Abstract
Alterations in KRAS, CDKN2A (p16), TP53, and SMAD4 genes have been major drivers of pancreatic carcinogenesis. The clinical course of patients with pancreatic cancer in relation to these driver alterations has not been fully characterised in large populations. We hypothesised that pancreatic carcinomas with different combinations of KRAS mutation and aberrant expression of CDKN2A, p53, and SMAD4 might show distinctive recurrence patterns and post‐operative survival outcomes. To test this hypothesis, we utilised a multi‐institutional cohort of 1,146 resected pancreatic carcinomas and assessed KRAS mutations by droplet digital polymerase chain reaction and CDKN2A, p53, and SMAD4 expression by immunohistochemistry. Multivariable hazard ratios (HRs) and 95% confidence intervals (CIs) for disease‐free survival (DFS) and overall survival (OS) were computed according to each molecular alteration and the number of altered genes using the Cox regression models. Multivariable competing risks regression analyses were conducted to assess the associations of the number of altered genes with specific patterns of recurrence. Loss of SMAD4 expression was associated with short DFS (multivariable HR, 1.24; 95% CI, 1.09–1.43) and OS times (multivariable HR, 1.27; 95% CI, 1.10–1.46). Compared to cases with 0–2 altered genes, cases with three and four altered genes had multivariable HRs for OS of 1.28 (95% CI, 1.09–1.51) and 1.47 (95% CI, 1.22–1.78), respectively (p trend < 0.001). Patients with an increasing number of altered genes were more likely to have short DFS time (p trend = 0.003) and to develop liver metastasis (p trend = 0.006) rather than recurrence at local or other distant sites. In conclusion, loss of SMAD4 expression and an increasing number of altered genes were associated with unfavourable outcomes in pancreatic cancer patients. This study suggests that the accumulation of the four major driver alterations can confer a high metastatic potential to the liver, thereby impairing post‐operative survival among patients with pancreatic cancer.
Keywords: cohort studies, oncogenes, pancreatectomy, pancreatic ductal adenocarcinoma, outcome
Introduction
Alterations in KRAS, CDKN2A (p16), TP53, and SMAD4 genes have been major drivers involved in pancreatic carcinogenesis [1, 2, 3]. Clinical studies suggest that each molecular alteration may be a biomarker for an unfavourable prognosis of pancreatic cancer patients with a modest increase in the mortality hazard [4, 5, 6, 7, 8, 9]. Further clinical data implicate that the mortality hazard among pancreatic cancer patients is elevated according to an increasing number of the driver gene alterations, suggesting additive tumour‐promoting effects of those drivers [5, 6, 7, 8, 9, 10]. However, the reported associations of alterations of the four driver genes with survival outcomes have been heterogeneous largely due to a lack of data derived from large populations.
Accumulating mechanistic evidence links alterations of the four driver genes to an aggressive phenotype of pancreatic cancer, characterised by high potentials of invasion, metastasis, and immune evasion [2, 3, 11, 12, 13, 14, 15, 16, 17]. However, there has been a paucity of data on how combinations of the four altered genes may affect tumour behavioural patterns in pancreatic cancer populations or mouse models. For instance, it is largely unknown whether the accumulation of those molecular alterations provokes premature post‐operative recurrence through a hematogenous or lymphogenic mechanism. In a pooled analysis of 20 randomised clinical trials investigating adjuvant chemotherapy, disease‐free survival (DFS) was well correlated with overall survival (OS) among patients with resected pancreatic cancer [18], suggesting the importance of regulating tumour progression during the short‐term post‐operative period. A better understanding of tumour behaviour according to molecularly defined subtypes would help us to refine post‐operative surveillance programmes and develop new treatment strategies for this highly lethal malignancy [19]. We hypothesised that pancreatic cancers with different combinations of the four major driver alterations might show distinctive recurrence patterns and post‐operative survival outcomes.
To test our hypothesis, we assessed KRAS mutations by droplet digital polymerase chain reaction (ddPCR) and CDKN2A, p53, and SMAD4 expression by immunohistochemistry (IHC) in 1,146 patients within a large multicentre cohort of resected pancreatic tumours with comprehensive data on clinical and pathological parameters. The current study found that combinations of the four driver alterations were associated with distinct recurrence patterns and post‐operative outcomes.
Materials and methods
Study population
Within the GTK (Good To Know) Pancreatic Cancer consortium [20], we collected data on consecutive patients who underwent surgical resection of pancreatic carcinoma between 2005 and 2017 at Cancer Institute Hospital, Japanese Foundation for Cancer Research, The University of Tokyo Hospital, or Keio University Hospital (all in Tokyo, Japan). Among those patients, we included 1,146 cases with pancreatic cancer where surgical specimens of the tumours were available for molecular analyses. We excluded (1) patients with ductal adenocarcinoma variants, including intraductal papillary mucinous neoplasm‐derived carcinoma, undifferentiated carcinoma, and colloid carcinoma, (2) patients with mixed tumours (e.g. mixed ductal‐neuroendocrine carcinoma), (3) patients having advanced cancer of other origins concomitantly at the time of the index surgery, or (4) patients with 30‐day or in‐hospital mortality. In analyses of post‐operative recurrence, we further excluded patients with a resected metastatic lesion, R2 resection margin (macroscopic residual tumour), or no available cross‐sectional imaging following the index surgery. The patients were followed until death or the end of follow‐up (30 April 2022), whichever came first.
The current study was designed and conducted according to the guidelines in the Helsinki Declaration. Given the retrospective study design, informed consent was obtained from all patients on an opt‐out basis. The study was approved by the ethics committees at Cancer Institute Hospital, Japanese Foundation for Cancer Research, The University of Tokyo, and Keio University School of Medicine (all in Tokyo, Japan); and was registered with the UMIN registry (the University Hospital Medical Information Network registry; registration number, UMIN000044027).
Assessment of KRAS mutation status
As previously described [20], we conducted ddPCR for mutations in KRAS codons 12, 13, and 61 using the QX200 system (Bio‐Rad Laboratories, Hercules, CA, USA). Variant allele frequency of KRAS was calculated as the ratio of the number of KRAS‐mutant droplets to that of droplets including KRAS‐mutant signal and/or KRAS‐wild type signal. As previously described [20], tumours were classified as KRAS‐mutant when the variant allele frequency was ≥1% and otherwise as KRAS‐wild type.
Assessment of SMAD4, p53, and CDKN2A (p16) expression
Based on a standardised protocol, we constructed tissue microarrays (TMAs) from surgical specimens of the primary pancreatic carcinomas. Up to four tumour cores, approximately 2 mm in diameter, were included from each case. We conducted IHC of SMAD4 [21], p53 [21], and CDKN2A (p16) using 4‐μm‐thick TMA sections in an automated staining system (BOND‐MAX; Leica Biosystems, Wetzlar, Germany) with the use of BOND Polymer Refine Detection Kit (Leica Biosystems). The primary monoclonal antibodies used were clone B‐8 (mouse; dilution, catalogue number sc‐7966, 1:100, Santa Cruz Biotechnology, Dallas, TX, USA) for SMAD4, clone DO‐7 (mouse; dilution, catalogue number M700101‐2, 1:2000, Agilent Technologies, Santa Clara, CA, USA) for p53, and clone EPR1473 for CDKN2A (p16) (mouse; dilution, catalogue number ab108349, 1:1000, Abcam, Cambridge, UK). A single pathologist (YM), blinded to other data, conducted all IHC assessments of the tumours based on the previously established criteria (Figure 1) [5, 22, 23]. For SMAD4, tumours were defined as ‘intact’ when there was any nuclear or cytoplasmic staining in carcinoma cells or ‘lost’ when there was a complete loss of staining [5, 22]. Tumours showing heterogeneous nuclear p53 staining were classified as ‘intact’, whilst those showing diffuse nuclear staining or complete loss of nuclear staining were defined as ‘aberrant’ [5]. Nuclear expression levels of CDKN2A (p16) were classified as ‘intact’ when any ductal epithelial cells exhibited nuclear expression above the cytoplasmic background; otherwise, they were classified as ‘lost’ [23]. In a subset of cases (n = 312), expression status was independently examined by a second pathologist (KHar), and the agreements between the two pathologists were reasonable for all three markers with κ coefficients of 0.81 [95% confidence interval (CI), 0.74–0.87] for SMAD4, 0.92 (95% CI, 0.88–0.96) for p53, and 0.65 (95% CI, 0.55–0.76) for CDKN2A (p16). Based on our preliminary results suggesting no substantial intratumour heterogeneity in expression levels of SMAD4, p53, and CDKN2A in most of the invasive components of pancreatic carcinomas, we used TMA sections when available and used whole tissue sections for tumours that were not available for the TMAs or were not assessable in the TMA slides. The concordances between the scores for different cores in a single case were almost perfect with κ coefficients of 0.97 (95% CI, 0.95–0.99) for SMAD4 (n = 837), 0.988 (95% CI, 0.979–0.998) for p53 (n = 854), and 0.96 (95% CI, 0.93–0.99) for CDKN2A (n = 857). Regarding IHC of SMAD4, p53, and CDKN2A, we did not observe significant staining differences between surgically resected specimens and fine‐needle aspiration specimens from selected cases of pancreatic cancer (supplementary material, Figure S1).
