Introduction

Chimeric antigen receptor (CAR) T cell therapy has demonstrated promising but inconsistent potency in the treatment of hematological malignancies,1–5 with sustained remissions remaining elusive for the majority of patients. Since phenotypic features of infused CAR T cells correlate with clinical potency,6–9 optimizing cell culture conditions may improve the function of infused cells.

Several previous studies have demonstrated advantages of generating CAR T cell products from selected T cells rather than unselected peripheral blood mononuclear cells (PBMC). One limitation of using unselected PBMC is that contaminating myeloid cells negatively impact cultured T cells,10 11 a problem that can be overcome by starting manufacturing with enriched T cells.11–13

In addition to improving manufacturing yields, T cell selection also appears to increase the potency of the final product. CD3+ enrichment generates CAR T cells with a higher fraction of naïve and central memory phenotype and improved cytotoxic function compared with products generated from unselected PBMC.14 In the clinical setting, products made from CD4/CD8-selected cells were associated with improved CAR T cell expansion, peak cytokine levels, and clinical responses compared with those generated from unselected PBMCs.12 Additionally, T-cell selection minimizes the risk that contaminating tumor cells are transduced with the CAR construct, potentially leading to antigen masking and subsequent relapse by functionally antigen-negative disease.15

Controlling the relative fractions of CD4+ and CD8+ cells in the infused product may have additional benefits. Patients with B-cell malignancies have highly variable ratios of CD4+ and CD8+ cells in the blood6 16 and in CAR T cell products.6 12 In animal models, infusing separately cultured CD4+ and CD8+ CAR T cells at a defined ratio significantly improves antitumor activity compared with unselected T cells.16 17 In the clinical setting (lisocabtagene maraleucel (liso-cel)), defined-ratio products yield clinical safety and efficacy outcomes that compare favorably with unselected CAR T cell products.1 2 4

The apparently superior efficacy of liso-cel compared with tisagenlecleucel (tisa-cel),18–20 another CD19-targeted 4-1BB-containing CAR T cell product that involves simple CD3+ enrichment during manufacturing, supports the concept of a more controlled CD4:CD8 ratio of infused CAR T cells. Although a CD4:CD8 ratio of 1:1 yielded the best results in animal models, 3:1 and 1:3 (but not 9:1 or 1:9) ratios led to improved survival compared with unfractionated cells,17 suggesting that there is an optimal effective range of CD4:CD8 ratios.

Collectively, these data supported our decision to manufacture clinical products through parallel CD4+ and CD8+ cell cultures, subsequently formulated for infusion at a 1:1 ratio, in a phase 1/2 clinical trial evaluating third-generation CD20-targeted CAR T cells (NCT03277729).21 22 Surprisingly, we observed suboptimal growth of CD8+ cell cultures from the initial patients in this trial. Based on the physiological role played by CD4+ cells during CD8+ cell priming and clonal expansion under some conditions,23–25 we hypothesized that CD8+ cell proliferation may be impaired in the absence of CD4+ support during ex vivo cell culture and explored the impact of combining CD4+ and CD8+ cells at various defined ratios at the initiation of cell manufacturing. We report here that CD8+ cells manufactured in separate cultures exhibit inferior ex vivo expansion, a hypofunctional phenotype, and aberrant gene expression signature, all of which can be avoided through coculture with CD4+ cells. The mechanisms underlying these findings involved both cytokine-mediated and contact-dependent mechanisms involving CD40L-CD40 and CD70-CD27 interactions.

ResultsCD8+ expansion and final cell product composition depend on initial CD4:CD8 ratio

To evaluate the impact that combining CD4+ and CD8+ T cells at different ratios at culture initiation has on expansion and CD4:CD8 composition of the final CAR T cell product, we isolated CD4+ and CD8+ cells from cryopreserved PBMCs collected from healthy donors and patients with relapsed B-cell lymphomas. The cells were activated, mixed at various CD4:CD8 ratios, transduced with 1.5.3-NQ-28-BB-z (third-generation anti-CD20 CAR) lentiviral vector, restimulated at day 7 with an irradiated CD20+ cell line to boost growth and enrich CAR+ cells, and harvested at days 14–15, concordant with the manufacturing process used at the initiation of our ongoing clinical trial. We measured expression of CD4, CD8, and truncated CD19 (tCD19, a marker of transduction encoded in the lentiviral vector26 (online supplemental figure 6D) by flow cytometry and counted cells at day 7 prior to restimulation and at the end of production on day 14.

