Live imaging of neolymphangiogenesis identifies acute antimetastatic roles of dsRNA mimics

Abstract

Long-range communication between tumor cells and the lymphatic vasculature defines competency for metastasis in different cancer types, particularly in melanoma. Nevertheless, the discovery of selective blockers of lymphovascular niches has been compromised by the paucity of experimental systems for whole-body analyses of tumor progression. Here, we exploit immunocompetent and immunodeficient mouse models for live imaging of Vegfr3-driven neolymphangiogenesis, as a versatile platform for drug screening in vivo. Spatiotemporal analyses of autochthonous melanomas and patient-derived xenografts identified double-stranded RNA mimics (dsRNA nanoplexes) as potent inhibitors of neolymphangiogenesis, metastasis, and post-surgical disease relapse. Mechanistically, dsRNA nanoplexes were found to exert a rapid dual action in tumor cells and in their associated lymphatic vasculature, involving the transcriptional repression of the lymphatic drivers Midkine and Vegfr3, respectively. This suppressive function was mediated by a cell-autonomous type I interferon signaling and was not shared by FDA-approved antimelanoma treatments. These results reveal an alternative strategy for targeting the tumor cell-lymphatic crosstalk and underscore the power of Vegfr3-lymphoreporters for pharmacological testing in otherwise aggressive cancers.

Keywords: GEMM melanoma models; dsRNA nanoplexes; midkine; neolymphangiogenesis; premetastatic niche.

Figure 1. Identification of antilymphangiogenic compounds in Vegfr3^Luc genetically engineered mouse models (GEMMs)

1. Schematic representation of the Vegfr3^Luc ‐GEMM (MetAlert) mice to assess melanomas driven by melanocytic‐specific induction of oncogenic Braf^V600E in a Pten‐deficient background (1), as well as to monitor xenografts of human cells (2) and patient‐derived specimens (PDX, 3).

2. Luciferase‐based imaging of drug response in Vegfr3^Luc;Tyr:CreERT2;Braf^V600E; Pten^{flox / flox} mice. Panels labeled as "basal" and "induced" correspond to the bioluminescence of animals prior and 5 weeks after administration of 4OH‐tamoxifen (5 mM, topical administration, 3 consecutive days) for the induction of melanomas. Right panels: Treatment with anti‐PD‐L1 antibody (αPD‐L1; clone 10F.9G2, 3 weeks) or the corresponding control IgG (200 µg/dose, twice per week, 3 weeks); vemurafenib (Vem, 50 mg/kg, oral once per day, 3 weeks), or BO‐110 (BO, 0.8 mg/kg, twice per week, 3 weeks). Scale: p/s/cm2/sr (×10⁶).

3. Growth curves of Vegfr3^Luc;Tyr:CreERT2;Braf^V600E; Pten^{flox / flox} melanomas treated with αPD‐L1 or IgG (200 µg/dose, 2 doses/week; left panel), or with vemurafenib (Vem, 50 mg/kg, daily dose), BO‐110 (BO, 0.8 mg/kg, 2 doses/week) or vehicle control (V, daily dose) as indicated. Data correspond to the average tumor size ± SD at the indicated time points. Red arrows mark the initiation of treatment (n = min 5 mice per condition). Two‐way ANOVA statistics. P = 0.0012 (αPD‐L1), P = 0.0007 (BO‐110) and P = 0.0353 (vemurafenib).

4. Histological visualization of lymphatic vessel density (dual Lyve11 Prox1 staining) in representative sections of tumors of Vegfr3^Luc;Tyr::CreERT2;Braf^V600E;Pten^{flox / flox} melanomas in mice treated with vehicle (V) or 4 doses of BO‐110 (BO, 0.8 mg/kg). Double‐positive (Lyve1, Prox1) vessels were pseudocolored to green to ease the visualization. See also images for lung lymphatic vessels in Fig EV1C. Red arrowheads indicate Lyve1‐Prox1‐positive lymphatic vessels.

