Ubiquitin-Specific Protease 6 Functions as a Tumor Suppressor in Ewing Sarcoma through Immune Activation

Ewing sarcoma is the second most common bone cancer, predominantly affecting the pediatric population (1, 2). Although patients with localized disease have a 5-year survival rate of approximately 75%, metastatic and recurrent patients have a dismal survival of 20%. This low survival rate reflects the inefficacy of standard therapy, which includes radiotherapy, surgery, and chemotherapy. Patient burden is exacerbated by the lack of markers that can reliably predict recurrence. Ewing sarcoma is driven by pathognomonic translocations between the RNA-binding protein EWS and Ets family transcription factors, most commonly FLI1. Agents targeting EWS–FLI1 or its downstream effectors have failed to yield effective treatments in recent clinical trials (3, 4). Furthermore, immunotherapy has had limited success in Ewing sarcoma, in part because the immune tumor microenvironment (TME) is poorly characterized (5).

ES, like other sarcomas, has generally been regarded as immunologically “cold,” with few tumor-infiltrating T lymphocytes (TIL; refs. 5–7). However, recent studies reveal that TILs actually vary significantly, and that high CD8⁺ TIL levels are associated with greatly improved overall survival (8, 9). In one study, survival was approximately 40% at 55 months versus approximately 90% at 100 months for patients with low versus high CD8⁺ TIL levels, respectively (8). Macrophage infiltration and polarization has also been found to affect Ewing sarcoma pathogenesis and progression (7, 10), though the factors contributing to these effects are unknown. Thus, elucidating the mechanisms of TIL recruitment and immunomodulation of the TME represents a window of opportunity for preventing recurrence and improving long-term survival. In the aforementioned study, expression profiling of chemokines revealed that CD8⁺ TIL abundance correlated with production of the IFN γ-inducible factors CXCL9/CXCL10 and CCL5 (8). Receptors for these chemokines (CXCR3 and CCR5, respectively) are highly expressed on CD8⁺ TILs, other cytotoxic immune cells such as natural killer (NK) cells, and myeloid lineages (11–13). Collectively, these immune features (IFNγ response, CXCL9/10 production, and high CD8⁺ TIL recruitment) define a “hot” TME, and predict both improved overall survival and response to checkpoint inhibitor immunotherapy across multiple cancers (14, 15). However, the tumor-intrinsic factors that confer this favorable immune phenotype, in Ewing sarcoma or any other cancer, are largely unknown.

We recently identified ubiquitin-specific protease 6 (USP6) as a novel regulator of IFN signaling in Ewing sarcoma. USP6 functions by directly deubiquitinating Jak1, leading to its stabilization and activation, with resultant phosphorylation of the STAT1 and STAT3 transcription factors (16, 17). We found that USP6 is sufficient to trigger an IFN response gene signature, both in Ewing sarcoma cells and other neoplastic models (17). USP6 further renders Ewing sarcoma cells hypersensitive to exogenous Type I and II IFNs, as reflected by heightened and prolonged STAT1 phosphorylation, and synergistic induction of IFN-stimulated genes. However, how USP6 and its activation of IFN signaling impact Ewing sarcoma pathogenesis remains unexplored. Indeed, IFNs can exert either anti- or protumorigenic effects depending on the cancer context (18–20).

To date, only protumorigenic functions have been ascribed to USP6. USP6 is the key etiologic agent in several benign mesenchymal neoplasms, including nodular fasciitis (NF) and aneurysmal bone cyst (ABC), where it undergoes promoter-swapping translocations that result in overexpression of wild type USP6 (21). We have developed cellular and animal models of NF and ABC, and shown that overexpression of USP6 in candidate cells of origin (fibroblasts and pre-osteoblasts) is sufficient to induce formation of tumors that recapitulate molecular, histological, and clinical features of NF and ABC (22, 23). In these models, activation of STAT3, NF-κB, and β-catenin by USP6 was essential for tumorigenesis, as genetic or pharmacologic inhibition of these pathways significantly attenuated xenograft growth (16, 23, 24). In addition, we recently reported that an IFN signature is triggered by USP6 in NF, both in primary patient samples and in our cellular model (17). How IFN signaling affects NF pathogenesis is completely unknown. It is of interest to note, however, that NF lesions invariably regress, although the mechanism underlying this spontaneous regression remains unexplored.

