A novel immunogenic mouse model of melanoma for the preclinical assessment of combination targeted and immune-based therapy

Both targeted therapy and immunotherapy have been used successfully to treat melanoma, but the development of resistance and poor response rates to the individual therapies has limited their success. Designing rational combinations of targeted therapy and immunotherapy may overcome these obstacles, but requires assessment in preclinical models with the capacity to respond to both therapeutic classes. Herein, we describe the development and characterization of a novel, immunogenic variant of the BrafV600ECdkn2a−/−Pten−/− YUMM1.1 tumor model that expresses the immunogen, ovalbumin (YOVAL1.1). We demonstrate that, unlike parental tumors, YOVAL1.1 tumors are immunogenic in vivo and can be controlled by immunotherapy. Importantly, YOVAL1.1 tumors are sensitive to targeted inhibitors of BRAFV600E and MEK, responding in a manner consistent with human BRAFV600E melanoma. The YOVAL1.1 melanoma model is transplantable, immunogenic and sensitive to clinical therapies, making it a valuable platform to guide strategic development of combined targeted therapy and immunotherapy approaches in BRAFV600E melanoma.

The development of targeted therapies and immunotherapies in recent years has revolutionized the landscape of cancer treatment, particularly melanoma. The most notable clinical successes in melanoma include immune checkpoint inhibitors of PD-1 and CTLA-4 [1][2][3][4][5][6][7][8] , and targeted inhibitors of the MAPK/ERK pathway; specifically dual inhibition of BRAF V600E and MEK [9][10][11][12][13][14][15] . However, resistance to targeted therapies and low response rates to immunotherapies have prompted great interest in combining these therapeutic strategies. While combination therapies are now being evaluated in clinical trials, most are performed on the basis of observed clinical success of individual therapies, with limited understanding of how these therapeutic classes interact with one another. As such, little judgement can be made about optimal combinations and scheduling, or which patients to target with various combinations. Emerging evidence suggests that therapies targeting the MAPK/ERK pathway may also impact on anti-tumor immune responses [16][17][18] , and hence a thorough understanding of these interactions is paramount for the strategic design of efficacious targeted and immune therapy combinations.
The Yale University Mouse Melanoma (YUMM) series of cell lines can be efficiently grown and studied in immunocompetent C57BL/6 mice, and importantly, have been derived from genetically modified mice bearing mutations commonly found in human melanoma 19 . These models provide an immunocompetent and clinically relevant setting in which to study targeted and immune therapy combinations. However, as these lines were generated through the introduction of a small number of oncogenic driver mutations, they are poorly T Scientific RepoRts | (2019) 9:1225 | https://doi.org/10.1038/s41598-018-37883-y cell immunogenic due to a low somatic mutational burden [20][21][22] ; a major challenge for mouse models genetically engineered in this way 23,24 . Melanoma, in particular, is a highly mutated and immunogenic cancer 25 , expressing numerous neoantigens that have the capacity to stimulate strong immune responses [26][27][28] . The remarkable success of immunotherapies in the treatment of melanoma, in contrast to other solid cancers, is due in part to high inherent immunogenicity and acquired immunosuppressive mechanisms 29 . Hence, weakly immunogenic mouse models do not capture the full characteristics of human melanoma. The YUMM1.1 line, derived from mice bearing a BRAF V600E mutation and deficient for Cdkn2a and Pten, is poorly immunogenic due to low neoantigen expression, and resistant to immunotherapy due to low inflammatory and chemotaxis gene signatures [20][21][22] . In the present study we show that expression of ovalbumin (OVA) was sufficient to alter the susceptibility of YUMM1.1 tumors to host T cell mediated control. The adoptive transfer of OVA-specific CD8 + T cells (OT-I T cells), as well as immune checkpoint blockade therapy, further enhanced tumor control. Checkpoint inhibitors were ineffective against the parental YUMM1.1 model, indicating the expression of OVA, and enhanced T cell engagement, sensitizes this model to immunotherapy. Importantly, the response of this tumor line to standard-of-care BRAF and/or MEK inhibition was equivalent to that observed in human BRAF V600E melanoma, consistent with the parental YUMM1.1 line 21 . Collectively, our data highlights the utility of YOVAL1.1 as a preclinical model for examining the complex interactions of targeted therapies and the immune system, providing a valuable platform to better guide clinical application of novel and existing therapy combinations in BRAF V600E melanoma.

