Rationally Designed Transgene-Encoded Cell-Surface Polypeptide Tag for Multiplexed Programming of CAR T-cell Synthetic Outputs

The successful application of chimeric antigen receptor (CAR) T cells to treat a broad array of hematologic malignancies represents the leading edge of using cellular genetic engineering in medicine (1–3). Currently, the variability of therapeutic responses and severity of toxicities attributable to CAR T cells are considerable, necessitating efforts to refine this treatment modality. Inconsistencies in therapeutic product composition, such as product-to-product heterogeneity in transgene frequency and expression level, contribute to variable clinical outcomes (4–6). The purification of T-cell products of defined cellular composition as well as uniform and prescribed levels of transgene expression would refine the potency and safety attributes of these products (7). Systems to deliver large and complex genetic payloads also would enable the engineering of cells that recognize multiple targets or carry genetic constructs that augment the therapeutic index of CAR T cells (8). Although currently used lentiviral vectors are highly efficient and safe, they have limited payload capacities (6, 9). Multiplexed engineering, using modules packaged into multiple viral vectors for cotransduction, necessitates a facile means of identifying cells housing the complete combinatorial genetic payload.

To this end, a limited number of selection transgenes with utility in clinical cell product manufacturing have been developed (3, 10). Expressed selection transgenes typically enrich for cells with chromosomally integrated transgenic payloads (3). Historically, selection transgenes encoded bacterial antibiotic phosphotransferases that render gene-modified mammalian cells resistant to antibiotic selection, but these also render cells immunogenic (11, 12). More recently, mutations in human genes have been described that confer resistance to cytotoxic pharmaceutical drugs (13). While effective for selection, the use of these transgenes in a multiplexed manner is limited as they require significant payload allocation in vectors and create complexity by necessitating compatible selection drug cocktails.

Transmembrane proteins can serve as selection markers for immunomagnetic or FACS-based purification of homogenous cellular products. Our group has deployed a surface tag, EGFRt, that is human, minimally immunogenic, orthogonal to T cells, small in size, and paired with pharmaceutical-grade selection antibodies (14, 15). This truncated form of human EGFR serves as an in vitro selection marker, allows for in vivo tracking of the therapeutic products by flow cytometry, and can act as a suicide trigger if activated via infused cetuximab (Erbitux). This construct is currently being used in numerous CAR T-cell clinical trials, with excellent performance attributes reported to date (ClinicalTrials.gov, NCT01683279, 01815749, 02051257, 02146924). It also is used in lisocabtagene maraleucel, the CD19 CAR T-cell product under FDA consideration for commercial approval (16).

We posited that a second cell-surface barcode tag to complement EGFRt would support a two-vector cotransduction platform and enhance the therapeutic acuity of multiplexed cellular products. As a candidate marker, we chose HER2 because it is not natively expressed on T cells, it is human in origin, and its interaction with the commercially available pharmaceutical-grade mAb trastuzumab is defined (17). We truncated HER2 to lack a cytoplasmic tail but retain a minimal extracellular conformational epitope that is recognized by trastuzumab; this form of HER2 we designated HER2t. We further modified this truncated HER2 to enhance trastuzumab epitope accessibility, thereby generating HER2t variants with tunable trastuzumab labeling avidities. We combined HER2t- with EGFRt-containing lentiviral CAR vectors and demonstrated that HER2t-CAR-A x EGFRt-CAR-B–redirected T cells can be manufactured, purified, and tracked in vivo. This combinatorial system allows for combining other transgene-encoded devices with CARs for improved potency and safety in next-generation clinical products.

All cell lines were maintained in RPMI1640 (Irvine Scientific; catalog no. 21875034) supplemented with 2 mmol/L l-glutamine (Cellgro; catalog no. 25-005-CI), 25 mmol/L HEPES (Irvine Scientific; catalog no. 9319), and 10% heat-inactivated FBS (Hyclone; catalog no. SH30070.03), unless otherwise noted.

Transduced cell lines were created by plating 1 × 10⁶ cells in 0.5 mL RPMI1640 media containing protamine sulfate (1:100 dilution, APP Pharmaceuticals; catalog no. 22905). Plated cells were transduced at a multiplicity of infection (MOI) of 1. The Raji eGFP:ffluc cell line [transduced with a lentiviral construct expressing eGFP fused to ffluc (18)] was kindly provided by Stanley Riddell [Fred Hutchinson Cancer Research Center (FHCRC, Seattle, WA)] in 2012. The Raji CD19KO cell line was created by transfection of pX330 plasmid (Addgene; catalog no. 42230) containing CD19 CRISPR guide sequence using an Amaxa Nucleofection Kit (Lonza, catalog no. VCA-1003; Program M-013) followed by clonal expansion of the CD19-negative Raji population.

K562 target cell lines were kindly provided by Stanley Riddell (FHCRC) in 2005. K562 cells were transduced with lentiviral constructs encoding the full-length target antigen CD19 (NCBI Entrez Gene: 930) or CD20 (NCBI Entrez Gene: 931) and selected using CD19 or CD20 microbeads, respectively (Miltenyi Biotec; CD19: catalog no. 130-050-301; CD20—catalog no. 130-091-104), according to the manufacturer's instructions. HER2t-expressing K562 were similarly transduced and selected using biotin-conjugated trastuzumab and anti-biotin microbeads (Miltenyi Biotec, catalog no. 130-090-485).

H9 T-lymphoblast (H9 T cells; obtained in 2007), Raji (obtained in 2005), and 293T (obtained in 2006) cell lines were from the ATCC. 293T cell lines were maintained in DMEM (Thermo Fisher Scientific; catalog no. 11960044), supplemented with 2 mmol/L l-glutamine, 25 mmol/L HEPES, and 10% heat-inactivated FBS. H9 T-cell lines expressing the HER2t variants or T2A-HER2t or T2A-EGFRt appended CD20CAR, EGFR806CAR, or hB7H3CAR were created by lentiviral transduction as described above.

