Multiomic Characterization Reveals a Distinct Molecular Landscape in Young-Onset Pancreatic Cancer

PURPOSE Using a real-world database with matched genomic-transcriptomic molecular data, we sought to characterize the distinct molecular correlates underlying clinical differences between patients with young-onset pancreatic cancer (YOPC; younger than 50 years) and patients with average-onset pancreatic cancer (AOPC; 70 years and older). METHODS We analyzed matched whole-transcriptome and DNA sequencing data from 2,430 patient samples (YOPC, n = 292; AOPC, n = 2,138) from the Caris Life Sciences database (Phoenix, AZ). Immune deconvolution was performed using the quanTIseq pipeline. Overall survival (OS) data were obtained from insurance claims (n = 4,928); Kaplan-Meier estimates were calculated for age- and molecularly defined cohorts. Significance was determined as FDR-corrected P values (Q) < .05. RESULTS Patients with YOPC had higher proportions of mismatch repair–deficient/microsatellite instability-high, BRCA2-mutant, and PALB2-mutant tumors compared with patients with AOPC, but fewer SMAD4-, RNF43-, CDKN2A-, and SF3B1-mutant tumors. Notably, patients with YOPC demonstrated significantly lower incidence of KRAS mutations compared with patients with AOPC (81.3% v 90.9%; Q = .004). In the KRAS wild-type subset (n = 227), YOPC tumors demonstrated fewer TP53 mutations and were more likely driven by NRG1 and MET fusions, whereas BRAF fusions were exclusively observed in patients with AOPC. Immune deconvolution revealed significant enrichment of natural killer cells, CD8+ T cells, monocytes, and M2 macrophages in patients with YOPC relative to patients with AOPC, which corresponded with lower rates of HLA-DPA1 homozygosity. There was an association with improved OS in patients with YOPC compared with patients with AOPC with KRAS wild-type tumors (median, 16.2 [YOPC-KRASWT] v 10.6 [AOPC-KRASWT] months; P = .008) but not KRAS-mutant tumors (P = .084). CONCLUSION In this large, real-world multiomic characterization of age-stratified molecular differences in pancreatic ductal adenocarcinoma, YOPC is associated with a distinct molecular landscape that has prognostic and therapeutic implications.


INTRODUCTION
][7][8][9][10][11][12] Emerging data indicate that smoking, 4,5 alcohol use, 13 obesity, 14 and family history 13,15,16 are risk factors for YOPC.YOPC also skews toward male sex 7,11 although rates in women-particularly Black womenare rising faster than in men. 6,7,12espite modern multimodal therapy. 17However, clinically annotated tumor profiling database studies such as the Know Your Tumor study have demonstrated that patients with PDAC experience longer survival when receiving therapies matching actionable mutations compared with nonmatched therapies. 18Moreover, The Cancer Genome Atlas analysis of PDAC revealed that, excluding KRAS and CDKN2A, 42% of patients could be candidates for molecularly informed clinical trials. 19The increasing armamentarium of precision medicine approaches for patients with PDAC emphasizes the critical need to understand tumor-level molecular differences between patients with YOPC and AOPC, which might inform personalized therapy in this subset of patients.
Efforts to describe molecular differences between YOPC and AOPC have been hampered by a lack of real-world, largescale matched genomic and transcriptomic data, leading to conflicting conclusions between studies.For instance, Raffenne et al 3 found no substantial differences in the mutational landscape between patients with YOPC and AOPC, whereas others have identified higher SMAD4 mutation rates, increased activation of the TGF-b pathway, 20 and differential expression of CDKN2A and FOXC2 in YOPC compared with AOPC. 21Despite these differences, some unifying signals have emerged, particularly that patients with YOPC harbor fewer oncogenic somatic KRAS mutations but more pathogenic germline mutations than patients with AOPC. 16,19,20Further complicating our understanding of this question are the conflicting survival outcomes observed in these studies, with many indicating that patients with YOPC have improved survival, 9,11,22,23 but others showing either shorter or no difference in survival compared with patients with AOPC. 3,5,10,15,20,24Together, these results illustrate gaps in our understanding of the genomic and transcriptomic correlates underlying clinical differences between patients with YOPC and AOPC.
In the present study, we analyzed a real-world multiinstitutional cohort of 2,430 sequenced tumors-including 292 YOPC-to characterize the distinct molecular landscape associated with YOPC compared with AOPC and better understand molecular correlates underlying the divergent clinical outcomes in patients with YOPC.