We observed significantly higher fold expansion at both day 7 and day 14 of CD8+ CAR T cells mixed with CD4+ cells compared with CD8+ cells cultured in isolation, for both healthy donors and patients (figure 1A,B, online supplemental figure 1). The expansion of CD8+ cells was positively correlated to increasing fractions of CD4+ cells, and CD4+ cell expansion was largely unaffected by coculture. The fraction of CD4+ and CD8+ cells at the end of manufacturing was proportional to the starting ratios (figure 1C–F, online supplemental figure 2).

Figure 1

Effect of coculture of CD4+ and CD8+ CAR T cells on cell growth and composition of cell products. CD4+ and CD8+ T cells were isolated from apheresis products from healthy donors or patients, activated by anti-CD3/CD28 beads, cultured as either CD4+ cells only, CD8+ cells only, or a 60:40 CD4:CD8 ratio of mixed cells. Cells were transduced with 1.5.3-NQ-28-BB-z anti-CD20 CAR lentiviral vector and restimulated on day 7 with irradiated CD20+ LCL cells. At day 7 prior to restimulation and again at day 14, cells were analyzed by flow cytometry for CD4, CD8, and tCD19 transduction marker expression. The fold expansion of CD4+ and CD8+ T cell subsets expanded in separate (red) or in mixed (blue) cultures is shown for cells harvested (A) at day 7 prior to restimulation (n=23: 17 patients (open circles) and 6 healthy donors (filled circles)), or (B) at day 14 following restimulation (n=22: 16 patients (open circles) and 6 healthy donors (filled circles)). Bars represent the median±SEM. For C–F, cells were cultured as in A, B but at a variety of CD4:CD8 ratios from healthy donors (n=4 for 90:10, 80:20, and 70:30, n=6 for ratios 60:40, 50:50, and 40:60, n=3, for 30:70, n=2 for 20:80 and 10:90) or lymphoma patients (n=13 for day 7 ratios 90:10 and 80:20, n=17 for 70:30, 60:40, 50:50, and 40:60, n=5 for 30:70; day 14 n was equivalent to day 8 n−1). The ratio of CD4+ to CD8+ cells at day 7 (C, D) or day 14–15 (E, F) for total T cells (orange) or gated on tCD19+ T cells (green) is shown. The median ratios with IQR are shown along with individual values. NS, not significant.

The final CD4:CD8 ratios of the transduced (tCD19+) T cells were similar to those of the total cell population (figure 1C–F), excluding the possibility that there were differences in transduction efficiency between CD4+ and CD8+ cells. The starting ratio that yielded a final median CD4:CD8 ratio of tCD19+ cells closest to 1 was 50:50 for healthy donors both at day 7 (0.96) and day 14 (0.86), and 70:30 for patients at both day 7 (1.1) and day 14 (0.97) (figure 1C–F).

The superior CD8+ cell expansion in the presence of CD4+ cells allowed us to modify the manufacturing process to eliminate the restimulation step and shorten the cell culture time to 8–9 days. Using this process, we similarly found that CD8+ CAR T cells expanded in mixed cultures exhibited approximately a threefold higher median expansion than cells cultured in the absence of CD4+ cells (online supplemental figure 3).

To evaluate whether these results were a function of the CD20 third generation CAR construct, we repeated the experiments using a second-generation CD19-targeted CAR (hCD19-BB-z). We similarly observed significantly higher fold expansion in mixed cultures of CD8+ T cells transduced with the CD19 CAR or even an empty vector, comparable to the effect with the CD28-41BB CD20-targeted CAR (online supplemental figure 4). We conducted additional experiments using second-generation CAR T cells with a CD28 costimulatory domain. As with the previous experiments, we found that CD8+ T cells modified with the 1.5.3-NQ-28-z CD20 CAR exhibited significantly higher fold expansion when cocultured with CD4+ cells than when cultured separately (online supplemental figure 5).