Figure 1.
Representative immunohistochemistry images of (A) SMAD4, (B) p53, and (C) CDKN2A (p16) and MTAP expression. Asterisks indicate tumour cells with an aberrant ‘lost’ pattern of the corresponding marker. Tumours with diffuse nuclear p53 expression were classified as aberrant (B, middle). Note that cancer cells showed loss of both CDKN2A (C, middle) and MTAP (C, right) expression, whereas pancreatic intraepithelial neoplasm lesions (daggers) preserved expression for these markers. Bar = 250 μm.
We conducted methylthioadenosine phosphorylase (MTAP) IHC in a random sample of 208 cases, as the MTAP gene is located at chromosome 9p21.3 close to CDKN2A, and thus, loss of MTAP expression has shown to be a good surrogate for homozygous deletions of CDKN2A [24]. The primary antibody, clone 2G4 (mouse; dilution, catalogue number H00004507‐M01, 1:100, Abnova, Taipei, Taiwan), was used for MTAP immunostaining. All the tumours with complete loss of MTAP expression (65 out of 208 cases) were categorised as ‘lost’ expression of CDKN2A.
Targeted sequencing of genomic DNA
Previously reported sequencing data from 53 patients with pancreatic cancer were reanalysed in this study [10, 25]. Detailed methods of targeted sequencing were described previously [10, 25]. In brief, DNA was extracted from formalin‐fixed paraffin‐embedded slides of surgically resected pancreatic cancer using the QIAamp DNA FFPE Tissue Kit (Qiagen, Venlo, Netherlands) according to the manufacturer's protocol. Amplicon libraries were prepared from 10 ng DNA per sample using the Ion AmpliSeq Cancer Hotspot Panel v2 and the Ion AmpliSeq Library Kit 2.0 (Thermo Fisher Scientific, Waltham, MA, USA). Sequencing variants for point mutations, insertions, and deletions were called using the Ion Reporter Software (Thermo Fisher Scientific).
Transcriptome analysis
As previously described [21], RNA was extracted from frozen tissues of 20 patients with pancreatic cancer and was analysed using a Clariom D Pico Assay (Thermo Fisher Scientific) with a GeneChip Scanner 3000 7G System (Thermo Fisher Scientific). The microarray data were deposited in NCBI's Gene Expression Omnibus database (accession number, GSE157353). Tumours were categorised into classical and basal‐like subtypes based on gene sets associated with Moffitt and colleagues' transcriptome subtypes [21, 26].
Statistical analysis
The statistical analysis is detailed in the Supplementary methods [27, 28, 29, 30, 31, 32, 33, 34]. All statistical analyses were performed using the SAS software (version 9.4; SAS Institute, Cary, NC, USA), and the two‐sided α level of 0.005 was used for statistical significance [34].
Results
Within our multicentre cohort, we examined the tumour status of the four driver genes in 1,146 patients with resected pancreatic cancer (supplementary material, Figure S2). During the median follow‐up time of 55.9 months (interquartile range, 42.8–83.7 months) for all censored cases, 812 (71%) patients were deceased. Clinical and pathological characteristics of the patients with pancreatic cancer, by alteration of each gene and the number of altered genes, are summarised in Table 1 and supplementary material, Table S1, respectively. KRAS was wild type and mutant in 23 (2.0%) and 1,123 (98%) cases, respectively. CDKN2A expression was intact and lost in 168 (15%) and 978 (85%) cases, respectively. p53 expression was intact and aberrant in 583 (51%) and 563 (49%) cases, respectively. SMAD4 expression was intact and lost in 512 (45%) and 634 (55%) cases, respectively. There were no statistically significant differences in clinical and pathological characteristics between the molecular subgroups at the α level of 0.005. Figure 2A demonstrates different combinations of tumour status of the four major drivers in strata of the number of altered genes. We found reasonable correlations between expression status by IHC and mutation status by targeted sequencing among 53 patients with pancreatic cancer (supplementary material, Table S2). IHC showed 95% specificity and 61% sensitivity for TP53 mutation status. For CDKN2A and SMAD4 mutations, IHC showed high sensitivities (100 and 87%, respectively). In exploratory analyses, we assessed the relationship between the Moffitt tumour subtype and the number of altered genes using transcriptome data from 20 pancreatic cancer specimens and did not observe any significant association (supplementary material, Figure S3).
Table 1.
Clinical and pathological characteristics of resected pancreatic cancer cases, overall or by the tumour status of the four drivers
| KRAS mutation | CDKN2A (p16) expression | p53 expression | SMAD4 expression | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| All cases | Wild type | Mutant | Intact | Lost | Intact | Aberrant | Intact | Lost | |||||
| Characteristic* | (n = 1,146) | (n = 23) | (n = 1,123) | P | (n = 168) | (n = 978) | P | (n = 583) | (n = 563) | P | (n = 512) | (n = 634) | P |
| Mean age ± SD (years) | 67.1 ± 9.7 | 66.0 ± 13.7 | 67.2 ± 9.7 | 0.56 | 66.8 ± 9.4 | 67.2 ± 9.8 | 0.57 | 67.2 ± 9.9 | 67.1 ± 9.6 | 0.83 | 67.1 ± 9.7 | 67.2 ± 9.8 | 0.87 |
| Sex | 0.50 | 0.13 | 0.011 | 0.16 | |||||||||
| Female | 478 (42%) | 8 (35%) | 470 (42%) | 79 (47%) | 399 (41%) | 222 (38%) | 256 (45%) | 202 (39%) | 276 (44%) | ||||
| Male | 668 (58%) | 15 (65%) | 653 (58%) | 89 (53%) | 579 (59%) | 361 (62%) | 307 (55%) | 310 (61%) | 358 (56%) | ||||
| Year of diagnosis | 0.23 | 0.046 | 0.26 | 0.034 | |||||||||
| 2005–2010 | 331 (29%) | 3 (13%) | 328 (29%) | 57 (34%) | 274 (28%) | 181 (31%) | 150 (27%) | 135 (26%) | 196 (31%) | ||||
| 2011–2014 | 387 (34%) | 9 (39%) | 378 (34%) | 43 (26%) | 344 (35%) | 192 (33%) | 195 (35%) | 193 (38%) | 194 (31%) | ||||