5. Quantification of lymphatic and blood vessels density in tumors and lungs of Vegfr3^Luc; Tyr::CreERT2;Braf^V600E;Pten^{flox / flox} melanomas after treatment as indicated in B, C. Data correspond to the quantification of four fields per tumor, performed in biological triplicates. Statistical significance was determined by the Mann–Whitney t‐test.

6. Treatment with BO‐110 of human patient‐derived xenografts (PDX) implanted in Vegfr3^Luc nu/nu. 42 days after implantation (when systemic luciferase was detected), animals were randomized for treatment with vehicle (V) or with 0.8 mg/kg BO‐110 (BO, twice per week), and luciferase emission was acquired at the indicated times. Scale, p/s/cm²/sr (×10⁶).

7. Quantification of the inhibitory effect of BO‐110 (BO, 0.8 mg/kg, 2 doses/week I.P. administration, 11 weeks) on the growth of melanoma PDXs. Red arrows mark the initiation of treatment. Shown are mean tumor size in mm³ ± SD in biological triplicates. Statistical significance was determined by two‐way ANOVA. P = 0.0009.

8. Representative sentinel, axillary, and brachial lymph nodes (SLN, ALN, and BLN, respectively) of mCherry‐SK‐Mel‐147‐driven xenografts in Vegfr3^Luc nu/nu mice treated with vehicle (V) or four doses of BO‐110 (BO, 0.8 mg/kg) and imaged for mCherry fluorescence to assess metastatic potential as a function of treatment. Numbers in parenthesis correspond to mice with positive metastases in at least one LN (lymph node) with respect to the total animals analyzed per condition. Scale, p/s/cm²/sr (×10⁸). See also Fig EV1D and E for images and quantification of lymphatic vessels in lymph nodes.

Source data are available online for this figure.

Figure EV1. Identification of BO‐110 as an antilymphangiogenic agent in MetAlert mice, with controls for selective inhibition of Vegfr3‐Luc signaling

1. Impact of BO‐110 on Tyr:CreERT2; BRAF^V600E;Pten^{flox / flox}; Vegfr3^Luc mice. Upper panels correspond to optical photographs of animals treated with vehicle or with 6 doses of BO‐110 (0.8 mg/kg, twice a week, 3 weeks), and depilated to ease in the imaging. These same animals are shown in the bottom panels for luciferase emission centering on the tumor (1, 3) or on sentinel LN (2, 4). Scale, p/s/cm²/sr (×10⁶).

2. Histological assessment of the impact of BO‐110 (BO) on neolymphangiogenesis in melanomas generated in Vegfr3^Luc;Tyr::CreERT2;BRAF^V600E;Pten^{fl / fl}Pten^{flox / flox} mice. Panels correspond to Lyve1 detected by immunohistochemistry (IHC, brown signal) or to dual fluorescence staining for Lyve1 (red) or Vegfr3 (green). Right panel corresponds to merged signal of Lyve1 and Vegfr3 IF.

3. Right panel, histological visualization of lymphatic vessel density by costaining for Lyve1 (blue) and Prox1 (purple) in representative lungs of Vegfr3^Luc;Tyr::CreERT2;Braf^V600E;Pten^{flox / flox} melanomas treated with vehicle (V) or 4 doses of BO‐110 (BO, 0.8 mg/kg). Left panel, pseudocoloring in red of overlapped staining with anti‐Lyve1 and Prox1 antibodies (cells positive for both markers are highlighted with red arrowheads).

4. Right panel, histological analyses of lymphatic vessel density by Lyve1 and Prox1 dual staining in representative lymph nodes of animals in Fig 2B processed at the endpoint of the experiment (four doses of BO‐110 or vehicle control). Left panel, dual‐positive Lyve1 and Prox1 cells pseudocolored in red.

5. Quantification of lymphatic vessel (left) and blood vessel (right) density in lymph nodes of animals in Fig 2B (SK‐Mel‐147 tumors) processed at the endpoint of the experiment (four doses of BO‐110 or vehicle control). Data correspond to the quantification of four fields per tumor, performed in biological triplicates. Statistical significance was determined by the Mann–Whitney t‐test.

