sFRP2 Supersedes VEGF as an Age-related Driver of Angiogenesis in Melanoma, Affecting Response to Anti-VEGF Therapy in Older Patients

The progression of melanoma to distant metastatic sites increases with age. Age is the single most prognostic factor for patient survival (1), and older patients are at increased risk for the development of frequently incurable visceral metastases (2, 3). However, a mechanistic basis for age-related changes in tumor dissemination that underlie such inferior clinical outcomes is not fully understood (4). The formation of visceral metastases requires tumor access to the systemic circulation, and in some cancers, low levels of angiogenesis are highly prognostic of distant metastatic-free survival (5, 6). Angiogenesis is accomplished by the recruitment of endothelial cells to the tumor microenvironment, and this is thought to be initiated by tumor cell secretion of factors such as the VEGF. Other factors including platelet-derived growth factor and basic FGF (bFGF) are also thought to contribute to VEGF-mediated angiogenesis. Melanoma cells have been shown to increase their production of all of these factors (7, 8). VEGF is thought to be a key factor in the angiogenic pathway, binding to its cognate receptors, and driving signaling that results in endothelial cell proliferation and migration. However, in melanoma, angiogenesis is complex, and VEGF itself is not thought to be a predictor of poor prognosis in melanoma, as there is no change in the levels of VEGF secreted by metastatic versus primary melanoma lesions (9, 10). Furthermore, while the analyses of human melanoma lesions have shown a strong correlation between angiogenesis and the ability of a melanoma lesion to transition from a radial to a vertical growth phase (11), other indicators of vascularity are not as clear. For example, microvessel density is not thought to correlate to patient prognosis in melanoma in some studies (12), where others show that it does (13, 14). Finally, and further complicating these issues is the fact that melanoma cells themselves can form vascular networks, in a process known as vascular mimicry (15, 16).

Despite the complicated landscape of angiogenesis in cancer, bevacizumab, an antibody that is targeted against multiple isoforms of VEGF, has been tested in metastatic cancers of multiple origins, and approved for the treatment of some. Bevacizumab, better known under its trade name Avastin, was tested in melanoma in a clinical trial known as the Avastin in Melanoma (AVAST-M) melanoma trial (17, 18). In this multicenter, randomized, phase III trial, 1,343 patients were randomized into those who received Avastin for a year, versus those who were placed under observation. With a minimum 5-year follow-up, there were no significant differences for distant metastasis free survival, or overall survival and only a slight improvement in disease-free intervals. These analyses did not consider age as a variable. Our data show that advanced age of the patient can predict their response to therapy, either in a negative manner, such as an attenuated response to BRAF/MEK inhibitors (19) or in a positive manner, such as a stronger response to immunotherapy (20, 21). In that context, we wanted to know whether the age of the patient could also predict their response to bevacizumab. Further to that, we have previously shown that the aged dermal fibroblast secretome has multiple effects on melanoma progression including increases in genomic instability, resistance to targeted therapy, and metastasis. One of the factors responsible for increased metastatic dissemination is the secreted frizzled-related protein 2 (sFRP2) which is a noncanonical Wnt family member that is elevated in the serum of aged patients with melanoma (22). In cancer models such as breast cancer, sFRP2 stimulates angiogenesis. This function of sFRP2 occurs through a β-catenin–independent pathway, via activated calcineurin/NFAT (23), and has been shown to increase angiogenesis in other cancer models. Targeting sFRP2 with a mAb has been shown to overcome angiogenesis in breast cancer models (23), and our own data have shown that angiogenesis in established melanoma tumors in aged mice can also be targeted by antibodies against sFRP2 (22), but we have not fully explored the impact of age on angiogenesis. In this study, therefore, we examined the hypothesis that the age of the patient with melanoma could predict response to Avastin, and that the observed age-related resistance to Avastin might be due to the increase in sFRP2-driven angiogenesis.

