: 495 nm, OP-87763; KEYENCE).A minimum of 50-300 nuclei per case were evaluated.Copy number loss of KDM5D was defined as cases in which more than 90% of tumor cells showed absence of a red (KDM5D) signal, as previously described [15].

DNA extraction and digital droplet PCR
Genomic DNA was extracted from FFPE of SCLC tissues using the QIAamp DNA FFPE Tissue Kit (Qiagen, Venlo, Netherlands).Primers and probes used in ddPCR were as follows: PrimePCR™ ddPCR™ Expression Probe Assay, KDM5D (FAM probe, Human, Catalog ID: 10031252, Unique Assay ID: dHsaCPE50322209; Bio-Rad Laboratories, Hercules, CA, USA) and PrimePCR™ ddPCR™ Copy Number Assay, SPIN4 (HEX probe, Human, Catalog ID: 10031243, Unique Assay ID: dHsaCP2506729; Bio-Rad Laboratories).Probes were designed on KDM5D locus for FAM reporters and reference sites for HEX reporters (SPIN4: X chromosome).The QX200 system (Bio-Rad Laboratories) was used in the study.The reactions were performed with 20-μl mixtures of 10 μl of 2Â ddPCR Supermix for probes (no dUTP) (Bio-Rad Laboratories), primers (900 nmol/l), probes (250 nmol/l), and 50 ng of the genomic DNA sample.The QX200 droplet generator (Bio-Rad Laboratories) converted each reaction mix to droplets.Droplet-partitioned samples were then transferred to a 96-well plate, sealed, and cycled in a C1000 Touch thermal cycler (Bio-Rad Laboratories) under the following cycling protocol: 95 C for 10 min, followed by 40 cycles at 94 C for 30 s, 60 C for 60 s, and a 10-min incubation at 98 C.
The products were then read on the QX200 droplet reader (Bio-Rad Laboratories).At least two negative control wells with no genomic DNA template were included in each run.The data analysis was conducted with QuantaSoft droplet reader software v1.7.4 (Bio-Rad Laboratories).KDM5D/SPIN4 ratio <0.1 was used as an indicator of copy number loss of KDM5D.

Multiplexed fluorescent immunohistochemistry
Multiplexed fluorescent immunohistochemistry was performed using the tyramide signal amplification method with an Opal IHC kit (Akoya Biosciences, Marlborough, MA, USA) according to the manufacturer's instructions.Tissue sections, of thickness 4 μm, were cut from FFPE tumor specimens and then baked at 60 C onto adhesive glass slides for 30 min before deparaffinization.The primary antibodies used were anti-human CD8 (clone C8/144b; DAKO), anti-human T-bet (Santa Cruz Biotechnology, Dallas, TX, USA), and cytokeratin (clone AE1/AE3; DAKO).A high-pH target-retrieval solution (DAKO) was used for antigen retrieval, and immunoactive buffer (Matsunami Glass, Osaka, Japan) was used for antibody stripping.Opal 520, 540, and 650 were used for labeling CD8, T-bet, and cytokeratin, respectively.A horseradish peroxidaselabeled secondary detection system (EnVision Plus, DAKO) was used as a catalyst for fluorophoreconjugated tyramide.Heating at 95 C for 20 min was performed for primary antigen unmasking and antibody stripping after each fluorescent labeling.

Image analysis and quantification
Multiplexed fluorescent-labeled images of three randomly selected fields (669 μm Â 500 μm each) were captured with an automated imaging system (Vectra ver.3.0; Akoya Biosciences).An image analyzing software program (InForm; Akoya Biosciences) was used to segment cancer tissue into cancer cell nests (intratumoral) and the framework (stromal) region, and to detect immune cells with CD8 and T-bet expression.
Training sessions for tissue segmentation and phenotype recognition were repeated until the algorithm reached the level of confidence recommended by the program supplier (at least 90% accuracy) before performing the final evaluation.

