CD4 depletion therapy but not ICB immunotherapy increased CAR T cell efficacy against B16F10 melanoma

We previously reported that second-generation anti-TRP1 (TA99) CAR T cells had no therapeutic effect as a single-line therapy in the immunocompetent, non-immunogenic B16F10 melanoma model without preconditioning regimens.8 Armoring TA99 CAR T cells with Super2 and IL-33 cytokines, however, significantly prolonged overall survival but only delayed tumor growth by approximately 1 week (figure 1A–D). Using live luminescence imaging to track CAR T cell localization and abundance, we found that the loss of S233-armored CAR T-cell signal coincided with the loss of tumor control (online supplemental figure 1). This led us to determine whether inhibitory receptors limit CAR T cell responses. However, there was no difference in B16F10 tumor growth or overall survival when S233-armored CAR T cells were combined with anti-PD-L1, -CTLA-4, or -VISTA blockade (online supplemental figure 2). While CD4 T cells can enhance antitumor immunity, immunosuppressive subsets including Tregs and Tr1 cells can limit activation of tumor-specific CD8 T cells.23 24 To determine whether CD4 T cells limited CAR T cell responses, we treated B16F10 tumor-bearing mice with anti-CD4 antibody 3 days prior to and following transfer of unarmored TA99 or S233-armored TA99 CAR T cells (figure 1A). Consistent with previous reports, the depletion of CD4 T cells delayed tumor growth by 27% compared with untreated mice (no treatment, NT) (figure 1B,C). In contrast, the transfer of unarmored TA99 CAR T cells with or without CD4 depletion had no additional effect on tumor growth in comparison to NT or anti-CD4 treated mice, respectively. S233-armored CAR T cells delayed tumor growth by 44%, similar to or modestly better than CD4 depletion alone (figure 1B,C). The combination of S233-armored CAR T cells and CD4 depletion yielded the best tumor control, delaying tumor growth by 140% and enhancing overall survival compared with other groups (figure 1B–D), highlighting the importance of eliminating the suppressive CD4 T cells in improving CAR T cell efficacy.

Figure 1

Effects of unarmored and S233-armored chimeric antigen receptor (CAR) T cells on B16F10 tumor growth and survival in the presence or absence of CD4 T cells. (A) Experimental schematic. TA99 CAR T cells with or without Super2 IL-33 (S233) armor (7×106) were transferred into B16F10 tumor-bearing mice (3×105) on day eight post-tumor inoculation. Some groups received 200 µg αCD4 depleting antibody on days 5 and 11. (B) Individual and average tumor volumes measured every other day starting day six post-tumor inoculation. Data are plotted as mean+SD. Two-way analysis of variance analysis with Tukey post-hoc test. ****p<0.0001. (C) Tumor growth delay resulting from CAR T cell therapy. (D) Kaplan-Meier curve depicting survival probability and analyzed using the log-rank test. **p<0.01. (B–D) Data are from 1 of 2 representative experiments with 3–6 biological replicates per experiment.

S233-armored CAR T cells exhibit a highly activated effector phenotype compared with more stem-like unarmored CAR T cells

When used in a neoadjuvant setting, depletion of CD4 T cells prior to B16F10 surgical resection leads to the differentiation of tumor-specific memory T cells that provide protection against B16F10 re-challenge.21 This prompted us to determine the potential of TA99 CAR T cells to adopt a memory precursor phenotype 1-week post-transfer with or without CD4 depletion (figure 2A). Unarmored and S233-armored TA99 CAR T cells were found in the TDLN in similar overall numbers and largely maintained an MPEC phenotype defined by CD127+KLRG1− expression, comparable to their phenotype prior to transfer (figure 2B–D, online supplemental figure 3A,B). CD4 depletion did not significantly affect CAR T cell numbers in the TDLN but resulted in a small decrease in MPEC proportions of S233-armored TA99 CAR T cells (figure 2D).

