Loss of chromosome 9p21 is associated with a poor prognosis in adenosquamous carcinoma of the pancreas

Abstract Adenosquamous carcinoma of the pancreas (ASCP) is a rare histological subtype of pancreatic cancer with a poor prognosis and a high metastasis rate. However, little is known about its genomic landscape and prognostic biomarkers. A total of 48 ASCP specimens and 98 pancreatic ductal adenocarcinoma (PDAC) tumour specimens were sequenced to explore the genomic landscape and prognostic biomarkers. The homozygous deletion of the 9p21.3 region (including CDKN2A, CDKN2B, and MTAP) (9p21 loss) occurred in both ASCP and PDAC, and a higher frequency of 9p21 loss was observed in ASCP (12.5% vs 2.0%, P = 0.022). Notably, 9p21 loss was significantly associated with poor disease-free survival (DFS) in ASCP patients (mDFS (Median DFS) = 4.17 vs 7.33 months, HR (Hazard Ratio) = 3.70, P = 0.009). The most common gene alterations in patients with ASCP were KRAS (96%), TP53 (81%), CDKN2A (42%), SMAD4 (21%), CDKN2B (13%), and FAT3 (13%). The mutation rates of ACVR2A (6.25% vs 0%), FANCA (6.25% vs 0%), RBM10 (6.25% vs 0%), and SPTA1 (8.33% vs 1.02%) were significantly higher in ASCP than in PDAC. In conclusion, we have comprehensively described the genomic landscape of the largest cohort of ASCP patients to date and highlight that 9p21 loss may be a promising prognostic biomarker for ASCP, which provides a molecular basis for prognosis prediction and new therapeutic strategies for ASCP.


Introduction
Adenosquamous carcinoma of the pancreas ( ASCP ) , a rare histological subtype of pancreatic cancer, 1 , 2 constitutes ∼1-4% of all pancreatic cancers. 3 , 4According to the World Health Organization classification of pancreatic cancers, ASCP is characterized by a variable proportion of adenocarcinoma and squamous carcinoma components, accounting for at least 30% of tumour cells.Similar to pancreatic ductal adenocarcinoma ( PDAC ) , surgical excision is the only radical cure for patients . 5Since ASCP gener all y occurs in the body of the pancreas and the tumours ar e usuall y lar ger than those of PDA C , 6 sur gical r esection is more feasible for ASCP, with an R0 resection rate of 58%-67%. 2 , 7 Ho w e v er, ASCP tends to be more aggressive and has a poorer prognosis than PD AC .Moreover, it has a higher metastasis rate with unfavourable clinical outcomes. 4 , 6Ther efor e, finding ne w ther a peutic str ategies to impr ov e the pr ognosis and surviv al of patients with ASCP is essential.
To better understand the molecular pathology of ASCP, researc hers hav e a pplied next-gener ation sequencing ( NGS ) to examine the genomic landscape .T hese examinations ha ve increased our understanding of the molecular basis of ASCP.Fang et al .investigated 17 ASCP specimens using whole-exome sequencing and determined that KRAS , TP53 , and SMAD4 are the most fr equentl y m utated genes.Multiple 3p r egions in ASCP sho w ed significant cop y number losses compared with those in PDA C . 8Due to the rarity of this subtype, most of the available studies have been limited by small sample sizes.Mor eov er, since it has a poorer prognosis than PD AC , it is urgent to seek optimal tr eatments, including c hemother a py, r adiother a py, tar geted ther a py, and imm unother a py.The scarcity of tumour samples suitable for high-throughput sequencing analyses has hindered genomic studies of this fatal subtype .T hus , information about the genomic landscape, potential oncogenic driver mutations, pr ognosis-r elated gene alterations of ASCP, and their relationship with PDAC is limited.
In the present study, we investigated the genomic landscape and prognostic biomarkers in ASCP.Understanding the underlying factors contributing to poor survival might be beneficial to the treatment of ASCP.We also compared ASCP and PDAC data to explore the molecular differences between these two subtypes.

Ethics appro v al and consent to participa te
Ethics a ppr ov al was gr anted by the Ethics Committee of the First Affiliated Hospital of Xi'an Jiaotong University, based on the World Medical Association Declaration of Helsinki ( No. XJTU1AF2021LSK-053 ) .

