Facts and Hopes in Immunotherapy of Soft-Tissue Sarcomas

Immunotherapy for sarcoma has its origins in 1891 with William B. Coley injecting mixed toxins of the Streptococcus erysipelas and the Bacillus prodigiosus into patients with sarcoma (1). These early experiments demonstrated that immunotherapy was a possibility for the treatment of sarcomas. The early promise of immunotherapy was then placed on hold for over a century (2). In the interim, a revolution in our understanding of the genetics of soft-tissue sarcoma (STS) occurred that demonstrated the diversity and complexity of different types of STS (3, 4). It is the same genetic complexity that has complicated a one-size-fits-all immunotherapy approach for STS. In this review, we will discuss the current advances in our understanding of sarcoma immunobiology and the effects that immunotherapy has had in clinical trials. From the facts of where we currently stand, to the hope that exists for immunotherapy in STS, we are still very much at the beginning of the beginning.

The immune microenvironment in STS substantially differs from other tumors where immunomodulation effectively functions, such as in the case of melanoma. When compared with immune responsive tumors, STS demonstrated a median tumor mutational burden (TMB) of 2 mutations per DNA megabase (Mb; ref. 5), CD8⁺ expression was 23% ± 13% (6), and the PD-L1 expression was 6.6% in the largest series (7). In contrast, the expression in melanoma, a paradigmatic immune-sensitive tumor, was 14 mutations per DNA Mb in TMB, 42% ± 23% for CD8⁺ lymphocytes (8), and 35% for PD-L1 (9).

Sarcoma genomic heterogeneity complicates these statistics. The highest range of TMB expression is found in cutaneous angiosarcoma (10), undifferentiated pleomorphic sarcoma (UPS), leiomyosarcoma, sarcoma not otherwise specified, and myxofibrosarcoma with median TMB ranging from 2.2 to 401.4 mutations per DNA Mb. On the opposite end of the spectrum, synovial sarcoma, myxoid liposarcoma, solitary fibrous tumor, or alveolar rhabdomyosarcoma exhibit lower TMB, with a median of 1.7 and maximum TMB ranging from 7.5 to 28.4 mutations per DNA Mb (5).

Mismatch repair (MMR) deficiency occurs by hypermutation of MMR genes, germline MMR pathway mutation, or double somatic mutation in MMR genes and this is termed microsatellite instability (MSI). MSI-high phenotypes have the highest response probability to PD-1 inhibitors (11). Sarcomas are rarely MSI-high (0.78%; ref. 12), although patients should still be tested for this status.

Studies characterizing infiltrates of immune cells in sarcoma show a low number of tumor-infiltrating lymphocytes (TIL) compared with melanoma. The mean and SD of number of cells per gram of weight were as follows: 72 ± 15 for CD3⁺ and 42 ± 23 for CD8⁺ in melanoma, while they were 35 ± 9 for CD3⁺ and 23 ± 13 for CD8⁺ in sarcoma, respectively (6). In a meta-analysis carried out to examine the prognostic role of high TIL population in different cancers, including 52 studies and 12,447 patients, the authors reported a positive effect on prognosis for high CD3⁺ and CD8⁺ TILs. The CD8/Foxp3 ratio was the best prognostic indicator of risk of death with HR of 0.48 [95% confidence interval (CI), 0.34–0.68; ref. 13]. In contrast, in series including several STS subtypes, the prognostic role of CD3⁺ or CD8⁺ seems less prominent. In a series with 249 patients with STS, 83% of them localized, high number of CD20⁺ lymphocytes in tumor significantly correlated with a longer disease-specific survival. In multivariate analyses, high number of CD20⁺ lymphocytes was the only independent prognostic factor among TIL for disease-free survival (14).

