Adenosine Deaminase 1 Overexpression Enhances the Antitumor Efficacy of Chimeric Antigen Receptor-Engineered T Cells

Abstract

Chimeric antigen receptor (CAR) T cell therapy mediates unprecedented benefit in certain leukemias and lymphomas, but has yet to achieve similar success in combating solid tumors. A substantial body of work indicates that the accumulation of adenosine in the solid tumor microenvironment (TME) plays a crucial role in abrogating immunotherapies. Adenosine deaminase 1 (ADA) catabolizes adenosine into inosine and is indispensable for a functional immune system. We have, for the first time, engineered CAR T cells to overexpress ADA. To potentially improve the pharmacokinetic profile of ADA, we have modified the overexpressed ADA in two ways, through the incorporation of a (1) albumin-binding domain or (2) collagen-binding domain. ADA and modified ADA were successfully expressed by CAR T cells and augmented CAR T cell exhaustion resistance. In a preclinical engineered ovarian carcinoma xenograft model, ADA and collagen-binding ADA overexpression significantly enhanced CAR T cell expansion, tumor tissue infiltration, tumor growth control, and overall survival, whereas albumin-binding ADA overexpression did not. Furthermore, in a syngeneic colon cancer solid tumor model, the overexpression of mouse ADA by cancer cells significantly reduced tumor burden and remodeled the TME to favor antitumor immunity. The overexpression of ADA for enhanced cell therapy is a safe, straightforward, reproducible genetic modification that can be utilized in current CAR T cell constructs to result in an armored CAR T product with superior therapeutic potential.

INTRODUCTION

Despite revolutionizing the treatment of B cell malignancies,^1–4 chimeric antigen receptor (CAR) T cells have had disappointing efficacy against solid tumors.^5–7 A formidable roadblock to the use of CAR T cell in solid tumor treatment is the immunosuppressive tumor microenvironment (TME).^8,9 Among the variety of cellular populations, soluble factors, and structural proteins that contribute to a pro-TME,^10–12 the past decade has witnessed a focus on the ubiquitous purine nucleoside adenosine.^13–15 Originally identified as a molecule essential for preventing excessive tissue damage during inflammation,¹⁶ adenosine has since been discovered to reach micromolar concentrations in the TME,^17,18 especially due to an increase in the expression of adenosine-generating ectonucleotidases CD39 and CD73.^19,20 In contrast to the source of its generation, the immunostimulatory adenosine triphosphate, adenosine is a potent anti-inflammatory signal, suppressing the function of antitumor immune cells while simultaneously enhancing the activity of suppressor cells.²¹ Four adenosine receptors (ARs), A₁R, A_2AR, A_2BR, and A₃R, bind adenosine,²² and the predominant AR subtype present on T lymphocytes is A_2AR.^23,24 A_2AR stimulation by adenosine dampens T cell receptor activation through cyclic adenosine monophosphate-protein kinase A signaling, leading to severe inhibition of T cell effector function and proinflammatory cytokine production.^25–27

Previous studies have addressed the problem of adenosine accumulation in the TME by inhibiting the enzymatic activities of CD39 and CD73 to reduce adenosine production,^28,29 or by blocking A_2AR with antagonists to prevent the activation of A_2AR and downstream signaling pathways.^30,31 Both strategies ameliorate the immunosuppressive effect of adenosine on antitumor T cells, more specifically CD8⁺ CAR T cells, and restore their antitumor efficacy.^32,33

Adenosine deaminase is an enzyme involved in purine metabolism. It catalyzes the irreversible degradation of adenosine into inosine and is essential for a functional immune system by preventing the buildup of toxic metabolites.^34,35 One isoform, adenosine deaminase 1 (ADA), is widely expressed in most cells in the body, particularly lymphocytes and macrophages, and also has a costimulatory role for T cells.³⁶ Overexpression of ADA on CAR T cells is a potential strategy to reduce adenosine accumulation in the solid TME and ameliorate the hypofunction of CAR T cells caused by adenosinergic signaling.

ADA has a short half-life in circulation, which is a common challenge shared by other protein drugs.³⁷ One strategy to improve the retention time of protein drugs is to modify proteins with albumin-binding peptides.³⁸ Albumin is an abundant 67 kDa serum protein with an exceptionally long half-life (19 days in humans) due to its interactions with Fc natal receptor, and proteins bound to albumin exhibit increased in vivo persistence. Rather than encumbering a protein with albumin, a short peptide that binds albumin can take advantage of the unique serum longevity of albumin while minimizing the risk to protein functionality.³⁹

Another strategy to improve the retention time of ADA in vivo takes advantage of the extracellular matrix of solid tumors. Cancerous extracellular matrix remodeling can result in increased concentrations of collagen and uncontrolled, unorganized tumor growth causes hyperpermeable tumor vasculature.⁴⁰ This leakiness causes tumor-associated collagen to be exposed to molecules in the blood preferentially to other tissues.^41,42 The von Willebrand factor (VWF) is a hemostasis that tightly binds collagen I and III to initiate thrombosis upon blood vessel injury, and within VWF, the A3 domain is responsible for collagen binding.⁴³ The conjugation of VWF A3 to immune checkpoint inhibitors and fusion to interleukin-2 (IL-2) has been demonstrated to enhance the targeted delivery of the immunomodulatory agents to the TME.⁴⁴ ADA linked with the collagen-binding domain VWF A3 may preferentially accumulate in the TME for localized adenosine-degrading function.