6. In vivo imaging of the comparative impact of BO‐110 (0.8 mg/kg) (BO) or vehicle (V) on luciferase emission driven from the Vegfr3^Luc MetAlert mice bearing a SK‐Mel‐147 tumor, or from an unrelated promoter (SV40‐Luc) stably expressed in SK‐Mel‐147 melanoma cells.

7. Quantification of luciferase emission 24 h after treatment in the tumor and the indicated organs in animals treated as in (F). N = 6 mice per condition. Boxplots show the median, 25^th and 75^th percentiles, and the maximum and minimum signal. Luciferase signal was normalized to the corresponding vehicle control. Statistical significance was determined by ANOVA.

Source data are available online for this figure.

Figure 2. Inhibitory effects of BO‐110 on prolymphangiogenic factors

1. Luciferase‐based imaging of short‐term drug response in Vegfr3^Luc;Tyr::CreERT2; Braf^V600E; Pten^{flox / flox} mice. Panels labeled as "basal" and "induced" correspond to the bioluminescence of animals prior and 5 weeks after administration of 4OH‐tamoxifen (5 mM, topical administration, three consecutive days) for the induction of melanomas. Right panels: images of mice that were treated the day before with one dose of the indicated compounds: αPD‐L1 antibody (clone 10F.9G2) or the corresponding control IgG (200 µg/dose); vemurafenib (Vem, 50 mg/kg); or BO‐110 (BO, 0.8 mg/kg). Scale: p/s/cm2/sr (×10⁶).

2. Response of xenografts of mCherry‐labeled SK‐Mel‐147 in Vegfr3^Luc nu/nu lymphoreporter mice treated with one dose (24 h) or 4 doses of BO‐110 (BO, 0.8 mg/kg). Left panels correspond to Vegfr3‐Luciferase (neolymphangiogenesis) and right panels to mCherry fluorescence emission (tumor content). Scale, Vegfr3^Luc: p/s/cm²/sr (×10⁶) and mCherry: p/s/cm²/sr (×10⁹).

3. qRT–PCR analysis of relative mRNA levels of MDA5 16 h after treatment of HLEC with 0.5 µg/ml BO‐110 (BO), 10 µM vemurafenib (Vem), or vehicle control (V). Data correspond to the mean ± SD of three biological replicates. Statistical significance was determined by the t‐test.

4. qRT–PCR analysis of relative mRNA levels of VEGFR3 16 h after treatment of HLEC with 0.5 or 1 µg/ml BO‐110 (VO), 10 µM vemurafenib (Vem), or the corresponding vehicle control (V). Data correspond to the mean ± SD of three biological replicates. Statistical significance was determined by ANOVA.

5. Luciferase signal driven by FLT4 (VEGFR3)‐promoter transduced into HLEC treated with vehicle (v) or BO‐110 (BO) at the indicated doses (µg/ml) as indicated in Methods. Results were normalized to vehicle control. N = 4 biological replicates. Error bars correspond to mean ± SD. Statistical significance was determined by ANOVA.

6. Tubulogenic activity of HLECs in the presence of BO‐110. Images correspond to cells plated in Matrigel and imaged at the indicated time points after treatment with 0.5 µg/ml BO‐110. Complete time‐lapse imaging of this process is shown in Movie EV1 (Appendix).

7. Analysis of apoptotic cells at the indicated time points. HLEC cells were treated with vehicle (V) or 0.5 µg/ml BO‐110 (BO) for the indicated time points. Cells were collected and apoptosis was analyzed by flow cytometry as indicated in Methods. Data correspond to the mean ± SD of three experiments. Statistical significance was determined by the t‐test.

8. Relative mRNA levels of VEGFC and VEGFD in the indicated melanoma cell lines 8 h after treatment with vehicle (V) or 0.5 µg/ml BO‐110 (BO), as determined by qRT–PCR. Data correspond to the mean ± SD of three experiments. Statistical significance was determined by the t‐test.

Figure 3. Inhibitory effects of BO‐110 on Midkine

1. Inhibitory effect of the indicated doses of BO‐110 (in µg/ml) or vehicle (V) on MDK mRNA expression determined by qRT–PCR in SK‐Mel‐147 (16 h after treatment). Data correspond to average mRNA levels in three experiments with technical replicates normalized to vehicle control ± SD. Statistical significance was determined by ANOVA.

