Multipronged enhancement of CART cell activity against ROR1 neoantigen by inhibition of H3K27 methyltransferase EZH2

Summary

ROR1 is expressed across a broad spectrum of malignancies. While T lymphocytes armed with chimeric antigen receptor targeting ROR1 (CART-ROR1 cells) have reached clinical stage, the efficacy of this therapy has been limited. We report that inhibition of H3K27 methyltransferase EZH2 and, to a much lesser degree, the DNA methyltransferase DNMT1 and the H3K9 methyltransferase G9A promotes ROR1 expression by malignant cells. EZH2 inhibition also directly impairs the growth of some malignant cell populations. The combination of EZH2 inhibition and CART-ROR1 cells is much more effective than CART-ROR1 cells alone in eliminating malignant cells. While EZH2 inhibition variably enhances the expression of immune checkpoint receptors, it markedly increases the synthesis of T cell-activating soluble factors and promotes the cytotoxic phenotype of CART-ROR1 cells. These results indicate that EZH2 inhibition enhances the CART-ROR1 cell effectiveness by diverse mechanisms, including promotion of ROR1 expression by malignant cells, suggesting that the combination may prove clinically more effective than CART-ROR1 cell monotherapy.

Graphical abstract

Highlights

• •Inhibition of EZH2 and, much less of DNMT1 or G9A, boosts expression of ROR1 neoantigen
• •CART-ROR1 cells kill malignant cells expressing ROR1 below detection level by flow cytometry
• •CART-ROR1 cell/EZH2 inhibitor combination effectively eliminates ROR1-expressing cells

Wang et al. report that inhibition of EZH2 and, to lesser degree, other epigenetic modifiers promotes expression of ROR1 neoantigen on malignant cells. CART cells targeting ROR1 are able to detect the neoantigen at very low concentration and, in combination with EZH2 inhibition, highly effectively eliminate the ROR1-expressing cells.

Introduction

Ectopic expression of receptor tyrosine kinase-like orphan receptor 1 (ROR1) has emerged as a prime example of cancer cell reprogramming, fostering the survival and growth of malignant cells.^1,2,3,4 ROR1 is physiologically expressed, seemingly at low concentration, in the fetal liver, parathyroid, pancreatic islets, adipose tissue, gastrointestinal tract, and hematogones.² Aberrant ROR1 neoantigen expression is observed across a broad spectrum of malignancies, including carcinomas and lymphoid malignancies derived from mature B lymphocytes including mantle cell lymphoma (MCL) and diffuse large B cell lymphoma (DLBCL).^1,2,3,4 ROR1 expression is associated with aggressive disease and poor prognosis.^3,4

ROR1 has emerged as an attractive therapeutic target in multiple cancer types, including lymphoid malignancies.^1,2,5,6,7 Various immunotherapeutic approaches have been developed to target ROR1, including immunotoxin conjugates,^8,9,10,11 bispecific antibodies,^6,7,12 and T lymphocytes armed with chimeric antigen receptor targeting ROR1 (CART-ROR1 cells).^13,14,15 While CART-ROR1 cells seem to be the most promising therapy and show no detectable toxicity in normal tissues,^13,15,16 their therapeutic efficacy has been so far rather limited,¹⁶ indicating the need for combinatorial approaches.

Gene expression is tightly regulated by a variety of epigenetic modifiers.¹⁷ Members of the DNA methyltransferase (DNMT) family, including DNMT1, inhibit gene expression by silencing gene transcriptional activity at the DNA level. Histone modifiers also silence genes, primarily by targeting histone 3 (H3) lysine (K). While EZH2 is responsible for histone 3 lysine 27 (H3K27) methylation including trimethylation (me3),¹⁸ G9A catalyzes H3K9 monomethylation and dimethylation (me1/2).¹⁹

Results

Frequent and heterogeneous ROR1 expression in cancers

MCL cell lines display a diverse degree of ROR1 expression at the mRNA level, ranging from very strong to barely detectable (Figure 1A). Notably, even the MCL-RL cell line in which ROR1 protein was not be detected by either flow cytometry²⁰ or western blot (Figure 1B) expressed ROR1 mRNA at a very low concentration, just above the negative control of peripheral blood mononuclear cells (PBMCs). In turn, the Granta cell line expressed ROR1 mRNA at an overall low concentration, with the expression of ROR1 protein barely detectable by western blot (Figure 1B). In contrast, Jeko-1, REC-1, and Maver cell lines strongly expressed ROR1, as detected on the mRNA (Figure 1A) and protein levels (Figure 1B and Zhang et al.²⁰). We also examined expression of ROR1 protein in seven DLBCL cell lines (Figure 1B), with two cell lines (RIVA and LY-18) expressing ROR1 strongly, two (LY-8 and SC-1) expressing it weakly, and the remaining three displaying no detectable ROR1 expression. We confirmed this diverse pattern of ROR1 expression in selected cell lines using immunohistochemistry (IHC) with the MCL cell line Maver showing universal expression, the DLBCL line LY-8 showing staining of cell subsets, and the two lines MCL-RL (MCL) and LY-1 (DLBCL) displaying no detectable staining (Figure S1A).

Figure 1.

Heterogeneity of ROR1 expression by malignant lymphoid cells

(A) Expression of ROR1 mRNA in MCL cell lines detected by quantitative reverse-transcription PCR (RT-qPCR), with PBMCs from two healthy donors serving as negative control.

(B) ROR1 protein expression in MCL and DLBCL cell lines identified by western blotting. Expression of EZH2 is also depicted. Expression of β-actin (BACT) served as control.

(C) ROR1 mRNA expression in primary MCL cells retrieved from GEO datasets (GEO: GSE1329291, GSE23502, GSE347633, GSE425494, and GSE360005): a total of 89 primary patient samples (red) and control four cell lines: Maver, Jeko-1, Mino, and Z138 (blue).

(D) ROR1 mRNA expression in primary DLBCL cells retrieved from GEO: GSE117556 and GSE181063: 2,238 patient samples.

(E) Expression of ROR1 protein by a representative primary MCL cell population detected by flow cytometry. Upper (left): expression of the pan-hematopoietic cell marker CD45 (y axis) vs. the B cell marker CD19 (x axis); (middle): CD19 vs. CD5 marker expressed by MCL cells and T lymphocytes; (right): CD19 vs. ROR1 expression. Lower: CD5 and ROR1 expression by MCL cells.

(F) Expression of ROR1 protein by a representative DLBCL population, detected by flow cytometry. Upper left: CD45 (y axis) vs. CD19 (x axis) expression; (middle): CD19 vs. CD5 expression; (right): CD19 vs. ROR1 expression. Lower: negative CD5 and positive ROR1 expression by DLBCL cells.

(G) Schematic summary of diverse ROR1 protein expression in MCL, DLBCL, CLL/SLL, and MZL cells, detected by flow cytometry.

Our analysis of publicly available datasets of ROR1 mRNA expression in primary MCL and DLBCL cells revealed a highly heterogeneous expression of ROR1 (Figures 1C and 1D, respectively), with a subset of cases strongly expressing this neoantigenic receptor, a larger subset expressing it moderately, and the remainder showing borderline to non-detectable expression. We also evaluated by flow cytometry ROR1 protein expression in primary malignant cells from MCL, DLBCL, chronic lymphocytic leukemia/small lymphocytic lymphoma (CLL/SLL), and marginal zone lymphoma (MZL), with examples of strong ROR1 expression in MCL and DLBCL cells shown in Figures 1E and 1F, respectively. The overall results for all four lymphoma types examined are summarized in Figure 1G, indicating substantial heterogeneity of ROR1 protein expression within and among these types of lymphoid malignancies, with DLBCL cells expressing ROR1 the least and CLL/SLL cells the most frequently.

Our additional mining of publicly available datasets identified a similar, highly heterogeneous ROR1 expression pattern in cell lines derived from the broad spectrum of hematologic malignancies including plasma cell myeloma (Figure S1B), acute myeloid and lymphocytic leukemia (Figure S1C), and non-hematologic neoplasms including cancers of breast (Figure S1D), colorectum (Figure S1E), and ovary (Figure S1F). The plot of ROR1 mRNA expression vs. ROR1 gene copy number relative to ploidy in multiple cancer types (Figure S1G) further indicated frequent, yet highly heterogeneous ROR1 expression and also indicated that ROR1 gene amplification does not play a significant role as a booster of ROR1 expression.

ROR1-expressing cells are efficiently eliminated by CART-ROR1 cells in a ROR1-dependent manner

To determine the effect of CART-ROR1 cells on malignant cells, we co-cultured Jeko-1 cells, strongly expressing ROR1 (Figures 1A and 1B; Figure 2A, left), with CART-ROR1 cells for 24 h at an effector-to-target (E:T) ratio of 1:1 (Figure 2A, right) and the E:T ratios of 0.5:1 and 2:1 (Figure S3A). T lymphocytes from the same donor that did not express the CAR-ROR1 construct (“mock” T cells) served as control. CART-ROR1 cells killed the vast majority of Jeko-1 cells, whereas “mock” T cells had very limited effect. Target specificity for ROR1 of the CART-ROR1 cells was confirmed by analyzing their impact on viability of ROR1-expressing vs. ROR1-depleted model cells, with the former being eliminated and the latter remaining unaffected (Figure S2).

Figure 2.