In this study, we transduced CAR T cells to overexpress ADA in a single bicistronic vector with an anti-CD19 CAR. To improve the pharmacokinetic profile of overexpressed ADA, we proposed two ways to modify it through the incorporation of (1) albumin-binding domain or (2) collagen-binding domain. We proved that CAR T cells overexpressing ADA/modified ADA exhibited enhanced exhaustion resistance and better functions in adenosine-rich environment. CAR T cells overexpressing ADA and ADA with collagen-binding domain outperformed parental CAR T cells in an engineered ovarian carcinoma xenograft solid tumor model. We further validated ADA overexpression as a potential cancer treatment in a proof-of-concept syngeneic colon cancer solid tumor model, in which the overexpression of mouse ADA by cancer cells reduced tumor burden and remodeled the TME to favor antitumor immunity compared with control cancer cells.

MATERIALS AND METHODS

General methods for cell culture, antibodies, enzyme-linked immunosorbent assay (ELISA), and Western blotting analysis are detailed in the Supplementary Data. Protocols used for retroviral vector production, T cell transduction and expansion, surface immunostaining analysis, and intracellular cytokine staining analysis can be found in a previous publication.⁴⁵

Mice

Female 6–8-week-old NOD.Cg-Prkdc^scidIL2Rγ^tm1Wj1/SZ (NSG) were purchased from The Jackson Laboratory. All animal studies were performed in accordance with the Animal Care and Use Committee guidelines of the NIH (Bethesda, MD) and were conducted under protocols approved by the Animal Care and Use Committee of the USC.

Plasmid design

The retroviral vector encoding anti-CD19 CAR (CAR) was constructed based on the MP71 retroviral vector kindly provided by Prof. Wolfgang Uckert. The vector encoding anti-CD19 CAR with ADA (EC. 3.5.4.4) overexpression (ADACAR) was then generated from the anti-CD19 CAR. The insert for ADACAR consisted of the following components in frame 5′ to 3′ end: anti-CD19 CAR, P2A linker, human IL-2 leader sequence, and ADA. For the vectors encoding anti-CD19 CAR with albumin-binding ADA (CD19.aADA) or collagen-binding ADA (CD19.cADA), ADA is followed by a GS linker and the albumin-binding peptide SA21 or a collagen-binding domain VWF A3, respectively.^39,44 The lentiviral vectors encoding his-tagged ADA with albumin binding (FUGW.aADA.his) and ADA with CBD binding (FUGW.cADA.his) were based off the FUGW backbone. A His-tag was added to the C terminus of the albumin- or collagen-binding sequence.

aADA and cADA isolation and characterization

HEK-293T cells were transduced with FUGW.aADA.his or FUGW.cADA.his lentivirus for the stable expression of his-tagged aADA or cADA. Following successful transduction and expansion, the engineered cells were seeded in 10-mL plates in D10. Sixteen hours later, the cells were rinsed twice with phosphate-buffered saline (PBS) and then cultured for 48 h in 10 mL serum-free media. Supernatants were subsequently collected, clarified, and then centrifuged in 10 kDa isolation columns (Sigma) for 1 h at 5,000g 4°C. The remaining supernatant was purified for his-tagged protein using Dynabeads (Thermo) according to the manufacturer's instructions, and standard bicinchoninic acid assay (Sigma) was used for quantification of the purified proteins. Specific activity was measured using an ADA Activity Kit (Sigma) according to the manufacturer's instructions. Binding assays were conducted as previously described for the albumin- and collagen-binding domains.^39,44 Briefly, 96-well plates (Maxisorp) were coated with 10 ng/mL human serum albumin (HSA) (Sigma), mouse serum albumin (MSA) (Sigma), or collagen III (Sigma) in PBS for 4 h at 4°C, after which HSA- and MSA-coated wells were blocked with 1% ovalbumin in PBS with 0.05% Tween 20 (PBS-T) and collagen-coated wells blocked with 2% bovine serum albumin (BSA) in PBS-T for 1 h at room temperature (RT). Then, wells were washed with PBS-T and incubated with increasing concentrations (0, 5, 25, 50, 150, 300 nm in duplicates) of aADA or cADA for 1.5 h at RT. After three washes with PBS-T, wells were incubated for 1 h at RT with rabbit anti-hADA (BosterBio) antibody, followed by washes and 1 h RT incubation with secondary horseradish peroxidase-conjugated anti-rabbit antibody. After washes, enzyme concentrations were detected with tetramethylbenzidine substrate by measurement of the absorbance at 450 nm with subtraction of the absorbance at 570 nm. The apparent Kd values were obtained by nonlinear regression analysis in Prism software (version 7; GraphPad Software) assuming one-site–specific binding.

Tumor model and adoptive cellular transfer

Six- to 8-week-old NSG mice were injected subcutaneously with 3.5 M SKOV3.CD19. Tumor volume was determined by caliper measurement (L × W²/2). Once tumors reached an average size of 50–80 mm³, CAR T cells were injected intravenously at the dose of 3M CAR+ cells per group. Tumor sizes were measured three times a week and mice were euthanized when they displayed obvious weight loss, ulceration of tumors, or tumor size larger than 750 mm³.

