Preclinical Evaluation of Chimeric Antigen Receptors Targeting CD70-Expressing Cancers

Cell-based immunotherapies have shown promising clinical outcomes in recent years. Adoptive transfer of tumor-infiltrating lymphocytes (TIL) can achieve durable complete regression of metastatic melanoma (1). Alternatively, redirecting autologous T cells with chimeric antigen receptors (CAR) or alpha-beta T cell receptors (TCR) against tumor-associated antigens also mediated long-term durable remissions of late-stage cancers refractory to standard therapies (2–4). The advantage of CARs is that they can target antigens expressed on the cell surface without MHC restriction, therefore making them more widely applicable. However, treating solid cancers using CAR immunotherapies remains challenging due to the paucity of safe, clinically effective tumor-associated cell-surface antigens identified to date (5, 6).

CD70 was initially identified as the ligand for CD27, a costimulatory receptor involved in T-cell proliferation and survival (7, 8). Studies on CD70 expression revealed the same restricted expression profile in both humans and mice (9, 10), with CD70 only found on a small percentage of activated T cells and antigen presenting cells in draining lymph nodes during viral infection (10). Interestingly, a number of human tumors can also express CD70, including solid cancers such as clear cell renal cancer (RCC), glioblastoma, and hematological malignancies (11–14). The mechanism of overexpression of CD70 on tumors remains unknown, but a recent study suggested the dysregulated pVHL/HIF pathway may be involved in RCC (15). Nonetheless, due to its restricted expression pattern on normal tissues and overexpression in cancers, CD70 may be an attractive therapeutic target. Approaches using antibody–drug conjugates directed against CD70 have shown some antitumor activity in vitro, and demonstrated some clinical responses against RCC and non-Hodgkin's lymphoma in a phase I clinical trial (13, 16). In addition, T cells genetically engineered with a CAR consisting of full-length CD27 coupled with the CD3-zeta signaling domain (CD3-zeta) has shown CD70-specific tumor recognition (17). These studies suggested that CD70 could be a potential immunotherapeutic target. However, due to CD70 expression by activated lymphoid tissues, the potential for “on-target, off-tumor” toxicity may hinder potential clinical applications, as has been observed when targeting CD19 on malignant and normal B cells with a CAR (2, 18). Therefore, in this study, we have compared the antitumor reactivities of anti-human CD70 CARs in which different portions of CD27 were fused with various costimulatory signaling domains from 41BB and/or CD28, and CD3-zeta, and established a murine model to test the potential for on-target, off-tumor toxicities. Our study demonstrates that anti-human and anti-murine CD70 CARs were effective in vitro and in vivo, respectively. However, some treatment-related toxicities were also observed, although they appeared to be reversible and tolerable in the murine model. Among the anti-human CD70 CAR constructs, CD27 without its intracellular signaling domain, fused with the costimulatory of 41BB and then CD3-zeta, designated as CD27-41BB-zeta, appeared to be most active in vitro. Therefore, plans are under way to test this anti-CD70 CAR in patients with advanced, refractory tumors expressing CD70.

C57BL/6J and NSG mice (The Jackson Laboratory) were maintained per protocols in the NIH animal facility. Murine tumor lines B16 and B16/mCD70 (retrovirally transduced with murine CD70) were maintained in RPMI 1640 (Life Technologies) with 10% fetal bovine serum (FBS; Life Technologies). All mouse studies were approved by the National Cancer Institute Animal Care and Use Committee.

Tumor lines from RCC patients were established and maintained in DMEM (Life Technologies), including 10% FBS, 10% tryptose phosphate (Sigma), 1× insulin-transferrin-selenium (Life Technologies) and 1 × serum pyruvate (Life Technologies). Melanoma tumor lines were maintained in RPMI 1640 (Life Technologies) with 10% FBS. All cell lines included in the study were generated at Surgery Branch, NCI, and tested and identities confirmed by HLA genotyping. The cell lines were maintained in the cell culture only when they were needed in the experiments and usually kept in culture for approximately a month, and mycoplasma testing were done routinely using mycoplasma detection kit (Lonza). The cell lines were reassessed for HLA and antigen expression by flow cytometry and coculture assays when they were thawed for each experiment.