Epstein–Barr virus–transformed lymphoblastoid cell lines [TM-LCL, kindly provided by Stanley Riddell (FHCRC) in 2010] were made from peripheral blood mononuclear cells (PBMC) as described previously (19). NS0-IL15 (obtained in 2005) cells were generated as described previously (20).

All cell lines were authenticated by short tandem repeat profiling matched to the DSMZ (Deutsche Sammlung von Mikroorganism un Zellkulturen, Braunschweig, Germany) Database (University of Arizona Genetics Core, Tucson, AZ). Cell supernatants were sent for third party Mycoplasma testing at the FHCRC Research Cell Bank (RCB). Samples were processed and evaluated using the MycoProbe Kit from R&D Systems (catalog no. CUL001B) following manufacturer's instructions and protocols. Samples were prepared for Mycoplasma assays monthly. Tumor cells were used for assays anywhere from passage 1 to 20.

Fluorochrome-conjugated antibodies specific for CD4 (SK3; catalog no. 344604), CD8 (RPA-T8; catalog no. 301040), CD45 (HI30; catalog no. 304028), CD45RO (UCHL1; catalog no. 983102), and CD62 L (DREG-56; catalog no. 304806), and fluorochrome-conjugated streptavidin (catalog no. 405204) were obtained from BioLegend. Antibody isotype controls (catalog nos. 554121 and 553454) were obtained from BD Biosciences. Cetuximab (Erbitux, Eli Lilly) and trastuzumab (Herceptin, Genentech) were purchased from the Seattle Children's Hospital pharmacy. Cetuximab and trastuzumab were biotinylated using an EZ-Link Sulfo-NHS-biotinylation kit (Thermo Fisher Scientific; catalog no. 21425) according to manufacturer's instructions or directly conjugated to antigen-presenting cell (APC) by BD Biosciences. Flow cytometry was performed on a FortessaLSR (BD Biosciences), and the percentage of cells in a region of analysis was calculated using FlowJo, Version 10.0.7. Histogram percentage results were determined by histogram subtraction on FlowJo. Cells were washed twice in PBS (Gibco, catalog no. 10010-023) prior to staining for 30 minutes at room temperature with fluorochrome-conjugated antibodies at predetermined optimum concentrations, washed twice with PBS post-stain, and then resuspended in fixation buffer (BioLegend, catalog no. 420801) prior to analysis on the FortessaLSR.

The third-generation self-inactivating (SIN) 41BB-CD3ζ CD19CAR-T2A-EGFRt_epHIV7 (CD19CAR-EGFRt) lentiviral construct was described previously (21). The CD20CAR-T2A-EGFRt_epHIV7 (CD20CAR-EGFRt) contains a Leu16 (murine anti-human CD20) single-chain fragment variable (scFv; ref. 22) fused to the human IgG4Hinge-CH3 (119aa) spacer domain portion of IgG4 (23), the same transmembrane and signaling components of the CD19CAR (CD28 transmembrane-41BB-CD3ζ), appended to T2A-EGFRt.

HER2t was synthesized by PCR splice overlap extension (HER2t(CHP)GMCSFFwd: AATAGCTAGCGCCGCCACCATGCTTCTCCTGGTGACAAGCCTTCTGCTCTGTGAGTTACCACACCCAGCATTCCTCCTGATCCCATGCCACCCTGAGTGTCAG and HER2t Rvs: AATAGCGGCCGCTCAGATGAGGATCCCAAAGACCAC) using pDONR223-ERBB2 (Addgene; catalog no. 23888) as template and the epHIV7 transfer plasmid as recipient. The final HER2t cDNA consists of the human GMCSF leader peptide (GMCSFRss) fused in frame to domain IV (aa. 563–652) and the transmembrane spanning components of HER2 (aa. 653–675). Models of HER2t were rendered using UCSF Chimera (PDB 1N8Z). We replaced EGFRt in the CD19CAR-T2A-EGFRt_epHIV7 construct with HER2t by PCR splice overlap extension and Gibson cloning (NEB; catalog no. E5510S; ref. 24). Specifically, PCR (NEB; catalog no. M0492S) was used to amplify two DNA fragments containing a CAR-T2A (epHIV7 Fwd: CAGATCCAAGCTGTGACCG and T2A Rvs: CCTAGGGCCGGGATTCT) and T2A-HER2t(Her2t(CHP)GMCSFFwd: AATAGCTAGCGCCGCCACCATGCTTCTCCTGGTGACAAGCCTTCTGCTCTGTGAGTTACCACACCCAGCATTCCTCCTGATCCCATGCCACCCTGAGTGTCAG and epHIV7 Rev: CGATAAGCTTGATATCGAATTCCTGC) and Gibson assembled into the epHIV7 backbone, creating CD19CAR-T2A-HER2t_epHIV7 (CD19CAR-HER2t). The T2A-HER2t and T2A-EGFRt appended CD20CAR, EGFR806CAR, and hB7H3CAR constructs were created by swapping the CD19scFv of the CD19CAR-T2A-HER2t_epHIV7 or CD19CAR-T2A-EGFRt_epHIV7 with the respective CD20, EGFR806, or hB7H3 scFv.