In the current study, we investigated the effects of USP6 and IFN signaling on Ewing sarcoma tumorigenesis, and uncovered an unexpected tumor inhibitory function for USP6. In Ewing sarcoma cells in vitro, USP6 was capable of directly inducing expression of immunostimulatory chemokines, and also mediated their synergistic induction upon treatment with exogenous IFNs. Conditioned medium (CM) from USP6-expressing Ewing sarcoma cells was competent to enhance migration of multiple immune cell lineages, and coculturing enhanced activation of NK cells and primary monocytes-derived lineages in vitro. Furthermore, USP6 inhibited Ewing sarcoma tumor growth in xenografted nude mice, coincident with increased intratumoral chemokine production, and immune cell infiltration and activation. Consistent with this antitumorigenic effect in mice, we found that high USP6 mRNA expression was associated with immune cell infiltration and significantly improved event-free and overall survival in patients with Ewing sarcoma. Together, our studies provide the first demonstration of a tumor-suppressive function for USP6, and reveal that it functions to promote an immunogenic TME through enhanced IFN signaling.

RD-ES, A673, SK-N-MC, and TC-71 Ewing sarcoma cell lines and their HA-USP6-expressing and Jak1/STAT1 CRISPR derivatives were cultivated as previously described (17). NK92 was from the ATCC. Cell line identity was confirmed by STR analysis and EWS–FLI1 expression. All cells were screened for Mycoplasma every 2–4 weeks. Primary human monocytes and T cells from peripheral blood were obtained from the Human Immunology Core at the University of Pennsylvania. Monocytes, CD4⁺ and CD8⁺ T cells from de-identified healthy adult human donors were purified from apheresis material by negative selection using RosetteSep Human Monocyte, CD4⁺ or CD8⁺ Enrichment Cocktails (StemCell Technologies) according to the manufacturer's instructions.

Reagents were obtained from: doxycycline, ClonTech (#8634–1); Jak Inhibitor I (#420099) and PS-1145 (P6624), Sigma. IFN α and β were from PBL Assay Science (#11410–2 and #11200–1), and IFNγ from PeproTech (#300–02). Lipofectamine 2000 was from Life Technologies. Flow antibodies/reagents were from BioLegend: Zombie UV (#423107), HA (#901526), IFNAR1 (#21370–3), CD11b (#101208), MHCII (I-A/I-E; #107620), CD86 (#105036), CD11c (#117318), CD45 (#103116 or #103128), F4/80 (#123116), Ly6c (#128033), Ly6g (#127622), mouse Fc Trustain block (#101319), NK1.1 (#108724), anti-mouse CD69 (#104514), and CD25 (#102036). IFNGR (10338-MM05-A) was from Sino Biologic.

CXCL10 promoter reporter constructs were generously provided by Dr. David Proud (University of Calgary). Reporters were co-transfected with constructs encoding Renilla luciferase and USP6 into HeLa cells; IFNγ was added, and dual luciferase assays were performed the following day (Promega Dual-Luciferase #E1910). Firefly luciferase/Renilla ratio was calculated, then expressed relative to the value for cells expressing the WT promoter with USP6 and treated with IFNγ.

Semiquantitative chemokine antibody array was purchased from RayBiotech (Array #G5–4). USP6/RD-ES cells were serum starved, with or without doxycycline and IFNγ (50 ng/mL) for 24 hours. The antibody array was incubated with CM overnight at 4°C, and processed per manufacturer's instructions. Signal intensities were quantified using LI-COR Odyssey.

For the CXCL10 ELISA, USP6/RD-ES or USP6/A673 cells were seeded in growth medium, and treated with or without doxycycline and IFNγ (5 ng/mL). CM was collected the following day and subjected to ELISA (BioLegend, #439904) per manufacturer's instructions. All samples were analyzed in technical triplicate; data represent n = 3 for USP6/RD-ES, and n = 2 for USP6/A673.

CM was prepared by starving cells in serum-free medium with or without doxycycline and IFNγ (50 ng/mL) for 24 hours. Immune cells (2.5–5.0E5/well in 25 μL serum-free medium containing 0.1% BSA) were seeded in Transwell chambers (Neuroprobe #101; 5 micron pore size). CM or control medium was dispensed into the distal chamber, and cells were allowed to migrate for 2 hours at 37°C. The upper side of the filter was washed three times with PBS to remove non-migrated cells; the plate/filter was then centrifuged at low speed, and cells in the bottom chamber counted either by flow cytometry, or manually using a hemocytometer. All samples were assayed in technical triplicate, for n = 3–5 experiments.

