Hu8F4-CAR T cells with mutated Fc spacer segment improve target-specificity and mediate anti-leukemia activity in vivo

Hu8F4 is a T cell receptor (TCR)-like antibody with high affinity for leukemia-associated antigen PR1/HLA-A2 epitope. Adapted into a chimeric antigen receptor (CAR) format, Hu8F4-CAR is comprised of the Hu8F4 scFv, the human IgG1 CH2CH3 extracellular spacer domain, a human CD28 costimulatory domain, and the human CD3ζ signaling domain. We have demonstrated high efficacy of Hu8F4-CAR-T cells against PR1/HLA-A2-expressing cell lines and leukemic blasts from AML patients in vitro. Previous studies have shown that modification of the Fc domains of IgG4 CH2CH3 spacer regions can eliminate activation-induced cell death and off-target killing mediated by mouse Fc gamma receptor (FcgR)-expressing cells. We generated Hu8F4-CAR(PQ) with mutated Fc receptor binding sites on the CH2 domain of Hu8F4-CAR to prevent unwanted interactions with FcgR-expressing cells in vivo. The primary human T cells transduced with Hu8F4-CAR(PQ) can specifically lyse HLA-A2+ PR1-expressing leukemia cell lines in vitro. Furthermore, both adult donor-derived and cord blood-derived Hu8F4-CAR(PQ)-T cells are active and can eliminate U937 leukemia cells in NSG mice. Herein, we demonstrate that modification of the IgG1-based spacer can eliminate Fc receptor-binding-induced adverse effects and Hu8F4-CAR(PQ)-T cells can kill leukemia in vivo.


Introduction
The advent of chimeric antigen receptor (CAR) therapy has ushered in a wave of immunotherapeutic advancements, particularly in the treatment of hematologic malignancies (1)(2)(3)(4)(5).CAR therapy has achieved profound e cacy and multiple approvals by the US Food & Drug Administration, particularly in the form of therapies targeting CD19 and B-cell maturation antigen (BCMA) expressed in hematologic malignancies(6).In the case of acute myeloid leukemia (AML), our group has identi ed and characterized the leukemia-associated antigen, PR1, a nonameric, HLA-A2-restricted peptide derived from serine proteases proteinase 3 and neutrophil elastase (7)(8)(9).PR1 is overexpressed on myeloid leukemia cells, and we have demonstrated the ability and e cacy of targeting the PR1 peptide bound to HLA-A2 on AML blasts by using PR1 peptide as a vaccine in a clinical trial, PR1-speci c cytotoxic T lymphocytes, using a rst-in-class T cell receptor-like monoclonal antibody named Hu8F4, and a bispeci c CD3/Hu8F4 antibody (10)(11)(12)(13)(14). Furthermore, we have previously reported on the development and in vitro e cacy of the Hu8F4 monoclonal antibody as a 2nd generation CAR therapy named Hu8F4-CAR (15).
In the last 30 years, the development of CAR therapies has encompassed a variety of different structural and signaling components each of which confer different capabilities and advantages(16-18).Hu8F4-CAR consists of the Hu8F4 scFv (single chain variable fragment) VL2 and VH regions, the human IgG1 CH2CH3 extracellular spacer domain, a transmembrane and intracellular human CD28 costimulatory signaling domain, and human CD3z intracellular signaling domain.We have demonstrated the e cacy of Hu8F4-CAR-T cells in vitro against PR1/HLA-A2-expressing cell line targets and leukemic blasts from AML patients.However, in a leukemia xenograft mouse model, Hu8F4-CAR-T cells lacked activity and the ability to persist in vivo (data not shown).Interestingly, others have reported that non-signaling extracellular spacer domains can trigger activation-induced cell death (AICD) upon binding of Fc gamma receptor (FcgR)-expressing cells in vivo leading to activation and death of the CAR-T cells, thus limiting their persistence and activity in vivo (19,20).Additionally, such interactions with cells expressing FcgR in vivo could result in off-target killing by activated CAR-T cells.Previous studies have shown that modifying these Fc domains such as those within the IgG4 CH2CH3 regions to eliminate FcgR binding restored CAR in vivo antitumor effects and persistence (20).Since Hu8F4-CAR indeed contains an extracellular nonsignaling Fc domain of IgG1 (IgG1 CH2CH3), we mutated the CH2 domain within the spacer to prevent these interactions with FcgR-expressing cells in vivo and termed this mutated Hu8F4-CAR as Hu8F4-CAR(PQ).While others have reported the ability to abolish Fc domains in IgG4 to prevent activation-induced cell death, few have reported the ability to accomplish this in an IgG1-based CAR therapy.Herein, we report that abrogating FcgR-binding capabilities within the IgG1 spacer region preserves Hu8F4-CAR(PQ)'s speci c recognition of PR1/HLA-A2 and is active against PR1-expressing targets in vitro and in vivo.