| 2015–2017 | 428 (37%) | 11 (48%) | 417 (37%) | 68 (40%) | 360 (37%) | 210 (36%) | 218 (39%) | 184 (36%) | 244 (38%) | ||||
| Tumour location | 0.32 | 0.25 | 0.55 | 0.85 | |||||||||
| Head of the pancreas | 733 (64%) | 17 (74%) | 716 (64%) | 114 (68%) | 619 (63%) | 368 (63%) | 365 (65%) | 326 (64%) | 407 (64%) | ||||
| Body to tail of the pancreas | 413 (36%) | 6 (26%) | 407 (36%) | 54 (32%) | 359 (37%) | 215 (37%) | 198 (35%) | 186 (36%) | 227 (36%) | ||||
| Histological type | 0.020 | 0.43 | 0.052 | 0.18 | |||||||||
| Adenocarcinoma | 1116 (97%) | 20 (87%) | 1096 (98%) | 162 (96%) | 954 (98%) | 573 (98%) | 543 (96%) | 495 (97%) | 621 (98%) | ||||
| Adenosquamous carcinoma | 30 (2.6%) | 3 (13%) | 27 (2.4%) | 6 (3.6%) | 24 (2.5%) | 10 (1.7%) | 20 (3.6%) | 17 (3.3%) | 13 (2.1%) | ||||
| Tumour differentiation † | 0.23 | 0.045 | 0.26 | 0.31 | |||||||||
| Well to moderate | 650 (58%) | 9 (45%) | 641 (58%) | 106 (65%) | 544 (57%) | 343 (60%) | 307 (57%) | 280 (57%) | 370 (60%) | ||||
| Poor | 466 (42%) | 11 (55%) | 455 (42%) | 56 (35%) | 410 (43%) | 230 (40%) | 236 (43%) | 215 (43%) | 251 (40%) | ||||
| Stroma type | 0.69 | 0.55 | 0.013 | 0.24 | |||||||||
| Non‐scirrhous | 753 (66%) | 16 (70%) | 737 (66%) | 107 (64%) | 646 (66%) | 403 (69%) | 350 (62%) | 327 (64%) | 426 (67%) | ||||
| Scirrhous | 393 (34%) | 7 (30%) | 386 (34%) | 61 (36%) | 332 (34%) | 180 (31%) | 213 (38%) | 185 (36%) | 208 (33%) | ||||
| Lymphatic invasion | 0.58 | 0.033 | 0.45 | 0.65 | |||||||||
| Absent/mild | 735 (64%) | 16 (70%) | 719 (64%) | 120 (71%) | 615 (63%) | 380 (65%) | 355 (63%) | 332 (65%) | 403 (64%) | ||||
| Moderate/marked | 411 (36%) | 7 (30%) | 404 (36%) | 48 (29%) | 363 (37%) | 203 (35%) | 208 (37%) | 180 (35%) | 231 (36%) | ||||
| Venous invasion | 0.39 | 0.009 | 0.49 | 0.46 | |||||||||
| Absent/mild | 396 (35%) | 6 (26%) | 390 (35%) | 73 (43%) | 323 (33%) | 207 (36%) | 189 (34%) | 171 (33%) | 225 (35%) | ||||
| Moderate/marked | 750 (65%) | 17 (74%) | 733 (65%) | 95 (57%) | 655 (67%) | 376 (64%) | 374 (66%) | 341 (67%) | 409 (65%) | ||||
| Neural invasion | 0.20 | 0.006 | 0.76 | 0.071 | |||||||||
| Absent/mild | 404 (35%) | 11 (48%) | 393 (35%) | 75 (45%) | 329 (34%) | 208 (36%) | 196 (35%) | 195 (38%) | 209 (33%) | ||||
| Moderate/marked | 742 (65%) | 12 (52%) | 730 (65%) | 93 (55%) | 649 (66%) | 375 (64%) | 367 (65%) | 317 (62%) | 425 (67%) | ||||
| Mean tumour size ± SD (cm) | 3.4 ± 1.6 | 3.6 ± 1.5 | 3.4 ± 1.6 | 0.65 | 3.2 ± 1.3 | 3.5 ± 1.6 | 0.046 | 3.4 ± 1.6 | 3.4 ± 1.6 | 0.99 | 3.3 ± 1.5 | 3.5 ± 1.6 | 0.086 |
| Positive lymph nodes | 0.47 | 0.14 | 0.38 | 0.080 | |||||||||
| 0 | 355 (31%) | 8 (35%) | 347 (31%) | 63 (37%) | 292 (30%) | 180 (31%) | 175 (31%) | 176 (34%) | 179 (28%) | ||||
| 1–3 | 460 (40%) | 11 (48%) | 449 (40%) | 62 (37%) | 398 (41%) | 244 (42%) | 216 (38%) | 197 (38%) | 263 (41%) | ||||
| ≥4 | 331 (29%) | 4 (17%) | 327 (29%) | 43 (26%) | 288 (29%) | 159 (27%) | 172 (31%) | 139 (27%) | 192 (30%) | ||||
| UICC cancer stage | 0.48 | 0.032 | 0.67 | 0.24 | |||||||||
| I | 298 (26%) | 8 (35%) | 290 (26%) | 59 (35%) | 239 (24%) | 157 (27%) | 141 (25%) | 148 (29%) | 150 (24%) | ||||
| II | 499 (44%) | 11 (48%) | 488 (44%) | 66 (39%) | 433 (44%) | 258 (44%) | 241 (43%) | 216 (42%) | 283 (45%) | ||||
| III | 301 (26%) | 4 (17%) | 297 (26%) | 38 (23%) | 263 (27%) | 145 (25%) | 156 (28%) | 127 (25%) | 174 (27%) | ||||
| IV | 48 (4.2%) | 0 | 48 (4.3%) | 5 (3.0%) | 43 (4.4%) | 23 (4.0%) | 25 (4.4%) | 21 (4.1%) | 27 (4.3%) | ||||
| Resection margin status | 0.82 | 0.11 | 0.47 | 0.89 | |||||||||
| R0 | 888 (77%) | 19 (83%) | 869 (77%) | 139 (83%) | 749 (77%) | 448 (77%) | 440 (78%) | 400 (78%) | 488 (77%) | ||||
| R1 | 253 (22%) | 4 (17%) | 249 (22%) | 28 (17%) | 225 (23%) | 131 (22%) | 122 (22%) | 110 (21%) | 143 (23%) | ||||
| R2 | 5 (0.4%) | 0 | 5 (0.5%) | 1 (0.6%) | 4 (0.4%) | 4 (0.7%) | 1 (0.2%) | 2 (0.4%) | 3 (0.5%) | ||||
| Neoadjuvant therapy | 0.52 | 0.27 | 0.17 | 0.19 | |||||||||
| None | 999 (87%) | 19 (83%) | 980 (87%) | 142 (85%) | 857 (88%) | 516 (89%) | 483 (86%) | 439 (86%) | 560 (88%) | ||||
| Chemotherapy ‡ | 147 (13%) | 4 (17%) | 143 (13%) | 26 (15%) | 121 (12%) | 67 (11%) | 80 (14%) | 73 (14%) | 74 (12%) | ||||
| Adjuvant therapy | 0.30 | 0.68 | 0.84 | 0.97 | |||||||||
| None | 239 (21%) | 7 (30%) | 232 (21%) | 33 (20%) | 206 (21%) | 123 (21%) | 116 (21%) | 107 (21%) | 132 (21%) | ||||
| Chemotherapy ‡ | 907 (79%) | 16 (70%) | 891 (79%) | 135 (80%) | 772 (79%) | 460 (79%) | 447 (79%) | 405 (79%) | 502 (79%) | ||||
UICC, Union for International Cancer Control.
Percentage indicates the proportion of cases with a specific clinical or pathological characteristic in all cases or in each stratum of molecular subtypes. Total percentages may not equal 100% due to rounding.
Tumour differentiation was assessed only for adenocarcinomas.
These categories include chemoradiotherapy.
Figure 2.
Associations of the number of altered genes with survival times among patients with resected pancreatic cancer. (A) Different combinations of tumour status of the four major drivers in strata of the number of altered genes. (B and C) Kaplan–Meier survival curves of DFS and OS times, respectively, according to the number of altered genes (0–2 and 3 versus 4 genes). (D and E) Meta‐analyses of associations of the number of altered genes (3–4 genes compared to 0–2 genes) with DFS and OS times, respectively. Squares and horizontal lines indicate multivariable‐adjusted HRs and 95% CIs, respectively. The area of the square reflects cohort‐specific weight (inverse of the variance). Diamonds indicate pooled multivariable‐adjusted HRs (centre) and 95% CIs (width). HRs were adjusted for the same set of covariates as Table 2. JFCR, Japanese Foundation for Cancer Research; KU, Keio University; UT, The University of Tokyo.