Impact of mixed cultures on CAR T cell immunophenotype

Given these profound effects of CD4+ cells on CD8+ CAR T cell expansion, we inquired whether CD8+ cells might also exhibit phenotypic and functional differences. As noted above, we employed two manufacturing processes: (1) a process of 14–15 days involving a restimulation step with CD20+ target cells at day 7 and (2) a process of 8–9 days without a restimulation step. While we have since adopted the second process for use both in the laboratory and in our clinical trial, the first process provides an opportunity to assess CAR T cell phenotype following antigen encounter, as occurs after cell infusion. We, therefore, present immunophenotypic and functional results for both processes throughout this report.

At both day 8 (without restimulation) or day 14 (following CD20+ restimulation), CD8+ CAR T cells cultured in the absence of CD4+ cells exhibited lower postexpansion levels of markers associated with memory and a less-differentiated state, and higher levels of exhaustion and terminal differentiation markers, compared with cocultured cells (figure 2A,B,D,E, online supplemental figure 6). CD19 CAR T cell products with a higher fraction of CD27+/PD-1– cells within the CD8+ T cell subset have been shown to yield clinical remissions more frequently in patients with CLL.6 We found that percentages of CD27+/PD-1– CD8+ cells were significantly increased in mixed CD4:CD8 cultures compared with CD8+ CAR T cells expanded separately (figure 2C,F). CD4+ CAR T cell immunophenotypes were largely similar regardless of the culture method (online supplemental figure 7). We observed a similar pattern of decreased memory markers and higher exhaustion markers among hCD19-BB-z CAR CD8+ T cells (online supplemental figure 8) and 1.5.3-NQ-28-z CAR T cells (online supplemental figure 9) from isolated cultures, indicating that these differences were not specific to CD20 CAR constructs, and were seen in CAR T cells across a variety of costimulatory domains (CD28-only, 4-1BB-only, or CD28-4-1BB).

Figure 2

Phenotypic differences between CD8+ CAR T cells cultured alone or those cultured with CD4+ T cells. CD4+ or CD8+ enriched PBMC from healthy donors (filled circles) or patients (open circles) were stimulated, then either mixed at a 60:40 CD4:CD8 ratio or maintained in separate cultures, then transduced with 1.5.3-NQ-28-BB-z CD20 CAR lentiviral vector. (A–C) Cells were harvested on day 8 of cell culture without restimulation, or (D–F) restimulated with irradiated CD20+ LCL cells on day 7 and harvested on day 14. Markers of memory and differentiation (A, D), exhaustion (B, E), or percentage of CD27+ PD1─ cells (C, F) were measured by flow cytometry. Gating strategy and representative histograms are shown in online supplemental figure 6. Data represent the fold increase in geometric mean fluorescence intensity (MFI) over isotype control, gated on viable CD8+ tCD19+ CAR T cells at day 7 (n=14: 5 patients and 9 healthy donors for A, B and n=9: 4 patients and 5 healthy donors for (C) or at day 14 (n=13: 7 patients and 6 healthy donors). P values were determined using paired two-tailed t-tests for markers meeting criteria for normality based on D’Agostino and Pearson or Shapiro-Wilk tests, or Wilcoxon matched-pairs signed rank test for markers not meeting normality criteria. CAR, chimeric antigen receptor; NS, not significant; PBMC, peripheral blood mononuclear cell.

Hypofunctional phenotype of CD8+ CAR T cells expanded in absence of CD4+ cells

At both day 8 and day 14, FACS-sorted CD8+ CAR T cells (1.5.3-NQ-28-BB-z) cultured without CD4+ cells secreted significantly lower levels of IFN-γ, IL-2, and TNF-α on stimulation with CD20+ Raji lymphoma cells than those expanded in mixed cultures (figure 3A,C). Likewise, proliferative capacity of CD8+ CAR T cells was lower when cultured without CD4+ cells (figure 3B,D). In contrast, cytokine secretion and proliferative capacity of CD4+ cells were minimally impacted regardless of culture method (online supplemental figure 10). The hCD19-BB-z and 1.5.3-NQ-28-z transduced cells similarly exhibited lower levels of cytokine secretion (though TNF-α for hCD19-BB-z and IFN-γ for 1.5.3-NQ-28-z did not reach statistical significance) and less proliferation in CD8+ CAR T cells cultured in the absence of CD4+ cells (online supplemental figures 11,12E,G). We also evaluated granzyme B secretion and cytotoxicity in second-generation CD20 CD8+ CAR T cells but found no differences between CD8+ CAR T cells expanded separately versus in CD4+ coculture (online supplemental figure 12F,H), indicating that these cells are not fully dysfunctional but retain cytotoxic function despite impaired cytokine secretion and proliferative capacity.

