Tamoxifen mechanically reprograms the tumor microenvironment via HIF-1A and reduces cancer cell survival

Abstract

The tumor microenvironment is fundamental to cancer progression, and the influence of its mechanical properties is increasingly being appreciated. Tamoxifen has been used for many years to treat estrogen-positive breast cancer. Here we report that tamoxifen regulates the level and activity of collagen cross-linking and degradative enzymes, and hence the organization of the extracellular matrix, via a mechanism involving both the G protein-coupled estrogen receptor (GPER) and hypoxia-inducible factor-1 alpha (HIF-1A). We show that tamoxifen reduces HIF-1A levels by suppressing myosin-dependent contractility and matrix stiffness mechanosensing. Tamoxifen also downregulates hypoxia-regulated genes and increases vascularization in PDAC tissues. Our findings implicate the GPER/HIF-1A axis as a master regulator of peri-tumoral stromal remodeling and the fibrovascular tumor microenvironment and offer a paradigm shift for tamoxifen from a well-established drug in breast cancer hormonal therapy to an alternative candidate for stromal targeting strategies in PDAC and possibly other cancers.

Keywords: GPER; HIF‐1A; tamoxifen; tumor microenvironment.

Figure EV1. Tissue proteomics—the tumors in KPC mice treated with 2 mg tamoxifen show changes in their protein content

1. A

   In total, 110 proteins show statistically significant (P < 0.05) changes, of which 45 are upregulated (by 50%; red) and 30 downregulated (by 50%; blue).

2. B, C

   Enriched (P < 0.05) Gene Ontology Biological Processes (GO‐BP) for proteins upregulated (B) and downregulated (C) by the tamoxifen treatment.

Figure EV2. Toxicity curve and proliferation of PSCs under tamoxifen treatment

1. Killing curve for tamoxifen doses.

2. Quantification of cell counting percent relative to time 0—PSCs proliferation.

Data information: Error bars are SEM. **P < 0.01, ***P < 0.0001, n.s. is not significant, t‐test. N = 3 experimental replicates and more than 15 fields of view analyzed per condition.

Figure EV3. RNA Sequencing of PSCs—Gene ontology analysis of differentially expressed genes—DEG (upregulated and downregulated)

13 downregulated and 15 upregulated GO terms selected based on the number of DEGs included in each GO term are shown. Color and circle size represent false discovery rate (FDR) and fraction of DEGs assigned to each GO term to total number of DEGs, respectively.

Figure 1. Tamoxifen decreases hypoxia and increases vascularization

1. A

   Immunofluorescence images of PDAC tissues from KPC mice treated with vehicle control of tamoxifen, scale bar 100 μm.

2. B–D

   (B, D) Quantification of GLUT1 (hypoxia marker) and CD31 (endothelial cell marker). Control (n = 5), 2 mg (n = 5), and 5 mg (n = 4). In all cases, bars represent mean ± SEM. (C) Relative values of protein levels for Glut1 in PDAC tumors assessed by proteomic analysis (6 mice for control and 2 mg and 3 mice for 5 mg, and samples were analyzed in duplicates).

3. E, F

   Expression levels of DEGs relevant to response to hypoxia (left) and blood vessel morphogenesis (right). The values were normalized by tubulin family genes.

4. G

   Immunofluorescence images of PDAC tissues from KPC mice treated with vehicle control and 5 mg of tamoxifen, scale bar 100 μm.

5. H

   Quantification of HIF‐1A in PDAC tissues. Control (n = 5) and 5 mg (n = 4). In the box‐and‐whisker plot, the central box represents values from the lower to upper quartile. The middle line represents the mean. The vertical line extends from the minimum to the maximum value.

6. I

   qPCR levels of HIF‐1A in PSCs, normalized to RPLP0 and relative to control.

7. J

   Western blot bands for protein expression in PSCs (p‐Tmod is post‐translational modification). The plot shows the quantification of the sum of band intensities corresponding to isoform 1, isoform 2, isoform 3, and post‐transcriptionally modified HIF‐1A (n = 8 control and n = 8 tam).

Data information: All histogram bars represent mean ± SEM; *P < 0.05, **P < 0.01, ***P < 0.001 (t‐test for H and J; ANOVA and Tukey's post hoc test for B, D, I). For (A and G), n ≥ 5 sections per animal. Results collected during 3 or more separate experiments.

Figure 2. Tamoxifen reduces LOX‐L2 levels in PSCs and PDAC tissues

1. qPCR levels of LOX‐L2 in PSCs, normalized to RPLP0 and relative to control.