Generation of mutated Fc receptor binding sites on Human IgG1 CH2CH3 domain
Site-directed mutagenesis was conducted using two sets of primers to alter amino acid residues ELLG (IgG1 CH2CH3 amino acid (AA)# 6-9) to PVA and N (CH2CH3 AA# 180) to Q. ELLG mutagenesis primers were CCGTGCCCAGCACCTCCCGTGGCCGGAACCGTCAGTCTTC (sense) and GAAGACTGACGGTTCCGGCCACGGGAGGTGCTGGGCACGG (antisense).Q mutagenesis primers were GAGGAGCAGTACCAGAGCACGTACCGT (sense) and ACGGTACGTGCTCTGGTACTGCTCCTC (antisense).Mutagenesis was conducted using the QuikChange Lightning Site-Directed Mutagenesis Kit from Agilent Technologies (Santa Clara, CA).Mutagenesis was con rmed via Sanger sequencing.

Generation of mutated Hu8F4-CAR(PQ)-T cells
Fc-mutated Hu8F4-CAR(PQ)-T cells were generated using healthy adult donor peripheral blood mononuclear cells (PBMCs) or umbilical cord blood mononuclear cells (CBMCs) obtained from the University of Texas MD Anderson Cancer Center Blood Bank.The mononuclear cells were isolated via density gradient centrifugation using Histopaque-1077 Hybri-Max from Sigma-Aldrich (St Louis, MO).
Because HLA-A2 − (negative) cells are required to generate Hu8F4-CAR-T cells, HLA-A2 status of each donor was con rmed using 5µL FITC-conjugated anti-human HLA-A2 (clone BB7.2) antibody from BioLegend (San Diego, CA) and assessed on a ow cytometer.To activate the T cells, two days prior to transduction (Day − 2), 24-well non-tissue culture-treated plates were coated with 0.5mL per well of a 1µg/mL cocktail of anti-human CD3 (clone HIT3a) and anti-human CD28 (clone CD28.2) from BioLegend (San Diego, CA) each for four hours.Following this, each well was blocked with complete medium consisting of 223mL Click's Medium from Irvine Scienti c (Santa Ana, CA), 222mL of RPMI 1640 from Corning (Corning, NY), 50mL of heat-inactivated fetal bovine serum from Gibco (Waltham, MA), and 5mL penicillin/streptomycin solution from Cytiva (Marlborough, MA) for 30 minutes and 1 million HLA-A2 − (negative) PBMCs were plated per well and incubated at 37 o C overnight.The following day (Day − 1), PBMCs were stimulated with 50IU/mL rhIL-2 from R&D Systems (Minneapolis, MN) and a separate 24well non-tissue culture-treated plate was coated with 7µL human RetroNectin from Takara (Kusatsa, Japan) in 1mL of sterile PBS per well and incubated overnight at 4 o C. The day of transduction (Day 0), RetroNectin solution was removed, blocked with 1mL complete medium for 30 minutes, and 2mL of Hu8F4-CAR(PQ)-packaged retrovirus supernatant was added to each well.Upon adding retrovirus to each well, the plate was centrifuged at 2000g for 90 minutes.Retroviral supernatant was aspirated and 2mL of complete CAR-T culture medium supplemented with 100IU/mL rhIL-2, and 0.25-0.75x10 6/2mL/well activated and stimulated T cells were added to each RetroNectin-and retrovirus-coated well and centrifuged at 1000g for 30 minutes before beginning CAR-T culture.Medium was changed every 2-3 days with complete CAR-T culture medium supplemented with 100IU/mL rhIL-2.Transduction e ciency, phenotypic assessment, and cytotoxicity assay was conducted at day 7 post-transduction.

