Epigenetic engineering for optimal chimeric antigen receptor T cell therapy

2.1 Memory T cell differentiation

Memory T cell formation principles have primarily been elucidated using mouse viral infection models.¹² Naïve CD8⁺ T cells differentiate into effector T cells upon initial antigen stimulation. After pathogen clearance, most of the effector T cells undergo apoptosis, while a part of them acquire memory T cell properties and survive for a long time. As T cells are usually stimulated in vitro using anti-CD3 mAb in CAR-T cell generation, CAR-T cells acquire a memory T cell phenotype at the time of infusion. Memory T cells progressively differentiate during multiple rounds of cell division. The differentiation process is accompanied by gradual phenotypic changes, and several representative differentiation stages are referred to as stem cell-like memory T cells (T_SCM, CD45RA⁺CCR7⁺CD62L⁺CD27⁺CD28⁺IL-7R⁺CD95⁺), central memory T cells (T_CM, CD45RA⁻CCR7⁺CD62L⁺), and effector memory T cells (T_EM, CD45RA⁻CCR7⁻CD62L^+/−).^{13, 14} Human T_SCM, located at the apex of the differentiation hierarchy, can usually maintain themselves through self-renewal cell division for a decade-long period.¹⁵ However, CAR-T cells are terminally differentiated in a much shorter period because of massive expansion during in vitro culture and after antigen encounter in vivo in most cases (Figure 1). As CAR-T cell persistence and therapeutic efficacy are closely correlated, enhancing the self-renewal capacity of T cells is a rational intervention to elicit a durable antitumor response.^{16, 17}

Multiple studies have identified key transcription factors that control effector differentiation memory formation, and progressive differentiation of memory T cells.¹⁸ Epigenetic profiles control the activity of these transcription factors by regulating their accessibility to cognate genomic regions. In fact, the epigenetic landscape is globally altered upon T cell phenotypic changes.^19-22 As most epigenetic modifiers have nonredundant roles in T cell functions, modulation of single epigenetic enzymes could globally alter T cell differentiation status.

2.2 T cell exhaustion

Chronic antigen stimulation of cytotoxic T cells impairs their proliferation and cytokine production capacity, which is frequently accompanied by upregulation of multiple immune checkpoint molecules, including programmed cell death 1 (PD1), TIM3, and LAG3. This state is referred to as T cell exhaustion. Because a detailed definition of T cell exhaustion has been reviewed elsewhere,²³ we will discuss the epigenetic profiles of exhausted T cells. Exhausted T cells possess a chromatin landscape distinct from that of conventional memory and effector T cells.²⁴ Importantly, these epigenetic profiles remain unchanged even after chronic antigen exposure resolution or immune checkpoint blockade therapy (ICT).^25-27 These findings indicate that T cell exhaustion is fundamentally regulated by epigenetic mechanisms rather than by the expression of immunoinhibitory receptors. Recent studies have revealed the significance of several transcription/epigenetic factors, including TOX,²⁸ NR4A,²⁹ and DNMT3A,²⁷ for the formation of exhaustion-related epigenetic profiles, which could be attractive targets to prevent or abrogate T cell exhaustion programs. Another important finding is that exhausted T cells contain a subset of T cells with an early memory phenotype, called precursor exhausted T cells (T_PEX).^30-32 Precursor exhausted T cells highly express the memory-related transcription factors TCF7, ID3, and BCL6, but not effector-associated factors such as B lymphocyte-induced maturation protein-1 (Blimp1) or ID2, compared to terminally exhausted T cells. They also possess surface marker phenotypes distinct from those of terminally exhausted cells: high expression of SLAMF6/Ly108 and CXCR5, and low expression of TIM3.³⁰ As the T_PEX population alone can regain effector functions upon ICT, at least transiently, it would be a reasonable strategy to expand or maintain this subset of exhausted T cells to augment the therapeutic efficacy of ICT (Figure 2).

3.1 DNA methylation

DNA-methyltransferase 3A plays a pivotal role in effector differentiation of naïve T cells. T cells upregulate the expression of DNMT3A after antigen stimulation, which then downregulates naïve-associated genes by de novo DNA methylation of their promoter regions.¹⁹ A part of the effector T cells become memory T cells through successful demethylation of some naïve-associated genes such as Sell, Ccr7, and Tcf7. Genetic KO of Dnmt3a in mouse CD8⁺ T cells promotes memory formation by enhancing the demethylation efficiency of these loci.^{33, 34} In addition, the role of DNMT3A in exhausted T cells has been investigated. Chronic antigen stimulation causes progressive DNA methylation of genes expressed in early memory T cells, which contributes to the formation of gene expression profiles in terminally exhausted T cells. Although Dnmt3a-KO T cells did not prevent the upregulation of immune checkpoint molecules, they maintained early memory phenotypes and regained effector functions upon PD1 blockade, suggesting that inhibition of Dnmt3a supports the maintenance of a T_PEX population under chronic antigen exposure.²⁷ Based on these findings, the effect of DNMT3A inhibition on CAR-T cells has been explored. CRISPR/Cas9-mediated KO of DNMT3A in human CAR-T cells enhanced their proliferative capacity and maintained a T_CM phenotype after repeated stimulation, which resulted in superior control of tumor progression in vivo.³⁵