Dermal fibroblast cell lines were obtained from Biobank at Coriell Institute for Medical Research (Camden, NJ). The fibroblasts were cultured in DMEM (Invitrogen) supplemented with 10% FCS and 4 mmol/L l-glutamine. HMVEC-dBlAd cells were obtained from Lonza and cultured in EGM-2MV media (CC-3202, Lonza). Murine Yumm1.7 melanoma cells were cultured in DMEM supplemented with 10% FCS and 4 mmol/L l-glutamine. WM35 melanoma cells were maintained in in MCDB153 (Sigma)/L-15 (Cellgro; 4:1 ratio) supplemented with 2% FBS and 1.6 mmol/L CaCl₂ (tumor growth media). All the cell lines were cultured at 37°C in 5% CO₂. Short tandem repeat profiling was done for melanoma cells and compared against our internal control of over 200 melanoma cell lines as well as control lines such as HeLa and 293T and the results are available upon request. Mycoplasma testing was carried out using a Lonza MycoAlert assay at the University of Pennsylvania Cell Center (Philadelphia, PA) Services.

In a 24-well plate, 5,000 cells were plated in triplicate per day of measurement. Cells were treated with recombinant sFRP2 (200 ng/mL, R&D, catalog No. 6838-FR) without any replacement of addition of fresh medium. Every 2–3 days, cells were counted using hemocytometer and the total cell number in the well was recorded. Graphs were plotted using percent proliferation based on day 0.

HMVEC-dBlAd cells were purchased from Lonza as stated. Cells were thawed and allowed to grow for two passages before the assay was performed as described previously (24). Briefly, 48-well plates were covered with 90 μL Matrigel (concentration >10 mg/mL; BD Biosciences, catalog No. 354234) and allowed to solidify at room temperature for 30 minutes. Human dermal microvascular endothelial cells (HMVECs) were aliquoted at 50,000 cells per well, resuspended in 400 μL of media, and added dropwise to each well. Young and aged media were concentrated at approximately 10× using 3000 Da MWCO filters (UFC900324, Amicon, EMD Millipore) and reconstituted to 1× with DMEM containing 10% FCS. Reagents were purchased from rsFRP2 (200 ng/mL, 6838-FR, R&D Systems), Wnt5a (200 ng/mL, 645-WN, R&D Systems), anti-sFRP2 neutralizing antibody (15 μg/mL, MABC539, EMD Millipore). bFGF at 35 ng/mL is used as positive control. Plates were incubated at 37°C for 12–18 hours. Tubules will start to appear at 10–12 hours and start to contract over time. Images were taken at various intervals to determine optimal tubule formation.

Patient samples for this study were collected under Institutional Review Board (IRB) exemption approval (protocol EX21205258-1), and were derived from patients in the AVAST-M clinical, in which all patients participated with written informed consent, under approval of an IRB, in accordance with the Declaration of Helsinki guidelines. Formalin-fixed, paraffin-embedded sections were deparaffinized using xylene followed by rehydration through series of alcohol washes and finally PBS. Heat-mediated antigen retrieval was performed using citrate based retrieval buffer (Vector Labs, H-3300). Samples were blocked in peroxide blocking buffer (Thermo Fisher Scientific), followed by protein block (Thermo Fisher Scientific) and incubated in appropriate antibody at 4°C overnight in a humidified chamber. Following day, samples were washed and incubated with a biotinylated secondary antibody (Thermo Fisher Scientific) followed by Streptavidin-HRP incubation. Samples were then washed in PBS and incubated in 3-amino-9-ethyl-1-carboazole chromogen and counterstained with Mayer hematoxylin, rinsed in dH₂O and mounted in Aquamount. For mouse samples to be incubated with mouse antibodies, samples were blocked for an additional hour with mouse on mouse block (MKB-2213, Vector Labs). Primary antibodies used were as follows: CD105/Endoglin (1:100, MAB1320, R&D Systems), LYVE1 (1:50, ab14917, Abcam). sFRP2 (1:100, ab137560, Abcam), VEGF (human: 1:100, M727329-8, DAKO; mouse: 1:50, sc7269, Santa Cruz Biotechnology), CD31 (1:100, GA61061-2, DAKO), mCherry (1:500, NBP2-25157, Novus Biologicals), biotinylated Wnt5A (500 ng/mL, BAF645, R&D Systems), NG2 (1:250, ab129051, Abcam), Desmin (1:250, ab15200, Abcam), ICAM1 (1:250, ab179707, Abcam), CD34 (1:250, ab185732, Abcam).