Patients and tumour samples
Written informed consent was obtained from all participants or their legal guardian/next of kin, allowing for the collection and use of their tumor samples.Tumour samples from 48 patients with ASCP and 98 patients with PDAC were collected at the First Affiliated Hospital of Xi'an Jiaotong Univ ersity fr om 2012 to 2020.P athological dia gnoses of ASCP wer e confirmed according to the 5 th edition of the World Health Organization Classification of Tumours and ensured that both adenocarcinoma and squamous car cinoma components w er e contained in eac h sample.P athological sections were cut from formalin-fixed paraffin-embedded ( FFPE ) tumour blocks for subsequent use.Both tumour tissue and matched normal blood samples were collected from each patient.Before sequencing, the samples w ere review ed b y tw o experienced pathologists who e v aluated the FFPE tumour samples for tumour cell percentages in the sequencing laboratory.Complete medical records included patient age, gender, pathological reports, operation time, surgical approach, and survival information.

Sequencing data analysis and tumour mutational burden
Genomic mutations were identified using an NGS-based YuanSu™ ( OrigiMed, Shanghai, China, a College of American Pathologistsaccredited and Clinical Laboratory Improvement Amendmentscertified laboratory ) gene panel, covering all coding exons of 450 cancer-related genes in solid tumours.At least 50 ng of cancer tissue DN A w as extr acted fr om eac h 40 mm FFPE tumour sample using a DNA extraction kit ( QIAamp DNA FFPE Tissue Kit; Qiagen, Hilden, Germany ) according to the manufacturer's protocols.Genomic alterations ( GAs ) were identified using the procedure described by Cao et al .: single nucleotide variants ( SNVs ) were identified using MuTect v1.7, and insertion deletions ( indels ) were identified using PINDELV0.2.5 . 9According to the ExAC, 1000 Genomes, dbSNP build 155, and ESP6500SI-V2 databases, variants with population frequencies > 0.1% wer e gr ouped as singlenucleotide pol ymor phisms ( SNPs ) and excluded from further analysis .T he remaining variants were annotated using ANNOVAR and SnpEff v.3.0.Copy number v ariations ( CNVs ) wer e identified using Control-FREEC ( v9.7 ) with the following parameters: window = 50 000 and step = 10 000.Gene fusions were detected using an in-house pipeline.Gene r earr angements wer e assessed using Integr ativ e Genomics Viewer . 10Our sequencing data were compared with those of The Cancer Genome Atlas ( TCGA ) database to eliminate bias, and it was found that the m utation fr equency of our sequencing data was basically the same as that of the TCGA-PAAD cohort ( Fig. S1, see online supplementary material ) .
Tumour mutational burden ( TMB ) was estimated by counting coding somatic mutations, including SNVs and indels, per megabase of the sequence examined for each patient.As cutoffs for categorizing the TMB status of ASCP have not been defined, we used the criteria established in a pr e vious study for other types of tumours . 11In this study, the median TMB was 2.75 mutations/megabase ( muts/Mb ) ; TMB-L was defined as < 10 muts/Mb and TMB-H as ≥10 muts/Mb of sequenced DNA.

Differential expression and functional enrichment analyses
In the TCGA-PAAD cohort, the Wilcoxon test was utilized to identify differ entiall y expr essed genes between the 9p21 loss and 9p21 wild type ( WT ) gr oups, and the r esults wer e visualized in a volcano plot.The Gene Ontology ( GO ) annotation and Kyoto Encyclopedia of Genes and Genomes ( KEGG ) pathway analysis were performed with DAVID 6.8 bioinformatics resources and plotted by using the ggplot2 pac ka ge in R. Lists of genes involved in various signalling pathways wer e acquir ed fr om the KEGG database ( http:// www.genome.jp/kegg/ pathway.html ) and are provided in Table S1, see online supplementary material.Gene set enrichment analysis ( GSEA ) was performed using GSEA software ( version 4.1.0) .To e v aluate the tumour imm une micr oenvir onment, singlesample GSEA was carried out by using the "GSVA" R pac ka ge.

