Stiff extracellular matrix activates the transcription factor ATF5 to promote the proliferation of cancer cells

Abstract

Cancer tissues are stiffer than normal tissues. Carcinogenesis stiffens the extracellular matrix (ECM) of cancerous tissues, to which cancer cells respond by activating transcription factors, such as YAP/TAZ, Twist1, and β-catenin, which further elevate their malignancy. However, these transcription factors are also expressed in normal tissues. Therefore, inhibiting these factors in order to treat cancer may lead to severe side effects. Here, we show that activating transcription factor 5 (ATF5), highly expressed in tumors, is activated by ECM stiffness and promotes the proliferation of cancer cells, including that of pancreatic cancer cells and lung cancer cells. In addition, ATF5 suppressed the expression of early growth response 1 (EGR1), thereby accelerating cancer cell proliferation. Stiff ECMs trigger the JAK-MYC pathway which activates ATF5. JAK activation was actomyosin independent whereas MYC induction was actomyosin dependent. These results demonstrate the critical role played by ATF5 in the mechanotransduction process seen in cancers.

Keywords: Biophysics; Cancer; Microenvironment.

Graphical abstract

Figure 1

ATF5 is highly expressed in cancer tissues and localizes in the nuclei by stiff matrices (A) ATF5 mRNA levels in pancreatic adenocarcinoma (PAAD), lung adenocarcinoma (LUAD), breast invasive carcinoma (BRCA), bladder urothelial carcinoma (BLCA), and ovarian serous cystadenocarcinoma (OV), with corresponding normal tissues. The GEPIA2 database was used with the following values for differential analysis: Log2FC-cutoff (0.58) and p value (0.05). (B) Immunofluorescent staining of ATF5 and nuclei in KP4, A549, AsPC1, and SUIT2 cells cultured on collagen gel (soft) or collagen-coated glass (stiff) substrates. (C) Quantification of the relative intensity of ATF5 in the nuclei to that in the cytoplasm, quantified from (B); n = 27 cells in 3 experiments. (D) Immunofluorescent staining of ATF5 and nuclei in KP4 cells on genipin-mixed collagen gel or collagen-coated glass (glass) substrates. Genipin increases the stiffness of collagen gels, as depicted by the stiffness of gels with 0, 0.01, 0.1, and 10 mM genipin being approximately 0.0292, 0.267, 1.49, and 12.5 kPa, respectively. (E) Relative intensity of ATF5 in the nuclei to that in the cytoplasm, quantified from (D); n = 45 cells in 3 experiments. Scale bar = 100 μm; mean with S.D. and each data point is shown; †, statistical significance (p < 0.05) determined via Student’s t test; ‡, statistical significance (p < 0.05) determined via Welch’s t-test; ¥, statistical significance (p < 0.05) determined via Wilcoxon rank-sum test. For multiple comparisons, we analyzed significance using the Bonferroni correction.

Figure 2

Stiff matrix triggers the proliferation of cancer cells via ATF5 (A) GSEA of upregulated genes for hallmark gene sets in KP4 or A549 cells on collagen-coated plastic (stiff) substrates compared with the cells on collagen gel (soft) substrates. NES, normalized enrichment score; NOM p-val, nominal p value; FDR q-val, false discovery rate q value. (B) GSEA of G2M checkpoint or E2F targets highlighted in (A). G2M checkpoint and E2F targets are associated with cell proliferation. (C) Phase-contrast images of KP4 or A549 cells on collagen gel (soft) or collagen-coated plastic (stiff) substrates. (D) Relative cell number, analyzed from (C); n = 3 experiments. (E) Western blot of ATF5 and β-actin in KP4 or A549 cells transfected with negative control RNA, siATF5-1, or siATF5-2 on collagen-coated plastic dishes. Relative intensity of ATF5 to β-actin is shown; n = 3 experiments. (F) GSEA of upregulated genes for hallmark gene sets in KP4 or A549 cells transfected with negative control RNA or siATF5-1 on collagen-coated plastic dishes. NES, normalized enrichment score; NOM p-val, nominal p value. FDR q-val, false discovery rate q value. (G) Phase-contrast images of KP4 or A549 cells transfected with negative control RNA, siATF5-1, or siATF5-2 on collagen-coated plastic dishes. (H) Relative cell number, analyzed from (G); n = 3 experiments. Scale bar = 100 μm; mean with S.D. and each data point is shown; ∗, statistical significance determined with 95% confidence interval. For multiple comparisons, we analyzed significance using the Bonferroni correction.