In vitro cytotoxic assessment of Hu8F4-CAR(PQ)-T cells
Freshly transduced Hu8F4-CAR(PQ)-T cells were assessed for functional cytotoxic activity via a calcein-AM-based cytotoxicity assay.Target cells included the U937 wildtype (PR1/HLA-A2 − ), U937 A2 + (HLA-A2transduced, PR1 + ), and T2 cell line (TAP-de cient cell line able to present exogenous peptides).T2 target cells were pulsed with 30µg of either PR1 (VLQELNVTV) or HIVgag (SLYNTVATL) peptide, then labeled with 5µg/mL calcein-AM from Corning (Corning, NJ).A suspension of 50µL containing 5000 target cells in RPMI 1640 medium supplemented with 10% heat-inactivated fetal bovine serum and 1% penicillin/streptomycin were plated per well in a 96-well V-bottom plate.With the same suspension medium, 50µL of effector cells were co-incubated for 3.5 hours with target cells.Effector cell numbers depended on effector-to-target (E:T) cell ratios of either 8:1, 5:1, 4:1, 2:1, 1:1, 0.5:1, or 0.25:1.Cytotoxicity was measured via calcein-AM released from lysed cells into the supernatant via uorescence reading (Ex 485nm / Em 528nm) on an Agilent BioTek Cytation3 plate reader (Winooski, VT).Control wells to calculate uorescence in test wells included control wells bearing only calcein-AM labeled target cells (to measure spontaneous release of calcein-AM) and wells bearing only calcein-AM-labeled target cells with a Triton X-100 detergent from Sigma-Aldrich (St Louis, MO) added (to measure maximal release of calcein-AM).Cell lysis was calculated via the following formula: % lysis = ((test sample RFU -SR RFU)/(MR -SR)) * 100, where RFU = relative uorescence unit, SR = spontaneous release, and MR = maximal release.Data was acquired on the BioTek Gen 5 software, analyzed via Microsoft Excel, and plotted using GraphPad Prism software.
In vivo bioluminescence (BLI) was performed using a cooled charge-coupled device camera system (IVIS Imaging System 100 from Xenogen /Caliper Life Sciences, Alameda, CA) weekly after the inoculation of U937-A2-Luc/GFP cells.Mice were injected with 75 mg/kg D-luciferin (Beetle Luciferin Potassium Salt, Regis Technologies, Morton Grove, IL) in 100 µL PBS subcutaneously near the scapula and were placed in the light-tight chamber of the imaging system under iso urane anesthesia.Dorsal luminescent images were acquired 10 minutes after D-luciferin injection.The total signal in the ROI (photons/s) was quanti ed using Living Image software (version 4.7; PerkinElmer.Inc).The same ROI was applied to all images acquired sequentially in a single-imaging session for a given mouse.The growth ratio was plotted against days after cell inoculation.

