High TPX2 expression correlates with poor outcome in advanced pancreatic cancer patients treated with gemcitabine-based chemotherapy

To examine a potential role of TPX2 in PDAC, we first interrogated two previously published independent expression datasets [13, 14] (n = 291 in total) and found a strong negative impact of high TPX2 expression on patients´ survival (Supplementary Fis. S2A, B). Moreover, PDAC tissue showed significantly decreased TPX2 gene methylation levels compared to normal pancreatic tissue (Supplementary Fig. S2C), as a potential reason for its overexpression. In addition, we did not detect significant TPX2 expression levels in non-neoplastic tissue adjacent to invasive PDAC, like exocrine parenchyma, exocrine pancreas with reactive changes and initial chronic pancreatitis, chronic pancreatitis, fat tissue or neural tissue (Supplementary Fig. S3A–H).

Patient characteristics in the advanced PDAC cohort

To further explore the role of TPX2 in advanced PDAC, we examined the tumour tissue of 139 patients from two independent study cohorts for its expression by immunohistochemistry and assessed its correlation with the patients´ clinicopathological characteristics including outcome. Seventy-one patients were treated with erlotinib-containing palliative chemotherapy within the AIO-PK0104 trial [20]; 68 patients were treated with erlotinib-free chemotherapy within translational biomarker trials [21]. Seventy-nine patients were male, 60 were female. Median patient age was 62.5 years. Median follow-up was 19.4 months (95% CI 12.2–26.6) (95% CI 18.9–45.9) for PFS and 32.4 months for OS. Fifty-one patients had a Karnofsky performance score (KPS) of lower than or equal to 80 (Table 1). Most of the patients (n = 80) had poorly differentiated tumours (grade G3 or G4) which was significantly associated to inferior patient outcome (Supplementary Table S4). Ninety-nine patients received a gemcitabine-based 1st-line palliative chemotherapy, whereas 40 received a non-gemcitabine-based 1st-line regimen (capecitabine for the 30 patients from the AIO-PK0104 study cohort and 5-fluorouracil or other agents for the ten patients from the translational trials cohort), which conferred significantly shorter OS and PFS times (Table 1 and Supplementary Table S4).

Table 1 Comparison of clinicopathological patient characteristics as well as TPX2 expression between the non-gemcitabine-based and the gemcitabine-based palliative treatment subgroups in the advanced PDAC cohorts.TPX2 expression in the advanced PDAC cohort

We detected high TPX2 expression in 13.7% of all tumour samples (Fig. 1), which was significantly associated with shorter patient PFS (5.9 vs 2.0 months, P < 0.001, HR 2.74, 95% CI 1.63–4.61, Supplementary Table S5) and OS (9.3 vs 4.4 months, P < 0.001, HR 2.52, 95% CI 1.52–4.16, Supplementary Table S5). On subgroup analyses, high TPX2 expression retained its strong negative prognostic impact on PFS and OS in the patients treated with 1st-line gemcitabine-based palliative chemotherapy (PFS 8.1 vs 2.0 months, HR 5.25, 95% CI 2.71–10.17, P < 0.001; OS 10.5 vs 3.8 months, HR 4.36, 95% CI 2.36–8.07, P < 0.001; Fig. 2a, b), whereas no significant differences were detected in the patients treated with non-gemcitabine-based 1st-line regimens (PFS 2.5 vs 1.7 months, HR 0.89, 95% CI 0.31–2.61, P = 0.83, OS 6.7 vs 7.3, HR 0.92, 95% CI 0.32–2.63, P = 0.88; Fig. 2c, d). A separate analysis of both cohorts confirmed the findings from the entire cohort: high TPX2 expression correlated significantly with shorter PFS and OS in patients which received 1st-line gemcitabine-based treatment but not in non-gemcitabine 1st-line treated patients in the AIO-PK0104 study and the translational trials cohort (Supplementary Table S5).

Fig. 1: Differential expression of TPX2 in pancreatic cancer.

Immunohistochemical detection of TPX2 expression in exemplary PDAC cases showing (a) high TPX2 expression and (b) low TPX2 expression. 200-fold magnification. Scale bars indicate 50 µm.

Fig. 2: High TPX2 expression is associated with dismal prognosis in 1st-line gemcitabine-treated pancreatic cancer patients.

Univariate analyses (Kaplan–Meier curves and log-rank tests) for PFS and OS in the gemcitabine 1st-line chemotherapy subgroup (a, b) as well as in the non-gemcitabine 1st-line chemotherapy subgroup (c, d), according to TPX2 expression. Crossed lines indicate censored cases.

In Cox multivariate regression analyses adjusting for tumour differentiation grade, KPS and type of 1st-line chemotherapy (where appropriate), we identified high TPX2 expression as independent negative prognostic tissue biomarker for PFS and OS in the 1st-line gemcitabine-treated patient subgroup (P < 0.001 each, Supplementary Table S6) but not in the non-gemcitabine 1st-line treatment subgroup as expected. Using cross-tabulations, we did not detect associations between high TPX2 expression and gender, age group, KPS, tumour grade nor type of 1st-line chemotherapy (Supplementary Table S7).