Figure 3

ATF5 activated by stiff matrix suppresses EGR1 expression (A) Venn diagram showing the overlap of candidate genes upregulated on stiff matrix and downregulated by siATF5 (left), or the overlap of candidate genes downregulated on stiff matrix and upregulated by siATF5 (right) in KP4 cells. (B) qPCR of candidate genes in (A) in KP4 cells on collagen gel (soft) or collagen-coated plastic (stiff) substrates. CFAP54 was not detected by qPCR in KP4 cells. β-actin (ACTB) was used as an internal control; n = 1 experiment. We used two primer pairs to cover all transcript variants in PRKCQ (V1, 2, 3, 4, 5, 6, 7) and PIFO (V1, 2). (C) qPCR of ATF5 and EGR1 in KP4 cells on collagen gel (soft) or collagen-coated plastic (stiff) substrates. β-actin (ACTB) was used as an internal control; n = 3 experiments. (D) Western blot of EGR1 and β-actin in KP4 cells on collagen gel (soft) or collagen-coated plastic (stiff) substrates. Relative intensity of EGR1 to β-actin is shown; n = 3 experiments. (E) qPCR of EGR1 in KP4 cells transfected with negative control RNA, siATF5-1, or siATF5-2 on collagen-coated plastic dishes. β-actin (ACTB) was used as an internal control; n = 4 experiments. (F) qPCR of EGR1 in KP4 cells transfected with ZsGreen (control) or EGR1 vector for EGR1 overexpression on collagen-coated plastic dishes. β-actin (ACTB) was used as an internal control; n = 4 experiments. (G) Phase-contrast images of KP4 cells transfected with ZsGreen or EGR1 vector for EGR1 overexpression on collagen-coated plastic dishes. (H) Relative cell number analyzed from (G); n = 3 experiments. Scale bar = 100 μm; mean with S.D. and each data point is shown; ∗, statistical significance determined with 95% confidence interval. For multiple comparisons, we analyzed significance using the Bonferroni correction.

Figure 4

Stiff matrix activates ATF5 via pJAK (A) Immunofluorescent staining of ATF5 and nuclei in KP4 cells treated with DMSO or Y320 on collagen-coated glass plates. Y320 is reported to inhibit JAK phosphorylation. Representative images from drug screening. The value shows the relative ATF5 intensity in nuclei to cytoplasm (DMSO = 1). (B) GSEA of JAK2 DN V1 DN gene set in KP4 or A549 cells on collagen-coated plastic (stiff) substrates compared with the cells on collagen gel (soft) substrates. NES, normalized enrichment score; NOM p-val, nominal p value. FDR q-val, false discovery rate q value. (C) Western blot of pJAK and β-actin in KP4 cells on collagen gel (soft) or collagen-coated plastic (stiff) substrates. Relative intensity of pJAK to β-actin is shown. n = 3 experiments. (D) Immunofluorescent staining of ATF5 and nuclei in KP4 cells treated with DMSO or JAK inhibitor I on collagen-coated glass dishes. (E) Relative intensity of ATF5 in the nuclei to that in the cytoplasm, quantified from (D). n = 27 cells in 3 experiments. (F) qPCR of EGR1 in KP4 cells treated with DMSO (the same data in Figure 6C) or JAK inhibitor I on collagen-coated plastic dishes. β-actin (ACTB) was used as an internal control; n = 3 experiments. Scale bar = 100 μm; mean with S.D. and each data point is shown; ∗, statistical significance determined with 95% confidence interval; ‡, statistical significance (p < 0.05) determined with Welch’s t-test.

Figure 5

Stiff matrix activates ATF5 via MYC (A) GSEA of hallmark MYC targets V2 gene set in KP4 or A549 cells on collagen-coated plastic (stiff) substrates, compared with the cells on collagen gel (soft) substrates. (B) Immunofluorescent staining of ATF5 and nuclei in KP4 cells treated with DMSO or MYC inhibitor MYCi361 on collagen-coated glass plates. Representative images from drug screening. The value shows the relative ATF5 intensity in the nuclei to that in the cytoplasm (DMSO = 1). (C) Western blot of MYC and β-actin in KP4 cells on collagen gel (soft) or collagen-coated plastic (stiff) substrates. Relative intensity of MYC to β-actin is shown; n = 4 experiments. (D) Immunofluorescent staining of ATF5 and nuclei in KP4 cells treated with DMSO or MYCi361 (5 or 10 μM) on collagen-coated glass dishes. (E) Relative intensity of ATF5 in the nuclei to that in the cytoplasm, quantified from (D). n = 27 cells in 3 experiments. (F) qPCR of EGR1 in KP4 cells treated with DMSO or MYCi361 (5 or 10 μM) on collagen-coated plastic dishes. β-actin (ACTB) was used as an internal control; n = 3 experiments. (G) Western blot of MYC and β-actin in KP4 cells treated with DMSO or JAK inhibitor I on collagen-coated plastic dishes. Relative intensity of MYC to β-actin is shown; n = 4 experiments. (H) Immunoprecipitation with control IgG or anti-ATF5 antibody followed by western blotting with anti-MYC or anti-ATF5 antibody in KP4 cells on collagen-coated plastic dishes. Input sample is shown together. Representative data of 3 experiments are shown. Scale bar = 100 μm; mean with S.D. and each data point is shown; ∗, statistical significance determined with 95% confidence interval; ‡, statistical significance (p < 0.05) determined via Welch’s t-test. For multiple comparisons, we analyzed significance using the Bonferroni correction.