Figure 3

CD8+ CAR T cells cultured in the absence of CD4+ cells have impaired cytokine secretion and proliferation in vitro. CD4+ and CD8+ enriched PBMC from patients (open circles) or healthy donors (filled circles) were stimulated, then either mixed at a 60:40 CD4:CD8 ratio or maintained in separate cultures, transduced with 1.5.3-NQ-28-BB-z CD20 CAR, and expanded. Cells were harvested on day 8 without restimulation (A, B) or restimulated with irradiated CD20+ LCL cells on day 7 and harvested on day 14 (C, D). FACS-sorted CD8+ tCD19+ T cells labeled with Cell Trace Violet (CTV) were incubated with irradiated CD20+ Raji-ffLuc cells (1:1 ratio). (A, C) supernatants were harvested at 24 hours and the indicated cytokines were measured by Luminex assay (n=9: 4 patients and 5 healthy donors for day 7; n=12: 4 patients and 8 healthy donors for day 14). (B, D) proliferation of the sorted cells after 4 days based on CTV dilution was assessed by flow cytometry. representative histograms are shown in the left panel, and summary data of geometric MFI ratio of unstimulated to stimulated cells and % divided cells are shown in the right panels (n=13: 4 patients and 9 healthy donors for day 7; n=6: 2 patients and 4 healthy donors for day 14). P values were determined using paired two-tailed t-tests for samples meeting criteria for normality based on D’Agostino and Pearson or Shapiro-Wilk normality test, or Wilcoxon matched-pairs signed RANK test for samples not meeting normality criteria. CAR, chimeric antigen receptor; MFI, mean fluorescence intensity; PBMC, peripheral blood mononuclear cell.

To evaluate the impact of separate versus mixed CD4+/CD8+ CAR T cell cultures on in vivo anti-lymphoma activity, we employed a CD20+ Raji mouse xenograft model, using suboptimal CAR T cell doses to distinguish differences in activity, since larger cell doses are curative using separate cell cultures infused at a 1:1 ratio.26 We found that mice treated with third-generation CD20 CAR T cell products manufactured in CD4+/CD8+ cocultures exhibited superior tumor control compared with mice treated with equivalent doses of CD4+ and CD8+ CAR T cells expanded separately and infused at a 1:1 ratio (figure 4, online supplemental figure 13).

Figure 4

Impact of mixed versus separate CD4+ and CD8+ CAR T cell cultures on in vivo antitumor activity. NSG mice (n=8 per experimental group, n=5 for untransduced and no treatment groups, n=1 for no tumor group) bearing 7 day Raji-ffLuc tumors, received a suboptimal dose of 2×106 tCD19+ 1.5.3-NQ-28-BB-z CD20-targeted CAR T cells cultured at a 60:40 CD4:CD8 ratio or expanded in separate parallel CD4+ and CD8+ cultures and formulated at a 1:1 ratio prior to injection. The mixed cells were at a 1:1.9 CD4:CD8 ratio at the time of infusion. Untransduced T cells expanded in mixed cultures were included as a control. Mice were imaged twice weekly by bioluminescence imaging. (A) Experimental schema. (B) Average tumor burden per group over time as measured by total body bioluminescence. The mean luminescence values±SEM are shown. The tumor burden over time was greater in the separate group compared with the mixed group based on a two-way repeated measures ANOVA, time x treatment group, F (3, 42)=34.64, p<0.0001; time factor, F (1.158, 16.21)=40.05, p<0.0001; treatment group factor, F (1, 14)=37.24, p<0.0001. (The overall model including no treatment, untransduced, separate, and mixed groups showed: time x treatment group, F (9, 84)=21.49, p<0.0001; time factor, F (1.054, 29.51)=76.63, p<0.0001; treatment group factor, F (3, 28)=35.70, p<0.0001; for one mouse who died between day 15 and day 19 in the no-treatment group, the day 15 body flux value was carried forward as the day 19 value). (C) Dorsal images of each mouse at day 19 are shown. ANOVA, analysis of variance; CAR, chimeric antigen receptor.

