Targeted Therapy Response Linked to NR2F1 Expression

NR2F1 is upregulated in melanoma tumors following BRAFi and MEKi therapy. NR2F1 is associated with tumor dormancy in several tumor models (27, 29), and the BRAFi + MEKi drug-tolerant state exhibits traits similar to cellular dormancy, prompting us to test the link between NR2F1 and targeted therapy response. First, we analyzed published RNA-Seq datasets from patients with melanoma treated with BRAFi + MEKi (30–32). Sample expression and metadata were retrieved from the Gene Expression Omnibus (GEO) database, and analysis was limited to patients treated for up to 20 days with the combination of BRAFi (dabrafenib) and MEKi (trametinib) (Figure 1A). In the datasets from Kwong et al. (30) and Song, Hugo, Lo, and co-authors (32), NR2F1 mRNA levels were significantly higher (P = 0.033 and P < 0.001, respectively) in BRAFi + MEKi on-treatment samples compared with patient-matched pre-treatment samples (Figure 1A). Using the Tsoi et al. melanoma drug-tolerant states dataset (33), we analyzed NR2F1 expression in the transition between phenotypic states and targeted therapy tolerance. NR2F1 expression was enhanced in the undifferentiated cell state (Figure 1B), 1of 4 melanoma states identified as highly invasive and resistant to BRAFi + MEKi. To further connect NR2F1, MRD, and therapy tolerance, we analyzed the publicly available single-cell RNA-Seq (scRNA-Seq) dataset from Rambow et al. (12). This dataset compares 4 drug-tolerant states (neural crest stem cell, invasive, “starved-like” melanoma cell, and pigmented) that are present in MRD following BRAFi + MEKi in BRAF-mutant melanoma patient–derived xenografts (PDXs). NR2F1 was selectively upregulated in the MRD-invasive state of DTP cells (Figure 1C and Supplemental Figure 1A; supplemental material available online with this article; https://doi.org/10.1172/JCI178446DS1).

NR2F1 is highly expressed in melanoma lesions of patients on BRAFi + MEKi therapy. (A) Normalized expression of NR2F1 for baseline or before treatment (Pre) and early during treatment (On) tumor samples from patients who received BRAFi + MEKi combination therapy in datasets from Song et al. (32) and Kwong et al. (30). (B) Box plot of NR2F1 RNA-Seq gene expression data for melanoma cell lines categorized across cell states according to the dataset from Tsoi et al. (33). (C) Violin plot of NR2F1 expression levels by cell state in a scRNA-Seq dataset of PDX melanoma following BRAFi + MEKi treatment, according to data from Rambow et al. (12). Cell types present during MRD are shown. NCSC, neural crest stem cell; SMC, “starved-like” melanoma cell. (D) Tumor volume in mice bearing 1205Lu-tdTomato–labeled xenografts following continuous BRAFi + MEKi (PLX4720 200 ppm + PD0325901 7 ppm) for 3 weeks. (E) Representative images of detection of tdTomato fluorescence representing tumor size in xenografts after 3 weeks on BRAFi + MEKi therapy and plot showing the mean nuclear intensity of NR2F1 protein expression in 1205Lu-tdTomato cells by immunofluorescence of tumor xenografts compared with the no-drug-treatment control group. Scale bars: 100 µm. Data are presented as the mean ± SD. *P < 0.05 and ***P < 0.001, by unpaired, 2-tailed Student’s t test.

Additionally, NR2F1 expression was assessed in the Song et al. study (32), which compares transcriptomes of patient-derived tumors on targeted therapy with MAPKi-induced cell states in human melanoma cell lines. The dynamic transcriptomic states are defined as subpopulations of DTP, DTP proliferating cells (DTP-prolif), and BRAFi/MEKi- or BRAFi/MEKi/ERKi-resistant sublines after short-term (2 days) or long-term (weeks/months) treatment (32). Using average expression data taken from this study, we observed that NR2F1 levels were increased after MAPK pathway inhibition in BRAF-mutant cell lines independently of their dynamic transcriptomic states (Supplemental Figures 1, B and C) (32). Together, these data suggest that NR2F1 is expressed in the tumors of patients with melanoma with MRD.