HA-USP6/RD-ES (which express GFP) and NK92 were cocultured at the indicated ratios overnight in the presence of 200 U/mL IL2 (BioLegend # 589108). To monitor killing, viable HA-USP6/RD-ES cells were quantified by flow cytometry, monitoring GFP⁺/propidium iodide–excluded cells. To monitor NK92 activation and degranulation, surface levels of CD69 and CD107a were quantified in the CD45⁺/GFP⁻ population by flow cytometry. RNA was also isolated from the cocultures and subjected to RT-qPCR.

All procedures were performed under Institutional Animal Care and Use Committee–approved protocols. J:NU mice (4–8 weeks; The Jackson Laboratory) were pre-treated with drinking water containing doxycycline (BioWorld #40410005–2; 1 mg/mL) and 2.5% sucrose for 5–7 days before xenograft. Mice were subcutaneously injected with indicated cells (2.5E6) resuspended 1:1 in PBS:Matrigel (Corning #356234). Doxycycline water was changed twice weekly for the duration of the experiment. Two protocols were used to determine effects of USP6 on tumor growth. In the first, the primary goal was to document effects on terminal mass at a common timepoint, and mice were sacrificed approximately 3–4 weeks after xenografting. In the second, the goal was to assess the time required to reach a defined volume (2,000 mm³). Tumors were measured 2–3 times per week using digital calipers. Tumor volume was calculated [Volume = (π/6)(Length×Width²)], where length is the longest measurement and width the shortest.

Following tumor excision, representative sections of tumor were digested with Miltenyi tumor dissociation kit (#130–096–730). Digested samples were subjected to surface staining for the indicated markers. Cells were resuspended in Cell Staining Buffer (BioLegend #420201) and blocked with either mouse or human Fc Trustain block (BioLegend #101319 or #422302) at room temperature in the dark. Staining with antibodies was subsequently performed at 4°C in the dark for 20 minutes. For experiments monitoring intracellular HA-USP6, after surface staining of the indicated markers, cells were washed fixed in 4% formalin for 20 minutes at room temperature in the dark. Cells were then permeablized with BioLegend intracellular staining permeabilization wash buffer (#421002), followed by staining with anti-HA for 20 minutes at room temperature in the dark. Samples were washed and resuspended in Cell Stain Buffer. The HA-negative gate was set using HA-USP6 Ewing sarcoma cells without doxycycline treatment, which lack expression of HA-USP6. Data were collected using LSR II or Cytoflex flow cytometers, and analyzed with FlowJo.

All data represent results from at least 3 biological replicates. GSEA and microarray analysis were performed as previously described (17). GO datasets were acquired from https://www.gsea-msigdb.org/gsea/msigdb/index.jsp. For the Kaplan–Meier analysis, only the primary untreated patient samples in GSE17679 were included. To assess immune infiltration in primary Ewing sarcoma samples, the CIBERSORT algorithm (25) was adapted by curating the gene lists for each immune lineage, then applying GSEA to compare enrichment of each lineage in patients with USP6^high versus USP6^low Ewing sarcoma. The ratio-paired t test was used to determine significance (P <0.05) for all samples that were comparing the effects of +dox or +IFNγ to +IFNγ+dox. All other comparisons were done using a paired t test.

We recently reported that USP6 is associated with an IFN response gene signature in primary patient samples of Ewing sarcoma, nodular fasciitis (NF), and germ cell tumors. Among the genes, most consistently associated with high USP6 expression across these neoplasms were chemokines, that is, chemotactic cytokines. In two independent Ewing sarcoma patient datasets, high USP6 expression was strongly associated with multiple genesets for leukocyte chemotaxis and chemokine signaling (Supplementary Fig. S1). Across all four genesets, ligands for the chemokine receptors CXCR3 (CXCL9 and CXCL10) and CCR5 (CCL4, CCL5, and CCL8) emerged within the core enrichment set (Supplementary Fig. S2). These same genesets and chemokines were also found to be associated with high USP6 expression in germ cell tumors and in USP6-translocated NF (Supplementary Fig. S3A), suggesting that USP6 may play a direct role in their induction.

To determine whether USP6 was capable of directly triggering expression of these chemokines, studies were performed in Ewing sarcoma cell lines inducibly expressing USP6 (17). As previously described, the USP6 locus is highly methylated in all immortalized cell lines examined, resulting in its silencing and necessitating generation of cell lines ectopically expressing USP6 in a doxycycline-inducible manner, at physiologic levels (17). In the patient-derived Ewing sarcoma cell line RD-ES, upon addition of doxycycline, USP6 triggered the same chemotaxis and chemokine signatures as was observed in the USP6^high primary Ewing sarcoma tumors (Supplementary Fig. S3B). Together, these results demonstrate that high USP6 expression is associated with chemokine production across multiple tumor types in patients, and that USP6 is capable of directly inducing their expression in vitro.