Imm unohistoc hemistry
Imm unohistoc hemical ( IHC ) staining of FFPE tissues was performed using anti-pr ogr ammed death-ligand 1 ( PD-L1 ) antibodies ( clone 22C3, Cat.M3653, DAKO ) .Dilutions ( 22C3; 1 : 50 ) of the primary antibodies were used for antigen detection.All slides were counterstained with hematoxylin.The PD-L1 le v el was reported as the combined positive score ( CPS ) , which was defined as the number of PD-L1-positive cells divided by the total tumour cells, multiplied by 100.CPS < 1 was defined as PD-L1 negative; 1 ≤ CPS < 10 was defined as PD-L1 with low expression; and CPS ≥ 10 was defined as PD-L1 with high expression.Genes were captured and sequenced, with a mean depth of 800 ×, using Illumina NextSeq 500 ( Illumina, CA, USA ) as described by Frampton et al . 12IHC for p16 status was also conducted using anti-human p16 mouse monoclonal antibody ( clone MX007 ) in tissues from patients with 9p21 loss and 9p21 WT.

Sta tistical anal yses
All analyses were performed using R software version v 4.0.3.Fisher's exact test was used for the association analysis of categorical variables.Student's t test and the Wilcoxon rank-sum test were used for the association analysis of normally and nonnormally distributed data, respectively.The Kruskal −Wallis test was used to analyse the association between groups of nonpar ametric data.A Cox pr oportional hazards r egr ession model was used to analyse the relationship between predictor variables and the time to an e v ent, suc h as ov er all surviv al ( OS ) or diseasefr ee surviv al ( DFS ) .Statistical significance was set at P < 0.05.

Pa tient char acteristics
A total of 48 patients with ASCP and 98 patients with PDAC were recruited for this study.The clinical characteristics of the patients ar e pr o vided in Table 1 .Of 48 patients , 28 ( 58.3% ) were male and 20 ( 41.7% ) were female .T he mean age of the patients w as 58 y ears ( r anging fr om 33 to 83 years ) .The median TMB was 2.75 muts/Mb ( r anging fr om 0.5 to 21.7 m uts/Mb ) , wher eas a TMB v alue > 10 muts/Mb was detected in only one patient.The ratio of males to females among the 98 patients with PDAC was ∼1 : 1, distal metastasis was the most common metastasis type, and tumour stages II and IV were most common.Detailed information on the pr ognosis and tr eatment of ASCP is shown in Table S2, see online supplementary material.

Histological pathology of ASCP
ASCP is defined as a lesion containing both adenocarcinoma and squamous carcinoma within the same tumour ( Fig. 1 A-C ) .Adenocarcinoma forms incomplete or complex glands, with a high nuclear-to-cytoplasmic ratio and prominent nucleoli.The squa-

9p21 loss may be a new predictor of a poor prognosis in ASCP
Mutation co-occurrence could provide information for drug combination ther a py and medication instruction, ther efor e gene comutation was analysed ( Fig. 4 A ) .We obtained 19 pairs of related mutations, including 15 co-occurring mutation pairs and 4 m utuall y exclusiv e m utation pairs.KRAS m utations and m utations in ARID2 and SMAD4 were mutually exclusive, and TP53 m utations wer e m utuall y exclusiv e with m utations in MDM2 and BRAC2 .Ho w e v er, se v er al m utations, suc h as ARID1A/SPINK1 , CDKN2A/CDKN2B , MTAP/CDKN2B , FRS2 / MDM2 , BAP1 / SPTA1 , and CDKN2B/MTAP, co-occurred to a significant degree.In particular, MTAP , CDKN2A and CDKN2B were adjacent at the 9p21.3region, and CNV analysis sho w ed a loss in this region in our cohort ( Fig. 2 C ) .We r e vie wed the original sequencing data from patients who carried 9p21 loss and show the somatic copy number alterations in Fig. 4 B and Fig. S4 ( see online supplementary material ) .To further assess the occurrence of 9p21 loss, we performed m utual exclusivity anal ysis in PAAD fr om TCGA.Co-deletion of CDKN2A / B and MTAP was fr equentl y observ ed in 183 patients with CNV ( 52/183 ) .

