Differential Expression of VEGFA Isoforms Regulates Metastasis and Response to Anti-VEGFA Therapy in Sarcoma

VEGFA is a potent regulator of neovascularization that contributes to tumor growth and metastasis via its receptors VEGFR1/Flt-1, VEGFR2/KDR, and VEGFR3. The function-blocking antibody for VEGFA, bevacizumab (Avastin), and several small-molecule VEGFR inhibitors are now in clinical use (1). However, VEGFA-targeted therapies have not achieved universal success and one of the remaining challenges is to identify biomarkers enabling prospective selection of the most appropriate patients for treatment (2).

VEGFA exists as multiple isoforms generated through alternative splicing and proteolysis (3, 4). This generates forms of VEGFA that are either freely diffusible/soluble (e.g., VEGF121 and 110) or bound by different degrees to the extracellular matrix (ECM) that limits diffusion (e.g., VEGF165, 189, 206). Recent retrospective analyses of data from several large phase III clinical trials have found a link between high concentrations of soluble VEGFA isoforms in plasma and poor prognosis, but also improved response to bevacizumab (Avastin), where progression-free survival (PFS) was measured (5–9). However, little is known about underlying mechanisms. Preclinical studies have shown that tumor cell expression of VEGF120 (mouse equivalent of human VEGF121) influences vascular development during embryonic development, and also response to therapy in tumors in vivo (10–13). In tumors, VEGF120 expression promotes the formation of leaky vasculature, with low pericyte coverage and high interstitial fluid pressure (13); characteristics that are linked to increased metastasis, the key factor regulating patient survival (14, 15).

In this study, we directly investigated the putative link between expression of soluble VEGFA isoforms, metastasis, and response to VEGFA-targeted therapy in mouse models of fibrosarcoma, utilizing tumor cells engineered to express only single isoforms of VEGFA. For fibrosarcoma cells expressing the more soluble VEGFA isoforms, and VEGF120 in particular, mice with subcutaneous tumors had increased plasma VEGFA and lung metastasis that was sensitive to anti-VEGFA therapy. Our data indicate anti-VEGFA therapy predominantly inhibits metastasis through effects on the primary tumor, as survival within the lung following intravenous injection of VEGF120-expressing tumor cells was insensitive to anti-VEGFA therapy. However, survival of VEGF120-expressing tumor cells in the lung after intravenous injection was strongly dependent on VEGFR1 activity. PlGF2 was identified as a potential VEGFR1 ligand mediating these effects in VEGF120-expressing tumor cells.

Detailed Materials and Methods can be found within the Supplementary Methods section.

All experiments were conducted in accordance with the United Kingdom Home Office Animals (Scientific Procedures) Act 1986, with local ethical approval and recently published guidelines (16). CB17-SCID (CB17/Icr-Prkdc^scid/IcrIcoCrl) mice were obtained from Charles River, and maintained as a colony at the University of Sheffield (Sheffield, United Kingdom). C57Bl6/SCID (B6.CB17-Prkdcscid/SzJ) mice were obtained from Charles River. C57BL6/J VEGFR1-ΔTKD mice (17) were generously provided by Prof. M Shibuya (Tokyo Medical and Dental University, Tokyo, Japan) via Dr. K.H. Plate (Johan Goethe University Medical School, Frankfurt, Germany). These were crossed to generate C57Bl6/SCID mice homozygous for vegfr1 or vegfr1-ΔTKD (VEGFR1-WT and VEGFR1-ΔTKD mice respectively).

The majority of experiments were conducted in C57Bl6/SCID mice. Experiments examining the effect of B20-4.1.1 and cediranib were conducted in CB17-SCID mice. Levels of lung metastasis from subcutaneous tumors or survival of cells on intravenous injection were not found to be significantly different between the two strains of mice (unpublished data). Mouse fibrosarcoma cell lines expressing either VEGF120, 164 or 188 (fs120, fs164, fs188, respectively) or all isoforms (fsWT) have been described previously and were derived from day 13.5 embryos from a mixed Swiss background and retrovirally transformed with hRAS and SV40 (12). Variants expressing luciferase 2 (fs120-LS, fs164-LS, fs188-LS and fsWT-LS) for optical imaging were developed as described in Supplementary Methods. Cell lines were cultured from passage 8 after authentication in house (morphology, VEGFA isoform expression, SV40 and hRAS expression levels) to a maximum of passage 30 (typically 2–3 months). Subcutaneous tumors were established by injection of 10⁶ cells into the rear dorsum of mice. For treatment with control IgG (BE5) or anti-VEGFA (B20-4.1.1; Genentech Inc.; refs. 18, 19), tumors were grown to 200 mm³ and mice treated with 5 mg/kg antibody twice weekly by intraperitoneal injection up to a maximum endpoint tumor size of 1,000 mm³. For an experimental model of metastasis involving intravenous injection of tumor cells, mice were injected with 5 × 10⁴ cells into a tail vein and tumor cell arrest within the lung capillary network or survival or colonization of cells in lung tissue investigated at 5 minutes, 48 hours, or ≥21 days after injection, respectively. In each case, presence of tumor cells in lung tissue was determined by analysis of luciferase activity or immunostaining, as described below.

