Mutational landscape of the patient cohort

A total of 639 patients were included in this study and KRAS mutation was present in 595 (93.1%) patients, as shown in Additional file 4: Table S3_Patient Cohort and KRAS Status. 473 KRASmut patients and 40 KRASWT patients underwent curative surgery, whose overall clinical information were shown in Additional file 4: Table S4_Clinical Information of the Patient Cohort.

As shown in Additional file 4: Table S5_Core Gene Pathway Alterations in KRASmut and KRASWT PDAC, the 595 KRASmut PDAC included 491 (82.5%) age-related subtype, 92 (15.5%) HRD subtype and 12 (2.0%) MMRD subtype, while the 44 KRASWT PDAC included 26 (59.1%) age-related subtype, 17 (38.6%) HRD subtype and 1 (2.3%) MMRD subtype. Compared with KRASmut PDAC, KRASWT PDAC were more likely to be the HRD subtype (38.6% vs. 15.5%, P < 0.001), showed lower alteration frequencies in the three traditional pathways (TP53, cell cycle and TGFβ), comparable frequencies in Trithorax genes, and higher frequencies in NOTCH pathway, Hedgehog pathway and DNA modification pathway.

Core gene pathway alterations and association with tumor differentiation in KRASmut PDAC

The tumor differentiation status of each sample was judged by two independent pathologists from the Department of Pathology of Shanghai Ruijin Hospital with official pathology reports. The differentiation status of PDAC was according to defined WHO criteria, including the presence of tubular structures or solid growth, the presence of mucin, nuclear polymorphism and number of mitoses. Specifically, tumors with more than 30% of areas showing features of poor differentiation were categorized as poor differentiated PDAC. Examples of the histological features of well, moderate and poor differentiated PDAC are shown in Additional file 5: Fig. S2.

As shown in Table 1, among TP53 and other eleven pathways analyzed, TP53 was the only one related to poor tumor differentiation in KRASmut PDAC. Notably, the rate of TP53 missense mutation in poor differentiated PDAC was 50.7% (vs. 36.1% in moderate and well differentiated PDAC, P = 0.001), while the frequency of TP53 truncating mutation was comparable between the two groups (18.7% vs. 20.4%, P = 0.612). Same results were obtained in patients with age-related PDAC, where the rate of TP53 missense mutation was 50.9% in poor differentiated PDAC (vs. 40.0% in moderate and well differentiated PDAC, P = 0.021), as shown in Additional file 4: Table S6_Core Gene Pathway Alterations and Association with Tumor Differentiation in age-related KRASmut PDAC.

We also conducted multivariate Logistic regression to validate the influence of core gene pathway alterations on tumor differentiation in KRASmut PDAC. As shown in Additional file 4: Table S7_Multivariate Logistic Regression of Core Gene Pathway Alterations for Tumor Differentiation in KRASmut PDAC, TP53 missense mutation was the only alteration related to poor tumor differentiation in multivariate analysis (OR 1.848 [1.290–2.647], P = 0.001).

Table 1 Core Gene Pathway Alterations and Association with Tumor Differentiation in KRASmut PDACCore gene pathway alterations and association with lymph nodes involvement and distal metastasis according to tumor size in KRAS mut PDAC

At a median number of 13 lymph nodes examined per patient (range, 5–55), lymph nodes (LNs) involvement in resectable KRASmut PDAC only negatively correlated with WNT pathway alteration (LNs positive rate: 37.2% in WNTmut cases vs. 54.0% in WNTWT cases, P = 0.036) and Hedgehog pathway alteration (LNs positive rate: 16.7% in Hedgehogmut cases vs. 53.4% in HedgehogWT cases, P = 0.012), as shown in Table 2 and Additional file 4: Table S8_Other Gene Pathway Alterations and Association with Lymph Nodes Involvement. However, after stratification according to tumor size, we found that in the small-sized group (≤ 2 cm), the LNs involvement rate of patients with TP53 missense mutation remained as high as 54.8% compared with TP53WT cases (vs. 23.5%, P = 0.010). In cases with small-sized tumors, the LNs involvement rate of patients with TP53 truncating mutations (27.8% vs. 23.5% in TP53WT cases, P = 0.736), cell cycle pathway (50.0% vs. 34.7% in cell cycleWT cases, P = 0.310) and TGFβ pathway (41.7% vs. 36.1% in TGFβWT cases, P = 0.712) alterations was only slightly higher than relevant wild-type cases, without significant differences.