The prognostic role of CD8⁺ is unclear, as in a series of 163 STS (81% localized at diagnosis) high CD8⁺ lymphocytes (cutoff 137 cell/mm²) significantly correlated with poor disease-free survival (P = 0.031) and overall survival (OS, P < 0.001) in the univariate analysis (15). This inconsistent outcome could be related to the high heterogeneity inherent to the inclusion of several STS subtypes in this study. Nevertheless, a few studies have focused on the prognostic impact of TIL in specific histotypes, however, with inconsistent results. Contradictory findings were seen, for example, in synovial sarcoma: while in one study with 36 patients with synovial sarcoma with high CD8⁺ or Foxp3 lymphocytes correlated with better prognosis (16), another study including 22 patients with synovial sarcoma showed worse prognosis for patients with high CD8⁺ lymphocytes (17). Additional series exploring the prognostic role of TIL in malignant peripheral nerve sheath tumors (MPNST; ref. 18) and cutaneous angiosarcoma (19) also showed divergent prognostic outcome. Besides, the largest series exploring the prognostic impact of TIL in sarcoma included 809 samples of STS and gastrointestinal stromal tumor (GIST), did not find any prognostic correlation among the translocation-associated sarcomas. In contrast, the authors reported a significant better prognosis for OS (P = 0.02) and progression-free survival (PFS, P = 0.01), with increasing lymphocyte infiltration among the nontranslocation-associated sarcomas. In Addition, the OS was significantly worse with increasing CD56⁺ (P = 0.03) or PD-1⁺ (P = 0.05) TILs (20). Of note, patients with positive immunostaining for both CD8 and FOXP3 had better OS compared with those negative for FOXP3 in dedifferentiated liposarcoma (P = 0.002) or MPNST (P = 0.002). In myxoid liposarcoma this correlation was inverse (P < 0.001; ref. 20).

Lymphocytic infiltrates have also been explored by gene expression. Sarcomas with a complex genome expressed high levels of genes related to T-cell infiltration and antigen presentation. For instance, CD3 and IL7 receptor (CD127) expression was significantly higher in leiomyosarcoma and UPS than in translocation-related sarcomas. A trend toward higher expression in nontranslocation-related sarcoma was also noted for IDO, CD4, CD27, and CCR5 (21). In addition, T-cell receptor (TCR) clonality has been found to be correlated to PD-1 (P = 0.007) and PD-L1 (P = 0.003) expression (21). Immune signatures characterizing the type of immune infiltration in different sarcomas were analyzed on the basis of The Cancer Genome Atlas (TCGA) mRNA data using 203 genes involved in immune response. Thus, UPS and myxofibrosarcoma had the highest median macrophages scores, dedifferentiated liposarcoma had highest CD8⁺ scores, and somatic leiomyosarcoma exhibited the highest PD-L1 score (4). Of note, macrophages, monocyte-derived phagocytic cells, play a crucial role in tumor immunomodulation. Tumor-associated macrophages (TAM) can mediate for anticancer effects or for tumor progression depending on their polarization. In general, M1-polarized macrophages mediates anticancer effects through adaptive immunity mechanisms and M2-polarized macrophages suppress adaptive immunity, favoring tumor progression, tumor angiogenesis, increase extracellular matrix breakdown, and tumor invasion (22). TAMs have been associated with poor survival in myxoid liposarcoma (23), gynecologic (24) and nongynecologic leiomyosarcomas (24, 25), solitary fibrous tumor (26), and UPS (27). Targeting the colony-stimulating factor 1 receptor (CSF1R), a protein that facilitates the differentiation of monocytes into TAMs and promotes their survival within the tumor, with CSF1R inhibitors (28, 29) could be a therapeutic strategy to be considered for these histologic subtypes in future trials. Natural killer cells were the only cell type to correlate significantly with disease-specific survival. Table 1 lists examples of studies that analyzed prognostic impact of immune infiltrates in STS.

Using a transcriptomic analysis of the microenvironment cell population, which measure the expression of eight immune and two stromal cell populations (30), STS can be classified into five different sarcoma immune classes (SIC). Each SIC exhibited a different profile, from A (immune desert), which showed the lowest expression of gene signatures of immune cells and vasculature expression, to E (immune and tertiary lymphoid structures) characterized by the highest expression of genes related to immune cells. In the middle, C (vascularized) was characterized by a high expression of endothelial-related genes. SIC B and D have expressed mixed profiles between A and C or C and E. Of note, grouping sarcomas into these five classes based on different profile expression of tumor microenvironment resulted in prognostic impact. Thus, patients with SIC A showed worse OS than SIC D (P = 0.048) or SIC E (P = 0.025). Furthermore, this genomic immune signature had predictive role in a prospective series treated with pembrolizumab. The overall response rate (ORR) was 50%, 25%, 22%, 0%, and 0% for SIC E, D, C, B, and A, respectively. Patients harboring SIC E had significantly higher ORR with pembrolizumab (P = 0.026). A more detailed analysis revealed a significant correlation of survival with B-cell lineage signature, whereas CD8⁺ signature did not significantly correlate with survival.