Impact of CART-ROR1 cells on survival of ROR1-expressing MCL cells

(A) Effect of CART-ROR1 cells vs. “mock” T cells from the same donor on Jeko-1 cells detected after 24-h co-culture, with the effector-to-target (E:T) ratio of 1:1. CD19 marks Jeko-1 cells, CD3 CART-ROR1 and “mock” T cells. Shown in left is strong ROR1 expression by Jeko-1 cells.

(B) Effect of CART-ROR1 cells vs. “mock” T cells on MCL Maver cells analyzed at the depicted time points, with the E:T ratio of 0.2:1. Depicted in uppermost left is strong ROR1 expression by Maver cells.

(C) Effect of CART-ROR1 cells vs. “mock” T cells on primary MCL cells from two MCL patients. The co-cultures were performed at an E:T ratio of 0.5:1 for 24 h. Bone marrow-derived stromal cells were also present to support the growth of primary MCL cells. p value: ∗∗, < 0.005.

(D) Effect of the EZH2i tazemetostat on the growth of MCL cells. Results of MTT conversion assay with the index MCL-patient derived cell lines after 10-day incubation in either vehicle (DMSO) or depicted concentrations of the EZH2i.

We next co-cultured Maver, another MCL cell line strongly expressing ROR1, with the CART-ROR1 and “mock” T cells for 24 h and 7 and 14 days at a low E:T ratio of 0.2:1 (Figure 2B). While the 24 h co-culture under such E:T ratio resulted in limited CART-ROR1 cell-mediated killing of the target cells, elimination of the target cells was much more pronounced by day 7. CART-ROR1 cells were, however, unable to completely eliminate the target MCL cells at this E:T ratio, as indicated by their disappearance by day 14 associated with persistence of Maver cells. Overall, Maver cells were somewhat less sensitive to CART-ROR1 cells than Jeko-1 cells, as determined by direct comparison in 24 h assay at the same E:T ratios, but still were essentially completely eliminated at the higher E:T ratios after 72 h (Figure S3).

Of note, analysis of the residual surviving Maver cells revealed that they did not express ROR1 detectable by flow cytometry on day 14 but gradually regained strong expression of the receptor (Figure S4A). Importantly, retreatment of Maver cells with CART-ROR1 cells at a high E:T ratio eradicated them (Figure S4B), suggesting that their non-detectable ROR1 in Figure S4A reflects not loss of but diminished expression below the level of flow cytometry sensitivity, yet these cells could be eliminated by CART-ROR1 cells at the higher E:T ratios. Accordingly, CART-ROR1 cells, but not the matched “mock” T cells, eliminated the weakly ROR1-expressing subset of Granta cells (Figure S5), further indicating that weak ROR1 expression is recognized by the CART-ROR1 cells. Strikingly, we found that CART-ROR1 cells also kill, in a strictly ROR1-dependent fashion, the MCL-RL cells in which the expression of ROR1 cannot not be detected by either western blotting (Figure 1B) or flow cytometry (Zhang et al.²⁰ and Figure S6), further stressing the potency of CART-ROR1 cells. As shown in Figure S7, co-culture of MCL-RL cells with three different well-characterized (Figure S8) CART-ROR1 cell populations with their matching “mock” T cells at the depicted E:T ratios resulted in a time-dependent killing of MCL-RL cells by the CART-ROR1 cells but not “mock” T cells. On day 3 of the co-culture, a marked MCL-RL cell elimination was visible for one of the CART-ROR1 cell populations (D31328) at a higher E:T ratio of 0.2:1. By day 5, this cytotoxic effect not only became even more pronounced for this CART-ROR1 population but became visible also for a second CART-ROR1 population (D17526). By day 7, profound killing of MCL-RL cells was seen for all three CART-ROR1 cell populations at the 0.2:1 ratio, proportionately to their degree of expression of the CAR construct (Figure S8A). The elimination also became very pronounced for the most potent CART-ROR1 D31328 cells at the lower E:T ratio of 0.05:1. These results indicate that CART-ROR1 cells specifically recognize and efficiently eliminate malignant cells expressing ROR1 at concentrations below the threshold of detection by standard methods.

To demonstrate that CART-ROR1 are also able to eliminate primary ROR1-expressing malignant cells, we co-cultured them with MCL cells with confirmed ROR1 expression (Figure S6) from two patients, using matched “mock” T cells as control. Similar to MCL cell lines, these cells were killed in a ROR1-specific manner (Figure 2C).

EZH2 inhibition impairs the growth of a subset of MCL cell populations

Given that tazemetostat (EPZ-6438), an inhibitor of H3K27 methyltransferase EZH2, suppresses growth of not only follicular lymphoma tumors harboring activating EZH2 mutations but also those with wild-type EZH2²¹ and that EZH2 is abundantly expressed in MCL cells,²² we investigated if tazemetostat affects growth of MCL cells. As shown in Figure 2D, the EZH2 inhibitor (EZH2i) diversely impacted growth of the index MCL cell lines, with profound inhibition of MCL-RL, moderate inhibition of SP-49, and no inhibitory effect on Granta and Jeko-1 cells. These results indicate that in some MCL cases, EZH2i may potentially complement the therapeutic effect of CART-ROR1 cells by directly impairing growth of MCL cells.

Inhibition of epigenetic gene silencers promotes expression of ROR1 in malignant cells

While normal mature B cell counterparts of MCL and DLBCL do not express ROR1 and various mechanisms may regulate its expression, we tested if epigenetic reprogramming may be responsible for aberrant ROR1 expression by these malignant cells. To test this possibility, we exposed Granta cells (∼8% of which express flow cytometry detectable ROR1; Figure S2) to either EZH2i or GSK3685032, a next-generation selective inhibitor of DNA methyltransferase DNMT1 (DNMT1i). While the DNMT1i displayed a negligible effect on ROR1 expression (Figures 3A and 3B), the EZH2i induced marked ROR1 expression in a time-dependent manner. The combination of the DNMTi and EZH2i was even more effective, with EZH2i clearly being the main contributor. Extension of cell treatment with the EZH2i for up to 15 days further increased the ROR1 expression (Figure S9), with only minor differences in the effectiveness seen among the drug’s concentrations of 1, 2, and 5 μM, of which 2 μM was selected for further studies.

Figure 3.

Effect of epigenetic silencers’ inhibitors on ROR1 expression by malignant cells

(A) Granta cells were treated with either the DNMT1i GSK3685032 at 1 μM, EZH2i tazemetostat at 2 μM, or the GSK3685032 (1 μM) + tazemetostat (2 μM) combination vs. drug vehicles for the listed number of days and analyzed for ROR1 expression by flow cytometry. Blue profiles: ROR1 expression in cells cultured with the drug vehicle(s), red profiles: ROR1 expression in cells cultured with the drugs.

(B) Schematic depiction of the drug impact on ROR1 expression calculated from the profiles shown in (A). The plotted results reflect the analysis of single samples per each depicted condition.

(C) Impact of EZH2i on ROR1 expression in Granta cell subsets with non-detectable and detectable ROR1 expression at baseline. Flow cytometry cell-sorted Granta cell differing in ROR1 expression as either non-detectable (ROR1⁻) or detectable (ROR1⁺) were exposed to EZH2i tazemetostat at 2 μM for the listed numbers of days and examined for gain in ROR1 expression by western blot. Unsorted (parental) Granta cells were used as control. Expression of β-actin (βACT), the EZH2 target H3K27me3, and total H3 served as additional controls.

(D) Induction of ROR1 expression by EZH2i in primary MCL cells. Patient’s CD19⁺CD5⁺ MCL cells, circulating in peripheral blood, were analyzed by multi-color flow cytometry for ROR1 expression on day 1 and day 5 of co-culture in triplicates with EZH2i at 2 μM (red profiles) or control medium (blue profiles).

(E) Schematic depiction of EZH2i-mediated ROR1 expression induction calculated from the flow cytometry profiles shown in (D). Relative ROR1 expression fold-increase (solid black columns) was calculated from baseline ROR1 expression (open columns) of 17.9% on day 1 and 20.6% on day 5 and the EZH2i-mediated expression of 18.1% on day 1 and 37.7% on day 5. p value: ∗∗∗, < 0.0001.

To confirm that the ROR1 induction is an on-target effect of tazemetostat, we also applied valemetostat, a structurally distinct inhibitor of EZH1/2. In the dose- and time-dependent studies, both inhibitors displayed highly comparable ROR1-inducing capability, with tazemetostat being somewhat more potent (Figure S10).

To determine which subset of Granta cells (either with flow cytometry-detectable or non-detectable ROR1 expression; Figure S5) was the source of the EZH2i-inducted ROR1-expressing cells, we separated ROR1-expressing and -non-expressing cells (Figure S11). Both Granta cell subpopulations upregulated ROR1 expression in response to EZH2i, as determined by western blot (Figure 3C) and flow cytometry (Figure S12A). Notably, the EZH2i-induced ROR1 expression was rather stable, because the withdrawal of the inhibitor for 4 days did not affect ROR1 expression in any of the Granta cell subpopulations (Figure S12B), and, in a separate experiment with parental Granta cells, no effect on ROR1 expression was observed upon withdrawal for up to 10 days (Figure S13).

EZH2i promoted ROR1 expression even in primary MCL cells, as a 5-day exposure markedly (Figure 3D) and statistically significantly (Figure 3E) upregulated ROR1 expression in these cells.