To test the association of NR2F1 with MRD in preclinical models, we identified cell lines with detectable NR2F1 expression (Supplemental Figure 1D) and treated xenografts from the NR2F1-expressing BRAF-mutant 1205Lu cell line with BRAFi (PLX4720) and MEKi (PD0325901). As expected, BRAFi + MEKi caused tumor regression leading to MRD by 3 weeks (Figure 1D). NR2F1 levels were enriched in MRD xenografts treated with BRAFi + MEKi compared with control tumors, as detected by immunofluorescence at the 3-week post-treatment MRD time point (Figure 1E and Supplemental Figure 1E). Additionally, we probed for the expression of NR2F1 in BRAFi + MEKi–tolerant cell lines that we have previously described (24, 34). Publicly accessible RNA-Seq analysis (24) showed increased mRNA expression of NR2F1 in two A375 BRAFi + MEKi tolerance models, CRT34 and CRT35, compared with parental A375 cells (Supplemental Figure 1F). Additionally, there was a modest but detectable increase in protein expression of NR2F1 in CRT34 and CRT35 cells cultured for 24 hours in the presence or absence of BRAFi + MEKi (Supplemental Figure 1G). Together, these data associate NR2F1 upregulation with drug tolerance and MRD following targeted therapy in cutaneous melanoma.

NR2F1 overexpression minimizes tumor inhibition effects by BRAFi and MEKi treatment. To determine whether NR2F1 is sufficient to enhance drug tolerance, we induced the expression of NR2F1 in BRAF-V600E 1205Lu, WM793, and A375 melanoma cell lines and treated them with or without BRAFi + MEKi (Figure 2A). LacZ/β-gal was used as a control for the inducible protein expression system (Supplemental Figure 2A). No other members of the NR2F family were upregulated when NR2F1 was overexpressed in melanoma cell lines (Supplemental Figure 2B). In colony growth assays, inducible expression of NR2F1 did not alter melanoma cell growth in the absence of BRAFi + MEKi but was able to partially rescue colony growth in the presence of BRAFi + MEKi (Figure 2B and Supplemental Figure 2C). We further tested whether NR2F1 expression altered cell proliferation following BRAFi + MEKi by measuring S-entry phase using 5-ethynyl-2’-deoxyuridine (EdU) staining. BRAFi + MEKi–treated cells that overexpressed NR2F1 exhibited higher EdU staining than did non-NR2F1-overexpressing cells on targeted therapy (Figure 2C). Additionally, as BRAFi + MEKi treatment induces cell death (35), we tested whether NR2F1 overexpression affected cell death by measuring propidium iodide (PI) uptake via real-time IncuCyte analysis. NR2F1 overexpression significantly reduced PI uptake during BRAFi + MEKi treatment (Figure 2D).

NR2F1 overexpression minimizes tumor inhibition effects by BRAFi and MEKi therapy. (A) NR2F1 protein levels in human BRAF-mutant melanoma cell lines expressing DOX-inducible NR2F1, 1205LuTR-NR2F1, WM793TR-NR2F1, and A375TR-NR2F1 after 72 hours of treatment using BRAFi + MEKi + DOX (1 μmol/L PLX4720 + 35 nmol/L PD0325901 + 100 ng/mL DOX). (B) Colony assay for cell lines overexpressing DOX-inducible NR2F1 after 1 week of treatment using BRAFi + MEKi + DOX. Original magnification, ×20. (C) Detection of S-phase cell-cycle arrest following EdU staining for the BRAF-mutant human melanoma cell lines listed above after 72 hours of treatment with BRAFi + MEKi + DOX. *P < 0.05, by Tukey’s test. (D) PI uptake over cell confluence in the BRAF-mutant human melanoma cell lines listed above after 72 hours of treatment with BRAFi + MEKi + DOX as determined by IncuCyte analysis. *P < 0.05 and **P < 0.01, by Tukey’s test. (E) Representative images of 3D tumor spheroids of the human BRAF-mutant melanoma cell lines 1205LuTR-NR2F1, WM793TR-NR2F1, and A375TR-NR2F1 after 72 hours of treatment with BRAFi + MEKi + DOX (1 μmol/L PLX4720 + 35 nmol/L PD0325901 + 100 ng/mL DOX). 3D tumor spheroids were stained with calcein-AM (7 μmol/L) for cell viability evaluation. Scale bars: 100 μm. (F) Scheme of coculturing of tdTomato cells overexpressing DOX-inducible NR2F1 WT (94) and GFP cells overexpressing a DOX-inducible dominant-negative form of NR2F1 (38) harboring a C141S point mutation within its DNA-binding domain (C141S). Cells were mixed at a ratio of 1:1 and then cocultured for 72 hours with or without BRAFi + MEKi + DOX (1 μmol/L PLX4720 + 35 nmol/L PD0325901 + 100 ng/mL DOX). Cocultures from F were collected and analyzed by FACS for tdTomato and GFP positivity. (G) The percentage of tdTomato and GFP positivity was compared with DMSO. **P < 0.01 and ***P < 0.001, by 2-way ANOVA. Data are presented as the mean ± SD.