Ex vivo analysis

Tumor, spleen, and bone marrow tissue were harvested from mice and filtered through 70 μm nylon strainers (BD Falcon, Franklin Lakes, NJ, USA) for single-cell suspensions. Blood from cardiac puncture was immediately transferred to TAC buffer for red blood cell lysis. The filtered cells and blood samples were washed and incubated with 1% BSA in PBS. Cells were then stained for fluorescent-activated cell sorting (FACS). For peripheral blood analysis, mice were tail bled. A portion of the collected whole blood was added to TAC buffer for FACS staining, and the remainder was left undisturbed at RT for 30 min. The coagulated blood samples were then centrifuged at 1,000g for 10 min, and the supernatant was immediately transferred to a clean polypropylene tube for assessment of inosine concentration in serum using an Inosine Assay Kit (Sigma-Aldrich, St. Louis, MO, USA).

Statistical analysis

Statistical analysis was performed in GraphPad Prism version 5.01. The differences among three or more groups were determined with one-way analysis of variance (ANOVA) with Tukey's posttest for multiple comparisons. Tumor growth curves were analyzed using two-way ANOVA with Tukey's posttest for multiple comparisons. Mice survival curves were evaluated by the Kaplan–Meier analysis (log-rank test with Bonferroni correction). A p-value <0.05 was considered statistically significant. Significance of findings were defined as: ns = not significant, p > 0.05; *p < 0.05; **p < 0.01; ***p < 0.001, ****p < 0.0001. In vitro assays are performed in triplicate and representative of at least two independent experiments.

RESULTS

Design, generation, and validation of CAR T cells engineered to secrete ADA

A preliminary combination study in which commercial bovine ADA was injected systemically into tumor-bearing mice in conjunction with CAR T cells has shown that ADA injections initially enhanced tumor growth inhibition, but this did not culminate in improved overall survival (Supplementary Fig. S1A–C). The results indicate that sustained administration of ADA may be necessary to engender long-term benefits. Therefore, we engineered CAR T cells to secrete ADA constitutively to circumvent the challenges associated with continuous exogenous delivery of ADA to the TME. Retroviral vector constructs used for our anti-CD19 CAR (CAR19) T cells, anti-CD19 CAR with ADA secretion (CAR19.cADA), anti-CD19 CAR secreting ADA with albumin-binding domain (CAR19.aADA), and anti-CD19 CAR secreting ADA with collagen-binding domain (CAR19.cADA) are depicted in Fig. 1A.

Figure 1.

Design and characterization of CAR and ADA expression in transduced PBMCs. (A) Schematic representations of parental anti-CD19 CAR T cells (CAR19) and ADA overexpressing anti-CD19 CAR T cell (CAR19.ADA, CAR19.aADA, CAR19.cADA) constructs. (B) Expression of the four CAR constructs. T cells were stained with biotinylated goat anti-mouse antibodies followed by APC-conjugated streptavidin to detect CAR expression on the cell surface. (C, D) Overexpression of ADA in the supernatant from 48 h T cell-only culture or coculture with SKOV3.CD19 at a 1:1 effector T cell:tumor cell ratio was analyzed by ELISA. (E) Western blot analysis of ADA. T cells were cultured in the presence of Brefeldin A followed by cell lysis and protein quantification using standard BCA assay. Around 20 μg total protein was then used for anti-ADA immunoblotting. (F) Characterization of modified ADA. *p < 0.05, **p < 0.01. ADA, adenosine deaminase 1; APC, antigen presenting cell; BCA, bicinchoninic acid; CAR, chimeric antigen receptor; ELISA, enzyme-linked immunosorbent assay; PBMC, peripheral blood mononuclear cell.

Two days after T cell activation, cells were transduced with retroviral vectors and expanded for 12 days before cryopreservation. All constructs were efficiently transduced into and expressed by T cells and CAR expression levels are shown in Fig. 1B. Before experiments, we thawed T cells and normalized CAR expression levels among groups. Cell culture supernatants and cell lysates were assessed for ADA and modified ADA. A significantly higher concentration of ADA in the cell culture supernatants of ADA-engineered CAR T cells was observed by ELISA (Fig. 1C, D), both when T cells were cultured alone and cocultured with antigen-presenting tumor cells. We further validated the presence and size of CAR T cell-expressed aADA and cADA by immunoblot following protein secretion-inhibited T cell culture (Fig. 1E). Cell lysates were stained for ADA and, as expected, endogenous ADA expression was identified for T cell groups, and the aADA and cADA were identified at their respective molecular weights, 44 and 62 kDa. Intracellular ADA and ecto-ADA flow cytometry analysis did not reveal differences in ADA expression levels (Supplementary Fig. S2A–C), likely due to the endogenous production and transport of ADA. To produce the necessary quantities of purified protein for binding analysis, we stably transduced HEK-293T cells to secrete his-tagged aADA or cADA. Following protein isolation and quantification, purified aADA and cADA were evaluated for target binding, specific activity, and size (Fig. 1F; Supplementary Fig. S3A–D). The modified ADAs of the predicted molecular weight bound their targets with nanomolar affinities similar to those previously reported for SA21 and VWF A3 fusion proteins,^39,44 and retained the catalytic function of ADA.