Monoclonal antibodies (mAb), including FITC-labeled anti-mouse CD3, anti-mouse CD45.1, and anti-human CD8 Abs, PE-labeled anti-mouse and anti-human CD70 Abs, PE-cy7-labeled anti-human CD3 Ab, APC-labeled anti-mouse CD27 Ab, APC-cy7-labeled anti-mouse CD8, and purified anti-mouse CD3 and anti-mouse CD28 Abs were purchased from BD Pharmingen. APC-labeled anti-human CD27 Ab was purchased from eBiosciences.

Genes encoding seven human anti-CD70 CARs, including cDNAs for full-length CD27 (flCD27) or truncated CD27 (trCD27; including extracellular and transmembrane portions of CD27, aa1-211) fused with CD28 and/or 41BB signaling domains and CD3-zeta were constructed and cloned into the pMSGV1 plasmid (Fig. 1A). Retroviral production and transduction were the same as described previously (19). Briefly, 293gp cells were transfected with 9 μg of anti-CD70 CARs and 4.5 μg of plasmid RD114 using Lipofectamine 2000 (Life Technologies; 60 μL). Two days later, the supernatants were harvested and used to transduce anti-CD3 stimulated PBL. PBL from allogeneic donors were stimulated with soluble OKT-3 (50 ng/mL) and rhIL2 (300 IU/mL) for 2 days before transduction was performed. The stimulated cells were added to 24-well plates initially coated with RetroNectin (Takara) and subsequently precoated with retrovirus by spinoculation (2,000× g, 32°C, 2 hours) at 5 × 10⁵/mL. The plates were then centrifuged at 1,000 × g for 10 minutes, and incubated overnight at 37°C in a 5% CO₂ incubator. This procedure was repeated the next day and cells were split as necessary to maintain cell density between 0.5 and 1 × 10⁶ cells/mL. Transduction efficiency was estimated by analyzing human CD27 expression on retrovirally transduced cells and comparing this to mock-transduced T cells. To test their reactivity, retrovirally transduced cells (1 × 10⁵) were cocultured with 5 × 10⁴ human tumor lines with or without CD70 expression at 37°C, 5% CO₂ overnight. The supernatants were harvested and tested for IFNγ secretion by ELISA (Thermo Fisher Scientific).

NSG mice were injected subcutaneously with 1.5 × 10⁵ 2654R human renal cancer cells. Eighteen days after inoculation, when tumors were approximately 5 mm in diameter, mice received 6 × 10⁶ intravenous human T cells retrovirally transduced with anti-CD70 CARs or control T cells, followed by intraperitoneal administration of 200,000 IU of rhIL2 per day for 3 days. Each group included 5 randomly assigned tumor bearing mice, and all tumor measurements were performed by a blinded impartial observer.

cDNA encoding murine CD27 fused with murine CD3-zeta signaling domain (mCD27-zeta) was constructed in the pMSGV1 plasmid. To produce retrovirus, 293gp cells were transfected with 9 μg of pMSGV1-mCD27-zeta and 4.5 μg of plasmid pCL-Eco (Addgene) using Lipofectamine 2000 (Life Technologies; 60 μL). Two days later, the supernatants were harvested and used to transduce activated mouse T cells. Splenocytes from C57BL/6J mice were harvested and CD3⁺ T cells were isolated by negative selection using a mouse pan–T-cell isolation kit II (Miltenyi Biotec). Murine CD3 T cells were then stimulated with plate-bound anti-mouse CD3 (1 μg/mL), and soluble anti-mouse CD28 (1 μg/mL) and rhIL2 (30 IU/mL) for 2 days before transduction was performed. Transduction of stimulated cells was performed similarly as described above by spinoculation. Transduction efficiency was determining by analyzing mouse CD27 expression on retrovirally transduced T cells, compared with cells just stimulated with anti-CD3, anti-CD28 and rhIL2. To assess the reactivity of mCD27-zeta, retrovirally transduced murine T cells (1 × 10⁵) were cocultured with B16/mCD70 or B16 (5 × 10⁴) at 37°C, 5% CO₂ overnight and supernatant was harvested and tested for mouse IFNγ secretion by ELISA (R&D Systems).

