An ‘off-the-shelf’ fratricide-resistant CAR-T for the treatment of T cell hematologic malignancies

CAR Design

CD7-CAR was generated by using commercial gene synthesis of an anti-CD7 single chain variable fragment (scFv) sequence found in patent WO2003051926_A2. The scFv was cloned into the backbone of a 3rd generation CAR with CD28 and 4-1BB internal signaling domains in the pELNS-Ef1α lentiviral vector (a kind gift from Dr. Carl June, University of Pennsylvania)¹⁴. The construct was modified to express the extracellular domain of hCD34 via a P2A peptide to enable both detection of CAR following viral transduction and, if required, purification of CAR-T using anti-hCD34 magnetic beads. Similarly, constructed CAR-T targeting CD19 were generated using an scFv obtained from Roguska et al and were used as a non-targeting control¹⁵.

Viral vector production

To produce lentivirus, the Lenti-X 293T Cell Line (Takara Bio, Mountain View, CA) was transfected with CAR lentiviral vector and the packaging plasmids, pMD.Lg/pRRE, pMD.G, pRSV.Rev^{16, 17} using the CalPhos™ Mammalian Transfection Kit (Takara) per the manufactures instructions. Virus was harvested 36 hrs. post transfection, filtered to remove cell debris, and concentrated by ultracentrifugation for 90 mins at 25 000 rpm, 4° C (Optima LE-80K Ultracentrifuge, Beckman Coulter, Indianapolis I.N). Virus was re-suspended in phosphate buffered saline, snap frozen in liquid nitrogen and stored at −80 °C in single use aliquots.

CRISPR/cas9 gene editing

Guide RNA were designed and validated for activity by Washington University Genome Engineering & iPSC Center (Supplemental table 1). Plasmids encoding gRNA (400 ng, Addgene 43860) and spCas9 (500 ng, Addgene 43945) were electroporated into the leukemia cell line, K562, using the nucleofector 4D (Lonza. NJ) in 20 μl solution P3 (program FF-120).

RNA guides were commercially synthesized (Trilink Biotechnologies San Diego, CA), incorporating 2′-O-methyl and 3′ phosphorothioate bases at the three terminal bases of the 5′ and 3′ ends of the gRNA to protect from nuclease activity¹⁸ Full guide sequences can be found in supplemental table 2. Streptococcus pyogenes Cas9 (spCas9) mRNA (5meC, Ψ) was purchased from Trilink Biotechnologies.

Gene edited CAR-T

T cells were cultured in Xcyte media supplemented with 50 U/mL IL-2 and 10ng/ml IL-15 in the presence of anti-CD3/CD28 beads (Bead to cell ratio 3:1). On day +2 post activation, beads were removed and 4×10⁶ T cells were electroporated in 100 μl buffer P3 with 15 μg spCas9 (Trilink CA.) and 20 μg of each gRNA (Trilink) using a nucleofector 4D, program EO-115. Cells were transduced with CAR7 or CAR19 (control) lentiviral particles in the presence of polybrene (Sigma Aldrich. St Louis MO) (final conc. 6 μg/ml) on day +3. Cells were expanded for an additional 6 days prior to use in downstream experiments.

Targeted deep sequencing

The CD7 locus was amplified with primers F_gcctgcgtgggatctacctgaggca and R_ AGCTATCTAGGAGGCTGCTGGGGGC. The TRAC Locus was amplified with F_ TGGGGCAAAGAGGGAAATGA R_ GTCAGATTTGTTGCTCCAGGC. PCR products were sequenced using the Illumina MiSeq platform (San Diego CA). Editing efficiencies were determined as a percentage of sequencing reads with indels aligned to reads obtained from WT cells.

Cell lines

CD7 positive T-ALL cell lines, MOLT-3 (ACC 84), MOLT-4 (ACC 362), HSB-2 (ACC 435) and CCRF-CEM (ACC 240) were obtained from directly from DSMZ-German collection of Microorganisms and Cell cultures (Leibniz, Germany). The cell lines were mycoplasma tested and characterized by DSMZ. CCRF-CEM cells were transduced with EF1α^CBR-GFP lentivirus. GFP positive cells were sorted and cloned to establish the CCRF-CEM^CBR-GFP cell line.

