Background
In the last 5 years, chimeric antigen receptor (CAR) T cells have emerged from bench to bedside and made headlines in clinical trials at a number of academic institutions [
1‐
4]. CARs are recombinant receptors that specifically target tumor-surface antigens. Once the CARs are transfected into T cells, the cells acquire supraphysiologic properties and act as “living drugs” [
5,
6]. Multiple iterations of CARs have been developed, mainly focusing on intracellular signaling modules, which are deemed crucial for CAR design [
7,
8]. To achieve appropriate costimulatory signals so as to activate effector T cells, improve response, and prolong persistence, many different kinds of costimulatory receptors can be incorporated (e.g., CD28 [
9,
10], 4-1BB [
11,
12], OX40 [
13], ICOS [
14], and CD27 [
15]), alone or in tandem [
16]. However, the effect of non-signaling extracellular modules, such as hinge and TM domains, on the proliferation of the transduced T cells and therapeutic efficacy of CARs remains largely unclear [
17].
A hinge domain is a structure between the targeting moiety and the T cell plasma membrane [
18]; these sequences are generally derived from IgG subclasses (such as IgG1 and IgG4), IgD and CD8 domains, of which IgG1 has been most extensively used [
19‐
21]. Currently, studies of the hinge domain mainly focus on the following four aspects: (1) reducing binding affinity to the Fcγ receptor, thereby eliminating off-target activation [
19,
21]; (2) enhancing the single-chain variable fragment (scFv) flexibility, thereby relieving the spatial constraints between tumor antigens and CARs, in turn promoting synapse formation between the CAR T cells and target cells; for example, to overcome steric hindrance in MUC1-specific CAR, a flexible and elongated hinge of the IgD isotype can be inserted [
20]; (3) reducing the distance between an scFv and the target epitope, for example, anti-CD22 CAR needs a hinge domain to exert optimal cytotoxicity [
22]; and (4) facilitating the detection of CAR expression using anti-Fc reagents. Nevertheless, the influences of the hinge domain on CAR T cell physiology are not well understood.
To better understand the effect of the hinge domain on CAR T cells, we generated two versions of CARs, with or without a hinge domain, targeting CD19, mesothelin, PSCA (prostate stem cell antigen), MUC1, and HER2 (human epidermal growth factor receptor 2), respectively [
23‐
31]. We systematically compared the effect of the hinge domains on the growth kinetics, cytokine production, and cytotoxicity of CAR T cells in vitro and in vivo. We revealed that the incorporation of a hinge into CAR constructs can substantially increase the CAR T cell percentage during the in vitro culture period, enhance the invasiveness of CAR T cells. In addition, we found that anti-CD19 CAR T cells with or without a hinge domain have similar abilities to eliminate leukemia cells, whereas a hinge domain can enhance the in vivo antitumor activity of anti-mesothelin CAR T cells.
Methods
Cells and culture conditions
NALM6-GL (acute lymphoblastic leukemia line, stably transfected with GFP and luciferase) and A549-GL (human lung cancer cell line, stably transfected with GFP and luciferase) cell lines were cultured in RPMI-1640 (Gibco, Life Technologies). HEK293T cells used for lentivirus production were cultured with DMEM (Gibco, Life Technologies). DMEM and RPMI-1640 media were supplemented with 10% heat-inactivated FBS (Gibco, Life Technologies), 10 mM of HEPES, 100 U/ml of penicillin, 100 μg/ml of streptomycin, and 2 mM of l-glutamine (Gibco, Life Technologies).
Human peripheral blood mononuclear cells (PBMCs) from healthy donors were obtained from Guangdong General Hospital, after obtaining informed consent and for research use only. Pan T cells were enriched from the PBMCs using a “Pan T cell isolation kit” (Miltenyi Biotec, Germany). CD8+ T cells were positively isolated from the Pan T cells using “CD8 microbeads” (Miltenyi Biotec, Germany); the unlabeled cells that passed through the column were collected, representing the CD4+ T cells.
Construction of chimeric antigen receptors (CARs)
CD19.28z, Meso.28z, HER2.28z, and PSCA.28z CARs were constructed by linking sequences from a signal peptide derived from GM-CSF (GenBank; AAA58735.1, aa 1–19) to the corresponding antigen-specific single-chain variable fragment (scFv); these CARs do not have a spacer domain. Anti-CD19 scFv derived from FMC63 monoclonal antibody, anti-Mesothelin scFv derived from SS1 monoclonal antibody, anti-HER2 scFv derived from FRP5 monoclonal antibody, and PSCA scFv derived from humanized 1G8 monoclonal antibody. The CARs were codon-optimized and chemically synthesized using Genscript. A spacer containing a hinge domain and a CH3 domain derived from human IgG4 (GenBank; AAC82527.1, aa 98–329) was included in PSCA-H.28z CAR; an additional IgD hinge spacer was included in MUC1-H.28z CAR. The scFv of anti-MUC1 CAR derived from HMFG2 monoclonal antibody. The resulting products were sub-cloned into a Pwpxld-based lentiviral backbone plasmid encoding the transmembrane and intracellular domains of CD28 (aa 153–220) and the intracellular domain of CD3-ζ (aa 52–163). CD19-H.28z CAR and Meso-H.28z CAR were constructed by overlapping PCR, adding the hinge domain and the CH3 of IgG4 to the 3′ end of the scFv.