Patient Samples
Two thousand four hundred thirty histologically confirmed PDAC samples were identified in the Caris Life Sciences database (Phoenix, AZ) with matched DNA sequencing, whole-transcriptome sequencing (WTS), and immunohistochemistry (IHC) data.We stratified these specimens into YOPC (younger than 50 years at diagnosis; n 5 292) and AOPC (70 years and older; n 5 2,138).Among YOPCs, 179 were metastases and 113 were primary tumors; among AOPCs, 1,167 were metastases and 967 were primary tumors.

Next-Generation Sequencing
Tumor enrichment was achieved using manual microdissection of formalin-fixed, paraffin-embedded (FFPE) sections that were marked for areas with an at least 20% tumor content.Next-generation sequencing (NGS) was performed on genomic DNA using the NextSeq or NovaSeq 6000 platforms (Illumina, Inc, San Diego, CA).For NextSeqsequenced tumors, a custom-designed SureSelect XT assay was used to enrich 592 whole-gene targets (Agilent Technologies, Santa Clara, CA).For NovaSeq-sequenced tumors, a hybrid pull-down panel of baits designed to enrich for >700 clinically relevant genes at high coverage and read depth was used, along with a separate panel to enrich for an additional >20,000 genes at lower depth.Genetic variants were detected with >99% confidence and were categorized by board-certified molecular geneticists as previously described. 25Tumor mutational burden (TMB)high was defined as ≥10 mutations/Mb.

IHC
FFPE sections on glass slides were stained for PD-L1 (clone SP142 [Spring Bioscience, Pleasanton, CA]) using automated staining techniques, per the manufacturer's instructions, and were optimized and validated per Clinical Laboratory Improvement Amendments/College of American Pathologists and International Organization for Standardization requirements.Staining was identified as positive if its intensity on the membrane of the tumor cells was ≥21 (on a semiquantitative scale of 0-3: 0 no staining, 11 weak staining, 21 moderate staining, or 31 strong staining) and the percentage of positively stained cells was ≥5%.

Mismatch Repair Deficiency/Microsatellite Instability-High Status
Multiple test platforms were used to determine mismatch repair deficiency (dMMR)/microsatellite instability-high (MSI-H) status of the tumors profiled, including fragment analysis (FA, Promega, Madison, WI), IHC (MLH1, M1 antibody; MSH2, G2191129 anti-body; MSH6, 44 antibody; and PMS2, EPR3947 antibody [Ventana Medical Systems, Tucson, AZ]), and NGS.The three platforms generated highly concordant results as previously reported 26 ; in the rare cases of discordant results, dMMR/MSI-H status was determined in the order of IHC, FA, and NGS.

WTS
mRNA was isolated from manually microdissected areas of FFPE sections with a tumor content of at least 10%.Whole-transcriptome sequencing (WTS) was performed using the Illumina NovaSeq platform (Illumina, Inc, San Diego, CA) and the Agilent SureSelect Human All Exon V7 bait panel (Agilent Technologies, Santa Clara, CA); transcripts per million were reported.Gene fusions were detected using the ArcherDX fusion assay (ArcherDX, Boulder, CO) and Illumina MiSeq platform (Illumina MiSeq, San Diego, CA) as previously described. 27Immune cell fractions were calculated from transcriptomic data using quanTIseq 28 and xCell. 29Gene set enrichment analysis (GSEA) and Metascape pathway analysis were performed on WTS data. 30,31HLA genotyping was performed using arcasHLA, an in silico tool that infers HLA genotypes from RNA sequencing data. 32If a single HLA genotype was detected, the specimen was classified as homozygous, which can occur because of parental homozygosity or HLA loss of heterozygosity.