Results
Expression of the immunogen, ovalbumin, in YUMM1.1 tumor cells promotes T cell-mediated tumor control. The YUMM series of mouse melanoma cell lines are reported to be poorly T cell immunogenic in vivo due to low neoantigen expression [20][21][22] . Consistent with this, we found no significant difference in the growth kinetics or overall survival of YUMM1.1 tumors grown in immunocompetent C57BL/6 or immunodeficient NOD scid gamma (NSG) mice; which are T and B cell deficient and lack functional NK cells due to a null mutation in the IL-2 receptor common gamma chain (Fig. 1a). While these tumors induced the recruitment of IFNγ−producing NK cells ( Supplementary Fig. 1a,b), this was not sufficient to control tumor growth. This was despite the fact that in vitro, NK cells could kill YUMM1.1 tumor cells and secrete IFNγ, which up-regulated MHC-I on the tumor cells ( Supplementary Fig. 1c). Furthermore, while YUMM1.1 tumors express MHC I in vivo ( Supplementary Fig. 1d) we speculate that, in the absence of sufficient neo-antigen expression on YUMM1.1 tumor cells, an anti-tumor T cell response was limited.
Human melanoma is inherently immunogenic due to a high neoantigen load 25 . To establish a model that would mimic the immunogenicity of melanoma, the YUMM1.1 line was retrovirally transduced to stably express OVA and GFP, and sorted for both low and high GFP expression (Fig. 1b). Both YUMM1.1 cells transduced with OVA, or an empty vector control (YV1.1), were resistant to OVA-specific OT-I T cell-mediated killing (Fig. 1c). However, pre-treatment of OVA-transduced YUMM1.1 tumor cells with IFNγ to induce H-2K b expression and presentation of the OVA peptide SIINFEKL, sensitized them to OT-I T cell killing (Fig. 1c). OVA stimulates a strong CD8 + T cell response and when expressed at high levels on tumor cells, can prevent successful engraftment of tumors in C57BL/6 mice due to immune-mediated rejection 30 . Thus, we utilized the low OVA-expressing population for our in vivo studies, referred to here as YOVAL1.1 (YUMM1.1-OVA-Low).
We first examined the growth kinetics of the YOVAL1.1 tumor line in both C57BL/6 and NSG mice. Compared to that observed in NSG mice, growth of the YOVAL1.1 tumors was significantly slower in C57BL/6 mice, with a difference in median overall survival of 12 days (40 days versus 28 days; Fig. 1d). Notably, growth of the YV1.1 empty vector control tumors in C57BL/6 mice was comparable to YOVAL1.1 tumors grown in NSG mice (Fig. 1d). Furthermore, in Rag1 −/− mice, which have a functional innate immune system but lack T and B cells, the YOVAL1.1 tumors grew out in a similar manner to that observed in NSG mice (Fig. 1e). YOVAL1.1 tumor growth in C57BL/6 mice was also significantly delayed following the adoptive transfer of OVA-specific OT-I T cells (Fig. 1f). Collectively these data support a role for T cells in mediating the control of YOVAL1.1 tumor growth in vivo.

Expression of the immunogen, ovalbumin, in YUMM1.1 tumor cells alters the tumor immune microenvironment.