In our primary analyses, we examined associations of the status of the four driver genes with survival outcomes among pancreatic cancer patients (Table 2 and supplementary material, Table S3). Loss of SMAD4 expression was associated with DFS [multivariable hazard ratio (HR), 1.24; 95% CI, 1.09–1.43] and OS (multivariable HR, 1.27; 95% CI, 1.10–1.46). The status of KRAS mutation, CDKN2A expression, or p53 expression was not associated with survival times in multivariable Cox regression analyses (p > 0.06). The number of altered genes was inversely associated with DFS and OS times: compared to cases with 0–2 altered genes, cases with three and four altered genes had multivariable HRs for DFS of 1.22 (95% CI, 1.04–1.42) and 1.32 (95% CI, 1.10–1.59), respectively (p trend = 0.003); and multivariable HRs for OS of 1.28 (95% CI, 1.09–1.51) and 1.47 (95% CI, 1.22–1.78), respectively (p trend < 0.001). Kaplan–Meier survival curves demonstrated similar associations of the gene alterations with DFS and OS (Figure 2B,C for the number of genes and supplementary material, Figure S4 for each gene). In meta‐analyses, we observed no statistically significant heterogeneity in the survival associations of the number of altered genes between the institutional cohorts (p heterogeneity > 0.58, Figure 2D,E) and, thus, pooled the institutional cohorts for all survival analyses. When we examined different combinations of altered genes in cases with two or three altered genes, we observed no differential associations with DFS and OS times by the combinations (supplementary material, Table S4). The limited number of KRAS‐wild type cases precluded a robust statistical assessment of different combinations in this subgroup (data not shown). In addition, we investigated the additive effect of loss of SMAD4 expression on specific combinations of the other driver alterations (supplementary material, Table S5). Loss of SMAD4 expression was associated with shorter DFS and OS if additionally present in tumours with KRAS and CDKN2A alterations. Such additive effect of SMAD4 loss was not observed in tumours with other combinations of the other driver alterations. We observed no statistical interaction between the four driver alterations in relation to survival times at the α level of 0.005 (p interaction > 0.007, supplementary material, Figure S5). Among 907 patients receiving adjuvant chemotherapy, the regimens were S‐1 in 469 patients (52%), gemcitabine‐based in 337 (37%), others in 93 (10%), and unknown in 8 (0.9%). In stratified analyses, similar survival associations of SMAD4 expression and the number of altered genes were observed irrespective of receipt of adjuvant chemotherapy (p interaction > 0.19, supplementary material, Table S6). When we examined a specific regimen (S‐1 or gemcitabine‐based), we also observed no statistical interactions in relation to survival times (p interaction > 0.12).
Table 2.
The tumour status of the four major drivers and survival among patients with resected pancreatic cancer
| DFS | OS | |||||||
|---|---|---|---|---|---|---|---|---|
| No. of cases | No. of events | Univariable HR (95% CI) | Multivariable HR* (95% CI) | No. of cases | No. of events | Univariable HR (95% CI) | Multivariable HR* (95% CI) | |
| KRAS mutation | ||||||||
| Wild type | 23 | 15 | 1 (reference) | 1 (reference) | 23 | 13 | 1 (reference) | 1 (reference) |
| Mutant | 1,070 | 850 | 1.23 (0.74–2.05) | 1.06 (0.63–1.78) | 1,123 | 799 | 1.27 (0.73–2.20) | 1.07 (0.62–1.86) |
| P | 0.43 | 0.83 | 0.40 | 0.81 | ||||
| CDKN2A (p16) expression | ||||||||
| Intact | 163 | 125 | 1 (reference) | 1 (reference) | 168 | 113 | 1 (reference) | 1 (reference) |
| Lost | 930 | 740 | 1.22 (1.00–1.47) | 1.10 (0.91–1.33) | 978 | 699 | 1.32 (1.08–1.61) | 1.21 (0.99–1.48) |
| P | 0.046 | 0.35 | 0.006 | 0.065 | ||||
| p53 expression | ||||||||
| Intact | 557 | 438 | 1 (reference) | 1 (reference) | 583 | 404 | 1 (reference) | 1 (reference) |
| Aberrant | 536 | 427 | 1.10 (0.96–1.26) | 1.07 (0.94–1.23) | 563 | 408 | 1.13 (0.98–1.30) | 1.12 (0.98–1.29) |
| P | 0.17 | 0.32 | 0.082 | 0.10 | ||||
| SMAD4 expression | ||||||||
| Intact | 490 | 373 | 1 (reference) | 1 (reference) | 512 | 346 | 1 (reference) | 1 (reference) |
| Lost | 603 | 492 | 1.20 (1.05–1.38) | 1.24 (1.09–1.43) | 634 | 466 | 1.25 (1.08–1.43) | 1.27 (1.10–1.46) |
| P | 0.008 | 0.002 | 0.002 | 0.001 | ||||
| No. of altered genes | ||||||||
| 0–2 | 350 | 263 | 1 (reference) | 1 (reference) | 362 | 238 | 1 (reference) | 1 (reference) |
| 3 | 487 | 391 | 1.29 (1.10–1.51) | 1.22 (1.04–1.42) | 513 | 367 | 1.36 (1.15–1.60) | 1.28 (1.09–1.51) |
| 4 | 256 | 211 | 1.36 (1.13–1.63) | 1.32 (1.10–1.59) | 271 | 207 | 1.52 (1.26–1.83) | 1.47 (1.22–1.78) |
| P trend † | <0.001 | 0.003 | <0.001 | <0.001 | ||||
The multivariable Cox regression model initially included age, sex, year of diagnosis, tumour location, histological type, cancer stage, resection margin status, receipt of neoadjuvant chemotherapy, and receipt of adjuvant chemotherapy. Backward elimination with a threshold p of 0.05 was conducted to select variables for the final models. The variables that remained in the final models are described in supplementary material, Table S2.
P trend was calculated by entering the number of altered genes (continuous, 0–4) in the Cox regression model.
In secondary analyses, we examined recurrence patterns of resected pancreatic cancer (supplementary material, Table S7) and their correlation with the number of altered genes. The number of altered genes was positively associated with the risk of distant metastasis [multivariable subdistribution HR 1.35 (95% CI, 1.11–1.64) and 1.23 (95% CI, 0.98–1.55) for three and four altered genes (versus 2), respectively; p trend = 0.096], but not with the risk of local recurrence (p trend > 0.39, Table 3). When we further examined specific patterns of distant metastasis, multivariable competing risks regression analyses demonstrated a stronger association of the number of altered genes for liver metastasis (p trend = 0.006) than for non‐liver metastasis (p trend > 0.35). Cumulative incidence curves of the recurrence patterns by the number of altered genes are presented in Figure 3. To investigate whether pancreatic carcinomas harbouring only KRAS mutation among the four drivers showed specific recurrence patterns, we examined recurrence patterns of KRAS‐mutated tumours with or without other driver alterations (aberrant expression of CDKN2A, p53, and/or SMAD4) and did not observe any significant difference between them (supplementary material, Table S8).
Table 3.
The number of altered genes in relation to incidences of specific recurrence patterns among patients with resected pancreatic cancer
| No. of altered genes | ||||
|---|---|---|---|---|
| 0–2 | 3 | 4 | P trend * | |
| Distant versus local † | ||||
| Distant only (n = 563) | ||||
| n | 159 | 271 | 133 | |
| Univariable SHR (95% CI) | 1 (reference) | 1.40 (1.16–1.68) | 1.27 (1.02–1.59) | 0.037 |
| Multivariable SHR (95% CI) ‡ | 1 (reference) | 1.35 (1.11–1.64) | 1.23 (0.98–1.55) | 0.096 |
| Distant + local (n = 128) | ||||
| n | 36 | 57 | 35 | |
| Univariable SHR (95% CI) | 1 (reference) | 1.30 (0.86–1.95) | 1.41 (0.90–2.23) | 0.11 |
| Multivariable SHR (95% CI) ‡ | 1 (reference) | 1.28 (0.86–1.93) | 1.39 (0.88–2.19) | 0.12 |
| Local only (n = 174) | ||||
| n | 68 | 63 | 43 | |
| Univariable SHR (95% CI) | 1 (reference) | 0.63 (0.45–0.88) | 0.87 (0.60–1.25) | 0.47 |
| Multivariable SHR (95% CI) ‡ | 1 (reference) | 0.61 (0.43–0.85) | 0.84 (0.58–1.22) | 0.40 |
| Specific site of distant metastasis | ||||
| Liver (n = 283) § | ||||
| n | 65 | 137 | 81 | |
| Univariable SHR (95% CI) | 1 (reference) | 1.55 (1.17–2.05) | 1.63 (1.19–2.23) | 0.003 |
| Multivariable SHR (95% CI) ‡ | 1 (reference) | 1.44 (1.08–1.92) | 1.58 (1.15–2.16) | 0.006 |
| Peritoneum (n = 146) § | ||||
| n | 39 | 76 | 31 | |
| Univariable SHR (95% CI) | 1 (reference) | 1.37 (0.95–1.98) | 1.01 (0.64–1.58) | 0.72 |
| Multivariable SHR (95% CI) ‡ | 1 (reference) | 1.37 (0.95–1.98) | 1.01 (0.64–1.58) | 0.72 |
| Lung (n = 150) § | ||||
| n | 49 | 70 | 31 | |
| Univariable SHR (95% CI) | 1 (reference) | 0.93 (0.65–1.33) | 0.76 (0.49–1.18) | 0.36 |
| Multivariable SHR (95% CI) ‡ | 1 (reference) | 0.94 (0.65–1.34) | 0.76 (0.49–1.19) | 0.36 |
| Lymph node (n = 131) § | ||||
| n | 37 | 63 | 31 | |
| Univariable SHR (95% CI) | 1 (reference) | 1.22 (0.83–1.79) | 1.10 (0.69–1.73) | 0.52 |
| Multivariable SHR (95% CI) ‡ | 1 (reference) | 1.22 (0.83–1.79) | 1.10 (0.69–1.73) | 0.52 |
SHR, subdistribution hazard ratio.