Mice were injected intraperitoneally with d-luciferin (Perkin Elmer) and bioluminescence of the thoracic region was measured using an IVIS Lumina II (Perkin Elmer) imaging system. Alternatively, mice were killed and individual lobes of excised lungs were imaged. Bioluminescence output measured as photons per second per cm² of tissue that radiate into a solid angle of one steridian (p/s/cm²/sr) or as photon per second (p/s) was calculated using Living Image software (Perkin Elmer). For analysis of luciferase activity in lung lysates, luciferase activity was measured, corrected for protein content, and expressed as p/s/μg protein.

Tumors were formalin-fixed and paraffin-embedded (FFPE) or frozen in optimal cutting temperature compound and sectioned. Staining for CD31 in FFPE sections and manual counts of vascular density were performed as described previously (13). CD34 and α-smooth muscle actin (αSMA) immunofluorescence staining of FFPE sections was used to establish the percent of vessels associated with perivascular cells. αSMA-positive cells were counted as perivascular if they were in direct contact with CD34-positive vascular staining and adopted a flattened and/or elongated morphology. Anti-CD34 rather than anti-CD31 was used in combination with the αSMA in FFPE sections for technical reasons. CD31 and CD34 stainings of the vasculature were found to overlap almost completely in frozen sections and provide virtually identical estimates of vascular density (Supplementary Fig. S1). For immunofluorescent staining of SV40 (fibrosarcoma cells), CD11b, Ki67, phospho S139-γH2AX, CD31, laminin, and collagen-I, frozen sections were cut and stained after methanol/acetone fixation. Detailed methods can be found in the Supplementary section.

Lysates from three-dimensional multicellular fs120 or fs188 cell spheroids were analyzed on mouse angiogenesis protein profiler arrays (R&D Systems). Spot intensity for PlGF2, VEGFA, and VEGFB was quantified using enhanced chemiluminescence (ECL) on a C-digit blot scanner (LI-COR) and Image Studio (LI-COR). PIGF2 levels were further characterized in plasma, tumors, and lung homogenates of tumor-bearing animals using an anti-mouse PlGF2 ELISA (R&D Systems).

Tumors were homogenized in cell lysis buffer and proteins separated using SDS-PAGE before collagen-I and laminin were detected using Western blot analysis and ECL. Laminin and collagen-I were measured using densitometry and normalized with respect to GAPDH. Migration of serum-starved cells was measured across 8-μm pore membranes coated on the top surface with laminin or collagen-I toward serum in the presence of control IgG or B20-4.1.1. Prior to analysis of single-cell migration using live video microscopy, cellular adhesion was titrated against coating concentrations of laminin and collagen to ensure maximal adhesion was reached. Live video microscopy of single-cell migration was performed in the presence or absence of the function blocking anti-mouse and human VEGFA antibody B20-4.1.1 or control IgG BE5 on laminin or collagen-I–coated plastic. Speed and distance migrated was calculated as described in detail in the Supplementary Methods section.

Data were plotted and analyzed using GraphPad Prism 6.0e (GraphPad Prism Software Inc). Pairwise analysis was made using an unpaired Student t test with a Welch correction or a Mann–Whitney test for parametric and nonparametric data, respectively. For parametric analysis of grouped data, ANOVA followed by a Sidak multiple comparison test was used. For nonparametric data, a Kruskal–Wallis test followed by a Dunn multiple comparison test was used. Kaplan–Meier analysis was used to measure significance between times to reach a predetermined humane endpoint as a surrogate for survival. In some experiments, data were combined from more than one cohort of mice to reach statistical power and to ensure similar data were obtained from separate litters. Where data were combined from more than cohort, they were normalized against the mean of a common parameter before statistical analysis; see figure legends for details.

Metastasis was assessed in whole lungs excised from mice bearing subcutaneous tumors after inoculation with fs120-LS, fs164-LS or fs188-LS cells, or their wild-type counterparts expressing all VEGFA isoforms (fsWT-LS), at the endpoint size of approximately 1,000 mm³ (Fig. 1A). Luciferase activity in lung lysates showed that mice bearing fs120-LS tumors had significantly higher levels of metastasis than those bearing fsWT-LS or fs188-LS tumors (Fig. 1A); lung luminescence was 10- to 20-fold higher for fs120-LS than for fsWT-LS or fs188-LS (P < 0.001 and P < 0.05, respectively). Metastasis for fs164-LS was closest to fs120-LS, with lung luminescence significantly higher than for fsWT-LS (Fig. 1A, P < 0.01). Differences in metastasis were not due to differences in time from tumor inoculation to endpoint as this was very similar between the different tumor lines with only fsWT-LS tumors taking a slightly longer time to reach endpoint (Fig. 1B). In all subsequent studies, we focused on comparing the fs120-LS and fs188-LS cells, as they represent the extremes of soluble versus ECM-bound isoforms.

