sFRP2 in the aged microenvironment drives melanoma metastasis and therapy resistance

Abstract

Cancer is a disease of ageing. Clinically, aged cancer patients tend to have a poorer prognosis than young. This may be due to accumulated cellular damage, decreases in adaptive immunity, and chronic inflammation. However, the effects of the aged microenvironment on tumour progression have been largely unexplored. Since dermal fibroblasts can have profound impacts on melanoma progression, we examined whether age-related changes in dermal fibroblasts could drive melanoma metastasis and response to targeted therapy. Here we find that aged fibroblasts secrete a Wnt antagonist, sFRP2, which activates a multi-step signalling cascade in melanoma cells that results in a decrease in β-catenin and microphthalmia-associated transcription factor (MITF), and ultimately the loss of a key redox effector, APE1. Loss of APE1 attenuates the response of melanoma cells to DNA damage induced by reactive oxygen species, rendering the cells more resistant to targeted therapy (vemurafenib). Age-related increases in sFRP2 also augment both angiogenesis and metastasis of melanoma cells. These data provide an integrated view of how fibroblasts in the aged microenvironment contribute to tumour progression, offering new possibilities for the design of therapy for the elderly.

Extended Data 1. Characterization of young and aged fibroblasts

1 × 10⁶ Yumm1.7 cells were injected into the tail vein of young (8 weeks, n=10 mice) and aged (52 weeks, n=10 mice) mice, and 3 weeks later, lungs were assessed for metastatic colonies. Samples were analyzed by H&E staining. Number of mice with metastatic colonies in the lungs was quantified in graph (a). Proliferation rate of aged and young fibroblasts was measured by simple cell counts over a period of 12 days. Analysis of variance (ANOVA) is insignificant (p=0.234) (b). Young and aged fibroblasts were assessed for basal β-galactosidase activity after 5 passages in culture. Representative images from 2 cell lines are shown for young and aged fibroblasts, magnification 100× (c). Staining of fibroblasts in skin reconstructs with α-SMA-1 to assess persistence of fibroblasts in cell culture. Representative images, magnification 150× (d). WM793 melanoma cells were grown in organotypic 3D skin reconstructs built with 3 different fibroblast cell lines derived from healthy young (25–35) and healthy aged (55–65) individuals. Representative images, magnification 150×. Invasion was quantified using NIS Element software. Analysis of variance (ANOVA) was performed (p=0.007). Holm-Sidak multiple comparisons comparing each young cell line with each aged cell line indicated p<0.05 (e). WM793 melanoma cells exposed to conditioned media from young and aged fibroblasts were assessed for proliferation using simple cell counts. Repeated measures analysis of variance (ANOVA) was calculated between samples (p=0.006). Bonferroni’s multiple comparisons test on day 7 and 9 was performed to obtain adjusted p value [day 7 (p=0.047), day 9 (p=0.0004)] (f). Multiple melanoma cells were allowed to form spheroids followed by treatment with conditioned media from aged or young fibroblasts for 48h. The spheroids were then examined for their ability to invade in a collagen matrix (g). Data are represented as mean ± s.d. for each graph (b,e,f).

Extended Data 2. sFRP2 promotes invasion and angiogenesis

Conditioned media from young fibroblasts treated with either control (PBS), or rsFRP2 (200ng/ml) was used to pre-treat WM793 melanoma cells for 48h. Invasion was assayed using a Boyden Chamber assay over 24–72h. Two-tailed unpaired t test was performed (p=0.033) (a). Conditioned media from aged fibroblasts treated with sFRP2 blocking antibody (15µg/ml) for 72h was used to pre-treat WM793 melanoma cells for 48h. The invasion of melanoma cells was assessed in a Boyden Chamber assay for 24–72h. Two-tailed unpaired t test was performed (p=0.035) (b). Young mice (8 weeks, 10 /group) were injected subcutaneously with Yumm1.7 cells. After palpable tumor appeared, mice were treated with rsFRP2 (200ng/mL) for 30 days and examined for angiogenesis using CD31 staining. Representative images, magnification 400× (c,d). Aged mice (52 weeks, n= 5 /group) were injected subcutaneously with Yumm1.7 cells and treated with either control IgG2aκ or sFRP2 blocking antibody (1mg/kg) for 3 weeks. Tumors were examined for angiogenesis by CD31 staining. Representative images, magnification 400× (e, f). Data are represented as mean ± s.d. for each graph (a,b).