Further studies (Figure 4) showed that DNMT1i and EZH2i can also induce ROR1 expression detectable by flow cytometry in MCL-RL cells, with EZH2i and the drug combination being more effective than DNMT1i inhibitor alone. The inhibitors upregulated ROR1 expression in a similar pattern in MCL SP49 cells displaying high baseline ROR1 expression and DLBCL LY-8 cells showing moderate baseline expression of the receptor.

Figure 4.

Effect of epigenetic silencers’ inhibitors on ROR1 expression in diverse MCL and DLBCL cells

The depicted MCL (MCL-RL and SP49) and DLBCL (LY-8) cells were treated for 6 days with either the DNMT1i GSK3685032 (1 μM), the EZH2i tazemetostat (2μM), or the GSK3685032 (1 μM) + tazemetostat (2 μM) combination vs. drug vehicles and analyzed for ROR1 expression by flow cytometry. Blue profiles: ROR1 expression in cells cultured with the drug vehicle(s), red profiles: ROR1 expression in cells cultured with the drugs. The results reflect single sample per each drug vs. the corresponding drug vehicle.

Considering that the H3K9me1/2 methyltransferase G9A is known to co-operate with EZH2 in gene silencing,²³ we examined the effect of its inhibition in two MCL cell lines, using EZH2i as a control (Figure S14). While the G9A inhibitor (G9Ai) promoted ROR1 expression in both ROR1 baseline-low Granta and ROR1 baseline-high SP49 cells, the ROR1 induction was much weaker compared to that by EZH2i, further highlighting the effectiveness of EZH2i in the induction.

EZH2 and H3K27me3 expression in primary MCL and DLBCL tissues

Because our index MCL and DLBCL cell lines universally expressed EZH2 (Figure 1B) and inhibition of its enzymatic activity boosted ROR1 expression (Figures 3 and 4), to determine EZH2 expression in primary cells and, perhaps more importantly, indirectly its activity status, we analyzed MCL and DLBCL tissues by IHC for expression of this histone modifier and its product H3K27me3. As shown in Figure S15, we detected the expression of both in the vast majority of cases. Among MCL cases, 18 displayed strong EZH2 expression and 5 weak expression, while in 3 cases, the protein was undetectable. In DLBCL cases, 31 expressed EZH2 strongly and 3 weakly, with no negative cases. Similarly, H3K27me3 expression was detected in all MCL and DLBCL cases, with strong expression seen in 25 and 22 and weak expression in 1 and 17 cases, respectively, highlighting its essentially universally expression in these lymphomas.

EZH2 binds to ROR1 gene promoter and methylates the promoter-associated H3K27

To demonstrate that the induction of ROR1 expression by EZH2i stems from the direct EZH2-mediated silencing of ROR1, we examined EZH2 binding to the gene’s promoter. The methyltransferase was indeed bound to the promoter in a much higher abundance in ROR1 low-expressing MCL-RL than in high-expressing Jeko-1 cells (Figure S16A). Similarly to EZH2, the H3K27me3 mark was also detected at the ROR1 promoter in Jeko-1 and MCL-RL cells, in an amount inversely proportional to ROR1 expression by these cells (Figure S16B). While EZH2i did not affect EZH2 binding to the promoter, it profoundly diminished H3K27me3 binding, further supporting the notion that EZH2 directly inhibits the expression of ROR1.

EZH2i promotes ROR1 expression in non-lymphocytic malignant cells

Given that aberrant, ectopic expression of the ROR1 neoantigen has been identified in a broad spectrum of malignancies, including plasma cell myeloma and ovarian carcinoma,^1,2,3,4 we examined two myeloma cell lines (KMS11 and RPMI-8226) and one ovarian carcinoma cell line (OVCAR3) for baseline and EZH2i-promoted ROR1 expression. As shown in Figure S17, while KMS11 and OVCAR3 lines displayed baseline ROR1 expression that was markedly upregulated by EZH2i, ROR1 expression could not be detected in RPMI-8226 cells, either at the baseline or after exposure not only to EZH2i but also to DNMT1i and G9Ai. Although limited in scope, these studies, nevertheless, confirm that ROR1 protein is expressed in cancers different from lymphomas and indicate that EZH2i is capable of enhancing ROR1 expression in some but not all cases of such malignancies.

EZH2i enhances elimination by CART-ROR1 cells of MCL cells, both strongly and weakly expressing ROR1

Given the targetability of the ROR1 neoantigen by CART-ROR1 cells (Figure 2) and EZH2i-mediated boosting of ROR1 expression (Figures 3 and 4), we examined the effects of combining CART-ROR1 cells with EZH2i on ROR1 highly expressing Jeko-1 cells. With cells cultured alone serving as control, these Jeko-1 were co-cultured at three different E:T ratios with CART-ROR1 cells in four different experimental setups: (1) only CART-ROR1, (2) CART-ROR1 and EZH2i applied simultaneously, (3) pretreatment of Jeko-1 target cells for 5 days with EZH2i but removal of the drug before addition of CART-ROR1 cells, and (4) Jeko-1 cell pretreatment with EZH2i and its continued presence after CART-ROR1 addition. When evaluated on day 5 of the co-culture (Figures 5A and 5B), CART-ROR1 cells alone effectively eliminated MCL cells at the E:T ratio of 0.2:1. However, a beneficial effect of EZH2i was observed at lower E:T ratios, most significantly when EZH2i was used in the pretreatment and continuously present during the co-culture. A similar beneficial effect of EZH2i addition was also observed (Figure S18) in the killing of Maver cells that also strongly expresses ROR1.

Figure 5.

Impact of EZH2 inhibition on CART-ROR1 cell-mediated elimination of MCL cells strongly and weakly expressing ROR1

(A) Jeko-1 cells strongly expressing ROR1 were cultured at the depicted diverse experimental conditions, with CART-ROR1 cells added at the listed different E:T ratios without or with EZH2i tazemetostat at 2 μM, applied in three different manners: present during the MCL/CART-ROR1 cell co-culture (& EZH2i), used to pretreat the MCL cells for 5 days directly prior to the co-culture (/EZH2i), or applied in the pretreatment as well as present in the culture medium (/EZH2i & EZH2i). Relative proportions (percentages) of Jeko-1 vs. CART-ROR1 cells were determined by flow cytometry after 5-day culture, using B cell-specific anti-CD19 and T cell-specific anti-CD3 antibodies. Arrows highlight the augmenting effect of EZH2i on CART-ROR1-mediated elimination of the target Jeko-1 cells.

(B) Schematic of CART-ROR1-mediated changes in the percentage of Jeko-1 cells relative to Jeko-1 cells cultured alone, calculated from the data depicted in (A). p values: ∗∗, <0.01; ∗∗∗∗, <0.0001.

(C) MCL-RL cells expressing ROR1 below the level of detection by standard methods, were co-cultured with CART-ROR1 cells at the listed E:T ratios without or with EZH2i tazemetostat (2 μM), either present during the co-culture (& EZH2i), used to pretreat the MCL cells for 5 days (/EZH2i), or applied in the pretreatment and present in the co-culture medium (/EZH2i & EZH2i). Relative percentages of MCL-RL vs. CART-ROR1 cells were determined by flow cytometry after 5-day culture, using CD19- and CD3-specific antibodies. Arrows highlight the augmenting effect of EZH2i on CART-ROR1-mediated elimination of the MCL-RL cells.

(D) Schematic summary of CART-ROR1-mediated changes in percentage of MCL-RL cells relative to MCL-RL cells cultured alone, calculated from the data shown in (C). p values: ∗∗∗, <0.001; ∗∗∗∗, <0.0001.

To determine the effect of CART-ROR1/EZH2i combination vs. CART-ROR1 cells on malignant cells with very low ROR1 expression, we targeted MCL-RL cells (Figures 5C and 5D). Although CART-ROR1 cells were somewhat less effective against MCL-RL cells than against Jeko-1 and Maver cells, the addition of EZH2i resulted in a much higher degree of MCL-RL cell elimination for all the E:T ratios examined, compared with CART-ROR1 cells alone. Interestingly, pretreatment of MCL-RL cells with EZH2i was essentially as effective as the pretreatment combined with continuous presence of the inhibitor, likely reflecting the direct cytotoxic effect of EZH2i on MCL-RL cells, as shown in Figure 2D. Qualitatively similar results indicating higher potency of CART-ROR1 cell/EZH2i combination compared with CART-ROR1 cells alone (Figure S19), were obtained with DLBCL LY-8 cells weakly expressing ROR1 (Figures 1A andS1A). The in-depth analysis indicated that the maximal elimination of LY-8 cells by CART-ROR1 cells alone and, in particular, in combination with EZH2i, was obtained at the higher E:T concentrations and was also time dependent, with the peak achieved on day 5.

Baseline and EZH2-triggered ROR1 expression is not universal among malignant cells and is absent in immune lymphocytes

Next, we examined two other DLBCL cell lines—JP and DLCL2 with non-detectable ROR1 expression (Figure 1A)—to determine if EZH2i treatment would induce ROR1 expression in these cells. However, neither EZH2i or DNMT1i alone nor the drug combination was able to induce ROR1 expression detectable by flow cytometry (Figure S20A). To assure that ROR1 is not expressed in these cells below the level of detection by this standard method, we co-cultured JP and DLCL2 cells, both untreated and pretreated with EZH2i, with CART-ROR1 cells in the absence and presence of the inhibitor. Co-culture in any of these conditions did not affect the growth of these DLBCL cells, in contrast to the control ROR1-expressing Jeko-1 cells efficiently killed by CART-ROR1 cells, to large degree in an EZH2i-exposure-dependent manner (Figure S20B). These findings indicate that not all DLBCL cells are capable of expressing ROR1, similar to plasma cell myeloma cells (Figure S17).