C57BL/6J mice were injected subcutaneously (s.c.) with 0.5 × 10⁶ B16 or B16/mCD70 tumors. Ten days after inoculation, when tumors were approximately 5 mm in diameter, mice received up to 10⁷ intravenous (i.v.) murine T cells retrovirally transduced with mCD27-zeta or stimulated, untransduced T cells as a control, followed by intraperitoneal (i.p.) administration of 200,000 IU of rhIL2 per day for 3 days. Splenocytes from pmel-1 TCR transgenic mice, that are reactive to gp100 in both B16 and B16/mCD70 tumors, were activated in vitro in the presence of 1 μmol/L hgp100_25–33 peptide and 30 IU/mL rhIL2 for 7 days. As a positive adoptive cell therapy control, a total of 10⁶ activated pmel T cells were given to mice i.v. along with recombinant vaccinia virus encoding hgp100 (2 × 10⁷ pfu) and the regimen of i.p. rhIL2 as above. Where specified, sublethal total body irradiation (TBI; 500 cGy) was given to mice immediately prior to cell transfer. Mice were given from 10⁴ to 10⁷ retrovirally transduced murine T cells when assessing minimal effective treatment dosage. All treatment groups were randomly assigned, and all tumor measurements were performed by a blinded impartial observer.

T cells from Ly5.1 congenic mice were stimulated with plate-bound anti-murine CD3 and soluble anti-mouse CD28 Abs for 2 days, and retrovirally transduced with mCD27-zeta. Four days after transduction, T cells (10⁷) were injected i.v. to C57BL/6J mice, followed by i.p. administration of 200,000 IU of rhIL2 per day for 3 days. Splenocytes were harvested from day 3 to day 12, and the presence of transferred cells was determined by expression of CD45.1-positive cells. Measurements of weight, absolute blood cell counts, blood chemistry, and the number of splenocytes and serum cytokines were determined in both tumor-bearing and non–tumor-bearing mice.

C57BL/6J mice were injected i.v. with murine T cells retrovirally transduced with mCD27-zeta or untransduced control T cells, followed by i.p. administration of 200,000 IU of rhIL2 per day for 3 days. Thirty-two days after transfer, mice were immunized with vaccinia virus encoding either OVA or gp100. Splenocytes from treated and control animals were harvested 7 days after immunization, and cultured in vitro in the presence of 1 μmol/L OVA_257–264 or gp100_25–33 peptides for 7 days, and then tested for IFNγ production of by coculturing T cells with LPS-stimulated lymphoblasts pulsed with either peptide.

Wilcoxon rank-sum test was used to compare tumor slopes between each treatment groups, and log-rank test was used to analyze survival.