We previously reported that USP6 not only induces an IFN gene signature by itself, but that it also magnifies the response to exogenous type I and II IFNs (17). Among the genes with the greatest synergistic induction were ligands for CXCR3 and CCR5: USP6 or IFNα individually led to modest increases in levels of these chemokines, whereas expression was increased 2–4 orders of magnitude in USP6-expressing cells treated with IFNα (Supplementary Fig. S3B). The chemokines exhibiting the greatest synergistic induction were CXCL9/CXCL10/CXCL11 and CCL4/CCL5/CCL8/CCL20. As mentioned above, CXCL10 and CCL5 production was correlated with CD8⁺ TIL recruitment and improved survival in patients with Ewing sarcoma, and CXCL10 has been associated with TIL recruitment and improved prognosis across multiple other malignancies (8, 11, 12, 14). USP6′s effect on these chemokines was highly selective, as it did not impact expression of chemokines with protumorigenic, immunoinhibitory functions (Supplementary Fig. S3B; ref. 13).

These RNA-seq results were validated by RT-qPCR in USP6/RD-ES, focusing on CXCL10 and CCL5 as representative ligands for CXCR3 and CCR5, respectively. Induction of USP6, or treatment with either IFNα, IFNβ, or IFNγ individually modestly induced CXCL10 and CCL5 expression, whereas all three IFNs elicited synergistic induction of these chemokines in the presence of USP6 (Fig. 1A and B). To further characterize this IFN-based synergy, we focused on type II IFNγ because our previous work showed that type I IFNs (particularly IFNβ) induce apoptosis of USP6-expressing RD-ES through activation of the TRAIL pathway, thus complicating analysis (17). We tested a range of IFNγ doses and found that USP6 amplified chemokine expression at all doses: At a low dose of 0.5 ng/mL, IFNγ had no effect on control RD-ES cells, but induced a 10,000-fold increase in CXCL10 and 100-fold increase in CCL5 expression in USP6/RD-ES (Fig. 1B). At a 100-fold higher dose of IFNγ (50 ng/mL), USP6 still enhanced CXCL10 and CCL5 expression more than 2 orders of magnitude (Fig. 1B). Similar effects were observed in multiple independent patient-derived Ewing sarcoma cell lines, including A673, TC71, and SK-N-MC cells, though the specific chemokines induced and the magnitude of the effect varied between the cell lines (Fig. 1C; Supplementary Fig. S4A). The extent of induction was also dependent on USP6 dosage as RD-ES cells expressing higher levels of USP6 exhibited greater CXCL10 induction, both basally and in response of IFNγ (Supplementary Fig. S4B). Furthermore, screening of CM by ELISA (for CXCL10) or a chemokine antibody array (for CCL4/5/8/20) confirmed that synergy of induction could be observed at the protein level (Fig. 1D and E; Supplementary Fig. S4C). Together, these data suggest that USP6 serves to fine tune and amplify the cellular response to IFNs, and that it could play a pivotal role in modulating the immune TME through production of chemokines.

We next explored the mechanisms underlying the synergistic chemokine induction, focusing predominantly on CXCL10, given its strong association with TIL recruitment across cancers. Jak1-STAT1 and NF-κB have previously been shown to regulate expression of CXCL10 and CXCL9 (26). CRISPR deletion of Jak1 or STAT1, but not STAT3, dramatically suppressed the synergistic induction of CXCL10 and CXCL9 by USP6+IFNγ (Fig. 2A; ref. 17). As an alternative means of probing their contributions, pharmacological inhibitors of the pathways were used. Induction was almost completely suppressed by a pan-Jak inhibitor, and partially suppressed by the NF-κB inhibitor PS1145 (Fig. 2B), indicating that Jak1-STAT1 assumes a dominant role in CXCL10 induction.