2. qRT–PCR analysis of relative mRNA levels of MDA5 16 h after treatment of SK‐Mel‐147 with the indicated doses of BO‐110 (in µg/ml) (BO). Data correspond to the mean ± SD of three experiments with three technical replicates. Statistical significance was determined by ANOVA.

3. Analysis of apoptotic cells at the indicated time points. SK‐Mel‐147 cells were treated with vehicle (V) or 0.5 µg/ml BO‐110 (BO) for the indicated time points. Cells were collected, and apoptosis was analyzed by flow cytometry as indicated in Methods. Data correspond to the mean ± SD of three experiments. Statistical significance was determined by the t‐test.

4. Luciferase signal driven by MDK promoter transduced into SK‐Mel‐147 cells treated with vehicle (v) or BO‐110 (BO) as indicated in Materials and Methods. Results were normalized to vehicle control. Data correspond to the mean ± SD of four biological replicates. Statistical significance was determined by ANOVA.

5. MDK secretion by ELISA in SK‐Mel‐147 melanoma cells treated with vehicle (V) or 0.5 µg/ml BO‐110 (BO). Data correspond to the mean ± SD of three biological replicates. Statistical significance was determined by two‐way ANOVA.

6. Immunohistochemical analysis of MDK repression (pink staining) in SK‐Mel‐147 xenografts and PDX lesions after treatment with BO‐110 (BO, 0.8 mg/kg, 2 doses/week). Histological staining in tumors extracted from animals treated with vehicle control (V) is included as a reference. Nuclei were counterstained with hematoxylin.

Figure EV2. MDA5 induction and Mdk repression by BO‐110 in a panel of melanoma cell lines

1. Inhibitory effect of the indicated doses of BO‐110 (in µg/ml) on MDK mRNA expression determined by qRT–PCR in the indicated melanoma cell lines 16 h after treatment. Data correspond to average mRNA levels of three experiments with three technical replicates normalized to vehicle control (V) ± SD. Statistical significance was determined by ANOVA.

2. qRT–PCR analysis of relative mRNA levels of MDA5 16 h after treatment of the indicated melanoma cell lines with the 0.5 or 1 µg/ml of BO‐110 (BO) or vehicle (V). Data correspond to the mean ± SD of three experiments with three technical replicates. Statistical significance was determined by ANOVA.

3. Percentage of apoptotic cells (Annexin‐V staining detection by flow cytometry) after treatment of the indicated cell lines with vehicle (V) or 0.5 µg/ml BO‐110 (BO) for the indicated time points. Data correspond to the mean ± SD of three experiments. Statistical significance was determined by the t‐test.

Figure EV3. Omic analyses of the BO‐110 mechanism of action

1. Heatmap showing differentially deregulated signaling cascades in SK‐Mel‐147 treated with 1 µg/ml B0‐110 for 4 or 10 h (versus vehicle‐treated controls). Data correspond to mRNA expression profiles analyzed by GSEA using the Hallmark gene sets. The scale indicates the normalized enrichment score (NES).

2. Venn diagrams depicting common and specific Hallmarks gene sets significantly enriched in a GSEA in cell lines SK‐Mel‐147 and HLEC upon BO‐110 treatment (left: upregulated; right: downregulated; FDR < 0.25). See also Datasets EV3 and EV4 for additional detail.

3. Venn diagrams as in B, but estimated for REACTOME gene sets. See also Datasets EV5 and EV6 for the complete gen set list.

4. Enrichment plots for the indicated signaling cascades gene sets upon BO‐110 treatment in SK‐Mel‐147 or in HLEC.

Figure 4. High‐throughput analysis reveals IFN induction as a key component of the BO‐110 mechanism of action

1. Heatmaps summarizing proteomic analyses (iTRAQ) performed in SK‐Mel‐28 and SK‐Mel‐147 after treatment with BO‐110 (1 µg/ml, 10 h). Shown are Hallmark gene sets with NES > 1. See also Datasets EV1 and EV2 for additional information.