Samples were fixed in 4% paraformaldehyde for 20 minutes followed by 1-hour treatment with blocking buffer (0.2% each of triton-x, BSA, gelatin and casein and 0.02% sodium azide). Cells were incubated in primary antibody and incubated overnight at 4°C. Following day, cells were washed in PBS and incubated with appropriate secondary antibodies (Alexa fluor series, Invitrogen, 1:20,00) at room temperature for 1 hour. Cells were washed in PBS, incubated with DAPI (Invitrogen, 1:10,000) and mounted in Prolong Gold anti-fade reagent (Invitrogen). Images were captured on a Leica TCS SPII scanning laser confocal system. Primary antibodies used were as follows: Ki67 (1:400, Cell Signaling Technology, catalog No. 9027), sFRP2 (1:100, MAB6838, R&D Systems), biotinylated Wnt5A (500 ng/mL, BAF645, R&D Systems).

All animal experiments were approved by the Institutional Animal Care and Use Committee (IACUC; IACUC #112503X_0, IACUC #112510X_0) and were performed in an Association for the Assessment and Accreditation of Laboratory Animal Care accredited facility. Yumm1.7 (1 × 10⁵ cells)-overexpressing mCherry or mCherry-Wnt5a were injected subcutaneously into aged (52 weeks) and young (8 weeks old) C57/BL6 mice (Charles River, catalog No. 556). Mice were treated as follows with rsFRP2 (200 ng/mL, tail vein, twice weekly), anti-sFRP2 antibody (1 mg/kg, tail vein, once weekly), anti-VEGF (1 mg/kg, tail vein, once weekly) or PBS, IgG2aκ, and goat IgG as control. Tumor sizes were measured every 3–4 days using digital calipers, and tumor volumes were calculated using the following formula: volume = 0.5 × (length × width²). Time-to-event (survival) was determined by a fivefold increase in baseline volume (∼1,000 mm³) and was limited by the development of skin necrosis. Mice were euthanized and primary tumors and lungs harvested. Half of the tissue was embedded in paraffin and half in optimal cutting temperature compound (O.C.T, Sakura) and flash frozen for sectioning. Lungs were sectioned and stained with mCherry (NBP2-25157, Novus Biologicals) to determine melanoma metastasis. All reagents injected in live mice were tested for endotoxin levels at University of Pennsylvania Cell Center (Philadelphia, PA) Services using The Associates of Cape Cod LAL test.

The RNAseq and Clinical dataset for skin cutaneous melanoma (25) was downloaded from The Cancer Genome Atlas (TCGA; http://cancergenome.nih.gov/). Normalized mRNA expression was analyzed by quartiles. Patient ages were grouped into categories (<45, 45–65, and >65 years). High versus low categories (sFRP2 and VEGFa) were stratified on the basis of the 25 percentile of high- or low-expressing tumor samples.

Wnt5a overexpression plasmid was prepared by cloning Wnt5a from pLNCX2 plasmid published previously (26) into lentiviral pLU-MCS-IRES-mCherry plasmid. Sequencing-based verification of all plasmids was performed at the Genomics facility at The Wistar Institute (Philadelphia, PA). Lentiviral production was performed according to the protocol suggested by the Broad Institute. Briefly, 293T cells are plated at 70% confluency and cotransfected with short hairpin RNA plasmid and the lentiviral packaging plasmids (pCMV-dR8.74psPAX2, pMD2.G). pLKO.1 empty vector was used as a control. For transduction, cells were treated with lentivirus overnight and allowed to recover for 24 hours before selection using puromycin (1 μg/mL).

