TIMP1 expression underlies sex disparity in liver metastasis and survival in pancreatic cancer

Abstract

Sex disparity in cancer is so far inadequately considered, and components of its basis are rather unknown. We reveal that male versus female pancreatic cancer (PC) patients and mice show shortened survival, more frequent liver metastasis, and elevated hepatic metastasis-promoting gene expression. Tissue inhibitor of metalloproteinases 1 (TIMP1) was the secreted factor with the strongest male-biased expression in patient-derived pancreatic tumors. Male-specific up-regulation of systemic TIMP1 was demonstrated in PC mouse models and patients. Using TIMP1-competent and TIMP1-deficient PC mouse models, we established a causal role of TIMP1 in determining shortened survival and increased liver metastasis in males. Observing TIMP1 expression as a risk parameter in males led to identification of a subpopulation exhibiting increased TIMP1 levels (T1HI males) in both primary tumors and blood. T1HI males showed increased risk for liver metastasis development not only in PC but also in colorectal cancer and melanoma. This study reveals a lifestyle-independent sex disparity in liver metastasis and may open new avenues toward precision medicine.

Graphical abstract

Figure 1.

PC-afflicted males show decreased survival and increased metastasis specifically in the liver. (A) Relative 1-, 3-, and 5-yr survival rates of female (n = 101,357) and male (n = 101,227) PC patients. Data were derived from the ECIS database. A paired samples t test was employed for statistics. (B) Survival of female (n = 22) and male (n = 34) PDAC-bearing KPC mice. The log-rank test was employed for statistics. (C–F) The proportion of female (n = 18) and male (n = 23) patients with metastatic PC (M1, UICC stage IV) with (yes) or without (no) metastases in the liver (C), LN (D), lung (E), or peritoneum (F). The χ² test was employed for statistics. (G–K) Disease recurrence in the pancreas (G), LN (H), lung (I), peritoneum (J), or liver (K) of female (n = 106) or male (n = 112) PDAC patients diagnosed without synchronous metastases (M0) and treated with primary tumor resection. The log-rank test was employed for statistics. *, P ≤ 0.05 (A–K).

Figure S1.

Inclusion and exclusion of PC patients from different cohorts to analyze site-specific metastasis development. (A) Of 493 total participants, 41 patients with metastatic PC (M1) were included for analysis of the sex-dependent metastatic organotropism of PC. (B) Of 453 patients with PC, 218 patients with comparable disease stage and therapy were included for analysis of sex-dependent, site-specific disease recurrence after primary tumor resection. adj., adjuvant; TNM, tumor, node, metastasis. (C) Female (n = 106; left) and male (n = 112; right) PDAC patients from three independent cohorts (Toronto, Munich, and TCGA cohorts) show similar disease recurrence after primary tumor resection. Log-rank statistics were employed for statistics.

Figure S2.

Inclusion and exclusion of PDAC patients to analyze sex-dependent gene expression in liver and pancreas, sex-dependent metastasis-promoting gene expression in the liver of healthy control mice, and TIMP1 expression in exocrine pancreas cells localized adjacent to injured tissue areas. (A) Of 349 transcriptomic samples from previously published RNA-seq data (Moffitt et al., 2015), 23 samples from liver metastases of PDAC patients were included for analysis of sex-dependent gene expression in PC-primed livers. (B) Relative hepatic mRNA expression of metastasis-promoting genes in healthy female (n = 6) and male (n = 4) control mice. Student’s t test was employed for statistics. Mean ± SEM. (C) Of 96 transcriptomic samples from Bailey et al. (2016), 67 samples with known survival data were included for analysis of sex-dependent mRNA expression in PancTs and its correlation with survival. (D) Representative pictures of immunohistochemical (IHC) TIMP1 (red) staining (left) and H&E staining (right) in serial pancreatic tissue sections of a male mouse killed 24 h after the first cerulein injection. TIMP1 expression in exocrine pancreatic cells localized adjacent to damaged cells (arrows) is increased in injured tissue (below dotted line) as compared with noninjured tissue (above dotted line). Scale bars, 100 µm. (E and F) Relative mRNA expression of TIMP1 (E) and relative DNA accessibility of the TSS as well as of a promoter region 1,553 bp upstream of the TSS of the Timp1 gene (F) in exocrine pancreatic cells isolated from control, cerulein-treated (pancreatitis), or PDAC-afflicted (KP^flC) male mice. Mean fold change with SE of mRNA expression as well as DNA accessibility data, including statistical information (adjusted P values), were derived from a previously published RNA-seq and ATAC-seq dataset (Alonso-Curbelo et al., 2021), respectively. *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001 (B, E, and F).