Chromium release assay

CAR-T were incubated with MOLT-3, MOLT4, HSB2 or CCRF cell lines (1 × 10⁴ total cells/well) at an effector:target [E:T] ratio ranging from 25:1 to 0.25:1 in RPMI supplemented with 5% fetal calf serum. Chromium-51 release assays were performed as described previously.¹⁹

In vitro primary T-ALL killing assay

Primary T-ALL from consented patients were obtained from the Siteman Cancer Center (IRB #201108251). Informed consent was obtained from all subjects. Primary cells were labelled with 150 nM carboxyfluorescein succinimidyl ester (CFSE) (sigma Aldrich, MO) to enable distinction between T-ALL blasts and CAR-T. Labeled cells were co-incubated at 1:1 ratio with either CD7^ΔCART7, UCART7 or their respective CD19 controls for 24 hours in Xcyte media supplemented with 50 U/mL IL-2 and 10ng/ml IL-15, 50 ng/ml SCF, 10 ng/ml IL-7, and 20 ng/ml FL3TL prior to FACS analysis. Absolute cell counts of viable target cells were quantified by flow cytometry using 7-aminoactinomycin D and SPHERO AccuCount fluorospheres (Spherotech Inc., Lake Forest, IL, USA) as previously described.²⁰ Data were analyzed using FlowJo V10.

Fratricide assay

WT T cells were cultured in Xcyte media supplemented with 50 U/mL IL-2 and 10ng/ml IL15 in the presence of anti-CD3/CD28 beads (bead to cell ratio 3:1). Beads were removed after 48 hours and T cell were transduced with lentivirus particles to express GFP. Seventy-two hrs. post transduction, T cells were sorted for GFP using flow cytometry and co-incubated with CD7^ΔCART7 or CD7^ΔCART19 at a ratio of 1:1 for 24hrs in Xcyte media supplemented with 50 U/mL IL-2 and 10ng/ml IL-15. Percent GFP+ cells were calculated as a percentage of total viable cells, quantified by flow cytometry using 7-aminoactinomycin D.

T cell phenotype analysis

Cultured T cells were washed in PBS/0.1% BSA and re-suspended at 1×10⁶ cells in 50 uL Brilliant Buffer (BD Biosciences) supplemented with 4% rat serum for 15 minutes at 4C. Cells were then incubated for 30 minutes at 4°C in 100 μL of Brilliant Buffer using the following antibody fluorophore conjugates (all from BD Biosciences unless otherwise noted): CD7 BV421, CD4 BV510, CCR4 BV605 (BioLegend), CD8 BV650, CD196 BV786 (BioLegend), CD3 AF488, CD45RA PerCPCy5.5, CD183 PE, CD197 PE-CF594, CD185 PE-Cy7 (BioLegend), and CCR10 APC (R&D Systems). Full details of fluorophore conjugated antibodies can be found in supplemental table 5. Cells were then washed twice in PBS/0.1% BSA and data acquired on a ZE5 (Yeti) cytometer (BioRad/Propel Labs). Compensation and analyses were performed on FlowJo V10 (TreeStar) using fluorescence minus one (FMO) controls. Statistical analyses were performed on GraphPad Prism 7 using 2-way ANOVA with Bonferroni post-hoc corrections.

Animal models

Animal protocols were in compliance with the regulations of Washington University School of Medicine Animal Studies Committee. Six to ten week old NOD.Cg-Prkdc^scid Il2rg^tm1Wjl/SzJ (NSG) were used in all mice experiments. Both male and female mice were used in all experiments and randomly assigned to a treatment group.