2. Western blot levels of LOX‐L2 in PSCs (n = 3 experimental replicates).

3. Expression of LOX family genes obtained from RNA‐seq data in control and tamoxifen‐treated PSCs (n = 3 experimental replicates). Expression value was normalized by tubulin family genes. Asterisk means significant differences (P < 0.05). Mann‐Whitney U‐test.

4. qPCR levels of LOX family in PSCs, normalized to RPLP0 and relative to control.

5. Relative values of protein levels for LOX members in PDAC tumors assessed by proteomic analysis (6 mice for control and 2 mg and 3 mice for 5 mg, and samples were analyzed in duplicates).

6. Immunofluorescence images of PDAC tissues from KPC mice treated with vehicle (control), and 2 mg and 5 mg of tamoxifen, scale bar 50 μm.

7. Quantification of LOX‐L2 for images in (F). n = 5 (control), 4 (2 mg), and 3 (5 mg), and n > 10 sections per animal.

8. qPCR levels of LOX‐L2 and HIF‐1A in PSCs, normalized to RPLP0 and relative to 1 kPa.

9. qPCR levels of LOX‐L2 and HIF‐1A in PSCs, normalized to RPLP0 and relative to control.

10. Quantification of average forces applied by PSCs on pillars. BBI = blebbistatin. In the box‐and‐whisker plot, the central box represents values from the lower and upper quartile. The middle line represents the mean. The vertical line extends from the minimum to the maximum. Three experimental repeats.

11. qPCR levels of LOX‐L2 in PSCs, normalized to RPLP0 (60S acidic ribosomal protein P0) and relative to control.

Data information: In all cases, histogram bars represent mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001 (t‐test for B, D, I; ANOVA and Tukey's post hoc test for A, G, H, J, K). For (A, D, H, I, J, and K) three replicates collected in at least three different experiments.

Figure 3. Tamoxifen decreases collagen fiber thickness, length, and alignment

1. A

   Images of Matrigel collagen gels previously remodeled by PSCs, second harmonic generation signal for fibrillar collagen (green) and F‐actin (red), scale bar 100 μm.

2. B

   Fiber thickness color‐code map in a represented through the BoneJ plugin where larger spheres fit along fibers represent greater thickness, scale bar 100 μm.

3. C

   SHG fibrillar collagen images used for calculation of alignment through the FFT algorithm. Insets show FFTs of fibrillar collagen‐I images, representing alignment with respect to the elliptical distribution of the FFT central maxima. Circular behavior (values approaching 1) represents no aligned orientation and lower values represent fiber orientation as alignment is displayed as a power distribution orthogonal to the orientation direction. Scale bar 20 μm.

4. D–F

   Quantification of fiber thickness, length, and alignment for images in (A–C).

5. G

   Quantification of collagen fiber thickness using the BoneJ algorithm for images in (B and H).

6. H

   Representative images of Matrigel collagen gels previously remodeled by PSCs, second harmonic generation signal for fibrillar collagen (green) and F‐actin (red), scale bar 100 μm.

Data information: n ≥ 6 matrices per condition. In the scatter plot in (E), each point represents a section. In the box‐and‐whisker plot in (D), the central box represents values from the lower to upper quartile. The middle line represents the mean. The vertical line extends from the minimum to the maximum value. Histogram bars (F and G) represent mean ± SEM. ***P < 0.001 (t‐test for D, E, F; ANOVA and Tukey's post hoc test for G). Three experimental replicates for all panels.

Figure 4. Tamoxifen decreases synthesis and secretion of collagen and MMP‐2 in PSCs and PDAC tissues

1. A

   qPCR levels of collagen in PSCs, normalized to RPLP0 and relative to control.

2. B

   Western blot analysis of collagen normalized to total protein and relative to control.

3. C

   Quantification of time‐lapse collagen synthesis and deposition by PSCs.

4. D

   Representative immunofluorescent images used for the quantification in (C), scale bar 20 μm, collagen‐I was assessed with a specific primary antibody staining.

5. E

   Relative values of protein levels for collagen in PDAC tumors assessed by proteomic analysis (6 mice for control and 2 mg and 3 mice for 5 mg, and samples were analyzed in duplicates).

6. F

   qPCR levels of MMP‐2 in PSCs, normalized to RPLP0 and relative to control.

7. G

   MMP‐2 activity on control and tamoxifen‐treated PSCs assayed by gelatin zymography; above signal intensity of the representative bands used for the quantification presented in the plot below.

8. H, I

   Immunohistochemistry images and quantification of MMP‐2 levels in PDAC tissues from KPC mice treated with vehicle (control), and 2 and 5 mg of tamoxifen, scale bar 100 μm (n = 5 (control), 4 (2 mg), and 3 (5 mg) and n ≥ 5 sections per animal).