Incorporation of ADA, aADA, or cADA into CAR T cells does not comprise in vitro effector functions

ADA-overexpressing CAR T cells exhibited comparable cytotoxicity, proliferation, and interferon gamma (IFN-γ) expression level compared with parental CD19 CAR T cells (Fig. 2; Supplementary Fig. S4A, B). The cytolytic function of the CAR T cells was assessed in 16 h cytotoxicity assays, in which CAR T cells and SKOV3.CD19 cells were cocultured at effector-to-target cell ratios of 1:1, 5:1, and 10:1 (Fig. 2A). The cytotoxicity assays revealed commensurate cytotoxicity among the groups. ADA has been reported to increase the proliferation of T cells,^46,47 but this phenomenon was not observed in our Ki67 staining proliferation studies nor CFSE staining (Fig. 2B; Supplementary Fig. S4A), possibly due to higher concentrations of ADA achieved through exogenous ADA supplementation in previous studies than through ADA overexpression from our modified CAR T cells. Intracellular expression level of IFN-γ after 6 h coculture with target cells was also comparable among CAR-expressing groups (Fig. 2C; Supplementary Fig. S4B). Thus, ADA overexpression does not hinder in vitro effector functions, but potential costimulatory effects were not observed.

Figure 2.

ADA overexpression does not compromise CAR T cell in vitro cytotoxicity, effector cytokine production, and proliferation. (A) Cytotoxicity of the four CAR groups against target cells. CAR T cells were cocultured for 16 h with SKOV3-CD19 cells at 0.2:1, 1:1, 5:1, and 10:1 effector-to-target ratios, and cytotoxicity against SKOV3-CD19 was measured. NT T cells were used as a control. (B) The four groups of CAR T cells were cocultured for 24, 48, 72, and 96 h with SKOV3-CD19 cells at 1:1 effector-to-target ratios, and Ki67 expression of CD3⁺ cells was measured. NT T cells were used as a control. (C) T cells were plated 6 h before SKOV3-CD19 cells were added at 1:1 effector-to-target ratios and brefeldin A was supplemented into the coculture. Six hours after coculture, IFN-γ expression of CD3⁺CD8⁺ cells was measured through intracellular cytokine staining, representative FACS scatterplots gated on CD3⁺CD8⁺ cells shown. NT T cells were used as a control. FACS, fluorescent-activated cell sorting; IFN-γ, interferon gamma.

ADA overexpression protects CAR T cells from the upregulation of the exhaustion-related inhibitory receptors, T regulatory induction, and adenosine immunosuppression

To assess the effect of ADA overexpression on protecting CAR T cells from the upregulation of exhaustion-related immunosuppressive receptors, parental and modified CAR T cells were cocultured with SKOV3.CD19 target cells in the presence or absence of 50 μM adenosine for 48 h, rechallenged with tumor cells (with or without 50 μM adenosine), and then analyzed at 96 h by flow cytometry for the expression of inhibitory receptors, PD-1, LAG-3, and TIM-3. We chose the concentration of 50 μM based on previous studies by the Albelda laboratory, in which a reduction of in vitro targeted cell lysis by human CAR T cell occurred at 50 μM adenosine.⁴⁸ Baseline precoculture exhaustion marker expression and phenotype proportions were similar across groups (Supplementary Fig. S4C). After 96 h, parental CAR19 T cells had an increase in the percentage of triple-positive (PD-1⁺TIM-3⁺LAG-3⁺) CD8⁺ cells when cultured with cancer cells in the presence of adenosine, whereas ADA-overexpressing CAR T cells were resistant to adenosine-mediated inhibitory receptor upregulation (Fig. 3A, B). Importantly, with or without adenosine supplementation, ADA-overexpressing CAR T cells had significantly lower percentages of triple-positive cells. Exogenous addition of 10 mU bovine serum ADA protected CAR19 T cells from adenosine-mediated exhaustion-related marker upregulation and reduced the proportion of triple-positive cells, although to a lesser extent than ADA overexpression. Furthermore, under the same repeated challenge conditions, ADA overexpression or exogenous addition improved CAR T cell cytotoxicity when assessed at 96 h (Fig. 3C).

Figure 3.

ADA overexpression prevents CAR T cell upregulation of inhibitory markers, preserves T cell effector function, and reduces T_reg differentiation. (A, B) CAR T cells were cocultured for 48 h with SKOV3-CD19 cells at a 1:1 effector-to-target ratio in the presence or absence of 50 μM supplemented adenosine, rechallenged with tumor cells (with or without 50 μM adenosine), and stained for the expression of exhaustion markers 48 h after rechallenge (96 h culture total). NT T cells (not shown) were used as a control. (C–F) CAR T cells were cocultured for 24 h with SKOV3-CD19 cells at a 1:1 effector-to-target ratio in the presence or absence of 50 μM supplemented adenosine and assessed for cytotoxicity (C), IFN-γ expression (D), IL-2 expression (E), and T_reg differentiation (F). Brefeldin A was added after 18 h of coculture for intracellular staining analysis of IFN-γ and IL-2 expression at 24 h. (G) Following a 24-h coculture of CAR T cells and SKOV3-CD19 cells at 1:1 effector-to-target ratio, T cells were stained for the expression of PD-L1. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001. IL-2, interleukin-2.