HER2t variants were synthesized as GeneArt Strings (Thermo Fisher Scientific) and directly cloned into the multiple cloning site of epHIV7. The following DNA sequences were inserted between the extracellular and transmembrane domains of HER2t: IgG4hinge (5′-gaatctaagtacggaccgccctgccccccttgccct-3′), CD28hinge (5′-tgtccaagtcccctatttcccggaccttctaagccc-3′), or G₃SG₃ (5′-ggtggaggcagcggaggtggc-3′).

The same cloning strategy as CD19CAR-T2A-HER2t_epHIV7 was used to create the CD19CAR-T2A-HER2tG_epHIV7 (CD19CAR-HER2tG) and chimeric cytokine receptor (CCR; CD122)-T2A-HER2tG_epHIV7 [CCR(CD122)-HER2tG; ref. 25]. Specifically for CCR(CD122)-T2A-HER2tG_epHIV7, the insert CCR2-T2A (epHIV7 pre NheI: GTGAACGTTCTTTTTCGCAACGGGTTTGCC and HER2t HpaI Rvs; GCTGCCTCCACCCGTTAA CGGGCTGGCTCTCTGC) and backbone HER2tG_epHIV7 were amplified (Mid HER2t Fwd: GACCTGGATGACAAGGGCTG and PreWPRE_R: CGATAAGCTTGATATCGAATTCCT) using a CCR(CD122)- and HER2tG epHIV7-containing plasmids as templates, respectively, and Gibson assembled as mentioned above. EGFRt and CCR(CD122) were synthesized as described previously (25, 26).

Construct-encoding lentiviruses were produced in 293T cells using the packaging vectors pCHGP-2, pCMV-Rev2, and pCMV-G as described previously (21). Briefly, 293T cells were transfected with packaging vectors and a transfer plasmid by complexing the plasmid DNAs with Lipofectamine 2000 (Life Technologies; catalog no. 11668-500). The following day, media was exchanged and 72 hours after the initial transfection, lentivirus was harvested and concentrated.

All vector maps are available upon request.

CD4⁺ and CD8⁺ bulk T cells were isolated from human PBMCs derived from blood discard kits of healthy donors (Puget Sound Blood Center) over Ficoll-Paque (Pharmacia Biotech, catalog no. 17144003). PBMCs from each donor were first positively selected for CD8⁺ T cells using CD8 microbeads (Miltenyi Biotec, catalog no. 130-045-201) and LS columns. The flow-through from the LS columns was subjected to a depletion step using LD columns to remove residual CD8 microbeads, then positively selected for CD4⁺ T cells using CD4 microbeads (Miltenyi Biotec, catalog no. 130-045-101) per the manufacturer's protocol.

Isolated cells were stimulated in RPMI1640 supplemented with 2 mmol/L l-glutamine, 25 mmol/L HEPES, 10% heat-inactivated FBS, 50 U/mL IL2 (Chiron Corporation; catalog no. 53905-991-01) for CD8⁺ T cells, 5 ng/mL IL7 (Miltenyi Biotec; catalog no. 130-095-363) for CD4⁺ T cells, 1 ng/mL IL15 (Miltenyi Biotec; catalog no. 130-095-765) for both CD4⁺ and CD8⁺ T cells, and anti-CD3/CD28 beads (Life Technologies; catalog no. 11132D). Primary CD4⁺ and CD8⁺ T cells were supplemented with the appropriate cytokines every other day. Primary T cells were transduced with CD19CAR-T2A-HER2t_epHIV7, CD19CAR-T2A-HER2tG_epHIV7, CD19CAR-T2A-EGFRt_epHIV7, CD20CAR-T2A-EGFRt_epHIV7 or a combination of these lentiviral vectors on day 3 after activation with anti-CD3/CD28 beads using protamine sulfate (1:100 dilution) and a MOI of 1 followed by centrifugation at 800 × g for 30 minutes at 32°C. The anti-CD3/CD28 beads were removed from culture 7 days after activation. CD19CAR-HER2t⁺, CD19CAR-HER2tG⁺, CD19CAR-EGFRt⁺, CD20CAR-EGFRt⁺, or dual CAR–positive cells in the transduction culture were enriched by immunomagnetic selection with biotin-conjugated trastuzumab (HER2t) or cetuximab (EGFRt) and anti-biotin microbeads (Miltenyi Biotech; catalog no. 130-090-485) as per the manufacturer's instructions. Selected T cells were expanded for 12 to 18 days after transduction by stimulation with irradiated (8,000 rad) TM-LCL at a T-cell: TM-LCL ratio of 1:7 in the presence of 50 U/mL IL2 (CD8), 5 ng/mL IL7 (CD4), and 1 ng/mL IL15. HER2tG⁺/EGFRt⁺ H9 T cells were transduced and sorted as described above.

This study was conducted in accordance with applicable Institutional Review Board requirements (protocols to acquire human cells approved by the Institutional Review Board of Seattle Children's Hospital, in accordance with the Declaration of Helsinki). All donors provided written informed consent in accordance with local regulatory review.

Four-hour chromium release assays were performed as described previously (9). Briefly, target K562 and Raji cells were labeled overnight with ⁵¹Cr (PerkinElmer; catalog no. NEZ030010MC), washed three times in PBS, and incubated in RPMI1640 without cytokine in triplicate at 5 × 10³ cells/well with T cells at various effector-to-target (E:T) ratios. Supernatants were harvested for γ-counting using a TopCount Scintillation Counter (PerkinElmer) and specific lysis was calculated as described previously (9).

For cytokine secretion, 5 × 10⁵ T cells were plated in RPMI1640 without cytokine in triplicate with target cells at an E:T ratio of 2:1 in a 96-well plate for 24 hours and supernatants were analyzed by cytometric bead array using a Bio-Plex Human Cytokine Panel (Bio-Rad; catalog no. 171-304070M) according to the manufacturer's instructions.