Statistical Analysis
Clinicodemographic features were compared using the chisquare test, with P < .05considered statistically significant.Comparative analysis of molecular alterations in the cohorts was analyzed using chi-square or Fisher's exact tests.Tumor microenvironment cell fractions were analyzed among cohorts using nonparametric Kruskal-Wallis testing.Because these closely related cohorts are only differentiated by age, P values of < .05were highlighted as relevant trends.For a more stringent analysis of the differences between AOPC and YOPC, P values were corrected for multiple hypothesis testing using the Benjamini-Hochberg method to avoid type I error and adjusted Q < .05 was considered statistically significant.

Clinical Outcomes Data
Real-world overall survival (OS) information was obtained from insurance claims data and calculated from the date of tissue collection to last contact.Kaplan-Meier estimates were calculated for YOPC and AOPC in the entire cohort of patients with clinical data in the Caris CODEai clinicogenomic database (n 5 4,928) and stratified by KRAS mutation status (n 5 3,116 patients with KRAS data; KRAS WT , n 5 393; KRAS MUT , n 5 2,723); these numbers differ from the molecular analysis since the database is constantly increasing in size.Significance was determined as log-rank P < .05.

Compliance Statement
This study was approved by the Institutional Review Board at the University of Miami and conducted in accordance with guidelines of the Declaration of Helsinki, Belmont report, and US Common rule.Per 45 CFR 46.101(b)(4), this study used retrospective, deidentified clinical data and no patient consent was necessary from the patients.

Data Availability
Data presented in this study are not publicly available because of data size and patient privacy but are available on reasonable request from the corresponding author.

Clinicodemographic Characteristics
At the time of molecular analysis, 2,430 patient samples were annotated with genomic and transcriptomic data.A total of 4,928 patients had available clinical outcomes data in the most recent query of the Caris CODEai clinicogenomic database, from which Kaplan-Meier curves were generated.

Comparative Molecular Landscape of YOPC and AOPC
Previous studies have reported differing prevalence of molecular alterations 3,19,20,33,34 and a preponderance of germline mutations in BRCA1/2 and MMR genes in patients with YOPC compared with patients with AOPC. 16owever, direct comparisons between YOPC and AOPC are scarce and have used smaller cohorts. 3,20We analyzed clinically relevant pathogenic/likely pathogenic mutations and copy number alterations in tumors from patients with YOPC and AOPC from this real-world cohort (Appendix Table A1).

Differentially Regulated Signaling Pathways in Tumor Transcriptomes From Patients With YOPC Versus AOPC
To better understand how these genomic differences between YOPC and AOPC tumors influence downstream oncogenic and tumor microenvironment signaling, we performed GSEA comparing whole-tumor transcriptomes in YOPC versus AOPC.A relatively narrow number-that is, total of 20-of genes were significantly differentially expressed (P < .05;Q < .25) between YOPC and AOPC (Fig 3A ; Appendix Table A2).The top genes more highly expressed in YOPC included carboxypeptidase B (CPB2), plasminogen (PLG), prothrombin (F2), and genes for fibrinogen alpha and beta chains (FGA/FGB), whereas plasminogen activator inhibitor 2 (SERPINB2) and interferon gamma (IFNG) had significantly higher expression in AOPC.We then used a less stringent P value cutoff (P < .25) in Metascape pathway analysis to clarify the transcriptomic nuances of these agestratified PDAC cohorts.This analysis revealed that YOPC tumor transcriptomes were significantly enriched in pathways related to blood clotting cascade, extracellular matrix, cancer pathways, cytokine/inflammatory response, and angiogenesis (Fig 3B).