Bioinformatic analysis of the cap analysis of gene expression data
Quantitative transcription start site (TSS)-level expression profiles were obtained from the 17 SCCs, for which the cap analysis of gene expression (CAGE) assay was performed as previously described [25].In brief, the CAGE reads were aligned to the reference genome (hg19) with a high mapping quality of ≥20.
The aligned CAGE reads were counted in each region of the FANTOM5 robust peaks [26], a reference set of TSS regions, as raw signals for promoter activity.The expression levels of individual TSSs were quantified as counts per million.Inactive TSS regions, with counts per million ≤1 in more than 77% of samples, were filtered out [27].A differential expression analysis was performed using R (version 4.2.1;https://www.r-project.org/;R Foundation for Statistical Computing, Vienna, Austria) with the edgeR package [28].Differentially expressed TSSs were defined as fold change cutoff >2 and false discovery rate <0.05.To characterize genes that were correlated with KDM5D copy number loss that was detected using FISH, gene set enrichment analysis was performed using the g:Profiler website [29], which includes gene set databases such as Gene Ontology [29,30], TRANSFAC [30], and The Human Protein Atlas [31].The distances between the samples for the selected genes were calculated in terms of Euclidean distances for counts per million, and average linkage clustering was performed.

Statistical analysis
The correlations between the results of ddPCR and FISH were assessed using Spearman's rank correlation.Categorical variables were analyzed using Fisher's exact or chi-square test.To determine prognostic significance, we performed Kaplan-Meier survival analysis.These statistical analyses were performed using GraphPad Prism ® version 9.0 (GraphPad Software, San Diego, CA, USA).Results with p < 0.05 were considered statistically significant.Associations between KDM5D and other genes were assessed using Spearman's rank correlation.A series of statistical analyses were performed in R with the edgeR package [28].

Clinicopathological characteristics of the study cohort
The clinicopathological characteristics of the 173 male patients with SCC whose KDM5D copy number was examined are shown in Table 1.The median age of the patients with SCC was 70.5 years; 159 patients (91.9%) were heavy smokers (Brinkman index >400) and the remaining 14 patients were never or light smokers (Brinkman index ≤400).Histologically, 117 (67.6%) and 56 (32.4%) cases were keratinizing and nonkeratinizing type SCCs, respectively (Figure 1A,B).Immunohistochemically, no tumor expressed TTF-1, whereas all tumors exhibited diffuse positivity for p40 (Figure 1C).Epstein-Barr virus-encoded RNA was positive in a patient with nonkeratinizing type SCC, which led to the pathological diagnosis of lymphoepithelial carcinoma (Figure 1D).Furthermore, 71 (40%), 58 (33%), and (25%) tumors had no, low, and high expression of PD-L1, respectively.Genetically, five tumors (2.9%) harbored KRAS mutations.Among five patients with SCCs harboring

Copy number loss of KDM5D in lung carcinomas
First, as a control, copy number loss of KDM5D was evaluated in SCLCs using ddPCR, a method that can absolutely quantify the KDM5D copy number without the need for standard curves.In SCLC, the median levels of KDM5D and SPIN4 were 608 and 2256 (copies/μl), respectively (supplementary material, Table S1).Copy number loss of KDM5D (KDM5D/SPIN4 ratio <0.1) was identified in 10 tumors (15%) using ddPCR, and there was correlation between the results of the copy number loss of KDM5D by ddPCR and FISH (p < 0.0001) (supplementary material, Figure S2).We then determined the frequency of copy number loss of KDM5D in TMA samples of SCCs using FISH (Figure 1E,F).Among the 173 patients with SCC, KDM5D copy number loss was detected in 75 SCCs (43%).

CAGE data analyses reveal that KDM5D copy number loss correlates with cell cycle in SCC
To further explore the mechanism by which copy number loss of KDM5D contributes to the phenotype in SCC, we performed bioinformatic analysis of comprehensive transcriptome data using the CAGE sequence.In total, 44,529 TSSs were identified (supplementary material, Table S2).First, we compared the expression of TSS region 1 of KDM5D in tumors with and without KDM5D copy number loss.The results indicated that the expression of TSS region 1 of KDM5D was significantly lower in tumors with KDM5D copy number loss than in wild-type tumors (p < 0.0001).Next, we comprehensively investigated the TSSs of whole genes that were positively or negatively correlated in tumors with KDM5D copy number loss.In the datasets, 208 TSSs were differentially expressed between SCCs with and without KDM5D copy number loss (Figure 2A and supplementary material, Table S3).Gene set enrichment analysis revealed that upregulated genes in tumors with the KDM5D copy number loss are associated with the cell cycle terms 'mitotic spindle organization', 'microtubule cytoskeleton organization involved in mitosis', and 'nuclear division', which is consistent with the finding that the loss of KDM5D expression is associated with acceleration of the cell cycle and mitotic entry [15].On the contrary, downregulated genes in tumors with KDM5D copy number loss were associated with the immune response terms