Prognostic analysis of clinical characteristics
Of the 48 patients investigated during our study, 40 had a complete follow-up with the best objectiv e r esponse assessment as well as DFS and OS e v aluations .T he correlations between clinical c har acteristics and pr ognosis wer e anal ysed, r e v ealing that tumour stage, tumour location, and distal metastasis were significantl y corr elated with DFS ( Fig. S6A-C, see online supplementary material ) .Lymph node metastasis was correlated with DFS, but the correlation did not reach statistical significance ( Fig. S6D ) .Univ ariate anal ysis confirmed that distal metastasis was a prognostic factor for DFS and OS ( Table S3, see online supplementary material ) .As observed, the higher the tumour stage and the larger the size, the worse the prognosis .Moreo ver, surgery was significantly related to DFS and OS ( Fig. S6E and F ) .T hus , the prognosis of patients after radical surgery could be significantly impr ov ed.Furthermor e, we conducted univ ariate and m ultiv ariate Cox prognostic models for DFS and OS ( Table S3 and Fig. 5 ) .Univariate Cox proportional hazard analysis sho w ed that several clinical c har acteristics wer e r elated to surviv al: these included 9p21 loss, tumour location, operation, tumour size, lymph node metastasis, and distal metastasis ( Table S3 ) .Multivariate Cox proportional hazard analysis showed that 9p21 ( loss/WT ) ( P = 0.012 ) , gender ( female/male ) ( P = 0.046 ) , CA19-9 ( 37 < CA19-9 ≤ 100/ ≤ 37 ) ( P = 0.028 ) , and distal metastasis ( yes/no ) ( P < 0.001 ) ar e highl y correlated with DFS ( Fig. 5 A ) , and 9p21 ( loss/WT ) ( P = 0.029 ) and C A19-9 ( 37 < C A19-9 ≤ 100/ ≤ 37 ) ( P = 0.019 ) were associated with OS ( Fig. 5 B ) .These results suggest that 9p21 loss could be an independent prognostic factor for OS and DFS.

9p21 loss alters gene expression and functional enrichment
We performed differential expression analysis on RNA sequencing data from the TCGA-PAAD cohort.P value < 0.05 and | fold change | > 1 were considered the cut-off criteria based on the TCGA cohort.A total of 272 differ entiall y expr essed genes ( DEGs ) wer e identified, including 179 downregulated genes and 93 upregulated genes ( Table S4, see online supplementary material ) .These results were visualized by a volcano plot ( Fig. S7A, see online supplementary material ) .Importantly, we found that CDKN2A and CDKN2B were significantl y downr egulated in the top 30 DEGs in 9p21 loss patients compared to 9p21 WT patients ( Fig. S7B ) .Furthermore, the expr ession le v els of CDKN2A , CDKN2B , and MTAP were depicted using a violin plot ( Fig. S7C ) .Collectiv el y, these findings provide compelling evidence that 9p21 loss is associated with the downregulation of CDKN2A and CDKN2B gene expression.
The CDKN2A gene ( c hr omosome 9p21 ) encodes the protein p16 INK4a ( also known as p16 ) through alternative exon usage, and this gene is the second most commonly inactivated tumour suppressor gene in cancer and is lost in the majority of chordomas . 13o provide experimental evidence complementing our omics data analysis, we further investigated the protein expression of p16 in tumour tissues with 9p21 loss and 9p21 WT using IHC.Our results sho w ed negative p16 expression in samples harbouring 9p21 loss, while positive p16 expression was observed in 9p21 WT samples ( Fig. 6 ) .Furthermor e, KEGG anal ysis sho w ed that the DEGs w ere involved in several important pathways that promote tumour progression, including the PI3K-Akt signalling pathway, p53 signalling pathway, extracellular matrix ( ECM ) -receptor interaction, and focal adhesion ( Fig. S7D and Table S5, see online supplementary material ) .This finding suggests that 9p21 loss may be associated with a poor prognosis.
We also observed that 9p21 loss was not strongly associated with TMB in ASCP and PDAC ( Fig. S8A, see online supplememtary material ) .In addition, Han et al .performed immune deconvolution analysis of the bulk RNA-seq data from TCGA-PAAD by a ppl ying Micr oenvir onment Cell Populations ( MCP ) -counter . 14or PAAD patients with frequent 9p21 loss ( vs 9p21-WT tumours ) , we observed a remarkable decrease in the abundance of T cells, natural killer ( NK ) cells, myeloid dendritic cells, monocytic linea ge, fibr oblasts, endothelial cells, c ytotoxic lymphoc ytes, CD4 T cells, and B lineage ( Fig. S8B ) .Moreover, we compared the levels of immune cells between patients with 9p21 WT and patients with 9p21 loss based on ssGSEA algorithms .T he results demonstrated that patients with 9p21 WT had significantly higher le v els of "activated" dendritic cells ( aDCs ) , B cells , CD8 + T cells , cytolytic activity, HLA, inflammation-pr omoting factors, neutr ophils, T cell co-stimulation, Tfh, T h1 cells , and TILs than patients with 9p21 loss ( Fig. S8C ) .GO term enrichment revealed that the DEGs betw een the tw o groups w er e significantl y r elated to the imm une r esponse, inflammatory r esponse, and extr acellular matrix or ganization in the BP category; extracellular exosome and plasma membrane in the cellular component ( CC ) category; and heparin binding and extracellular matrix structural constituent in the MF category ( Fig. S9A and Table S5, see online supplementary material ) .GSEA also sho w ed that imm une-r elated pathways wer e enric hed in patients with 9p21 WT but wer e not enric hed in patients with 9p21 loss ( Fig. S9B, C ) .These results indicate that 9p21 loss promotes the formation of a 'cold' tumor microenvironment ( TME ) in PAAD.Our results sho w ed a higher rate of 9p21 loss in ASCP than in PD AC , whic h indir ectl y suggests that ASCP might be related to the tumour immune microenvironment and a decreased response to immunotherapy.