Table 2 Core Gene Pathway Alterations and Association with Lymph Nodes Involvement

Interestingly, as shown in Table 2, the LNs involvement rate of large-sized (> 3 cm) resectable tumors with TP53 missense mutation was lower than those without TP53 mutation (LNs positive rate: 51.2% in TP53missense cases vs. 62.7% in TP53WT cases, P = 0.169), though without significant difference. It was also found that in resectable cases, the rate of TP53 missense mutation was higher in large-sized (> 3 cm) tumors (48.3% vs. 36.9% in small-sized tumors, P = 0.083), as shown in Additional file 4: Table S9_Frequency of Core Gene Pathway Alterations according to Tumor Size in KRASmut PDAC. We assumed that firstly, those with TP53 missense mutation that didn’t metastasize in the early stage were tumors with specifically lower invasiveness, therefore had lower LNs involvement rate. Secondly, if early metastasis hadn’t occurred, KRASmut PDAC with TP53 missense mutation tended to grow faster. Certainly, these two hypotheses need further validation.

With regard to distal metastasis, the metastatic rate of patients with TP53 mutations (20.4% vs. 17.5%, P = 0.389), cell cycle pathway (20.9% vs. 19.0%, P = 0.642) and Trithorax genes (22.0% vs. 18.6%, P = 0.384) alterations was only slightly higher than relevant wild-type cases, without significant differences, as shown in Table 3 and Additional file 4: Table S10_Other Gene Pathway Alterations and Association with Distal Metastasis. TGFβ pathway alteration was the only potential factor significantly associated with distal metastasis (metastatic rate: 26.8% in TGFβmut cases vs. 17.1% in TGFβWT cases, P = 0.011). The impact of TGFβ pathway alteration on distal metastasis seemed widely reported although its detailed characteristics remained controversial [22, 23]. After stratification according to tumor size, patients with TP53 missense mutation instead of truncating mutation showed significant higher metastatic rate than those without TP53 mutation (metastatic rate: 20.5% in KRASmutTP53missense PDAC vs. 2.9% in KRASmutTP53WT PDAC, P = 0.030), indicating that TP53 missense mutation probably led to early invasiveness and metastasis in KRASmut PDAC patients.

Table 3 Core Gene Pathway Alterations and Association with Distal Metastasis

Similar results were obtained in age-related subtype of KRASmut PDAC. TP53 missense mutation in small-sized age-related cases was significantly associated with higher LNs involvement rate (LNs positive rate 53.6% in TP53missense cases vs. 25.8% in TP53WT cases, P = 0.029) and distal metastasis (metastatic rate: 17.6% in TP53missense cases vs. 0% in TP53WT cases, P = 0.025), as shown in Additional file 4: Table S11_Core Gene Pathway Alterations and Association with Lymph Nodes Involvement in Age-related PDAC, Additional file 4: Table S12_Frequency of Core Gene Pathway Alterations according to Tumor Size in Age-related KRASmut PDAC, and Additional file 4: Table S13_Core Gene Pathway Alterations and Association with Distal Metastasis in Age-related PDAC.

Core gene pathway alterations and association with patients’ survival in resectable KRASmut cases

The follow-up data of 458 patients with KRASmut PDAC and 38 patients with KRASWT PDAC who underwent curative surgery were used for survival analysis. At a median follow-up time of 22.4 months (range, 1.6–52.0 months), reduced DFS (12.8 vs. 16.1 months, HR 1.262 [0.829–1.922], P = 0.278) and OS (22.5 vs. 36.9 months, HR 1.545 [0.956–2.498], P = 0.076) were found in KRASmut cases compared with KRASWT cases, as shown in Fig. 1A and B.

Fig. 1

Kaplan-Meier curves for patients (A, B) with KRAS mutation, (C, D) with KRAS G12D mutation in KRASmut cases and (E, F) with alterations of 2–3 pathways among TP53, cell cycle pathway and TGFβ pathway. DFS and OS were displayed as median [95% CI]. NA indicated not available