To further dissect the contributions of these pathways, we used a luciferase reporter construct driven by the proximal CXCL10 promoter comprising 972 base pairs upstream of the transcription start site (27). USP6+IFNγ cooperatively activated this reporter in transiently transfected HeLa cells, although the effects were much less pronounced than the dramatic synergy observed with endogenous CXCL10 (Fig. 2C). The reason for this reduced response is unknown, but possibilities are that cis elements outside of the proximal promoter are required, or that the transiently transfected reporter is not properly chromatinized in a manner required for synergy. Irrespective of the cause, we probed the contribution of various response elements within the promoter, which include two STAT1 homo-dimer–binding sites, two NF-κB sites, an AP-1 site, and an IFN-sensitive response element (ISRE), which binds to the STAT1–STAT2–IRF9 complex. As shown in Fig. 2C, the first STAT1 site, both NF-κB sites, and the ISRE all contributed to activation of the CXCL10 promoter by USP6+IFNγ, whereas the AP-1 site was largely dispensable.

The array of chemokines induced by USP6 has pleiotropic functions, including the ability to stimulate migration of myeloid, NK and T cells (13, 28). To assess their functionality, Transwell assays were performed on both primary and immortalized immune cells. CM from RD-ES or A673 cells either expressing USP6 or treated with IFNγ induced a modest increase in migration of THP-1, a monocytic cell line (Fig. 2D) and primary human monocytes (Fig. 2E). However, CM from USP6-expressing cells treated with IFNγ triggered a cooperative 20–100-fold increase in migration. Although it remains to be determined which specific chemokines/factors in this complex cocktail are responsible, it is likely that multiple factors function in concert. Indeed, although recombinant CXCL10 (reCXCL10) alone was a poor chemoattractant for THP-1, when combined with CCL5 a cooperative induction was observed (Fig. 2D). Enhanced migration was also induced in primary human CD8⁺ T cells, although the magnitude of enhancement was not as pronounced (Supplementary Fig. S5). We also tested whether CM could stimulate migration of NK cells using the NK92 cell line. However, contrary to prior reports we did not observe significant migration even with reCXCL10, perhaps due to low levels of CXCR3 in our isolates. Nevertheless, in sum these data confirm that USP6 expression in Ewing sarcoma cells can stimulate immune cell chemotaxis, both by itself and in the presence of ectopic IFNγ.

We also probed whether hyper-responsiveness of USP6-expressing cells to IFN might arise from effects on IFN receptor (IFNR) levels. As shown in Fig. 3A, levels of the IFNGR1 were elevated in the presence of doxycycline, both under basal and IFNγ-stimulated conditions. RT-qPCR confirmed that this effect was post-transcriptional (Fig. 3B). Flow cytometry revealed that surface IFNGR1 was significantly increased upon expression of HA-tagged USP6 (Fig. 3C and D). Similar effects were seen with the type I IFN receptor, IFNAR1 (Supplementary Fig. S6A–S6C). Thus, elevated levels of the IFNRs and Jak1 likely both contribute to the hyper-responsiveness of USP6-expressing cells to IFNs.

These results prompted us to investigate USP6′s effects on Ewing sarcoma tumor growth and the immune TME. Immunocompetent genetically engineered mouse models that accurately recapitulate human Ewing sarcoma disease are lacking (29). The field is thus limited to xenografting patient-derived cell lines into immunodeficient strains, such as nude or NSG mice. We first subcutaneously xenografted into nude mice clonal RD-ES cell lines expressing high or intermediate levels of USP6 [USP6 (high) and USP6 (med), respectively; ref. 17], in the absence or presence of doxycycline in the drinking water. USP6 inhibited growth of RD-ES xenografts in a dose-dependent manner (Fig. 4A). Although the effects did not reach statistical significance, there was a clear trend toward suppression of tumor growth by USP6. Indeed, only one out of five animals developed a tumor in the doxycycline-treated USP6 (high)/RD-ES cohort. In striking contrast, USP6 had no effect on RD-ES tumor growth when xenografted into NSG mice (Fig. 4B). Because the key difference between nude and NSG mice is that the latter lack NK cells and functional macrophages/dendritic cells (DC), this strongly suggests that the immune system plays a significant role in mediating the antitumorigenic effect of USP6 in nude mice. Indeed, these results indicate that in the absence of NK cells and macrophages/DCs, USP6 exerts no discernable cell-autonomous inhibitory effect on Ewing sarcoma tumor growth in NSG mice.