To confirm that tumor cells could colonize the lung and develop macrometastatic deposits, we injected tumor cells intravenously as an experimental model of metastasis. In this case, colonization of the lungs by tumor cells could be detected by bioluminescent imaging of live mice (Fig. 1C), in addition to imaging of excised intact lung lobes (Fig. 1D). Consistent with metastasis from the primary subcutaneous tumor, bioluminescence within the thoracic regions of mice injected intravenously with fs120-LS cells was detected at significantly higher levels than in mice injected with fs188-LS cells by day 21 after injection (P < 0.01, Fig. 1E). Mice injected intravenously with fs120-LS cells reached a predetermined humane endpoint (>10⁷ p/s within the thoracic region), significantly sooner than mice injected with fs188-LS cells (P < 0.05, Fig. 1F). At day 40, the majority of fs120-LS–injected mice had reached their humane endpoint (8/9), whereas only 5 of 10 of fs188-LS–injected mice had reached the endpoint by day 50 (Fig. 1F).

We then investigated whether the anti-mouse/human VEGFA function-blocking recombinant antibody B20-4.1.1, directed to amino acids 8–109 of VEGFA that are common to both VEGF120 and VEGF188 (18, 19), had variable effects on metastasis dependent on differential tumor cell expression of VEGFA isoforms. Mice bearing subcutaneous fs120-LS or fs188-LS tumors were treated with B20-4.1.1 when tumors reached approximately 200 mm³, and lungs were removed when subcutaneous tumors reached the endpoint size of 1,000 mm³. B20-4.1.1 decreased metastasis to the lung from fs120-LS tumors significantly (3-fold, P < 0.05), but had no significant effect on metastasis from fs188-LS tumors (Fig. 2A).

Mice with subcutaneous fs120-LS tumors had higher plasma concentrations of VEGFA than mice with fs188-LS tumors (P < 0.05; Fig. 2B). As circulating VEGFA could act at sites of VEGF receptor expression in the host, we investigated whether there was a role for VEGF receptor 1 (VEGFR1) in the elevated lung metastasis observed from subcutaneous fs120-LS versus fs188-LS tumors, as activity of VEGFR1 in host cells has been shown to be involved in the early stages of lung colonization (20, 21). For this, tumor cells were injected subcutaneously into mice with inactivated VEGFR1, due to a deleted tyrosine kinase domain (VEGFR1-ΔTKD mice), or into VEGFR1-WT control mice (17) and metastasis measured as before. The time taken for subcutaneous fs120-LS or fs188-LS tumors to reach 1,000 mm³ and tumor vascular density were not significantly different between VEGFR1-WT and VEGFR1-ΔTKD mice (Supplementary Fig. S2A–S2C). However, lung metastasis from subcutaneous fs120-LS tumors was significantly decreased by approximately 5-fold (P < 0.05) when cells were injected into VEGFR1-ΔTKD mice compared with VEGFR1-WT mice (Fig. 2C). In contrast, inactivation of VEGFR1 had no significant effect on lung metastasis from subcutaneous fs188-LS tumors, with a tendency towards an increase rather than a decrease (Fig. 2C).

To investigate whether VEGFA (via B20-4.1.1 treatment) and VEGFR1 activity affect survival within the lung after tumor cells have entered the circulation, we injected fs120-LS or fs188-LS tumor cells intravenously via the tail vein of mice and measured bioluminescence of excised lung lobes 48 hours later. As we found for metastasis from subcutaneous tumors (Fig. 2C), intravenous injection of fs120-LS cells into VEGFR1-ΔTKD mice decreased lung bioluminescence by approximately 5-fold compared with VEGFR-WT mice (P < 0.01, Fig. 2D and E), whereas lung bioluminescence following intravenous injection of fs188-LS cells was unaffected by host activity of VEGFR1 (Fig. 2E). Lung bioluminescence at 48 hours following intravenous injection of fs188-LS cells was lower than that for fs120-LS cells (P < 0.01) in the wild-type mice (Fig. 2E), as we had previously found at 21 days (Fig. 1E) and for metastasis from the subcutaneous tumors (Figs. 1A and 2A). Overall, the pattern of lung bioluminescence observed following intravenous injection of fs120-LS and fs188-LS cells into wild-type or VEGFR1-ΔTKD mice was very similar to that found for metastasis from subcutaneous tumors (comparing Fig. 2E and C). Coupled to the lack of an obvious influence of VEGFR1 on the subcutaneous primary tumor itself (Supplementary Fig. S2), these data suggest that fs120-LS survive better than fs188-LS cells in lung tissue within the first 48 hours, at least partially via a VEGFR1-dependent mechanism, which could account for their differing metastatic rates from subcutaneous inoculation of cells.