2. REACTOME gene set analysis of protein changes in SK‐Mel‐28 and SK‐Mel‐147 cell lines treated with BO‐110 (1 µg/ml, 10 h).

3. Type I IFN mRNA induction (IFNA2 and IFNB1) in SK‐Mel‐147 melanoma cells treated for 16 h with the indicated amounts of BO‐110 (in mg/ml). Data correspond to the mean ± SD of three experiments with three technical replicates normalized to vehicle control. Statistical significance was determined by ANOVA.

4. Heatmap showing differentially deregulated signaling cascades in SK‐Mel‐147 and HLEC treated with 1 µg/ml B0‐110 for 10 h (versus vehicle‐treated controls). Data correspond to mRNA expression profiles analyzed by GSEA using the Hallmark gene sets. The scale indicates the normalized enrichment score (NES). See also Dataset EV3.

5. Heatmap depicting expression changes in interferon‐related genes (GO:0034340) in SK‐Mel‐147 melanoma cells (left panel) and HLEC (right panel) treated with vehicle or 0.5 g/ml of BO‐110 for 10 h. CD274, LAG3 and PDCD1 genes were also included as a reference.

Figure 5. Mechanistic analyses of the repressive activity of BO‐110 on melanoma‐induced neolymphangiogenesis

1. IFNB1 mRNA induction analyzed by qPCR at the indicated times after BO‐110 treatment (0.5 µg/ml) of SK‐Mel‐147 melanoma cells or HLEC (left and right graphs, respectively). Data correspond to the mean ± SD of three experiments with three technical replicates normalized to vehicle control.

2. Quantification of the impact of BO‐110 as single agent or in the presence of the indicated blocking antibodies for type I interferon (IFNB1 or IFNAR1). Upper graphs show the effect of these agents on MDK mRNA levels in SK‐Mel‐147 melanoma cells. Similar treatments were performed on HLEC for the analysis of VEGFR3 mRNA (middle graphs) and tube formation capacity (lower graphs). Data correspond to the mean ± SD of 3 biological replicates in triplicate.

3. BO‐110‐driven blockade of the tube‐forming capacity of HLEC and rescue with anti‐IFNAR1 blocking antibodies. Images correspond to cells plated in Matrigel and imaged 8 h after treatment with 0.5 µg/ml BO‐110. See also Fig EV4B, for additional results with anti‐IFNAR1 and anti‐IFN‐β blocking antibodies.

4. Growth of B16 melanoma xenografts in siblings of Ifnar1 ^+/+, Ifnar1 ^+/−, or Ifnar1 ^−/− mice. Treatment started 10 days after tumor cell implantation. BO‐110 was administered at 0.8 mg/kg, every third day for 2 weeks. N = 6 mice per condition. Graphs show the mean tumor size ± SD at each time point. Statistical significance was determined by two‐way ANOVA.

5. Histological analyses of lymphatic vessel density by Lyve1 (blue) and Prox1 (purple) in representative lymph nodes of animals in (D) processed at the endpoint of the experiment (four doses of BO‐110 or vehicle control). N = 6 mice per experimental condition. See Fig EV4C for a more complete view of these lymph nodes, where dual Prox1‐Lyve1‐positive cells were pseudocolored in red.

Figure EV4. IFN‐dependent inhibition of lymphangiogenesis by BO‐110

1. Binding sites for experimentally validated IRF‐related transcription factors in the promoters of MDK and FLT4 (VEGFR3). Data from Interferome and TRANSFAC (www.genexplain.com).

2. BO‐110‐driven blockade of the tube‐forming capacity of HLEC and rescue with anti‐IFNAR1 blocking antibodies (anti‐IFN‐β and anti‐INFAR1) or corresponding controls. Images correspond to cells plated in Matrigel and imaged 8 h after treatment with 0.5 µg/ml BO‐110.

3. Histological analyses of lymphatic vessel density defined by dual staining for Lyve1 (blue) and Prox1 (purple). To facilitate the visualization, cells positive for both markers were pseudocolored in red. Large magnifications of the doted squared areas can be found in Fig 5E.