Expression of Fc-mutated Hu8F4-CAR(PQ) on primary T cells from adult PBMCs
We have previously described the development of a 2nd generation chimeric antigen receptor (CAR) against the leukemia-associated antigen, PR1/HLA-A2, named Hu8F4-CAR (15).Hu8F4-CAR consists of the Hu8F4 scFv VL2 and VH regions, the human IgG1 CH2CH3 extracellular spacer domain, a transmembrane and intracellular human CD28 costimulatory signaling domain, and human CD3z intracellular signaling domain (Fig. 1A).We have demonstrated the e cacy of Hu8F4-CAR in vitro on primary T cells against PR1/HLA-A2-expressing leukemia cell lines as well as blasts from AML patients.However, in vivo, Hu8F4-CAR T cells did not show in vivo activity in a leukemia xenograft model (data not shown).Others have reported the length of nonsignaling spacers/stalks exerted a major effect on CAR-T cell effector function after target engagement, and, more important, reported that spacers/stalks with Fc domains require alterations to prevent in vivo interactions with cells expressing Fc gamma receptors (FcgR) (20).They reported that CAR-T spacer Fc domain interaction with cells in vivo expressing Fc gamma receptors resulted in off-target activation of CAR-T cells and led to activation-induced cell death (AICD) of the CAR-T cells and limited their persistence and anti-tumor activity in vivo (20).
To generate mutated Fc receptor binding sites to prevent activation-induced cell death of CAR-T cells reported in IgG4 extracellular spacer domains (20), we conducted mutagenesis on Fc binding sites on the human IgG1 CH2CH3 extracellular spacer domain of Hu8F4-CAR (Fig. 1B).We replaced the four amino acids ELLG (Glu-Leu-Leu-Gly) of the IgG1 CH2 with tri-amino-acid PVA (Pro-Val-Ala) and mutated an additional N (Asn) to Q (Gln) (Fig. 1C), creating Hu8F4-CAR(PQ).
To determine the function of Hu8F4-CAR(PQ), we transduced primary T cells from either healthy adult donor HLA-A2 − peripheral blood mononuclear cells (PBMC) or healthy umbilical cord blood mononuclear cells (CBMC) with retrovirus bearing Hu8F4-CAR(PQ) and measured transduction e ciency at day 7 posttransduction.We stained the cells with PR1/HLA-A2 tetramer, or CMV pp65/HLA-A2 tetramer as a control, and anti-human A niPure IgG F(ab') 2 which binds to the IgG1 spacer domain of Hu8F4-CAR.Data from 10 different experiments showed consistent, high transduction e ciency of Hu8F4-CAR(PQ) into primary T cells with a mean e ciency of 94.79% +/-4.144%S.D (Fig. 2A; representative example shown in ow plots).Additionally, no differences were seen in the transduction e ciencies of CD4 + T cells compared to CD8 + T cells with both exhibiting mean transduction e ciencies greater than 93% (Fig. 2B).The majority of Hu8F4-CAR(PQ)-transduced adult T cells were of memory phenotype, with effector memory (EM) and central memory (CM) cells averaging 36.59% and 31.12% of the total transduced cells, respectively (Fig. 2C).This is in contrast to non-transduced cells which displayed higher frequencies of terminal differentiated (TD) effector T cells and naïve T cells (data not shown).The prominence of memory T cells within the Hu8F4-CAR(PQ) transduced population could further enhance activity, function, and long-term persistence of the cells in vivo.
In terms of Hu8F4-CAR(PQ)-T cells derived from umbilical CBMCs, our data from three independent experiments, each with a different healthy umbilical CBMC donor, also show consistent, high transduction e ciency of Hu8F4-CAR(PQ) into umbilical cord blood T cells (mean 91.20% +/-9.885%S.D.) (Fig. 3A).As with the adult primary T cells, no signi cant differences were seen between the transduction e ciencies of CD4 + and CD8 + cord blood T cells (Fig. 3B).Phenotypically, umbilical CBMC-derived T cells transduced with Hu8F4-CAR(PQ) are primarily associated with effector phenotypes, predominantly TD effector and EM T cells (Fig. 3C).Not surprisingly, phenotypic differences were evident between Hu8F4-CAR(PQ)-T cells and non-transduced T cells (data not shown).Although there are differences, as noted above, on balance Hu8F4-CAR(PQ) is e ciently transduced and expressed in primary T cells and across phenotypically different T cell subsets.

Hu8F4-CAR(PQ)-T cells actively and speci cally target PR1expressing cell lines in vitro
To assess the functional activity of Hu8F4-CAR(PQ)-T cells after mutating the Fc binding capability in the CAR spacer domain, in vitro cytotoxicity assays were used to investigate whether Hu8F4-CAR(PQ)-T cells are active against PR1-expressing targets.To measure cytotoxic activity in PBMC-derived Hu8F4-CAR(PQ)-T cells, U937 WT (wildtype), U937-A2 + (HLA-A2-transduced), and the T2 cell line were used as targets.T2 cell line targets were pulsed with either PR1 peptide as speci c target or HIVgag peptide as an irrelevant target control.Hu8F4-CAR(PQ)-T effector cells and targets were combined in effector-to-target ratios of either 8:1, 4:1, 2:1, 1:1, or 0.5:1 and co-incubated for 3.5 hours before measuring cell lysis.As shown in Fig. 4A, Hu8F4-CAR(PQ)-T cells show speci c cytolytic activity against T2 cells pulsed with PR1 peptide as well as against the U937-A2 + cell line compared to T2 cells pulsed with HIVgag peptide and the U937 WT cell line, respectively.In addition, umbilical CBMC-derived Hu8F4-CAR(PQ)-T cells were assayed for cytotoxic activity against cell line targets.The U937 A2 + cell line was used as a target and CBMC-derived Hu8F4-CAR(PQ)-T cells or non-transduced T cells were used as effector cells in effector-totarget ratios of either 5:1 and 2:1or 2:1, 1:1, 0.5:1, and 0.25:1.As shown in Fig. 4B, there is strong activity of Hu8F4-CAR(PQ)-T cells against the U937-A2 + cell line.Altogether, Hu8F4-CAR(PQ) actively and speci cally eliminates PR1-expressing cell line targets in vitro.