of 12
A Ura et al 'adaptive immune response', 'immune effector process', and 'activation of immune response' (Figure 2B).We further evaluated the TSS-level expression profiles of the genes on the Y chromosome.The analysis involved 40 TSSs of 18 genes, including KDM5D.Overall, the expression of genes on the Y chromosome, especially around KDM5D including TXLNG2P and TTTY14, was significantly lower in tumors with KDM5D copy number loss than in wild-type tumors (Figure 2C).Lastly, we evaluated the TSS-level expression profiles of KDMs in tumors with and without KDM5D copy number loss.In addition to KDM5D, 50 TSSs from 14 genes were identified, namely, KDM1A, KDM1B, KDM2A, KDM2B, KDM3A, KDM3B, KDM4A, KDM4B, KDM4C, KDM4D, KDM5A, KDM5B, KDM5C, and KDM6A.
Overall, there were no significant associations between the expression of KDM5D and other genes of the KDM family at the TSS expression level (Figure 2D); however, there were significant inverse correlations in the expression of subsets of TSSs, including the higher expression of TSS region 1 of KDM1A (r = À0.610) and KDM1B (r = À0.557), and TSS region 3 of KDM1B (r = À0.828), in SCCs with KDM5D copy number loss (Figure 2E and supplementary material, Table S4).

Clinicopathological characteristics of SCC with KDM5D copy number loss
To explore the effects of clinicopathological characteristics of SCC with KDM5D copy number loss, the   1.SCCs with KDM5D copy number loss were significantly associated with lymph node metastasis (p = 0.0107) and late pathological stage (p = 0.0085).There was no significant correlation between copy number loss of KDM5D and other clinicopathological features such as age, smoking index, tumor size, pathological grade, lymphovascular invasion, KRAS mutation status, and PD-L1 expression status.We further evaluated histological features in SCC with KDM5D copy number loss.SCCs with KDM5D copy number loss were significantly associated with high stromal content ( p = 0.0254) (Figure 3A,B).There were no significant differences in other histological features including, nuclear area (supplementary material, Figure S1C), existence of tumor budding, and spread through alveolar spaces between tumors with KDM5D copy number loss and wild-type tumors.High expression of PD-L1 was identified in 23 (31%) patients with KDM5D copy number loss; however, a comparison between copy number loss of KDM5D and wild-type tumors revealed no significant differences in PD-L1 expression (Figure 3C).Multiplexed fluorescent immunohistochemistry analysis showed that CD8 + /T-bet + T cells equally infiltrated the stroma of tumors with and without copy number loss of KDM5D.In contrast, CD8 + /T-bet + T cells infiltrated tumors with KDM5D copy number loss less than wildtype tumors, but these differences did not reach statistical significance (Figure 3D-F).The results of the transcriptome and multiplexed fluorescent immunohistochemistry analyses are summarized in supplementary material, Table S5.

Clinical outcomes
With respect to clinical outcomes, the median follow-up period after surgery for all patients was 1,482 days.In our cohort, the overall survival (OS) and recurrencefree survival (RFS) rates were significantly associated with the pathological stage (OS, p = 0.0284; RFS, p = 0.0029) (supplementary material, Figure S3A,B).Patients with KDM5D copy number loss tended to have worse OS ( p = 0.1220) and RFS rates ( p = 0.1210), but these differences were not

of 12
A Ura et al significant (Figure 4A,B).Regarding other clinicopathological parameters, the OS and RFS rates were significantly associated with lymphovascular invasion (OS, p = 0.0002; RFS, p < 0.0001) in patients with SCC (Figure 4C,D), whereas there were no significant differences in histological types (supplementary material, Figure S3C,D) and expression level of PD-L1 (supplementary material, Figure S3E,F).In this cohort, 51 patients developed recurrence.Among them, five patients were treated with immune checkpoint inhibitors (ICIs): four patients died of disease (two patients with KDM5D copy number loss and low expression of PD-L1; one patient with KDM5D copy number loss and high expression of PD-L1; one patient with KDM5D wild type and low expression of PD-L1) and one patient with KDM5D copy number loss and high expression of PD-L1 was alive with disease (supplementary material, Figure S4).