In contrast to the influence of VEGFR1 activity, B20-4.1.1 treatment had no influence on fs120-LS survival in the lung 48 hours after intravenous injection (Fig. 2F). This suggests that a VEGFR1 ligand, other than VEGFA itself, is involved in promoting survival of tumor cells in the lung. Furthermore, it suggests that the inhibitory effect of B20-4.1.1 on metastasis from subcutaneous tumors (Fig. 2A) is via effects on the primary subcutaneous tumor.

Although our data examining tumor cell survival in the lung 48 hours after intravenous injection suggested that innate characteristics of the fs120-LS cells themselves are important for their increased lung metastasis over fs188-LS, the full metastatic process could also be influenced by additional factors arising from the primary tumor. To address this, mice were first injected subcutaneously with fibrosarcoma cells lacking luciferase expression (fs120 or fs188). When tumors reached approximately 200 mm³, counterpart fs120-LS or fs188-LS luciferase-expressing cells were injected intravenously, followed by measurement of bioluminescence in excised lungs after 48 hours, as before (Fig. 3A). In this case, lung bioluminescence following intravenous injection of tumor cells into wild-type mice was approximately 150-fold higher for fs120-LS cells versus fs188-LS cells (Fig. 3A), compared with an approximate 5-fold differential when there was no subcutaneous tumor present (Fig. 2E). Secreted factors and/or early cellular dissemination from fs120-LS subcutaneous tumors could have a priming effect on lung tissue, resulting in a favorable environment for survival of the intravenously injected tumor cells. By analyzing lung homogenates from a cohort of mice bearing small 200 mm³ subcutaneous tumors, we found that fs120-LS tumor cells had already disseminated to the lungs at this stage (Supplementary Fig. S3A). In contrast, pretreatment of mice with fs120 cell–conditioned medium prior to intravenous injection of fs120-LS cells was unable to increase cell survival in the lung (Supplementary Fig. S3B).

Survival of intravenously injected fs120-LS cells (but not fs188-LS) in the lung, when a primary subcutaneous tumor was present, still showed dependence on VEGFR1 activity (Fig. 3A). Similar to data obtained on intravenous injection of fs120-LS cells alone, B20-4.1.1 treatment did not inhibit survival of intravenous-injected fs120-LS cells in the lung in the presence of a subcutaneous tumor (Fig. 3B). This provides further support for the notion that blocking the activity of VEGFA within the primary tumor is the main mechanism by which B20-4.1.1 reduced lung metastasis from subcutaneous fs120-LS tumors (Fig. 2A).

As B20-4.1.1 had no effect on survival of intravenously injected tumor cells in the lung, we investigated whether VEGFR1 ligands, other than VEGFA itself, were preferentially produced by fs120-LS cells. Using a protein array–based method to screen lysates from multicellular spheroids of tumor cells (22), fs120 cells were found to express significantly higher levels of the VEGFR1 ligand PlGF2 than fs188 cells. VEGFA and VEGFB spot intensity was low and showed no significant differences between the cell lines (Supplementary Fig. S4A and S4B). The increased expression of PlGF2 was confirmed using ELISA of cell conditioned medium, with fs120 cells expressing significantly more PlGF2 than fs188 cells under hypoxic conditions (Fig. 3C). Although PlGF2 was not differentially expressed between fs120-LS and fs188-LS subcutaneous tumors (Supplementary Fig. S4C), mice bearing fs120-LS tumors did have significantly higher levels of PlGF2 in plasma and lungs than those bearing fs188-LS tumors (P < 0.05; Fig. 3D and E). As both VEGFA and PlGF2 concentrations in plasma were increased in fs120-LS tumor–bearing mice, this points to redundancy in activation of VEGFR1 in host tissue by VEGF family ligands, which would explain why B20-4.1.1 did not decrease survival of intravenously injected fs120-LS cells in the lung.

