Engineering CAR-T cells for radiohapten capture in imaging and radioimmunotherapy applications

Rationale: The in vivo dynamics of CAR-T cells remain incompletely understood. Novel methods are urgently needed to longitudinally monitor transferred cells non-invasively for biodistribution, functionality, proliferation, and persistence in vivo and for improving their cytotoxic potency in case of treatment failure. Methods: Here we engineered CD19 CAR-T cells (“Thor”-cells) to express a membrane-bound scFv, huC825, that binds DOTA-haptens with picomolar affinity suitable for labeling with imaging or therapeutic radionuclides. We assess its versatile utility for serial tracking studies with PET and delivery of α-radionuclides to enhance anti-tumor killing efficacy in sub-optimal adoptive cell transfer in vivo using Thor-cells in lymphoma models. Results: We show that this reporter gene/probe platform enables repeated, sensitive, and specific assessment of the infused Thor-cells in the whole-body using PET/CT imaging with exceptionally high contrast. The uptake on PET correlates with the Thor-cells on a cellular and functional level. Furthermore, we report the ability of Thor-cells to accumulate cytotoxic alpha-emitting radionuclides preferentially at tumor sites, thus increasing therapeutic potency. Conclusion: Thor-cells are a new theranostic agent that may provide crucial information for better and safer clinical protocols of adoptive T cell therapies, as well as accelerated development strategies.


Assembly of huC825 Reporter Gene
A geneblock encoding the membrane bound huC825s was purchased (Integrated DNA Technologies).The geneblock consists of the human CD4 endoplasmic reticulum signal sequence, the scFv huC825, and a CD4 transmembrane domain.Gibson assembly was used to clone the huC825s sequence into an SFG retroviral vector harboring the DNA sequence for the anti-CD19 19BBz CAR (huC825s-19BBz).The IgG4-CH2CH3 spacer domain with mutated Fc regions (hinge) was added to the geneblock using Gibson assembly between the scFv huC825 and the CD4 transmembrane domain; this sequence was also cloned into the 19BBz viral vector to generate huC825-19BBz.huC825-GFP was also generated by Gibson assembly of the huC825 sequence into a retroviral vector encoding eGFP.19BBz SFG retrovirus was generously provided by the Renier Brentjens lab (Memorial Sloan Kettering Cancer Center/MSK).Retroviral sequences were verified by sequence analysis (Eton Bioscience).

Cell Culture
All cell lines and human PBMC's were grown in RPMI-1640 medium (MSK Center Media Core).
All media was supplemented with 10% fetal bovine serum, 2 mMol L-glutamine, 100 IU/mL penicillin and 100 μg/mL streptomycin.Raji tumor cells were purchased from ATCC and transduced to express eGFP and firefly luciferase (Raji fluc/eGFP cells) using viral supernatant from Galv9 cells producing SFG retrovirus encoding the DNA for each of these proteins.All cell lines were regularly tested to ensure they were free of mycoplasma contamination.Separations were run on a 4.6 × 250 mm Gemini-NX C18 column (Phenomenex Inc., Torrance, CA).HPLC conditions were: mobile phase A, 10 mM pH 5 NH4OAc; B, CH 3 CN; 1.0 mL/min flow rate; λ = 254 nm; injection volume 5-20 μL; gradient 0% B to 40% B over 10 min.Samples of free radiometals, reaction mixtures, and purified products were diluted 1:5 in 5 mM DTPA or EDTA prior to analysis.With the exception of unsubstituted DOTA, multiple peaks may be observed due to diastereomeric / helical isomerism of radiometallated DOTA species, these isomers bind to the anti-[M]DOTA radiohapten capture platform (C825).