To determine the impact of OVA expression on the tumor microenvironment, we compared the immune infiltrate in parental (YUMM1.1), empty vector control (YV1.1) and OVA-expressing (YOVAL1.1) tumors 4 weeks following implant. We observed a significant increase in the frequency of major immune subsets, including CD8 + T cells, CD4 + T cells, T regulatory cells, NK cells, dendritic cells and macrophages, within YOVAL1.1 tumors compared to control YUMM1.1 and YV1.1 tumors ( Fig. 2a and Supplementary Figs 2-4). Indeed, immunohistochemical analysis of YOVAL1.1 tumors revealed markedly higher levels of infiltrating CD3 + T cells compared to YV1.1 tumors (Fig. 2b). Together these data indicate that YOVAL1.1 tumors can stimulate strong CD8 + T cell activity, which appears to contribute to immune-mediated tumor growth control in C57BL/6 mice. However, the induction of an anti-tumor T cell response was insufficient to cause complete tumor rejection, which may in part have been attributed to the observed increases in tumour associated T regulatory cell and/or macrophage frequency (Fig. 2a). Notably, expression of PD-1 and PD-L1 on the CD8 + TILs and YOVAL1.1 tumor cells, respectively, was also detected (Fig. 2c).
To confirm that YOVAL1.1 tumors did not escape immune control as a result of acquired resistance to T cell killing, we harvested tumors at endpoint (>1200 mm 3 ) and found they were sensitive to killing by in vitro activated OT-I T cells (Fig. 2d). Collectively these observations suggest that therapies aimed at overcoming these immunosuppressive mechanisms, such as checkpoint blockade, may be effective in this model.   Analysis of this combination strategy in C57BL/6 mice bearing parental YUMM1.1 tumors revealed that this model is poorly responsive, with no significant improvement in median survival relative to isotype control treated mice (Fig. 3a). However, we hypothesized that the enhanced capacity of YOVAL1.1 tumors to trigger an endogenous T cell response would correlate with improved response to immunotherapy.
To directly compare the two models, which grow with different kinetics in vivo, we inoculated mice with 2 × 10 6 YOVAL1.1 cells or 1.5 × 10 6 YUMM1.1 cells to establish tumors of the same size at day 15 post inoculation; the time at which treatment was started ( Supplementary Fig. 5). In contrast to YUMM1.1 tumors, there was a significant extension of survival of YOVAL1.1-bearing C57BL/6 mice co-treated with anti-PD-1 and anti-CTLA-4 therapy compared to isotype controls, with median survival of 57 days versus 40 days post tumor inoculation, respectively (Fig. 3b). Significantly, 10/12 mice treated with the combination therapy were still alive 50 days post tumor injection, compared to 0/12 mice in the control group. Notably, the YOVAL1.1 tumors were resistant to anti-PD-1 monotherapy (Fig. 3c), but responsive to anti-CTLA-4 monotherapy (Fig. 3d), indicating that the response to the combination therapy was predominately anti-CTLA-4 driven. The immunogenicity of YOVAL1.1 tumors, and the capacity of these tumors to respond to immunotherapy in vivo, makes this a valuable model to dissect how these immunotherapy approaches may be enhanced by target therapies. The response of YOVAL1.1 tumors to MAPK/ERK pathway-targeted therapy recapitulates that of human models. The YUMM series of cell lines were developed through the introduction of common mutations observed in human melanoma 21 . Specifically, YUMM1.1 has a BRAF V600E mutation and is sensitive to treatment with BRAF inhibitors (BRAFi) in vitro 21 . However, the response of YUMM1.1 to BRAFi, and standard-of-care therapy BRAFi plus MEKi, has not been reported in vivo, nor has the sensitivity to these inhibitors been compared to other common melanoma preclinical models. Hence, to determine the suitability of YOVAL1.1 as a model for studying targeted therapy, we examined its response to inhibitors of BRAF V600E and MEK. As a measure of sensitivity, we determined drug doses of PLX4720 (BRAF V600E inhibitor) and cobimetinib (MEK inhibitor) required for 50% growth inhibition (GI50). The PLX4720 and cobimetinib GI50s of YOVAL1.1 were not significantly different to those of the human BRAF V600E melanoma line, A375 (239 ± 50 nM and 4.6 ± 0.7 nM vs. 139 ± 9 nM and 4.7 ± 0.5 nM, respectively) (Fig. 4a). In contrast, B16 cells, which lack a clinically relevant genetic background, were not sensitive to these inhibitors (PLX4720 and cobimetinib GI50s 7,412 ± 675 nM and 68.3 ± 10.8 nM, respectively) (Fig. 4a).