Figure 2

Unarmored TA99 chimeric antigen receptor (CAR) T cells resemble memory precursor effector cells (MPECs) while S233-armored TA99 CAR T cells appear hyperactivated. (A) Experimental schematic. Tumor-draining lymph nodes (TDLNs) were collected on day 15 post-tumor inoculation. (B–C) Flow cytometric analysis of CD45.1+ CAR T cells. (B) Representative flow plots and (C) Counts of CD45.1+ CAR T cells, gated on CD8+CD44+. (D) Representative flow plots and proportions of CD45.1+ CAR T cell MPECs (defined by CD127+KLRG1−), gated on CD8+CD44+CD45.1+. (E–J) Representative plots and mean fluorescence intensity (MFI) of CAR MPECs in the TDLN at D15, gated on CD8+CD44+CD45.1+, (E) CD127, (F) CD62L, (G) CD44, (H) CD69, (I) LAG-3, and (J) PD-1 and Tim-3 expression. Data are from 1 of 2–4 representative experiments with 4–6 biological replicates per experiment and are plotted as mean+SD. Ordinary one-way analysis of variance with Holm-Šídák test. ****p<0.0001, ***p<0.001, **p<0.01, *p<0.05.

Unarmored TA99 CAR T cells expressed elevated stemness factors, including higher TCF-1 expression prior to adoptive transfer (online supplemental figure 3C) and higher expression of CD127 and CD62L, which are associated with CAR T cell persistence and proliferative capacity,30 31 in the TDLN 1 week after transfer (figure 2E,F). In contrast, S233-armored TA99 CAR T cells expressed lower CD127 and CD62L levels, particularly when combined with CD4 depletion (figure 2E,F). Additionally, S233-armored TA99 CAR T cells appeared more activated and expressed higher CD44 and CD69 compared with unarmored CAR T cells in the TDLN or CAR T cells prior to transfer (figure 2G,H, online supplemental figure 3B,C).

Unarmored and S233-armored TA99 CAR T cells were also abundant in the spleen of all groups, with unarmored TA99 CAR T cells present in similar MPEC proportions compared with TDLN (online supplemental figure 4A–C). However, there was a significant decrease in splenic S233-armored TA99 CAR T MPECs. In combination with CD4 depletion, we observed an increase in CD127−KLRG1+ SLEC proportions, suggesting that systemically distributed S233-armored CAR T cells are hyperactivated and are more differentiated terminal effector cells (online supplemental figure 4A–C).

PD-1 is rapidly induced following T cell activation32 and was found to be equally expressed in unarmored and S233-armored TA99 CAR T cells prior to transfer (online supplemental figure 3D). One week post-transfer, unarmored TA99 CAR T cells in the TDLN and spleen expressed little to no PD-1, and while they were abundant in the tumor, remained PD-1low, consistent with reduced T cell activity (figure 2J,I, online supplemental figure 4D–F). In contrast, S233-armored TA99 CAR T cells in the TDLN and spleen expressed higher proportions of PD-1+Tim3− cells, indicating that S233-armoring supported an activated CAR T cell phenotype (figure 2J–I, online supplemental figure 4D). CD4 depletion further increased the PD-1+Tim3− subset in S233-armored TA99 CAR T cells (figure 2J–I, online supplemental figure 4D), suggesting that host or CAR CD4 T cells limit the activation of S233-armored CAR CD8 T cells.

Chronic T cell activation induces expression of additional inhibitory receptors including Tim-3, LAG-3 and TIGIT, and co-expression of PD-1 and Tim-3 is often used to identify dysfunctional or exhausted T cells in tumors.33 A small proportion of S233-armored CAR T cells expressed LAG-3 and co-expressed PD-1 and Tim-3 in the TDLN and spleen, but PD-1+Tim-3+ exhausted cells were readily detected in B16F10 tumors (figure 2J–I, online supplemental figure 4D,F). CD4 depletion had little effect on PD-1, Tim-3, and LAG-3 expression in unarmored TA99 CAR T cells. CD4 depletion cooperated with S233-armoring to increase PD-1+Tim-3+ CAR T cells in the TDLN and spleen but did not further increase exhausted cells within B16F10 tumors (figure 2J–I, online supplemental figure 4D,F).