Since drug-tolerant cells may also gain motility and interact with the extracellular matrix to promote disease progression (36, 37), we investigated how NR2F1 influences tumor invasion. Using a 3D tumor spheroid outgrowth assay, we found that NR2F1 expression enhanced melanoma invasiveness in spheroids treated with BRAFi + MEKi (Figure 2E and Supplemental Figure 2D). To confirm that NR2F1 confers a survival benefit to melanoma cells in the presence of BRAFi + MEKi, we inducibly expressed either tdTomato-NR2F1-WT or a dominant-negative form of NR2F1, GFP-NR2F1-C141S, in 1205Lu-TR and WM793TR melanoma cells, which express the Tet repressor (TetR). NR2F1-C141S harbors a point mutation within its DNA-binding domain (38). When cocultured at a 1:1 ratio, tdTomato-NR2F1-WT cells outgrew the GFP-NR2F1-C141S–mutant cells in the presence of BRAFi + MEKi (Figure 2, F and G). In sum, our data suggest that NR2F1 confers a pro-growth and pro-invasion state in melanoma during targeted therapy.

Despite presenting a response evaluation criteria in solid tumors–defined (RECIST-defined) complete response to MAPK inhibitors (39), residual disease often persists, and the disease progresses in approximately 80% of patients (40–42). Thus, we tested whether NR2F1-expressing cells tolerate long periods of targeted therapy. Real-time IncuCyte monitoring of cell growth confirmed the ability of NR2F1-overexpressing cells to maintain growth following a 28-day exposure to BRAFi + MEKi across multiple melanoma lines (Figure 3A). In vivo, we observed that xenografts overexpressing NR2F1 responded to BRAFi + MEKi treatment but progressed faster over time than did noninduced tumors (Figure 3, B and C). Mice harboring tumors overexpressing NR2F1 showed poorer survival than did mice bearing control xenografts on targeted therapy (Figure 3D). Together, these data suggest that BRAF-mutant melanoma cells expressing high levels of NR2F1 tolerate targeted therapy and that tumors develop resistance more quickly.

NR2F1 overexpression promotes tumor relapse following BRAF and MEK inhibitors therapy. (A) IncuCyte live-cell analysis for DOX-inducible cells overexpressing NR2F1 after 4 weeks of treatment using BRAFi + MEKi + DOX (1 μmol/L PLX4720 + 35 nmol/L PD0325901 + 100 ng/mL DOX). Data show the percentage of cell confluence on the plate. (B) NR2F1 protein levels for the DOX-inducible NR2F1-expressing cell line 1205Lu-E2F-tdTomato(tdTW)-TR-NR2F1 after 72 hours of treatment with BRAFi + MEKi + DOX. (C) In vivo tumor growth curves and (D) survival of 1205Lu xenografts with DOX-inducible NR2F1 expression following BRAFi + MEKi (200 ppm PLX4720 + 7 ppm PD0325901 + 25 mg/mL DOX) treatment. ***P < 0.001 and ****P < 0.0001, by Kaplan-Meier analysis.