Specific immune checkpoint expression as PD-1/PD-L1 axis has not demonstrated convincing prognostic or predictive value in sarcoma. In one study with 105 (74% localized) patients with STS expressing intratumor PD-L1 in 65% of cases, a worse prognosis was observed. The expression of tumoral PD-L1 predicted for shorter OS (HR, 5.69; 95% CI, 2.558–12.700; P < 0.001) and event-free survival (HR, 3.27; 95% CI, 1.776–6.036; P < 0.001; ref. 31). However, in other studies including different sarcoma subtypes, no prognostic correlation could be established (21, 32, 33). Moreover, there were substantial differences for the same study among cases studied by tissue microarray or by whole sections from the block (34). Thus, the predictive value of PD-L1 expression remains uncertain, and caution should be used in clinical practice when utilizing this expression to make therapeutic off-label recommendations.

A genomic approach, addressing the gene expression of PD-L1, could circumvent the constraints of PD-L1 IHC. In one genomic array performed on 758 previously untreated sarcoma samples, 470 of them considered in the prognostic analysis, authors reported a significant correlation with metastasis-free interval (MFS). PD-L1 high expression had 5-year MFS of 61% (95% CI, 50–73), while PD-L1 low expression had 5-year MFS of 72% (95% CI, 63–83; P = 0.0037). This prognostic value has been confirmed in a validation set and PD-L1 expression had an independent prognostic value in the multivariate analysis, HR, 1.51 (95% CI, 1.06–2.16; P = 0.024; ref. 35). In TCGA analysis the highest PD-L1 expression was observed in leiomyosarcoma (4), while in another study, the highest expression was in UPS (21). In addition, a second study found significantly higher mRNA expression of PD-L1 in UPS (36). Table 2 lists examples of studies analyzing prognostic impact of PD-L1 expression in tumor cell in the sarcoma context.

The pioneer study was SARC028 phase II trial with pembrolizumab flat dose at 200 mg every 3 weeks, which was conducted in 12 academic centers in United States, and enrolled patients in two different cohorts, soft tissue and bone tumors. Among STS, the selected histologies chosen on the basis of prevalence were leiomyosarcoma, UPS, synovial sarcoma, and dedifferentiated/well-differentiated liposarcoma. The main endpoint was investigator-assessed objective response by RECIST 1.1, considering 25% of ORR as clinically meaningful and if less than 10% as ineffective. The authors reported objective response in 7 of 40 (18%; 95% CI, 7–33) patients accrued in STS cohort, with a median duration of response of 33 weeks. Responses were seen in four UPS, two dedifferentiated liposarcomas, and one synovial sarcoma. The median of PFS (mPFS) and OS (mOS) for patients with STS was 18 weeks (95% CI, 8–21) and 49 weeks (95% CI, 34–73), respectively. The authors concluded that pembrolizumab was clinically active in patients with UPS and dedifferentiated liposarcoma (37).

Tumor biopsies were required at baseline and after 8 weeks of treatment, which were crucial for gaining insight into the response to PD-1 blockade. Using a multiplex immunofluorescence including the following antibodies: PD-L1, CD3, CD8, PD-1, CD68, granzyme B, Foxp3, and CD45RO, a correlation between response and expression of different immune cells or receptors was analyzed. Higher density of immune cells significantly correlated with response. PD-L1 expression in tumor cells was detected in only two of 40 analyzed cases (5%), and these were the two responding patients diagnosed with UPS. Given the lack of power within this study, the only major conclusion was that more studies were warranted.

In the pharmacodynamics analysis comparing immune infiltrates from baseline and at week 8, the only remarkable changes were detected for two immune cell phenotypes: effector memory cytotoxic T cells (CD3⁺ CD8⁺ CD45RO⁺) that increased from 7.9% to 21.5% and regulatory T cells (CD3⁺ Foxp3⁺) or (CD3⁺ CD8⁺ Foxp3⁺) that increased from 3.7% to 8.3%. More interestingly, higher percentage of regulatory T cells at baseline was correlated with a significant longer mPFS: (40 vs. 8 weeks, P = 0.044), and similarly, higher percentage of cytotoxic T-cell infiltrates at baseline was correlated with a significant longer mPFS (40 vs. 8 weeks, 0.016).