Altered gene expression profiles of CD8+ cells expanded in absence of CD4+ help

In view of the marked differences in ex vivo expansion, immunophenotype, and function between CD8+ cells from mixed versus separate cultures, we hypothesized that the hypofunctional phenotype of CD8+ CAR T cells generated in the absence of CD4+ help is driven by an altered transcriptional program. We, therefore, analyzed transcriptional profiles of FACS-sorted CD8+ CAR T cells expanded in mixed cultures with CD4+ cells compared with those in separate CD8-only cell cultures. At both day 8 of cell culture, and at day 14 (7 days after restimulation), there were large numbers of differentially expressed genes (false discovery rate<0.05 and fold change ≥1.5), indicating profoundly different gene expression patterns. At both timepoints, CD8+ CAR T cells cultured separately expressed significantly higher levels of genes associated with exhaustion and dysfunction, and lower levels of genes associated with memory formation compared with CD8+ cells expanded in cocultures. Differentially expressed genes were strongly correlated between day 8 and day 14 (figure 5A,B, online supplemental figure 14A–C and dataset S1). A broad array of costimulatory molecules and inhibitory receptors were upregulated in the separately cultured CD8+ cells, a characteristic of dysfunctional T cells.27 Full RNA-seq gene expression data from day 8 and day 14 are available at the NCBI Gene Expression Omnibus (accession number GSE245427, https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE245427).28

Figure 5

CD8+ CAR T cells cultured in presence or absence of CD4+ T cells exhibit distinct transcriptional signatures. Gene expression profiles of flow-sorted CD8+ tCD19+ 1.5.3-NQ-28-BB-z CAR T cells cultured either with CD4+ cells (“mixed”) or alone (“separate”) were evaluated by bulk RNA-seq on day 8 (n=6: 2 patients and 4 healthy donors). (A) Volcano plot of false discovery rate (FDR) (–log10) versus fold change (log2) showing differentially expressed genes between mixed versus separate CD8+ CAR T cells (FDR<0.05 and |log2FC|≥0.585 [≥1.5 fold change]), with upregulated and downregulated genes shown in orange and blue, respectively. (B) Heat map of fold change (log2) of selected genes (all with FDR<0.05) between mixed versus separate CD8+ CAR T cells in the 6 individual subjects. (C) Normalized enrichment scores from GSEA using selected gene sets related to CD8 naïve, memory, effector, and exhausted cells from the MSigDB C7 database, using the rankings of differential expression p values for all the genes. Positive/negative scores indicate enrichment of the gene sets in upregulated or downregulated genes when comparing mixed versus separate CD8+ CAR T cells. The FDR was ≤0.06 for all the gene sets shown here. GEO datasets are indicated in parentheses. CAR, chimeric antigen receptor; GSEA, gene set enrichment analysis;

We assessed which pathways were over-represented by the differentially expressed genes. At day 8, CD8+ CAR T cells expanded in the presence of CD4+ cells exhibited stronger TCR signaling, CD28 costimulation, and IFN-γ signaling, and lower levels of G-protein coupled receptor signaling and IL-10 signaling (online supplemental figure 15). At day 14 following restimulation, CD8+ cells from mixed cultures were distinguished primarily by upregulation of pathways associated with proliferation and cell cycle (online supplemental figure 14D). We also performed gene set enrichment analysis (GSEA) using the rankings of differential expression p values for all the genes. At day 8, consistent with our immunophenotypic findings, CD8+ CAR T cells from mixed cultures overexpressed genes that are upregulated in less differentiated CD8+ T cells (figure 5C). At day 14, the GSEA results were more heterogeneous, with mixed CD8+ T cells sharing upregulated genes with less differentiated cells from some gene sets and more differentiated or exhausted cells from other gene sets (online supplemental figure 14E), possibly reflecting the overlap of genes differentially expressed in both activated and exhausted states.

Mechanisms underlying improved CD8+ cell growth in CD4+ cocultures

Because cytokines impact CD8+ cell proliferation29 and promote expansion of naïve and/or memory T cells,30 we hypothesized that some of the observed differences between separate versus mixed CD8+ cells might be mediated by cytokines secreted by CD4+ cells. We measured the concentrations of several cytokines in CD4-only, CD8-only, or mixed cultures at early timepoints during cell culture (days 1 and 4) and found that mixed cultures had higher levels of IFN-γ, IL-2, TNF-α, and IL-15 than CD8-only cultures at day 1, and higher levels of IFN-γ, TNF-α, and IL-21 at day 4 (online supplemental figure 16).