Extended Data 3. Treatment of aged tumor bearing mice with an α-sfrp2 antibody results in a lethal inflammation

Cytokine analysis of lungs in aged tumor-bearing mice (52 weeks, n=5 /group) treated with IgG2aκ or α-sFRP2 antibody (1mg/kg, once weekly for 3 weeks). RT-PCR demonstrates a difference in the lungs of mice treated with IgG2aκ or α-sFRP2 in cytokines CCL5 (a), IL4 (b), IL10 (c), CXCL9 (d), IFNγ (e), and IL2 (f). Early response inflammatory genes TNFα (g) and IL6 (h) were no longer significantly altered. Estimate of variance was performed for all genes. For all cytokines, two-tailed unpaired t test was performed *=p<0.05, **=p<0.02. Data are represented as mean ± s.d. for each graph.

Extended Data 4. β-catenin loss in the aged microenvironment

β-catenin expression in normal human skin from young and aged donors, with a focus on the fibroblast population (zoom) (a). β-catenin nuclear translocation in melanoma cells treated with conditioned media from aged as compared to young fibroblasts as measured by Western analysis (b) and a TOPFLASH assay. Two-tailed unpaired t test was performed to indicate statistical significance between treatment with young and aged media (p=0.023). Data are represented as mean ± s.d. (c). Immunofluorescent analysis of β-catenin in melanoma cells treated with media from young fibroblasts in which the β-catenin is knocked down (d).

Extended Data 5. Increase in oxidative stress in the aged microenvironment

APE1 expression in normal human skin as measured by immunohistochemistry. Slides were scored for intensity of stain (3-highest, 0-lowest, <35yr, n=12; >55yr, n=7). Representative images, magnification 400× (left) and 200× (right). Unpaired t test using rank sum (Mann Whitney) revealed statistical significance (p=0.009) (a). Western analysis of SOD3 (b) and PRDX6 (c) levels in conditioned media from young and aged fibroblasts. Immunofluorescent analysis of 8-oxo-dG in normal young and aged skin stained for oxidative stress marker 8-oxo-dG (red), smooth muscle actin (green) and DAPI (blue) (d). ROS activity in melanoma cells with APE1 knockdown, after exposure to aged media. Analysis of variance was performed for each cell line treatment [FS13 (p=0.0006); FS14 (p=0.004)]. For FS13, Two-tailed unpaired t test indicated significance (p<0.01) each shAPE1 cell line compared to control cells. For FS14, two-tailed unpaired t test indicated significance with p<0.05 each shAPE1 cell line compared to control cells (e). Data are represented as mean ± s.d. for each graph (a,e).

Extended Data 6. Gene expression analysis of melanoma cells exposed to aged fibroblasts reveals increases in DNA damage

γH2AX expression was analyzed in melanoma cells exposed to H₂O₂ using immunofluorescence (a). Microarray analysis of the gene expression profiles of melanoma cells exposed to young/middle (Y/M) and aged fibroblasts identified 63 genes commonly increased in 3 melanoma cell lines cultured with aged vs young fibroblasts (b). 33 genes involved in DNA damage response were significantly altered due to effects of aging microenvironment in three different melanoma cell lines. Color scale indicates expression levels relative to aged group (c).