We further characterized CAR T cell performance by evaluating cytokine production, T regulatory (T_reg) cell differentiation, and PD-L1 expression following 24 h cocultures with target cells in the presence or absence of 50 μM adenosine. In contrast to the 6 h IFN-γ expression results (Fig. 2C), ADA overexpression increased the percentage of CAR T cells positive for IFN-γ at 24 h (Brefeldin A added at 18 h). Although adenosine supplementation dampened IFN-γ expression in all groups, ADA-overexpressing CAR T cells maintained higher IFN-γ expression in the face of additional adenosine (Fig. 3D). CAR19.ADA, CAR19.aADA, and CAR19.cADA T cells had higher IL-2 expression than parental CAR T cells and, in contrast to CAR19 T cells, were resistant to adenosine-mediated IL-2 suppression (Fig. 3E). Exogenous ADA also mitigated adenosine's reductive effects on cytokine production. In conjunction with the preservation of effector functions, the overexpression and exogenous addition of ADA reduced the percentage of CD4⁺CD25⁺FOXP3⁺ T_reg cells (Fig. 3F). Lastly, coinhibitory receptor PD-L1 upregulation was assessed, as PD-L1 engagement on T cells was recently shown to promote self-tolerance and suppression of neighboring effector T cells in cancer,⁴⁹ and unmodified CAR 19 T cells had the highest percentage of CD8⁺PD-L1⁺ cells (Fig. 3G). The pronounced reduction in expression of critical T cell inhibitory receptors and T_reg differentiation, as well as enhanced adenosine immunosuppression resistance, suggest that ADA overexpression can augment CAR T cell resistance to immunosuppressive signaling in the TME.

ADA overexpression and cADA expression enhance CAR T cell antitumor activity and improve overall survival in a xenograft solid tumor model

Following the promising in vitro assessment of ADA-overexpressing CAR T cells, we evaluated the therapeutic potency of our treatments in an immunocompromised mouse model with xenograft ovarian cancer. Once subcutaneous SKOV3.CD19 tumors reached a volume of 50–80 mm³, 3 million CAR⁺ T cells were administered for treatment and tumor growths were monitored (Fig. 4A). All Nontransduced (NT) group mice reached endpoint by day 15, and by day 18 the CAR19 and CAR19.aADA groups had reached their median overall survival, achieving an increase in median OS of 6.7% and 20%, respectively, when compared with the control mice (Fig. 4C; Supplementary Fig. S5A). The CAR19.ADA- and CAR19.cADA-treated mice had significantly reduced tumor burden by day 13 when compared with all other groups, and experienced prolonged overall survival, both with an approximately doubled median OS compared with the control mice.

Figure 4.

The overexpression of ADA and cADA increases overall survival. (A) Schematic representation of experiment procedure for tumor challenge and adoptive cellular transfer. (B) Tumor growth curve for mice treated with NT or CAR T cells. Data were presented as mean tumor volume ± SEM at indicated time points (n = 7 for all groups). (C) Mouse survival curves for the different treatment groups were calculated using the Kaplan–Meier method (n = 7). (D) The percentage of human CD45⁺ T cells in the blood of SKOV3.CD19-bearing mice that were treated with NT or CAR T cells was analyzed by flow cytometry at day 12. (E) Ratio CD45⁺CD4⁺FOXP3⁺ T cells in peripheral blood. (F) Relative concentration of inosine in peripheral serum compared with inosine concentration in CAR19-treated mouse serum (n = 3 mice/group). *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001. SEM, standard error of the mean.

On day 12 posttreatment, three mice from each group were tail bled for circulating T cell analysis and serum inosine concentration measurements. CAR19.cADA and CAR19.cADA T cells were more prevalent in the blood (Fig. 4D). ADA overexpression also resulted in a decreased proportion of peripheral T_reg cells (Fig. 4E). Serum inosine levels were analyzed using the Inosine Assay Kit (Sigma), which revealed an increase in inosine concentration contingent on ADA or cADA overexpression (Fig. 4F).

ADA overexpression and cADA expression enhance intratumoral CAR T cell expansion

To study whether the enhanced antitumor efficacy observed in the CAR19.cADA and CAR19.cADA groups are correlated with intratumoral CAR T cells, we performed another animal study following a similar schematic (Fig. 5A). Once the subcutaneous SKOV3.CD19 tumors reached a volume of 50–80 mm³, 3 million CAR⁺ T cells were administered. On day 10 the mice were sacrificed and ex vivo analysis was performed. When compared with unmodified CAR T group, CAR19.cADA and CAR19.cADA groups displayed superior in vivo expansion as shown by increased percentages of CD45⁺ T cells in the tumor, blood, spleen, and bone marrow (Fig. 5B; Supplementary Fig. S5B). Cells harvested from the organs were further characterized for phenotype, PD-1 expression, and pCREB expression. Notably, only the CAR19.cADA tumor infiltrating lymphocytes (TILs) displayed an increased CD8⁺:CD4⁺ ratio (Fig. 5C). TILs from all the ADA-overexpressing groups displayed lower PD-1 expression compared with parental CAR T cells (Fig. 5D), whereas the intensities for pCREB were equivalent between groups (Supplementary Fig. S5C). Although the percentage of T_reg in the CD4⁺ helper T cell populations in the tumor and spleen were commensurate among the parental and modified CAR T cell groups (Supplementary Fig. S5D), the increase in CD8⁺ TILs for the cADA-overexpressing treatment resulted in a significantly higher CD8⁺:T_reg ratio in the TME (Fig. 5E).