Radiosynthesis of [ 111 In]In-Pr
No-carrier-added [ 111 In]InCl 3 (47.4MBq/1.28 mCi) in 50 μL of 0.05 M HCl (Nuclear Diagnostic Products Inc., Plainview, NY) was transferred to a metal-free 0.5 mL microcentrifuge tube, diluted with 50 μL of metal-free 0.5 M NH 4 OAc (pH 5.3), and mixed gently.To this was added 2.6 nmol of proteus-DOTA (Pr; DO3A-PEG 4 -[ nat Lu]Lu-DOTA-Bn; 2.6μL of 1mM solution in water) mixed gently by pipetting and placed in a heat block at 80 °C for 60 min.After cooling for 5 min, the entirety was gravity-loaded on a 30 mg Strata-X SPE cartridge (Phenomenex, Torrance, CA) that had been equilibrated with 1 mL of ethanol and 1 mL of water.Water (100 μL) was used to rinse the reaction tube and passed through the cartridge.The column was washed slowly dropwise with 200 μL of water and gently blown dry with nitrogen gas.The product was slowly eluted dropwise with 200 μL of ethanol, diluted as necessary in normal saline (Hospira, Lake Forest, IL) and sterile filtered to obtain "[ 111 In]In-Pr" [ 111 In]InDO3A-PEG 4 -[ nat Lu]LuDOTA-Bn (40.7 MBq, 1.10 mCi, 86% ndc yield, A M = 18.5 MBq/nmol, 500 μCi/nmol) at time of synthesis.Radio-HPLC of crude and purified material confirmed that no free radiometal remained (major isomer t R = 8.0 min, 97.5% radiochemical purity).

Radiosynthesis of [ 225 Ac]Ac-Pr
Carrier-free 225 Ac (2.146 x 10 6 GBq/g [5.80 x 10 4 Ci/g]) was obtained from NIDC, Oak Ridge National Laboratory (Oak Ridge, TN) as a dried nitrate residue.The 225 Ac nitrate was dissolved in 0.1 M Optima grade hydrochloric acid for subsequent radiochemistry.Ac-225 activity measurements were made at secular equilibrium prior to labeling using a CRC-55tR radioisotope calibrator (Capintec Inc., Florham Park, NJ) set at 775, and the displayed activity value multiplied by 5; samples were positioned at the bottom and center of the well for measurement.Reactions were typically run at 10-200 µCi scale; a representative process is as follows: [ 225 Ac]AcNO 3 in 0.1 M HCl (20 μL, 2.442 MBq / 66.0 μCi) was neutralized with 100 μL of 3 M ammonium acetate solution and 15 μL of L-ascorbic acid (150 g/L) in a glass reaction vial.
The reaction was incubated at 60 °C for 30 min and then purified using a CM Sephadex C-25 resin (Sigma-Aldrich) column pre-equilibrated with 6 mL of normal sterile isotonic saline solution (NSS).The reaction mixture was added to the column and eluted with 4 mL of NSS to recover all removable activity ([ 225 Ac]Ac-Pr); percentage of total activity eluted was used to determine yield and specific molar activity A M .After formulation and sterile filtration, 2.29 MBq (62.0 μCi) of [ 225 Ac]Ac-Pr was obtained with a non-decay corrected radiochemical yield of 94% and A M = 3.1 MBq/μmol (84 μCi/μmol).Good (>85%) yields were obtained up to A M ≈ 120 MBq/µmol (3 µCi/nmol).Optimally, in vivo doses were formulated at 185-370 kBq/mL (5-10 µCi/mL) in saline for a 200 µL injection volume.Doses for mice were typically <1 nmol and ~40-80 kBq (1-2 µCi) as specified elsewhere.

In vivo Experiments
All animal experiments were performed under a protocol approved by the Memorial Sloan Kettering Cancer Center Institutional Animal Care and Use Committee.For in vivo confirmation of huC825-19BBz anti-tumor activity and binding capacity for radiotracer, 5x10 5 Raji cells were implanted intraperitoneally (IP); two days later 5x10 5 T cells were also injected IP.On day 6, [ 86 Y]Y-ABD (3.7 MBq) was injected IP and radiotracer uptake was measured using PET/CT imaging (see below for specifics).Tumor burden was measured through detection of intravenously (IV) injected luciferin (GoldBio, 50 µL of a 20 mg/ml solution) and detected using an IVIS Spectrum In Vivo Imaging System (PerkinElmer).Total bioluminescent signal (BLI) was normalized to the measurement taken on day 2 and displayed as fold change in BLI.BLI measurements were taken every five days until mice reached endpoint.
To evaluate the sensitivity of the huC825-19BBz in vivo, CAR T-cells were stimulated with OKT3 in vitro overnight prior to implant.1x10 4 , 3x10 4 , 1x10 5  A modified version of the Raji SC model was utilized to characterize the huC825-19BBz CAR-T cells on a tissue level.This model followed the same procedure as above with one exception: T cells were injected fourteen days after tumor engraftment.[ 86 Y]Y-ABD was injected on day 7 and 14 post T cell injection and measured as above.