4. Quantification of Lyve1/Prox1 staining in lymph nodes of animals in Fig 5D and E, processed at the endpoint of the experiment (4 doses of BO‐110 or vehicle control). Data correspond to the mean ± SD of four biological replicates. Statistical significance was determined by the t‐test.

Source data are available online for this figure.

Figure 6. BO‐110 induces a systemic interferon response that inhibits Mdk blood levels and tumor‐induced lymphangiogenesis

1. A

   Impact of BO‐110 on tumor‐bearing mice implanted with syngeneic B16R2L (5 × 10⁵ cells). When tumors have an average size on 100 mm³, mice were randomized into two groups and treated with BO‐110 (BO) or vehicle (V), every second day for 2 weeks. The arrow indicates the start of the treatment. Tumor size was measured with a caliper at the indicated time points.

2. B

   Quantification of Vegfr3^Luc emission in the tumor area in mice in experiment (A).

3. C

   Vegfr3^Luc emission in the inguinal and brachial lymph nodes of mice in experiment (A).

4. D–F

   Quantification of Vegfr3^Luc emission in the spleen, liver, and lung of mice in experiment (A), respectively.

5. G

   ELISA analysis of Mdk blood levels in mice in experiment (A).

6. H

   ELISA analysis of mouse Ifn‐β blood levels in mice in experiment (A).

7. I

   Antitumoral effect of BO‐110 on xenografts of SK‐Mel‐147 (1 × 10⁶ cells) implanted in the back of Vegfr3^{Luc; nu / nu} nude mice. When tumors had an average size on 150 mm³, mice were randomized into two groups and treated with BO‐110 (BO, 0.8 mg/kg) or vehicle (V), every second day for 2 weeks. The arrow indicates the start of the treatment. Tumor size was measured with a caliper at the indicated time points, and tumor volume was calculated as indicated in Materials and Methods.

8. J

   Quantification of Vegfr3^Luc emission in the tumor area in mice in experiment (I).

9. K–N

   Quantification of Vegfr3^Luc emission in the lymph nodes, spleen, liver, and lung of mice in experiment (I), respectively.

10. O

    ELISA analysis of Mdk blood levels in mice from (I).

11. P

    ELISA analysis of mouse Ifn‐β blood levels in mice from (I).

Data information: For all panels in this figure, N = 4 mice per condition. Statistical significance was determined by two‐way ANOVA.

Figure EV5. Disseminated tumor cells at lymph node and lung lymphovascular metastatic niches

Histological analysis of lymph nodes in basal conditions or in mice implanted s.c with mCherry‐labeled SK‐Mel‐147 performed before tumor excision as in Fig 7A. Vegfr3 and mCherry were detected by immunofluorescence (green and red signaling, respectively). Panels 1–4 correspond to a larger magnification of different areas in these lymph nodes. Note that Vegfr3 concentrates in lymphatic endothelial cells (LECs), with undetectable levels in melanoma cells. T, tumor cells.

Figure 7. BO‐110 adjuvant treatment prevents metastatic melanoma relapse

1. Efficacy of BO‐110 as an adjuvant (preventing relapse after surgical removal of the primary lesion). Shown are representative examples of Vegfr3^Luc mice implanted with mCherry‐SK‐Mel‐147 and imaged for luciferase emission (prior to and after tumor removal). Animals were left to recover from surgery (4 days) and then treated for 2 weeks (four doses) with 0.8 mg/kg BO‐110 or vehicle control (n = 8 for control and n = 10 for treatment arm). Scale, p/s/cm²/sr ×10⁶.

2. mCherry emission from tumor cells of the animals in (A). Scale, p/s/cm²/sr ×10⁹.

3. The Kaplan–Meier survival curves of animals treated as in (A). 8 of 8 animals treated with vehicle (V) control had to be sacrificed for humane reasons 110 days after surgery. 9 of 10 animals in the BO‐110 arm (BO) remained tumor‐free 8 months after stopping treatment. The gray box marks the period of treatment with BO‐110. Statistical significance was determined by the logrank test.