We next performed experiments to evaluate the relative effects of soluble versus cell contact-dependent factors of CD4+ cells on CD8+ cells. Both-generation second and third-generation CD20-targeted CD8+ CAR T cells separated from CD4+ cells by a transwell insert exhibited inferior expansion to CD8+ cells in fully mixed cultures, but superior to CD8+ cells cultured in complete isolation (online supplemental figures 17A and 18A), suggesting that both cytokine-mediated and cell contact-dependent mechanisms contribute to ex vivo expansion. However, on antigen stimulation of CD8+ CAR T cells, improved proliferative capacity was only observed in fully mixed cultures, indicating that cell-cell contact was required (online supplemental figures 17B and 18B). The impact of soluble factors on cytokine secretion was less clear, with third generation CD8+ CAR T cells in transwell cultures demonstrating increased TNF-α and granzyme B, but not IL-2 or IFN-γ (online supplemental figure 17C,D) compared with cells in separate cultures, but no differences in cytokine secretion with second-generation CD8+ CAR T cells (online supplemental figure 18C,D).

With respect to immunophenotypic signatures, we evaluated the memory and exhaustion markers previously found to be significant between mixed versus separate CD8+ CAR T cells (figure 2A,B, online supplemental figure 9C,D) and found no significant difference in these markers between cells expanded in separate cultures or in transwell cultures for either second or third generation CD8+ CAR T cells (online supplemental figures 17E,F and 18E,F). Instead, differences were only observed between transwell and fully mixed cultures, indicating that soluble factors from CD4+ cells were not solely responsible for the observed phenotypic differences but rather that direct cell-cell contact was required. Cumulatively, these results suggested that although soluble factors from CD4+ cells contribute to superior CD8+ cell expansion ex vivo and possibly cytokine secretion, cell contact-dependent mechanisms are required for the improved function and less differentiated phenotype of CD8+ CAR T cells cultured with CD4+ cells.

In light of previous reports that activated CD8+ cells express CD40, and that CD40 ligation may provide a direct costimulatory signal independent of the well-known effects of CD40L on antigen-presenting cells (APCs),31–33 we hypothesized that the contact-dependent effect of CD4+ cells on CD8+ cells might be mediated through CD40L-CD40 interactions. We also hypothesized that given the improved outcomes reported in patients with higher levels of CD27+ cells,6 that CD70-CD27 interactions may also contribute. We first measured expression of these markers over time and found that in addition to the expected strong upregulation of CD40L on CD4+ cells on days 1–4, there was also a small but significant concurrent upregulation of CD40 on CD8+ cells (online supplemental figure 19). CD8+ cells also expressed robust levels of CD27, and both CD4+ and CD8+ cells expressed CD70, which was upregulated after activation, but present even on resting cells prior to stimulation.

We then tested the impact of CD40 and CD27 signaling on ex vivo expansion, phenotype, and function, using antagonistic anti-CD40L or anti-CD70 antibodies in mixed CD4+/CD8+ cultures to block receptor-ligand interactions, or by using agonistic anti-CD40 or anti-CD27 antibodies in isolated CD8+ cultures. Anti-CD40L or anti-CD70 antibodies significantly impaired CD8+ CAR T cell expansion in mixed cultures, and agonistic CD40 or CD27 antibodies increased CD8+ cell expansion in separate cultures (figure 6A–D), suggesting that CD40 and CD27 signals contribute to improved CD8+ ex vivo expansion.