Discussion
To the best of our knowledge, this study involved the largest cohort among ASCP genomic landscape studies and prognostic gene v ariant anal yses.Compar ed with pr e vious studies and the external TCGA-PAAD analyses, we observed a high frequency of KRAS , TP53 , CDKN2A , SMAD4 , and losses of 8q ( MYC ) , which is consistent with our report 8 ( Fig. S1 ) .We detected some additional CNVs that have not been reported, such as losses on 7p and 17q and gains on 12q.Lenkiewicz et al .detected common recurring PDAC driv er e v ents in eac h of the ASCP genomes, including CDKN2A and SMAD4 homozygous deletions, KRAS and TP53 mutations, and MYC amplifications.Ho w e v er, the CNV pr ofiles of eac h flow-sorted aneuploid population ov erla pped and included gains of 8q21.3qtel that contained the MYC locus, a focal amplicon at 15q23-q24.2 that included PKM and CD276, and a homozygous deletion at 5q21.3 targeting EFNA5 and FBXL17 . 15This indicates that the driv er m utation genes of the for eign population and Chinese population are consistent, but there are differences in gene CNV c hanges.Mor e importantl y, we observ ed a higher pr oportion of 9p21 loss in ASCP than in PDAC and r e v ealed that 9p21 loss is linked to a poor patient prognosis in ASCP, suggesting that 9p21 loss could function as a prognostic marker in ASCP.
Se v er al published studies hav e r e v ealed the existence of 9p21 loss in different types of solid tumours, and this loss may have a certain correlation with prognosis .T he most prominent homozygous deletions in cancer affect c hr omosome 9p21.3 and eliminate CDKN2A/B tumour suppressors, disabling a cell-intrinsic barrier to tumorigenesis.Barriga et al .demonstrated that 9p21.3 deletion not only fails to effectively prevent PDAC cell proliferation but also pr omotes imm une esca pe while disrupting both intr acellular and extracellular tumour suppression programmes . 16The results of our study also illustrated for the first time the important pr ognostic v alue of 9p21 loss in ASCP, whic h was consistent with pr e vious studies in other types of tumours.Our findings indicate that 9p21 loss is an adverse prognostic factor for ASCP.9p21 loss is one of the most frequent somatic copy number alterations that occurs in human cancers. 17 , 18The 9p21.3 r egion, whic h includes MTAP , CDKN2A , and CDKN2B , is pleiotropic, and SNPs in this region spanning the three genes are susce ptibility mark ers for several cancers . 19Two crucial tumour suppressor genes located in this r egion, CDKN2A and CDKN2B , hav e well-established r oles in cell pr olifer ation and apoptosis and are inactivated in a wide range of cancers . 20Additionall y, sim ultaneous deletion m utations for CDKN2B and CDKN2A ar e fr equentl y observ ed in human cancers, including pancreatic cancer . 21CDKN2B deletion is considered essential for pancreatic cancer de v elopment instead of co-deletion  with CDKN2A . 22Biallelic deletions of MTAP with the neighbouring CDKN2A ar e commonl y observ ed in 40% of pancr eatic cancers . 23her efor e, the pr ognostic factor for ASCP should be validated in larger sample sizes and polyethnic studies.In the present study, we found that all CDKN2B deletions occurred within the adjacent domains that included CDKN2A or CDKN2A/MTAP .The real mutant forms included losses in the 9p21.3region that were verified with actual sequencing data, which were significantly associated with DFS, indicating the important prognostic role of 9p21 loss in ASCP.Apart from their prognostic value, anti-tumour agents are also the focus of clinical r esearc h.In our study, we analysed the DEGs between patients with 9p21 loss and patients with 9p21 WT and found that the expression of CDKN2A and CDKN2B was significantl y upr egulated in patients with 9p21 WT compar ed to patients with 9p21 loss .T hese results indicate that 9p21 loss leads to the downregulation of CDKN2A and CDKN2B gene expression.The two gene products of the CDKN2A gene, p16 and p19ARF, hav e r ecentl y been linked to eac h of two major tumour suppr essor pathways in human carcinogenesis, the RB1 pathway and the p53 pathway . 24These results suggest that 9p21 loss results in CDKN2A and CDKN2B gene expr ession and then pr omotes tumour pr ogr ession.Of course, this is also a limitation of our study, and we will further investigate its mechanism and potential biological behaviour through basic experiments in future studies.
