Multiple Injections of Electroporated Autologous T Cells Expressing a Chimeric Antigen Receptor Mediate Regression of Human Disseminated Tumor

Adoptive transfer of CTLs has shown great promise in both viral infections and cancers. After many years of disappointing results with chimeric antigen receptor (CAR) T-cell therapy, improved culture systems and cell engineering technologies are leading to CAR T cells with more potent antitumor effects (1). Results from recent clinical trials indicate improved clinical results with CARs introduced with retroviral vectors (2, 3). Perhaps not surprisingly, these CAR T cells also exhibit enhanced toxicity (4, 5). Recent editorials have discussed the need for safer CARs (6, 7).

The receptor transfer strategies described above used retroviral vector transduction that results in stable genomic integration of the transgene. This allows for constitutive expression of the transgenic receptors. However, the integration of the provirus into the genome bears the risk of insertional mutagenesis and, at least theoretically, malignant transformation of the transduced cells. In addition, stable expression of the transgene may be a disadvantage when unintended cross-reactivity of the transgenic immunoreceptor results in severe adverse effects as reported recently (5, 8).

Here, we report that by combining a robust T-cell culture system (9) with the optimized mRNA CAR electroporation protocol described herein, we have developed a platform that has the potential to increase the therapeutic window with CARs that contain increasingly potent signaling domains. Using good manufacturing practice (GMP)–grade RNA encoding a CAR against mesothelin, a glycosyl-phosphatidylinositol–linked molecule that is overexpressed on ovarian and pancreatic cancer and mesothelioma (10), we show robust antitumor effects in preclinical models. Most notably, significantly prolonged survival and reduced tumor burden was observed in treated mice compared with control groups, even when using autologous T cells from a patient with advanced metastatic cancer. Electroporation of T cells with optimized RNA CARs provides a novel and cost-efficient platform for the treatment of cancer without the associated safety concerns of integrating gene vectors.

Mesothelin (ss1) and CD19-specific CARs (11, 12) were optimized as described in detail in Supplementary Materials and Methods.

Three RNA in vitro transcription (IVT) systems were used to optimize RNA expression in T cells as described in detail in Supplementary Materials and Methods.

Anonymous healthy donors donated lymphocytes at the University of Pennsylvania Apheresis Unit after informed consent under an Institutional Review Board–approved protocol, and T cells were purified by elutriation. In some experiments, we used cryopreserved T cells and tumor cells from the same patient. “Patient 108” had malignant mesothelioma. As part of an earlier clinical trial, this patient underwent leukapheresis and had tumor cells generated from his malignant pleural effusion. T cells were activated by addition of CD3/CD28 beads (Invitrogen) and a single cycle of stimulation as described (9). For the experiment shown in Fig. 5, patient 108 T cells were stimulated with irradiated artificial antigen-presenting cells expressing 4-1BBL and loaded with anti-CD3 monoclonal antibody (mAb) OKT3 and CD28 mAb 9.3 as described (13). T cells were maintained at a density of 0.8 × 10⁶ to 1 × 10⁶ cells/mL in RPMI 1640 with 10% FCS and 1% penicillin-streptomycin (R10) after bead stimulation.

Activated T cells were electroporated on day 10 of culture as described in Supplementary Materials and Methods.

A slightly modified version of a flow cytometry cytotoxicity assay was used (14).

Studies were performed as previously described with certain modifications (15, 16) as described in Supplementary Materials and Methods.

Analysis was performed with STATA version 10 (StataCorp) or Prism 4 (GraphPad Software). In vitro data represent means of duplicates, and comparisons of means were made via Mann-Whitney test. For comparison among multiple groups, Kruskal-Wallis analysis was performed with Dunn multiple comparison tests to compare individual groups. Survival curves were compared using the log-rank test with a Bonferroni correction for comparing multiple curves.