RT-qPCR of tumors confirmed that USP6 induced expression of CXCL9, CXCL10, and CCL5 in vivo (Fig. 4C). To investigate whether the increased immune infiltrates were associated with inhibition of tumorigenesis, tumors were isolated from the USP6 (med)/RD-ES cohorts, dispersed, then analyzed by flow cytometry [tumors in the USP6 (high)/RD-ES cohorts were too small to obtain sufficient material for analysis]. Strikingly, USP6 significantly increased total infiltrated immune cells, as monitored by the pan-leukocyte marker CD45 (Fig. 4D). Analysis of lineage markers revealed that USP6 increased intratumoral abundance of macrophages (F4/80⁺), DCs (CD11c⁺), and myeloid cells (CD11b⁺), but not monocytes (Ly6c⁺) or granulocytes (Ly6g⁺; Fig. 4D; Supplementary Figs. S7 and S8 for gating strategy).

USP6′s antitumorigenic effects were validated using two additional Ewing sarcoma cell lines, A673 (Fig. 5) and TC71 (Supplementary Fig. S9). For the USP6/A673 cells, 10 days after xenografting, palpable tumors were observed in 90% of control mice but only 50% of the doxycycline-treated cohort (Fig. 5A). USP6 prolonged the time to reach maximum tumor volume (2,000 mm³; Fig. 5B), with two “strong responders” in the +dox cohort failing to reach this endpoint (labeled with asterisks in Fig. 5A). Although tumor growth rates of the remaining mice in the +dox cohort ostensibly clustered with the −dox cohort, flow cytometric analysis of the tumors revealed that the former had a significantly increased fraction of dead cells (Fig. 5C; gray dots correspond to the strong responders; and Supplementary Fig. S8). Furthermore, as with the RD-ES xenografts, we observed significant increases in intratumoral levels of total leukocytes (CD45⁺), macrophages (F4/80⁺), and DCs (CD11c⁺), but not CD11b⁺ cells (Fig. 5C; Supplementary Figs. S8 and S9 for gating strategy). Notably, the two strong responder mice had the highest percentages of dead cells and immune cell infiltrates (gray dots in Fig. 5C–E). Markers of antigen-presenting cell (APC) activation were also examined: CD86 (a ligand for the co-stimulatory receptor CD28 on T cells), and MHC Class II (MHCII, which mediates antigen presentation). Among the total live population, cells that were positive for CD86 or MHCII were increased upon USP6 induction (Supplementary Figs. S10–S12). Expression of these markers was examined in classical macrophages (CD11b⁺ F4/80⁺) and DCs (CD11b⁺ CD11c⁺). In both lineages, USP6 induced a marked increase in the percentage of cells that were single and double positive for CD86 and MHCII (Fig. 5F; and Supplementary Figs. S10–S12).

We also examined infiltration of NK cells, the main cytolytic effectors present in nude mice. USP6 increased not only the intratumoral abundance of NK cells (Fig. 6A; Supplementary Fig. S13), but also their activation and proliferative potential (as monitored by CD69 and CD25/IL2Rα, respectively; Fig. 6A; Supplementary Fig. S14). Notably, the two strong responder mice exhibited the highest levels of NK cell infiltration, and CD69 and CD25 surface expression (gray dots in Fig. 6A). We were unable to detect enhanced IFNγ mRNA production in vivo, likely due to the low absolute numbers of intratumoral NK cells, which are the predicted main source of IFNγ.

The studies above demonstrate that USP6 is sufficient to increase intratumoral levels and activation of NK cells in vivo. To determine whether USP6 could directly induce their activation in vitro, coculture assays were performed using the human NK cell line, NK92. NK92 cells were cocultivated at various ratios with USP6/RD-ES (which express GFP), and survival was quantified by monitoring the GFP⁺/propidium iodide–excluded population. USP6/RD-ES were sensitive to killing by NK92 in the absence of doxycycline, but addition of doxycycline enhanced death at all effector:target (E:T) ratios (Fig. 6B). USP6 expression in RD-ES cells dramatically enhanced NK cell activation and degranulation, as measured by surface levels of CD69 and the degranulation marker CD107a, respectively (Fig. 6C). Furthermore, we speculated that because NK cells produce IFNγ upon activation, CXCL9/10 induction would be enhanced in doxycycline-treated co-cultures, due to hypersensitivity of USP6-expressing Ewing sarcoma cells to IFNγ. This was confirmed in Fig. 6D: cocultivation of USP6/RDE-ES with NK92 induced IFNγ production (independently of doxycycline), and CXCL9/CXCL10 induction was significantly increased upon USP6 expression.