Off target analysis

Statistical analysis

The determination of sample size and data analysis for this study followed the general guideline for animal studies²⁴. The distributions of time-to-death were described using Kaplan-Meier product limit method and compared by log-rank test. All the other in vivo data were summarized using means and standard deviations. The differences were compared using two-sample Student t-test, one-way ANOVA, or two-way ANOVA for repeated measurement data as appropriate, followed by ad-hoc multiple comparisons for between-group differences of interest (see also details for specific info of each Figure). Based on the law of diminishing returns, Mead (1998) recommended that a degree of freedom (DF) of 10-20 associated with error term in an ANOVA will be adequate for a pilot study to estimate preliminary information²⁵. The normality of data was assessed graphically using residuals and the similarity of variance across groups was also assessed visually by checking the estimated variance of each group. A logarithm transformation was performed as necessary to better satisfy the normality and homoscedasticity assumptions. The resultant p-values were adjusted by a step-down Bonferroni adjustment for multiple comparisons if needed. Comparing to the widely-used Bonferroni adjustment, a step-down method is more powerful (smaller adjusted p-values) while maintaining strong control of the familywise error rate. All analyses were two-sided and significance was set at a p-value of 0.05. The statistical analyses were performed using SAS 9.4 (SAS Institutes, Cary, NC).

CD7-CAR-T induce substantial fratricide

To generate the CD7-CAR-T (CART7), an anti-CD7 single chain variable fragment (scFv) was commercially synthesized and cloned into a 3rd generation CAR backbone with CD28 and 4-1BB internal signaling domains. The extracellular domain of hCD34 was added after a P2A peptide to enable both the detection of CAR following viral transduction and purification using anti-hCD34 magnetic beads (Fig 1a)²⁶. CAR-T targeting CD19 (CART19) were used as an irrelevant CAR-T control. Following transduction of T cells, there were significantly fewer CART7 than CART19 (Fig 1b). In addition, CART7 were biased towards a CD4 phenotype when compared to CART19 (Supplemental table 6).

Deletion of CD7 by CRISPR/Cas9

To prevent fratricide, we deleted CD7 in CAR-T using CRISPR/Cas9 gene-editing. Ten guide RNAs (gRNA) targeting CD7 were designed and validated for activity (Supplemental table 2). Plasmids encoding the gRNA and Cas9 were electroporated into the K562 leukemia cell line. CD7g4 and CD7g10 had the highest gene-editing efficiencies, as determined by targeted deep-sequencing across the CD7 locus (Fig 1c) and were selected for further investigation. CD7g4 and CD7g10 guides were commercially synthesized, incorporating 2′-O-methyl and 3′phosphorothioate bases at the three terminal bases of the 5′ and 3′ of the gRNA to protect from nucleases activity¹⁸. We next tested the efficacy of gene-editing by both CD7g4 and CD7g10 in human primary T cells. Activated T cells were electroporated with gRNA and Cas9 mRNA (Fig 1d) and CD7 expression analyzed by flow cytometry on day +7. CD7g4 was the most effective at deleting CD7 expression, reducing the percentage of CD7+ T cells from 98.8%±0.17 to 9.1%±1.74 (Fig 1e). Effective disruption of the CD7 locus was confirmed by targeted deep-sequencing with indels observed in 89.14% of CD7 sequence reads (Fig 1e). Only minimal loss of viability was observed 24hrs post electroporation (Fig 1e). As CD7g4 was most effective at deleting expression of CD7 in T cells, all future experiments were performed using CD7g4.

CD7^ΔCART7 prevents fratricide and kills T-ALL in vitro and in vivo

Next we generated CART cells deficient in expression of CD7. We performed CRISPR/Cas9 gene-editing of CD7 in primary T cells followed by transduction of CD7 edited T cells (CD7^Δ) with CAR7 (Fig 2a) to generate CD7^ΔCART7. We anticipated that there would be a low level of fratricide resulting from residual CD7 surface expression following gene-editing, and this was confirmed by a moderate reduction in CD7^ΔCART7 yield relative to CD7^ΔCART19 (7.5-fold vs. 12.6-fold expansion over 6 days Fig 2a). Autologous T cells transduced with GFP were effectively killed by CD7^ΔCART7 but not by CD7^ΔCART19 confirming the requirement for CD7 deletion in CART targeting CD7 (Fig. 2b). Finally, in contrast to CD7^ΔCART19, CD7^ΔCART7 effectively killed CD7+ T-ALL cell lines MOLT-4 (70% CD7+), MOLT-3 (96% CD7+) and HSB-2 (99% CD7+) as determined by 4hr Cr release assays (Fig 2c)