To determine whether CD4 depletion and/or S233-armoring has similar effects on CAR memory precursor and activation phenotypes in another CAR T cell solid cancer model, we evaluated NKG2D CAR T cells in the immunocompetent MC38 colon cell carcinoma model. Using a similar treatment scheme as B16F10, we adoptively transferred 7×106 NKG2D CAR T cells into C57BL/6 mice 8 days after intradermally injection of 106 MC38 cells with or without CD4 depletion (figure 3A). Unarmored and S233-armored NKG2D CAR T cells were found in similar frequencies (figure 3B,C). Like TA99 CAR T cells in the B16F10 model, while unarmored and S233-armored NKG2D CAR T cells largely exhibited an MPEC phenotype across all groups, S233-armoring resulted in reduced CD127 expression levels (figure 3C,D). Unarmored NKG2D CAR T cells expressed higher CD127, lower activation markers, CD44 and CD69, and did not express PD-1 or other inhibitory receptors, Tim-3, LAG-3, and TIGIT (figure 3D–I). In contrast, S233-armored NKG2D CAR T cells expressed higher activation markers, had substantial PD-1+Tim3− and PD-1+Tim3+ populations, and expressed high levels of LAG-3 and TIGIT (figure 3E–I). CD4 depletion had little effect on tumor control in the MC38 immunogenic model or expression of activation markers in NKG2D CAR T cells except it further increased PD-1+Tim3− cells without increasing PD-1+Tim3+ cells (online supplemental figure 5A, figure 3G). Like TA99 CAR T cells, unarmored splenic NKG2D CAR T cells were overwhelmingly the MPEC phenotype while S233-armored NKG2D CAR T cells had an increase in SLEC phenotype cells (online supplemental figure 5B,C). Unarmored NKG2D CAR T cells in the tumor lacked PD-1 expression unlike S233-armored NKG2D CAR T cells, which were either PD-1+Tim3− or PD-1+Tim3+ cells and expressed high levels of LAG-3 and TIGIT (online supplemental figure 5D–G). CD4 depletion did not alter MPEC, activation, or exhaustion phenotypes of either unarmored or S233-armored CAR T cells (online supplemental figure 5B–G).

Figure 3

Characterization of chimeric antigen receptor (CAR) memory precursor effector cell phenotypes from the tumor-draining lymph node (TDLN) in another model system (MC38). (A) Experimental schematic. TDLN were collected on day 15 post-tumor inoculation. (B–C) Flow cytometric analysis of CD45.1+ CAR T cells. (B) Representative flow plots and proportions of CD45.1+ NKG2D CAR T cells, gated on CD8+CD44+. (C) Representative flow plots and proportions of CD45.1+ NKG2D CAR T cells. (D–F) Representative histograms and mean fluorescence intensity (MFI) of NKG2D CAR memory precursor effector cells in the TDLN at D15, gated on CD8+CD44+CD45.1+, (D) CD127, (E) CD44, (F) CD69, (G) PD-1 and Tim-3, (H) LAG-3 and TIGIT expression. Data are from 1 of 2 representative experiments with 4–6 biological replicates per experiment and are plotted as mean+SD. Ordinary one-way analysis of variance with Holm-Šídák test. ****p<0.0001, ***p<0.001, **p<0.01, *p<0.05.

Collectively, these data indicate that unarmored CAR T cells preferentially express memory-associated but not activation markers in the spleen, TDLN, and tumor, while S233-armoring promotes an activated memory precursor phenotype in the TDLN and increased differentiation into terminal effector cells in the spleen and tumor. CD4 depletion had little effect on CAR T cell proportions in lymphoid organs but led to a reduction in tumor-infiltrating CAR T cells in both B16F10 and MC38 models. Yet CD4 depletion augmented the S233-armored CAR T cell-dependent tumor growth delay of B16F10, a non-immunogenic tumor but had little effect on MC38 (figure 2, online supplemental figures 4–6). In the immunogenic tumor MC38, CD4 depletion had little effect on CAR effector T cell phenotypes or tumor growth kinetics (figure 3, online supplemental figure 5).