9. J

   Relative values of protein levels for MMP‐2 in PDAC tumors assessed by proteomic analysis (6 mice for control and 2 mg and 3 mice for 5 mg, and samples were analyzed in duplicates).

10. K

    qPCR levels of MMP‐2 in PSCs, normalized to RPLP0 and relative to 1 kPa.

11. L

    qPCR levels of MMP‐2 in PSCs, normalized to RPLP0 and relative to control.

Data information: Histogram bars represent mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001 (t‐test for group of two data sets in B and G and ANOVA and Tukey's post hoc test for the rest). For (A, B, C, F, G, H, K, and L) 3 replicates collected in three different experiments.

Figure 5. Tamoxifen decreases fibronectin levels in PSCs and PDAC tissues

1. A

   qPCR levels of fibronectin (FN), fibronectin extracellular domain A (FN‐EDA), and fibronectin extracellular domain B (FN‐EDB) in PSCs, normalized to RPLP0 and relative to control.

2. B

   Western blot analysis of FN normalized to total protein and relative to control.

3. C

   Quantification of fibronectin intensity density of images presented in (D).

4. D

   FN fluorescence and SHG collagen images for matrigel collagen gels remodeled by PSCs.

5. E

   Immunohistochemistry images of PDAC tissues from KPC mice treated with vehicle control, and 2 mg, and 5 mg of tamoxifen (n = 5 (control), 4 (2 mg), and 3 (5 mg), and n ≥ 5 sections per animal).

6. F

   FN fiber thickness color‐code map in (E) represented through the BoneJ plugin.

7. G

   Relative values of protein levels for FN in PDAC tumors assessed by proteomic analysis.

8. H–J

   Quantification of fibronectin immunohistochemistry staining, fiber thickness, and alignment scored of images presented in (E).

Data information: In all cases, histogram bars represent mean ± SEM; *P < 0.05, **P < 0.01, ***P < 0.001 (t‐test for A, B, C; ANOVA and Tukey's post hoc test for H, I, J). For (A, B, C, H, I, and J) three replicates collected in three different experiments. Scale bars are 100 μm.

Figure 6. Tamoxifen treatment decreases proliferation and increases apoptosis in epithelial cells of PDAC tissues

1. Immunofluorescence images of PDAC tissues from KPC mice treated with vehicle control, and 2 mg of tamoxifen, scale bar is 100 μm. Upper panels: Ki67 staining is used as a surrogate of proliferation. White arrows indicate Ki67‐positive nuclei in epithelial cells. Lower panels: Cc3 staining shows the cells undergoing caspase‐3‐mediated apoptosis. Tamoxifen panels show higher levels of yellow staining, which indicates higher percentage of apoptotic epithelial cells.

2. Quantification of staining in panel (A) (n = 4 animals per condition, and n ≥ 5 sections per animal, two experimental repetitions). Histogram bars represent mean ± SEM; ***P < 0.001, t‐test.

Figure 7. Tamoxifen treatment decreases HIF‐1A levels and proliferation and increases apoptosis in Suit‐2 pancreatic cancer cells

1. A–F

   (A, E, F) Immunofluorescence images of control and tamoxifen‐treated Suit2‐007 cells, scale bars is 20 μm. Panel (A) represents HIF‐1A staining in hypoxia and non‐hypoxia conditions, panels (E and F) show Ki67 and Cc3 staining as markers of proliferation and caspase‐mediated apoptosis, respectively. Panel (E): red—F‐actin, green—Ki67, blue—nuclei. Tamoxifen negatively regulates HIF‐1A in hypoxia and non‐hypoxia conditions. (B, C, D) Quantification of immunofluorescence staining in panels (A, E, F). For quantification, eight fields of view (n > 50 cells) per condition. Histogram bars represent mean ± SEM; **P < 0.01, ***P < 0.001, t‐test. All panels include data collected during 3 independent experiments.

Figure 8. Model illustrating the pleiotropic changes of tamoxifen in PSCs and PDAC stroma

Top panel: In PSCs, tamoxifen reduces actomyosin contractility via GPER. This mechanically downregulates HIF‐1A that is the unifying factor through which tamoxifen reduces LOX‐L2, MMP‐2, and FN. GPER activation reduces the synthesis and secretion of collagen by PSCs. Tamoxifen also impairs LOX‐L2 collagen remodeling. The red color has been used to highlight the two main pillars by which tamoxifen acts on PSCs. Bottom panel: In PDAC tissues, tamoxifen reduces response to hypoxia and increases vascularization, and also reduces the amount and organization of collagen and FN in the ECM.