NR2F1 expression leads to enrichment of growth-related transcripts following BRAFi and MEKi therapy. To investigate how NR2F1 may lead to targeted therapy tolerance, we performed bulk RNA-Seq and gene set enrichment analysis (GSEA) on melanoma cells with or without NR2F1 overexpression that were treated with BRAFi + MEKi. In 3 melanoma cell lines (1205Lu, WM793, and A375), NR2F1 expression induced an enrichment in the hallmark gene sets for MYC targets variant 2 and mTORC1 signaling in the presence BRAFi + MEKi compared with treated control cells that did not overexpress NR2F1 (Figure 4, A and B, and Supplemental Figure 3A). NR2F1 overexpression also increased enrichment in the hallmark gene sets for a late response to estrogen (Supplemental Figure 3B). Some pathways were only altered in 2 of the 3 cell lines. For example, enrichment hallmark gene sets for MYC targets variant 1, DNA repair, E2F targets, G2/M checkpoint, mitotic spindle, and spermatogenesis were upregulated in NR2F1-expressing 1205Lu and A375 cells (Supplemental Figure 3C), whereas augmentation of cholesterol homeostasis, the unfolded protein response, IFN-α response, IFN-γ response, and hypoxia were altered in 1205Lu and WM793 (Supplemental Figure 3D). Additionally, we observed an NR2F1-dependent reduction in an adipogenesis gene set in WM793 and A375 cell lines (Supplemental Figure 3E).

NR2F1 overexpression upregulates cell proliferation and mTORC1 signaling following BRAFi + MEKi therapy. (A) Heatmap showing GSEA normalized enrichment scores (NES) for the hallmark gene set collection comparing NR2F1 expression with no DOX after 72 hours of treatment with BRAFi + MEKi + DOX (1 μmol/L PLX4720 + 35 nmol/L PD0325901 + 100 ng/mL DOX) in 1205LuTR-NR2F1, WM793TR-NR2F1, and A375TR-NR2F1 cell samples. NES values are displayed for enriched gene sets using a Benjamini-Hochberg FDR (BHFDR) cutoff of 0.05. (B) GSEA hallmark enrichment plots showing the mTORC1 response for the same cells and drug treatments as in A. (C) RPPA analysis of the BRAF-mutant human melanoma cell lines 1205LuTR-NR2F1 after 72 hours of treatment with BRAFi + MEKi + DOX. Results indicate the median-centered, log2-transformed group average RPPA expression data for targets with at least 25% change when comparing NR2F1 expression with no DOX after 72 hours of treatment. (D) Representative Western blot of the RPPA-identified proteins in C after treatment with BRAFi + MEKi for 72 hours (n = 2–3). Quantification of the band densitometry (phosphorylated proteins vs. total protein [e.g., RB (pS780)/RB] normalized to no BRAFi + MEKi + no-DOX conditions] is displayed under each band. Note that the NR2F1 blot is the same here and in Supplemental Figure 4A. (E) Heatmap showing hierarchical clustering of commonly enriched genes in mTORC1 GSEA results for all 3 cell line comparisons of BRAFi + MEKi + DOX. (F) MACS2 fold enrichment values are displayed for genes with NR2F1 ChIP-Seq binding in their promoter region (n = 17; each ChIP-Seq experiment is represented by a dot) from the genes in E.

Given the role NR2F1 in dormant and senescent cancer cells (27, 28), we further queried these Kim et al. RNA-Seq data with dormancy “down” and “up” gene signatures (43). We found no consistently significant changes in dormancy markers (other than NR2F1) across any of the cell lines comparing BRAFi + MEKi treatment with or without NR2F1 overexpression (Supplemental Figure 3, F–H). For example, ADAM10 from the “up” signature was significantly decreased in 1205Lu and WM793 cells, but was increased in A375 cells. In the dormancy “down” signature, CDKN3, CKS2, BUB1, FOXM1, BUB1B, and TK1 were all significantly increased in 1205Lu and WM793 cells, but either were either up- or downregulated (not significantly) in A375 cells. We further probed additional dormancy markers in the data from Fane et al. (9), again finding significant changes in only 2 of 3 cell lines (for example, increased DKK1 in A375 and WM793 cells) (Supplemental Figure 3G). Next, to address the possible role of NR2F1 overexpression in senescence, we queried these RNA-Seq data with the REACTOME Cellular Senescence gene pathway, finding that only IL-6 was significantly increased in all cell lines (Supplemental Figure 3H). Together, these data suggest that NR2F1 overexpression affects multiple components of dormancy, but not always consistently between cell lines, in melanoma treated with BRAFi + MEKi.