Further studies indicated that the immune T cell-rich PBMCs from five different donors, mitogen-activated to sustain their viability, did not express ROR1, either spontaneously or in response to EZH2i (Figure S21A). This finding was consistent in PBMCs with a baseline lack of H3K27me3 marks at the ROR1 gene promoter (Figure S21B). CART-ROR1 cells themselves also did not express ROR1, either at baseline, after treatment with EZH2i, or after co-culture in the presence of EZH2i with ROR1 high-expressing, EZH2i-pretreated Jeko-1 cells (Figure S22), further supporting the notion that immune T cells are not prone to express ROR1.

EZH2 inhibition diversely enhances expression of immune checkpoint receptors by CART-ROR1 cells

To examine the impact of EZH2i on the expression of key checkpoint receptors by CART-ROR1 cells, we co-cultured these cells with Jeko-1 cells, either alone or with EZH2i also used to pretreat Jeko-1 cells, and examined the receptors’ expression at baseline and on days 1, 3, and 6 of the co-culture. While only CD39 displayed baseline expression, three other receptors, PD-1, LAG-3, and TIM-3, were not detectable (Figure S23). However, all four became upregulated upon co-culture, with the peak of the expression seen on day 3, followed by a decrease for all but CD39 on day 6 (Figures 6A and 6B). While EZH2i significantly increased the expression of PD-1 and, to a much lesser degree, of LAG-3 and even less of CD39, it exerted almost no impact on TIM-3 expression.

Figure 6.

Effect of EZH2i on expression of key checkpoint receptors by CART-ROR1 cells

(A) Flow cytometry-detected expression of the depicted immune checkpoint receptors by CART-ROR1 cells after 3-day co-culture with index MCL Jeko-1 cells at the depicted E:T ratios in the presence or absence of EZH2i tazemetostat (2 μM)/EZH2i & EZH2i: the inhibitor applied in the MCL cell pretreatment and present in the co-culture.

(B) Schematic diagrams of expression of the depicted immune checkpoint receptors calculated from the profiles shown in (A) and in Figure S23B. p values: ∗, <0.05; ∗∗, <0.01; ∗∗∗∗, <0.0001.

EZH2 inhibition augments fitness and promotes cytotoxic phenotype of CD8⁺ CART-ROR1 cells

To explore if CART-ROR1 cell exposure to MCL cells and EZH2i affects their phenotype beyond expression of checkpoint receptors, we analyzed these effector T cells at the baseline and after 1-, 3-, and 6-day culture by flow cytometry, using a 14-color T cell-focused antibody panel. The results, summarized in Table S1 with the key results displayed in Figure S24, identified several changes in CART-ROR1 cell viability and phenotype induced by specific culture conditions. Accordingly, EZH2i alone sustained survival of the cells, preferentially of their CD8-expressing subset (Figure S24, days 3 and 6). Co-culture with Jeko-1 cells induced prompt and sustained expression of CD25 (IL-2Rα) and primarily stimulated a CD8⁺ subset, regardless of the presence or absence of EZH2i. The Jeko-1/EZH2i combination induced the expression of the cytotoxic markers CD56 and CD57 in the CD8⁺ subset (Table S1; Figure S24, day 6). These results indicate that CD8⁺ T cells within the CART-ROR1 cell population are the key effectors in eliminating target cells and that EZH2i supports their survival and cytotoxic capacity.

EZH2 inhibition boosts synthesis of CART-ROR1 cell-activating immune soluble factors

To examine the impact of EZH2 on synthesis of the factors, we analyzed the MCL/CART-ROR1 cell co-culture supernatants for the expression of 30 cytokines, growth factors, and chemokines. As shown in Figure S25A for Jeko-1 and Figure S25B for Maver, and summarized in Table S2, co-culture with CART-ROR1 but not “mock” T cells promoted the synthesis of the vast majority of the secreted factors examined (27 for Jeko-1 and 24 for Maver). Among the highly boosted factors were MIP-1α (CCL3) and MIP-1β (CCL4). These chemokines are produced by a spectrum of immune cell types, including activated T and B lymphocytes, and attract and stimulate T lymphocytes and other cells to promote immune responses.²⁴ Importantly, EZH2i significantly increased synthesis of both these stimulatory chemokines, with the profiles indicating that they were produced by Jeko-1 and CART-ROR1 but not Maver and “mock” T cells.

Among the other factors markedly increased by EZH2i in both Jeko-1 and Maver were IL-2 and IL-6, the cytokines that stimulate T cell growth and activity; IFN-γ, which promotes T cell cytotoxicity and Th1 polarity; and IL-13, potentially contributing to Th2 differentiation.^25,26 EZH2i also significantly enhanced synthesis of the chemokines IP-10 (CXCL10) and MIG (CXCL9) and the cytokine MCP-1, all of which are involved in attracting T lymphocytes and other immune cells.²⁴ There were also EZH2i-mediated increases of IL-10, a cytokine with multiple functions,²⁷ including activation of tumor-infiltrating cytotoxic CD8⁺ T cells and of IL-5, known mostly for inducing eosinophilia but also able to enhance anti-tumor responses.²⁶ Finally, EZH2i enhanced the synthesis of VEGF, adding to the plethora of pro-inflammatory and immune response-boosting functions promoted by EZH2i in the context of CART-ROR1 cell activity against MCL target cells.

EZH2 inhibition potentiates CART-ROR1-mediated suppression of tumor growth

To determine the effect of EZH2i on CART-ROR1 function in vivo, we developed xenotransplant NSG mouse models from the ROR1 high-expressing Maver and low-expressing MCL-RL cells. While both types of MCL tumors rapidly grew in the presence of “mock” T cells, CART-ROR1 cells profoundly suppressed tumor growth (Figures 7A and 7B). The tumor growth was further significantly suppressed by the addition of EZH2i; this suppressive effect was more pronounced for MCL-RL tumors, consistent with the additional direct inhibitory effect of EZH2i, identified in the in vitro studies (Figure 2D).

Figure 7.

Impact of EZH2 inhibition on the activity of CART-ROR1 cells in vivo

(A) Effect of EZH2i on growth of Maver tumors. NSG mice inoculated intravenously (i.v.) with 5 × 10⁶ click-beetle green luciferase (CBGL) gene-transduced Maver cells were treated for 7 days BID with either EZH2i tazemetostat at 250 mg/kg or the drug vehicle. Next (day 0 in the graph), the mice were examined using bioluminescence (BLI) imaging and injected i.v. with 1 × 10⁶ of CART-ROR1 or “mock” T cells from the same donor with continuation of the EZH2i applications. p value: ∗∗∗, <0.0005.

(B) Impact of EZH2i on growth of MCL-RL tumors. NSG mice inoculated i.v. with 5 × 10⁶ CBGL gene-carrying MCL-RL cells were treated for 7 days BID with either 250 mg/kg of tazemetostat or the drug vehicle. On day 0, the mice were examined by BLI imaging and injected i.v. with 5 × 10⁵ CART-ROR1 or same-donor “mock” T cells with continuation of the EZH2i therapy. p values: ∗, <0.05; ∗∗, <0.005; ∗∗∗, <0.0005.

(C) Microscopic images (400×) of representative mouse tissues visualized by H&E stain or the listed immunohistochemical stains to highlight MCL-RL cells (CD20 and cyclin D1) and CART-ROR1 cells (CD3, CD4, CD8, and TIA1).

(D) Schematic summary of identified CART-ROR1 cell subsets in livers. p, proliferating (Ki-67⁺) cells; np, non-proliferating (Ki-67⁻) cells; c⁺, CTLA4-expressing cells; c⁻, CTLA4-negative cells.

(E) Schematic summary of identified CART-ROR1 cell subsets in spleens. p, proliferating (Ki-67⁺) cells; np, non-proliferating (Ki-67⁻) cells; c⁺, CTLA4-expressing cells; c⁻, CTLA4-negative cells.

Immunophenotypic profiling of CART-ROR1 cell infiltrates of tumor tissues

We next performed a comparative histologic and immunophenotypic analysis of lymphocytic infiltrates in the livers and spleens of mice injected with the MCL-RL cells and treated with either “mock” T cells, CART-ROR1 cells, or the combination, harvested at the end of the experiment on day 42 (Figure 7C; representative images). This analysis confirmed a high MCL tumor burden in Mock T cell-treated mice, a much lower burden in CART-ROR1 cell-treated mice, and the lowest burden in CART-ROR1 cell-/EZH2i-treated mice, as detected by H&E staining and the expression of CD20 and cyclin D1. While livers of “mock” T cell-treated mice showed very few CD3⁺ T cells, including CD4⁺ and CD8⁺ subsets as well as cells expressing the TIA-1 marker characteristic for activated cytotoxic CD8⁺ T cells, such cytotoxic CD8⁺ T cells were frequent in livers from CART-ROR1 cell-treated mice and somewhat less frequent at this late therapeutic stage in the mice receiving the CART-ROR1 cell + EZH2i combination. Spleens displayed qualitatively similar results to livers but with a much higher accumulation of CD3⁺, CD8⁺, and TIA-1⁺ cells in CART-ROR1-treated mice and relatively much lower accumulation of these cytotoxic T cells in CART-ROR1 cell-/EZH2i-treated mice, suggesting, for both, a direct correlation between the frequency of residual tumor cells and CART-ROR1 cells.