PBMC-derived Hu8F4-CAR(PQ) -T cells eliminate U937 leukemia cells in NSG mice
Using the established U937 leukemia/NSG mouse model in our lab, we evaluated whether the anti-tumor function of Hu8F4-CAR(PQ)-T cells is dose dependent in vivo.A bioluminescence (BLI) imaging system was used to systemically study spatial-temporal pattern of U937-A2 + cells transduced with a lentiviral vector encoding the re y luciferase (Luc-GFP) genes (U937-A2 + -Luc).U937-A2 + -Luc cells (n = 5000 cells) were injected into NSG mice and monitored by BLI for 4 weeks post-infusion.Three different doses of PBMC-derived Hu8F4-CAR(PQ)-T cells were injected into the mice on day 3 when the U937-A2 + -Luc leukemia cells had engrafted and were detectable with BLI.Mice that received 10 million non-transduced T cells (NT), as a control, showed persistent engraftment of leukemia while mice receiving Hu8F4-CAR(PQ)-T cells displayed a signi cantly reduced leukemia burden (Fig. 5A).The overall survival of U937-A2 + -Luc-bearing mice was increased after treatment with higher cells doses of Hu8F4-CAR-T cells (4 and 10 million cells).Furthermore, PBMC-derived Hu8F4-CAR(PQ)-T cells can eliminate leukemia cells and prolong the overall survival of U937-A2 + -Luc -bearing mice (Fig. 5B).The Hu8F4-CAR(PQ)-T (human CD45 + CD3 + ) cells are present and detectable both 8-and 28-days post-infusion in blood (Fig. 5C).This data demonstrates that Hu8F4-CAR(PQ)-T cells exhibit speci c antitumor effects and persist in vivo.
Expanded CBMC-derived Hu8F4-CAR(PQ)-T cells are active against U937 leukemia cells in vivo.Previously we have shown that human umbilical CBMC-derived Hu8F4-CAR T cells were capable of killing leukemia cell lines in vitro (15).In the current study, Hu8F4-CAR(PQ)-T cells were generated from three different CBMC-donors and were subsequently infused into NSG mice with established U937-A2 + -Luc leukemia (Fig. 6).By week two, we can detect leukemia cell expansion in control mice (PBS and NT) by BLI (Fig. 6A).Similar results were seen with CBMC-derived Hu8F4-CAR(PQ)-transduced T cells from another donor (data not shown).Mice treated with 1 and 4 million of CBMC-derived Hu8F4-CAR(PQ)-T cells showed much slower expansion of U937-A2 + -Luc cells.More importantly, mice treated with 10 million CBMC-derived Hu8F4-CAR-T cells did not have any detectable U937-A2 + -Luc cells by week four and exhibited prolonged survival (Fig. 6B).We can detect the CBMC-derived Hu8F4-CAR(PQ)-T (human CD45 + CD3 + ) cells both 8-and 28-days post-infusion in blood (Fig. 6C).In conclusion, we have demonstrated that CBMC-derived Hu8F4-CAR(PQ)-T cells can protect mice against U937-A2 + -Luc leukemia cells, confer a signi cant survival advantage, and persist up to 28 days, a marked improvement from the initial Hu8F4-CAR construct, prior to Fc domain mutation.