Considering the evidence for B20-4.1.1 acting at the primary tumor site to decrease lung metastasis selectively from fs120-LS tumors, we investigated whether B20-4.1.1 had a VEGFA isoform-selective effect on subcutaneous tumor growth, vascular density or pericyte coverage that could account for this effect. B20-4.1.1 retarded tumor growth to a very similar extent in fs120-LS and fs188-LS tumor types (Fig. 4A–C). We found significant antivascular activity for B20-4.1.1 in both tumor types, as determined by CD31-positive (CD31^+VE) blood vessel counts, but no differential effect between the two tumor types. Specifically, vascular density was significantly reduced by approximately 2-fold in both fs120-LS and fs188-LS tumors (P < 0.001 and P < 0.01 respectively, Fig. 4D and E). There was a tendency for the percent of αSMA pericyte positive (αSMA^+VE) vessels to increase in fs120-LS tumors following B20-4.1.1 treatment but effects were insignificant in both tumor types (Fig. 4F and G). Pericyte-positive vessels were more numerous (P < 0.01) in control fs188-LS tumors than in control fs120-LS tumors (Fig. 4G), confirming previous results for subcutaneous tumors of nonbioluminescent cells (12).

As we did not identify any VEGFA isoform–specific effects of B20-4.1.1 on blood vessels within the primary subcutaneous tumors, we investigated their extracellular matrix (ECM) composition, as this has an influential role in regulating cell migration and metastasis (23). fs120 cells have higher levels of adhesion and spreading on laminin compared with collagen-I in vitro and they adopt a mesenchymal morphology on laminin versus collagen-I, suggesting increased migration on this matrix (24). Tumors derived from fs188 cells also have higher numbers of αSMA-positive fibroblasts than those from fs120 cells (12), suggestive of differences in ECM composition between the two tumor types in vivo (25). Analysis of protein extracted from subcutaneous tumor lysates using Western blot analysis showed that fs120-LS tumors expressed higher total levels of laminin compared with fs188-LS tumors with the reverse true for collagen-I (Fig. 5A and B). Laminin could be detected in protein extracts of fs120 and fs188 cell cultures in vitro. Collagen-I was detected in protein extracts of fs188 cells as high molecular weight aggregates that were absent in fs120 cell extracts. Lower molecular weight immunoreactive bands for collagen-I were detected at low levels in lysates of fs120 cells but these did not align with immunoreactive forms seen in fs188 cells or pooled tumor lysates, indicating fs120 cells do not make detectable levels of collagen-I (Supplementary Fig. S5). Immunofluorescence staining of tumors showed laminin is associated with the vasculature in both tumor types with very prominent intense staining in fs188-LS tumors (Fig. 5C) consistent with increased vascular maturation associated with greater αSMA–positive pericyte coverage (Fig. 4G; ref. 12). However, laminin was expressed more widely throughout the fs120-LS tumor stroma and with a greater staining intensity in comparison with fs188-LS tumors, accounting for the differences seen on Western blots between the two tumor types (Fig. 5C). Collagen-I was partially associated with the vasculature in fs188-LS tumors but also more widely distributed, whereas there was negligible collagen-1 staining in fs120-LS tumors (Fig. 5D).

We then investigated migration of fs120 and fs188 cells on laminin versus collagen-I matrices and its dependency on VEGFA using B20-4.1.1. Using transwells coated with collagen-I or laminin on the upper surface, serum-starved fs120 cells migrated more extensively towards the serum-containing bottom chamber than fs188 cells on laminin, whereas fs188 cells were more migratory on collagen-I (Fig. 5E and F). B20-4.1.1 inhibited fs120 cell migration on laminin, but not on collagen-I–coated membranes, and had no significant effect on migration of fs188 cells on either matrix (Fig. 5E and F). We also investigated single-cell migration on laminin or collagen-I–coated plastic in the presence of serum measured using live video microscopy, so that migration could be quantified more accurately and independently of chemotactic effects. In this case, fs188 cell migration speed and distance was found to be very similar on laminin and collagen and faster than fs120 cell migration on both matrices (Supplementary Fig. S6A and S6B). However, as for the transwell assay, only fs120 cell migration on laminin could be inhibited by B20-4.1.1 (Supplementary Fig. S6A and S6B).