As shown in Table 4, the multivariate survival analysis was conducted including 5 mutated pathways, sex, age, vascular invasion, tumor differentiation, tumor size, LNs involvement, resection margin status and adjuvant chemotherapy. It was demonstrated that in patients with KRASmut PDAC, KRASG12D mutation was associated with reduced DFS (11.8 vs. 14.0 months, P = 0.075) and OS (21.1 vs. 23.3 months, P = 0.139) compared with mutation of other KRAS codons without significant difference. TP53 mutation was associated with reduced DFS (10.6 months in TP53mut cases vs. 17.3 months in TP53WT cases, P = 0.005) and OS (21.0 months in TP53mut cases vs. 27.0 months in TP53WT cases, P = 0.047), and alterations of cell cycle pathway was also associated with reduced DFS (10.6 months in cell cyclemut cases vs. 13.9 months in cell cycleWT cases, P = 0.017) and OS (19.2 months in cell cyclemut cases vs. 23.3 months in cell cycleWT cases, P = 0.029) with significant difference. Alterations of TGFβ pathway was associated with reduced DFS and OS without significant difference, and that DFS and OS were comparable between patients with and without Trithorax genes alterations. The Kaplan–Meier curves for patients with KRASG12D mutation and with several pathway mutations were shown in Fig. 1C, D and E F, with HR calculated by Cox regression model after adjusting for comutations and clinical covariates.

Table 4 Core Gene Pathway Alterations and Association with Patients’ Survival in Resectable KRASmut PDACThe impact of TP53 missense mutation on patients’ survival

DFS or OS were comparable in patients with TP53 missense and truncating mutation (Fig. 2A and B). In patients with small-sized KRASmut PDAC, TP53 missense mutation was associated with reduced DFS (13.5 months in TP53missense cases vs. 21.8 months in TP53truncating cases, P = 0.046 by univariate log-rank test, HR 0.895 [0.311–2.576], P = 0.837 by Cox regression analysis) instead of OS (25.7 months in TP53missense cases vs. 28.4 months in TP53missense cases, HR 0.910 [0.300–2.761], P = 0.868), suggesting the potential early micrometastases in these cases (Fig. 2C and D).

Fig. 2

Kaplan–Meier curves for patients with TP53 missense or truncating mutation (A, B) in resectable cases, (C, D) in patients with small-sized tumor, (E, F) in patients who failed to receive chemotherapy and (G, H) in patients who received chemotherapy. DFS and OS were displayed as median [95% CI]. NA indicated not available

However, the impact of TP53 mutation subtypes on patients’ survival remained controversial. Although TP53 missense mutation induced potential early metastases, TP53 truncating mutation was reported to more negatively correlated with patients’ survival than TP53 missense mutation, [24, 25] so we then evaluated whether this was due to tumor’s sensitivity to chemotherapy. Interestingly, TP53 missense mutation was significantly associated with reduced DFS (6.6 months in TP53missense cases vs. 9.2 months in TP53truncating cases, HR 0.368 [0.200–0.677], P = 0.005) and reduced OS (9.6 months in TP53missense cases vs. 18.3 months in TP53truncating cases, HR 0.457 [0.248–0.842], P = 0.012) in patients who failed to receive adjuvant chemotherapy (Fig. 2E, F). In patients who received chemotherapy, while DFS were comparable between the two groups (12.9 months in TP53missense cases vs. 13.2 months in TP53truncating cases, HR 1.231 [0.860–1.761], P = 0.257), TP53 missense mutation was even associated with slightly extended OS compared with tumors with TP53 truncating mutation (24.2 months in TP53missense cases vs. 23.8 months in TP53truncating cases, HR 1.461 [1.005–2.124], P = 0.047), as shown in Fig. 2G, H. The decreased survival differences between the missense and truncating group and the increased differences between the wild-type and truncating group in patients who received adjuvant chemotherapy induced us to hypothesize that tumors with TP53 missense mutation were not associated with insensitivity to chemotherapy, compared with tumors with TP53 truncating mutation.

The impact of the Trithorax gene alterations on patients’ survival according to TP53 mutation status

The Trithorax gene encodes a large family of proteins which serve as active epigenetic regulators counteracting the repressive gene expression programmes guided by the the Polycomb group of proteins [26]. The transcriptional repression of the locus INK4A/ARF downstream of TP53-Rb signaling axis is a potential target of the Polycomb group of proteins, thus alterations of the Trithorax genes might result in the repression of TP53 function, [27] as illustrated in Additional file 5: Fig. S3.