Discussion
We generated a Hu8F4-CAR(PQ) construct with mutated Fc receptor binding sites on the CH2 domain of Hu8F4-CAR and demonstrated consistent and high transduction e ciency of the new construct into both CD4 + and CD8 + T cells.Our in vitro cytotoxicity data shows potent and speci c activity against PR1/HLA-A2 + targets.More importantly, Hu8F4-CAR(PQ)-T cells demonstrate strong anti-leukemic e cacy in vivo.
We modi ed the original Hu8F4-CAR design to resolve the lack of in vivo activity while maintaining the speci city and functionality of the CAR-T cells in vivo.One potential route to modify Hu8F4-CAR's design was described in the ndings Hudecek and others who reported the ability to restore in vivo function and persistence by mutating Fc-receptor binding domains with IgG-based spacer domains to prevent AICD and off-target activation (20,22).Use of human IgG (IgG1 and IgG4)-derived hinge region plus heavy chain constant regions (CH2CH3) as the extracellular spacer domain is a common practice in CAR design.These modi cations allow for straightforward detection of CAR expression using anti-Fc antibodies as well as exible length adjustability by removing either CH2 and/or CH3 domains while still maintaining a low immunogenicity pro le.Countering these advantages are the potential undesirable effects such as Fc-binding by FcgR-bearing cells, leading to off-target activation of CAR-T cells and nonspeci c target killing.In addition, antigen-independent activation due to binding of FcgR-expressing cells could result in AICD of the CAR-T cells and diminishing long term anti-tumor activity in vivo (20).Hudecek et al. described a modi cation of an ROR1-speci c CAR with an IgG4 constant heavy chains 2 and 3 (CH2CH3)-derived spacer (20).They replaced the known FcgR binding site in the CH2 of IgG4 with an IgG2 sequence that lacked such interaction.A similar effect was also observed in an earlier study with an IgG1-based anti-CD30-CAR by another group (22).Subsequently, Hudecek abrogated a conserved Nglycosylation site in the CH2 domain, which was implicated in additional FcgR binding by human monocytes, through single amino acid replacement (N into Q).The resulting modi ed ROR1-CAR maintained both target-speci c cytolytic activity in vitro as well as persistent anti-tumor functionality in an in vivo xenograft model (20).We took advantage of the highly conserved amino acid sequences among the different classes of human IgG CH2 domains and modi ed the Hu8F4-CAR's IgG1(CH2-CH3) spacer domain by replacing the known FcgR-binding site in IgG1 CH2 (ELLP) with the IgG2 CH2 sequence PVA, which reduced binding to FcgRI by 10 4 -fold (22).Following the strategy of others in IgG4 spacer domains, we also eliminated the N-glycosylation site, which is conserved between IgG1 and IgG4 (20).
The redesigned Hu8F4-CAR(PQ) construct maintains full cytotoxic activity in vitro against cell lines expressing PR1/HLA-A2 and is consistently and highly transduced into both CD4 + and CD8 + T cells.Interestingly, Hu8F4-CAR-transduced T cells derived from PBMCs show a preference towards memory phenotypes while non-transduced control T cells seem to show a diversity in T cell phenotypes, primarily naïve and terminally differentiated effector.Contrary to the original, unmodi ed Hu8F4-CAR T cells, which failed to sustain anti-leukemic activity in xenograft mice (data not shown), the Hu8F4-CAR(PQ)-T cells exhibited robust anti-leukemic activity in vivo at multiple dose levels.Notably high doses of Hu8F4-CAR(PQ)-T cells (10 million) could eliminate established leukemia (U937-A2 + ) completely.Decreased tumor burden correlated with the presence of Hu8F4-CAR(PQ)-transduced T cells in mouse peripheral blood taken during the early time points after the injection, suggesting that the modi cations made in the IgG1 spacer region led to a more sustainable CAR-T cell population.In vivo anti-tumor function of the modi ed Hu8F4-CAR(PQ)-T cells was evident not only with adult PBMC but also with CBMC, providing alternative cell source options for use in clinical settings.
In this study, we have shown that CAR-T cells using human IgG1-based spacer can be modi ed to eliminate Fc receptor-binding-induced adverse effects on its survival and functionality in vivo.This paves the way for using such CAR-T cell products in future clinical trials against cancers.Furthermore, Hu8F4-CAR(PQ)-T cells demonstrate consistent high transduction e ciency into both PBMC-and CBMC-derived T cells and are active against PR1-expressing leukemic targets in vitro and in vivo.

Figure 3 Human
Figure 3

Figure 4 Adult
Figure 4