We next sought to investigate how fs120-LS cells gain a survival advantage over fs188-LS cells in the lung during the first 48 hours after intravenous injection. We found that bioluminescence of fs120-LS cells in the lung was significantly higher than fs188-LS cells even at 5 minutes postinjection (4-fold; P < 0.01; Fig. 6A). Lung bioluminescence significantly decreased by 48 hours in both cell types by 400- to 700-fold (P < 0.01; Fig. 6A), with a similar differential between the cell types at 48 hours (6-fold; P < 0.01; Fig. 6A) as previously observed (see Fig. 2E). Immunofluorescence staining of lung sections to detect the SV40-positive (SV40^+VE) fibrosarcoma cells directly confirmed significantly fewer fs188-LS cells compared with fs120-LS cells at 5 minutes, with the majority still trapped within the vasculature, indicating differences in efficiency of vascular arrest within the lung (P < 0.05; Fig. 6B and C). VEGFR1 activity had no significant influence on cell numbers within the lung at 5 minutes (Fig. 6B). SV40 staining of fs120-LS and fs188-LS cells showed subpopulations of tumor cells that were highly distorted and disrupted, as well as more rounded cells (Fig. 6C), suggesting cancer cell trauma on encountering the microvasculature of the lung (26). To investigate whether there was a difference between the cell types, we measured γH2AX phosphorylation (p-γH2AX) in SV40^+VE nuclei by immunofluorescence staining (Fig. 6D) to detect activation of the DNA damage response pathway in response to mechanical trauma or nuclear deformation (27). We found that fewer fs120-LS cells were positive for nuclear p-γH2AX than fs188-LS cells at 5 minutes (P < 0.05, Fig. 6E) and this was independent of VEGFR1 activity (Fig. 6E). Immunofluorescence staining of lungs of mice 48 hours after intravenous injection of cells for Ki67, a marker of proliferation, showed an increase in Ki67/SV40 double-positive fs120-LS cells, (Fig. 6F), compared with fs188-LS cells (P < 0.01, Fig. 6G). These results demonstrate that fs120-LS cells can survive initial arrest in the vasculature of the lung better and subsequently proliferate more efficiently than fs188-LS cells.

CD11b antigen–positive (CD11b^+VE) leukocyte recruitment is an early event that supports survival of cells during metastasis and has been shown to require stromal activity of VEGFR1 (28). Five minutes after intravenous injection of fs120-LS or fs188-LS cells, no increase in total numbers of CD11b^+VE cells was detected compared with control lungs and there was no difference between VEGFR-WT and ΔTKD mice (Fig. 7A). CD11b^+VE cells clustered around fs120-LS and fs188-LS cells equally in the lungs of mice at 5 minutes (Supplementary Fig. S7A and S7B). At 48 hours, CD11b^+VE cell counts were significantly increased in lungs containing fs120-LS cells in VEGFR1-WT mice (P < 0.001) but not in VEGFR1-ΔTKD mice (Fig. 7A). Conversely, injection of fs188-LS cells did not increase the levels of CD11b^+VE cells in the lungs at 48 hours and there was no dependence on the VEGFR1 TKD status of the mice (Fig. 7A). B20-4.1.1 treatment did not alter the numbers of CD11b^+VE cells within the lungs of mice injected intravenously with fs120-LS cells at 48 hours (Fig. 7B). These results show that the presence of fs120-LS cells but not fs188-LS cells in lung tissue leads to an increase in CD11b^+VE immune cells that is dependent on VEGFR1 stromal activity.

We have found that metastasis of fibrosarcomas to the lung from subcutaneous tumors is strongly influenced by which isoforms of VEGFA are expressed by the cancer cells, with the more soluble isoforms associated with increased metastasis. Moreover, results clearly showed that tumor cells expressing VEGF120 versus VEGF188 had increased survival in the lung 48 hours after intravenous injection and that the presence of a VEGF120-expressing primary tumor enhanced this effect. Survival of fs120-LS tumor cells in the lung after 48 hours was associated with stromal activity of VEGFR1 yet was not influenced by inhibition of VEGFA itself using the antibody B20-4.1.1. In contrast, B20-4.1.1 did significantly decrease lung metastasis from primary fs120-LS subcutaneous tumors, with an insignificant effect on fs188-LS tumors (summarized in Supplementary Fig. S8).

Our previous studies in mice have shown that expression of short isoforms of VEGFA (120/164) lead to changes in the primary tumor microenvironment known to associate with metastasis including reduced pericyte coverage and increased interstitial fluid pressure (12, 13); important factors influencing cell survival during metastasis (15, 29). In addition, our studies in vitro have shown that VEGF188-expressing tumor cells have increased levels of apoptosis and decreased proliferation compared with VEGF120-expressing cells (24). In the current study, we show that VEGF120-expressing cells have increased survival and proliferation in the lung following intravenous injection. A recent study published by Benedetto and colleagues showed that VEGF189 overexpression in human MDA-MB-231 breast cancer cells increased apoptosis and decreased establishment of experimental lung metastases in mice after intravenous injection (30). Our current results are consistent with these data and furthermore show that differential effects of the isoforms on lung metastasis are apparent when VEGFA in tumor cells is under the control of its constitutive promoter, ensuring physiologic regulation of its production.