We have previously reported that anti-mesothelin ss1 scFv CARs with combinations of CD3ζ, CD28, and 4-1BB activation domains are highly and stably expressed in T cells when introduced using lentiviral vector technology (11). Human T cells were activated for 10 days as previously described (9), and as the cells returned to a near resting state, they were electroporated with RNA encoding the ss1 scFv with the previously described combinations of signaling moieties. We found that the level of transgene expression was not uniform (Supplementary Fig. S1), as T cells electroporated with CAR bearing CD3ζ alone (ss1-z) showed the highest transgene expression, followed by nearly equivalent levels of ss1-28z (CD28 + CD3ζ) and ss1-bbz (4-1BB + CD3ζ) expression. Because “second-generation” CARs containing costimulation domains seem superior in several preclinical and early-stage clinical trials when expressed with viral vector systems (11, 12, 17, 18), we decided not to optimize expression of the “first-generation” ss1-z CAR. Rather, the second-generation ss1-bbz and CD19-bbz CARs were chosen for further optimization using RNA electroporation because they are being tested in a clinical trial using lentiviral vector technology (Clinicaltrials.gov NCT00891215).

Structural modification of noncoding regions by incorporation of two repeats of 3′ untranslated regions (UTR) from β-globulin and longer poly(A) sequences has been shown to enhance RNA stability, translational efficiency, and the function of RNA-transfected dendritic cells (19). However, these strategies have not been systematically evaluated in RNA-electroporated T cells. To test if this approach applies to human T lymphocytes, we modified our IVT vector (pGEM-ss1bbz.64A) by adding 5′UTR (SP163) or 3′UTR [two repeats of 3′UTR derived from human β-globin (2bgUTR) or a prolonged poly(A) (150A) sequence as shown in Fig. 1A]. The SP163 translational enhancer is derived from the 5′UTR of the vascular endothelial growth factor gene and is reported to increase expression levels 2- to 5-fold compared with promoter alone (20). RNA made from these constructs was electroporated into stimulated T cells. As shown in Fig. 1B, compared with our basic IVT construct containing a 64-poly(A) tract, addition of 3′UTR from β-globulin (2bgUTR) and longer poly(A) (150A) tailing enhanced the transgene expression, especially when combined (2bgUTR.150A). In contrast, incorporation of the SP163 sequence at the 5′ end of ss1-bbz repressed transgene expression, which might be due to reduced capping efficiency when the SP163 sequence was added.

The 5′ cap located at the end of mRNA molecule consists of a guanine nucleotide connected to the mRNA via a 5′ to 5′ triphosphate linkage. Several cap structures have been described, including caps 0 and 1 (21). Several methods have been used to incorporate the 5′ cap structure onto the transgene and poly(A) tail construct. Commercially available systems incorporate cap 0 or 1 using cotranscriptional or enzymatic approaches to produce capped mRNA. This process is important to optimize to enhance translational efficiency and because of the considerable expense of the various capping systems (see Supplementary Materials and Methods). RNA made using the different capping systems was electroporated into stimulated T cells, and the transgene expression was monitored by flow cytometry (Fig. 2A and B). The results showed that the transgene expression of T cells electroporated with RNA capped by anti-reverse cap analogue (ARCA) was 3-fold higher than regular cap (RC) analogue capped RNA at 4 hours. The transgene persistence of ARCA capped RNA was also improved, as at day 5 after electroporation >50% of the T cells still expressed the CAR as shown in Fig. 2B.

We next compared enzymatic addition of caps 0 and 1 to nonenzymatic addition of the ARCA. The potential advantage of using the capping enzyme (CE) system is that this approach includes CE and mScript 2′-O-methyltransferase that work together to produce the cap1 structure, which is very similar to ARCA and provides superior translation efficiency in many in vivo systems. To evaluate the efficiency of cap 0 or 1 RNA encoding ss1-bbz, human T cells were electroporated with RNA made by ARCA, CE, cap1 CEs, or CEs plus additional poly(A). As shown in Fig. 2C, the CAR expression using cap1 RNA electroporation was equivalent to ARCA IVT mRNA. The transgene expression was further enhanced by incorporation of the longer poly(A) tail, consistent with the results in Fig. 1.

One potential functional advantage of optimized IVT RNA is that CAR expression could be sustained, as translation of additional CAR could lead to more persistent expression and overcome downregulation induced by target recognition or homeostatic expansion. Activated T cells were electroporated with various RNA preparations encoding ss1-bbz, and then cocultured with K562-meso or control K562-CD19 target cells for 2 days (Supplementary Fig. S2). T cells electroporated with ARCA and CE1 or CE1+A capped ss1-bbz RNA could still maintain their transgene expression after being stimulated with the K562-meso cell line compared with the same T cells cocultured with control target cells. In contrast, T cells electroporated with ss1-bbz RNA capped by the RC analogue did not have detectable CAR on the surface after cocultured with antigen-bearing target.