P trend was calculated by entering the number of altered genes (continuous, 0–4) in the competing risks regression model.
Three recurrence patterns (distant only versus distant and local versus local only) were treated as competing risk events.
The multivariable competing risks regression model initially included age, sex, year of diagnosis, tumour location, histological type, cancer stage, resection margin status, receipt of neoadjuvant chemotherapy, and receipt of adjuvant chemotherapy. Backward elimination with a threshold p of 0.05 was conducted to select variables for the final models.
In an analysis of a specific site of distant metastasis, recurrence at other sites only was treated as a competing risk event.
Figure 3.
Cumulative incidence curves of recurrence at specific sites among patients with resected pancreatic cancer according to the number of altered genes. (A) Distant metastasis (with or without local recurrence), (B) local recurrence only, (C) liver metastasis, and (D) recurrence at sites other than the liver.
Discussion
In this large multi‐institutional cohort study of patients with resected pancreatic cancer, we have shown that the number of altered genes among the four major drivers is inversely associated with post‐operative DFS and OS times. These findings suggest that those driver genes play a pivotal role in pancreatic cancer progression after surgical resection as well as in pancreatic carcinogenesis. Of note, pancreatic tumours with a greater number of the altered genes were more likely to exhibit liver metastasis rather than local recurrence or recurrence at other distant sites. The survival associations of the altered genes were consistently observed in an analysis limited to patients receiving adjuvant chemotherapy. Our data underline not only the potential of those fundamental molecular alterations as biomarkers for survival among patients with pancreatic cancer but also the importance of designing specific treatment strategies for pancreatic carcinomas with the unfavourable genetic and epigenetic profile defined by the four tumour markers.
Pancreatic cancer is a collective disease entity consisting of molecularly heterogeneous neoplasms that evolve on the basis of sequential accumulation of genetic and epigenetic aberrations and represent distinctive tumour behavioural patterns according to their molecular architecture [1, 2, 3, 35, 36, 37]. Experimental and clinical studies attest that alterations of the four major driver genes (KRAS, CDKN2A, TP53, and SMAD4) can individually or collectively exhibit cancer‐promoting properties through accelerating cell proliferation, migration, and invasion [13, 14, 15, 16, 17, 38, 39, 40, 41]. For example, SMAD4 serves as a mediator in the TGFB1 (TGF‐β) signalling pathway, which has been involved in the progression of pancreatic intraepithelial neoplasia to invasive adenocarcinoma [16, 42]. In addition, those molecular alterations potentially form the immunosuppressive microenvironment, as illustrated by preclinical and clinical data supporting the roles of KRAS‐ and TP53‐related signalling pathways in immune evasion of pancreatic cancer [11, 12, 13, 43]. In line with these mechanistic data, the current study points to the roles of SMAD4 expression and accumulation of the four major driver alterations in tumour progression following clinical diagnosis and surgical resection. Few large cohort studies have investigated the responsiveness of pancreatic cancer to conventional chemotherapy regimens (i.e. 5‐fluorouracil‐ or gemcitabine‐based regimens) by the status of the four major driver alterations [44, 45, 46]. In the current study, the high mortality hazard associated with those driver alterations appeared to be unamenable to the administration of standard adjuvant chemotherapy. The four major driver alterations have long been untargettable, and no molecular‐targeted agents have been approved, resulting in poor prognosis of pancreatic cancer patients. However, inhibitors selectively targeting KRAS G12C mutation (e.g. sotorasib) have emerged as promising therapeutics [47, 48] and have opened opportunities for precision oncology for pancreatic cancer. In addition, investigations are underway to examine the effectiveness of agents on tumours harbouring other driver mutations, such as TP53 mutations [49, 50]. Our findings implicate the importance of developing agents targeting the signalling pathways altered by the four primary drivers.
In our study, pancreatic cancer with an increasing number of driver alterations was more likely to represent hepatic metastasis at an earlier stage of the post‐operative course. Our large multicentre study likely provides robust evidence supporting the roles of the four major driver genes in determining post‐operative tumour progression patterns as well as overall mortality hazard. Interestingly, the number of altered driver genes was not associated with any histopathological features of pancreatic cancer, including the number of positive lymph nodes and histopathological lymphatic invasion. Therefore, it is considered that pancreatic tumours with accumulated molecular drivers may metastasise through the hematogenous mechanism rather than the lymphogenic mechanism. Accumulating biallelic loss of cell‐cycle regulation genes, including CDKN2A, TP53, and SMAD4, has contributed to a stepwise acceleration in cell cycle progression in pancreatic cancer [51]. A previous study has suggested that an increasing number of genetic alterations of KRAS, CDKN2A, TP53, and SMAD4 confer niche independency in engineered organoid models of pancreatic cancer [52]. Another mechanistic study using pancreatic cancer organoids has shown that SMAD4 inactivation is required for the collective invasion of tumour cells [53]. These lines of evidence suggest that the four driver alterations in pancreatic carcinoma cells may co‐ordinately drive highly metastatic phenotype. Future basic research is warranted to better understand the mechanisms underlying the development of liver‐specific metastasis of pancreatic cancer.
In pancreatic carcinomas, SMAD4 has been inactivated through homozygous deletion in 35% and a monoallelic mutation with loss of the other allele in 20% [54]. CDKN2A alterations in pancreatic carcinomas occur based on homozygous deletion in 40%, a monoallelic mutation with loss of the other allele in 40%, and hypermethylation in the gene promoter in 15% [54, 55]. In our NGS (next‐generation sequencing)‐based analysis of 53 pancreatic cancer specimens, the prevalence of mutations in SMAD4 and CDKN2A genes was 28 and 18%, respectively. This result is likely reasonable because the mutation calling by targeted sequencing approach underestimates the alterations in SMAD4 and CDKN2A as this assay cannot detect uncommon mutations, large genomic deletion, or hypermethylation in promoter regions. In our IHC‐based assessment, loss of expressions of SMAD4 and CDKN2A was found in 55 and 85%, respectively. Of note, IHC of SMAD4 and CDKN2A could detect genetic mutations in corresponding genes with high sensitivity (87 and 100%, respectively). SMAD4 IHC has been a powerful tool for detecting the genetic status of SMAD4 [5, 22]. Previous studies have also suggested that IHC may be a surrogate method to detect CDKN2A promoter methylation and subsequent gene silencing in various cancer types. For example, a previous study of 766 colorectal carcinomas has shown that loss of immunohistochemical expression of CDKN2A is more common in CDKN2A‐methylated carcinomas than in CDKN2A‐unmethylated carcinomas (61% versus 8.7%, respectively) [56]. Collectively, IHC may be a simple and useful method in assessing SMAD4 and CDKN2A alterations in pancreatic cancer.
On the other hand, we should acknowledge a limitation of the IHC‐based approach for assessing TP53 status. The TP53 gene is inactivated in 75% of pancreatic adenocarcinoma cases by loss of one allele coupled with an intragenic mutation in the second allele [54]. In the current study, the prevalence of aberrant p53 status examined by IHC was lower (51%) than expected. Indeed, our IHC‐based assessment could predict TP53 mutation status with high specificity (95%) but with moderate sensitivity (61%), suggesting that some TP53 mutated tumours might be misclassified as intact by IHC. This result is in line with previous studies reporting the correlations between IHC‐based findings and TP53 mutation status in several types of malignancies [57, 58]. Taken together, IHC is not a perfect method but a surrogate for estimating intragenic mutations of the TP53 gene.
