Fully human antibody VH domains to generate mono and bispecific CAR to target solid tumors

Background Chimeric antigen receptor (CAR) T cells are effective in B-cell malignancies. However, heterogeneous antigen expression and antigen loss remain important limitations of targeted immunotherapy in solid tumors. Therefore, targeting multiple tumor-associated antigens simultaneously is expected to improve the outcome of CAR-T cell therapies. Due to the instability of single-chain variable fragments, it remains challenging to develop the simultaneous targeting of multiple antigens using traditional single-chain fragment variable (scFv)-based CARs. Methods We used Humabody VH domains derived from a transgenic mouse to obtain fully human prostate-specific membrane antigen (PSMA) VH and mesothelin (MSLN) VH sequences and redirect T cell with VH based-CAR. The antitumor activity and mode of action of PSMA VH and MSLN VH were evaluated in vitro and in vivo compared with the traditional scFv-based CARs. Results Human VH domain-based CAR targeting PSMA and MSLN are stable and functional both in vitro and in vivo. VH modules in the bispecific format are capable of binding their specific target with similar affinity as their monovalent counterparts. Bispecific CARs generated by joining two human antibody VH domains can prevent tumor escape in tumor with heterogeneous antigen expression. Conclusions Fully human antibody VH domains can be used to generate functional CAR molecules, and redirected T cells elicit antitumoral responses in solid tumors at least as well as conventional scFv-based CARs. In addition, VH domains can be used to generate bispecific CAR-T cells to simultaneously target two different antigens expressed by tumor cells, and therefore, achieve better tumor control in solid tumors.


INTRODUCTION
Chimeric antigen receptors (CARs) typically consist of an extracellular antigen-binding domain in the form of a single-chain fragment variable (scFv), a transmembrane domain and signaling molecules such as costimulatory endodomains and CD3ζ chain. [1][2][3] Expression of CARs in T cells enables specific targeting of surface antigens in a Major Histocompatibility Complex-independent manner and associated T cell activation. 4

While classical
CARs use scFvs as the antigen-binding moiety, other ligands fused with signaling molecules of the T-cell receptor complex can also trigger phosphorylation events in T cells. 6 For example, engineering of natural receptors such as NKG2D and CD27 fused with CD3ζ have been shown to redirect T cell specificity. 7 8 Ligands to receptors such as interleukin (IL)-13Rα2 have also been engineered to redirect T cell specificity towards glioblastoma. 9 More recently, synthetic antigen binding moieties as exemplified by a 'monobody' based on the type III domain of fibronectin have also been shown to serve as a robust platform to generate CAR molecules. 10 Therefore, using alternative binding moieties to replace scFvs to generate CAR remains a critical area because scFvs are frequently unstable and showing intrinsic tendency to self-aggregation, which may lead to tonic signaling and loss of function of CAR-T cells in vivo. 11 12 A scFv molecule is composed of paired antibody (Ab) light chain and heavy chain variable domains (V L and V H ) that are fused into a single polypeptide chain via a short flexible linker. 11 13 14 Heavy-chain-only Abs without light chains have been reported in camelids and cartilaginous fish, [15][16][17] and shown to exhibit strong and specific antigen binding. 13 Single domains targeting BCMA (B-cell maturation antigen) have been developed to generate BCMA-specific CAR-T cells for the treatment of multiple myeloma. 18 Whether human-derived V H -only domains can be used as a CAR to target antigens expressed in solid tumors is unknown.
Treatment failure and/or disease recurrence after CAR-T cell therapy can be caused by epitope or antigen loss. 10 In particular, the inherently heterogeneous expression pattern of antigens in solid tumors can easily cause tumor escape after targeted Open access immunotherapy. 10 19 20 Therefore, targeting multiple tumor-associated antigens (TAAs) is generally expected to improve the outcome of CAR-T cell therapy in solid tumor. 10 19 However, including multiple scFvs within a CAR causes protein instability and decreases binding specificity and affinity. V H domain-only format of CARs provide an ideal solution for multiple antigen targeting because V H domains have smaller size and may easily fold correct 3D structure compared to scFv molecules.
Here, we explored the use of Humabody V H domains derived from a transgenic mouse to develop CARs that target prostate-specific membrane antigen (PSMA) 21 22 and mesothelin (MSLN). 23 We found that Humabodybased CARs exhibited comparable or superior antitumor activity compared with traditional scFv CARs. Moreover, we demonstrated that Humabodies were suitable for constructing bispecific CAR-T cells, which can significantly better control tumors with heterogeneous antigen expression.