Based on the above results and other data (data not shown), we concluded that RNA capped with ARCA or with cap1 and a long poly(A) tail is the best RNA production system among the RNAs tested. For large-scale GMP production of IVT RNA, when the production cost is also considered, cap1 is preferred.

RNAs prepared from both plasmids bearing parental or internal ORF-free CAR sequences were electroporated into T cells, and it was found that the transgene expression from the RNAs with internal ORF-free electroporated T cells was equivalent to the T cells electroporated with RNAs with parental sequences (Supplementary Fig. S3) at 20 hours after electroporation. However, substantial prolongation of CAR expression was observed in activated T cells electroporated with clinical-grade RNA generated from internal ORF-free pD-A.ss1.OF or pD-A.19.OF RNAs using the CE system that incorporated both cap1 and prolonged poly(A) into the IVT RNAs (Fig. 3). Transgene expression of the optimized IVT RNA could be detected as long as 7 days after RNA electroporation for both meso and CD19 RNA CARs as shown in Fig. 3C.

Previous studies have shown that 4-1BB is upregulated on CD8⁺ T cells after T-cell receptor stimulation (22). We incubated bulk T cells electroporated with ss1-bbz or CD19-bbz RNA with target cells expressing either mesothelin or CD19, and found robust upregulation of 4-1BB, particularly on CD8⁺ T cells, which was target specific (Fig. 3A). The T cells expressing RNA CARs also secreted substantial amounts of interleukin-2 (IL-2) and translocated CD107a on target-specific recognition (Fig. 3B and D). Finally, in a flow-based lytic assay, we found that both CD19 (19.OF) and ss1 (ss1.OF) CAR RNA–electroporated T cells could specifically lyse target cells efficiently (Supplementary Fig. S4).

A pilot experiment was first conducted to evaluate the therapeutic potential of T cells expressing optimized RNA CARs in mice bearing large pre-established tumors. Mesothelin-positive tumors were established in NSG mice as previously reported (11). Sixty-six days after tumor inoculation, 10 × 10⁶ to 15 × 10⁶ ss1-bbz RNA CAR–electroporated T cells from a healthy donor were injected intratumorally, twice weekly for 2 weeks. The biweekly administration schedule was based on the in vitro expression data shown in Fig. 3. As seen in Fig. 4A, the tumors regressed in the mice treated with ss1 RNA–electroporated T cells, whereas progressive tumor growth was observed in the control group of mice. At the time the mice were sacrificed on day 98, tumor size was substantially smaller in all of the mice treated with electroporated T cells than that of the mice treated with saline (Supplementary Fig. S5). These results indicate therapeutic potential of multiple injections of RNA CAR T cells; however, they are not as potent in the same tumor model using lentiviral transduced T cells, where two intratumoral injections of T cells were able to cure most mice (11).

We developed the M108-Luc model to test if RNA CAR–electroporated T cells are capable of treating mice bearing large disseminated tumors. M108 parental cells were stably transduced with firefly luciferase to allow for bioluminescence imaging (BLI), and in preliminary experiments, we found that NSG mice develop widely disseminated disease with progressive ascites and that all mice die or become moribund by day 100 (data not shown). NSG mice (n = 18) were injected with M108-Luc, and they were randomized into three i.p. treatment groups. On day 58 day after tumor injection, when all mice had large vascularized tumors with ascites and metastatic nodules lining the peritoneal cavity, ss1-bbz RNA CAR–electroporated T cells from a healthy donor were injected (i.p.) into the mice, twice weekly, for 2 weeks. As a control for CAR specificity, a group of mice was injected with CD19-bbz RNA CAR T cells, and another group was treated with saline. Tumor burden in the ss1-bbz RNA CAR group progressively decreased from baseline on day 53. Furthermore, on day 78 after tumor inoculation, the tumor growth in the ss1-bbz RNA CAR T-cell–treated group was significantly repressed (P < 0.01) compared with both saline or CD19-bbz RNA T-cell–treated groups (Fig. 4B). In a side by side experiment, a mouse treated with ss1-bbz CAR T cells expressed using a lentiviral vector exhibited a more robust treatment effect (Fig. 4C), similar to our previously published data (11). However, the ss1-bbz RNA CAR T-cell–treated group had a survival advantage and a significant slowing of tumor growth between days 72 and 92, at which point all of the control mice died from tumor progression (Fig. 4C).