Furthermor e, as pr ecision medicine adv ances, the c har acterization of the genomic landscape will dir ectl y influence clinical ther a peutic decision making.P atients with tar getable alter ations r eceiving matc hing tar geted ther a py surviv e 1 year longer than those receiving standard therapies . 25In our ASCP cohort, 48% of patients harboured potential actionable alterations, including CDKN2A/B , PIK3CA , NF1 , FBXW7 , STK11 , ATM , BRCA2 , FGFR1 , PTEN , and KRAS mutations.As mentioned above, CDK inhibitors could be a ther a peutic option for patients with CDKN2A/CDKN2B muta-tions.Patients with a BRCA2 mutation reportedly respond well to single-a gent pol y ( ADP-ribose ) pol ymer ase ( PARP ) inhibitors . 26The US Food and Drug Administration has approved the application of PARP inhibitors, such as olaparib, for germline BRCA-mutated pancreatic cancers . 27Patients with ATM mutations, which play an important role in the HR pathwa y, ma y also benefit from PARP inhibitors.8][29] NF1 inactivation leads to the activation of the mT OR pathwa y and promotes tumour cell gr owth; ther efor e , mT OR inhibitors can be used in potential ther a pies for NF1 -deficient patients . 30KRAS is the most common oncogene in pancr eatic cancers, and se v er al inhibitors tar geted to KRAS G12C hav e been tested in clinical trials, including ada gr asib and sotor asib, whic h hav e shown some e vidence of efficacy . 31The high proportion of patients with actionable alterations in our ASCP cohort could provide reliable evidence for targeted ther a py and impr ov e the outcomes of patients with ASCP.
In addition to targeted therap y, immunotherap y is an important treatment method for patients with tumours.TMB is a pr edictiv e biomarker in cases of high values, indicating a high rate of response to immune checkpoint inhibitors ( ICIs ) . 32Studies have also verified that immunotherap y sho ws promising results for PDAC patients with high TMB . 33Ho w e v er, TMB in ASCP is not yet fully understood.In our study, the TMB status in the 48 patients with ASCP was assessed: only one patient ( 2.1% ) was defined as TMB-H, a similar proportion to that among PDA C patients. 34 , 35D-L1 expression is another important biomarker of ICIs.In previous studies, ∼11% of ASCPs were positive for PD-L1, with a 10% cut-off value for PD-L1 positivity 25 .In our cohort, 12.5% ( 4/32 ) of patients had high PD-L1 expression with a cut-off of CPS ≥ 10.
Furthermor e, we explor ed whether m utated genes and alter ed signalling pathways can influence the TMB values in ASCP.We found that ACVR2A , BRC A2 , C ASP8 , CDKN2A , EPHB1 , FRS2 , and SPTA1 mutations led to a significantly higher TMB, whereas KRAS and STK11 mutations led to significantly lo w er TMB.Several studies hav e pr oposed that 9p21 loss may be related to the failure of imm unother a py and result in worse outcomes. 14 , 36In our study, 9p21 loss seemed to be more common in ASCP than in PD AC , whic h indir ectl y suggests that ASCP has a worse prognosis and mor e adv erse imm une r eactivity.In the TCGA-PAAD cohort, GSEA sho w ed that imm une-r elated pathw ays w er e enric hed in patients with 9p21 WT but were not enriched in patients with 9p21 loss.In addition, patients with 9p21 WT had significantly higher levels of aDCs , B cells , CD8 + T cells , cytolytic activity, HLA, inflammationpr omoting factors, neutr ophils, T-cell co-stim ulation, Tfh, Th1 cells, and tumour-infiltr ating l ymphocyte ( TIL ) than patients with 9p21 loss .T hese findings indicated that tumours with 9p21 loss sho w ed a cold tumour-like imm unophenotype, whic h may be related to the suppression of immune cell recruitment, the promotion of T-cell activation and imm unor egulatory factor expr ession, and the upregulation of immunosuppressive signals, which jointly induce the cold immunophenotype of tumours with 9p21 deletion.Mor e imm unother a py data need to be gather ed in pancreatic cancer to evaluate the influence of 9p21 loss in the future.