Seven anti-human CD70 CAR retroviral vectors were constructed and transduced into anti-CD3 stimulated normal donor peripheral blood lymphocytes (PBL) to evaluate their expression and antitumor activities. As shown in Fig. 1A, full-length CD27 (flCD27) fused with the CD3 zeta signaling domain (CD3-zeta) with or without CD28 and/or 41BB signaling domains were constructed, designated as flCD27-zeta, flCD27-CD28-zeta, flCD27-41BB-zeta, and flCD27-CD28-41BB-zeta, as several studies demonstrated CD28 and 41BB signaling domains could augment antitumor reactivities and in vivo persistence (20–24). Similarly, three anti-human CD70 CARs using truncated CD27 (trCD27, aa 1–211; i.e., CD27 without its intracellular domain) were also constructed, shown as trCD27-CD28-zeta, trCD27-41BB-zeta, and trCD27-CD28-41BB-zeta. PBL retrovirally transduced with each of the seven vectors showed various degrees of CD27 expression (Fig. 1B). As might be expected, PBL successfully transduced with these CAR were devoid of CD3⁺CD70⁺ cells. flCD27-zeta appeared to be the best with ∼92% of cells expressing CD27 in comparison with ∼70% of CD27 expression on T cells transduced with trCD27-41BB-zeta, flCD27-41BB-zeta, or flCD27-28-41BB-zeta (and 0.26% of stimulated but untransduced PBL). Because mock-transduced or untransduced T cells still expressed certain level of CD27, positivity of CD27 expression in these anti-CD70 CAR-expressed T cells was only counted when expression level was above the level of mock-transduced or untransduced T cells. This may lead to underestimate transduction efficiency of anti-CD70 CARs. Downregulation of CD3 on these transduced T cells was also observed, and consistently seen with high transduction efficiency, suggesting that CD3 internalization may occur with introduced CD3-zeta chain signaling domain. Surprisingly, only 45%, 16%, and 3% were CD27 positive when T cells were transduced with trCD27-28-zeta, trCD27-28-41BB-zeta, and flCD27-28-zeta, respectively. To test antigen-dependent tumor recognition by these anti-human CD70 CARs, a panel of CD70-negative tumor lines (624mel and 938mel), and their stably transduced CD70 expressing counterparts (624/CD70 and 938/CD70), and RCC tumor lines naturally expressing high to low levels of CD70 (RCC HC, RCC RO, RCC DS, and RCC MW, respectively; Supplementary Fig. S1 and ref. 25) were included. All of the CARs, except for flCD27-28-zeta, demonstrated specific anti-CD70 reactivity, as they only recognized CD70-positive tumor lines, but not CD70-negative tumor lines (Fig. 1C). The control vector did not show any reactivity against these tumors. Among the candidate CARs, trCD27-41BB-zeta appeared to possess the highest antitumor reactivity by IFNγ production. In addition to flCD27-28-zeta, lower anti-CD70 reactivity was also observed in T cells transduced with trCD27-28-zeta, suggesting that signaling through CD28 may have adverse effects on these constructs. Based on transduction efficiency and antitumor reactivity of these CARs, we decided to compare the efficacy of flCD27-zeta and trCD27-41BB-zeta in vivo. The CD70-positive human renal tumor line, 2654R, was injected subcutaneously into NSG mice and once palpable, treated with human T cells transduced with either flCD27-zeta or trCD27-41BB-zeta. A curative effect was observed in both groups as shown in Fig. 2A and B. None of the control groups showed any delayed tumor growth. Our results suggest that either CAR could be effective in treating CD70-positive tumors in vivo. However, because normal tissues of NSG mice do not express human CD70, we could not assess the treatment-related toxicities using this model. Therefore, a mouse tumor model with tumors expressing murine CD70 was generated to address these issues.