Discussion
Enrichment of repetitive and palindromic sequences on the Y chromosome makes bioinformatic studies of copy number alterations on the Y chromosome challenging [15].Thus, in the present study we used FISH, the gold standard for detecting copy number changes in tumor cells, and successfully determined that approximately 40% of the patients had KDM5D copy number loss in SCC.Furthermore, the transcriptome analysis results indicated that the expression of TSS region 1 of KDM5D was reduced in SCCs with KDM5D copy number loss, suggesting that copy number loss of KDM5D could lead to the absence of KDM5D transcripts and the subsequent loss of KDM5D protein expression in SCC.Reduced expression of KDM5D in human cancer cells has been reported in various organs such as the prostate [15], kidney [32], stomach [33], and lung [34].Notably, the downregulation of KDM5D induces an aggressive phenotype in prostate and gastric carcinomas [15,35].Male patients with KDM5D deficiency in NSCLC tumors show a substantially increased risk of death [34].These findings are consistent with our clinicopathological results, showing patients with KDM5D copy number loss tended to exhibit worse OS.Histologically, stromal content was exclusively different in SCCs with and without KDM5D copy number loss, and there were no apparent aggressive histological features in SCC with KDM5D copy number loss.This suggests that SCC with KDM5D copy number may be characterized by its unique tumor microenvironment (TME).Moreover, activated ATR signaling with the loss of KDM5D expression can be exploited to elicit synthetic lethality of ATR inhibition in cancer cells [15].The discovery of clinically targetable genomic alterations in NSCLC has revolutionized the treatment of patients with NSCLCs.Furthermore, such patient stratification has started to permeate therapeutic strategies for SCLCs based on the expression of transcription factors including ASCL1, NEUROD1, and POU2F3 [36].However, ongoing clinical trials to advance the development of targeted therapies for SCC are limited.Thus, patients with SCCs harboring copy number loss of KDM5D may be candidates for treatment with ATR inhibitors.
Our transcriptome data demonstrated that TSSs of genes on the Y chromosome around KDM5D were significantly lower in tumors with KDM5D copy number loss than wild-type tumors, implying that somatic loss of the Y chromosome including the KDM5D locus may occur in tumors with KDM5D copy number loss.Similarly, a previous study revealed a deficiency of KDM5D with low expression of the other eight Y chromosome transcripts because of a partial somatic deletion in NSCLC [34].Consequently, our findings may be affected by not only the copy number loss of KDM5D, but also aberration in other genes on the Y chromosome; however, such genes play a limited role in carcinogenesis [37].Alternatively, SCCs with KDM5D copy number loss exhibited high expression of TSS region 1 of KDM1A.Thus, the interplay of KDM5D and KDM1A may have resulted in the clinicopathological findings of the present study, because the final effect on chromatin structure and genome function in most cancers is not strictly dependent on one histone modification [38].
Our multiplex immunohistochemical results along with transcriptome data indicate that SCCs with copy number loss of KDM5D can be 'cold tumors', which are characterized by the paucity of tumor T-cell infiltration and usually do not respond to immunotherapy.This observation is inconsistent with that of previous studies, which showed that dysfunction of the DNA damage response pathway in tumor cells is capable of activating an immune response possibly by increasing the level of neoantigen and tumor immunogenicity [39,40].Currently, the molecular mechanism underlying the paucity of tumor T-cell infiltration within tumors and subsequent downregulation of genes associated with the immune response in SCCs with copy number loss of KDM5D is unclear.However, lossof-function RNF45 mutations reduce CC chemokine ligand (CCL4) levels and prevent CD8 + T cell infiltrations, leading to noninflamed TME in high-frequency microsatellite instability colorectal adenocarcinomas [41].Furthermore, some of the neoantigenic mutations, particularly driver mutations, can paradoxically act as antitumor immunity suppressors through the original gene function [41].Such 'paradoxical neoantigenic mutations' may occur in SCCs with KDM5D copy number loss.Alternatively, our results indicate that the histology of stroma in SCCs with and without KDM5D copy number loss is different, suggesting that unique cellular constituents within the TME in SCCs with KDM5D copy number loss affect the distinct immunological attributes.Further studies with other approaches such as spatial transcriptomics will be needed to elucidate the regulation of the TME in SCCs with KDM5D copy number loss.
A limitation of this study is that we detected KDM5D copy number loss in male patients with SCC using FISH.This method has been successfully used to identify prostate adenocarcinoma with KDM5D copy number loss [15].However, the association of copy number loss of KDM5D and loss of protein expression of KDM5D using immunohistochemistry was not evaluated in FFPE specimens.The immunohistochemical analysis could be useful in clinical settings to select patients for this new targeted therapy.Therefore, development of a KDM5D-specific antibody is needed.Finally, only five patients with KDM5D copy number loss were treated with ICI in the present study.Thus, it is still unclear whether patients with KDM5D copy number loss show positive or negative responses to treatment with ICI.However, predictive markers of ICI efficacy have been gradually explored from the expression of intermolecular interactions within tumor cells to the expression of various molecules and cells in the TME, and have been extended to the exploration of circulating and host systemic markers [42].Thus, further clinical analyses using multifactorial synergistic predictive markers are needed to determine the response to ICI in patients with KDM5D copy number loss and patient stratification for determining optimal therapy for lung cancer.
In summary, FISH-based KDM5D screening revealed that approximately 40% of male patients with SCC have copy number loss of KDM5D.Although additional studies are required to clarify the clinicopathological effects of KDM5D loss in SCC through external validation, patients whose tumors show this distinct phenotype, especially in a subgroup of patients without PD-L1 expression, could benefit from targeted therapies involving ATR inhibitors.