Figure 6

CD4+ cells augment CD8+ T cell growth and function through both CD40L-CD40 and CD27-CD70 interactions. CD8+ cells from healthy donors (filled circles) or patients (open circles) were activated with αCD3/CD28 beads, cultured separately (A, C, E, G, I, J) or at a 60:40 ratio with CD4+ cells (B, D, F, H), transduced with 1.5.3-NQ-28-BB-z CD20 CAR lentivirus, and expanded in the presence of plate-bound agonistic anti-CD40 (A, E, I), plate-bound agonistic anti-CD27 antibody (C, G, J), soluble antagonistic anti-CD40L (B, F), or soluble antagonistic anti-CD70 antibody (D, H). (A–D) Fold expansion of CD8+ cells at day 8 is shown. (E–H) At day 8, cells were harvested, labeled with Cell Trace Violet (CTV), and restimulated with irradiated CD20+ Raji-ffLuc cells (1:1 ratio). Proliferation of the CD8+ cells after 4 days based on CTV dilution was measured by flow cytometry, with geometric MFI ratio of stimulated to unstimulated cells shown. (I–J) Supernatants from E, G were harvested 24 hours after restimulation, and levels of the indicated cytokines were measured by Luminex. Data represent mean values (±SEM). For (A), n=15 (6 patients and 9 healthy donors); for (B), n=14 (5 patients and 9 healthy donors); for (C), n=7 healthy donors; for (D), n=9 healthy donors; for (E, F, I), n=9 (2 patients and 7 healthy donors); for (G, H, J), n=7 healthy donors. P values were determined using paired two-tailed t-tests for samples meeting criteria for normality based on D’Agostino and Pearson or Shapiro-Wilk normality test, or Wilcoxon matched-pairs signed rank test for samples not meeting normality criteria. MFI, mean fluorescence intensity; NS, not significant.

We found that neither blocking CD40L/CD40 or CD70/CD27 interactions between CD4+ and CD8+ CAR T cells with antagonistic antibodies, nor treatment of CD8+ cells with agonistic anti-CD40 or anti-CD27, recapitulated the phenotypic changes observed between separate versus mixed-culture CD8+ cells (online supplemental figure 20). However, antagonistic anti-CD40L or anti-CD70 antibodies reduced the proliferative capacity of CD8+ CAR T cells in mixed cultures (figure 6F,H). Agonistic anti-CD40 or anti-CD27 antibodies did not impact proliferative capacity, IL-2 secretion, or granzyme B secretion of CD8+ CAR T cells but did increase IFN-γ and TNF-α secretion (figure 6E,G,I,J). Taken together, the data suggest that CD40 and CD27 signals on CD8+ CAR T cells received from CD40L and CD70 on CD4+ cells contribute to the improved ex vivo expansion and effector function observed in cocultured CD8+ CAR T cells but do not explain the less differentiated phenotype, which appears to require both soluble factors as well as direct cell contact with CD4+ cells.

Discussion

The various methods of T cell selection, activation, provision of supplemental cytokines, and culture duration currently used in CAR T cell manufacturing likely impact T cell function,6–9 16 17 34–36 but the optimal conditions are not known. T-cell enrichment avoids the deleterious effects of contaminating myeloid cell subsets and minimizes the risk of transducing circulating tumor cells that can occur with unselected PBMC.11 12 15 Infusion of CAR T cells at a defined CD4:CD8 ratio leads to superior outcomes in preclinical studies compared with products generated from unselected PBMC,16 17 and while no direct comparisons have been made in human clinical trials, CAR products with defined 1:1 ratios compare favorably with those generated from unselected PBMC.1 2 4 However, the need to generate, maintain, and qualify two separate cultures for each patient adds cost and complexity to the manufacturing process and increases the risk of a nonconforming product.

In this study, we evaluated whether CAR T cell manufacturing could be improved by combining CD4+ and CD8+ cell at a defined ratio at the initiation of cell cultures. It is well established that, in vivo, CD4+ cells play an important role in CD8+ T cell priming, memory formation and maintenance, effector differentiation, and sustaining functionality during chronic antigen exposure,32 37–40 and that these effects are primarily mediated through APCs. However, the direct impact of CD4+ cells on CD8+ cells during ex vivo CAR T cell culture, which does not depend on APCs, has not been well characterized. Our results reveal that CD4+ coculture with CD8+ cells markedly enhances not only CD8+ cell ex vivo expansion, but also phenotype and function. CD8+ cells cultured in the absence of CD4+ cells have a more exhausted phenotype, correlating with inferior in vitro proliferation, cytokine secretion, and in vivo antitumor activity. Importantly, these findings were reproducible across a variety of types of CAR constructs.

The observed phenotypic and functional differences were driven by distinct transcriptional programs. CD8+ cells cultured alone expressed lower levels of genes associated with memory formation and higher levels of genes associate