To characterize the CART-ROR1 cell populations from MCL-RL tumors in greater detail, we prepared a tissue microarray (TMA) from the selected T cell-enriched areas of the livers (Figure 7D) and spleens (Figure 7E) of the mice treated with CART-ROR1 alone or CART-ROR1/EZH2i combination and immunophenotyped them using sequential immunofluorescence. In addition to CD3 and T cell subset-specific CD4 and CD8 antibodies, we applied antibodies to examine T cell markers of functional polarization (T-bet, CCR4, FoxP3, CCR6, and IL-23R) and inhibition (CTLA-4), as well as CART-ROR1 cell proliferation status (Ki-67). While the study confirmed the predominance of cytotoxic CD8⁺ T cells over helper CD4⁺ T cells, it revealed a number of additional features of the CART-ROR1 T cells. First, the overall T cell subset profiles were qualitatively similar for CART-ROR1 cells alone vs. CART-ROR1 cell/EZH2i combination, with an exception of a relatively small population of T helper 17 (Th17: CD3⁺, CD4⁺, CCR6⁺, and IL-23R^+/−) cells seen in the former but not the latter. Second, there were frequent T cells, which did not express either CD4 or CD8. Third, Th1 (CD3⁺, CD4⁺, and T-bet⁺), Th2 (CD3⁺, CD4⁺, and CCR4⁺), and regulatory T (Treg: CD3⁺, CD4⁺, and FoxP3+) cells were essentially absent. Fourth, all identified T cell subpopulations (CD8⁺, CD4⁻/CD8⁻, and Th17) contained both non-proliferative (Ki-67⁻) and proliferative (Ki-67⁺) components, with the latter being usually more frequent. Finally, only small subsets of the T cell subpopulations expressed CTLA-4.

Discussion

ROR1 has emerged as an attractive therapeutic target in a broad spectrum of cancers, including lymphoid malignancies.^1,2,5,6,7 Although small-molecule inhibitors of ROR1 have been identified^28,29 and newer compounds have also been developed,^30,31,32 the inability of purified ROR1 to directly utilize ATP suggests that ROR1 may be enzymatically inactive. While ROR1 degraders show preclinical efficacy^33,34; their clinical usefulness remains to be established. Several immunotherapeutic approaches have also been developed, including immunotoxin conjugates,^8,9,10,11 bispecific antibodies,^6,7,12 and CART cells.^13,14,15,16 While the CART-ROR1 cells, clinically, have shown no toxicity to normal tissues,^13,15,16 their efficacy in cancer patients has proven limited,¹⁶ indicating that combinatorial therapies are needed.

In this study, we highlight the widespread yet heterogeneous expression of ROR1 among diverse malignancies and show that the inhibition of EZH2 and, to a much lesser degree, of DNMT1 and G9A, promotes ROR1 expression by malignant cells. The inhibition also impairs growth of some malignant cell populations. Elimination of ROR1-expressing malignant cells by CART-ROR1 cells is very effective and can be profoundly enhanced by addition of EZH2i, especially against cells with low ROR1 expression and at low E:T ratios. While EZH2i augments expression of immune checkpoint receptors by CART-ROR1 cells, it also supports their survival and cytotoxic phenotype and markedly increases synthesis of immune soluble factors.

Our searches of the publicly available mRNA expression datasets not only confirm the frequent ROR1 expression across hematopoietic and non-hematopoietic malignancies examined but also highlight marked quantitative differences in the expression of the ROR1-encoding mRNAs within given cancer types. We expanded this latter observation in a group of lymphomas by demonstrating that ROR1 expression ranges from non-detectable to dim, moderate, and bright. Strikingly, we also found that malignant cells can express ROR1 at levels below the detection thresholds of flow cytometry and western blotting, yet are still recognized by CART-ROR1 cells, leading to ROR1-specific elimination of these malignant cells, potentially significantly expanding the population of cancer patients who may benefit from CART-ROR1 cell immunotherapy. A similar observation was recently made by others in regard to CD70 antigen, considered a potential CART-cell therapeutic target in solid tumors.³⁵ Our other key finding is that the inhibition of epigenetic silencers, foremost of EZH2, promotes ROR1 expression, including in some malignant cells, with flow cytometry non-detectable baseline ROR1 expression. This finding may also have translational implications because the density of targeted receptor on malignant cells directly corresponds with the effectiveness of CART cell therapy,^36,37 and the loss or diminished expression of the target is one of the main mechanisms of acquired resistance to CART cells.^38,39

Our findings indicate that malignant MCL cells, rather than CART-ROR1 cells, are the main direct target of EZH2 inhibition. In addition to boosting their ROR1 expression, the inhibition directly impairs the growth of some MCL populations, potentiating the therapeutic effect of the CART-ROR1 cell + EZH2i combination. The critical impact of pretreating malignant cells with EZH2i in enhancing their elimination by CART-ROR1 cells further strengthens this conclusion. Recent studies by others on the effect of EZH2i on the activity of CART cells targeting the conventional B cell-specific receptor CD19 (CAER-CD19)^40,41 found no appreciable effect of pretreatment of such cells, in contrast to pretreatment with EZH2i of target cells, providing yet another piece of evidence that EZH2i acts primarily by enhancing the immunogenicity of malignant cells. However, direct positive effects of EZH2i on immune T cell function were also noted, using in vitro and murine tumor models.^40,42 Accordingly, EZH2 inhibition was found to reduce the emergence of Treg cells, promote the development of memory CD8⁺ T cell phenotype, augment CART cell recruitment and interaction within the tumor microenvironment, and reduce CART cell exhaustion.⁴⁰ It also enhanced the anti-tumor activity of native immune T cells, including promoting expression of the self-renewal T cell transcription factor TCF1.⁴²

Noteworthy, an analysis of T cells from lymphoma patients treated with CART-CD19 cell + EZH2i combination showed an increase in CD8⁺ T cells with naive and memory phenotypes and a decrease in terminally differentiated CD4⁺ and CD8⁺ cells, directly correlating with the effectiveness of CART cell therapy.⁴³ These findings combined with overall 100% response rate in thirteen patients, including complete remission in ten (77%), and no excess toxicities as compared with CART-CD19 cell therapy alone, support the notion that the CART cell + EZH2i combination may also emerge as a therapy against neoantigens such as ROR1, superior to CART cell monotherapy.

While our results indicate that EZH2 inhibition mediates ROR1 expression in malignant cells, this occurrence is not universal, given that the selected DLBCL and plasma cell myeloma cell lines treated with EZH2i failed to express this neoantigen. The data also suggest that EZH2i augments preexisting ROR1 expression, in some cases, expression that is below the level of detection by standard immunophenotyping, rather than inducing de novo expression. Therefore, the development of a biomarker, such as the presence of EZH2 and/or H3K27me3 at the ROR1 promoter, may be required to prospectively identify cancer patients with non-detectable tumor ROR1 expression at baseline who may benefit from CART-ROR1 cell + EZH2i combination therapy. There may also be a concern that EZH2i may upregulate ROR1 expression in normal tissues, resulting in tissue damage by CART-ROR1 cells. Our findings that EZH2i-triggered ROR1 induction is selective among malignant cells and does not occur either in peripheral blood T cell-rich immune cells or in CART-ROR1 cells diminish to a large degree but do not fully eliminate such concern. Carefully designed and monitored clinical trials may be needed to directly address this issue. Usage of CART cells targeting only highly expressed ROR1 (expected in malignant but not normal cells) together with a second, highly expressed tumor-associated antigen (e.g., CD70 also upregulated by EZH2i³⁵) may be a consideration, should highly sensitive CART-ROR1 against ROR1 and the other antigen combined with EZH2i show toxicity against any normal tissue.

Recently, an association between tazemetostat therapy and excess risk of the development of myeloid malignancies has been noticed based on the results of the Symphony-1 clinical trial, leading to a withdrawal of this drug by Ipsen.⁴⁴ However, this association has not been seen with three other EZH2i used as single agents in clinical trials, which are ongoing in an uninterrupted manner, suggesting that the risk might be an off-target effect of tazemetostat. Furthermore, Symphony-1 was based on comparing tazemetostat/lenalidomide/rituximab triple-agent therapy with placebo. Lenalidomide increases the risk of myeloid malignancies,^45,46 presumably due to its immunosuppressive properties that permit outgrowth of spontaneously developed or therapy-induced malignant clones, with an exception of del 5q myelodysplatic syndrome where the drug shows therapeutic benefit.⁴⁷ It is, therefore, possible that the tazemetostat + lenalidomide combination, rather than tazemetostat alone, resulted in the excess risk identified. Accordingly, EZH2 inhibitors including tazemetostat augment, rather than suppressing immune response, as shown here and by others^40,41,42,43; so, such clones may not progress to an overt disease in the absence of lenalidomide. Finally, in contrast to the application of tazemetostat in follicular lymphoma in multiple 28-day cycles, therapy with EZH2i to boost CART cell effects could be short, substantially limiting any potential pro-oncogenic effect of the drug. All the above considerations suggest that benefits of the CART cell + EZH2i combination may well outweigh the potential risks.