Figure 2.

Metastasis-promoting conditioning of PDAC-primed livers occurs in a male-biased manner. (A) GSEA revealed male-specific up-regulation of fibrosis-, immune-, and lipid metabolism–related processes in the liver of PDAC patients. Genes with significant differential expression in the liver tissue of female (n = 9) and male (n = 14) PDAC patients (see Table S1) were employed as a gene list, whereas male versus female fold change in gene expression was used as the gene rank in GSEA. Biological processes (“nodes”) were significantly up-regulated in male livers (NES ≥2.0, FDR q-value ≤0.01), whereas no biological process was detected as down-regulated in male livers (NES ≤2.0, FDR q-value ≤0.01). The size of connections (“edges”) correlates with overlapping genes between the respective nodes. ECM, extracellular matrix; IGF, insulin-like growth factor; PI3K, phosphoinositide 3-kinase. (B) Volcano plot of transcriptomic data showing sex-dependent expression of metastasis-promoting genes (right) in liver metastases of female (n = 9) versus male (n = 14) PDAC patients. Student’s t test was employed to calculate P values. Genes were considered to be significantly up-regulated (red) or down-regulated (blue) in males after Benjamini-Hochberg adjustment (see also Table S1). αSMA, α-smooth muscle actin. (C) The median fold change in hepatic mRNA expression of metastasis-promoting genes of KPC mice with early-stage PDAC (female: n = 3; male: n = 5) normalized to healthy control mice (female: n = 4; male: n = 4). The Wilcoxon signed-rank test was employed for statistics. (D) Quantification of hepatic macrometastases after challenge of control (female: n = 3; male: n = 4) or KPC mice (female: n = 4; male: n = 4) by i.v. inoculation of tumor cells. Data were derived from two independent experiments. Student’s t test and the Mann-Whitney test for independent variables, respectively, were employed for statistics, depending on normal distribution. Data are presented as mean ± SEM. (E) Quantification of the proportion of macrometastases in relation to total hepatic metastases (macro- plus micrometastases) after challenge of control (female: n = 5; male: n = 6) or cerulein-treated mice (female: n = 6; male: n = 6) by i.v. inoculation of tumor cells. Data were derived from three independent experiments. Student’s t test was employed for statistics. Data are presented as mean ± SEM; *, P ≤ 0.05; ***, P ≤ 0.001 (C–E).

Figure 3.

Male-biased pancreatic TIMP1 expression is associated with a male-specific increase of TIMP1 plasma levels in pancreatic disease. (A) Volcano plot (left) of transcriptomic data showing sex-dependent gene expression in PancT tissue of female (n = 33) and male (n = 34) PDAC patients. The Mann-Whitney test was employed to calculate P values. Genes were considered to be significantly (sign.) up-regulated (red) or down-regulated (blue) in males after Benjamini-Hochberg adjustment (see also Table S2). Median male versus female fold change expression of genes encoding for secreted factors most significantly up-regulated in males (right). (B and C) Relative pancreatic TIMP1 mRNA expression in healthy female (n = 4) or male (n = 3) mice as well as in PDAC-afflicted female (n = 4) and male (n = 4) KPC mice (−: control mice; ++: advanced PC; B) or in mice treated with (+, female: n = 3; male: n = 5) or without (−, female: n = 5; male: n = 4) cerulein (C), respectively. The Mann-Whitney test (B) and Student’s t test (C) were employed for statistics, respectively. Up-regulation of TIMP1 expression in male KPC mice was close to significant (P = 0.057). Median ± interquartile range (B and C). (D and E) Correlation of relative pancreatic TIMP1 mRNA expression with TIMP1 plasma levels in female (n = 8) or male (n = 7) KPC mice (D) or in female (n = 8) or male (n = 9) mice of the cerulein-based mouse model (E). Spearman’s correlation was employed for statistics. (F) Plasma TIMP1 levels of female (n = 5) and male (n = 5) healthy control or female (+: n = 3; ++: n = 5) and male (+: n = 7; ++: n = 5) KPC mice (−: control mice; +: early PancTs; ++: advanced PC). Student’s t test was employed for statistics. Mean ± SEM. (G) Plasma TIMP1 levels of female and male mice treated with (+, female: n = 9; male: n = 10) or without (−, female: n = 8; male: n = 8) cerulein. Student’s t test was employed for statistics. Mean ± SEM. Data were derived from at least three independent cerulein-based mouse experiments (C, E, and G). (H) Plasma TIMP1 levels of female (n = 5) and male (n = 7) healthy control subjects or female (n = 17) and male (n = 14) patients with PDAC (UICC stages I and II). Student’s t test was employed for statistics. Median ± SEM; *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001 (B–H).