Intratumoral Immune Deconvolution and HLA Landscape in YOPC Versus AOPC
Because of the enrichment of select pathways and somatic alterations with diverse immunologic repercussions in YOPC, we sought to determine differences in the tumor immune microenvironment between YOPC and AOPC using quanTIseq immune deconvolution. 28While there were no significant differences in rates of TMB-high tumors, PD-L1 positivity (via IHC), or immune checkpoint gene expression between the cohorts (Appendix Figs A1A and A1B), there was a statistically significant enrichment in computationally inferred signatures for natural killer (NK) cells (P 5 .009;Q 5 .039),CD8  4B).
We then used xCell deconvolution 29 to further compare CD8 1 T-cell subsets between cohorts, which revealed no differences in effector or central memory CD8

DISCUSSION
To our knowledge, the present study represents the largest pragmatic molecular comparison of YOPC versus AOPC.Our data reinforce previously observed epidemiologic distinctions between patients with YOPC and AOPC, 4,5,7,11 specifically its male preponderance and association with active smoking behaviors in patients with YOPC, and conclusively reveal a higher incidence of KRAS WT tumors in YOPC.Within this KRAS WT subset, we uncovered distinct molecular vulnerabilities when stratifying by age-that is, MET and NRG1 fusions in YOPC-KRAS WT and BRAF fusions in AOPC-KRAS WT .Among the unstratified cohort, tumors from patients with YOPC demonstrated higher rates of alterations in HRR genes, higher prevalence of dMMR/MSI-H, and enrichment of NK cells and na ïve CD8 1 T cells.Finally, our data reconcile conflicting previous evidence by demonstrating improved survival in patients with YOPC compared with patients with AOPC, which may not only reflect the reduced prevalence of the virulent oncogenic drivers KRAS, SMAD4, and CDKN2A in tumor genomes of YOPC patients but also be driven by the significantly longer survival of YOPC-KRAS WT versus AOPC-KRAS WT patients.
While the success of targeted and immune-based therapies has significantly lagged in PDAC compared with other solid tumors, the Know Your Tumor study illustrated the oncologic importance of molecularly matched targeted therapies in patients with advanced PDAC. 18To that end, our data provide a biologic map of the distinct molecular vulnerabilities in patients with YOPC that might be exploited therapeutically.2][43] Conversely, the significant reduction in HLA-DPA1 homozygosity-which has been previously associated with dampened antigen presentation and checkpoint blockade efficacy 44 -and associated increases in computationally inferred adaptive immune subpopulations (ie, NK and CD8 1 T cells) in YOPC suggest a less immunosuppressive and more immunostimulatory microenvironment.While the impact of greater numbers of intratumoral na ïve CD8 1 T cells in YOPC is unclear, higher circulating levels of these cells have been associated with improved prognosis in other solid tumors, for example, non-small-cell lung cancer. 45These findings underscore the need for deeper investigation and functional characterization of cellautonomous and nonautonomous immunologic repercussions in YOPC tumors.The differentially expressed transcriptome we observed in patients with YOPC in the current study, however, is not strongly consistent with previous-albeit underpowered-studies that revealed enrichment in pathways predominantly related to hedgehog signaling and hypoxia in YOPC. 3,20This lack of concordance might be attributable to our substantially larger cohort size and/or the inherent heterogeneity of patients enrolled in this pragmatic real-world study capturing data with wide geographic and clinicodemographic variability, which present novel insights into the genotype-immunophenotype chasm in YOPC. 46r study has several limitations.First, while the classification of YOPC and AOPC into age cutoffs of <50 and ≥70-year was informed in previous studies, [3][4][5] this arbitrary distinction may underestimate subtle molecular differences in patients with YOPC.Second, while several of the reported genomic differences did not achieve significance by multiple hypothesis testing, we felt it important to report these novel signals with the recognition that our study compares molecular determinants in two closely related PDAC patient populations differentiated solely by a 20-year age gap.Further validation of the subtle molecular features distinguishing these cohorts is warranted.Third, the lack of clinical annotation (eg, performance status, resection status, stage, BMI, and multimodality treatment information) in the Caris CODEai data set precluded our ability to perform multivariable analyses to account for confounding by these clinical parameters.
0][21]    ).AOPC expression is set to 1 for each gene.There were no statistically significant differences in immune gene expression between YOPC and APOC (determined using the Mann-Whitney U test).(C) Fold-change median CD8

FIG 1 .FIG 2 .FIG 3 .
FIG 1.Molecular landscape of YOPC and AOPC.(A) Oncoprints displaying the pathogenic molecular alteration pattern of YOPC (n 5 292) and AOPC (n 5 2,138).Columns represent tumor samples.Rows represent individual molecular biomarkers, whose percentages in the cohort are described in the boxes to the left of oncoprints.Pink, expressed or amplified; green, mutated; gray; not altered; light gray, data not available.(B) Bar graph showing statistically significant differential molecular alterations in YOPC (blue bars, n 5 292) versus AOPC (red bars, n 5 2,138).P values (chi-square or Fisher's exact tests) and FDR-adjusted Q values are indicated above the compared groups for each molecular alteration.AOPC, average-onset pancreatic cancer; FDR, false discovery rate; YOPC, young-onset pancreatic cancer.