CD4 depletion increases CAR memory T cell differentiation

Next, we evaluated CAR memory T cells in the TDLN, spleen, and skin at least 30 days after tumor resection surgery (figure 5A). Unarmored TA99 CAR memory T cells were found in the TDLN and spleen at low numbers; however, they were significantly increased when mice were also treated with CD4 depletion (figure 5B,C, online supplemental figure 6A). In contrast, S233-armored TA99 CAR T cells were largely absent at a memory timepoint in the TDLN and spleen unless CD4 T cells were depleted (figure 5B,C, online supplemental figure 6A). Unarmored NKG2D CAR memory T cells also developed in the immunogenic MC38 model and enhanced by CD4 depletion (online supplemental figure 6B). While CD4 depletion did not improve the ability of S233-armored NKG2D CAR T cells to delay MC38 tumor growth, it was able to increase S233-armored NKG2D memory CAR T cells (online supplemental figures 5A, 6B). We phenotyped CAR memory T cells by flow cytometry and found that most CAR memory T cells in the TDLN preferentially differentiated into CD69−CD62L+ TCM cells irrespective of S233 armoring or CD4 depletion (figure 5D,E). A small proportion of CAR memory T cells from CD4-depleted mice were CD69+CD103+, characteristic of tissue-resident memory cells (figure 5D–G). The proportion of CAR TRM cells did not differ in the presence or absence of S233 armoring (figure 5F,G). Unarmored TA99 CAR memory T cells were found in the skin of treated mice, and CD4 depletion increased their frequency (figure 5H,I). All CAR memory T cells in the skin were CD69+CD103+ TRM cells (figure 5J,K). In contrast, few S233-armored TA99 CAR T cells populated the skin, irrespective of CD4 depletion (figure 5H–J). Together, these data show that unarmored CAR T cells have the capacity to form circulating and resident T cell memory and that their numbers are increased in the absence of CD4 T cells. S233-armoring impedes their ability to efficiently form memory, particularly in the skin.

Figure 5

Phenotypic characterization of chimeric antigen receptor (CAR) memory T cell subsets in the tumor-draining lymph nodes (TDLN) and skin. (A) Experimental schematic. B16F10 tumors were resected on day 18 post-tumor inoculation, and TDLN, spleen, and skin were harvested at least 30 days post-surgery (D48+ post-tumor inoculation). (B–G) Flow cytometric analysis of CD45.1+ CAR T cells in the TDLN on day 30+ post-surgery. (B) Representative flow plots and (C) counts of CD45.1+ CAR T cells, gated on CD8+CD44+. (D) Representative flow plots and (E) frequency of CD45.1+ CAR T cell memory subsets, gated on CD8+CD44+CD45.1+. (F) Representative flow plots and (G) frequency of CAR T cell expression of CD69 and CD103, gated on CD8+CD44+CD45.1+CD62L−CD69+. Data are from three combined experiments with 8–27 total mice per group for TDLN. (H–K) Flow cytometric analysis of CD45.1+ CAR T cells in the skin on day 30+ post-surgery. (H) Representative flow plots and (I) frequency of CD45.1+ CAR T cells, gated on CD8+CD44+. (J) CD69 and CD103 expression and (K) frequency of CD69+CD103+ CAR T cells in the skin, gated on CD8+CD44+CD45.1+. Data are from two combined experiments with 4–16 total mice per group for skin. Data are plotted as mean+SD and analyzed using ordinary one-way analysis of variance with Holm-Šídák test. ****p<0.0001, ***p<0.001, **p<0.01, *p<0.05.