As we observed increased S-phase entry of NR2F1-overexpressing cells during BRAFi + MEKi therapy (Figure 2C), we probed the RNA-Seq data for the Kyoto Encyclopedia of Genes and Genomes (KEGG) cell-cycle gene signature in all cell lines. GSEA analysis of the KEGG cell-cycle pathway displayed a positive enrichment of cell-cycle genes with NR2F1 overexpression in BRAFi + MEKi–treated cells (Supplemental Figure 3I). We probed for genes that were altered during NR2F1 overexpression in BRAFi + MEKi–treated cells and found that numerous genes were markedly upregulated in 2 of the 3 cells lines. For example, CDKN1A was only substantially increased in 1205Lu and WM793 cells, whereas MCM genes, CDC genes, ORC genes, BUB genes, and others were only substantially increased in 1205Lu and A375 cells (Supplemental Figure 3J). Thus, NR2F1 overexpression has a restorative effect on multiple components of the cell cycle during BRAFi + MEKi.

In parallel to RNA-Seq analyses, we tested the effect of NR2F1 on survival-related signaling pathways by reverse-phase protein array (RPPA) analysis. Consistent with the RNA-Seq and GSEA results (Figure 4, A and B, and Supplemental Figure 3), NR2F1-overexpressing melanoma cells treated with BRAFi + MEKi displayed high mTORC1 pathway readout (phosphorylated S6 [p-S6]) and higher expression of cell-cycle and survival regulators such as cyclin B1, PLK1, p-RB1, p-S6, and Wee1 than did BRAFi + MEKi–treated, non-NR2F1-overexpressing cells (Figure 4C). Altered levels of RPPA-identified proteins involved in the cell cycle and survival were validated by Western blotting and subsequent quantification (Figure 4D and Supplemental Figure 4A) (43). Overall, our data indicate that NR2F1 maintained cell growth in drug-tolerant melanoma subjected to targeted therapy, in part, through enhancement of the level of mTORC1 signaling during BRAFi + MEKi treatment.

To further assess NR2F1 modulation of mTORC1 signaling, we queried the RNA-Seq data and identified commonly enriched genes in the Hallmark mTORC1 gene set for 3 independent cell lines overexpressing NR2F1 and treated with BRAFi + MEKi. We found 21 significantly overlapping genes (Supplemental Figure 4B) involved in mTORC1 pathways, such as the PDK1 and HSPD1 genes, that were upregulated in the presence of NR2F1 overexpression during BRAFi + MEKi treatment (Figure 4E). We then queried publicly available MACS2 (44) NR2F1 ChIP-Seq binding data from the Gene Transcription Regulation Database (45) and found that NR2F1 bound to the promoter region of 17 of the 21 genes identified including PDK1 and HSPD1 (Figure 4F). Together, these data suggest that mTORC1 signaling is partially rescued after NR2F1 upregulation during BRAFi + MEKi treatment, likely occurring by NR2F1 binding to the promoter region of genes that promote mTORC1 signaling and inducing their transcription.

Targeting mTORC1 decreases MRD during BRAFi + MEKi treatment. A previous study has shown that inhibition of mTORC1 signaling delays the onset of acquired resistance to BRAFi and/or MEKi (46). On the basis of the partial maintenance of mTORC1 signaling, we tested the efficacy of 2 mTOR inhibitors (AZD2014 and rapamycin) to target NR2F1-overexpressing cells (47). In NR2F1-induced cells, mTORi alone did not affect growth; however, a triple combination using BRAFi + MEKi + mTORi prevented colony formation (Figure 5A). Real-time monitoring assays of cell growth over 3 weeks indicated that NR2F1-overexpressing cells responded better to the triple combination treatment (Figure 5B), suggesting a potential therapeutic strategy to target DTP melanoma cells. Western blot analysis showed that mTORi inhibited p-S6 levels in NR2F1-overexpressing cells on BRAFi + MEKi (Supplemental Figure 5A), further supporting our proposed model, whereby DTP cells were associated with NR2F1 overexpression and persistent mTOR signaling.