Lentiviral production and transduction
Pwpxld (encoding the CARs or GFP), Pspax2 (expressing the three required lentiviral proteins), and PMD.2G (encoding the lentiviral envelope protein) were transfected into HEK293T cells with PEI reagent (Life Technologies). The lentiviral supernatants were collected 48 and 72 h after transfection and were passed through a 0.4 μm filter. Human Pan T cells, CD8+ T cells, or CD4 T cells were stimulated with microbeads loaded with anti-human CD3, anti-human CD2, and anti-human CD28 antibodies (Miltenyi Biotec, Germany) in a 3:1 bead:cell ratio for 72 h and cultured in RPMI 1640, supplemented with 10% FBS, 10 mM of HEPES, 100 U/ml of penicillin, 100 μg/ml of streptomycin, 2 mM of l-glutamine, r-human IL-2 (300 IU/ml, PeproTech, CT, USA), and r-human IL-15 (5 ng/ml, PeproTech, CT, USA). On day 3 after cell activation, the activated T cells were transduced with lentiviral supernatants and Polybrene 8 μg/ml (Takara) once; then, the cells were washed with PBS 3 times to completely remove any residual lentiviral supernatants. The cells were then resuspended in complete medium to achieve the expansion. The microbeads were removed on day 5. Fresh medium was added every 2 days to maintain an appropriate cell density ranging from 5 × 105 to 1 × 106 cells/ml.
Flow cytometry
All of the samples were analyzed with an LSR Fortessa or C6 (BD Bioscience), and the data were analyzed using FlowJo software. CAR T cells were detected by GFP, and T cell phenotypes were evaluated via CD3 PE-cy7 (clone OKT3, eBioscience), CD3 BV421 (clone UCHT1, BD), CD8a PE (clone HT8a, eBioscience), CD8a PE-cf594 (clone RPA-T8, BD), and CD4 APC (clone OKT4, eBioscience). All FACS plots representing the CAR T cell phenotypes were gated on CD3 and GFP double positive cells.
In vitro tumor-killing assays and cytokine-release assays
The target leukemia cells, NALM6-GL, were co-cultured with GFP T, 19.28z, or 19-H.28z T cells at the indicated E:T ratios in triplicate in U-bottomed 96-well plates for 18 h; A549-GL cells were co-cultured with GFP T, Meso.28z T, or Meso-H.28z T cells, each well of 96-well plates contained 200 μl supernatants. One hundred microliter supernatants from the wells with E:T ratios of 1:1 were collected and used for detecting the concentrations of IL-2 and IFN-γ using an ELISA kit (eBioscience). The luciferase substrate d-luciferin (potassium salt, 150 μg/ml, Cayman Chemical, USA) was added to a 100 μl/well, and the viability of the target cells was monitored by a microplate reader at a 450-nm excitation wavelength. The background luminescence was negligible (<1% of the signal from the wells with only target cells); therefore, the target cell viability (%) was calculated as (experimental signal/maximal signal) × 100%, and the killing percentage was calculated as 100% − viability percentage.
Transwell cell migration assay
For the 19.28z T and 19-H.28z T cells, Nalm6 cell lysates were used as a chemoattractant in the lower chamber, while for the Meso.28z T and Meso-H.28z T cells, A549 cell lysates were used as a chemoattractant in the lower chamber. The T cells were cultured in an insert coated with Matrigel for 24 h, and the cells that transmigrated to the lower chamber were counted by flow cytometry. The percentage of migration was calculated as follows: (CAR T cells migrating through the Matrigel chamber membrane/total CAR T cells in insert membrane before assay begin) × 100.