Transcriptome‐based subtypes of pancreatic cancer (i.e. basal‐like subtype, squamous subtype, and quasi‐mesenchymal subtype) have been associated with poor prognosis and liver metastasis [26, 51, 59, 60, 61, 62, 63, 64, 65, 66]. In our exploratory analysis of 20 pancreatic carcinomas, we did not observe any significant association between the number of altered genes and basal‐like tumour subtypes. However, Bailey et al have attested to the importance of TP53 mutation in increasing squamous subtype‐related gene expressions in pancreatic cancer [66], although TP53 mutation is not necessarily accumulated in pancreatic cancer with squamous or basal‐like features [67, 68]. Integrated whole‐genome analysis has demonstrated that TP53 mutation, CDKN2A loss, and mutated‐KRAS amplification are associated with basal‐like pancreatic cancer subtype, whereas complete SMAD4 loss is correlated with classical subtype [62]. Future comprehensive analyses of genetic and epigenetic alterations of KRAS, CDKN2A, TP53, and SMAD4 genes may unveil a potential link between transcriptome‐based subtypes and those driver alterations in pancreatic cancer.
The use of reliable assays for KRAS mutation was a strength of our study. Low tumour cellularity due to abundant desmoplastic stroma characteristic of pancreatic cancer [69, 70] has hampered DNA‐based molecular annotation in clinical tumour samples [14]. As previously described [20], the use of ddPCR allowed for the sensitive detection of KRAS mutations in decades‐old FFPE (formalin‐fixed paraffin‐embedded) tissue samples of pancreatic cancer. The current study has other notable strengths, including the large sample size based on three independent patient populations. The large sample size allowed us to conduct various subgroup analyses with sufficient statistical power. In this multicentre study, we observed no statistically significant heterogeneity between the institutional cohorts in our main findings, supporting the generalisability of our data. The availability of comprehensive clinical data on pre‐ and post‐operative chemotherapy regimens and various survival outcomes (e.g. various recurrence types and all‐cause death) was another strength.
The current study has several limitations. First, there might be unmeasured confounding factors in our survival analyses. Nonetheless, our multivariable models included a variety of clinical and pathological characteristics, including established prognostic factors in patients with pancreatic cancer, and the adjustment did not alter the results materially. Second, a lack of gold‐standard criteria for IHC assessments of CDKN2A, p53, and SMAD4 expression is another limitation; nonetheless, it is more likely that misclassifications of the expression status would have driven the survival associations of the altered expression levels toward the null hypothesis. Third, our study population consisted of patients with resected pancreatic ductal adenocarcinoma. Therefore, the usefulness of the driver alterations as prognostic biomarkers should be validated in patients with locally advanced or distant adenocarcinomas and other adenocarcinoma variants (e.g. intraductal papillary mucinous neoplasm‐derived carcinoma and colloid carcinoma). Finally, a vast majority of the study population was Japanese. Therefore, our findings should be validated in independent populations with racial diversity.
We have demonstrated the feasibility of IHC‐based assessments of these driver alterations in small tissue samples obtained via fine‐needle aspiration. As described, ddPCR has been shown to provide a reliable mutation calling in specimens with low DNA amounts. In the current practice, an increasing number of patients diagnosed with pancreatic cancer are referred to neoadjuvant chemotherapy. A further investigation is warranted to examine the potential of those simple and cost‐effective analyses in identifying patients with highly aggressive tumours who are less likely to benefit from neoadjuvant chemotherapy and thus are candidates for upfront surgery.
In summary, the current study suggests that the accumulation of the four major driver alterations involved in pancreatic carcinogenesis can confer a high metastatic potential to tumours, thereby impairing post‐operative survival among patients with pancreatic cancer. The elevated mortality hazard associated with an increasing number of the altered genes appeared not to be offset by the standard chemotherapy regimens used in the current clinical practice. Therefore, further research is warranted to identify targetable molecular and immune parameters in resected pancreatic cancer with the accumulated molecular alterations and, thereby, improve clinical outcomes of patients with this highly lethal tumour subgroup.
Author contributions statement
YM, MTak, MTan and TH planned the study. All authors collected, analysed and interpreted the data. YM and TH drafted the manuscript, and all authors participated in revising the manuscript. YM, MTak, TH, TS, YN and TT obtained funding. YM, TH, TU, KTak and MS supervised the study. All authors approved the final submitted manuscript.
Ethics approval and consent to participate
This study was designed and conducted according to the guidelines in the Helsinki Declaration and was approved by the ethics committees at Cancer Institute Hospital, Japanese Foundation for Cancer Research, The University of Tokyo, and Keio University School of Medicine (all in Tokyo, Japan). Informed consent was obtained from all patients on an opt‐out basis.
Supporting information
Supplementary methods
Figure S1. Microscopic images of H&E and immunohistochemistry staining in a fine‐needle aspiration specimen from a patient with pancreatic cancer
Figure S2. Flow diagram of the selection of patients with resected pancreatic cancer in a multi‐institutional cohort
Figure S3. Association between the Moffitt tumour subtype and the number of altered genes
Figure S4. Kaplan–Meier survival curves of patients with pancreatic cancer according to altered genes/proteins
Figure S5. Interactions of altered genes in relation to survival patients with pancreatic cancer
Table S1. Clinical and pathological characteristics of pancreatic cancer cases by the number of altered genes
Table S2. Correlations of tumour status based on immunohistochemical expression of CDKN2A (p16), p53, and SMAD4 with mutation calls by targeted sequencing
Table S3. The tumour status of the four drivers and survival among patients with pancreatic cancer (the final multivariable models)
Table S4. Specific combinations of the four major driver alterations and survival among patients with resected pancreatic cancer
Table S5. SMAD4 expression status in addition to specific combinations of the other three major driver alterations and survival among patients with resected pancreatic cancer
Table S6. The tumour status of the four drivers and survival among patients with pancreatic cancer, stratified by administration of adjuvant chemotherapy
Table S7. Recurrence patterns of resected pancreatic cancer by the number of altered genes
Table S8. Recurrence patterns of resected pancreatic cancer harbouring KRAS mutation with or without other driver alterations
Acknowledgements
The authors would like to thank the following collaborators for their valuable support in tissue processing and/or data collection: Satoko Baba and Motoyoshi Iwakoshi, Department of Pathology, Cancer Institute Hospital, Japanese Foundation for Cancer Research, Tokyo, Japan; Kikuko Kaji, Department of Hepato‐Biliary‐Pancreatic Medicine, Cancer Institute Hospital, Japanese Foundation for Cancer Research, Tokyo, Japan; Kei Sakuma, Department of Pathology, Graduate School of Medicine, The University of Tokyo, Tokyo, Japan; Noriko Koga, Hepato‐Biliary‐Pancreatic Surgery Division, Department of Surgery, Graduate School of Medicine, The University of Tokyo, Tokyo, Japan; and the staff of the Fourth Laboratory of Department of Pathology in Keio University School of Medicine, Tokyo, Japan. This work was supported by Japan Society for the Promotion of Science (JSPS) KAKENHI Grants (JP20K07414 to YM, JP21K15393 to MTak, JP19K08362 and JP22H02841 to TH, and JP21K15368 to TS), by the Practical Research for Innovative Cancer Control Program from AMED (JP21ck0106557 to YN), and by Grants from Takeda Science Foundation (to TH), Daiwa Securities Health Foundation (to TT), and Pancreas Research Foundation of Japan (to TT).
No conflicts of interest were declared.
Contributor Information
Yohei Masugi, Email: masugi@z6.keio.jp.
Tsuyoshi Hamada, Email: hamada-tky@umin.ac.jp.
Data availability statement
The data sets generated and/or analysed during the current study are available from the corresponding authors upon reasonable request. The microarray data were deposited in NCBI's Gene Expression Omnibus database (accession number, GSE157353).