Extended Data 7. DDR response is increased in melanoma cells exposed to aged fibroblasts

Skin reconstructs made with young or aged fibroblasts were stained with anti 53BP1 (a) or γH2AX (b) and analyzed by immunofluorescence. FS5 melanoma cell line treated with conditioned media from young and aged fibroblasts showing DNA damage as measured by a comet assay. Two-tailed unpaired t test with Welch’s correction was performed between young and aged treatments (p=0.039). Data are represented as mean ± s.e.m. (c). Percent ROS activity remaining after NAC treatment of aged fibroblasts. Spearman’s correlation between dose and percent inhibition is significant [p=0.043, r=−0.700] Data are represented as mean ± s.d. (d). Knockdown of SOD3 in young fibroblasts as analyzed by Western blotting (e). Young fibroblasts (2003,071-032 and AG11732) were treated with rsFRP2 (200ng/ml) for 72h and this conditioned media was used to treat melanoma cells for 48h. Cells were assessed for DNA damage by γH2AX (f). Aged fibroblasts (AG13004 and AG11726) were treated with α-sFRP2 (15µg/ml) for 72h and this conditioned media was used to treat melanoma cells for 48h. Cells were assessed for DNA damage by γH2AX (g).

Extended Data 8. Analysis of sFRP2, β-catenin, MITF, 8-oxo-dG, APE1, and 53BP1 in individual patients

Multiple melanoma samples from aged patients (red bars), and young patients (black bars) were compared. Bars represent average staining intensity (3- highest, 0-lowest) in all patients (n=9/group) for indicated proteins (a). Dot plots of staining intensity (0–3+) in individual patient samples for: sFRP2 (b) β-catenin (c) MITF (d) APE1 (e), 8-oxodG (f) and 53BP1 (g). pvalues for each graph obtained by Mann Whitney tests.

Extended Data 9. β-catenin predicts for sensitivity to vemurafenib

Melanoma spheroids were embedded in collagen and treated with 1µM PLX4720 in the presence of conditioned media from either young or aged fibroblasts. After 48h, spheroids were assessed for cell death by staining with ethidium homodimer (magnification, 40×) (a). In cells intrinsically sensitive to vemurafenib in culture, β-catenin expression is increased (b). Knockdown of β-catenin in Yumm1.7 cells decreases their sensitivity to PLX4720. Spearman’s correlation between dose and percent proliferation is significant in CTRL cells [p<0.0001, r=−1.000] whereas shCTNNB1 cells indicated no significant changes in curve after treatment [p=0.948, r=0.03] (c). Young mice (8 weeks, n=10 /group) were injected with rsFRP2 (200ng/mL, twice weekly) and skin was examined for β-catenin levels by immunohistochemistry. Representative images, magnification 200× (d). Yumm1.7 tumors were injected in aged mice pre-treated with α-sFRP2 antibody (1mg/kg, once weekly). Mice were then administered either control or 417mg/kg PLX4720 laced chow. Analysis of variance is significant between treatments (p<0.0001). Two-tailed unpaired t test using rank sum (Mann Whitney) was performed on tumor volumes on day 25 (1 week after treatment). Results were significant in sFRP2 treatment (p=0.036) and insignificant in IgG2aκ treatment (p=0.057) (e). Patient samples show a continuum of decreased response in relation to age, Spearman’s correlation between percent response and age is significant [r=0.243, p=0.035] (f). Data are represented as mean ± s.d for each graph (c,e).

Figure 1. The aged microenvironment promotes phenotype switching

Growth of 1×10⁶ YUMM1.7_mCherry cells subcutaneously injected in young (8 weeks, n=10) and aged (52 weeks, n=10) C57/BL6 mice. (ANOVA p<0.0001; Multiple comparison’s test using Bonferroni correction, p<0.0001 after day 19)(a). CD31 expression from (a) (magnification 200×). Graph indicates percent of mice with extensive angiogenesis (multiple CD31+ vessels/ field) (b). mCherry staining in mouse lungs (positive cells circled in yellow; magnification 400×) (c). WM35 melanoma cells in skin reconstructs built with young or aged fibroblasts (n=3, magnification 150×). Invasion was quantified using ImageJ (ANOVA p=0.0006) (d). Ki67 staining from (d) (magnification 400×) (e). Proliferation of WM35 melanoma cells in conditioned media from young or aged fibroblasts (n=4 fibroblasts/group) ANOVA, p=0.004; Holm Sidak correction [day 6 (p=0.002), day 8 (p=0.034)] (f). Invasion of melanoma spheroids exposed to aged or young conditioned media (n=2 fibroblasts/group, magnification 40×, two-tailed unpaired t-test, p=0.041) (g). Boyden chamber invasion of FS5 melanoma cells treated with young vs. aged fibroblast media (magnification 150×, two-tailed unpaired t-test, p=0.043) (h). Data represented as mean±s.d. (a,d,g,h).