To investigate the molecular differences between ASCP and PD AC , m utational c har acteristics wer e anal ysed, and the fr equencies of mutated genes between ASCP and PDAC wer e compar ed.Our results demonstrated that the genomic characteristics in ASCP and PDAC had some similarities, but there were also differences .T he mutational frequencies of ACVR2A , FANCA , RBM10 , and SPTA1 in ASCP were significantly higher than those in PD AC .ACVR2A , FANCA , and RBM10 m utations wer e detected only in patients with ASCP.It seemed that ACVR2A , FANCA , RBM10 , and SPTA1 mutations might have important functions in tumour development for different subtypes and different molecular characteristics in Chinese patients.We also observed a higher frequency of Q61X in PDAC than in ASCP, which means that the molecular mechanism of KRAS mutations may be different between the two subtypes.A larger cohort is needed for further verification.
ASCP has a poor pr ognosis, and ther e is limited data to predict prognosis and to guide optimal treatment.T herefore , we examined the associations between clinical c har acteristics and prognosis in all respects .T he screening and identification of ne w pr ognostic biomarkers in patients with pancr eatic cancer, especially those with ASCP, is ur gentl y needed for clinical decision-making.Tumour location, CA19-9 le v el, and tumour size have been identified as preoperati ve, inde pendent, predicti ve risk factors for poor prognosis in patients with PDA C . 37A r etr ospectiv e study r e v ealed that for patients with advanced pancreatic cancer, tumour stage, chemotherapy, circulating regulatory T cells, CA19-9 le v els, CA125 le v els, and KRAS G12D and G12V mutations ar e significantl y associated with OS . 38Lymph node metastasis appears to correlate with a poor prognosis in patients with pancr eatic adenocarcinoma, and sur gical r esection has a better pr ognosis than nonsur gical r esection for patients with r esectable pancreatic cancer. 39 , 40The r esults of univ ariate Cox r egr ession analyses in our study also sho w ed that surgery w as significantly associated with DFS and OS.Although there was no statistical significance, our data also sho w ed that lymph node metastasis tended to be associated with an adverse prognosis .T herefore , earl y and accur ate dia gnosis is crucial for the outcome of patients after radical surgery.In addition, tumour stage and distal metastasis were significantly correlated with DFS, but OS was not significant.One possible reason is that the follow-up time was not long enough to gather reliable data for the long-term response.Inter estingl y, the r esults of univ ariate Cox r egr ession anal yses and m ultiv ariate anal ysis both sho w ed that 9p21 loss was related to poor prognosis and could function as an independent risk factor for poor prognosis in patients with ASCP.An increasing number of studies have suggested that 9p21 loss is related to poor survival. 14 , 41 , 42Ov er all, pr ompt interv ention and detection of patients with ASCP, and especially attention and measures for patients with 9p21 loss, are needed to prolong life expectancy.
In summary, we compr ehensiv el y inv estigated the genomic variations in ASCP and explored the specific mutation patterns of ASCP and pr ognostic-r elated molecular markers, pr oviding a molecular basis for differential diagnosis, prognosis prediction, and the exploration of new treatment options for ASCP.Ho w e v er, limited information is available for targeted and immunological ther a py in ASCP.We look forw ar d to obtaining mor e tr eatment data for the benefit of patients and to further exploring the tumour immune microenvironment to predict immunotherapy responses for patients.Our results displayed a higher rate of 9p21 loss in ASCP than in PD AC .Notably, 9p21 loss was significantly associated with a poor prognosis in ASCP patients .T he results of our study also illustrated for the first time the important prognostic value of 9p21 loss in ASCP and that ASCP might have a more adv erse imm une r eactivity than PD AC .