The study, Alliance A091401, a noncomparative randomized phase II trial, randomized 85 patients to received nivolumab versus nivolumab plus ipilimumab in progressing STS patients after at least one previous systemic line. The main endpoint was investigator-assessed confirmed objective response by RECIST 1.1. Confirmed partial responses were observed in 2 of 38 evaluable patients (5%). mPFS was 1.7 months (95% CI, 1.4–4.3) and the estimated 6-month PFS rate was 15% (38). Of note, the two arms of the trial were noncomparator by design, so direct comparison of both arms was not possible.

Because of poor results observed with monotherapy, investigators have explored combinations for a more efficient immunomodulation, trying to convert the sarcoma microenvironment into T-cell–inflamed tumor. Alliance A091401 is the only reported randomized trial testing nivolumab alone against ipilimumab and nivolumab in STS. There were six objective responses of 38 evaluable patients (16%; 92% CI, 7–30) in the combination arm. Responses were seen in UPS (2), leiomyosarcoma (2), angiosarcoma (1), and myxofibrosarcoma (1). The mPFS was 4.1 months (95% CI, 2.6–4.7) and the mOS was 14.3 months (95% CI, 9.6–not reached; ref. 38).

Angiogenesis mediators as VEGFA have a well-known function promoting neoangiogenesis, while preventing immune response (39). A phase II clinical trial examining axitinib 5 mg twice daily with pembrolizumab 200 mg starting on day 8 and then every 3 weeks was investigated in 33 patients with STS. There were 69% of patients with at least two previous lines, 51% that had received previous tyrosine kinase inhibitors, and 36% diagnosed with alveolar soft-part sarcomas (ASPS). The endpoint was 3-month PFS rate of 40%. The 3-month and 6-month PFS rates were 65.6% (95% CI, 46.6–79.3) and 46.9% (95% CI, 29.2–62.8), respectively. Of 32 evaluable patients, 8 (25%) had partial response and 9 (28%) had stable disease. Six responders were ASPS, one was epithelioid sarcoma, and another was leiomyosarcoma. The median duration of response was 29 weeks, while the mOS was 18.7 months. Neither PD-L1 positivity nor high TIL showed statistical correlation with PFS or partial response. Interestingly, angiogenic plasmatic activity at baseline was more likely to respond to this regimen (40). Intriguingly, ASPS seems to be sensitive to immunotherapy-based regimens, even if this histologic subtype does not exhibit an immune-sensitive microenvironment. In fact, it has been reported that ASPS has a TIL ratio lower than nontranslocation-related sarcomas (20) and the TMB is also lower compared with other histologic subtypes, such as synovial sarcoma or Ewing sarcoma (41). In this latter study, a MMR deficiency signature was associated with the activity of anti-PD-L1 in ASPS, however, further studies in larger series of cases are required to validate these observations. Besides, it is important to mention that some genes normally expressed in the context of ASPSCR1-TFE3 seemed to be involved in pathways of immune surveillance, immune regulation by chemokines, and focal adhesion. An example is CCL4, a gene positioned in the TFE3 neighboring cytogenetic band chr17q21 that signals through the receptor CCR5 and regulates, among other functions, macrophage migration (42). This aspect could help understand, in part, the sensitivity of ASPS to anti-PD-1/PD-L1 inhibitors.

A similar approach was used in the IMMUNOSARC trial, a phase I/II trial exploring the combination of sunitinib plus nivolumab in some patients with STS and bone sarcoma. Data from phase II part of STS cohort was recently presented with 50 patients accrued. The recommended scheme derived after phase I part was sunitinib 37.5 mg on a daily basis for the first 15 days, from then on nivolumab was administered at 3 mg/kg every 2 weeks and sunitinib dose was given at 25 mg per day. The main endpoint was 6-month PFS rate and the accrual was limited to UPS, synovial sarcoma, epithelioid sarcoma, angiosarcoma, clear cell sarcoma, extraskeletal myxoid chondrosarcoma, solitary fibrous tumors, and ASPS. The reported 6-month PFS rate was 50%, while the 6-month OS was 77% (median not reached). The response rate following central assessment was 11%, stable disease 61%, and progressive disease 28%. ASPS comprised 6% of patients (43).

The combination of doxorubicin and pembrolizumab was tested in a phase I/II trial exploring the concept of immune death induced by doxorubicin. This later was recommended at 75 mg/m² along with pembrolizumab. The response rate was distributed as partial response 22%, stable disease 59%, and progressive disease 19%. The mPFS was 8.1 months (95% CI, 6.3–10.8) being superior to historic controls considered by the authors, 4.1 months (95% CI, 3.0–6.6; ref. 44).