Figure 4.

Increased liver metastasis and shortened survival of PDAC-afflicted males is TIMP1 dependent. (A and B) Correlation of plasma TIMP1 with hepatic SAA mRNA expression in female (SAA1: n = 12; SAA2: n = 9) and male (SAA1: n = 15; SAA2: n = 9) KPC mice (A) or in female (n = 6) or male (n = 8) mice of the cerulein-based mouse model (B). Data were derived from three independent experiments (B). Spearman’s correlation was employed for statistics. (C) Correlation of plasma TIMP1 with hepatic SAA mRNA expression in female (n = 7) and male (n = 11) PDAC patients. Spearman’s correlation was employed for statistics. (D) Median fold change in hepatic mRNA expression of pro-metastatic genes in TIMP1-competent (+/+) or TIMP1-deficient (−/−) female (+/+: n = 3; −/−: n = 4) and male (+/+: n = 5; −/−: n = 4) KPC mice with PDAC normalized to female (+/+: n = 4; −/−: n = 3) and male (+/+: n = 4; −/−: n = 3) control mice. For the attribution of the individual genes, see also Fig. 2 C. The Wilcoxon signed-rank test was employed for statistics. (E) SAA1 mRNA expression of untreated (UNT; n = 3) or TIMP1-stimulated (n = 6) HepaRG cells. Data were derived from three independent experiments. A one-sample t test against reference value 1 was employed for statistics. Mean ± SEM. (F) Quantification of hepatic macrometastases (right) after challenge of TIMP1-deficient control mice (n = 3) or TIMP1-deficient (n = 3) or TIMP1-competent (n = 4) KPC mice by i.v. inoculation of tumor cells. Data were derived from two independent experiments. The Mann-Whitney test was employed for statistics. Mean ± SEM. (G) Quantification of the proportion of macrometastases in relation to total hepatic metastases (macro- plus micrometastases) after challenge of TIMP1-deficient female (n = 6) or male (n = 5) control mice or TIMP1-deficient (−/−) or TIMP1-competent (+/+) cerulein-treated female (−/− and +/+: n = 6) and male (−/− and +/+: n = 6) mice by i.v. inoculation of tumor cells. Data were derived from three independent experiments. Student’s t test was employed for statistics. Mean ± SEM. (H) Overall survival of TIMP1-competent PDAC-bearing female (n = 22) and male (n = 34) mice and of TIMP1-deficient female (n = 11) and male (n = 11) PDAC-bearing mice. The log-rank test was employed for statistics. Ρ_s, Spearman’s correlation coefficient; *, P ≤ 0.05; **, P ≤ 0.01 (A–H).

Figure 5.