References
- 1. Wood LD, Canto MI, Jaffee EM, et al. Pancreatic cancer: pathogenesis, screening, diagnosis, and treatment. Gastroenterology 2022; 163: 386–402.e1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Connor AA, Gallinger S. Pancreatic cancer evolution and heterogeneity: integrating omics and clinical data. Nat Rev Cancer 2022; 22: 131–142. [DOI] [PubMed] [Google Scholar]
- 3. Hayashi A, Hong J, Iacobuzio‐Donahue CA. The pancreatic cancer genome revisited. Nat Rev Gastroenterol Hepatol 2021; 18: 469–481. [DOI] [PubMed] [Google Scholar]
- 4. McIntyre CA, Lawrence SA, Richards AL, et al. Alterations in driver genes are predictive of survival in patients with resected pancreatic ductal adenocarcinoma. Cancer 2020; 126: 3939–3949. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Qian ZR, Rubinson DA, Nowak JA, et al. Association of alterations in main driver genes with outcomes of patients with resected pancreatic ductal adenocarcinoma. JAMA Oncol 2018; 4: e173420. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Hayashi H, Kohno T, Ueno H, et al. Utility of assessing the number of mutated KRAS, CDKN2A, TP53, and SMAD4 genes using a targeted deep sequencing assay as a prognostic biomarker for pancreatic cancer. Pancreas 2017; 46: 335–340. [DOI] [PubMed] [Google Scholar]
- 7. Oshima M, Okano K, Muraki S, et al. Immunohistochemically detected expression of 3 major genes (CDKN2A/p16, TP53, and SMAD4/DPC4) strongly predicts survival in patients with resectable pancreatic cancer. Ann Surg 2013; 258: 336–346. [DOI] [PubMed] [Google Scholar]
- 8. Yachida S, White CM, Naito Y, et al. Clinical significance of the genetic landscape of pancreatic cancer and implications for identification of potential long‐term survivors. Clin Cancer Res 2012; 18: 6339–6347. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Schlitter AM, Segler A, Steiger K, et al. Molecular, morphological and survival analysis of 177 resected pancreatic ductal adenocarcinomas (PDACs): identification of prognostic subtypes. Sci Rep 2017; 7: 41064. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Yokose T, Kitago M, Matsuda S, et al. Combination of KRAS and SMAD4 mutations in formalin‐fixed paraffin‐embedded tissues as a biomarker for pancreatic cancer. Cancer Sci 2020; 111: 2174–2182. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Ischenko I, D'Amico S, Rao M, et al. KRAS drives immune evasion in a genetic model of pancreatic cancer. Nat Commun 2021; 12: 1482. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Leinwand J, Miller G. Regulation and modulation of antitumor immunity in pancreatic cancer. Nat Immunol 2020; 21: 1152–1159. [DOI] [PubMed] [Google Scholar]
- 13. Hashimoto S, Furukawa S, Hashimoto A, et al. ARF6 and AMAP1 are major targets of KRAS and TP53 mutations to promote invasion, PD‐L1 dynamics, and immune evasion of pancreatic cancer. Proc Natl Acad Sci U S A 2019; 116: 17450–17459. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Buscail L, Bournet B, Cordelier P. Role of oncogenic KRAS in the diagnosis, prognosis and treatment of pancreatic cancer. Nat Rev Gastroenterol Hepatol 2020; 17: 153–168. [DOI] [PubMed] [Google Scholar]
- 15. Journo S, Goldberg AK, Sokol ES, et al. Genomic alterations drive metastases formation in pancreatic ductal adenocarcinoma cancer: deciphering the role of CDKN2A and CDKN2B in mediating liver tropism. Oncogene 2022; 41: 1468–1481. [DOI] [PubMed] [Google Scholar]
- 16. Gough NR, Xiang X, Mishra L. TGF‐β signaling in liver, pancreas, and gastrointestinal diseases and cancer. Gastroenterology 2021; 161: 434–452.e15. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Escobar‐Hoyos LF, Penson A, Kannan R, et al. Altered RNA splicing by mutant p53 activates oncogenic RAS signaling in pancreatic cancer. Cancer Cell 2020; 38: 198–211.e8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Nie RC, Zou XB, Yuan SQ, et al. Disease‐free survival as a surrogate endpoint for overall survival in adjuvant trials of pancreatic cancer: a meta‐analysis of 20 randomized controlled trials. BMC Cancer 2020; 20: 421. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Siegel RL, Miller KD, Fuchs HE, et al. Cancer statistics, 2022. CA Cancer J Clin 2022; 72: 7–33. [DOI] [PubMed] [Google Scholar]
- 20. Suzuki T, Masugi Y, Inoue Y, et al. KRAS variant allele frequency, but not mutation positivity, associates with survival of patients with pancreatic cancer. Cancer Sci 2022; 113: 3097–3109. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Ogawa Y, Masugi Y, Abe T, et al. Three distinct stroma types in human pancreatic cancer identified by image analysis of fibroblast subpopulations and collagen. Clin Cancer Res 2021; 27: 107–119. [DOI] [PubMed] [Google Scholar]
- 22. Wilentz RE, Su GH, Dai JL, et al. Immunohistochemical labeling for dpc4 mirrors genetic status in pancreatic adenocarcinomas: a new marker of DPC4 inactivation. Am J Pathol 2000; 156: 37–43. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Wilentz RE, Geradts J, Maynard R, et al. Inactivation of the p16 (INK4A) tumor‐suppressor gene in pancreatic duct lesions: loss of intranuclear expression. Cancer Res 1998; 58: 4740–4744. [PubMed] [Google Scholar]
- 24. Chapel DB, Schulte JJ, Berg K, et al. MTAP immunohistochemistry is an accurate and reproducible surrogate for CDKN2A fluorescence in situ hybridization in diagnosis of malignant pleural mesothelioma. Mod Pathol 2020; 33: 245–254. [DOI] [PubMed] [Google Scholar]
- 25. Fujita Y, Matsuda S, Sasaki Y, et al. Pathogenesis of multiple pancreatic cancers involves multicentric carcinogenesis and intrapancreatic metastasis. Cancer Sci 2020; 111: 739–748. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Moffitt RA, Marayati R, Flate EL, et al. Virtual microdissection identifies distinct tumor‐ and stroma‐specific subtypes of pancreatic ductal adenocarcinoma. Nat Genet 2015; 47: 1168–1178. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Cochran WG. The combination of estimates from different experiments. Biometrics 1954; 10: 101–129. [Google Scholar]
- 28. DerSimonian R, Laird N. Meta‐analysis in clinical trials. Control Clin Trials 1986; 7: 177–188. [DOI] [PubMed] [Google Scholar]
- 29. Hamada T, Cao Y, Qian ZR, et al. Aspirin use and colorectal cancer survival according to tumor CD274 (programmed cell death 1 ligand 1) expression status. J Clin Oncol 2017; 35: 1836–1844. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Marubini E, Valsecchi MG. Analysing Survival Data from Clinical Trials and Observational Studies. John Wiley and Sons: New York, 1995. [Google Scholar]
- 31. Gray RJ. A class of K‐sample tests for comparing the cumulative incidence of a competing risk. Ann Stat 1988; 16: 1141–1154. [Google Scholar]
- 32. Fine JP, Gray RJ. A proportional hazards model for the subdistribution of a competing risk. J Am Stat Assoc 1999; 94: 496–509. [Google Scholar]
- 33. Kohl M, Plischke M, Leffondré K, et al. PSHREG: a SAS macro for proportional and nonproportional subdistribution hazards regression. Comput Methods Programs Biomed 2015; 118: 218–233. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34. Benjamin DJ, Berger JO, Johannesson M, et al. Redefine statistical significance. Nat Hum Behav 2018; 2: 6–10. [DOI] [PubMed] [Google Scholar]
- 35. Ho WJ, Jaffee EM, Zheng L. The tumour microenvironment in pancreatic cancer – clinical challenges and opportunities. Nat Rev Clin Oncol 2020; 17: 527–540. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. O'Kane GM, Grünwald BT, Jang GH, et al. GATA6 expression distinguishes classical and basal‐like subtypes in advanced pancreatic cancer. Clin Cancer Res 2020; 26: 4901–4910. [DOI] [PubMed] [Google Scholar]
- 37. N Kalimuthu S, Wilson GW, Grant RC, et al. Morphological classification of pancreatic ductal adenocarcinoma that predicts molecular subtypes and correlates with clinical outcome. Gut 2020; 69: 317–328. [DOI] [PubMed] [Google Scholar]