Ex-vivo Flow Cytometry Time Course
NSG mice injected with huC825-19BBz CAR-T cells were sacrificed on d 4, d 7, d 11, and d 14 post T cell injection using the previously described Raji SC model.Tumors were collected and analyzed by flow cytometry to determine the ratio of T cells to tumor cells overtime.huC825-19BBz CAR-T cells were identified as CD3 + , CD45 + , CAR + using the same antibodies as detailed in the main text.Raji tumor cells were identified as GFP + and CD20 + through use of an anti-CD20 Alexa647 antibody (BioLegend, 302318).

Image-Based Biodistribution
Volumes of interest (VOIs) were defined using a combination of manual and semi-automatic segmentation techniques (VivoQuant or 3D Slicer v4.10); CT anatomical guidance was used to segment the blood/heart contents, lungs, liver, stomach contents, intestine contents, kidneys, skeleton, muscle, and tumor tissue.PET guidance was used to determine activity within the urinary bladder. .The organ-level level time-integrated activity coefficients were assumed, in first order, to be independent of species, and were input into MIRDcalc dosimetry software [3] to obtain absorbed dose estimates for humans using the ICRP adult reference phantoms.

Autoradiography and Immunohistochemical Staining
Tissues were frozen in OCT (Sakura Finetek USA, Inc.) immediately after collection, sectioned at 5 microns thickness, mounted on slides and fixed in 10% neutral buffered formalin (NBF) prior to staining.Other tissues were fixed in 10% NBF, processed in alcohol and xylene, embedded in paraffin, sectioned at 5 microns thickness and mounted on slides prior to staining.
Immunohistochemistry was performed on a Leica Bond RX automated stainer using Bond reagents (Leica Biosystems, Buffalo Grove, IL), including a polymer detection system (DS9800, Novocastra Bond Polymer Refine Detection, Leica Biosystems, Buffalo Grove, IL).The chromogen was 3,3 diaminobenzidine tetrachloride (DAB), and sections were counterstained with hematoxylin.After heat-induced epitope retrieval in a citrate buffer (performed for formalin-fixed paraffin embedded sections only), staining was performed with a primary antibody against CD3 (clone SP162, Abcam ab135372) applied at a concentration of 1:200.Adjacent sections were hematoxylin-and eosin-stained for morphologic evaluation.Stained slides were evaluated by a board-certified veterinary pathologist (Laboratory of Comparative Pathology, MSKCC).
For the enumeration of T cells in selected tissue and tumor sections, CD3 + cell density maps were created with a diameter of 0.5 µm per cell nucleus using the software Qupath v0.3.2.
To get an estimation of the total number in a tissue, the average of 10 sections was extrapolated to the full tumor volume.
Radiolabeling of [ 86 Y]Y-ABD, [ 111 In]In-Pr, and [ 225 Ac]Ac-Pr Radiochemistry was performed in appropriately shielded chemical fume hoods equipped with electronic flow monitoring and sliding leaded glass windows.A CRC-55tR dose calibrator was used to measure radioactivity using manufacturer-recommended calibration settings (Capintec Inc., Florham Park,NJ).Buffers and water used for radiochemical synthesis were treated with 5% (w/v) Chelex ion exchange resin (BT Chelex 100 Resin, Bio-Rad Inc., Hercules, CA) to remove adventitious heavy metals, or were tracemetal grade.Plasticware (such as pipette tips and microcentrifuge tubes) was trace metal grade or PCR grade.Radio-HPLC was performed on a Shimadzu Prominence HPLC system (Shimazdu Scientific Instruments, Somerset, NJ) comprised of an LC-20AB dual pump module, DGU-20A3R degasser, SIL-20ACHT autosampler, SPD-20A UV-Vis detector and a Bioscan Flow-Count B-FC-1000 with PMT/NaI radioactivity detector in-line.
and 3x10 5 huC825-19BBz cells were implanted subcutaneously over the left and right shoulder in 40 µL of 50:50 RPMI:Matrigel solution.[ 86 Y]Y-ABD (3.7 MBq) was injected via tail vein and the radiotracer uptake was measured 3 hours post tracer injection by PET/CT.