Identification of a subpopulation of male PDAC patients with increased TIMP1 expression in PancTs accounting for sex differences in PDAC survival. (A) TIMP1 mRNA expression in the pancreatic primary tumors of male PDAC patients separated by two-step cluster analysis into males with low (T1^LO; n = 18) or high (T1^HI; n = 16) TIMP1 expression. Student’s t test was employed for statistics. (B) Probability of overall survival of female (n = 33) or male PDAC patients with low (PancT/T1^LO; n = 18) or high (PancT/T1^HI; n = 16) TIMP1 mRNA expression in pancreatic primary tumors. Log-rank statistics were employed for statistics. The difference in overall survival between PancT/T1^LO and PancT/T1^HI male PDAC patients was close to significant (P = 0.052). (C) Probability of overall survival of male PDAC patients with high TIMP1 mRNA expression in the pancreatic primary tumor (PancT/T1^HI; n = 16) compared with all other patients (n = 51). Log-rank statistics were employed for statistics. HRs with 95% confidence intervals were determined by Cox regression analysis between PancT/T1^HI males and all other PDAC patients. (D) HRs with 95% confidence intervals for overall survival between a subpopulation of male PDAC patients with increased expression of TIMP1, VAMP7, GYG1, FUCA1, or TMEM30A (determined by two-step cluster analysis) and all other PDAC patients, respectively. Cox regression analysis was employed for statistics. *, P ≤ 0.05; ***, P ≤ 0.001 (A–D).

Figure S3.

Liver metastasis development in female PDAC patients separated by TIMP1 as well as inclusion and exclusion of PDAC patients for analysis of liver metastasis development. (A) Probability of overall survival of female PDAC patients with low (PancT/T1^LO; n = 28) or high (PancT/T1^HI; n = 5) TIMP1 mRNA expression in the pancreatic primary tumor. The log-rank test was employed for statistics. (B) Of 183 samples from FireBrowse.org (PAAD), 25 samples from patients with early-stage nonmetastatic PC treated with margin-negative primary tumor resection and subsequent chemotherapy were included for analysis of liver metastasis development after primary tumor resection. TNM, tumor, node, metastasis. (C) Of 493 total participants, 31 patients diagnosed with early-stage nonmetastatic PDAC and known blood TIMP1 levels were analyzed for liver metastasis development after primary tumor resection.

Figure 6.

A subpopulation of male PDAC patients with increased TIMP1 levels show earlier development of metastases specifically in the liver. (A) TIMP1 mRNA expression in pancreatic primary tumors of male PDAC patients separated by two-step cluster analysis into males with low (T1^LO; n = 16) or high (T1^HI; n = 11) TIMP1 expression. Student’s t test was employed for statistics. Medians (continuous lines) and interquartile ranges (dotted lines). (B) Probability of liver metastasis–free survival of male PDAC patients (M0, R0) with high TIMP1 mRNA expression in the pancreatic primary tumor (PancT/T1^HI; n = 5) compared with all other patients (combined female and T1^LO males; n = 20). The log-rank test was employed for statistics. (C) HRs including 95% confidence intervals for liver metastasis–free survival between a subpopulation of male PDAC patients with increased expression of TIMP1, VAMP7, GYG1, FUCA1, or TMEM30A (determined by two-step cluster analysis) and all other PDAC patients, respectively. Cox regression analysis was employed for statistics. (D) Plasma TIMP1 levels of male PDAC patients separated by two-step cluster analysis into males with low (Plasma/T1^LO; n = 9) or high (Plasma/T1^HI; n = 5) plasma TIMP1 levels, respectively. Student’s t test was employed for statistics. Medians (continuous lines) and interquartile ranges (dotted lines). (E) Probability of liver metastasis–free survival of male PDAC patients with high TIMP1 plasma levels (Plasma/T1^HI; n = 5) as compared with all other patients (combined female and Plasma/T1^LO males; n = 25). The log-rank test was employed for statistics. *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001 (A–E).

Figure 7.

Validation of the male-specific clinical value of TIMP1 to predict metastasis development specifically in the liver of primary tumor-resected PDAC patients across multiple patient cohorts. (A) Combination of clinical data from primary tumor resected PDAC patients from three independent studies/cancer centers by (1) quantification of TIMP1 in pancreatic tumors on the level of mRNA (TCGA cohort) or protein (Toronto cohort) or in blood plasma (Munich cohort), (2) clustering of patients based on TIMP1 levels in each cohort via two-step cluster analysis, and (3) grouping of T1^HI or T1^LO patients from each cohort in one combined population. (B and C) Probability of recurrence-free survival in the liver (B) or in entities other than the liver (i.e., pancreas, LN, lung, peritoneum; C) of female (n = 78), T1^LO (n = 50), or T1^HI (n = 36) male PDAC patients. The log-rank test was employed for statistics. HRs with 95% confidence intervals were determined by Cox regression analysis between T1^HI males and all other PDAC patients. (D) Probability of liver metastasis-free survival of male (n = 86), T1^LO (n = 43), or T1^HI (n = 35) female PDAC patients. The log-rank test was employed for statistics. ***, P ≤ 0.001 (A–D).