CCR-mediated proliferation was assayed using mock (CAR negative), CD19CAR-EGFRt_epHIV7 transduced and purified, or CD19CAR-EGFRt_epHIV7 transduced and purified/CCR(CD122)-HER2tG_epHIV7 cotransduced CD8⁺ bulk T cells. T cells were transduced and stimulated with anti-CD3/CD28 beads as described above. On day 15 after anti-CD3/CD28 stimulation, T cells washed of exogenous cytokine(s) were plated in a 48-well plate at 10⁶ cells/well. Mock (CAR negative), CD19CAR-EGFRt_epHIV7 transduced and purified, or CD19CAR-EGFRt_epHIV7 transduced and purified/CCR(CD122)-HER2tG_epHIV7 cotransduced CD8⁺ bulk T cells were cultured in the presence or absence of 50 U/mL IL2 and 0.5 ng/mL IL15. Viability and proliferation were measured every 2 to 3 days until the end of the study using a custom T-cell assay on the NucleoCounter NC-3000 (Chemometec) according to the manufacturer's instructions.

All mouse experiments were approved by the Seattle Children's Research Institute (SCRI, Seattle, WA) Animal Care and Use Committee (protocol number 00104). NOD.Scid.IL2RγCnull (NSG) mice were obtained from The Jackson Laboratory or bred in-house. Mice used in experiments were female, 6 to 10 weeks of age, and an average of 20.7 ± 1.2 g at the start of the study.

Six- to 10-week-old NSG mice were injected subcutaneously with 5 × 10⁶ live NS0-IL15 cells on day 0 to provide a systemic supply of human IL15 in vivo. Later the same day, mice were intravenously injected with 1 × 10⁷ of either HER2tG/EGFRt negative (mock) or HER2tG- and/or EGFRt-selected T cells. Bone marrow was harvested from sacrificed animals 14 days later, and cell suspensions were analyzed by flow cytometry using live/dead stain (Thermo Fisher Scientific; catalog no. L23105); anti-CD45, anti-CD4, and anti-CD8; biotinylated trastuzumab; and APC-conjugated cetuximab (see “Antibodies and flow cytometry” for details).

eGFP:ffluc expressing Raji or eGFP:ffluc CD19 CRISPR Raji target cells (0.5 × 10⁶) were injected intravenously into 6- to 10-week-old mice. Two and 8 days later, 1 × 10⁷ CAR-HER2tG⁺, CAR-EGFRt⁺, CD20CAR-EGFRt or dual CAR–expressing T cells (1:1 ratio CD4:CD8) were injected intravenously, as described previously (27). 25 mice in total were used for this study, with 5 mice per group, which has been tested previously in our lab and shown to demonstrate significant differences in flux and survival among treatment groups at end of study. Mice were distributed between groups so that the average and spread of tumor flux from the mice was similar between groups. Mice with the highest and lowest flux value were excluded from the study.

Bioluminescent imaging was performed weekly by intraperitoneal injection of 4.29 mg/mouse d-luciferin (Xenogen; catalog no. XR-1001) to mice anesthetized with isoflurane and imaged 10 minutes post d-luciferin injection using the IVIS Spectrum Imaging System (PerkinElmer). Luciferase activity was analyzed using Living Image Software Version 4.3 (PerkinElmer) and photon flux was analyzed within regions of interest.

Statistical analyses were conducted using Prism Software (GraphPad). Data are presented as mean ± SD or SEM as stated in the figure legends. Student t test was conducted as a two-sided paired test with a confidence interval of 95% and results with a P value less than 0.05 were considered significant.

Predicted HLA targets were assessed by using the MHC-I antigenic peptide processing prediction algorithm NetCTL, which can be found here: http://www.cbs.dtu.dk/services/NetCTL/. Potential MHC ligands were found within the endogenous sequence of HER2 and did not overlap with the introduced G₃SG₃linker of HER2tG (Supplementary Table S1).

Cell lysis was carried out in RIPA buffer containing protease inhibitor cocktail (Cell Signaling Technology; catalog no. 9806S). Cell lysates were analyzed by BCA assay (Pierce; catalog no. 23225) and equally loaded onto gels (50 μg protein per lane). Western blots were probed with primary mouse anti-human CD247 (CD3ζ; BD Pharmingen; 1:500 dilution, catalog no. 51-6527GR). Secondary IRDye 800CW goat anti-mouse (LI-COR; 1:1,000 dilution, catalog no. 26-32210) was added as per the manufacturer's instructions. Blots were imaged on the Odyssey Infrared Imaging System (LI-COR). Ratios were calculated by determining the band intensity of the CAR CD3ζ protein and dividing it by the intensity of the endogenous CD3ζ protein.

Genomic DNA was isolated from 5 × 10⁶ H9 or primary T cells using PureLink Genomic DNA Mini Kit (Invitrogen; catalog no. K182001). DNA concentrations determined using a Nanodrop (Thermo Fisher Scientific). Genomic DNA concentrations were adjusted to 12.5 ng/μL. For each 20 μL droplet digital PCR (ddPCR) reaction, 50 ng of genomic DNA was used. ddPCR reactions, targeting woodchuck hepatitis virus posttranscriptional response element (WPRE; target) and albumin (reference), were setup using ddPCR Supermix for Probes (No dUTP; Bio-Rad; catalog no. 1863024) as per manufacturer's recommendations. WPRE-F ATACGCTGCTTTAATGCCTTTG, WPRE-R GGGCCACAACTCCTCATAAA, WPRE-Probe: 5′6FAM/TCATGCTAT/ZEN/TGCTTCCCGTATGGCT/3′IABkFQ/. Albumin ALB-F: GCTGTCATCTCTTGTGGGCTGT, ALB-R: ACTCATGGGAGCTGCTGGTTC ALB probe:/5′HEX/CCTGTCATG/ZEN/CCCACACAAATCTCTCC/3′IABkFQ/. Briefly droplets were generated using the autoDG (Bio-Rad) and thermocycled on a C1000 Touch (Bio-Rad). Cycling parameters are as follows: activation 95°C for 10 minutes, denaturation 94°C for 30 seconds and annealing/extension 60°C for 1 minute (40 cycles), enzyme deactivation 98°C for 10 minutes. Droplets were read using the QX200 Droplet Reader. All genomic DNA samples were measured in triplicate. Droplet counts were visualized, and thresholds were manually set above the negative droplet population using Quantasoft 1.7.4.