The ability of USP6 to induce immune infiltration and activation in Ewing sarcoma xenografts led us to examine its association with immune infiltrates in human Ewing sarcoma tumors using an adaptation of the CIBERSORT algorithm that was recently developed to assess the presence of immune lineages in complex tumor samples (25). This algorithm curates approximately 500 immune markers to define 22 distinct immune lineages/activation states. In two independent Ewing sarcoma patient datasets, high USP6 expression was found to be associated with enhanced immune infiltration (Supplementary Fig. S15). Remarkably, USP6^high tumors exhibited increased abundance of multiple antitumorigenic lineages, including CD8⁺ and CD4⁺ (resting memory) T cells, and activated DCs, NK, and γδT cells. In contrast, levels of protumorigenic lineages such as M2 macrophages were not associated with high USP6 expression (Supplementary Fig. S15).

We next examined whether high USP6 expression was associated with differential prognosis. When patients were stratified by USP6 levels, comparing the top approximately 1/3 to the bottom approximately 2/3, those with high USP6 expression exhibited dramatically improved event-free and overall survival (Fig. 7A). After 5 years, patients with high USP6 levels had a nearly 80% survival rate, whereas those with low expression had an approximately 40% survival rate. It was not possible to confirm USP6 protein levels due to inavailability of antibodies suitable for IHC. Nevertheless, taken together these results suggest that in patients with Ewing sarcoma, as in our xenograft model, USP6 stimulates immune infiltration and suppresses cancer growth.

Hallmark features of a hot TME are induction of IFN response signature, CXCL10 production, and infiltration of CD8⁺ T lymphocytes (8, 11, 14). In aggregate, these features predict both improved overall survival and response to checkpoint inhibitors, across dozens of cancer types. Yet, remarkably little is known of why a subset of patients spontaneously exhibits this favorable immune phenotype. This study demonstrates that high USP6 expression is associated with all of these hot TME hallmarks in patients with Ewing sarcoma, and that USP6 is sufficient to drive an IFN response signature and CXCL10 expression in vitro, as well as intratumoral immune cell recruitment and activation in vivo.

Our work further provides mechanistic insights into how USP6 functions, and reveals for the first time an antitumorigenic role for USP6 in human malignancy. We show that in the context of Ewing sarcoma, USP6 functions to amplify IFN signaling, leading to immune infiltration/activation and inhibition of tumor growth. We envisage that USP6 engenders a potent positive feed-forward immunostimulatory loop as follows (model, Fig. 7B): USP6 directly induces Jak1 stabilization/STAT1 activation, leading to low level chemokine production, which we speculate stimulates recruitment of myeloid, NK, and T cells; these factors may also enhance immune cell activation and proliferation. USP6 expression in Ewing sarcoma can also directly enhance NK cell activation, and we predict that USP6 would also promote activation of T cells in vivo due to their ability to stimulate their migration as well as activate APCs in the TME. Activated NK and T cells produce IFNγ, which could feed back on the USP6-expressing Ewing sarcoma cells to synergistically enhance production of CXCL10 and other chemokines (due to their upregulated expression of IFNAR, IFNGR, and Jak1). Immune cell recruitment and activation would thereby be amplified, with resultant tumor cell elimination. We posit that this potent feed-forward response underlies the association of high USP6 expression with immune infiltration and dramatically improved survival in patients with Ewing sarcoma.

Although we show that USP6 has pleiotropic effects on multiple immune lineages in vitro and in vivo in nude mice, the full extent of its impact on the immune TME in Ewing sarcoma in vivo remains to be enumerated. As mentioned, in vivo studies are restricted to xenografting of human Ewing sarcoma cell lines into immunocompromised mice. Nevertheless, despite using a cross-species immunocompromised model, USP6 was able to elicit an antitumor immune response. Although nude mice have allowed us to probe effects on NK and myeloid lineages, they lack mature T lymphocytes. Thus, validating USP6′s effects on acquired immune cells in future in vivo studies will require allografting/xenografting of primary T cells. It is highly likely that USP6 will be able to promote T-cell recruitment and activation in vivo, given that USP6 induces robust production of CXCL10, that CM from USP6-expressing Ewing sarcoma cells can stimulate migration of primary T lymphocytes, and that Ewing sarcoma cells expressing USP6 can induce APC activation in vivo (as monitored by CD86 and MHCII surface upregulation). This notion is further supported by analysis of Ewing sarcoma patient samples, which indicates that high USP6 expression is associated with infiltration of CD8⁺/CD4⁺ lymphocytes and activated DCs (Supplementary Fig. S15). Interestingly, USP6 is also associated with infiltration of γδT lymphocytes in silico, which were recently reported to be the most significant favorable prognostic intratumoral immune lineage, across dozens of cancer types, in tens of thousands of samples (30).