For in vitro studies, a Student t test or Wilcoxon rank-sum test (Mann–Whitney) was performed for two-group comparison. Estimate of variance was performed, and parameters for the t test were adjusted accordingly using Welch correction. ANOVA or Kruskal–Wallis test with post hoc Bonferroni’s or Holm–Sidak adjusted P values was used for multiple comparisons. For in vivo studies, repeated measures ANOVA was calculated between samples. The Holm–Sidak correction was performed. Kaplan-Meier and Cox regression was utilized for univariate and multivariate survival analyses, respectively For other experiments, Graphpad/Prism8 was used for plotting graphs and statistical analysis. Data were represented as ± SEM. Significance was designated as follows: *, P < 0.05; **, P < 0.01; and ***, P < 0.001.

To determine the effects of aging on angiogenesis, we first stained whole tumor samples from young and aged patients with melanoma for CD31, to assess blood vessel density throughout the tumor. We found that the percentage of CD31 coverage within tumors was significantly increased in patients over the age of 65 as compared with patients under the age of 65 (Fig. 1A, additional representative staining; Supplementary Fig. S1A). Next, we examined neoangiogenesis and lymphangiogenesis in a syngeneic mouse model of melanoma. Yumm1.7 cells (27), derived from the Braf^V600E/Cdkn2a^−/−/Pten^−/− mouse model of melanoma (28) were injected into the dermis of young (8 weeks) or aged (52 weeks) C57/BL6 mice and serial sections of primary tumors were stained for CD105 and LYVE1. Previous studies of ours have shown that CD31 is increased within aged melanoma mouse tumors (22); however, while CD31 is a commonly utilized pan-endothelial cell marker, it does not distinguish LYVE1-positive lymphatic endothelium from CD105⁺ blood microvascular endothelium (29). Moreover, CD105 is thought to be a more specific marker than CD31 for malignancy-associated blood microvasculature (30, 31) and has a stronger prognostic significance than traditional vascular markers (32). In a cohort of genetically identical mice inoculated with identical melanoma cells, tumors in the aged microenvironment showed increased CD105⁺ blood microvascular density (Fig. 1B, additional representative staining; Supplementary Fig. S1B) with no age-related differences in LYVE1-positive lymphatic vessel density (Fig. 1C). To determine whether age-related changes in the fibroblast secretome may account for differences in tumor angiogenesis, adult HMVECs were exposed to the conditioned media of dermal fibroblasts from young (<45 years) and aged (>55 years) healthy donors from the Baltimore Longitudinal Study of Aging (33). The secretome of aged fibroblasts significantly increased the proliferative index of HMVECs (Fig. 1D) as well as the formation of capillary structures in a previously validated in vitro tubule formation assay (Fig. 1E and F; ref. 34). In contrast, HMVECs treated with young fibroblast conditioned media or unconditioned media formed less complex capillary structures. Such data support the role of the aged dermal fibroblast secretome as a proangiogenic microenvironment.