We found the expression analysis results of key checkpoint receptors PD-1, TIM3, LAG-3, and CD39 by CART-ROR1 cells rather surprising. While these receptors with exception of CD39 were not expressed at the baseline and became strongly expressed upon co-culture with malignant target cells including an additional increase by EZH2i, foremost for PD-1 and, to lesser degree, TIM-3, their expression did not seem to adversely affect CART-ROR1 cell function, suggesting their limited inhibitory activity, at least in the experimental setting of short-term in vitro co-culture with target cells. Our findings that EZH2i supports viability of CD8⁺ CART-ROR1 cells and fostered their expression of CD56 and CD57 cytotoxic markers provide further evidence of the EZH2i-enhanced functional fitness of CART-ROR1 cells. Moreover, EZH2i had a profound boosting effect on the synthesis of cytokines, chemokines, and growth factors. While these EZH2i-upregulated soluble mediators, in particular cytokines, may exert diverse effects^24,25,26 including promotion of immunosuppression,²⁵ they are also capable of enhancing T cell activity and stimulating immune response and inflammatory processes and, hence, of augmenting the effectiveness of CART-ROR1 cells.^24,25,26 Accordingly, IL-2 and IL-6 can promote the activation and expansion of both CD4⁺ and CD8⁺ T cells, and together with IFN-γ and IL-10²⁷ can stimulate the cytotoxic activity of the latter T cell subset, although IL-10 can also display immunosuppressive properties.²⁷ The EZH2i-mediated marked enhancement of the synthesis of chemokines MIP-1α/CCL3, MIP-1β/CCL4, IP-10/CXCL10, and MIG/CXCL9 and of MCP-1/CCL2 cytokine, all involved in attracting T lymphocytes and other immune cells, may further foster the elimination of malignant cells by CART-ROR1 cells. Others have seen EZH2i-triggered increases in IL-2, IL-6, GM-CSF, and soluble CD137 ligand in co-cultures of murine CART-CD19 cells with follicular lymphoma-like tFL-P6 cells.⁴⁰ We also observed a marked EZH2i-mediated increase in GM-CSF synthesis in CART-ROR1 cell co-culture with Jeko-1 but not with Maver cells, suggesting that the given cytokines may be produced in the context of specific malignant cells. While there could be a concern of EZH2i-driven increased risk of high-grade cytokine release syndrome (CRS), results of the above-mentioned clinical study with CART-CD19 cell + tazemetostat combination indicate no such increase.⁴³

Immunophenotypic analysis of the livers and spleens (the organs where MCL-RL cells formed tumors) collected at the end of the experiment showed similar findings regardless of the presence or absence of EZH2i therapy, with the rather small number of samples (n = 2–5 per group) representing a limitation of the analysis. There was marked predominance of CD8⁺ T cells over CD4⁺ T cells within the remaining CART-ROR1 cells, suggesting that the former T cell subset is mainly responsible for the elimination of malignant MCL cells. Expression of TIA-1, a marker of T cells with an active cytotoxic program, by the subset of T cells, further supports this conclusion. Multiplexed immunophenotyping additionally identified a relatively large population of CD3⁺ T cells that were CD4- and CD8-negative. This possibly arose by the downregulation of CD8 receptor upon immune response against the targeted malignant cells, as seen in normal CD8⁺ T cells to prevent excessive tissue damage and emergence of autoimmunity.²⁵ Many of the CD4⁻CD8⁻cells, as well as other identified T cell populations, expressed Ki-67, indicating cell proliferation and, hence, ongoing immune stimulation. Limited expression by the T cell subsets of the immunosuppressive receptor CTLA-4 further supported this conclusion. Among the overall sparse CD4⁺ T cells, there were negligible subsets of Treg, Th1, or Th2 cells. There was a relatively small Th17 population seen in the CART-ROR1 cells alone but not in the CART-ROR1 cell + EZH2i therapy setting; the significance of this finding is unclear, given the dual pro- and anti-cancer roles of such cells.

In summary, we demonstrate that CART-ROR1 cells are highly active against malignant target cells, even if they express ROR1 below the threshold of detection by flow cytometry. ROR1 expression on malignant cells and the related cytotoxic effectiveness of CART-ROR1 cells in eliminating ROR1-expressing malignant cells is markedly boosted by EZH2i. These findings may have important translational implications by increasing the population of cancer patients who may benefit from CART-ROR1 therapy as well as improving the efficacy of CART-ROR1 cells by combining them with EZH2i therapy.

Limitations of the study

While the study generated compelling evidence that the CART-ROR1 cell and EZH2i combination should be highly effective in therapy of ROR1-expressing malignancies, most experiments were performed in vitro. Our in vivo study was based solely on a short-term lymphoma xenotransplant model. While we provided several lines of evidence for ROR1 specificity of CART-ROR1 cells, documenting that ROR1-expressing malignant cells lose susceptibility to CART-ROR1 cells after ROR1 expression abrogation would provide additional evidence for ROR1 specificity of the malignant cell killing.

Resource availability

Lead contact

Requests for further information, resources, and reagents should be directed to and will be fulfilled by the lead contact, Mariusz A. Wasik (mariusz.wasik@fccc.edu).

Materials availability

Non-commercially available materials can be provided upon request. There is no associated code.

Data and code availability

• •Data reported in the paper will be shared by the lead contact upon request.
• •This paper does not report original code.
• •Any additional information required to reanalyze the data reported in this paper is available from the lead contact.

Acknowledgments

This study was supported, in part, by grants from the National Cancer Institute (R01CA228457), Lyell Immunopharma, Christopher M. Fuga Memorial Foundation, and Fox Chase Cancer Center Institute for Cancer Research.

Author contributions

S.W., C.L., and M.A.W. designed research; S.W., C.L., P.K.G., R.F, S.Z.-E., D.R., A.K., H.L., and K.N. performed research; B.S., F.A., D.N., J.B., D.A.B., M.J., J.F.-B., K.N., and M.A.W. analyzed the data; H.F. provided patients’ samples; B.B. and G.L. generated CART-ROR1 cells and critically reviewed the manuscript; and V.D. edited and M.A.W. wrote the manuscript.

Declaration of interests

The authors have filed a patent application PCT/US24/44753, entitled “Methods and compositions for increasing ROR1 expression.”

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
PE anti-human ROR1 Antibody	Biolegend	Cat #: 357804
FITC anti-human CD19 Antibody	Biolegend	Cat #: 363008
APC anti-human CD3 Antibody	Biolegend	Cat #: 317318
Brilliant Violet 421™ anti-human EGFR Antibody	Biolegend	Cat #: 352911
PE/Cyanine7 anti-human CD279 (PD-1) Antibody	Biolegend	Cat #: 367414
Brilliant Violet 605™ anti-human CD366 (TIM-3) Antibody	Biolegend	Cat #: 345018
Brilliant Violet 421™ anti-human CD223(LAG-3) Antibody	Biolegend	Cat # 369314
PE anti-human CD39 Antibody	Biolegend	Cat # 328208
ROR1 (D6T8C) Rabbit mAb	Cell Signaling	Cat #: 16540
EZH2 (D2C9) Rabbit Monoclonal Antibody	Cell Signaling	Cat #: 5246
Histone H3K27me3 antibody (pAb)	ACTIVE MOTIF	Cat #: 39155
Histone H3 (96C10) Mouse Monoclonal Antibody	Cell Signaling	Cat #: 3638
β-actin (13E5) Rabbit Monoclonal Antibody	Cell Signaling	Cat #: 4970
CD20	DAKO	GA60461-2
Cyclin D1	DAKO	GA08361-2
CD3	DAKO	GA50361-2
CD4	DAKO	IR64961-2
CD8	DAKO	GA62361-2
EZH2 and DNMT1 inhibitors
EZH2 inhibitor Tazemetostat (EPZ-6438)	Selleckchem	S7128
DNMT1 inhibitor (GSK3685032)	Selleckchem	E1046
G9A inhibitor (UNC0646)	Selleckchem	S7020
Biological samples
Primary human MCL cells and tissues	Stem Cell and Xenografts Core, University of Pennsylvania and residual diagnostic cells and tissues	N/A
CAR-ROR1 cells and Mock T cells	generated by Lyell Immunopharma	N/A
D17526 PLD05 CAR-ROR1	N/A	N/A
D17526 Mock	N/A	N/A
D31328 PLD05 CAR-ROR1	N/A	N/A
D31328 Mock	N/A	N/A
D47294 PLD05 CAR-ROR1	N/A	N/A
D47294 Mock	N/A	N/A
Critical commercial assays
RNeasy kit	Qiagen	Cat No./ID: 74004
SuperScript IV First-Strand Synthesis System	Thermo Fisher Sci	Cat No./ID: 18091050
PowerUp SYBR™ Green Master Mix	Thermo Fisher Sci	Cat No./ID: A25742
MTT Promega	Promega	Cat# 11465007001
Invitrogen Cytokine 30-Plex Human Panel	Thermo Fisher Sci	LHC6003M
Experimental models: Human cell lines
Jeko-1	ATCC	N/A
REC-1	ATCC	N/A
Maver	a gift from Dr. Lynn Wang	N/A
MCL-RL	developed in MA Wasik’s laboratory	N/A
Mino	ATCC	N/A
SP49	a gift from Dr. Raymond Lai	N/A
Granta	a gift from Dr. Lynn Wang	N/A
SUDHL-4	a gift from Dr. Yibin Yang	N/A
SUDHL-6	a gift from Dr. Yibin Yang	N/A
RIVA	a gift from Dr. Yibin Yang	N/A
LY-1	a gift from Dr. Yibin Yang	N/A
LY-8	a gift from Dr. Yibin Yang	N/A
LY-18	a gift from Dr. Yibin Yang	N/A
SC-1	a gift from Dr. Yibin Yang	N/A
KMS11	a gift from Dr. David Allman	N/A
OVCAR3	a gift from Dr. Denise Connolly	N/A
Experimental models: Organisms/strains
Mouse: NOD-SCID-ƴc−/− (NSG)	Stem Cell and Xenografts Core, Univ. Pennsylvania	N/A
Primers	Forward Sequence (5′-3′)	Reverse Sequence (5′-3′)
ROR1	GAGGCAACCAAAACACGTCAGAG	GGCACACTCACCCAATTCTTCC
GAPDH	ACCCAGAAGACTGTGGATGG	TCTAGACGGCAGGTCAGGTC