In melanoma patients and in preclinical models using human xenografts, combined inhibition of BRAF and MEK achieves synergistic anti-cancer responses 9,[12][13][14] . Importantly, synergy between PLX4720 and cobimetinib was also observed in the YOVAL1.1 model. In vitro the combination synergistically halted the proliferative activity of the YOVAL1.1 line, whereas no such synergy was observed in the B16 line (Fig. 4b). PLX4720, cobimetinib and their combination decreased P-ERK levels in YOVAL1.1 and A375 cells, but not in the non-sensitive B16 line (Fig. 4c). In vivo, the combination of BRAF and MEK inhibition significantly improved survival of YOVAL1.1-bearing C57BL/6 mice, with a median overall survival of 40 days on treatment, compared to 12 days for vehicle-treated mice (Fig. 4d), which was similar to the response of the parental YUMM1.1 line (Supplementary Fig. 6). Taken together, these data demonstrate that YOVAL1.1 tumors respond to MAPK/ERK pathway inhibition with similar sensitivity to that of human preclinical models. This highlights the utility of YOVAL1.1 tumors as a clinically relevant in vivo model for studying responses to these targeted therapies in combination with immune-based approaches.

Discussion
Immunotherapy and targeted therapy have both been immensely successful in extending the life of melanoma patients. However, the majority of patients treated with targeted therapy eventually relapse and approximately 50-70% of patients treated with immune checkpoint therapy do not respond 6,[10][11][12][13]15 . In addition to inhibiting tumor intrinsic growth pathways, it is now well known that targeted therapies also impact anti-cancer immune responses [16][17][18] . Understanding these interactions is paramount in the strategic design of immune and targeted therapy combinations and this requires physiologically relevant, preclinical models that are both immunogenic, and responsive to standard-of-care therapies. Here, we describe YOVAL1.1 as a novel mouse model of melanoma that is suitable for evaluating in vivo immune interactions in response to targeted therapy.
Chicken ovalbumin (OVA) is widely used as a model antigen in T cell biology. Numerous mouse cancer models, including the commonly used B16 melanoma cell line, have been modified to express OVA to aid in enhancing and tracking tumor-specific T cell responses 31,32 ; however, these models lack the genetic background commonly found in human melanoma. In contrast, the recently developed, YUMM series of cell lines carry the relevant genetic background and are fast becoming the preferred syngeneic model of melanoma [20][21][22]33 . In this study, we have introduced OVA into YUMM1.1 cells, to enhance in vivo immune interactions and thus have created a melanoma OVA model antigen system that is more clinically applicable.
Low immunogenicity is a known major challenge for genetically engineered mouse models 23,24 . Indeed, we found YUMM1.1 to be poorly immunogenic in vivo. The introduction of OVA to generate the YOVAL1.1 cell line was sufficient to sensitize the line to endogenous T cell control, but did not cause complete tumor rejection, supporting the presence of immunosuppressive mechanisms in this model. Consistent with this, we found an abundance of regulatory T cells within these tumors, in addition to expression of the immunosuppressive checkpoint molecules, PD-1 and PD-L1, on the T cells and tumor cells respectively. The loss of PTEN in this model is also a likely contributor to such strong immunosuppression, as PTEN loss in melanoma is associated with increased production of immunosuppressive cytokines and resistance to T cell-mediated immunotherapies [34][35][36] . Interestingly, the enhanced immunogenicity of YOVAL1.1 rendered the model amenable to checkpoint blockade with anti-CTLA-4, but not anti-PD-1, despite PD-1 and PD-L1 being expressed in the microenvironment, and anti-PD-1 demonstrating superior results to anti-CTLA-4 in the clinic 37 . This observed resistance to PD-1 blockade therapy is comparable to observations reported previously in a YUMM model with the same genetic background 22 . Recently, the combination of anti-CTLA-4 and anti-PD-1 was shown to be superior to anti-PD-1 monotherapy 6 , and it is currently unclear whether this added benefit is due to complementary actions of the inhibitors, or a subset of anti-PD-1-resistant patients who are responsive to anti-CTLA-4. Our data suggests the latter is possible, given that this model responds equally well to anti-CTLA-4 with, and without, the addition of anti-PD-1. The primary anti-tumor mechanism of CTLA-4 checkpoint blockade remains controversial. In addition to enhancing T cell priming through blockade of inhibitory interactions between antigen presenting cells and T cells 38 , anti-CLTA-4 therapy may also deplete T regulatory FOXP3 + cells in the tumor microenvironment [39][40][41] . Importantly, this model provides a unique platform to dissect these mechanisms, which may provide insight into which patients are most likely to respond to anti-CTLA-4 therapy alone. Conversely, the innate resistance of the model to anti-PD-1 therapy may offer insight into mechanisms contributing to such resistance. Given the high toxicity 42 and significant monetary costs 43 of combined immune checkpoint therapies, there is great value in stratifying patients who will receive benefit from single agents or novel combination approaches.