Figure 5.

ADA-overexpressing CAR T cells exhibit enhanced in vivo expansion. (A) Schematic representation of experiment procedure for tumor challenge and adoptive cellular transfer. NSG mice were subcutaneously challenged with 3.5 × 10⁶ SKOV3.CD19 tumor cells. Once the tumors reached a size of 50–80 mm³ (day 0), 3 × 10⁶ CAR⁺ T cells were adoptively transferred through intravenous injection. (B) The percentage of human CD45⁺ T cells in the tumor, blood, spleen, and bone marrow of SKOV3.CD19-bearing mice that were treated with NT or CAR T cells was analyzed by flow cytometry at day 10 (n = 3/group). (C) CD8⁺:CD4⁺ ratios of T cells in the tumor, blood, spleen, and bone marrow. CAR19.cADA.CBD cells exhibit a higher CD8⁺:CD4⁺ T cell ratio at the local tumor site. (D) PD-1 expression in TILs. (E) Ratio of CD45⁺CD8⁺:CD45⁺CD4⁺FOXP3⁺ cells in the tumor tissue (n = 3 mice/group). *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001.

ADA overexpression reduces tumor burden and remodels the TME to favor antitumor immunity in a syngeneic solid tumor model

Adenosine precipitates tumor tolerance and promotion by binding to ARs on a variety of immune and stromal cellular populations.^21,28 While immunocompromised preclinical models enable the study of human T cells and cancer cells, corroborating ADA overexpression in a syngeneic model is necessary to address the multifaceted role adenosine plays in the TME. In a straightforward ADA gene therapy proof-of-concept study, we transduced CT26 murine colon carcinoma cells to express green fluorescent protein (GFP) (CT26.GFP) or GFP and mADA (CT26.GFP.mADA) and sorted for GFP⁺ cells of comparable intensity (Fig. 6A). We verified mADA overexpression (Fig. 6B) and commensurate growth rates for the two engineered CT26 cell lines (Fig. 6C). We then challenged BALB/c mice with subcutaneously inoculated CT26.GFP or CT26.GFP.mADA. Regardless of the starting tumor inoculum (1M or 3M cancer cells), the CT26.GFP.mADA groups exhibited significantly smaller tumor burden by day 25 when compared with the CT26.GFP groups (Fig. 6E). Weights of the mice were monitored and no toxicities related to mADA overexpression were observed (Fig. 6D).

Figure 6.

Mouse ADA expression reduces the growth rate of CT26 subcutaneous tumors and remodels the TME to favor antitumor immunity. (A) CT26 cells were stably transduced with lentiviral vectors expressing FUGW or FUGW.mADA, and sorted with flow cytometry for comparable GFP levels. (B) Overexpression of mADA in cell culture supernatant was measured by ELISA. (C) In vitro growth kinetics of CT26.GFP and CT26.GFP.mADA cells. (D, E) BALB/c mice were injected subcutaneously with 1 × 10⁶ or 3 × 10⁶ CT26.GFP or CT26.GFP.mADA cancer cells and monitored for tumor growth (D) (n = 5 mice/group) and weight (E). (F–K) Fourteen days after CT26.GFP or CT26.GFP.mADA tumor cells were inoculated into BALB/c mice, tumors were excised, homogenized, and analyzed by flow cytometry for surface marker expressions (n = 5 mice group), data are shown as the percentage of CD45⁺ cells, except for (G), which is shown as the percentage of CD45⁺CD8⁺ cells. *p < 0.05; **p < 0.01; ***p < 0.001. GFP, green fluorescent protein.

We performed subsequent in vivo studies to characterize immune cell populations in CT26.GFP and CT26.GFP.mADA tumors. Fourteen days after tumor inoculation, mice (5/group) were sacrificed and tumor tissue extracted, homogenized, and stained for cell surface markers. The mADA-overexpressing CT26 tumors had an increased percentage of CD8⁺ T cells and a lower percentage of which were coexpressing PD-1 and TIM-3 (Fig. 6F, G). CT26.GFP.mADA tumors also had a higher proportion of NK cells (CD45⁺CD3⁻NK1.1⁺; Fig. 6H). Concurrently, CT26.GFP.mADA had a lower prevalence of several immunosuppressive immune populations, including T_reg cells (CD45⁺CD4⁺CD25⁺FOXP3⁺; Fig. 6I), CD206 expressing macrophages (CD45⁺CD11b⁺F4/80⁺CD206⁺; Fig. 6J), and monocytic myeloid-derived suppressor cells (mMDSC) (CD45⁺CD11b⁺Ly6C⁺Ly6G⁻; Fig. 6K), whereas there was no significant difference in granulocytic MDSC (CD45⁺CD11b⁺Ly6C⁺Ly6G⁺, data not shown). The immune cell infiltration analysis indicates that ADA fosters antitumor immunity by increasing the presence of effector cells and decreasing the presence of T_reg cells, immunosuppressive macrophages, and mMDSCs.