Image co-registration, segmentation, and Monte Carlo dosimetry simulation.Dosimetry estimates were based on co-registered [ 86 Y]Y-ABD serial PET/CT images obtained at 14 d post T cell administration.The [ 86 Y]Y-ABD PET images were originally quantified in units of %IA/mL, decay corrected to time of administration (as-described previously) and therefore represent biological clearance over time.The CT scan acquired at the 16 h time point was used as the "fixed" volume to which the PET/CT images at other time points (1, 3, or 36 h) were co-registered.Tumor, organsof-interest, and residual tissues were contoured using a combination of manual and semiautomatic segmentation techniques in 3D Slicer image analysis software.After manual alignment of the CT images, rigid registration (3 rotational and 3 translational degrees of freedom) was performed, followed by B-spline deformable registration using the Plastimatch extension of 3D Slicer.The deformable registration step used a mean squared error cost function, a 2x2x2 image subsampling rate, a 7.5 mm grid size, and was performed in a single stage.After successful co-registration of the CT images with the 16 h CT, the transforms were applied to the respective PET images.The PET images were then resampled to the same image dimensions, spacing, and origin.The resampled [ 86 Y]Y-ABD PET images were then used forecast the effective activity concentration (i.e., representing biological clearance and physical decay) of therapeutic analogs -[ 177 Lu]Lu-ABD, [ 90 Y]Y-ABD, or [ 225 Ac]Ac-Pr -under the assumption that biological clearance kinetics are equivalent among these variants, viz: is the decay-corrected activity concentration of [ 86 Y]Y-ABD in the voxel indexed by i, j, and k; λ I is the physical decay constant for radionuclide I (i.e., [ 177 Lu]Lu-ABD, [ 90 Y]Y-ABD, or [ 225 Ac]Ac-Pr); t is the time post-administration; and, is the forecasted activity concentration of radionuclide I.A time-integrated activity coefficient image (in units of Bq*s/Bq*mL) was obtained for each radionuclide as follows: the �� was voxelwise integrated in time via the trapezoidal method up to the last measured time point, after which clearance was assumed to occur by physical decay only.The �� at the first measured time point.The time-integrated images were used as the voxel source term in the PHITS Monte Carlo simulations used to compute the equivalent dose coefficients for [ 177 Lu]Lu-ABD or [ 90 Y]Y-ABD.The simulation geometry was defined by the segmentation described above.Where available, each segment was assigned material composition and density defined by the International Commission on Radiological Protection (ICRP) Publication 89; in case a specific tissue was unavailable in the ICRP tables, the elemental composition and density of soft tissue was used.For [ 225 Ac]Ac-Pr, a Monte Carlo simulation was not run; the equivalent dose coefficients [Sv/mCi administered] were computed by assuming local energy deposition from radiations produced by 225 Ac and all radioactive progenies.In all cases, a radiation weighting factor of unity for beta particles/photons was used for weighting the absorbed doses to obtain equivalent dose coefficients [Sv/mCi administered].For alpha particles, a radiation weighting factor of 5 was used (which reflects the relative biological effectiveness of alpha particles for deterministic endpoints).Estimated dosimetry for human administration.Murine organ-level time integrated activity coefficients [Bq*s/Bq] were derived as the product of the corresponding segment volumes and the mean time-integrated activity concentration, normalized to the administered activity (vide supra)

TABLE S1 . Specific and non-specific binding of huC825 expressing T cells with [ 111 In]In- Pr
. B max , sites/cell; K d , equilibrium dissociation constant, [nM].Representative data shown.Experiments were performed in triplicate at 37 °C.n = 3 donors.

TABLE S2 . Quantification of CD3 + T-cells in selected tissues.
Tissues of selected organs and tumor tissue from Figure3A.Absolute number of CD3 + cells were counted on QuPath v 0.3.0 and absolute number of CD3 + cells per mm 3 were calculated, leading to an estimation of the total amount of CD3 + cells.

TABLE S3 .
Projected murine [ 86 Y]Y-ABD doses.Absorbed dose per organ or tissue was calculated using the dosimetry described in Figure5.