HER2t was designed to lack the HER2 carboxyl cytoplasmic tail domain and the membrane distal ligand-binding domain yet retain a conformationally intact minimal binding epitope for trastuzumab, the HER2 transmembrane domain, and a signal sequence to facilitate surface expression (Fig. 1A and B). The resultant HER2t_epHIV7 was packaged into a third-generation SIN lentiviral vector (21) and transduced into HER2-negative K562 erythroleukemia cells at a low MOI, yielding a subpopulation (13.8%) of HER2t⁺ cells (Fig. 1C). The transduced K562 population was subjected to immunomagnetic purification using biotinylated trastuzumab in combination with anti-biotin microbeads (26), which consistently enriched HER2t⁺ cells to ≥95% (Fig. 1C). Titration experiments revealed that as little as 1.2 ng biotinylated trastuzumab was sufficient to maximally label 1 × 10⁶ HER2t⁺ cells (Fig. 1D). Collectively, these data show that HER2t is expressed on the cell surface and can be bound by trastuzumab.

To evaluate the performance of HER2t immunomagnetic positive selection, we transduced T cells with lentiviral vectors housing the coding sequence of a CAR and HER2t separated by a T2A ribosome skip polypeptide sequence (Fig. 2A; ref. 25). We also assessed the performance of HER2t as an immunomagnetic selection marker relative to our previously described genetic tag EGFRt. Human CD4⁺ or CD8⁺ T cells (Supplementary Fig. S1A) were stimulated with anti-CD3/CD28 beads, transduced with either a CD19CAR-T2A-HER2t_epHIV7 or a CD19CAR-T2A-EGFRt_epHIV7 lentiviral vector, and expanded for 12 to 18 days after transduction. On day 14 of expansion, T cells were selected for HER2t or EGFRt using biotinylated trastuzumab and cetuximab, respectively, and anti-biotin microbeads (CD8, Fig. 2B; CD4, Supplementary Fig. S1B). Transduced CD4⁺ and CD8⁺ T cells were consistently enriched to uniform purity (from 22%–92% CD19CAR-HER2t⁺ or CD19CAR-EGFRt⁺ prior to immunomagnetic selection to ≥90% after selection; CD8, Fig. 2B; CD4, Supplementary Fig. S1B).

We found that HER2t-selected T cells expressed higher levels of CAR protein than their EGFRt-selected counterparts, as assessed by Western blot analysis of whole cell lysates using an anti-CD3ζ (1.8-fold more for CD8⁺ T cells and 3-fold more for CD4⁺ T cells, Fig. 2C and Supplementary Fig. S1C, respectively; full Western blots, Supplementary Fig. S2A and S2B). Similar trends in CD20CAR, EGFR806CAR, and hB7H3CAR protein expression were found between HER2t- and EGFRt-selected CAR T cells, generated using H9 T cell lines (Supplementary Fig. S3A and S3B; flow of HER2t- or EGFRt-selected cells; Supplementary Fig. S3C). We also used ddPCR of WPRE, a component of all the lentiviral vectors, to quantify vector copies in genomic DNA isolated from primary CD4⁺ and CD8⁺ CD19CAR HER2t- and EGFRt-selected T cells, and HER2t- or EGFRt-selected H9 T cells expressing the CD20CAR, EGFR806CAR, or hB7H3CAR. We found that although the HER2t-selected CD4⁺ and CD8⁺ CD19CAR T cells incorporated greater numbers of vector copies per cell than their CD19CAR-EGFRt counterparts (Supplementary Fig. S4A), there was not a defined trend in vector copy-number difference between HER2t- and EGFRt-selected CD20CAR, EGFR806, or hB7H3CAR H9 T cells (Supplementary Fig. S4B). Our protein expression data suggest that positively selected HER2t-CAR cells require more HER2t surface density to achieve threshold loading for selection compared with EGFRt-CAR cells. We hypothesized that this feature could be exploited to balance CAR expression levels that achieve desired functional activation thresholds required by effector T cells.

We next sought to study the influence on T-cell function of the selection tags HER2t and EGFRt, when linked through ribosome skip peptides to a CAR, following immunomagnetic positive selection. Cytotoxicity analyses by 4-hour chromium release showed that each CAR-redirected T-cell subset (HER2t- or EGFRt-selected) conferred similar levels of specific lysis against CD19⁺ Raji cells and K562 tumor cells transduced to express CD19 (CD8, Fig. 3A; CD4, Supplementary Fig. S5A and S5B). Cytokine production by CAR⁺ T cells in response to stimulation with Raji and K562/CD19⁺ cells showed similar selective stimulation by CD19⁺ cells (CD8, Fig. 3B; CD4, Supplementary Fig. S5C).