Figure 2. sFRP2 promotes metastasis during aging

Western analysis of sFRP2 (media) and β-catenin (cell lysates) in fibroblasts (a). Invasion of melanoma spheroids (magnification 40×), treated with rsFRP2 (200ng/ml, 48h, two-tailed unpaired t-test, p=0.046) (b). Melanoma spheroids treated with media from aged fibroblasts pre-treated with either IgG2aκ, or α-sFRP2 monoclonal antibody (15µg/ml, 48h, magnification 40×, two-tailed unpaired t-test, p=0.023) (c). sFRP2 in mouse serum (n=10/group): young, aged, and young + 200ng/mL rsFRP2. ANOVA (p=0.015), two-tailed unpaired t-test [PBS vs rsFRP2 (p=0.045), young vs aged mice (p=0.007)] (d). 1×10⁶ mCherry labeled Yumm 1.7 cells were injected intravenously into PBS or rsFRP2 (200ng/mL) treated young mice (6–8 weeks, n=10/group). Graph indicates percent positive lungs recorded as IVIS fluorescence intensity where highest intensity is 3+(red), i.e., multiple metastatic foci, 0-lowest intensity, i.e., no metastases (green) (e). Yumm 1.7 tumors in young mice (6–8 weeks, n=10/group) treated with either PBS or rsFRP2 (200ng/mL) or in aged mice (52 weeks, n=5/group) treated with either control IgG2aκ, or α-sFRP2 monoclonal antibody (1mg/kg) were assessed for CD31 (magnification 400×) (f). Western analysis of non-phosphorylated (active) β-catenin in melanoma cells treated with rsFRP2 for 48h (g). β-catenin expression in young and aged human skin (<35yr, n=12; >55yr, n=7). Slides were scored for positivity (3-highest, 0-lowest), magnification 400×, unpaired t-test using rank-sum (p=0.019) (h). sFRP2 and β-catenin in Yumm1.7 tumors in young and aged mice (magnification 600×) (i). Aged tumor-bearing mice (>52 weeks, n=5/group) treated with α-sFRP2 monoclonal antibody (1mg/kg). Tumors were stained for β-catenin (magnification 400×) (j). sFRP2 ELISA in shCTNNB1 vs. shCTRL fibroblast media. Two-tailed unpaired t-test [2003-071-032 (p=0.015), TP113 (p=0.008)] (k). Invasion of melanoma spheroids treated with TP113 shCTNNB1 conditioned media for 48h (magnification 40×) (l). Data represented as mean±s.d. (b,c,d,h,k).

Figure 3. Loss of APE1 renders melanoma cells more sensitive to oxidative stress in an aged microenvironment

β-catenin, APE1 and MITF in melanoma cells exposed to aged or young fibroblast media (a). APE1 in Yumm1.7 tumors in aged and young mice, percent positive tumors is quantified (magnification 400×) (b). ROS levels in multiple young and aged fibroblasts (two-tailed unpaired t-test, p=0.022) (c). Percent positive 8-oxo-dG Yumm1.7 tumors implanted in aged and young mice (magnification 400×) (d). Western analysis of WM35 melanoma cells after APE1 knockdown (e). ROS levels in shAPE1 melanoma cells in absence (two-tailed t-test, p=0.012), and presence (two-tailed t-test, p=0.008) of conditioned media from aged fibroblasts (f). 8-oxo-dG levels assessed in tumors from young mice treated with either PBS or rsFRP2 and aged mice treated with either IgG2aκ or α-sFRP2 (magnification 400×). (g). Schematic of sFRP2 effects in melanoma cells exposed to aged or young fibroblasts (h). Data represented as mean±s.d. (c,f).