We were able to identify at which point metastasis is sensitive to anti-VEGFA therapy by comparing metastasis to the lung from subcutaneous tumors with tumor cell survival after intravenous injection. VEGF120 expression led to increased lung metastasis from subcutaneous tumors, which was sensitive to B20-4.1.1, and increased survival of tumor cells after intravenous injection, which was insensitive to B20-4.1.1. We found no VEGFA isoform–selective effect of B20-4.1.1 on subcutaneous tumor growth, vascular density, or pericyte investment that could account for the preferential effect of B20-4.1.1 on fs120-LS tumor metastasis from the subcutaneous site, although the antibody did significantly decrease vascular density in both fs120-LS and fs188-LS subcutaneous tumors, which is likely to have contributed to the antimetastatic effect observed in the fs120-LS tumors. Characterization of the ECM within subcutaneous tumors revealed differences in laminin versus collagen-I content, with fs120-LS tumors having higher levels of laminin and lower levels of collagen-I compared with fs188-LS tumors. The distribution of laminin and collagen was also different between fs120-LS and fs188-LS tumors, with laminin and collagen having a greater degree of association with the vasculature in fs188-LS tumors. This is consistent with increased vascular maturation and decreased vascular permeability in fs188-LS tumors (12), which could contribute to their lower metastatic propensity. Laminin was expressed at higher levels in fs120-LS tumors compared with fs188-LS tumors and was distributed throughout the tumor tissue not just localized to the vasculature. Migration of fs188 cells on both laminin and collagen in vitro was insensitive to B20-4.1.1. In contrast, fs120 cell's migration on laminin, but not collagen, was sensitive to B20-4.1.1. Considering the widespread prominence of laminin in fs120-LS tumors, this could at least partially explain the particular sensitivity of this tumor type to the antimetastatic effect of B20-4.1.1. Why fs120 cells have a dependence on VEGFA for migration on laminin requires further investigation. Interaction between VEGFA, VEGF receptors and the coreceptor neuropilin-1 (NRP1) can regulate migration and this has been reported to be VEGFA isoform dependent in some cell types (3, 31, 32). Such a selectivity of interaction could be involved in our observations, as our previous studies have shown that fs120 and 188 cells express VEGFR1 and 2 and NRP1 to varying extents (24).

In addition to effects in the primary tumor, we also identified that there was increased survival of fs120 versus fs188 cells within the lung 48 hours after intravenous injection, which was insensitive to treatment with B20-4.1.1. There were also more fs120-LS cells arrested within the lung vasculature in the first 5 minutes after intravenous injection compared with fs188-LS cells. Previous long-standing studies have shown that overcoming the initial trauma of forces associated with capillary arrest is a significant rate-limiting step in metastasis (26). Correlations between increased metastasis and greater cell deformability are also well established (33, 34) and more recently this has been linked to activation of γH2AX phosphorylation, a marker commonly used to detect DNA damage (27). Our data indicate that fs120-LS cells have greater resistance to trauma incurred through vascular arrest than fs188-LS cells, demonstrated by lower activation of the DNA damage marker p-γH2AX. Consistent with the in vitro characteristics of the tumor cells (24), the proportion of proliferative tumor cells in the lung, within the first 48 hours after intravenous injection, was significantly greater for fs120-LS than fs188-LS cells. We have previously shown fs120 cells have greater actinomyosin contractility and display a more amoeboid morphology in vitro, unlike fs188 cells that display more orthodox mesenchymal features (24). This could explain the increased capacity of fs120 cells to resist mechanical trauma during circulation and capillary arrest. Furthermore fs120 cells exhibit greater plasticity and can shift between amoeboid and mesenchymal states (23), a characteristic that also potentially favors their escape through matrix barriers (35). Relatively long-term adaptive changes within the tumor cells to expression of a single VEGFA isoform leading to these differences between fs120 and fs188 cells could explain why treatment with B20-4.1.1 over the 48 hours examined in this study could not inhibit survival of intravenously injected luciferase-expressing cells within the lung. Further studies are required to investigate these processes.

Survival of tumor cells in the lung can be supported by recruitment of CD11b cells (28, 36); a process that is also dependent on stromal VEGFR1 (20, 28). We found that intravenous injection of fs120 cells, but not fs188 cells, increased levels of CD11b cells in the lung at 48 hours and this was dependent on stromal VEGFR1 activity. VEGFR1 dependency was also seen on metastasis to the lung from subcutaneous fs120 tumors, although in this case there were clearly additional primary tumor-related factors influencing metastatic spread, as the presence of a subcutaneous fs120 tumor also increased 48-hour survival of intravenously injected cells in the lung. Although an increase in plasma VEGFA and PlGF2 was detected in mice bearing fs120-LS tumors, we could not replicate the 48-hour survival effect by injecting conditioned medium as reported in the literature in other model systems (28, 37). We cannot rule out the possibility that the primary tumor releases other tumor-derived factors that are not present in cell-conditioned medium, such as exosomes, growth factors, or cells derived from the activated stroma (38). However, metastasis of fs120-LS cells to the lung from subcutaneous fs120-LS tumors was an early event, suggesting that cells already within the lung may increase survival of cells that arrive subsequently. The function-blocking VEGFA antibody, B20-4.1.1 had no effect on survival of intravenously injected cells in the lung and did not inhibit the increase in CD11b cells. This re-enforces our data showing B20-4.1.1 inhibits VEGF120-dependent processes predominantly in the subcutaneous primary tumor for its antimetastatic effect, as well as re-enforcing a role for VEGF120 expression in conferring fibrosarcoma cells with additional advantages during the early stages of survival within the tissue of the lung that are not sensitive to anti-VEGFA therapy. As our data show, there is a high degree of redundancy in VEGFR1 ligands in mice with VEGF120-expressing tumors, particularly in plasma and lung tissue, hence targeting VEGFR1 directly may inhibit survival of disseminated cells within the lung, as found in previous studies of other tumor types (39, 40). In a pilot study, we found that treatment with the VEGFR1/2 tyrosine kinase inhibitor cediranib tended to decrease survival of fs120-LS cells within the lung 48 hours after intravenous injection but this effect was not statistically significant (Supplementary Fig. S9). We conclude that either more selective or potent VEGFR1 inhibitors would be more effective or that further resistance mechanisms to VEGFA–VEGFR–targeted therapies are present.