MATERIALS AND METHODS Generation of V H domains
Crescendo Mouse 17 was immunized with PSMA and MSLN recombinant proteins. Spleens and lymph nodes were harvested, cloned into a phagemid vector and selected by phage display. Outputs were screened for specific target binding and further characterized.

CAR construction
The following antigen-binding moieties were used: scFv derived from the J591 Ab specific for PSMA; human V H domain specific for PSMA (PSMA-V H ); scFv derived from a MSLN-specific Ab Amatuximab; human V H domain specific for MSLN (MSLN-V H ). All ligands were assembled with the CD8α hinge and transmembrane domain, the CD28 costimulatory domain and CD3ζ intracellular signaling domain and cloned into the SFG retroviral vector. 24 A FLAG-tag was incorporated after the antigen ligand to detect the expression of CARs by an anti-FLAG Ab. Dual specific (PSMA and MSLN) CARs were also generated by linking the two V H domains. The corresponding CARs were called J591, PSMA-V H , MSLN scFv, MSLN-V H and PSMA-V H /MSLN-V H . Retroviral supernatants were produced by transfection of 293 T cells with the retroviral vectors, the RD114 envelope from RDF plasmid and the MoMLV gag-pol from PegPam3-e plasmid. Supernatants were collected 48 hours and 72 hours after the transfection and filtered with 0.45 µm filter. 24 Cell lines Tumor cell lines PC-3, C4-2 (prostate cancer) and Aspc-1 (pancreatic cancer) were purchased from ATCC (American Type Culture Collection). All tumor cell lines were cultured with RPMI-1640 (Gibco) supplemented with 10% Fetal bovine serum (Sigma), 2 mM GlutaMax (Gibco) and penicillin (100 units/mL) and streptomycin (100 µg/mL; Gibco). All cells were cultured at 37°C with 5% CO 2 . PC-3 cell line was transduced with retroviral vectors encoding PSMA or MSLN to make PC- 3-PSMA  and PC-3-MSLN. PC-3-PSMA, PC-3-MSLN and Aspc-1 were transduced with retroviral vectors encoding Firefly-Luciferase-eGFP (FFluc-eGFP) gene.
Western blot CAR-T cells were incubated with 2 µg anti-FLAG Ab in 100 µL PBS for 20 mins on ice and then with 2 µg goat antimouse secondary Ab for another 20 mins on ice. Cells were then incubated in the 37°C water bath for the selected time points and then lysed with 2 x Laemelli buffer for 10 mins. Cell lysates were then separated in 4% to 15% 10 well SDS-PAGE gels and transferred to polyvinylidene difluoride membranes at 75V for 120 mins (Bio-Rad). Blots were examined for human CD3ζ (Santa Cruz Biotechnology), p-Y142 CD3ζ (Abcam), pan-ERK (BD Biosciences), and pan-Akt, p-S473 Akt, and p-T202/ Y204 MAPK (Cell Signaling Technology) with 1:1000 dilution in 5% TBS-Tween milk. Membranes were incubated with HRP-conjugated secondary goat anti-mouse or goat anti-rabbit IgG (Santa Cruz) at a dilution of 1:3000 and imaged with the ECL substrate kit (Thermofisher) on the ChemiDoc MP System (Bio-Rad) according to the manufacturer's instructions. 26