FIG 5 .
FIG 5. OS of patients with YOPC and AOPC stratified by KRAS MUT and KRAS WT .(A) Kaplan-Meier curves depict the OS of patients with YOPC (blue line, n 5 787) versus patients with AOPC (red line, n 5 2,753) in the entire PDAC cohort with clinical outcome data (n 5 4,141 total).(B) and (C) All PDAC cases with KRAS mutation data available were stratified by KRAS status.Kaplan-Meier curves depict the OS of patients with YOPC (blue line, n 5 98) versus patients with AOPC (red line, n 5 295) with KRAS WT tumors (n 5 393 total; B) and patients with YOPC (blue line, n 5 347) and AOPC (red line, n 5 2,376) KRAS MUT tumors (n 5 2,723 total; C).AOPC, average-onset pancreatic cancer; OS, overall survival; PDAC, pancreatic ductal adenocarcinoma; YOPC, young-onset pancreatic cancer.
FIG A1.Immune landscape of YOPC and AOPC.(A) Percentage of cases positive for immunotherapy biomarkers TMB-H and PD-L1(1) for YOPC (blue bars, n 5 273-292 [a small number of patients lacked PD-L1 IHC or dMMR/MSI-H analysis]) and AOPC (red bars, n 5 2,028-2,138 [a small number of patients lacked PD-L1 IHC or dMMR/MSI-H data]).Statistical analysis was performed using the chi-square or Fisher's exact test.(B) Fold-change gene expression levels in TPM of immune checkpoint genes between YOPC (blue bars, n 5 284) and AOPC (red bars, n 5 2,089).AOPC expression is set to 1 for each gene.There were no statistically significant differences in immune gene expression between YOPC and APOC (determined using the Mann-Whitney U test).(C) Fold-change median CD8 1 T-cell fractions (YOPC v AOPC) calculated by xCell immune deconvolution.P value determined using the Kruskal-Wallis test.Significant P value (<.05) indicated by italics.AOPC, average-onset pancreatic cancer; dMMR, mismatch repair deficiency; IHC, immunohistochemistry; MSI-H, microsatellite instability-high; TMB-H, tumor mutational burden-high; TPM, transcripts per million; YOPC, young-onset pancreatic cancer.

TABLE 1 .
Clinicodemographic Characteristics of Patients With PDAC From the Caris Life Sciences Database

1
FIG 4. Intratumoral immune populations and HLA landscape in YOPC and AOPC.(A) Computationally inferred intratumoral immune population between YOPC (n 5 284) and AOPC (n 5 2,089).The heatmap indicates fold change (YOPC v AOPC) in median immune fraction according to quanTIseq.P values were determined using the nonparametric Kruskal-Wallis test.Asterisks indicate P < .05,with Q values shown to the right.Italicized/bolded Q values indicate Q < .05. (B) For cell types with median values of "0" (ie, monocytes, CD4 1 T cells, CD8 1 T cells, and myeloid dendritic cells), the percentage of tumors with nonzero immune infiltrates were compared.(C) Differences in HLA landscape inferred from WTS data in YOPC (blue bars, n 5 284) compared with AOPC (red bars, n 5 2,089).P values (chi-square or Fisher's exact test) and FDR-adjusted Q values are indicated above compared groups for each HLA gene.AOPC, average-onset pancreatic cancer; FDR, false discovery rate; WTS, whole-transcriptome sequencing; YOPC, young-onset pancreatic cancer.

TABLE A2 .
List of Differentially Expressed Genes Between YOPC and AOPC