Figure 1 .
Figure 1.Histopathological features of ASCP ( hematoxylin and eosin staining ) .( A ) Irregular glandular structures and sheets of tumours were observed in the fibrous stroma.( B ) In the adenocarcinoma component, atypical cells formed incomplete or complex glands .T he nuclear-to-cytoplasmic ratio was r elativ el y high, and nucleolar staining was significant.( C ) The squamous cell carcinoma components demonstr ated typical ker atin pearls and single-cell keratinization.The polygonal tumour cells displayed abundant eosinophilic cytoplasm and distinct borders.( D ) Histopathological features of PDAC ( hematoxylin and eosin staining ) .

Figure 2 .
Figure 2. Genomic landscape of 48 ASCP patients.( A ) The features of GAs.The panel shows the matrix of mutations coloured according to the type of m utation.Eac h column denotes an individual patient and each row represents a gene .( B ) T he distributions indicate GAs. ( C ) Onco-driver genes r epr esented in the ASCP samples.( D ) Recurrent copy number variations.Red represents copy number gains and blue represents losses.

Figure 3 .
Figure 3.Comparison of genetic c har acterization between ASCP and PD AC .( A ) The profiles of GAs.Dark red represents 48 ASCP patients and green r epr esents 98 PDAC patients .T he panel shows the matrix of mutations coloured according to the type of mutation.Green: substitution/indel; red: gene amplification; blue: gene homozygous deletion; y ello w: fusion/r earr angement; pur ple: truncation.( B ) Genes with a significant differ ence in m utation frequenc y betw een ASCP and PD AC .( C ) Comparison of the 9p21 homologous deletion ratio in ASCP and PD AC .* P < 0.05.

Figure 4 .
Figure 4. Genetic c har acteristics and pr ognosis of 9p21 loss.( A ) Gene co-m utations in ASCP.( B ) Example of copy number v ariation data of c hr omosome 9. Z and C r epr esent two software algorithms.Genomic region is presented according to the assay used ( purple = Z-score, red = C-score ) and ordered according to genomic position.Blue spots/purple lines and green spots/red lines indicate the average log-ratio in Z and C, respectively.Lines indicate the av er a ge log-r atio in a segment.( C , D ) , Ka plan −Meier anal ysis of disease-fr ee surviv al and ov er all surviv al for 9p21, r espectiv el y.The hazard ratio was obtained using the Cox proportional hazards test.Loss, homozygous deletion type.

Figure 6 .
Figure 6.Expression of p16 protein in ASCP specimens.( A ) IHC for p16-positive status in patients with 9p21 WT. ( B ) IHC for p16-negative status in patients with 9p21 loss.

Table 1 .
Clinicopathological c har acteristics of patients with ASCP and PD AC .