Inhibition of BRAF and MEK is standard-of-care targeted therapy for BRAF V600E melanoma, and the clinical response rate of BRAF V600E melanoma to combined BRAF/MEK inhibition is around 70% 44  now substantial evidence that these inhibitors alter anti-tumor immune responses [16][17][18] , potentially impacting on the efficacy of immunotherapy. To comprehensively evaluate the effects of these inhibitors on the immune system using preclinical models, it is essential that the model system is sensitive to these therapies. The response of YOVAL1.1 to BRAF and MEK inhibition recapitulates that of human BRAF V600E A375 melanoma, demonstrating its suitability over the commonly used B16 syngeneic melanoma model. Importantly, these drugs demonstrated target specificity in YOVAL1.1, as evidenced by a reduction in P-ERK, and markedly decreased tumor progression in vivo. The expression of OVA in this model make it an ideal platform to evaluate the effects of these inhibitors on anti-tumor endogenous T cell responses, as well as adoptively transferred OVA-specific T cells.
In addition to being immunogenic and sensitive to clinical therapies, YOVAL1.1 tumors are transplantable, making the model a simple and effective tool to trial a range of novel and existing therapy combinations and scheduling. We therefore propose that the YOVAL1.1 melanoma model will provide a valuable platform to guide strategic development of combined targeted therapy and immunotherapy approaches in BRAF V600E melanoma. blot of phospho-and total-ERK (P-ERK and T-ERK) expression in YOVAL1.1, A375 and B16 treated with DMSO, 1 μM PLX4720 (BRAFi), 10 nM cobimetinib (MEKi) or BRAFi + MEKi for 48 hours, representative blot of n = 3. All images were cropped from different parts of the same blot, with exposure times differing according to antibody and cell line (see Supplementary Fig. 7). (d) Tumor growth and survival in YOVAL1.1bearing C57BL/6 mice treated daily, 6 days/week, with dabrafenib (BRAFi) plus trametinib (MEKi). Survival is measured as time for tumors to reach >1200 mm 3 , log-rank (Mantel-Cox) test, n = 7. All error bars show ±SEM. ***p < 0.001, ****p < 0.0001. of anti-PD-1 or isotype was administered once per week for 3 weeks starting 11 days post tumor inoculation. For OT-I T cell transfer, mice were given 4 Gy total body irradiation (using X-RAD iR-160; Precision X-Ray, North Branford, CT) on day 16 post tumor inoculation, followed by intravenous administration of 1 × 10 7 primary OT-I T cells. Mice were given 50,000 IU IL-2 by intraperitoneal injection daily for 5 days post OT-I T cell transfer.
statistical analysis. One-way analysis of variance (ANOVA) with Tukey's multiple comparisons tests and unpaired t-tests were performed using GraphPad PRISM. Kaplan-Meier survival was compared using a log-rank (Mantel-Cox) test. All experiments were performed in at least three biological replicates. Error bars show ±SEM.

Data Availability
All data generated or analysed during this study are included in this published article (and its Supplementary Information files).