To determine whether these changes in angiogenesis were driven by classical angiogenic drivers, specifically VEGF, we queried the importance of VEGF signaling in the aged microenvironment. Intriguingly, increased patient age was associated with decreased expression of VEGF (Fig. 2A) as well as its two primary receptors, VEGFR1 (Fig. 2B) and VEGFR2 (Fig. 2C) in an age-stratified analysis of patient with melanoma gene expression in TCGA. This decrease in VEGF with age led us to hypothesize that age-related differences in response to bevacizumab (Avastin), an anti-VEGF antibody, might exist. We tested this hypothesis by analyzing data from the AVAST-M randomized controlled phase III trial evaluating the role of adjuvant bevacizumab in patients with resected melanoma at high risk of recurrence (17). Higher VEGF scores in this group also correlated with improved survival, potentially due to the availability of target for the drug (Fig. 2D). Next, we performed a post hoc trial analysis for the interaction between age and response to treatment. Younger patients who received adjuvant bevacizumab had a significantly longer disease-free interval [age <45 years: HR 0.71, 95% confidence interval (CI), 0.53–0.96], while there was no impact of therapy on disease recurrence in older patients (age >65 years: HR 1.00, 95% CI, 0.72–1.38; Fig. 2E). Similar patterns of treatment effect trended strongly for the outcome of overall survival as well (Supplementary Fig. S2A). Finally, given that BRAF status has been associated with response to Avastin (18), we examined the data with an emphasis on BRAF status, and age. We found that the effect of Avastin is significant for young patients harboring the BRAF mutation, although the young patients who were BRAF wild-type (WT) showed a similar nonsignificant trend. The fact that the data did not reach significance could also be due to the fact that far fewer young patients are BRAF WT. Older patients showed no benefit from bevacizumab regardless of their BRAF status (Supplementary Fig. S2B). We also queried the TCGA database for the expression of VEGF and its receptors according to BRAF status and age. We found significant correlations between age and VEGF and most of its receptors, but for the most part, these did not separate out by BRAF status (Supplementary Fig. S2C), supporting the observations in Supplementary Fig. S2B. We found that the effect of Avastin is significant for young patients who are BRAF mutant, although the young patients who are BRAF WT respond in a similar, albeit nonsignificant manner. Given that there was no therapeutic benefit for BRAF mutant patients of older age ranges, we hypothesize that age and not BRAF status is the predominant determinant of response to Avastin. Overall, however, it is worth noting that BRAF WT tumors show significant changes in VEGF and its receptors and BRAF WT expression is associated with a poor HR in patients on Avastin treatment.

Because increased age-related angiogenesis could not be explained by VEGF changes, and older patients did not appear to benefit from adjuvant bevacizumab, we hypothesized that other factors were driving angiogenesis in the elderly. sFRP2 is an established proangiogenic factor (23, 35) and we have previously demonstrated that sFRP2 is an important component of the aged dermal fibroblast secretome, and is associated with increased CD31 staining in tumors in aged mice (22). The treatment of HMVECs with recombinant sFRP2 was sufficient to increase cell proliferation (Fig. 3A) and its addition to young conditioned media increased capillary formation in vitro to levels observed with treatment with aged conditioned media (Fig. 3B and C). HMVEC treatment with aged conditioned media and a neutralizing α-SFRP2 antibody was sufficient to inhibit in vitro angiogenesis (Fig. 3B, bottom).

These data suggested that sFRP2 may take over as the predominant proangiogenic factor during aging. To further support this, analysis of melanoma samples in the TCGA database demonstrates that there is a positive association between CD105 and sFRP2, while VEGF expression inversely correlates with CD105 expression (Fig. 3D and E). Furthermore, we observed that sFRP2 is increased in ulcerated tumors, which have large amounts of blood vessels, whereas VEGF showed no such correlation to ulceration (Supplementary Fig. S3A and S3B). We assessed the impact of sFRP2 on neoangiogenesis in vivo, specifically by assessing CD105 staining. We injected Yumm1.7 cells into the dermis of young (8 weeks) C57/BL6 mice treated with recombinant sFRP2 (rsFRP2; 200 ng/mL, tail vein, twice weekly) or PBS control. In parallel, we also injected Yumm1.7 cells into the dermis of aged (52 weeks) C57/BL6 mice treated with anti-sFRP2 antibody (1 mg/kg, tail vein, once weekly) or with IgG2aκ control. Subsequently, we stained these tumors for intratumoral CD105⁺ microvascular density, which was significantly greater in young mice treated with rsFRP2 (Fig. 3F) with no difference in lymphatic vessel density (Fig. 3G). Neutralizing sFRP2 in aged mice was sufficient to reduce CD105⁺ microvascular density, without any quantifiable changes in lymphatic vessel density. Notably, there was no difference in tumor size with rsFRP2 or α-sFRP2 treatment (Supplementary Fig. S3C and S3D), nor increased human melanoma cell proliferation with rsFRP2 treatment in vitro (Supplementary Fig. S3E), supporting a selective effect on endothelial and not melanoma cell proliferation, and excluding any potential bias between tumor size and vascularity. Finally, we also confirm previous data showing that treating young mice with recombinant sFRP2 drives melanoma cell metastasis from the skin to the lung, and that inhibiting sFRP2 in aged mice inhibits metastatic progression to the lung (Supplementary Fig. S3F and S3G). Together, these data showed that the observed age-related increase in angiogenesis was likely due to sFRP2.