Conversely, DNA demethylation through TET2 is associated with upregulation of effector-associated genes (e.g., Ifng, Tnf, and Gzmb) following antigen exposure.³⁶ Tet2 KO in murine CD8⁺ T cells promoted memory formation and enhanced pathogen control upon secondary antigen encounter.³⁷ Fraietta et al.¹⁶ reported that a CAR-T cell clone in which the CAR transgene was inserted into TET2 gene locus and ablated with its expression showed substantial expansion in a patient with chronic lymphocytic leukemia and induced complete response. They further confirmed that TET2 inhibition conferred a proliferative advantage on repeatedly stimulated CAR-T cells. TET2 also appears to play an essential role in the immunosuppressive functions of myeloid cells within the tumor microenvironment (TME). Tet2 KO, specifically in myeloid-lineage cells, endowed tumor-associated macrophages with pro-inflammatory gene expression profiles and promoted tumor infiltration of effector T cells.³⁸ These studies suggest that pharmacological inhibition of TET2, although not developed at present, could enhance antitumor immunity by reversing immunosuppressive TME. Although DNMT3A and TET2 play reciprocal roles in DNA methylation, inhibiting both enzymes contributes to T cell longevity through different mechanisms. Further studies are required to elucidate how these factors coordinately regulate T cell properties.

3.2 Repressive histone marks

As discussed above, terminal T cell differentiation is accompanied by decreased expression of memory-associated genes. As this is accomplished, at least in part, by the deposition of repressive histone marks, inhibition of the repressive epigenetic factors targeting these genes could promote the maintenance of long-lived memory T cells. We have recently identified that genetic KO of PR domain zinc finger protein 1 (PRDM1), encoding Blimp1, helped to maintain the functional properties of early memory T cells in human CAR-T cells.³⁹ Although Blimp1 itself seems to lack histone-modifying enzymatic activity,⁴⁰ it organizes the formation of repressive histone marks by interacting with several other epigenetic enzymes such as G9a (H3K9 methyltransferase), HDAC1/2 (histone deacetylase), and PRMT5 (protein arginine methyltransferase).⁴¹ In addition to its role as a master regulator for B cell terminal maturation, Blimp1 is closely associated with effector T cell functions.⁴² In CD8⁺ T cells, T cell receptor (TCR) stimulation and subsequent cytokine signaling induce the expression of Blimp1.^{43, 44} Blimp1 promotes terminal T cell differentiation by negatively regulating memory-associated genes.

Blimp1 disruption in CAR-T cells significantly remodeled chromatin openness in approximately 7000 genomic regions, resulting in altered expression levels of more than 2000 genes. Blimp1-deficient CAR-T cells showed a less-differentiated memory phenotype, which resulted in better persistence after infusion. As expected, Blimp1 KO attenuated the effector functions of CAR-T cells, as indicated by reduced granzyme B and perforin production. Despite this disadvantage, Blimp1-KO CAR-T cells showed potent antitumor responses in vivo by acquiring a long-lived potential.

SUV39H1 is a histone methyltransferase that catalyzes H3K9me3 independent of the Blimp1 complex.⁴⁵ During effector differentiation of CD8⁺ T cells, SUV39H1 contributes to the downregulation of naïve-associated genes by adding H3K9 trimethylation in their promoter regions.⁴⁶ SUV39H1-KO CD8⁺ T cells showed enhanced memory formation and attenuated effector functions. In contrast, SUV39H1 also plays a role in repressing effector gene expression in tumor-infiltrating T-lymphocytes (TILs).⁴⁷ In a mouse tumor model, the expression of SUV39H1 was upregulated in TILs. Treatment with the SUV39H1 inhibitor F5446 enhanced the effector functions of TILs through upregulation of effector genes. These studies suggest that SUV39H1 promotes the initial step of the naïve-to-effector transition; however, it suppresses effector T cell functions at the terminally differentiated stage.