S233-armored CAR T cells in combination with CD4 depletion enhanced antigen-specific endogenous TRM cells in the TDLN and skin

Mice treated with CD4 depletion therapy and S233-armored CAR T cells exhibited the most significant protection against B16F10 rechallenge (figure 4F,G) despite the limited generation of CAR memory T cells (figure 5). This led us to examine endogenous tumor-specific memory T cells. Both CD4 depletion and cytokine armoring can activate endogenous tumor-specific T cells.3 29 We previously reported that S233-armored CAR T cell treatment increased the number of tumor-specific endogenous effector T cells identified by TRP2-tetramer (TRP2Tet+) staining within the tumor.8 Endogenous TRP2Tet+ T cells expressing MPEC markers were observed on day 15 following either S233-armored CAR T cell treatment or CD4 depletion (online supplemental figure 9). We next quantified and characterized TRP2Tet+ memory T cells (figure 6A). TRP2Tet+ memory T cells were absent from the TDLNs without CD4 depletion (figure 6B,C). However, S233-armored CAR T cell treatment in combination with CD4 depletion significantly increased the overall frequency and absolute numbers of TRP2Tet+ T cells compared with CD4 depletion alone or in combination with unarmored TA99 CAR T cells (figure 6B,C). Greater than 60% of TRP2+ memory T cells in the TDLN expressed tissue residence markers, CD69 and CD103 (figure 6E,F).

Figure 6

Phenotypic characterization of endogenous tumor-specific memory T cell subsets in the tumor-draining lymph nodes (TDLN) and skin. (A) Experimental schematic. B16F10 tumors were resected on day 18 post-tumor inoculation, and TDLN and skin were harvested at least 30 days post-surgery (D48+ post-tumor inoculation). (B–F) Flow cytometric analysis of endogenous TRP2-tetramer+ CD8 T cells in the TDLN on day 30+ post-surgery. (B) Representative flow plots and (C) counts of TRP2-Tet+ CD8 T cells, gated on CD45.1−CD8+CD44+. (D) Representative flow plots of TRP2-Tet+ CD8 memory T cell subsets, gated on CD45.1−CD8+CD44+. (E) Representative flow plots and (F) frequency of TRP2-Tet+ T cell expression of CD69 and CD103, gated on CD8+CD44+CD45.1−CD62L−CD69+. Data are from three combined experiments with 8–27 total mice per group for TDLN. (G–I) Flow cytometric analysis of TRP2-tetramer+ T cells in the skin on day 30+ post-surgery. (G) Representative flow plots and (H) frequency of TRP2-Tet+ T cells, gated on CD8+CD44+CD45.1−. (I) Frequency of TRP2-Tet+ TRM cells in the skin, gated on CD8+CD44+CD45.1−CD103+CD69+. (J–K) Melanoma-associated vitiligo (MAV) phenotype and prevalence. (J) Representative images of CD4 depletion and S233-armored chimeric antigen receptor (CAR) T cell treated mice. (K) Quantification of MAV prevalence. Data are from two combined experiments with 4–16 total mice per group for skin. Data are plotted as mean+SD and analyzed using ordinary one-way analysis of variance with Holm-Šídák test. ****p<0.0001, ***p<0.001, **p<0.01, *p<0.05.

Like the TDLN, CD4 depletion alone or in combination with TA99 CAR T cells induced TRP2+ T cells in the skin. However, CD4 depletion combined with S233-armored CAR T cells further increased skin TRP2Tet+ T cells (figure 6G,H). Skin TRP2Tet+ T cells were predominantly CD69+CD103+ TRM cells (figure 6I). B16F10 tumor-bearing mice treated with neoadjuvant CD4 depletion therapy results in melanoma-associated vitiligo (MAV) in 60%–70% of treated mice.29 MAV-affected mice are protected against secondary tumors, consistent with the improved overall survival observed for MAV-affected melanoma patients, highlighting the significance of TRM cells in durability tumor immunity.21 34 We found that combining CD4 depletion with S233-armored CAR T cells resulted in 100% MAV prevalence and substantially increased TRM numbers in the skin and TDLN (figure 6J,K). Together these data indicate that combination therapies that include depletion of CD4 T cells and S233-armored CAR T cells can improve durable antitumor immunity via induction of tumor-specific tissue-resident memory.