- 38. Lillemoe HA, Passot G, Kawaguchi Y, et al. RAS/TP53 co‐mutation is associated with worse survival after concurrent resection of colorectal liver metastases and extrahepatic disease. Ann Surg 2022; 276: 357–362. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39. Tiwari A, Tashiro K, Dixit A, et al. Loss of HIF1A from pancreatic cancer cells increases expression of PPP1R1B and degradation of p53 to promote invasion and metastasis. Gastroenterology 2020; 159: 1882–1897.e5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40. Hu S, Chen X, Xu X, et al. STRAP as a new therapeutic target for poor prognosis of pancreatic ductal adenocarcinoma patients mainly caused by TP53 mutation. Front Oncol 2020; 10: 594224. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41. Collins MA, Bednar F, Zhang Y, et al. Oncogenic Kras is required for both the initiation and maintenance of pancreatic cancer in mice. J Clin Invest 2012; 122: 639–653. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42. David CJ, Huang YH, Chen M, et al. TGF‐β tumor suppression through a lethal EMT. Cell 2016; 164: 1015–1030. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43. Maddalena M, Mallel G, Nataraj NB, et al. TP53 missense mutations in PDAC are associated with enhanced fibrosis and an immunosuppressive microenvironment. Proc Natl Acad Sci U S A 2021; 118: e2025631118. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44. Hoyer K, Hablesreiter R, Inoue Y, et al. A genetically defined signature of responsiveness to erlotinib in early‐stage pancreatic cancer patients: results from the CONKO‐005 trial. EBioMedicine 2021; 66: 103327. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45. Fei N, Wen S, Ramanathan R, et al. SMAD4 loss is associated with response to neoadjuvant chemotherapy plus hydroxychloroquine in patients with pancreatic adenocarcinoma. Clin Transl Sci 2021; 14: 1822–1829. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46. Sinn M, Sinn BV, Treue D, et al. TP53 mutations predict sensitivity to adjuvant gemcitabine in patients with pancreatic ductal adenocarcinoma: next‐generation sequencing results from the CONKO‐001 trial. Clin Cancer Res 2020; 26: 3732–3739. [DOI] [PubMed] [Google Scholar]
- 47. Kwan AK, Piazza GA, Keeton AB, et al. The path to the clinic: a comprehensive review on direct KRAS(G12C) inhibitors. J Exp Clin Cancer Res 2022; 41: 27. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 48. Hong DS, Fakih MG, Strickler JH, et al. KRAS(G12C) inhibition with sotorasib in advanced solid tumors. N Engl J Med 2020; 383: 1207–1217. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 49. Hsiue EH, Wright KM, Douglass J, et al. Targeting a neoantigen derived from a common TP53 mutation. Science 2021; 371: eabc8697. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 50. Sicklick JK, Kato S, Okamura R, et al. Molecular profiling of cancer patients enables personalized combination therapy: the I‐PREDICT study. Nat Med 2019; 25: 744–750. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 51. Connor AA, Denroche RE, Jang GH, et al. Integration of genomic and transcriptional features in pancreatic cancer reveals increased cell cycle progression in metastases. Cancer Cell 2019; 35: 267–282.e267. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 52. Seino T, Kawasaki S, Shimokawa M, et al. Human pancreatic tumor organoids reveal loss of stem cell niche factor dependence during disease progression. Cell Stem Cell 2018; 22: 454–467.e456. [DOI] [PubMed] [Google Scholar]
- 53. Huang W, Navarro‐Serer B, Jeong YJ, et al. Pattern of invasion in human pancreatic cancer organoids is associated with loss of SMAD4 and clinical outcome. Cancer Res 2020; 80: 2804–2817. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 54. Hruban RH, Adsay NV, Esposito I, et al. Pancreatic ductal adenocarcinoma. In: WHO Classification of Tumors of the Digestive System (5th edn). International Agency for Research on Cancer (IARC): Lyon, 2019; 322–323. [Google Scholar]
- 55. Hustinx SR, Leoni LM, Yeo CJ, et al. Concordant loss of MTAP and p16/CDKN2A expression in pancreatic intraepithelial neoplasia: evidence of homozygous deletion in a noninvasive precursor lesion. Mod Pathol 2005; 18: 959–963. [DOI] [PubMed] [Google Scholar]
- 56. Shima K, Nosho K, Baba Y, et al. Prognostic significance of CDKN2A (p16) promoter methylation and loss of expression in 902 colorectal cancers: cohort study and literature review. Int J Cancer 2011; 128: 1080–1094. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 57. Takami H, Yoshida A, Fukushima S, et al. Revisiting TP53 mutations and immunohistochemistry – a comparative study in 157 diffuse gliomas. Brain Pathol 2015; 25: 256–265. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 58. Singer G, Stöhr R, Cope L, et al. Patterns of p53 mutations separate ovarian serous borderline tumors and low‐ and high‐grade carcinomas and provide support for a new model of ovarian carcinogenesis: a mutational analysis with immunohistochemical correlation. Am J Surg Pathol 2005; 29: 218–224. [DOI] [PubMed] [Google Scholar]
- 59. de Andrés MP, Jackson RJ, Felipe I, et al. GATA4 and GATA6 loss‐of‐expression is associated with extinction of the classical programme and poor outcome in pancreatic ductal adenocarcinoma. Gut 2023; 72: 535–548. [DOI] [PubMed] [Google Scholar]
- 60. Dreyer SB, Rae S, Bisset K, et al. The impact of molecular subtyping on pathological staging of pancreatic cancer. Ann Surg 2023; 277: e396–e405. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 61. Dreyer SB, Upstill‐Goddard R, Legrini A, et al. Genomic and molecular analyses identify molecular subtypes of pancreatic cancer recurrence. Gastroenterology 2022; 162: 320–324.e324. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 62. Chan‐Seng‐Yue M, Kim JC, Wilson GW, et al. Transcription phenotypes of pancreatic cancer are driven by genomic events during tumor evolution. Nat Genet 2020; 52: 231–240. [DOI] [PubMed] [Google Scholar]
- 63. Collisson EA, Bailey P, Chang DK, et al. Molecular subtypes of pancreatic cancer. Nat Rev Gastroenterol Hepatol 2019; 16: 207–220. [DOI] [PubMed] [Google Scholar]
- 64. Singh S, Hasselluhn MC, Neesse A. A tangled tale of molecular subtypes in pancreatic cancer. Gut 2019; 68: 953–954. [DOI] [PubMed] [Google Scholar]
- 65. Somerville TDD, Xu Y, Miyabayashi K, et al. TP63‐mediated enhancer reprogramming drives the squamous subtype of pancreatic ductal adenocarcinoma. Cell Rep 2018; 25: 1741–1755.e7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 66. Bailey P, Chang DK, Nones K, et al. Genomic analyses identify molecular subtypes of pancreatic cancer. Nature 2016; 531: 47–52. [DOI] [PubMed] [Google Scholar]
- 67. Cao L, Huang C, Cui Zhou D, et al. Proteogenomic characterization of pancreatic ductal adenocarcinoma. Cell 2021; 184: 5031–5052.e26. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 68. Hayashi A, Fan J, Chen R, et al. A unifying paradigm for transcriptional heterogeneity and squamous features in pancreatic ductal adenocarcinoma. Nat Cancer 2020; 1: 59–74. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 69. Hosein AN, Brekken RA, Maitra A. Pancreatic cancer stroma: an update on therapeutic targeting strategies. Nat Rev Gastroenterol Hepatol 2020; 17: 487–505. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 70. Puleo F, Nicolle R, Blum Y, et al. Stratification of pancreatic ductal adenocarcinomas based on tumor and microenvironment features. Gastroenterology 2018; 155: 1999–2013.e3. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Supplementary methods
Figure S1. Microscopic images of H&E and immunohistochemistry staining in a fine‐needle aspiration specimen from a patient with pancreatic cancer
Figure S2. Flow diagram of the selection of patients with resected pancreatic cancer in a multi‐institutional cohort
Figure S3. Association between the Moffitt tumour subtype and the number of altered genes
Figure S4. Kaplan–Meier survival curves of patients with pancreatic cancer according to altered genes/proteins
Figure S5. Interactions of altered genes in relation to survival patients with pancreatic cancer
Table S1. Clinical and pathological characteristics of pancreatic cancer cases by the number of altered genes
Table S2. Correlations of tumour status based on immunohistochemical expression of CDKN2A (p16), p53, and SMAD4 with mutation calls by targeted sequencing
Table S3. The tumour status of the four drivers and survival among patients with pancreatic cancer (the final multivariable models)
Table S4. Specific combinations of the four major driver alterations and survival among patients with resected pancreatic cancer
Table S5. SMAD4 expression status in addition to specific combinations of the other three major driver alterations and survival among patients with resected pancreatic cancer
Table S6. The tumour status of the four drivers and survival among patients with pancreatic cancer, stratified by administration of adjuvant chemotherapy
Table S7. Recurrence patterns of resected pancreatic cancer by the number of altered genes
Table S8. Recurrence patterns of resected pancreatic cancer harbouring KRAS mutation with or without other driver alterations
Data Availability Statement
The data sets generated and/or analysed during the current study are available from the corresponding authors upon reasonable request. The microarray data were deposited in NCBI's Gene Expression Omnibus database (accession number, GSE157353).