The above studies indicate that biweekly injections of RNA CAR T cells can control advanced flank and i.p. tumors, and that the inhibition is dependent on the CAR specificity, as T cells expressing the CD19 RNA CAR were not effective. However, the T cells in those experiments were obtained from healthy donors and were allogeneic to the tumor. Because allogeneic antitumor effects were observed with repeated long-term administration of RNA CAR T cells (data not shown), autologous peripheral blood mononuclear cells from the patient from whom the M108 tumor was derived were used. T cells were stimulated and electroporated using GMP grade RNA. Thirty NSG mice were randomized into three i.p. treatment groups, as depicted in the diagram in Supplementary Fig. S6. Mice were inoculated with M108-Luc (i.p.) on day 0 and treated with ss1-bbz or CD19-bbz RNA CAR T cells or with saline control, and tumor burden was monitored by serial BLI and body weight as indicated. Therapy was started on day 56 when the tumor was advanced based on the finding of ascites on physical examination and high BLI signals. Tumor burden was dramatically reduced in the group treated with T cells electroporated with ss1-bbz RNA CAR T cells, whereas the tumor continued to grow in the control mice treated with either CD19-bbz RNA CAR T cells or saline (Fig. 5A and C). Even in this setting, where the T cells are autologous to the tumor, there was still a modest CD19 CAR treatment effect, which may be due to the RNA backbone, as this is unlikely to be related to the CD19 scFv CAR given that there were no B cells in these mice. However, after the first six doses of T cells, imaging revealed a lower mean change tumor bioluminescence in the ss1 CAR mice (39%) compared with both the CD19 CAR (244%) and the saline mice (237%; P < 0.001). The 50% median survival after T-cell injection was significantly (P < 0.05) greater in the ss1 CAR mice (73 days) compared with the CD19 CAR (62 days) and saline mice (36 days; Fig. 5B). After the first six doses were given, the mean change in total body weight was lower in the ss1 RNA CAR mice (1.62 g) compared with both the CD19 CAR (6.21 g) and the saline mice (11.4 g; P < 0.001; Fig. 5C). Although we observed disease stability and even “cures” by imaging in some of the ss1 CAR mice, tumor eventually recurred. Despite giving an additional eight doses of treatment, tumor progression was observed in the ss1 CAR mice. Thus, repeated injections of ss1-bbz RNA CAR T cells can provide a survival benefit for advanced disseminated tumors. The mechanism for tumor recurrence in spite of continued therapy is under investigation.

The goal of these experiments was to determine the therapeutic potential of activated T cells expressing electroporated RNA CARs. The main point of our article is that mRNA CARs provide a platform that is expected to be safer and more economical than retroviral or lentiviral vectors for the evaluation of new targets. In the event of toxicity, injections of RNA CAR T cells can be terminated, and the toxicity would be expected to rapidly abate. However, the RNA CAR T cells have a substantial treatment potential, especially in compartmentalized tumors such as mesothelioma. RNA CAR T cells are expected to complement therapies currently being developed with retroviral and lentiviral CARs.

Our approach was to first optimize RNA expression and then test a multiple dosing strategy in robust tumor models. This is the first report indicating that retargeted T cells can have potent in vivo antitumor effects without the use of an integrating vector system. Using optimized IVT mRNA, we show that RNA CAR T cells have potent antitumor effects on advanced flank and intraperitoneal tumors. Further, as far as we are aware, our studies are the first to show that autologous T cells obtained from a patient with advanced cancer can be engineered and shown to control metastatic tumor in robust preclinical models.

We and others have previously shown that RNA electroporation can modify T-cell function in vitro (23 24 25). Mitchell and colleagues (26) reported that T cells can be functionally modified by RNA transfection of the chemokine receptor CXCR2 to migrate efficiently toward a variety of CXCR2-specific chemokines in vitro and in vivo. Yoon and colleagues (27) recently showed that adoptive transfer of Her2/neu RNA CAR–electroporated T cells in the SKOV3 xenograft model led to a decreased rate of tumor growth compared with transfer of mock-transfected T cells. A recent report showed the feasibility of mRNA transfection of a CD19 chimeric receptor into a natural killer cell line, but without a preclinical model or demonstration of in vivo effect (28). Our studies are the first to report the in vivo regression of large advanced tumors and survival extension using RNA-electroporated T cells.