2 of 12 A
Ura et al Institutional Review Board of Juntendo University (Approval No. 2019123).

5 of 12 KDM5D
loss in lung squamous cell carcinoma © 2023 The Authors.The Journal of Pathology: Clinical Research published by The Pathological Society of Great Britain and Ireland and John Wiley & Sons Ltd.J Pathol Clin Res 2024; 10: e350

Figure 1 .
Figure 1.Pathological findings of SCC.Representative hematoxylin and eosin staining of (A) well-differentiated and (B) poorly differentiated SCC.(C) Representative image of diffuse immunohistochemical staining of p40 in a poorly differentiated SCC.(D) A case of lymphoepithelial carcinoma was Epstein-Barr virus-encoded RNA positive, as determined using in situ hybridization.(E and F) FISH analysis using a two-color lysine-specific demethylase 5D (KDM5D) (Yq11/red)/CenX (green) probe mix.Representative images showing (E) loss of KDM5D in a SCC and (F) KDM5D-positive SCC.

Figure 2 .
Figure 2. TSS-level expression of KDM5D in SCC.(A) Two distinct clusters were observed in clustering based on the KDM5D copy number status.Red and green indicate loss of KDM5D and wild-type tumors, respectively.(B) Top 20 terms of gene set enrichment analysis, performed using g:Profiler.The upregulated genes in SCC with copy number loss of KDM5D were associated with the term 'cell cycle', whereas downregulated genes in SCC with copy number loss of KDM5D were associated with 'immune response'.(C) TSS-level expression profiles of the genes on the Y chromosome.Genes are listed in a linear order on the chromosome.The expression of TSSs of genes on the Y chromosome was lower in tumors with KDM5D copy number loss (red) than in wild-type tumors (green).(D) Hierarchical clustering of Spearman's rank correlation among all TSSs of KDM family genes.(E) Scatter plots and Spearman's rank correlation of TSSs of KDM5D, KDM1A, and KDM1B.Note the inverse correlations between the expression of TSS region 1 of KDM1A (r = À0.610) and KDM1B (r = À0.557), and TSS region 3 of KDM1B (r = À0.828).

7 of 12
KDM5D loss in lung squamous cell carcinoma © 2023 The Authors.The Journal of Pathology: Clinical Research published by The Pathological Society of Great Britain and Ireland and John Wiley & Sons Ltd.J Pathol Clin Res 2024; 10: e350 clinicopathological factors were evaluated in the 173 SCCs.The clinicopathological characteristics of 75 SCCs with KDM5D copy number loss are summarized in Table

Figure 3 .
Figure 3. Clinicopathological characteristics of SCC with copy number loss of KDM5D.Tumor stromal content was categorized as (A) high (≥50% of the whole tumor area) and (B) low (<50%).(C) Representative image of high expression of PD-L1 in SCC with KDM5D copy number loss.(D and E) Representative staining of multiplex immunohistochemistry of cytokeratin (yellow), CD8 (green), and T-bet (red) in SCC (D) with and (E) without KDM5D copy number loss.(F) The number of infiltrating CD8 + /T-bet + T cells in stroma was similar between SCC with KDM5D copy number loss and wild-type tumors (left), whereas the number of infiltrating CD8 + /T-bet + T cells within tumor in SCC with KDM5D copy number loss was lower than that in wild-type tumors (right).

Figure 4 .
Figure 4. Kaplan-Meier curves of OS and RFS of patients with SCC harboring KDM5D copy number loss.OS and RFS according to the status of (A and B) KDM5D copy number and (C and D) existence of lymphovascular invasion.

Table 1 .
Clinicopathological characteristics of male patients with lung squamous cell carcinoma