To test the hypothesis that sFRP2 superseded VEGF as the predominant angiogenic factor during aging, we examined serial sections of primary Yumm1.7 allografts from untreated young or aged C57/BL6 mice for their VEGF and sFRP2 status. As observed with the human data, increasing age led to a decrease in VEGF expression, and we also observed a parallel increase in sFRP2 expression (Fig. 4A and B). To test whether increased sFRP2 could impact response to anti-VEGF therapies, we treated young mice with an antibody against VEGF-A [bevacizumab is species specific and cannot neutralize murine VEGF-A (36)] in the presence and absence of sFRP2. Anti-VEGF reduced angiogenesis in mouse tumors in young mice in vivo; however, when administered in the presence of rsFRP2, there was no reduction in angiogenesis, and indeed blood vessel density was increased, just as it is in the presence of sFRP2 alone (Fig. 4C and D, additional representative staining in Supplementary Fig. S4). Tumor size was unaffected by treatment (Fig. 4E), and there was little toxicity with the treatments (Fig. 4F). This confirmed our hypothesis that when sFRP2 levels increase, as they do in aging, anti-VEGF therapy is no longer effective, and provides a basis for the observed lack of benefit from adjuvant bevacizumab in older patients recruited to the AVAST-M trial. Overall, our data suggest that angiogenesis is increased in the aged tumor microenvironment, but that this angiogenesis is an sFRP2 rather than VEGF-driven process. Because VEGF is no longer the predominant angiogenic factor in older patients, targeting this pathway may have less impact in older versus younger patients.

Angiogenesis in melanoma is complex, and we show in this study that age-related changes in the tumor microenvironment further limit the efficacy of current therapies targeting angiogenesis. We have shown that angiogenesis and growth are uncoupled in melanoma, such that tumors in aged mice have more angiogenesis, and metastasize at much great rates, but actually grow far more slowly than tumors in young mice (22). This may be due to the fact that the factors that drive angiogenesis in aging (e.g., sFRP2) also increase other factors such as Wnt5A which we have shown previously is a powerful suppressor of growth and proliferation, and promotes a slow-cycling phenotype in melanoma allowing cells to resist therapy (26). Angiogenesis thus becomes only one of a myriad of changes wrought by the aged microenvironment, and we have shown that these changes encompass everything from biophysical changes (37) to immune microenvironment changes (21) to changes in lipid secretion and uptake (19). In addition to sFRP2, we have found other factors in the secretome of aged fibroblasts that may also have an impact on angiogenesis. These include RARRES1 and CLU. Intriguingly, RARRES1 has been shown to inhibit VEGF (38), further supporting our theory that VEGF levels are decreased during aging. Whether or not this affects sFRP2 levels, or whether sFRP2 can induce RARRES1 remains an open question, and a topic of further study. We also identified CLU as elevated in the aged secretome. This first seemed counterintuitive as CLU or clusterin, has been shown to increase angiogenesis in ovarian cancer; however, the role of CLU is dependent upon VEGF (39). It is possible that the increase in CLU is a futile attempt to compensate for VEGF decreases. Therefore, the significance of clusterin in age and sFRP2-induced angiogenesis is somewhat unclear. The roles of both of RARRES and CLU, and their potential regulation by sFRP2 and/or other age-related factors remain interesting areas to pursue.