Experimental model details

Patient-derived lymphoma cells and tissues

Jeko-1 and REC-1 cell lines were obtained from the American Type Culture Collection (ATCC). Maver and Granta cell lines were were obtained from Lynn Wang, University of Chicago, SP49 from Raymond Lai, University of Alberta, Edmonton, Canada, SUDHL-4, SUDHL-6, RIVA, LY-1, LY-8, LY-18, and SC-1 from Yibin Yang, Fox Chase Cancer Center, KMS11 from David Allman, University of Pennsylvania, and OVCAR3 from Denise Connolly, Fox Chase Cancer Center. MCL-RL cell line developed in MA Wasik’s laboratory fully preserves the phenotype, mutational profile, and DNA methylation pattern of the parental MCL cells, as described previously.⁴⁸ The cell lines were authenticated by flow cytometry based immunophenotyping and by STR DNA profiling method performed by LabCorp, using PowerPlex 16 HS system (Promega) with the generated STR profiles comparatively analyzed by employing The Cellosaurus STR Similarity Search tool. They were cultured in the 10% FBS RPMI1640 medium, and checked periodically to exclude mycoplasma contamination, as described previously.^48,49,50 Viably frozen primary MCL cells were obtained from cell collection bank of the Stem Cell and Xenografts Core, University of Pennsylvania. To enhance survival of the primary cells in culture, we supported them with “feeder” stromal bone marrow cells, as we have done before.⁵¹ Fresh lymphoma cells were analyzed for ROR1 expression as a part of flow cytometry-based diagnostic immunophenotyping. Tissue studied by IHC were on the left-over formalin-fixed, paraffin-embedded lymphoma tissues, also collected to establish diagnosis. The studies were performed under the Fox Chase Cancer Center IRB-approved protocol (ID: 22-9065MODCR12).

CART-ROR1 cells

Cryopreserved CD4⁻or CD8-positively selected healthy donor-derived T cells were sourced from AllCells. CD4⁺ and CD8⁺ cells were thawed and mixed in a 1:1 ratio in OpTmizer T-cell expansion medium (Gibco) supplemented with human IL-2, IL-7 and IL-15 and activated using 1:100 TransAct (Miltenyi) in G-Rex plates (Wilson Wolf). 24 h after activation cells were mixed with predetermined titers of a lentiviral vector (Lyell Immunopharma) to achieve a 40–60% transduction rate. The lentiviral vector encodes for an MND promoter controlling expression of a ROR1-specific CAR, a P2A self-cleaving peptide, and a truncated human EGFR including domains III, IV, and transmembrane domain (EGFRt). The CAR consists of the anti-ROR1 R12 scFv, an IgG4 linker, a CD28 transmembrane domain, and the 4-1BB and CD3z signaling domains. Binding specificity of the ROR1 construct was examined using the Retrogenix assay (Charles Rivers Laboratories) and yielded negative results for off-target binding. T-cells were expanded for 7 days after activation before cryopreservation in 5% CryoStor (Stemcell Technologies).

Method details

Quantitative (q)RT-PCR

RNA was extracted (RNeasy kit; Qiagen) and reverse-transcribed using SuperScript IV First-Strand Synthesis System (Thermo Fisher). The relative target cDNA copies were quantified by real-time PCR with PowerUp SYBR Green Master Mix (Thermo Fisher Sci). The fold difference in RNA levels was calculated on the basis of the difference between C_T values (ΔC_T) obtained for the investigated and control mRNA.

Western blotting

Western blotting was performed using antibodies against ROR1, EZH2, H3K27me3, total H3 histone, total ß-actin (all from Cell Signaling Technology) and GAPDH (Proteintech), according to standard protocols.^20,52

Flow cytometry to detect ROR1 expression in patient-derived samples

Cell surface expression of ROR1 and CD5 in lymphoma patient-derived cells was analyzed using PE-conjugated anti-human ROR1 and APC-conjugated anti-human CD5 antibodies or control isotype-matched IgG antibodies (Biolegend, San Diego, CA). Flow analyses were done on an FACSCanto 10-color Flow Cytometer (Becton Dickinson Biosciences, San Jose, CA). Data were analyzed using FlowJo v10.8.0 software (TreeStar, Ashland, OR).

Cell count assay

MCL cells (2 × 10⁴ cells/well) were cultured in 96-well plates (Corning, Inc.) for 24 h with either Mock T cells or CART-ROR1 cells. The cells were suspended in trypan blue–containing PBS and microscopically counted using a hemocytometer.

Cell growth assay

Cells were plated in 96-well plates (3–5x10³ cells/well) in triplicates and labeled with 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT, Promega) at 5 mg/mL for 4 h and solubilized with 10% SDS in 0.01 M HCl. The optical density (O.D.) of the culture supernatant, corresponding to the MTT conversion-mediated change in supernatant color, was determined at 570 nm using a Titertek Multiskan reader (Titertek Instruments).

Multiparametric flow cytometry-based analysis of CART-ROR1 cell-MCL cell co-cultures

Antibodies for flow cytometry analysis were purchased from Biolegend (San Diego, CA), including PE anti-human ROR1, FITC anti-human CD19, APC anti-human CD3, Brilliant Violet 421 anti-human EGFR, PE/Cyanine7 anti-human CD279 (PD-1), Brilliant Violet 605 anti-human CD366 (TIM-3), Brilliant Violet 421 anti-human CD223(LAG-3), and PE anti-human CD39 Antibodies. Cells were washed once in PBS, supplemented with 2% fetal bovine serum and stained for 15 min at room temperature. In all analyses, the population of interest was gated based on forward vs. side scatter characteristics followed by singlet gating, and live cells were gated using Propidium Iodide (PI) solution. Time gating was included for quality control. Flow cytometry was performed on a FACSymphony A5 flow cytometer (BD BioSciences) and analyzed with FlowJo v10.8.0 software. All tests were performed in triplicates to permit statistical analysis and assure validity of the findings. Relative target cell survival was calculated as the percentage of viable CD19⁺ cells in co-culture normalized to the corresponding tumor-only control, which was set to 100%. Specifically, relative survival was calculated as: (viable CD19⁺ cells in co-culture/viable CD19⁺ cells in tumor-only control) ×100. Statistical significance was determined using two-way ANOVA with Tukey’s multiple-comparisons test on log-transformed data. Residual normality and homogeneity of variance were assessed using the Shapiro–Wilk and Brown–Forsythe tests, respectively. p values <0.05 were considered statistically significant.

ChIP-qPCR

Cells were treated with EZH2 inhibitor tazemetostat at 2μM or the drug’s vehicle for 5 days, harvested for chromatin immunoprecipitation (ChIP) using ChIP-qPCR kit from the DiagenodeiPeal, according to the manufacturer’s protocol. Briefly, cells were cross-linked, lysed, and chromatin was sheared to an average size of approximately 100–500 bp. Sheared chromatin was immunoprecipitated with antibodies against EZH2 (Cell Signaling Technology: D2C9/#5246) or H3K27me3 (Active Motif: #39155), with normal IgG used as a negative control. After reversal of cross-linking and DNA purification, the enriched DNA fragments were analyzed by quantitative PCR using primers targeting the ROR1 promoter region (Forward primer: 5′-AAAGTTAAACCAGGAGAAATGG-3′, Reverse primer: 5′-CTGGCAAGATTTATTGAGGATT-3′). ChIP-qPCR results were calculated as fold enrichment relative to IgG control and normalized to input chromatin.

Cytokine concentrations measurements

Effector and target cells were co-incubated at a 0.2:1 ratio in media for 72h. Supernatant was harvested and analyzed by 30-plex Luminex array (Luminex Bio-Rad Bio-Plex 200) according to the manufacturer’s protocol (Invitrogen, Liftechnologies, LHC6003).

Immunohistochemical of MCL and DLBCL tissue analysis

mmunohistochemical staining of the patient-derived MCL and DLBCL formalin-fixed, paraffin-embedded lymphoma tissues was performed as previously described.⁵¹ In brief, the slides were heat-treated for antigen retrieval and incubated with IHC-validated antibodies against EZH2 and H3K27me3 (both from Cell Signaling, Cat. No. 5246 and 9733, respectively). Tissues from xenotransplant MCL-RL tumors were stained with standard primary antibodies from Dako against CD20 (B-cell marker), Cyclin D1 (CCND1; pathognomonic MCL marker) and T-cell markers: CD3, CD4. CD8, and TIA1. Hematoxylin-eosin (H&E)-stained slides were generated for histologic examination. The stained slides were evaluated using light microscopy, and the images were acquired using a microscope attached to a camera (Leica). The results were scored as negative (no staining: score 0), weak staining (score 1), and moderate to strong staining (scores 2 and 3, respectively).