A curiosity of our findings is that the dominant effect of USP6 in Ewing sarcoma is tumor suppressive, contrasting its oncogenic activity in NF. We confirmed that this was not a peculiarity of Ewing sarcoma, as USP6 also inhibited tumor growth and stimulated immune infiltration in a disparate tumor model, 293T embryonic kidney cells (Supplementary Fig. S16A–A16C). We speculate that whether USP6′s pro- or antitumorigenic function dominates depends on the oncogenic driver present (model, Supplementary Fig. S17). Ewing sarcoma is driven by the potent oncogenic driver EWS-FLI1, and in this context the oncogenic functions of USP6 may be superfluous; its antitumorigenic activity, mediated through IFN signaling, is therefore its dominant manifestation. In contrast, translocated USP6 is the oncogenic driver in NF (Supplementary Fig. S17, bottom). We posit that in the early phase of the disease when the tumor is actively growing, USP6′s protumorigenic signaling (via STAT3 and NF-κB) dominates (16, 23, 24). However, USP6 simultaneously triggers IFN signaling (17) and infiltration of antitumorigenic immune lineages, which ultimately prevail to drive the spontaneous regression that typifies NF. Indeed, adapted CIBERSORT analysis and IHC of NF confirmed enrichment of CD4⁺ and CD8⁺ T lymphocytes, γδT cells, and activated DCs and NK cells, and but not T_regs (Supplementary Fig. S18A–S18C).

On a final note, we conjecture a role for USP6 in cancer immunoediting, the process by which the immune system monitors and modulates the immunogenicity and clearance of developing neoplasms. Cancer immunoediting comprises three stages: elimination, equilibrium, and escape (31). During the elimination phase, nascent neoplastic cells are effectively recognized and cleared by the immune system. Some cells may evade elimination and enter the equilibrium phase, which is considered a state of functional dormancy where there is no detectable increase in tumor burden. Elimination and equilibrium have been extensively dissected mechanistically in mouse models, which reveal a central role for IFN signaling, CXCL9/10 production, and both NK and T cells, although a degree of immunoediting can occur even in the absence of adaptive immunity (31–33). However, confirmation of these mechanisms in humans has been elusive, because elimination/equilibrium lack clinical manifestation and are thus difficult to monitor. Our longstanding studies of NF, a self-limited transient neoplasm, led us to uncover that USP6 can render these defining features of elimination. We thus propose that USP6 may represent the first identified human endogenous elimination factor, in both NF and Ewing sarcoma, explaining its association with significantly improved Ewing sarcoma patient survival. Our current study demonstrates that USP6 has pleiotropic effects on the immune system, with its potent antitumorigenic activity likely mediated through multiple immune lineages. USP6 increases the intratumoral abundance of activated macrophages/DCs and NK cells in mice. Recent work has highlighted the importance of these innate immune lineages in combating tumor growth, and suggested that shifting macrophage and NK activation can have a profound impact on the response to existing immunotherapeutics (33–35). Forthcoming studies will dissect how immune subsets are activated by USP6, and assess their contribution to tumor suppression. Furthermore, future work will seek to harness the powerful and pleiotropic immunostimulatory properties of USP6 for therapeutic benefit.

No disclosures were reported.

I.C. Henrich: Conceptualization, data curation, formal analysis, validation, investigation, methodology, writing–original draft, project administration, writing–review and editing. K. Jain: Conceptualization, data curation, formal analysis, validation, investigation, methodology, writing–original draft, project administration, writing–review and editing. R. Young: Formal analysis, investigation, methodology, writing–review and editing. L. Quick: Investigation, methodology. J.M. Lindsay: Investigation, methodology. D.H. Park: Investigation, methodology. A.M. Oliveira: Conceptualization, formal analysis, investigation, methodology, writing–review and editing. G.A. Blobel: Conceptualization, formal analysis, funding acquisition, writing–review and editing. M.M. Chou: Conceptualization, data curation, formal analysis, supervision, funding acquisition, investigation, methodology, writing–original draft, project administration, writing–review and editing.

The authors thank Dr. Florin Tuluc and colleagues at the flow cytometry at CHOP, the Human Immunology Core, which is co-funded by the Penn Center for AIDS Research (CFAR, P30 AI 045008), and Abramson Cancer Center support grant (P30CA016520). This work was funded by NIH/NCI grants CA168452 and CA178601 (to M.M. Chou), and T32 GM008076 (to I.C. Henrich), and Kids Beating Cancer, Inc.

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