Table 3 depicts outcomes of trials with anti-PD-1 alone or in combination in advanced STS. The combination of immune checkpoint inhibitors, especially with antiangiogenic agents, seemed to induce a clearly longer PFS in STS second-line treatment, compared with anti-PD-1 alone, or anti-CTL4A alone, or anti-angiogenic alone (45).

Cancer-testes antigens represent a family of antigens that arise from 276 genes (46). The most commonly studied within sarcomas are the New York esophageal tumor antigen (NY-ESO) and the MAGEA4 antigen (MAGE Family Member A4), and more recently, the PRAME (preferentially expressed antigen in melanoma; refs. 47, 48). This review will focus on the most mature clinical data for synovial sarcoma, as first demonstrated that NYESO is expressed in 80% of synovial sarcoma (49). Robbins and colleagues went on to genetically engineer lymphocytes that were reactive to NY-ESO, and in a phase I clinical trial found on an objective tumor response in 4 of 6 patients with synovial sarcoma (49). More recently, in a modified T-cell therapy called SPEAR (specific peptide enhanced affinity receptor) T cells were developed on the basis of HLA:02 status and NY-ESO expression that was published as a phase I trial (50). Of the 42 patients treated, 1 patient had a complete response and 14 had a partial response. The response rate was up to 50% in the cohort with high antigen expression and a conditioning regimen that included 30 mg/m² of fludarabine for 4 days and 1,800 mg/m² of cyclophosphamide for 2 days, with the response requiring a highly dose conditioning regiment. Of note, one case of aplastic anemia was seen with this regimen. This treatment is being further developed as a phase II clinical trial.

Most recently, a similar SPEAR T-cell targeting MAGEA4 in synovial sarcoma was presented at ESMO with an update at the Connective Tissue Oncology meeting (51, 52). Seven of 14 patients demonstrated a partial response, and six of the seven responders had durable responses to week 18 at the time of the data cutoff with most patients still on trial. Thirteen of 14 patients had clinical benefit and the same 13 of 14 had at least grade 1 cytokine release syndrome in response to treatment. The durability of this target and therapy is awaiting and this is formally being tested in a phase II registration trial. Finally, 1 patient on a high-dose chemotherapy expansion cohort with 30 mg/m² of fludarabine for 4 days and 1,800 mg/m² also developed aplastic anemia. What both patients here and above had in common were advanced age and extensive pretreatment. Therefore, clinical development utilizing high-dose conditioning regimens in elderly heavily pretreated patients undergoing SPEAR T-cell therapies should be approached with caution.

Talimogene Laherparepvec (T-VEC) is an oncolytic immunotherapy based on intratumoral injection of a modified self-replicating human herpes virus type 1 that causes tumor lysis and antigen release. In a single-center phase II study, 20 patients were treated with T-VEC and pembrolizumab with the primary endpoint being objective response rate at 24 weeks. The ORR was 30% at 24 weeks, with the responding histologies being cutaneous angiosarcoma, UPS, myxofibrosarcoma, epithelioid sarcoma, and an unclassified sarcoma. There is need for further development of T-VEC in a randomized trial based on these results (53).

T cells expressing MC.7.G5 TCR are able to kill a broad spectrum of tumor cell lines regardless of their HLA allomorph. This novel TCR acts through a protein called MR1 (major histocompatibility complex class I-related gene protein) and not through the largely described MHC, because anti-MR1, but not MHC I or MHC II antibodies blocked target cell recognition by MC.7.G5. Of note, unlike MHC, MR1 sequence seems partially conserved among individuals, opening new doors for the development of novel pan-cancer, pan-population T-cell–mediated cancer immunotherapy approaches (54).

Moreover, other immune checkpoint receptors, yet to be explored in a deeper way in sarcomas, such as LAG-3 or TIM-3, might also play an important role in this field. Therefore, following the recent immune classification of STS microenvironment, mainly those belonging to the SIC E, expressed these immune checkpoint proteins (LAG-3 and TIM-3; ref. 55). The expression of LAG-3 has been correlated with worse OS in STS and its inhibition impaired tumor growth in immunocompetent 3-methylcholanthrene–induced fibrosarcoma mouse models (15). On the other hand, and in these immunocompetent mouse models, anti-TIM-3 antibodies were more effective in combination with anti-CTLA-4 or anti-PD-1 antibodies, in comparison with monotherapy (56).