DISCUSSION

In this study, we engineered CAR T cells to overexpress ADA or express protein-modified ADA for enhanced pharmacokinetic profiles and demonstrated that the overexpression increased the amount of ADA secreted by the CAR19.ADA, CAR19.aADA, and CAR19.cADA T cells. The sizes, binding affinities, and enzymatic activities were confirmed for ADA with albumin-binding domain and ADA with collagen-binding domain. We found that ADA overexpression was not toxic to CAR T cells, nor did it impede in vitro effector functions, as the CAR groups had commensurate cytotoxicity, proliferation, and 6-h IFN-γ production. Discrepancies surfaced when performing inhibitory receptor and adenosine-supplemented studies. The upregulation of coinhibitory receptors and promotion of T_reg differentiation are well-documented consequences of A_2AR stimulation.^50–52 After repeated challenge cocultures with antigen-presenting cancer cells, ADA overexpression protected CAR T cells from inhibitory receptor upregulation and resulted in a significantly smaller percentage of triple-positive (PD-1⁺TIM-3⁺LAG-3⁺) CD8⁺ cells. In cultures supplemented with adenosine, ADA-overexpressing CAR T cells were resistant to exhaustion-related marker upregulation, whereas parental CAR19 T cells experienced an increase in the percentage of triple-positive cells. After the 96-h repeated challenge, with and without adenosine supplementation, ADA overexpression enhanced CAR T cell cytotoxicity. Studies adopting 24 h CAR T cell and tumor cell cocultures, in the presence or absence of adenosine, revealed that ADA overexpression increased CAR T cell effector cytokine (IFN-γ, IL-2) production, reduced T_reg differentiation, and downregulated PD-L1 expression. Our in vitro assays demonstrated that ADA overexpression protects CAR T cells from the upregulation of the exhaustion-related inhibitory receptors, T_reg induction, and adenosine immunosuppression.

In our engineered ovarian solid tumor model, ADA overexpression as unmodified ADA or cADA resulted in superior tumor growth control compared with parental CAR T cells. CAR19.cADA and CAR19.cADA T cells had robust in vivo expansion, persisting in tumor, blood, spleen, and bone marrow at significantly higher levels than control CAR T cells. Peripheral T_reg have been reported to correlate with poor prognosis in many solid tumor types,⁵³ and we observed a lower prevalence of peripheral T_reg for CAR19.ADA, CAR19.aADA, and CAR19.cADA T cell groups. Further characterization of the CD45⁺ T cells in the tumor revealed that TILs from all three ADA-overexpressing CAR variants had diminished PD-1 expression compared with CAR19 TILs. Only cADA overexpression favored CD8⁺ phenotype in the TME, and markedly increased the CD8⁺:T_reg ratio among TILs. CD8⁺:T_reg ratio can be used to predict response to PD-1 inhibition, and indicates a TME that fosters antitumor immunity.⁵⁴ CAR19.cADA and CAR19.cADA T cells significantly prolonged OS, resulting in approximately double median OS compared with NT and CAR19 control groups. Interestingly, aADA overexpression did not promote superior expansion nor increase OS. Although albumin-binding modifications hold promise for improving anticancer therapeutics and are included in clinically relevant constructs,⁵⁵ binding to albumin was shown to reduce the in vivo efficacy of some drugs.^56,57 We hypothesize that the albumin-binding modification abrogates the benefits of overexpressed ADA because of its binding to albumin. Despite increased half-life, binding to albumin may prevent ADA from accumulating and being accessible in the TME where T cells would benefit directly from ADA activity, and further studies would be required to test this theory. Throughout our in vivo studies, we monitored body weight and witnessed no differences between groups, indicating that the overexpression of ADA and modified ADA will not exacerbate CAR T cell treatment-associated toxicity.

Previously, researchers have developed pegylated form of adenosine deaminase isoenzyme ADA2 (PEG-ADA2) for cancer treatment and the administration of PEG-ADA2 slowed tumor growth in multiple syngeneic models, although survival data were not shown.^58,59 Despite recognition as isoenzymes, ADA and ADA2 differ in structure, cellular localization, and expression. ADA is a 41 kDa monomer expressed in all tissues (highest expression in lymphocytes) intracellularly or bound to CD26, whereas ADA2 is a 59 kDa protein that forms homodimers and is secreted primarily by myeloid cells.⁶⁰ Although ADA2 has higher serum stability, we introduced ADA into our CAR constructs rather than ADA2 for several reasons: (1) ADA is one-hundred fold more catalytically efficient^58,61; (2) ADA, through binding cell membrane receptor CD26, can act as a costimulatory molecule for T cells^36,47; (3) ADA is natively expressed by T lymphocytes, and is thus unlikely to cause toxicity upon its overexpression; and (iv) ADA gene therapy has proven to be effective for treating ADA deficiency severe combined immunodeficiency and demonstrated a good safety profile.⁶²

The therapeutic efficacies of CAR19.cADA and CAR19.cADA groups demonstrated the advantage of CAR T cells to locally deliver additional ADA to the TME. Crosslinking ADA with collagen-binding domain utilizes the abundant collagen in extracellular matrix of tumor structure and could allow more ADA enzymatic activity at tumor site to degrade adenosine. Compared with systemic delivery of ADA that requires frequent injections and wastes much injected ADA due to fast clearance from the body, ADA-overexpressing CAR T cells (CAR19.ADA, CAR19.cADA) present advantages of better therapeutic efficacy in terms of enhanced overall survival and convenience in administration.