High TPX2 expression is associated with poor prognosis in resected pancreatic cancer patients treated with adjuvant gemcitabine-based chemotherapy and correlates with gemcitabine resistance

As we detected a significant impact of TPX2 expression on outcome limited to gemcitabine-based treated patients with advanced PDAC, we examined whether similar effects existed in resected PDAC, in which gemcitabine is still a widely employed regimen for adjuvant treatment. Thus, we examined TPX2 expression in the tumour tissue of 400 resected PDAC patients and tested its effect on patient outcome according to the type of adjuvant treatment they received. The study cohort consisted of 204 women and 196 men (median age 68 years, range 22–87 years), of which 222 received gemcitabine-based adjuvant treatment (aG) and 178 received either non-gemcitabine-based or no adjuvant treatment (naG, Table 2). We detected high TPX2 expression at similar rates in both cohorts (aG 14.4%, naG 14.6%, P = 0.96, Table 2), which conferred significantly shorter DFS and OS times in the aGC cohort (DFS 7.6 vs 13.8 months, HR 2.56, 95% CI 1.62–4.04, P < 0.001; OS 16.2 vs 25.8 months, HR 1.56, 95% CI 1.02–2.37, P = 0.04, Fig. 3a, b). However, in the naG cohort, high TPX2 expression did not affect outcome (DFS 6.4 vs 7.4 months, HR 0.83, 95% CI 0.46–1.49, P = 0.53; OS 12.0 vs 13.4 months, HR 0.98, 95% CI 0.62–1.54, P = 0.93, Fig. 3c, d). Cox multivariate regression analyses adjusting for differentiation grade, age, disease stage and R-status confirmed TPX2 expression as an independent prognosticator in the aG cohort (DFS P < 0.001, OS P = 0.01, Supplementary Table S8). As there existed significant differences in some clinicopathological parameters such as age, UICC stage and R-status between both cohorts (Table 2), we employed a propensity-score matching approach to eliminate these imbalances. In the resulting, well-balanced cohort, consisting of 95 patients in each group (Supplementary Table S9), we clearly confirmed the findings from the unmatched cohorts (Supplementary Fig. S10A–D).

Table 2 Comparison of clinicopathological patient characteristics as well as TPX2 expression between the non-gemcitabine-based and the gemcitabine-based adjuvant treatment subgroups in the resected PDAC cohort.Fig. 3: High TPX2 expression is associated with dismal prognosis in resected pancreatic cancer patients treated with adjuvant gemcitabine-based chemotherapy.

Univariate analyses (Kaplan–Meier curves and log-rank tests) for DFS and OS in the gemcitabine-based adjuvant treatment cohort (a, b) and the non-gemcitabine-based adjuvant treatment cohort (c, d) according to TPX2 expression, as well as in the gemcitabine-based treated (e, f) and the non-gemcitabine-based treated (g, h) validation cohorts. Crossed lines indicate censored cases. Association of TPX2 expression and 14 gene-based gemcitabine-resistance expression signature (i). Correlation of TPX2 expression and the expression of gemcitabine-resistance-associated genes (j).

TPX2 expression in the validation dataset

To further verify our data, we employed expression data from TCGA firehose, comprising 149 patients with available data on DFS as a validation cohort. Sixty-eight patients received gemcitabine-based adjuvant therapy (aG); 81 received either none or no gemcitabine-based adjuvant treatment (naG, Table 3). Importantly, high TPX2 mRNA levels within the patients´ tumour tissue conferred significantly decreased DFS and OS times in the aG patients (DFS 9.6 vs 25.1 months, HR 4.89, 95% CI 2.38–10.07, P < 0.001; OS 16.4 vs 66.9 months, HR 4.10, 95% CI 1.69–9.97, P < 0.001, Fig. 3e, f), but not in the naG patients (DFS 7.5 vs 13.0 months, HR 1.64, 95% CI 0.37–3.08, P = 0.12; OS 21.7 vs 30.0 months, HR 1.94, 95% CI 0.93–4.08, P = 0.07, Fig. 3g, h). To examine whether TPX2 expression was directly linked to gemcitabine resistance, we calculated a gemcitabine-resistance score based on a 14 gene expression signature. Tumours showing high TPX2 expression levels displayed high gemcitabine-resistance scores (Fig. 3i), as high TPX2 expression co-segregated with the expression of genes known to be associated with gemcitabine resistance [18] (Fig. 3j). Thus, in the TCGA firehose legacy dataset [22], TPX2 expression positively correlated with the expression of genes associated with gemcitabine resistance, such as ribonucleoside-diphosphate reductase large subunit (RRM1) [23] and polo-like-kinase-1 (PLK1) [24], but also correlated inversely with the expression of dipeptidase 1 (DPEP1), a gene previously shown to increase gemcitabine sensitivity in vitro [25] (Supplementary Fig. S11).

Table 3 Comparison of clinicopathological patient characteristics as well as TPX2 expression between the non-gemcitabine-based and the gemcitabine-based adjuvant treatment subgroups in the resected PDAC validation cohort.