Figure 4. Melanoma cells in an aged microenvironment exhibit increased DNA damage markers

γH2AX and 53BP1 levels in Yumm1.7 tumors in aged and young mice (magnification 400×) (a), and in melanoma cells treated with conditioned media from young and aged fibroblasts (b). Comet assay analysis in melanoma cells treated with conditioned media from aged and young fibroblasts. Two-tailed unpaired t-test with Welch’s correction (young vs aged, p=0.002). Data represented as mean±s.e.m. (c). γH2AX in melanoma cells treated with conditioned media from aged fibroblasts pretreated with 20mM NAC (IF; 48h) (d), conditioned medium from young fibroblasts with SOD3 knock-down (IF; 72h) (e), or in melanoma cells with APE1 knockdown (Western; 48h) (f). γH2AX in melanoma cells exposed for 48h to conditioned media from young fibroblasts pre-treated with rsFRP2 or aged fibroblasts treated with α-sFRP2 antibody (72h) (g). 53BP1 in tumors from young mice treated with PBS or rsFRP2 and aged mice treated with IgG2aκ or α-sFRP2 (magnification 600×) (h). Serum sFRP2 ELISA in young and aged melanoma patients (N=8 young, N=15 aged; unpaired t-test with Welch’s correction, p=0.008). Data represented as mean±s.d. (i). Representative images from one young and aged patient for indicated proteins in the proposed pathway (magnification 400×) (j).

Figure 5. The aged microenvironment induces therapy resistance

Yumm 1.7 tumors in young (8 weeks, n=10 mice/treatment) or aged (52 weeks, n=10 mice/treatment) mice fed PLX4720 (417 mg/kg) chow. Tumors responded in young (p=6 ×10⁻⁵, two way paired t-test) (a), but not aged mice (p=0.361). Dotted lines indicate untreated controls (b). Melanoma spheroids were treated with 100nM H₂O₂ (48h), embedded in collagen and treated with young media containing 1µM PLX4720 (ANOVA, p<0.0001) (c). WM35 spheroids were embedded in collagen and treated with aged media containing 1µM PLX4720 and/or 20mM NAC. Invasion was measured after 48h (representative images, magnification 40×, one-way ANOVA, p=0.02). 20mM NAC (unpaired t-test, p=0.03) and 1µM PLX4720 (two-tailed unpaired t-test, p=0.006) compared to control (d). Live-dead staining of spheroids from (d). ANOVA, p<0.0001; Holm-Sidak’s multiple comparisons test indicated (p<0.0001) after pre-treatment with 20mM NAC in presence of either control or 1µM PLX4720 (e). β-catenin staining of biopsies (magnification 200×) from patients undergoing vemurafenib treatment (percentage indicates percent RECIST response) (f). Tabulation of patient samples correlating H-score (intensity of stain per field of cells) to RECIST response (two-tailed paired t-test, p=0.035). 30% is considered a responder by RECIST criteria (g). Young mice (n=10/group) with Yumm1.7 tumors were treated with rsFRP2 (200ng/mL) bi-weekly. PLX4720 (417/mg/kg) was administered once the tumor reached 500mm³. rsFRP2-treated (red line) vs control (blue line) young mice (ANOVA, p=0.009; Holm-Sidak corrected multiple comparisons, p<0.05 after day 12, mean±s.e.m.) (h). Yumm1.7 tumors were injected in aged mice (52 weeks, n=5/group) pre-treated with α-sFRP2 antibody (1mg/kg, once weekly). Mice were administered control or 417mg/kg PLX4720-laced chow. Tumor volume at day 25 is shown (unpaired t-test with Welch’s correction, p=0.048). (i). RECIST response in patients under 65 vs. over 65. Two sample Wilcoxon rank-sum (Mann Whitney) test indicated statistical significance (p=0.016) (j). Data represented as mean±s.d. (a,b,c,e,i,j).