To generate a mouse model, an anti-murine CD70 CAR was constructed by fusing full-length murine CD27 with murine CD3-signaling domain (mCD27-zeta). Murine T cells expressed high levels of CD27 after retroviral transduction with the anti-murine CD70 CAR compared with mock-transduced or untransduced T cells (Fig. 3A). Meanwhile, a stably transfected cell line, B16/ mCD70, was generated by transducing B16 melanoma (a murine CD70-negative tumor line recognized by pmel T cells; Fig. 3B; ref. 26), with murine CD70. As shown in Fig. 3C, pmel T cells produced mouse IFNγ when cocultured with either parental B16 tumor, or B16/mCD70. However, mCD27-zeta could confer specific, anti-CD70 reactivity only against B16/mCD70. To test the antitumor efficacy of mCD27-zeta in vivo, C57BL/6J mice were implanted with either B16 or B16/mCD70 subcutaneously, and then treated with mouse T cells retrovirally transduced with mCD27-zeta or other controls when tumors became palpable. While pmel T cells (given with concomitant vaccination and IL2) could reduce tumor burdens in irradiated mice carrying either B16 or B16/mCD70, mCD27-zeta-transduced T cells (given only with IL2) delayed tumor growth only in irradiated mice carrying B16/mCD70, in an antigen-specific manner (Fig. 3D). Mice that were treated with untransduced T cells or left untreated did not show any delay of tumor growth. In addition, treatment efficacy was dose dependent, with 1 × 10⁵ the lowest effective T-cell number in irradiated mice (Fig. 3E). In nonirradiated mice, however, treatment effects could only be observed when 1 × 10⁷ cells were transferred. This may be attributable to superior persistence of transferred cells seen in irradiated mice compared with nonirradiated mice (Supplementary Fig. S2A). Furthermore, irradiated B16/mCD70-bearing mice treated with 1 × 10⁵ to 1 × 10⁷ CAR T cells had significantly better survival than mice treated with no cells or 1 × 10⁷ mock-transduced T cells (Fig. 3F). Although curative effects could be achieved in mice that were treated with high cell doses (1 of 5 given either 1 × 10⁶ or 1 × 10⁷ cells), the majority of mice showed tumor growth inhibition rather than durable regressions. To delineate possible mechanisms of immune escape, we analyzed mCD70 expression on tumors progressing in treated mice (Fig. 3G). While tumors from control animals remained mCD70 positive, tumors from mCD27-zeta–treated mice completely lacked mCD70 expression, implicating antigen loss as a cause of immune escape. While efficacy data were interesting, this mouse model was primarily developed to analyze the in vivo toxicity of using a CD27-zeta CAR to target CD70.

To assess treatment-related toxicities, C57BL/6J mice were implanted with either B16 or B16/mCD70, treated with mCD27-zeta–transduced or control T cells, and body weight evaluated as an indicator of systemic cytokine toxicity. Irradiated B16/mCD70 mice receiving mCD27-zeta–transduced T cells showed significant lower body weight than those receiving untransduced T cells, pmel T cells, or left untreated from day 6 to day 10, but recovered 2 weeks after the cell transfer (Fig. 4A). On the other hand, no significant differences were observed in nonirradiated recipients of T-cell transfers. Interestingly, lower body weight was also observed in irradiated, mCD27-zeta–treated mice that were implanted with antigen-negative B16 tumors, which suggested that this transient toxicity results primarily from mCD27-zeta–transduced cells interacting with normal endogenous host cells.

To further define the toxicity of mCD27-zeta CAR cells independent of the effects of tumor growth, we conducted experiments in non–tumor-bearing C57BL/6J mice. Body weight, peripheral white blood cell (WBC) and lymphocyte counts, splenocyte counts, blood chemistries, and serum cytokines were evaluated for 2 weeks after treatment. Similar to tumor-bearing mice, irradiated mice that were treated with mCD27-zeta–transduced T cells showed significant lower body weight at days 6 and 8 than those that were treated with mock-transduced or untransduced T cells, but recovered approximately 2 weeks after cell transfer (Supplementary Fig. S2B). This effect was not observed in nonirradiated mice. As would be predicted, whole body irradiation, with or without CAR T-cell transfer dramatically decreased absolute WBC and lymphocyte counts as well as splenocyte counts compared with all nonirradiated groups (Fig. 4B and Supplementary Fig. S2C). Although blood chemistry did not show any differences among groups, we could detect IFNγ in mouse serum from day 3 to day 5 only in irradiated mice that were treated with mCD27-zeta CAR T cells (Fig. 4C). However, no differences could be detected in the very low levels of CD70 expression on splenocytes among these groups (Fig. 4D). Taken together, our data suggest that transient cytokine toxicity, as indicated by weight loss and elevated serum IFNγ levels, could be detected in mice undergoing whole body irradiation followed by mCD27-zeta CAR T-cell transfer, and this resulted primarily from reactivity against endogenous host cells.