Open access
Proliferation assay T cells were labeled with 1.5 mM carboxyfluorescein diacetate succinimidyl ester (CFSE; Invitrogen) and plated with tumor cells at an effector to target (E:T) ratio of 1:1. CFSE signal dilution from gated T cells on day 5 was measured using flow cytometry. 26 In vitro cytotoxicity assay Tumor cells were seeded in 24-well plates at a concentration of 2.5×10 5 cells/well overnight. CAR-T cells were added to the plate at an E:T of 1:5 without exogenous cytokines. Cocultures were analyzed 5-7 days following coculture to measure residual tumor cells and T cells by flow cytometry. Dead cells were recognized by Zombie Aqua Dye (Biolegend) staining while CAR-T cells were identified by CD3 staining and tumor cells by GFP. 26 CD69, PD-1 and Lag3 expression was measured by flow cytometry from day 0 to day 5 each day after coculture of CAR-T cells with tumor cells. For the granzyme-B staining, Golgi protein inhibitor (BD Biosciences) was added on day 1 of coculture for 6 hours. Cocultures were then first stained with Zombie Aqua Dye (Biolegend) and CD3 mAb, followed by fixation/permeabilization solution (BD Biosciences). Intracellular staining of granzyme-B was then conducted.
Cytokine analysis CAR-T cells (1×10 5 cells) were cocultured with 2.5×10 5 tumor cells in 24-well plates without exogenous cytokines. Supernatant was collected after 24 hours, and cytokines (interferon-γ (IFN-γ) and IL-2) were measured by using ELISA kits (R&D, Research And Development system) in duplicates following manufacturer's instructions. 26 Expression and purification of recombinant proteins A panel of recombinant proteins was produced, comprizing bispecific (2V H ) proteins that bind both PSMA and MSLN, monospecific V H protein binding PSMA, monospecific V H protein binding MSLN and a control scFv protein based on Amatuximab. Bispecific protein was made in two formats, one with a short flexible linker (G4S) 3 , aother one with a long flexible linker (G4S) 6 . Bispecific proteins were expressed in mammalian cells and purified by protein A binding. Monospecific proteins were His tagged at the C terminus, expressed in Escherichia coli and purified by His trap and size exclusion chromatography.
Binding and kinetic analysis Binding analyses were performed at 25°C using BIAcore 8K system. The instrument was run on 1 x HBS-EP + (BR100669) buffer and the data were analyzed using Biacore Insight Evaluation software. Recombinant human MSLN was diluted to 2 ug/mL in 10 mM sodium acetate buffer pH4.0 and immobilized on a CM5 sensor chip (contact time 120 s) using amine-coupling kit with accordance to the manufacturer's instructions. Humabody V H samples were tested for binding at 5 concentrations 3.7 nM, 11.1 nM, 33.3 nM, 100 nM and 300 nM using multicycle kinetics method. Each sample was injected for 100 s at the flow rate 35 µL/min and dissociated for 100 s. The antigen surface was regenerated by 20 s injection of 10 mM glycine pH 2.0. Recombinant human PSMA antigen with a human Fc tag was captured on a Protein G sensor. Humabody V H samples were tested in Single-cycle kinetics mode at increasing concentrations of 2.22 nM, 6.67 nM, 20 nM and 60 nM with 90 s association and 600 s dissociation time at the flow rate of 30 µL/min. Buffer injections were made to allow for double-reference subtraction. The sensor surface was regenerated with 10 mM glycine pH1.5 (GE Healthcare BR100354).To detect dual binding to MSLN and PSMA, human PSMA antigen surface was captured as above. Bispecific PSMA-MSLN Humabody constructs were captured on the PSMA surface by injecting 100 nM of each sample for 100 s at 35 µL/min flow rate. The capture was immediately followed by an injection of 300 nM recombinant human MSLN with 100 s contact time and 100 s dissociation. A PSMA-specific Humabody construct without a MSLNbinding arm was used as a control.
Xenograft murine models NSG (NOD scid gamma mouse) mice (6-8 weeks old) were injected intravenously through tail vein with either PC-3-PSMA-FFluc-eGFP, or PC-3-PSMA-FFluc-eGFP and PC-3-MSLN-FFluc-eGFP mixed at 1 to 1 ratio, or Aspc-1-FFluc-eGFP tumor cells of 1×10 6 cells per mice. Fourteen days later, CAR-T cells were injected intravenously through tail vein. For the high dose treatment, 4×10 6 CAR-T cells per mice were injected, while for the low dose treatment, 1×10 6 CAR-T cells per mice were injected. In the rechallenge experiments, mice were infused 1×10 6 tumor cells per mice on clearance of the previous tumor. Tumor growth was monitored by bioluminescence using IVIS (In Vivo Imaging Systems)-Kinetics Optical in vivo imaging system (PerkinElmer) (PSMA-V H and MSLN-V H part) or AMI(AMI Medical Imaging) Optical in vivo imaging system (Spectral instruments imaging) (PSMA-V H /MSLN-V H part).