Despite the efficacy of ADA overexpression shown in our study, the roles of ADA and its catalysis product inosine in cancer progression remain controversial. ADA blockade suppressed the progression of breast cancer in a 4T1 preclinical model,⁶³ and inosine, an A_2AR agonist, can facilitate immunosuppression.⁶⁴ Despite these findings, supplementation with inosine enhanced the antitumor efficacy of immune checkpoint inhibitors and ACT in multiple solid tumor models.^65,66 Interestingly, inosine has a unique signaling bias upon A_2AR binding, favoring the ERK1/2 pathway to PKA stimulation, and, relative to adenosine, inosine is approximately four orders of magnitude less potent at A_2AR (EC50 300 μM vs. 6 nM),^67,68 which may account for its differing effect on immunotherapies.

We also recognize that our NSG preclinical studies have shortcomings. We chose SKOV3 as our tumor model for its prevalent adenosine signaling,^19,69,70 and modified the cell line to stably express CD19. Studies using ADA-overexpressing CAR T cells targeting solid tumor antigens, such as CEA or HER2, could be conducted to better evaluate the clinical efficacy of ADA overexpression. Additionally, although the NSG model allows the study of human T cells and cancer cells, it fails to accurately represent the complex TME found in patients. Adenosine is a key suppressor of antitumor immunity, not only through its effect on T cells but also by thwarting the antitumor activity of natural killer cells and polarizing dendritic cells, macrophages, myeloid-derived suppressor cells, endothelial cells, and fibroblasts toward tolerogenic, protumor phenotypes.⁷¹ To address the dynamic interplay of adenosine in the TME, we assessed ADA overexpression in a syngeneic solid tumor model. In this proof-of-concept study, CT26 colon cancer cells expressing GFP or GFP and mADA secretion were inoculated subcutaneously into BALB/c mice. Without any treatment, tumors with mADA secretion grew at significantly slower rates in comparison to CT26.GFP tumors in vivo, suggesting the antitumor effects of overexpressed ADA in immunocompetent mice beyond its effect on CAR T cell treatment. In accordance with slower tumor growths, excised CT26.GFP.mADA tumors had an increase in CD8⁺ T cells and NK cells compared with CT26.GFP tumors. ADA protected CD8⁺ T cells from the upregulation of PD-1 and TIM-3 as shown by the lower frequency of PD-1⁺Tim-3⁺ double-positive cells. ADA-overexpressing tumors also had smaller proportions of several immunosuppressive cellular populations, including T_reg cells, CD206⁺ macrophages, and mMDSC, which indicated that ADA overexpression remodeled the TME to favor antitumor immunity.

The strategy of ADA overexpression in CAR T cells to enhance CAR T antitumor efficacy in solid tumors has potential translational value. The first human gene therapy was performed thirty years ago for ADA-deficient patients, in which autologous peripheral T cells were extracted and transduced with human ADA, followed by reinfusion into the patients.⁶² This track record proved the clinical safety profile of ADA gene therapy and allowed its potential use for other diseases.

ADA overexpression provides another option in combating adenosine immunosuppression in the TME. Extensive studies of A_2AR inhibition or CD73/CD39 blockade have solidified the adenosinergic pathway as a targetable checkpoint, but there are potential downsides to current approaches. A_2AR inhibitors fail to prevent immunosuppression caused by other ARs (i.e., adenosine inhibits dendritic cell activation through A₁R signaling).⁷² There are alternative adenosine production pathways,⁷³ which tumors may exploit to compensate for CD39/CD73 inhibition. T cell intrinsic modifications, such as the genetic ablation of the A_2AR or intracellular expression of PKA localization altering peptide,^33,48 enhance the CAR T cell product but fail to stymie adenosine signaling for other immune cell types in the TME. ADA, by directly targeting adenosine, has the potential to alleviate the widespread immunosuppression of adenosine irrespective of AR. The genetic addition of the relatively small ADA transgene (∼1 kb) can be easily incorporated into other immunotherapies, such as oncolytic viruses and cancer vaccines, and these other formulations will deepen our understanding of the therapeutic potential of ADA gene therapy for cancer. Spurred on by the success of our ADA-overexpressing CAR T cells and the favorable safety profile of ADA gene therapy, we hope to add ADA gene therapy to the cancer treatment armamentarium.

AUTHORS' CONTRIBUTIONS

Conception and design: Y.Q., Z.S.D., and P.W. Development of methodology: Y.Q., Z.S.D., X.C., and P.W. Acquisition of data: Y.Q., Z.S.D., X.C., M.M., G.C., H.W., J.L., F.H., and P.W. Analysis and interpretation of data: Y.Q., Z.S.D., X.C., and P.W. Writing, review, and/or revision: Y.Q., Z.S.D., X.C., and P.W. Study supervision: P.W.

AUTHOR DISCLOSURE

No competing financial interests exist.

FUNDING INFORMATION

This work was supported by grants from the National Institutes of Health (R01AI068978, R01CA170820, R01EB017206, and P01CA132681) and a grant from the Ming Hsieh Institute for Research on Engineering-Medicine for Cancer.

SUPPLEMENTARY MATERIAL

Supplementary Data

Supplementary Figure S1

Supplementary Figure S2

Supplementary Figure S3

Supplementary Figure S4

Supplementary Figure S5