Because successfully targeting CD70 could affect some T cells and APCs, we tested immune responses in treated mice. One month after cell transfer, treated mice were immunized with vaccinia-OVA or vaccinia-gp100. Splenocytes from treated mice were stimulated and tested for reactivity against OVA or gp100 peptides. As shown in Fig. 5A, 7 days after in vitro stimulation with OT-I_257–264, splenocytes from mice treated with mCD27-zeta or mock-transduced cells (with or without recipient preirradiation) were reactive to OT-I_257–264. Similar results were observed with mice immunized with vaccinia-gp100 (Fig. 5B), as T cells from mice immunized with vaccinia-gp100 were reactive to gp100_25–33, regardless which treatment the mice had received. Interestingly, T cells from mice given mCD27-zeta CAR T cells with irradiation appeared to be more reactive than other groups. Nonetheless, our data show that mice treated with anti-murine CD70 CAR T cells could still mount immune responses against exogenous antigens.

We have demonstrated in this study that using CD27, the natural receptor that engages CD70, we could construct CARs to target CD70-expressing tumors. CD70 can be expressed on activated normal lymphocytes, but we did not encounter in vitro “fratricide” by CD27-zeta CAR expressing T cells as an impediment to activating, transducing, and expanding populations of reactive, CAR-transduced T cells. We then compared 7 different anti-human CD70 CARs introduced in a replication-defective gamma-retrovirus for in vitro CD70-specific antitumor reactivity. These 7 CARs could all be categorized as second-generation CARs (flCD27-zeta, trCD27-28-zeta, and trCD27-41BB-zeta) or third-generation CARs (flCD27-28-zeta, flCD27-41BB-zeta, trCD27-28-41BB-zeta, and fCD27-28-41BB-zeta) given the costimulatory nature of the full-length CD27 receptor. Significant differences were observed when different costimulatory signaling domains were included in the constructs. T cells transduced with receptors containing only the 41BB cosignaling domain, i.e., trCD27-41BB-zeta and flCD27-41BB-zeta, showed high transduction efficiency and anti-CD70–specific reactivity, while T cells including only a CD28 cosignaling domain, such as trCD27-28-zeta and flCD27-28-zeta, showed poor receptor expression and lower anti-CD70 reactivity. Interestingly, adding the 41BB signaling domain (constructs trCD27-28-41BB-zeta and flCD27-28-41BB-zeta compared with trCD27-28-zeta and flCD27-28-41BB-zeta, respectively) could partially compensate for the poor performance of CD28 alone in these constructs. Our findings coincide with a recent publication in which the authors demonstrate that 41BB costimulation reduces, but CD28 costimulation induces, exhaustion of CAR-transduced T cells (27). Although the intracellular domain of CD27 may augment antitumor reactivity in vivo (28), including the full-length CD27 receptor in the CAR did not reverse the deleterious effect of the CD28 costimulatory domain alone in our in vitro study. Overall, the second-generation construct, trCD27-41BB-zeta, appeared to be the best among the constructs we tested in vitro. Comparing trCD27-41BB-zeta with flCD27-zeta in a xenograft experiment, both constructs showed similar curative effects against a naturally expressing CD70⁺ tumor. Recent studies using CD19 CAR treating multiple hematologic malignancies have shown dramatic clinical responses (2, 3, 29), and the in vivo expansion of CAR-transduced T cells correlated with clinical responses (29). In fact, multiple preclinical studies have demonstrated that the 41BB costimulatory domain can enhance in vivo persistence and survival of CAR-transduced T cells (21, 24). A recent study further suggests that 41BB promotes the growth of central memory T cells with enhanced fatty acid oxidation and mitochondrial biogenesis (30). Therefore, we have chosen trCD27-41BB-zeta for our future clinical studies.