Helfrich and colleagues (40) have shown that increased pericyte coverage can confer resistances to anti-VEGF treatment strategies, and this is consistent with our observed increases in CD31 staining in aged human patients. This study used the MT/RET mouse model of melanoma to uncover the changes that occur in the stabilization of blood vessels, and intriguingly, this mouse model is an excellent model in which to study aging-related changes. MT-RET transgenic mice are engineered to express the human ret transgene in melanocytes controlled by the mouse metallothionein-I promoter-enhancer (41). These mice are reported to spontaneously develop melanomas metastasizing to lymph nodes, lungs, brain, kidney, and spleen. This process involves the development of skin melanosis (100% just after birth), benign melanocytic tumors (100% after 6 months), and metastatic melanoma (65%–70% in mice 18 months old). The Helfrich study examined pericyte coverage in lesions with high versus low levels of angiogenesis, but it is unclear whether these studies took into account the age at which the highly angiogenic versus less angiogenic tumors developed, and whether there was an age-related component of angiogenesis in this model. It is important to note that in this study, they, like us, observed that angiogenesis was uncoupled from tumor size. We decided to test the expression of pericyte markers NG2, Desmin, along with the endothelial cell activation markers ICAM1 and CD34 to investigate age-related effects on vessel stabilization and normalization. NG2 and Desmin are substantially increased in the aged as compared with young tumor microenvironment. Furthermore, they are increased by sFRP2 treatment (Supplementary Fig. S5A–S5D). This is interesting given that previous studies have also shown that aged microvascular networks in healthy subjects have increased pericyte coverage compared with younger adults (42). Despite this increase in pericyte coverage, assessment of the activation markers ICAM1 and CD34 do not appear to change significantly in young or aged tumor microenvironments or in sFRP2-treated conditions in young mice (Supplementary Fig. S5E–S5H). These data suggest that while vessel stabilization is increased in the aged tumor microenvironment, activation is not necessarily affected.

While currently available antiangiogenic therapies target the VEGF pathways, age-dependent loss in VEGF, in tandem with increases in alternative angiogenic pathways, such as sFRP2, may limit efficacy of bevacizumab in older patients with melanoma, but could potentially still be an appropriate treatment in young patients. The AVAST-M trial demonstrated that adjuvant bevacizumab improved the disease-free interval compared with observation alone in the overall recruited patient population, although this early benefit did not translate to an overall survival gain. The trial was not designed to adjust for patient age in the randomization process, thus these post hoc data must be interpreted with caution. However, there does appear to be a striking association between patient age and risk of relapse after adjuvant bevacizumab. Additional data also suggest that using bevacizumab in conjunction with immunotherapy (specifically ipilimumab) results in an improved immune microenvironment (43), potentially increasing response to immunotherapy. It is possible that vessel stabilization as observed to a slight extent in the aged microenvironment and upon treatment with sFRP2 (increased NG2 and Desmin) could potentially lead to an increase in immune infiltration. This is in keeping with previous data from our lab and others that show that older patients actually perform better on immunotherapy than young patients (20, 21) and warrants further investigation of the possibility of combination therapy.

It remains to be determined whether sFRP2 can be targeted in human patients with melanoma. Alternatively, drug therapies targeting the downstream mediators of sFRP2 may be additional avenues for exploration. The aged secretome is characterized by disturbances in multiple proteins including increases in Wnt5A, which can also activate NFAT via CAMKII/calcineurin pathways (44, 45), suggesting sFRP2 and Wnt5A may work in concert during aging to increase angiogenesis. In fact, similar increases in angiogenesis were observed following treatment with rWnt5a in vitro (Supplementary Fig. S6A and S6B) and in Wnt5A-overexpressing Yumm1.7 tumors (Supplementary Fig. S6C and S6D). As calcineurin inhibitors (e.g., tacrolimus) have already been approved for human use, this may be a preferred strategy for inhibiting age-related angiogenesis by blocking a convergence of NFAT-driven mechanisms, and indeed have been suggested before (23). In conclusion, the aged tumor microenvironment promotes hematogenous melanoma dissemination, where loss of VEGF is offset with overexpression of proangiogenic sFRP2. Together, these data provide therapeutic rationale to consider patient age in decisions regarding selection of therapy.