As the Trithorax genes were reported to be altered in less frequent but up to 10% of PDAC and that some of its components like KDM6A and ARID1A might influence the development, differentiation and metastasis of PDAC, [4, 28,29,30] we analyzed the relationship between the Trithorax genes alterations and the survival of patients with different TP53 mutation status. As shown in Fig. 3, although the survival outcomes between patients with and without Trithorax genes alterations were comparable (Fig. 3A and B), Trithorax genes alterations contributed to reduced DFS (12.9 months in Trxmut cases vs. 18.6 months in TrxWT cases, HR 1.580 [0.976–2.559], P = 0.063) and OS (24.0 months in Trxmut cases vs. 31.8 months in TrxWT cases, HR 1.254 [0.757–2.078], P = 0.379) in KRASmut, TP53WT PDAC patients (Fig. 3C and D), higher DFS (16.7 months in Trxmut cases vs. 11.6 months in TrxWT cases, HR 0.512 [0.267–0.982], P = 0.044) and OS (25.1 months in Trxmut cases vs. 19.9 months in TrxWT cases, HR 0.606 [0.309–1.190], P = 0.146) in KRASmut, TP53truncating PDAC patients (Fig. 3E, F), and slightly higher DFS (11.1 months in Trxmut cases vs. 9.5 months in TrxWT cases, HR 0.639 [0.426–0.956], P = 0.030) and OS (22.0 months in Trxmut cases vs. 20.6 months in TrxWT cases, HR 0.608 [0.385–0.960], P = 0.033) in KRASmut, TP53missense PDAC patients (Fig. 3G, H), indicating the different properties between TP53 missense and truncating mutations.

Fig. 3

Kaplan–Meier curves for patients with Trithorax genes alterations (A, B) in resectable cases, (C, D) in TP53WT cases, (E, F) in patients with TP53 truncating mutation and (G, H) in patients with TP53 missense mutation. DFS and OS were displayed as median [95% CI].

Validation by Independent TCGA cohort and functional experiments

We managed to validate our conclusions in two ways. Firstly, we used a cohort from the TCGA database including 126 pancreatic cancer patients to validate the association of TP53 mutation status with tumor differentiation. Secondly, we conducted functional experiments in vitro to validate the invasive properties of TP53 missense mutations.

We collected a cohort from the TCGA database with 126 pancreatic cancer patients who had recorded clinical information and SNV analysis, as recorded in Additional file 3: Raw Data 3_Information of the TCGA cohort. There are 93 KRASmut cases and 33 KRASWT cases. As shown in Additional file 4: Table S14_Core Gene Alterations and Association with Tumor Differentiation in KRASmut PDAC in the TCGA cohort, TP53 missense mutation was presented in 0 (0%) well differentiated tumors, while the rate was 54.0% in moderate differentiated tumors (P = 0.001) and 46.7% in poor differentiated tumors (P = 0.007). On the contrary, TP53 truncating mutation was presented in 72.7% well differentiated tumors vs. 20.0% in moderate differentiated tumors (P = 0.001) and 33.3% in poor differentiated tumors (P = 0.036), indicating TP53 missense mutation instead of truncating mutation was associated with poor tumor differentiation.

Notably, tumors with more than 30% of areas showing features of poor differentiation were categorized as poor differentiated PDAC in our study. As shown in Additional file 4: Table S14_Core Gene Alterations and Association with Tumor Differentiation in KRASmut PDAC in the TCGA cohort, there seemed to be more moderate differentiated tumors in the TCGA cohort, probably because that PDAC was defined as moderate differentiated in our hospital only when there was less than 30% of areas showing features of poor differentiation.

Since the volume of the TCGA cohort was not large enough to validate the invasive properties of TP53 missense mutations, we next used two pancreatic cancer cell lines both with KRAS and TP53 missense mutation to perform functional experiments. One cell line was PANC-1 carrying a heterozygous KRASG12D allele (c.35G > A) and a heterozygous TP53R273H allele (c.818G > A). The other cell line was CFPAC-1 carrying a heterozygous KRASG12V allele (c.35G > T) and a heterozygous TP53C242R allele (c.724T > C). A non-specific si-TP53 was transfected into the two cell lines to knock out the TP53 missense mutants, mimicking TP53 truncating mutations. As shown in Fig. 4, the Transwell migration assays (Fig. 4A) and the wound healing assays (Fig. 4B) showed that TP53 knock-out was associated with lower invasiveness compared with TP53 missense mutations. Using specific primers, qRT-PCR was performed to validate the knock-out of TP53 missense allele by non-specific si-TP53 (Fig. 4C).

Fig. 4

TP53 knock-out was associated with lower invasiveness compared with TP53 missense mutations. A Transwell migration assay of PANC-1 and CFPAC-1 cells after knock-out of the TP53 missense mutation. B Wound healing assays performed using PANC-1 and CFPAC-1 cells after knock-out of the TP53 missense mutation. C Validation of the knock-out of the TP53 missense mutation by qRT-PCR.