Statistics
All data was calculated and represented as mean with SD. One-way analysis of variance (ANOVA) or two-way ANOVA analyses were performed to compare multiple groups. Two-tailed t-test was used to compare two groups. P value of less than 0.05 was significant. All calculations and figures were achieved by GraphPad Prism V.7 (La Jolla, California, USA).

Human V H domain-based CAR targeting PSMA is expressed and signals in T cells
We constructed the PSMA-specific CARs using the scFv from the J591 mAb (J591) and the PSMA binding human V H domain (PSMA-V H ) joined to the CD8α stalk, CD28 costimulatory domain and CD3ζ intracellular domain. A flag-based tag was incorporated into the cassettes to detect CAR expression by flow cytometry (figure 1A). Activated T cells were successfully transduced and expressed the CARs equally ( figure 1B,C). The CD19specific CAR (CD19) and non-transduced (NT) T cells were used as controls. On transduction, J591-T cells and PSMA-V H -T cells showed similar expansion in vitro when exposed to IL-15 and IL-7 cytokines, which was similar to CD19-T cells and NT-T cells ( figure 1D). Furthermore, no differences were observed in T cell composition as assessed by flow cytometry at day 12-14 of culture (figure 1E). We examined proximal signaling of CAR-T cells before and after CAR cross-linking mediated by an anti-Flag Ab. Phosphorylation of the CAR-associated CD3ζ as well as phosphorylation of Akt and ERK were equal in J591-T cells and PSMA-V H -T cells ( figure 1F). Therefore, a V H domain-based CAR is expressed and signals in T cells on cross-linking as observed for scFvbased CAR-T cells.

Open access
To further assess differences between PSMA-V H -T cells and J591-T cells, we used low doses of T cells (1×10 6 cells/ mouse) in tumor-bearing mice (figure 3D). We observed that PSMA-V H -T cells still eliminated tumor cells in vivo as J591-T cells ( figure 3E,F). In addition, we also observed similar V H CAR-T cell persistence in the peripheral blood, spleen and bone marrow compared with traditional scFvbased CAR-T cells at day 58 at the time of euthanasia ( figure 3G,H). Therefore, Humabody V H CAR-T cells demonstrated comparable antitumor effects to scFv-based CAR-T cells in vitro and in vivo.

MSLN-specific V H domain-based CAR-T cells demonstrate antitumor activity
To further assess the reproducibility of V H domain-based CARs, we tested a MSLN-specific Humabody V H . We

In vitro analysis of monovalent and bivalent V H domain recombinant proteins
To test whether the V H domains are suitable to construct bispecific CARs, two V H domains in tandem recombinant proteins linking PSMA-specific and MSLN-specific V H were generated ( figure 5A). To test whether the linkers had any effect on the target binding affinity, two different linkers were used: the (G4S) 3 linker ('short flexible linker') and a longer linker (G4S) 6   using either flexible linkers ( figure 5B). Similarly, analysis of binding to MSLN recombinant protein by SPR Biacore assay showed that the affinity of the MSLN-V H domain was not altered when the PSMA-V H was formatted with the MSLN-V H using either flexible linkers (figure 5C). In summary, these data show that V H modules in bispecific format are capable of binding their specific target with the same affinity as their monovalent counterparts.