R. Colling reports Oxford is part of the PathLAKE digital pathology consortium (PathLAKE is one of the UK Government’s 5 AI Centres of Excellence and has received in kind industry investment from Philips for digital pathology equipment, software and other services). M.R. Middleton reports receiving trial fees for work [institution received trial fees for work performed in the field of angiogenesis and cancer therapy from Molecular Partners, Pfizer, Astrazeneca, Novartis, and Merck (MSD)]. P. Corrie reports nonfinancial support from Roche Pharmaceuticals (Roche supplied bevacizumab free of charge for the purpose of treating patients in the AVAST-M clinical trial) and grants from Cancer Research UK (the AVAST-M trial was funded by a grant from Cancer Research UK, paid to Cambridge University Hospitals NHS Foundation Trust) during the conduct of the study. X. Xu reports other from CureBiotech Inc (cofounder) and other from Exio Bio (cofounder) outside the submitted work. A.T. Weeraratna reports grants from NIH/NCI (R01CA174646), NIH/NCI (R01CA207935), NIH/NCI (P01 CA114046), NIH/NCI (P50 CA174523), NIH/NCI (P30CA006973), NIH/NCI (P30CA010815), and NIH/NCI (CA232256) during the conduct of the study; other from Heathe Light Sciences (scientific advisory board) and Melanoma Research Foundation (scientific advisory board) outside the submitted work. No potential conflicts of interest were disclosed by the other authors.

M.E. Fane: Formal analysis, investigation, methodology, writing-original draft, writing-review and editing. B.L. Ecker: Formal analysis, investigation, methodology, writing-original draft, writing-review and editing. A. Kaur: Investigation, methodology, writing-original draft. G.E. Marino: Investigation. G.M. Alicea: Investigation. S.M. Douglass: Investigation. Y. Chhabra: Methodology, writing-review and editing. M.R. Webster: Investigation. A. Marshall: Data curation, formal analysis, investigation. R. Colling: Data curation, investigation. O. Espinosa: Data curation, investigation. N. Coupe: Data curation, formal analysis, investigation, writing-review and editing. N. Maroo: Formal analysis, investigation. L. Campo: Data curation, investigation. M.R. Middleton: Resources, data curation, formal analysis, investigation, writing-original draft, writing-review and editing. P. Corrie: Resources, writing-review and editing. X. Xu: Formal analysis, investigation, visualization. G.C. Karakousis: Formal analysis, investigation, writing-original draft, writing-review and editing. A.T. Weeraratna: Conceptualization, resources, formal analysis, supervision, funding acquisition, investigation, visualization, writing-original draft, project administration, writing-review and editing.

A.T. Weeraratna, G. Alicea, and G.E. Marino are supported by R01CA174746, and A.T. Weeraratna, M. Fane, and S.M. Douglass are supported by R01CA207935. X. Xu and A.T. Weeraratna are supported by P01 CA114046. X. Xu, G.C. Karakousis, and A.T. Weeraratna are also supported by P50 CA174523. M.R. Webster is supported by R00 CA208012-01. A.T. Weeraratna is also supported by R01CA232256, a Melanoma Research Alliance/L’Oréal Paris-USA Women in Science Team Science Award, the Wistar Science Discovery Fund, the E.V. McCollum Chair, and a Bloomberg Distinguished Professorship. Core facilities used in this grant are supported by P30CA010815 and P30CA006973. The AVAST-M study was funded by grants from Cancer Research UK (reference: C7535/A6408 and C2195/A8466). M.R. Middleton is supported by the NIHR Oxford Biomedical Research Centre. The views expressed are those of the authors and not necessarily those of the NHS, the NIHR, or the Department of Health.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.