Figure 6

ATF5 is activated in an actomyosin-dependent or independent manner (A) Immunofluorescent staining of ATF5, nuclei, and F-actin in KP4 cells treated with DMSO or Latrunculin A (actin polymerization inhibitor) on collagen-coated glass dishes. (B) Relative intensity of ATF5 in the nuclei to that in the cytoplasm, quantified from (A). n = 45 cells in 3 experiments. (C) qPCR of EGR1 in KP4 cells treated with DMSO (the same data in Figure 4F), Latrunculin A, or Blebbistatin (myosin II inhibitor) on collagen-coated plastic dishes. β-actin (ACTB) was used as an internal control; n = 3 experiments. (D) Western blot of pJAK and β-actin in KP4 cells treated with DMSO, Latrunculin A, or Blebbistatin on collagen-coated plastic dishes. Relative intensity of pJAK to β-actin is shown; n = 3 experiments. (E) Western blot of MYC and β-actin in KP4 cells treated with DMSO, Latrunculin A, or Blebbistatin on collagen-coated plastic dishes. Relative intensity of MYC to β-actin is shown; n = 3 experiments. (F) Immunofluorescent staining of ATF5, nuclei, and F-actin in KP4 cells treated with control IgG or AIIB2 (integrin β1 blocking antibody) on collagen-coated glass dishes. (G) Relative intensity of ATF5 in the nuclei to that in the cytoplasm, quantified from (F); n = 45 cells in 3 experiments. (H) qPCR of EGR1 in KP4 cells treated with control IgG or AIIB2 on collagen-coated plastic dishes. β-actin (ACTB) was used as an internal control; n = 3 experiments. (I) Western blot of pJAK and β-actin in KP4 cells treated with control IgG or AIIB2 on collagen-coated plastic dishes. Relative intensity of pJAK to β-actin is shown; n = 3 experiments. (J) Western blot of MYC and β-actin in KP4 cells treated with control IgG or AIIB2 on collagen-coated plastic dishes. Relative intensity of MYC to β-actin is shown; n = 3 experiments. Scale bar = 100 μm; mean with S.D. and each data point is shown; ∗, statistical significance determined with 95% confidence interval; ¥, statistical significance (p < 0.05) determined via Wilcoxon rank-sum test. For multiple comparisons, we analyzed significance using the Bonferroni correction.

Figure 7

ATF5 is highly localized in the nuclei of human and mouse pancreatic cancer cells in stiff tumors (A) Representative staining images of hematoxylin and eosin (HE) and immunohistochemistry for ATF5 and EGR1 in adjacent normal, and collagen-poor and collagen-rich tumor regions of human pancreatic ductal adenocarcinoma. The areas of black rectangles are magnified. (B) Proportions of cells negative for ATF5 (−), nuclear positive for ATF5 (Nuc(+)), and nuclear negative and cytoplasm-positive for ATF5 (Other(+)) in human pancreatic cancer tissues (n = 9 patients) were examined via immunohistochemistry, followed by quantification. (C) Proportions of cells negative for EGR1 (−), nuclear-positive for EGR1 (Nuc(+)), and nuclear-negative and cytoplasm-positive for EGR1 (Other(+)) in human pancreatic cancer tissues (n = 9 patients) were examined by immunohistochemistry, followed by quantification. (D) Representative staining images of hematoxylin and eosin (HE) and immunohistochemistry for ATF5 and EGR1 in control pancreatic tumors and those treated with AM80. The areas of black rectangles are magnified. Note that AM80 is reported to reduce tumor stiffness by inducing changes in the phenotype of CAFs. (E) Proportions of cells negative for ATF5 (−), nuclear positive for ATF5 (Nuc(+)), and nuclear negative and cytoplasm-positive for ATF5 (Other(+)) in tumor tissues obtained from control mice (n = 9) and those administered AM80 (n = 10) were examined via immunohistochemistry followed by quantification. (F) Proportions of cells negative for EGR1 (−), nuclear positive for EGR1 (Nuc(+)), and nuclear negative and cytoplasm-positive for EGR1 (Other(+)) on tumor tissues obtained from control mice (n = 9) and those administered AM80 (n = 10) were examined via immunohistochemistry followed by quantification. Scale bar = 100 μm; mean with S.D. and each data point is shown; †, statistical significance (p < 0.05) determined via Student’s t test; ‡, statistical significance (p < 0.05) determined via Welch’s t-test; ¥, statistical significance (p < 0.05) determined via Wilcoxon rank-sum test; §, statistical significance (p < 0.05) determined via paired t-test. For multiple comparisons, we analyzed significance using the Bonferroni correction.