One of the biggest drawbacks with T cells targeting overexpressing tumor antigens is the “on-target” toxicity against normal tissues (2, 31–34). The severity of these treatment-related toxicities is largely dependent on which normal tissues express the targeted antigen. T cells engineered with an anti-CD19 CAR, for instance, can effectively treat late-stage cancer patients with CD19⁺ B-cell malignancies. However, it can also cause a transient acute cytokine release syndrome as well as long-term eradication of normal CD19⁺ B cells in some patients (2, 35). These effects can be managed medically in nearly all patients and are considered tolerable in view of the efficacy of this cell transfer. Thus, analysis of antigen expression patterns and toxicity in an appropriate all-murine model can be valuable when vetting a potential tumor-associated antigen targeted by adoptive T-cell therapy. Similar to previous studies, we could only detect CD70 expression on a very small subpopulation of human peripheral blood cells, and murine lymphoid tissues, such as splenocytes, lymph nodes, bone marrow, and peripheral blood. An all murine model targeting mCD70 with a CAR using the mCD27 binding domain could elucidate the in vivo consequences of depleting this subpopulation of normal lymphocytes and possibly reveal other unsuspected CD70 expression on normal tissues, but further strict quantitative translation of toxicities may not be possible when homologous receptor components and host species also have to be changed. Therefore, for our in vivo murine toxicity studies, we used a basic CD27-zeta CAR that not only depleted CD70 expressing immune cells but also had demonstrated efficacy against tumor. Using this therapeutically effective anti-murine CD70 CAR, we demonstrated that acute toxicities such as weight loss and low lymphocyte counts occurred in the irradiated treatment group at the highest cell doses but resolved within 2 weeks after cell transfer. In addition, low levels of serum IFNγ could be detected in the irradiated treatment group during the first week after cell transfer. Our data suggest endogenous CD70-expressing cells may induce these short-term, self-limited toxicities. Analyses of blood chemistry and sequential histological studies of multiple tissues also did not show any consistent differences associated with the transfer of these CAR T cells. Moreover, the immune competence of the treated mice appeared to remain intact, because they could respond normally to immunization using vaccinia virus encoding OVA or gp100. This would imply that CD70 expression is not an obligate step in T-cell activation or APC function. Any impact on this small normal population of immune cells appears similar and analogous to that seen when clinically targeting the normal B-cell antigen CD19 with CAR T cells. No unacceptable toxicities were encountered in this preclinical evaluation of an anti-CD70 CAR using CD27 as the binding domain. The potential for additional toxicities with more complex receptor structures always exists, but this will be best addressed by careful phase I dose-escalation trials. Such a clinical trial (NCT02830724) with escalating cell doses of CAR-transduced autologous PBL in patients with late-stage CD70⁺ cancers is under way.

No potential conflicts of interest were disclosed.

Conception and design: Q.J. Wang, J.C. Yang

Development of methodology: Q.J. Wang, Z. Yu, K. Patel, N.P. Restifo, J.C. Yang

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Q.J. Wang, Z. Yu, K.-i. Hanada, K. Patel, D. Kleiner

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Q.J. Wang, K. Patel, D. Kleiner, N.P. Restifo, J.C. Yang

Writing, review, and/or revision of the manuscript: Q.J. Wang, Z. Yu, K. Patel, D. Kleiner, N.P. Restifo, J.C. Yang

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Q.J. Wang, N.P. Restifo, J.C. Yang

Study supervision: Q.J. Wang, J.C. Yang

The authors thank Drs. Steven A Rosenberg and Paul Robbins for thoughtful discussions.

This research was supported by the Intramural Research Program of the NIH, National Cancer Institute, Center for Cancer Research.

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.