There are a variety of nonintegrating approaches to engineer T cells (29). A temporary expression approach toward CAR immunotherapy, such as mRNA transfection, runs counter to our previous efforts and to those of most investigators in the field. However, the improving technology for RNA transfection may compliment the use of CARs that are stably expressed by integrating viral vector or transposon systems. By systematic comparison of 3′UTR and 5′UTR, incorporation of longer poly(A) tails, efficient capping of RNA, and removal of internal ORFs we were able to achieve high-level and longer expression of RNA CARs in electroporated T cells.

The prevailing paradigm in the adoptive transfer field is that long-term persistence of the cells within the patient is key to efficacy (30, 31). However, it is being increasingly realized that transferred cells can lose their ability to function within the tumor microenvironment rather quickly (32). Our data suggest that it may be possible to give multiple, more frequent injections of T cells that only temporarily express the transgenes of choice, avoiding the accumulation of CAR T cells that have become tolerized, and therefore achieve antitumor efficacy with an improved safety profile. Alternatively, the improving technology for RNA transfection may complement the use of CARs that are stably expressed by integrating viral vector or transposon systems.

Several adverse events have been observed and others are theoretically possible with CAR T-cell therapy. Two deaths have recently been reported following treatment with retrovirally modified CAR T cells, and the early toxic events have been related to systemic effects from cytokine release (4, 5). As a consequence of these clinical events, recent editorials have discussed the need for safer CARs (6, 7). Other toxicities encountered with stably transduced CAR T cells have been on-target, off-organ effects such as the depletion of normal B cells following CD19 CAR therapy, or the induction of hepatic toxicity following carbonic anhydrase IX therapy (8). Although not tested in this work, it is likely that repeated administration of RNA CARs would be required to elicit this form of toxicity, and that the toxicity would resolve following discontinuation of RNA CAR T-cell infusions. Finally, concerns over the lentiviral or retroviral introduction of CARs into CTLs include the known risk of malignant transformation from insertional mutagenesis (33, 34). As there is no integration into the host cell genome and the CAR expression is self-limited, these concerns are circumvented by mRNA transfection.

The primary potential limitation of CAR therapy is the relatively short persistence of RNA CARs. This can be expected to be exacerbated when the RNA CAR T cells are administered to hosts that have been lymphodepleted, which would be expected to result in the induction of homeostatic proliferation of the CAR T cells and, as a consequence, the accelerated loss of CAR expression at the T-cell surface. Thus, RNA CAR T cells may be more effective when given for compartmentalized tumors such as mesothelioma or central nervous system tumors. Furthermore, more frequent administration of RNA CARs may be required in lymphodepleted hosts.

In addition to providing a form of toxicity management discussed above, there are several potential opportunities for RNA CAR T-cell therapy. First, RNA CARs offer the potential to accelerate the pace of CAR development, by providing a flexible and more rapid path to the clinic, and thereby enabling an efficient iterative approach to optimize CAR design and potency. Based on the data in these studies, we plan to open a phase I trial testing anti-mesothelin RNA CARs. The regulatory approval process is less cumbersome with RNA CARs than with stably expressed CARs that require genomic integration. Clinical-grade mRNA is less costly to produce than integrating retroviral or lentiviral vectors, although more expensive than plasmid DNA that is being used in transfection or transposon-based protocols (2, 35). Second, it may be attractive to combine RNA CAR “knockdown” therapy using potent but potentially toxic CARs for remission induction, with consolidation and maintenance therapy using stably expressed CARs as a strategy to provide CAR cells with a potential for memory.

In summary, multiple injections of RNA-engineered T cells are a novel approach for adoptive cell transfer, providing a cost-efficient and flexible platform for the treatment of cancer diseases. In addition, this approach may increase the therapeutic index of T cells engineered to express powerful activation domains without the associated safety concerns of integrating viral vectors.

No potential conflicts of interest were disclosed.

We thank Drs. Katalin Kariko, Shuguang Jiang, and Michael Kalos (University of Pennsylvania) and Dr. Linhong Li (Maxcyte, Inc.) for their advice, and members of the Human Immunology Core (Abramson Cancer Center) for providing lymphocytes.

Grant Support: NIH grants R01CA120409 (C.H. June) and P01-CA-066726 (C.H. June and S.M. Albelda) and Alliance for Cancer Gene Therapy (C.H. June).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.