T-cell activation for cytokine secretion, integrating costimulation and requiring gene transcription, is a more complex signaling event than triggering for degranulation (28). The production of certain cytokines, in particular IL2, is therefore considered to more accurately reflect the quality of CAR signaling in T cells than degranulation. Relative to EGFRt-selected CD19CAR T cells, HER2t-selected CD19CAR CD8⁺ T cells cocultured with Raji cells produced more than 2-fold higher IL2 (P < 0.007), IFNγ (P < 0.0023), and TNFα (P < 0.05; Fig. 3C). Similar trends were seen for CD4⁺ T cells when co-cultured with Raji cells (Supplementary Fig. S5D). These data show that the selection method can influence CAR expression levels and can modulate the functional activation of T cells upon antigen stimulation in vitro.

We next tested HER2t as a cell-tracking cell-surface tag to identify infused CAR T cells in blood following dosing in an NSG mouse model (26). Initial tests revealed that HER2t afforded low resolution between HER2t⁺ T cells and their HER2t⁻ counterparts (Supplementary Fig. S6A and S6B), which prompted us to engineer the extracellular domain of HER2t to improve trastuzumab binding. We reasoned that engineered derivatives of HER2t in which the trastuzumab epitope is relocated away from the plasma membrane could increase epitope accessibility and enhance trastuzumab binding (29). To this end, we incorporated a small library of short spacer regions of synthetic or human origin [IgG4hinge (19), CD28hinge (28), and G₃SG₃] between the trastuzumab binding epitope sequence and the transmembrane amino acid domain of HER2t (Fig. 4A). The panel of HER2t spacer variants was transduced into H9 T cells that then were subjected to trastuzumab-biotin/anti-biotin microbead positive immunomagnetic selection. Of the spacers tested, the G₃SG₃ linker HER2t (designated HER2tG) afforded the highest gain, and broadest breadth, of trastuzumab binding as determined by flow cytometry and analysis of median fluorescence intensity (MFI; fold increase in MFI of HER2tG relative to HER2t: 6.3, IgG4hinge: 1.04, CD28hinge: 21.3; Fig. 4B). The MFI for the selected H9 T-cell populations tracked with higher vector copy number per cell as assessed by ddPCR (Fig. 4B). The G₃SG₃ linker of HER2tG did not introduce any predicted HLA targets (Supplementary Table S1).

The variant HER2tG was appended to the second-generation 41BB-ζ CD19CAR in epHIV7 (15) and subjected to lentiviral vector production. CD4⁺ or CD8⁺ T cells were then transduced with viral vectors containing CD19CAR-T2A-HER2tG_epHIV7 or CD19CAR-T2A-EGFRt_epHIV7 (55%–78% HER2tG⁺ or EGFRt⁺) and immunomagnetically selected according to the respective marker, resulting in uniform purity (≥90%) after selection (CD8, Fig. 4C; CD4, Supplementary Fig. S6C). Although HER2tG-selected cells showed a higher MFI than HER2t-selected cells in H9 T cells, CD19CAR-HER2tG–selected CD8⁺ T cells induced similar levels of cytolytic activity, and IFNy and TNFα production, relative to CD19CAR-HER2t–selected CD8⁺ T cells (Supplementary Fig. S7).

When appended to a CAR using ribosomal skip sequences, HER2tG-selected T cells, unlike with HER2t-selected T cells, expressed similar CAR protein levels as EGFRt-selected T cells (70% relative CAR zeta chain intensity relative to EGFRt-selected T cells) as assessed by CD3z-specific Western blot analysis (CD8, Fig. 4D, full Western blot analysis Supplementary Fig. S8A; CD4, Supplementary Fig. S8B). These data suggest that HER2t can be engineered to yield prescribed expression levels of ribosome skip peptide-linked therapeutic transgene payloads when subjected to immunomagnetic purification, showcasing the utility of HER2t and its spacer variants for calibrating transgene expression upon immunomagnetic selection.

We next asked whether the HER2tG and EGFRt selection markers could be deployed in separate lentiviral vectors, each housing a different CAR, to generate T cells expressing two distinct CARs and exhibiting bispecific tumor targeting function. Accordingly, we assessed whether the cotransduction of T cells with lentiviruses housing CD19CAR-T2A-HER2tG_epHIV7 and CD20CAR-T2A-EGFRt_epHIV7 (Fig. 2A) could efficiently yield dual-marked EGFRt⁺HER2tG⁺ T cells. Dual vector transduction generated a heterogeneous product consisting of a mixture of EGFRt⁻HER2tG⁻, EGFRt⁺HER2tG⁻, EGFRt⁻HER2tG⁺, and EGFRt⁺HER2tG⁺ T cells (Fig. 5A). Serial immunomagnetic positive selection with trastuzumab and cetuximab reagents consistently resulted in T-cell populations that were ≥90% HER2tG⁺EGFRt⁺ (Fig. 5A). Dual CAR–expressing CD8⁺ T cells exhibited robust antitumor functionality in tumor-lysis and cytokine-release assays (Fig. 5B and C). HER2tG⁺EGFRt⁺ CAR T cells lysed K562 tumor cells transduced to express either CD19 or CD20, or both CD19 and CD20 (Fig. 5B). Likewise, HER2tG⁺EGFRt⁺ CD8⁺ T cells could recognize and lyse Raji tumor cells in which CD19 expression was ablated by CRISPR knockout (CD19KO; Fig. 5B). The dual functionality of the HER2tG⁺EGFRt⁺ CAR CD8⁺ T cells was replicated in cytokine secretion assays, as demonstrated by IL2, IFNγ, and TNFα secretion in response to stimulation with K562 cells transduced to express either CD19 or CD20, or both (CD8, Fig. 5C; CD4, Supplementary Fig. S9). HER2tG also acted as a complementary cell-tracking marker to EGFRt in vivo, as shown in experiments with dual CAR–expressing CD4⁺ and CD8⁺ T cells engrafted into NSG mice, analyzed by flow analysis of harvested bone marrow specimens (Fig. 5D).