Figure S4.

Liver metastasis development in all PDAC patients separated by TIMP1, inclusion and exclusion of CRC and SKCM patients for analysis of liver metastasis development, histology of pancreatic lesions in the cerulein mouse model, and histology of micro- and macrometastases after experimental metastasis assays. (A and B) Probability of liver metastasis–free survival of all PDAC patients (males and females combined in A or subsequently subdivided by sex in B) separated by two-step cluster analysis according to high (female: n = 35; male: n = 33) or low (female: n = 43; male: n = 53) TIMP1 levels, respectively. The log-rank test was employed for statistics. HR with 95% confidence interval was determined by Cox regression analysis. (C and D) Inclusion and exclusion of stage-matched CRC (C) and SKCM (D) patients, respectively, for analysis of liver metastasis development. (E) Representative pictures of H&E staining of pancreatic tissue from cerulein-treated (pancreatitis) or vehicle-treated (control) mice killed 24 h after the first injection. Acinus cell injury is highlighted (arrows). Scale bars, 100 µm. (F) Representative pictures of an X-Gal–stained (blue) hepatic micrometastasis (upper panel) and macrometastasis (center panel), as well as a representative picture of H&E staining of a hepatic macrometastasis (lower panel), respectively. Scale bars, 250 µm (upper and center panels), 100 µm (lower panel). (G) List of genes with corresponding primer sequences and probes employed to perform real-time quantitative RT-PCR. (H) Sex determination of transcriptomic data from human liver metastases. Two-step cluster analysis of RNA-seq data from liver metastases (n = 25; Moffitt et al., 2015) based on relative DDX3Y (left) or XIST (center) expression resulted in the separation of two clusters with high (HI) or low (LO) expression, respectively. Combination of both two-cluster analyses revealed three populations (right). Population 1 (n = 14) had high expression of DDX3Y, a Y chromosome–encoded gene, and low expression of XIST, an X chromosome–encoded gene involved in X chromosome inactivation, which we identified as the male population. Population 2 (n = 9) had low expression of DDX3Y and high expression of XIST, which we identified as the female population. Population 3 (n = 2) had low expression of both genes, DDX3Y and XIST, and was excluded from further analyses because the sex of these two individuals could not be determined conclusively. ***, P ≤ 0.001 (A and B).

Figure 8.

Increased TIMP1 expression in tumors of male CRC or SKCM patients predicts early development of liver metastases. (A and D) TIMP1 mRNA expression in primary tumors of male CRC (A) or SKCM (D) patients separated by two-step cluster analysis into males with low (CRC: T1^LO, n = 18; SKCM: T1^LO, n = 53) or high (CRC: T1^HI, n = 11; SKCM: T1^HI, n = 9) TIMP1 expression. Student’s t test was employed for statistics. Medians (continuous lines) and interquartile ranges (dotted lines). (B and E) Probability of liver metastasis–free survival of male CRC (B) or SKCM (E) patients with high TIMP1 expression (CRC: T1^HI, n = 11; SKCM: T1^HI, n = 9) compared with all other patients (CRC: n = 42; SKCM: n = 103), respectively. The log-rank test was employed for statistics. HRs with 95% confidence intervals were determined by Cox regression analysis between T1^HI males and all other patients. (C and F) Probability of liver metastasis–free survival of female CRC (C) or SKCM (F) patients with high TIMP1 expression (CRC: T1^HI, n = 6; SKCM: T1^HI, n = 4) compared with all other patients (CRC: n = 47; SKCM: n = 108). The log-rank test was employed for statistics. **, P ≤ 0.01; ***, P ≤ 0.001 (A–F).