In vivo xenograft mouse model

Animal studies were performed under the protocol #803466 approved by the University of Pennsylvania Institutional Animal Care and Use Committee and performed in the University’s facility dedicated to immunodeficient mice. MCL-RL and Maver cells were injected into NSG male mice, 4–6 weeks of age. Twenty mice were injected with each of the cell lines and received subsequently Mock T cells (five mice) or CART-ROR1 cells (either alone: seven mice or in combination with EZH2 inhibitor: eight mice; please see the section below for more details).

In vivo tumor engraftment, treatment, and bioluminescence imaging

For both the MCL-RL and MAVER MCL models, 5×10⁶ cells transduced in vitro with Click-beetle green protein gene were inoculated into NSG mice via lateral tail vein injection on Day −7 to allow systemic engraftment. Treatment with EZH2i was initiated on Day −7 at a dose of 250 mg/kg, administered twice daily (BID) by oral gavage, and continued throughout the duration of the study. On Day 0, tumor-mice bearing MCL-RL cells received a single injection of 5×10⁵ CART-ROR1 or matched Mock T cells.

Sequential immunofluorescence staining and tissue imaging

Unless otherwise specified, all reagents used in the COMET instrument were proprietary to Lunaphore Technologies. Representative regions of formalin-fixed, paraffin-embedded murine livers and spleens assembled into a tumor microarray, as well as individual human tonsil tissues used for antibody optimization and as a positive control, were dewaxed in xylene and rehydrated through a graded ethanol series. Heat-induced epitope retrieval was performed using the 2100-Retriever apparatus (Aptum Biologicals, UK) with R-Universal Epitope Recovery buffer (Cat. #62719–20, Electron Microscopy Sciences), according to the manufacturer’s instructions. Tissue slides were then rinsed twice in PBS and maintained in Lunaphore’s Multistaining Buffer (MSB) until loaded into the COMET PA automated platform. Slides were processed using COMET’s sequential immunofluorescence (SeqIF) technology. Each tissue section underwent ten full SeqIF cycles, each cycle comprised detection of one to two biomarkers, image acquisition, and removal of primary and secondary antibody complexes by an elution step using Lunaphore’s Elution Buffer, as previously described.⁵³ All Lunaphore reagents and imaging chips were prepared and used according to the manufacturer’s instructions.

We designed a 11-plex marker panel using the following antibodies. From Lunaphore Technologies-Biotechne (Switzerland), anti-CD3 [clone LUN3] (cat. # MR10010), anti-CD8 [clone C8/144B] (cat. # MR10030), anti-CD4 [clone BL-155-1C11] (cat. # MR10020), anti-Ki67 [clone BLR021E] (cat. # MR10110). From Sigma-Prestige (MA, USA) anti-CCR4 (cat. # HPA031613). From Proteintech (IL-USA) anti-T-bet (cat. # 13700-1-AP) and anti-IL-23R (cat. # 27163-1-AP). From Santa Cruz (CA, USA) anti-HLA-DR (cat, # sc-53319). From Abcam (MA, USA) anti-CTLA-4 [clone CAL49] (cat. # MA5-16365). From R&D Systems (MN, USA) anti-CCR6 [clone 53103] (cat. # MAB195-100). From Invitrogen-Thermo Fisher (MA, USA) anti-FoxP3 [clone SP97] (cat. # MA5-16365) and nuclear counterstaining (DAPI, cat. #62248). Prior COMET processing all tissue sections were blocked using a commercial buffer (Section Block, Cat#62711, Electron Microscopy Sciences). Primary antibodies were diluted in MSB, while secondary antibody solutions were prepared in signal staining diluent (Cat#8112L, Cell Signaling Technology). Additionally, COMET’s proprietary blocking buffer was applied during the automated cycles before secondary antibody incubations. Before incorporating into the panel, each antibody was individually optimized using human tonsil or lymph node tissue as a positive control to determine the appropriate concentration, incubation time, and optimal elution conditions. Proper negative controls (e.g., secondary-only cycles) and elution-confirmation steps were included. Additional human lymph node tissue in the TMA was used as internal reference for all markers (not shown). After the final antibody panel was established, a control murine tissue sample was processed with the panel to confirm that anti-human antibodies did not cross-react with mouse antigenic epitopes. All antibody incubations and elution steps were performed at 37°C. Primary antibodies, applied singly or in pairs, were dynamically incubated for 4–8 min, while elution lasted 8 min per cycle. Secondary antibodies were purchased from Jackson ImmunoResearch: Goat anti-mouse Cy3, (Cat# 115–166-146) and Goat anti-rabbit Alexa Fluor 647 (Cat# 111–606-144) were co-incubated with DAPI for 2 min. Imaging was performed using the COMET system’s built-in epifluorescence microscope using DAPI-based autofocus. A 9.1 × 9.1 mm area was imaged using a 20× objective. Monochromatic 12-bit images were acquired with the following filter/exposure settings: DAPI (378/432 nm, 80 ms), TRITC (543/595 nm, 425 ms), and Cy5 (635/730 nm, 225 ms). Tissue autofluorescence (AF) was recorded for each channel prior to marker detection. The final output consisted of a multi-layer OME-TIFF file comprising 289 stitched and aligned fields of view. The file containing 12-bit monochromatic images, one per marker plus DAPI, was postprocessed using each channel’s corresponding AF images for digital background subtraction.

Quantification and statistical analysis

Statistical analysis of the in vitro studies

In each Figure, the analyzed data are presented as the mean ± standard error of the mean (SEM). Unless stated otherwise, two-tailed Student’s t-tests assuming variance homogeneity were used to calculate p values, and α = 0.05 was considered to indicate statistical significance. Key studies were recalculated using.

Statistical analysis of the in vivo studies

Tumor burden was quantified as total photon flux (photons/sec) within defined regions of interest, and values were tracked longitudinally for response assessment. Tumor response was calculated from longitudinal bioluminescence values for each animal. Mean and SEM values were calculated using standard propagation-of-error formulas applied to the individual animal measurements. The displayed statistical comparisons between groups were performed using two-sided, 2-sample Wilcoxon rank sum test.

Image analysis of sequential immunofluorescence staining

For image interrogation, we used the AI pre-trained Phenoplex workflow from the digital image analysis platform Visiopharm (V2025.02.1.18631 × 64). Briefly, cells were detected using pre-trained deep-learning-based nuclear detection (DAPI channel) and cell segmentation algorithms. Unique pixel intensity thresholds per marker channel were established through experienced human visual assessment and confirmation for each tissue cohort (liver or spleen) and applied consistently across the corresponding images to discriminate background from positive signal. Additionally, we excluded excessively bright artifacts. Depending on marker localization, pixel signals were assessed within nuclear or cytoplasmic compartments for each detected cell. Single-cell proteomics and spatial localization data per image were subsequently exported as TSV files for automated analysis using our Spatial Proteomics Analysis (SpPrAn) tool. To analyze the captured images, we used Spatial Proteomics Analysis (SpPrAn), a state-of-the-art bioinformatics pipeline to analyze data from antibody-based protein detection at subcellular resolution, exported from digital image analysis platforms (e.g., VISIOPHARM). SpPrAn adapts domain experts’ prior knowledge and quantifies pre-defined cell types from protein abundance profiles (e.g., positive or negative) and reconstructs their spatial distribution across the imaged tissue using each cell’s registered coordinates. Our pipeline further visualizes spatial distribution by generating graphical plots for those cell types. SpPrAn pipeline is available at https://github.com/comp-epi-lee-lab/spatial_proteomics/. The SpPrAn pipeline was implemented in Python 3.10 using the libraries: anndata (v0.10.8),⁵⁴ matplotlib (v3.9.1),⁵⁵ numpy (v2.0.0), pandas (v2.2.2),⁵⁶ and squidpy (v1.5.0).⁵⁷ For more details, see the pipeline’s GitHub repository. The dataset analyzed consisted of duplicate samples from murine liver (L) and spleen (S) specimens (see Material and Methods). Samples were grouped into two experimental conditions: CAR-T-treated (4L and 3S), and CAR+EZH2i-treated (2L and 2S). Using discrete marker profiles defined by the presence (value = 1) or absence (value = 0) of 11 protein markers, we defined 6 main immune cell types as follows: cytotoxic CD8 T cells [CD3⁺, CD8⁺, CD4⁻]; T helper subsets Th1 [CD3⁺, CD8⁻, CD4⁺, T-bet+, CCR4-, CCR6-], Th2 [CD3⁺, CD8⁻, CD4⁺, Tbet-, CCR4+, CCR6-], and Th17 [CD3⁺, CD8⁻, CD4⁺, Tbet-, CCR4±, CCR6+, IL23R±]; T regulatory cells (Treg) [CD3⁺, CD8⁻, CD4⁺, FoxP3+]; double-negative T cells (DN) [CD3⁺, CD8⁻, CD4⁻]. For all populations, Ki67 and CTLA-4 positivity was assessed. The SpPrAn pipeline included a pre-processing step that filtered out cell-segmentation artifacts producing erroneous readouts (values −1) detected during the marker-positivity screening, thereby removing defective objects (cells). We leveraged the domain expert-defined cell-types as phenotypes to label every cell in each sample, then calculated the proportion of each T cells sub-phenotype relative to the total number of detected CD3⁺ cells.

Published: September 25, 2026