Overall, we conclude that VEGF120 expression confers a metastatic advantage over VEGF188 expression at several levels including high vascular density coupled to more immature, permeable vasculature in the primary tumor, increased vascular arrest and proliferation within the lung, leading to increased survival that is associated with VEGFR1-dependent recruitment of myeloid cells, possibly driven by multiple VEGFR1 ligands including VEGFA and PlGF2. Not all of these advantages in the metastatic cascade can be inhibited by anti-VEGFA therapy, but we have identified the decrease in vascular density coupled to inhibition of cell migration as possible mechanisms by which anti-VEGFA therapy inhibits metastasis to the lung from subcutaneous primary fs120-LS tumors.

Our results support the clinical findings associating high levels of circulating soluble forms of VEGFA with poor prognosis in conjunction with a greater response to bevacizumab (Avastin) and provide a mechanistic basis for these associations. Although the clinical associations were not significant in all settings (1), they were sufficiently promising for initiation of a phase III trial incorporating prospective analysis of plasma VEGFA levels in Her2^−VE breast cancer to be initiated (MERiDiAN, NCT01663727/GO25632). VEGF pathway inhibitors are also becoming part of established treatment of soft tissue sarcoma (STS) following the PALETTE clinical trial (41). Other potential biomarkers of a good response to anti-VEGFA/VEGFR therapy include low levels of NRP-1 (2), which we previously found associated with VEGF120 tumor cell expression (24).

In summary, our data represent substantial progress in understanding the role of tumor VEGFA isoform expression in metastasis and point toward the utility of analyzing VEGFA isoform expression for predicting treatment outcome with bevacizumab. Furthermore, our results suggest that inhibition of VEGFR1 activity in combination with anti-VEGFA therapy may be particularly useful for treatment of tumors with high expression of soluble isoforms of VEGFA.

No potential conflicts of interest were disclosed.

Conception and design: W.R. English, C. Kanthou, G.M. Tozer

Development of methodology: W.R. English, D.V. Lefley, J.E. Hurrell, S.K. Lyons, C. Kanthou, G.M. Tozer

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): W.R. English, S.J. Lunt, D.V. Lefley, M. Dhingra, Y.-C. Lee, J.E. Hurrell, G.M. Tozer

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): W.R. English, M. Dhingra, Y.-C. Lee, K. Bingham, C. Kanthou, G.M. Tozer

Writing, review, and/or revision of the manuscript: W.R. English, S.J. Lunt, C. Kanthou, G.M. Tozer

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): W.R. English, S.J. Lunt, M. Fisher, D.V. Lefley, J.E. Hurrell

Study supervision: W.R. English, C. Kanthou, G.M. Tozer

Other (approved final version): G.M. Tozer

We would like to thank Genentech Inc. for the antibodies BE5 and B20-4.1.1 and Professors Shibuya and Plate for the VEGFR1-ΔTKD mice. We would also like to thank Susan Clark and Kay Hopkinson of the UoS Medical School flow cytometry core facility, Dr. Phil Watson and his team for assistance with fluorescence microscopy, Colin Gray for assistance with confocal microscopy, Maggie Glover and Jennifer Globe of the histology core facility, Dr. Toby Holmes for his assistance with the IVIS Lumina II, and the staff of the Biological Services Unit for their technical assistance.

W.R. English, S.J. Lunt, M. Fisher, D.V. Lefley, J.E. Hurrell, C. Kanthou, and G.M. Tozer received support from Cancer Research UK (program grant C1276/A9993), Yorkshire Cancer Research (equipment grant S311), and the University of Sheffield. M. Dhingra was supported by a fellowship from the Jean Shanks Foundation.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.