In vivo studies
NSI mice (NOD-scid-IL2Rg−/−mice, Guangzhou Institutes of Biomedicine and Health (GIBH)), aged 6–12 weeks were used to construct the xenograft tumor mouse models. All of the animal studies were carried out in accordance with instructional guidelines from the China Council on Animal Care and under protocols approved by the guidelines of the Ethics Committee of Animal Experiments at GIBH. Equivalent numbers of male and female mice were used. The NALM6-GL leukemia lines were intravenously inoculated into 2 × 105 cells, and the A549-GL carcinoma lines were inoculated into 5 × 105 cells subcutaneously (flank). The mice then received an adoptive transfer of CAR T cells intravenously 3–14 days later, as indicated in the individual experiments. To control the differences in transduction efficiency, non-transduced T cells were supplemented to ensure that both the number of CAR+ T cells, and the total number of T cells remained constant across all CAR T cell groups. The leukemia burden was evaluated using a cooled CCD camera system (IVIS 100 Series Imaging System, Xenogen, Alameda, CA, USA). The mice were injected intraperitoneally with d-luciferin firefly potassium salt at 75 mg/kg and then imaged 5 min later with an exposure time of 30 s. Quantification of the total and average emissions was performed using Living Image software (Xenogen).
Statistics
All graphs report the mean ± SEM. All statistical analyses were performed with Prism software version 6.0 (GraphPad). The statistical significance of all data was calculated using an unpaired Student’s t test with the Bonferroni correction for multiple comparisons, where applicable. P < 0.05 was considered significant and is designated with an asterisk in all figures.
Discussion
Despite the remarkable progress in CAR T cell-based immune therapy, several obstacles remain [
37,
38]. For example, the efficiency of CAR T cell expansion requires improvement. Recently, some groups have reported that an optimal CD4/CD8 ratio is important for the in vivo antitumor activity of CAR T cells, and the percentage of CD4+ CAR T cells is positively correlated with patient recovery rates [
39‐
42]. Because CD8+ T cells tend to be preferentially expanded in current T cell in vitro culture systems [
43], a method to promote the expansion of CD4+ T cells is urgently needed. Herein, we found that both IgG4-CH3 and IgD hinges were able to continuously increase the CAR T cell percentages and absolute cell numbers during the in vitro culture period [
20,
44], and even more importantly, the additional CAR T cells were mainly CD4+. However, when we isolated the CD4+ T and CD8+ T cells and cultured them separately in vitro, the increased-growth effect disappeared. The mechanism of the increased growth will be studied by us in the future.
To date, many different versions of anti-CD19 CARs have been used in clinical trials [
38]. The scFvs of these CARs are almost all derived from FMC63 mAb [
45], while various different hinge domains and costimulatory molecules have been used. For example, CTL019, which is the most widely used CAR in CD19+ leukemia and lymphoma treatment, has a CD8α hinge [
46,
47], while CD19RCD28 CAR uses a modified IgG4 hinge and Fc region [
48], and also has a hinge-deleted version [
19]. The functional differences between CD28 and 41-BB costimulatory molecules have already been well characterized [
49]; however, the influence of the hinge domain on anti-CD19 CARs has not been studied. Our results showed that the killing activities of 19.28z T and 19-H.28z T cells were similar whether in vitro or in vivo. There are several possible explanations of this result: the location of the CD19 epitope recognized by the FMC63 mAb is not membrane-proximal, there are no steric inhibitory effects between FMC63 mAb and its epitope, or the density of the CD19 molecule on tumor cells is high [
50]. These factors may also partially explain the greater popularity of CD19 in clinical trials compared with CD20 and CD22.
Mesothelin is a glycosyl–phosphatidyl inositol-linked cell surface glycoprotein, which is highly expressed in mesothelioma, and lung, pancreas, breast, ovarian, and other cancers, and has been used as a tumor antigen of CAR T cells in several trials [
51]. We have shown that Meso-H.28z T cells have a better tumor-eradication capacity than Meso.28z T cells in tumor-bearing mice, suggesting that the mesothelin epitope recognized by the scFv may be membrane-proximal, or that there exist steric inhibitory effects between scFv and its epitope, like in SM3 mAb (a MUC1-specific mAb) [
20]. Thus, a hinge is necessary to reduce the distance or ameliorate the steric inhibitory effects between the scFv and its epitope. Furthermore, a hinge can also improve the flexibility of the scFv, which may also be one of the reasons why Meso-H.28z T cells exhibit better anti-tumor activities than Meso.28z T cells. The importance of Ab flexibility has also been demonstrated in naive B cells which co-express cell surface IgM, which lacks a hinge, and IgD, whose elongated monomeric hinge is the longest of all Ab isotypes [
52]. As a result, IgD can assume a “T shape” in which Fab regions can engage Ag in virtually any orientation, making the scFv omni-directional.
Acknowledgements
Dr. Guohua Huang and Dr. Qiuhua Deng helped him to collect some data and gave several suggestions for the revision. So I added Dr. Huang and Dr. Deng into author list of this manuscript in the revision. However, I think I should thank them in the acknowledgment instead of giving them the authorship after carefully reading the BioMed Central authorship policy.