Effector T cells expressing two CARs have the potential to yield enhanced in vivo CAR T-cell antitumor activity relative to T cells housing single CAR counterparts when tumor CAR target antigen expression heterogeneity is present. To assess this, we tested the bispecific functionality of HER2tG⁺EGFRt⁺ CAR T cells in vivo in a systemic Raji cell tumor model (CD19⁺/CD20⁺) and a Raji cell antigen loss model in which a proportion of tumor cells did not express CD19 but retained CD20 expression (CD19⁻/CD20⁺). We injected NSG mice intravenously with CD19⁺/CD20⁺ Raji eGFP:ffluc cells (Fig. 6A), or a mixture of CD19⁺/CD20⁺ and CD19⁻/CD20⁺ Raji eGFP:ffluc cells (Fig. 6B), and followed this with treatment 2 and 8 days later with CD4⁺ and CD8⁺ T cells expressing CD19CAR-HER2tG, CD20CAR-EGFRt, or both (27). As expected, mock control (untransduced) T cells failed to reduce tumor biophotonic signal in both models. In contrast, all mice treated with CAR-expressing T cells exhibited a reduction in biophotonic tumor signal and consequent improved survival after treatment relative to mock-treated mice (Fig. 6A, P < 0.0001). However, in our model of antigen-escape relapse, in which the tumor inoculum was spiked with 30% CD19KO Raji cells, CD19CAR T cells alone were ineffective in eradicating disease, whereas dual CAR–expressing T cells completely cleared tumor cells (Fig. 6B). Analysis of tumor harvested prior to the onset of hindlimb paralysis showed progression of both CD19⁻ and CD19⁺ tumor cell populations for mock (nontransduced T cell)-treated mice, progression of CD19⁻ tumor cells for mice treated with CD19CAR alone, and no observable tumor in mice treated with the dual-CAR T cells (Fig. 6C; Supplementary Fig. S10). These data demonstrate that HER2tG⁺EGFRt⁺ bispecific CAR T cells can exhibit an “OR” gated Boolean output when confronted with antigen escape tumor variants.

The development of complementary transgene-encoded devices to improve the potency and safety of CAR T cells is an area of intensive research (1, 2, 4). Many of these genetic engineering strategies deploy genetic payloads that cannot be accommodated in a single vector (1, 2, 4). Here, we combined a CD19CAR-T2A-EGFRt_epHIV7 lentivirus vector with a lentivirus housing CCR(CD122), a CCR that mediates CAR T cell–intrinsic IL2/IL15 signaling, in a T2A-HER2tG configuration (25). Dual-transduced T cells exhibited exogenous γc cytokine–independent proliferation and survival over 30 days in culture (Fig. 7A). EGFRt⁺HER2tG⁺ CD8⁺ T cells exhibited cytokine-independent cell expansion at levels equivalent to cultures supplemented with recombinant IL2 and IL15 (Fig. 7A). Parallel flow analysis demonstrated that CCR2(CD122) selects for HER2tG⁺ outgrowth in cultures without exogenous cytokine addition (% HER2tG⁺ D15–D30; Fig. 7B). These data demonstrate the versatility of HER2tG in pairing genetic engineering elements to enhance the safety and potency of CAR-EGFRt⁺ T cells.

CAR-redirected T-cell tumor targeting has shown robust antitumor potency and meaningful clinical responses in patients with otherwise lethal refractory leukemia, lymphoma, and multiple myeloma (30). Despite these successes, ongoing challenges to enhancing the impact of this emerging therapeutic modality include mitigating toxicity, overcoming antigen escape, and achieving regressions of solid tumors (31, 32). To address such challenges, synthetic biology solutions are emerging as approaches to augment the potency of CAR T cells (33, 34). However, strategies to accommodate the genetic transfer of these large and often complex synthetic biology solutions are needed.

In this article, we describe a system to facilitate the delivery of complex, modular genetic payloads to T cells. We engineered a cell-surface tag of human origin, HER2t, that is biologically inert, compact, and traceable via flow cytometry using a pharmaceutical-grade antibody. Using structural modifications in HER2t, such as HER2tG, we were able to select cells with distinct levels of a linked vector-encoded therapeutic protein. The HER2tG tag could be used in combination with the previously described EGFRt tag, housed in independent lentiviral vectors, which enabled the purification of transduced cells containing integrated payloads of the two vectors and doubled the payload capacity for engineered cells. These paired tags have the potential to be deployed in other vector systems, including nonviral transposon vectors.

We showed the potential clinical utility of our system through the cotransduction of HER2tG⁺EGFRt⁺ bispecific CAR T cells that operationally functioned as an “OR” Boolean logic gate. The clinical experience of CD19CAR and CD22CAR monotherapy for acute lymphoblastic leukemia (ALL) supports the need to overcome escape from CAR recognition by epitope loss and/or down regulation of these targets (35). The simultaneous targeting of two targets constrains escape to tumor cells that have simultaneously lost two target epitopes on genetically independent gene products. This strategy to reduce antigen escape in pediatric ALL is the subject of an initiated clinical trial (PLAT-05; NCT03330691) in which CAR T-cell products are manufactured by cotransduction of a CD19CAR-T2A-HER2tG lentivirus and a CD22CAR-T2A-EGFRt lentivirus.

Multiplexed lentiviral transduction affords some advantages to “all-in-one” approaches. For instance, bispecific CARs, which employ a single CAR housing two tandemly arrayed scFvs in series (36, 37), place constraints on the positioning of the scFv relative to the T-cell plasma membrane. Proper “spacing” has been shown to be critical for biophysical alignment of the CAR's scFv to the target molecule epitope location for a functional signaling synapse to efficiently form. A clinical trial of a CD19xCD22 bispecific CAR revealed therapeutic inferiority that was not readily predictable based on preclinical testing (NCT03241940; ref. 38).