To assess the activity of CD7^ΔCART7 in a xenogeneic model of T-ALL, 5×10⁵ Click Beetle Red luciferase (CBR) labeled CCRF-CEM T-ALL (99% CD7+ by FACS) cells were injected I.V. into NSG recipients prior to infusion of 2×10⁶ CD7^ΔCART7 or non-targeting CD7^ΔCART19 control cells on day+4 (Fig 3a). In contrast to mice receiving CD7^ΔCART19, or mice injected with tumor only, mice receiving CD7^ΔCART7 had significantly prolonged survival (p=0.0003 Fig 3b) and reduced tumor burden as determined by bioluminescent imaging (BLI) (Fig 3c). CD7^ΔCART7 treated mice exhibited signs of GvHD and, consistent with previous reports^27–29, most mice eventually succumbed to leukemia. To assess efficacy of CD7^ΔCART7 against patient primary T-ALL cells, CAR-T were tested against patient derived xenografts. However, T-ALL blasts were only detectable in mice receiving tumor only and were eliminated in mice receiving either CD7^ΔCART7 or CD7^ΔCART19 suggesting that CD7^ΔCART7 and CD7^ΔCART19 maintain alloreactivity in vivo in NSG mice (Supplemental figure 1).

Double deletion of TRAC and CD7 in CART7 prevents fratricide and GvHD, and maintains robust CD7 directed T-ALL killing

To overcome alloreactive barriers that limit the use of non-self T cells, due to the risk of lethal GvHD, we generated CAR-T in which both CD7 and the TRAC were genetically deleted. The gRNA sequence, targeting TRAC, was obtained from Osborn et al³⁰. T cells were activated using anti-CD3/CD28 beads for two days prior to bead removal and electroporation with 20 μg of CD7g4, 20 μg of TRACg and 15 μg of spCas9 mRNA (Fig 4a). Multiplex CRISPR/Cas9 gene-editing resulted in the simultaneous deletion of CD7 and TRAC in 72.8%±1.92 of cells, as determined by FACS analysis (Fig 4a).

In keeping with recent nomenclature in the field, our CD7^ΔTRAC^ΔCART7 was termed universal CART7 or UCART7³¹. UCART7 was as effective as CD7^ΔCART7 at killing T-ALL cell lines in vitro. UCART7 had no proliferation defect when compared to CD7^ΔCART7, however, as observed with CD7^ΔCART7, UCART7 resulted in moderately reduced CAR-T proliferation and yield relative to the CD19 control CART (Fig 4b). Since incomplete gene-editing of TRAC leaves residual potentially alloreactive CD3⁺ CAR-T, these were depleted by negative selection using anti-CD3 magnetic beads on Day +8. Both UCART7 and CD7^ΔCART7 killed CD7+ T-ALL cell lines, MOLT3, CCRF-CEM and HSB-2 in vitro with equally high efficiencies demonstrating no loss of efficacy upon double deletion of CD7 and TRAC (Fig 4c). Interestingly, non-specific killing by UCART19 was attenuated at high effector to target (E:T) ratios when compared to CD7^ΔCART19 suggesting loss of alloreactivity following TRAC deletion.

We next tested the ability of UCART7 to kill primary T-ALL blasts in vitro. Due to the similarity of antigen expression between primary T-ALL and CAR-T, total leukocytes containing T-ALL from patients (83.9% – 99.6% CD7+ cells) were labeled with CFSE to clearly distinguish T-ALL from CAR-T. Patient cells were incubated with CAR-T at a ratio of 1:1 for 24hrs. Both CD7^ΔCART7 and UCART7 killed an average of 95% of cells across all three primary samples, relative to the respective CD19 control CAR-T (Fig 5a), thus demonstrating efficacy against human primary T-ALL in vitro.