Lysine-specific histone demethylase 1A (KDM1A), also known as lysine-specific histone demethylase 1 (LSD1), is a demethylase specific for mono- and dimethylated H3K4 and H3K9, thereby acting as both a transcriptional activator and repressor. Lysine-specific histone demethylase 1A promotes the self-renewal and proliferative capacity of cancer stem cells and is therefore considered a direct therapeutic target.⁴⁸ Another interesting finding on KDM1A inhibition in tumor cells is that it induces double-stranded RNA stress and subsequent IFN pathway activation, which results in enhanced tumor immunogenicity through upregulation of HLA class I expression in tumor cells.^{49, 50} A recent study has also elucidated the role of KDM1A in antitumor T cells. Liu et al. showed that KDM1A inhibition expanded exhausted T cells with a T_PEX phenotype within the TILs, suggesting that KDM1A promoted terminal T cell differentiation.⁵¹ Inhibition of KDM1A prior to anti-PD1 immunotherapy augmented the antitumor response through the expansion of a less differentiated exhausted T cell population that can respond to ICT.

3.3 Active histone marks

Histone lysine acetylation (H3K9ac or H3K27ac) and the trimethylated form of H3K4 (H3K4me3) are representative histone marks that promote gene expression. These histone modifications undergo genome-wide alteration upon memory formation and differentiation.^{20, 52} The BET bromodomain protein family, including BRD2, BRD3, BRD4, and BRDT, contains bromodomain signatures that bind to acetylated lysine residues in histones and regulate gene expression as an epigenetic reader.⁵³ We have previously reported that CAR-T cells treated with the BET bromodomain inhibitor JQ1 maintained an early memory phenotype during in vitro expansion and showed enhanced persistence in vivo.⁵⁴ As one of the mechanisms, BRD4 positively regulates the expression of the effector gene BATF by binding to its promoter region. JQ1- or RNA interference-mediated downregulation of BATF results in increased expression of memory markers and interleukin-2 (IL-2) production. Although BATF is an essential transcription factor that induces potent effector functions in CD8⁺ T cells,⁵⁵ its activity is not required during in vitro preparation of antitumor T cells. In more recent studies, Milner et al. investigated BRD4 target genes more comprehensively and showed that BRD4 transcriptionally regulates multiple genes necessary for effector T cell differentiation, including BATF, ID2, and ZEB2, further corroborating our findings.⁵⁶ Another group has reported that JQ1 reinvigorates CAR-T cells from the exhausted state.⁵⁷ JQ1 treatment promoted antigen-dependent proliferative capacity of CAR-T cells through maintaining surface expression levels of the CAR, which is frequently downregulated in repeatedly stimulated CAR-T cells. Intriguingly, JQ1 treatment also enhanced antitumor immunity by suppressing PD-L1 expression in tumor cells and tumor-associated macrophages.⁵⁸ These findings suggest that BET bromodomain proteins play pleiotropic roles in T cell functions.

Acetylated histones are erased by HDACs, which are composed of 18 genes and grouped into classes I–IV. Multiple studies have elucidated the inhibitory effects of class I HDACs (HDAC1, 2, 3, and 8) on memory and effector T cell function. Tay et al. showed that HDAC3 inhibited the cytotoxic T cell response at an early time point during T cell activation.⁵⁹ The HDAC3-specific inhibitor RGFP966 enhanced effector T cell functions through upregulation of the genes associated with cytotoxicity programs and effector differentiation. Conversely, HDAC3 inhibition negatively affected long-term T-cell persistence, which attenuated the recall T-cell response upon antigen rechallenge. Recent studies have shown that HDAC inhibitors can enhance the antitumor immune response by modulating several immune cell functions. Vorinostat and trichostatin A (class I/II inhibitor) decreased the infiltration of myeloid-derived suppressor cells into tumors, resulting in the activation of cytotoxic T cell functions.^{60, 61} The HDAC8-specific inhibitor PCI-34051 induced the expression of T cell-trafficking chemokines in hepatocellular carcinoma, resulting in the increased infiltration of CD8⁺ T cells within TME.⁶²

The histone methyltransferase DOT1L catalyzes mono-, di-, and trimethylation of H3K79. We previously reported that DOT1L inhibition in human T cells suppressed allogeneic T cell reactions.⁶³ CD8⁺ T cells treated with the DOT1L inhibitor SGC0946 showed increased levels of the ERK phosphatase DUSP6, which shortened the duration of ERK phosphorylation triggered upon TCR stimulation. Interestingly, DOT1L inhibition did not compromise the high-avidity T-cell response mediated by genetically overexpressed CAR or TCR, suggesting its potential usefulness in selectively suppressing off-target T-cell reactions. Another study investigated the kinetics of DOT1L-KO T cells in mice.⁶⁴ T cell-specific Dot1l deletion resulted in the loss of naïve CD8⁺ T cell populations and promoted premature differentiation into a T_CM phenotype, independent of antigen exposure. As Dot1l deletion also alters epigenetic profiles other than H3K79 methylation, it could regulate gene expression profiles in a complex manner by interacting with other epigenetic factors.