A single lentivirus can accommodate two CAR-coding sequences in tandem, separated by an IRES or ribosome skip sequence, although care is needed with redundant sequences to avoid splicing when the vector is packaged. Although a single lentivirus with two CARs is economical at one level, due to a single vector production benefit, these CARs are essentially “locked together” in how that vector is used. Parsing CARs into modules at the level of vectors; however, achieves an economy of production for the manufacture and release of products of multiple 2× combinations of CAR specificities. Our approach also enables mixing and matching of CARs of dual specificities from a collection of vectors to achieve a combination best suited for a given therapeutic setting; for example, our clinical trial for pediatric ALL using the combination of our CD19CAR-T2A-HER2tG and CD22CAR-T2A-EGFRt (NCT03330691). The same CD19CAR-T2A-HER2tG lentivirus is also being used in combination with a conformational epitope-specific EGFRCAR-T2A-EGFRT in a clinical trial that will study the capacity of the CD19CAR to drive the proliferation of T cells coexpressing the EGFRCAR for the purpose of expanding effector cells outside the solid tumor inhibitory microenvironment to large numbers for augmented solid tumor responses (STRIvE-01; NCT03618381; ref. 39).

Because of its compact size (113 amino acids) and predicted lack of interacting ligands, we believe HER2t has advantages over the use of human CD19, CD20, and CD34 as markers of T-cell transduction (14, 15). When the level of expression of a transgene requires calibration for safety and/or function, the spacer variants of HER2t may be useful in conjunction with immunomagnetic selection with trastuzumab. Moreover, the natural amino acid sequence of the extracellular domain of HER2t and the paucity of predicted potential MHC-I antigenic peptides in that sequence (Supplementary Table S1), make HER2t less likely to be a target of an anti-transgene antibody or T-cell rejection response in human hosts, a risk that may be higher in more complex compound chimeric proteins such as compound chimeras housing CD20 and CD34 epitopes (29).

Herein, we also provide an example of the combinatorial transduction of T cells with lentiviruses housing orthogonal payloads linked to either HER2tG or EGFRt. Our proof-of-concept approach was to pair a CAR with a CCR. The realm of synthetic biology–inspired solutions to improve the potency and safety of CAR T cells is in its infancy, and we are taking a modular approach, using CAR vectors linked to EGFRt and device vectors linked to HER2tG. Advanced synthetic technologies such as the recently described Syn-Notch transcriptional control system (33, 34) advance such context-dependent control of transgene expression, but they require a large payload capacity to incorporate inducible and constitutively expressed constructs, a challenge our system can accommodate. The ability to calibrate the expression of a CAR linked to HER2t variants, such as HER2tG, may be especially important in the Syn-Notch system or other systems that involve regulated transgene outputs. Here it is critical to purify transduced cells to be dual positive for EGFRt and HER2t to couple the CAR and device outputs in individual cells. Future iterations of engineered T cells are anticipated, and the combinatorial use of tagged vectors can accommodate the complexity of these strategies.

A.J. Johnson reports other support from Juno Therapeutics during the conduct of the study; other support from Juno Therapeutics outside the submitted work; and a patent for transgene genetic tags and methods of use with royalties paid. J. Wei reports grants from Juno Therapeutics during the conduct of the study. J.M. Rosser reports other support from Juno Therapeutics during the conduct of the study. A.N. Reid reports other support from Juno Therapeutics during the conduct of the study. M.C. Jensen reports other support from Juno Therapeutics during the conduct of the study; other support from Juno Therapeutics outside the submitted work; a patent for drug regulated transgene expression, pending, issued, and licensed to Juno Therapeutics, a Bristol Myers Squibb company, a patent for transgene genetic tags and methods of use pending to Juno Therapeutics, and a patent for truncated epidermal growth factor receptor (EGFRt) for transduced T cell selection issued and licensed to Juno Therapeutics; and has interests in Umoja Biopharma and Juno Therapeutics. M.C. Jensen is a seed investor and holds ownership equity in Umoja, serves as a member of the Umoja Joint Steering Committee, and is a board observer of the Umoja Board of Directors. M.C. Jensen also holds patents, some of which are licensed to Umoja Biopharma and Juno Therapeutics—relevant patents outlined above. No disclosures were reported by the other authors.

A.J. Johnson: Conceptualization, data curation, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. J. Wei: Conceptualization, data curation, formal analysis, validation, investigation, methodology, project administration. J.M. Rosser: Data curation, formal analysis, validation, investigation, visualization, methodology, writing–review and editing. A. Künkele: Conceptualization, resources, visualization, methodology. C.A. Chang: Data curation, formal analysis, validation, methodology. A.N. Reid: Data curation, formal analysis, validation. M.C. Jensen: Conceptualization, resources, supervision, funding acquisition, investigation, project administration, writing–review and editing.

This study was supported in part by institutional funds; NIH, NCI grant RO1 CA136551-05, to M.C. Jensen; and St. Baldrick's Foundation—Stand Up To Cancer Pediatric Dream Team Translational Research grant (SU2C-AACR-DT-27-17), to A.J. Johnson, C.A. Chang, and M.C. Jensen. Stand Up To Cancer is a division of the Entertainment Industry Foundation. Research grants are administered by the American Association for Cancer Research, the Scientific Partner of SU2C. The authors thank Kamila Gwiazda and Gabrielle Curinga for comments on this article and Charlotte Schubert for editorial assistance.

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.