Bispecific V H domain-based CAR-T cells demonstrate dual specificity
We constructed a bispecific V H domain CAR to facilitate CAR-T cells to specifically recognize two antigens simultaneously. We used the MSLN-V H and PSMA-V H domains fused with the short (G4S) 3

DISCUSSION
CARs approved by the Food and Drug Administration and those in clinical studies are mostly based on scFvbinding moieties. Here we demonstrated that monospecific human V H domain-based CAR-T cells achieved comparable antitumor effects both in vitro and in vivo as scFv-based CAR-T cells. Furthermore, V H domains combined in tandem to create bispecific molecules allowed the generation of effective CAR-T cells targeting two antigens.
Redirected T cell based on single-domain Abs have been recently proposed. 17 28 29 However, most of them are obtained from llamas or camelid-derived libraries. Biological therapeutic molecules with non-human sequence can cause immune responses. 18 28 Transgenic mouse technology has enabled the generation of biophysically robust fully human V H domains known as Humabody V H or Humabodies 30 which have the potential for use in CAR constructs while mitigating immunogenicity risk.
Despite the remarkable clinical activity of CAR-T cells in hematological malignancies, objective responses in patients with solid tumors are modest. 10 26 31-33 Heterogeneity of antigen expression is one of the main reasons causing tumor escape in solid tumors after targeted therapies. 10 19 20 Furthermore, murine-based scFv may cause immune responses especially in solid tumor patients who are usually less immunosuppressed compared with patients with liquid tumors. Targeting multiple TAAs and using human binding moiety in CAR molecules may improve the outcome of CAR-T cells in solid tumors. 10 Here, we demonstrated that human V H domains generated from a transgenic mouse might solve both issues of immunogenicity and tumor heterogeneity since bispecific CAR-T cells can be efficiently generated using two human V H domains in tandem.
In addition to the issue of heterogeneity in antigen expression, the complex inhibitory pathways of the tumor microenvironment in solid tumors mean that additional genetic modification of T cells would likely be required to enhance T cell trafficking and functions. 5 31 34-36 Generation of vector cassettes encoding multiple genes requires a significant optimization of the engineering strategies since the size of the entire cassette is limited. V H domains are a good alternative to scFv since they are approximately half the size.
Here, we have used two target antigens, PSMA and MSLN, that are currently under evaluation to treat mesothelioma, lung cancer, breast cancer, pancreatic cancer and prostate cancer via scFv-based CAR-T cells. [37][38][39] Our preclinical experiments validate the potential use of bispecific human V H domains targeting both PSMA and MSLN in these difficult to treat malignancies. It remains to be validated if dual or multiple targeting with V H domainbased CARs can be broadly applicable, and if targeting multiple antigens in solid tumors leads to increased potential for toxicity.
Additionally, we observed that V H domain-based CAR-T cells have comparable cytotoxicity and proliferative capacity as traditional scFv-based CAR-T cells. MSLN-V H -T cells showed even more profound antitumor effects as compared with mice treated with MSLN-scFv CAR-T cells. Interestingly, MSLN-V H showed lower affinity than MSLN-scFv (28 nM compared with 79pM) recapitulating what has been observed for other scFvs that very high affinity is not necessarily optimal for CAR-based targeting for some targets. [40][41][42] However, we cannot exclude that the observed superior antitumor activity of the MSLN-V Hbased CAR-T cells can be associated with the recognition of a different epitope rather than to different affinity. In summary, we have demonstrated that V H domain CAR-T cells in monospecific format achieved comparable antitumor response compared with traditional scFv-based CAR-T cells both in vitro and in vivo. Furthermore, bispecific V H domain CAR-T cells delivered potent anti-tumor effects demonstrating the potential to target solid tumors with heterogeneous antigen expression. These proofof-concept experiments lay the foundation for further development of human V H domain-based CAR-T cells in clinical trials. Competing interests GD has sponsored research agreements with Bluebird Bio, Cell Medica, and Bellicum Pharmaceutical. GD serves on the scientific advisory board of MolMed and Bellicum Pharmaceutical. CJ and BM are employees of Crescendo Biologics Ltd.
Patient consent for publication Not required.
Ethics approval The present studies in mice were approved by the Institutional Animal Care and Use Committee at the University of North Carolina at Chapel Hill, North Carolina, USA.
Provenance and peer review Not commissioned; externally peer reviewed.
Data availability statement All data relevant to the study are included in the article or uploaded as supplementary information.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.
Open access This is an open access article distributed in accordance with the Creative Commons Attribution 4.0 Unported (CC BY 4.0) license, which permits others to copy, redistribute, remix, transform and build upon this work for any purpose, provided the original work is properly cited, a link to the licence is given, and indication of whether changes were made. See https:// creativecommons. org/ licenses/ by/ 4. 0/.