In light of the anti-tumor activity we observed when either CD7^ΔCART and CD7^ΔCART19 were infused into primary T-ALL PDX bearing NSG mice (Supplemental figure 1), we tested the capacity of UCART7 to kill primary T-ALL in vivo without inducing an alloreactive Graft-vs.-Leukemia effect (GvL) or xenogeneic GvHD (Fig 6). Recipients of T cells edited to delete TRAC (TRAC^Δ) exhibited high tumor burden in both the blood and spleen (Fig. 6b) when compared to recipients of WT T cells (Day +48 spleen p <0.0001, blood p = 0.0001), demonstrating an inability of TRAC^Δ to induce allogeneic GvL. Furthermore, considerable expansion of alloreactive T cells (Fig 6a) and severe GvHD (mean clinical GvHD score = 5.66, Fig 6c) was observed in recipients of WT T cells. In contrast, GvHD was completely absent and T cells undetectable in mice receiving TRAC^Δ T cells (Fig 6c). T-ALL blasts were absent in peripheral blood of mice receiving UCART7 in comparison to mice receiving UCART19 with T-ALL comprising >56% of total CD45⁺ cells in these mice (p<0.0001), similar to the high tumor burden observed in PDX only controls (Fig 6b). Concordant results were observed in the spleen (Fig 6b, UCART7 <3% T-ALL vs. UCART19 = 85.87% T-ALL; p<0.0001), with TRAC^Δ T cells, UCAR19 and PDX only recipient mice exhibiting splenomegaly. In stark contrast UCART7 recipients had normal sized spleens (data not shown). Furthermore, unlike UCART19, UCART7 were detectable 6 weeks’ post injection as detected by the hCD34 epitope, demonstrating persistence of UCART7 in vivo (Supplemental Figure 2).

Off target nuclease activity

High efficiency gene-editing with CRISPR/Cas9 can induce undesirable off-target genetic changes that could have potentially detrimental effects on the biology of these T cells and subsequently on the recipients that receive UCART7 infusions. We used two different techniques to assess off-target genetic changes in human primary T cells, both of which rely on the insertion of a small double stranded oligodeoxynucleotide (dsODN) at DNA double strand breaks. The first protocol utilized a modified version of GUIDE-seq²³, without the inclusion of barcoded indexes, to specifically amplify target sites surrounding the inserted dsODN using PCR; the second technique used Integrated DNA Technologies (IDT) capture probes to enrich loci containing the target dsODN sequence. Both techniques use next-generation sequencing to identify the loci of inserted dsODN. To ensure identification of bona fide sites of off-target nuclease activity, each condition (CD7g4, TRACg and combined CD7g+TRACg) was performed in triplicate, generating an average 1.26×10⁶ sequencing reads per replicate. Loci with bidirectional sequencing reads >10× coverage were included in the analysis. First, we assessed the ability of each technique to identify on-target sites. Both GUIDE-seq and dsODN capture robustly identified sites of on-target activity, with each on-target site generating between 300× and 22,000× coverage across all three replicates in each condition (Table 1). Next, we assessed off target nuclease activity using the same stringency. GUIDE-seq revealed a single off target site in multiplex edited T cells (CD7g4+TRACg), an intronic insertion in RBM33, present in all three replicates. No off-target sites were observed when assessed by dsODN capture despite high coverage of on-target activity (Table 1). Upon relaxing the stringency of GUIDE-seq analysis to include potential off target sites present in two or more replicates, we identified four additional potential off-target sites for CD7g and four sites for TRACg. One additional sit of potential off-target activity for multiplexed CD7 and TRAC gene-editing was identified (Table 1). Loci identified by GUIDE-seq were not present in the data obtained using dsODN capture, nor were additional sites of off-target nuclease activity identified following analysis of these data with reduced stringency. In addition, dsODN capture allows the identification of gene rearrangements between multiple sites of on target nuclease activity, however, no gene rearrangements were observed between CD7 and TRAC following multiplex gene editing. These results suggest that high efficiency CRISPR/Cas9 gene-editing of primary T cells with CD7g4 and TRACg, individually or in combination, is not associated with significant off-target activity or on-target gene rearrangements, using these platforms.