Bispecific antibodies are emerging as a new class of immunotherapeutic agents (57, 58). At least three classes of bispecific antibodies are currently being developed: (i) cytotoxic effector cell redirectors; (ii) tumor-targeted immunomodulators; and (iii) dual immunomodulators. The cytotoxic effector cell redirectors engage a tumor-associated antigen with the complex CD3 T-cell coreceptor, thus guiding T-cell cytotoxic effect directly toward the tumor cells. As an example, orlotamab is a T-cell engager that targets both B7-H3 and CD3. Noteworthy, high expression of B7-H3 gene (CD276) has been described in STS, more precisely in dedifferentiated liposarcoma, UPS, and myxofibrosarcoma (4). Among the tumor-targeted immunomodulators, special interest should be reserved for tumor-targeting 4-1BB agonist, composed by a trimeric 4-1BB ligand, a Fab moiety targeting stromal fibroblast activation protein (FAP), and a silenced Fc domain that lacks affinity for C1q and FcγRs (59, 60), because FAP seems to be consistently expressed in STS (61, 62). Moreover, simultaneously targeting two immune checkpoints is a promising concept that can be achieved with dual immunomodulators. Several dual immunomodulators are being developed, targeting PD-1 and LAG-3 (e.g., MGD013 or FS118), PD-1 and TIM-3 (e.g., MCLA-134), or PD-1 and CTL-4 (e.g., XmAb20717). There efficacy in STS awaits formal testing.

In summary, we are still at the beginning of the beginning of our understanding of immunotherapy in sarcoma. From 1891 to the present, individual patients have been rendered disease free using immunotherapy approaches. Overall, combination regimens, based on immune checkpoint inhibitors, seemed to be more efficient compared with monotherapy (e.g., anti-PD-1 or anti-CTL4); specially the combination of anti-PD-1 with antiangiogenic agents. Moreover, modified T-cell therapies are currently being tested in specific STS subtypes with a significant clinical benefit for the patients. However, further studies are needed to describe novel antigens in other STS subtypes. Likewise, new therapeutic strategies, such as dual immunomodulators, deserve to be tested in the context of STS. It will take the combined partnership between the international sarcoma centers, patients, and basic scientists to fully understand and optimize the use of immunotherapy for STS.

J. Martín-Broto reports grants, personal fees, and other from PharmaMar (research funding for clinical studies; institutional) and Eisai (research funding for clinical studies; institutional), grants from IMMIX Biopharma, grants and other from Novartis (research funding for clinical studies; institutional), personal fees and other from Eli Lilly (research funding for clinical studies; institutional) and Bayer (research funding for clinical studies; institutional), and other from AROG (research funding for clinical studies; institutional), Lixte (research funding for clinical studies; institutional), Karyopharm (research funding for clinical studies; institutional), Deciphera (research funding for clinical studies; institutional), GlaxoSmithKline (research funding for clinical studies; institutional), Blueprint (research funding for clinical studies; institutional), Nektar (research funding for clinical studies; institutional), Forma (research funding for clinical studies; institutional), Amgen (research funding for clinical studies; institutional), and Daiichi Sankyo (research funding for clinical studies; institutional) outside the submitted work. D.S. Moura reports grants and non-financial support from PharmaMar (institutional grant for preclinical research) and Eisai (institutional grant for preclinical research), grants from Novartis (institutional grant for preclinical research) and Immix Biopharma (institutional grant for preclinical research), and non-financial support from Pfizer, Bayer, and Celgene outside the submitted work. B.A. Van Tine reports personal fees from Lilly (advisory board, speaker bureau), Janssen (advisory board, speaker bureau), Caris (consulting fee/advisory board), Novartis (consultant/adviser/speaker), CytRX (consultant/adviser/speaker), Plexxikon (consultant/adviser/speaker), Epizyme (consultant/adviser/speaker), Daiichi Sankyo (consultant/adviser), Adaptimmune (consultant/adviser/speaker), Bayer (consultant/adviser), Deciphera (consultant), Apexigen (consultant), and Polaris (board member), grants and personal fees from Pfizer (consultant/adviser/research grant) and GlaxoSmithKline (consultant/research grant/speakers bureau/travel expenses), and grants from Merck (research grant) and Tracon (research grant) outside the submitted work. No other potential conflicts of interest were disclosed.

B.A. Van Tine was supported by the NCI (RO1CA227115).