Rapamycin targets NR2F1-overexpressing drug-tolerant cells and delays tumor growth. (A) Colony assay for cell lines overexpressing DOX-inducible NR2F1 after 1 week of treatment with BRAFi + MEKi + mTORC1i (either 1 μmol/L AZD2014 or 1 μmol/L rapamycin). Original magnification, ×20. (B) IncuCyte live-cell analysis for DOX-inducible cells overexpressing NR2F1 after 4 weeks of treatment using BRAFi + MEKi + DOX. Cell growth was analyzed for percentage of cell confluence on the plate (representative of three independent experiments). (C) Images show the expression of NR2F1 in the MRD state following BRAFi + MEKi + rapamycin treatment (4 mg/kg) in vivo. Results are for mice bearing 1205Lu-tdTomato–labeled xenografts following continuous BRAFi + MEKi chow (200 ppm PLX4720 + 7 ppm PD0325901) for 3 weeks and then either control chow + vehicle (control), control chow + 4 mg/kg rapamycin (rapamycin), BRAFi + MEKi chow + vehicle (BRAFi + MEKi), or BRAFi + MEKi chow + rapamycin (BRAFi + MEKi + rapamycin) for 1 week. Plot on the right shows the mean nuclear intensity of NR2F1 protein expression in MRD tumor xenografts by immunofluorescence compared with the control. Statistical significance was determined by 2-way ANOVA. (D) Tumor volume results for mice bearing 1205Lu-tdTomato–labeled xenografts following continuous treatment with BRAFi + MEKi chow (200 ppm PLX4720 + 7 ppm PD0325901) until the tumors entered a state of MRD for several weeks (day 52 after BRAFi + MEKi), and were then given either continuous BRAFi + MEKi chow + vehicle (BRAFi + MEKi) or continuous BRAFi + MEKi chow + 4 mg/kg rapamycin twice per week (BRAFi + MEKi + rapamycin) for the duration of the experiment (treatment start indicated with dotted line on x axis). Tumor growth of the treated mice is shown. An “X” on a tracing denotes an animal that was euthanized for nonexperimental reasons. Data indicate the mean ± SD.

We tested the effect of mTORi on residual tumors following BRAFi + MEKi in vivo. Despite using an intermittent regimen of 2 days on/5 days off AZD2014 administered with continuous BRAFi + MEKi, we had to interrupt our initial in vivo experiments because of symptoms of drug toxicity in the mice, including lethargic behavior and 20% or more body weight loss in triple combination–treated animals (Supplemental Figure 5B). To manage toxicity while retaining efficacy, we tested BRAFi + MEKi + rapamycin combination treatment for 1 week in MRD tumors. Rapamycin (sirolimus) has been shown to be well tolerated in xenograft models and in patients in clinical trials (48, 49). We found that NR2F1 levels were downregulated in MRD xenografts following BRAFi + MEKi + rapamycin compared with BRAFi + MEKi–residual tumors as detected by immunofluorescence (Figure 5C and Supplemental Figure 5C).

Given these promising results with rapamycin, we tested the effects of BRAFi + MEKi + rapamycin versus BRAFi + MEKi by treating mice with rapamycin twice a week to avoid potential toxicity. Tumors were initially treated with BRAFi + MEKi. Upon reaching MRD when tumors expressed high NR2F1 levels (Figure 1E and Figure 5C) and tumor size was stable for several weeks, rapamycin was given in combination with BRAFi + MEKi. Rapamycin delayed tumor regrowth during BRAFi + MEKi, but did not significantly affect mouse weight compared with BRAFi + MEKi alone (Figure 5D, Tables 1 and 2, and Supplemental Figure 5D). Specifically, for mice with tumors that regrew to 100 mm3, the exponential regrowth rates in each treatment group and the difference in the regrowth rates between BRAFi + MEKi + rapamycin versus BRAFi + MEKi groups were estimated (Tables 1 and 2). Tumors in BRAFi + MEKi + rapamycin group grew significantly slower (P = 0.0005) with a 3% daily increase (95% CI: 2.1%–3.9%) versus a 5.2% daily increase (95% CI: 4.4%–6.1%) in the BRAFi + MEKi group. Altogether, these data suggest that inhibiting mTORC1 with rapamycin treatment improves targeted inhibitor therapy by targeting DTP melanomas that overexpress NR2F1 and decreases regrowth rates of MRD.

Estimated mean exponential growth rate after tumor onset