Unless otherwise noted, all FR+ and FR-negative cancer cell lines were, respectively, maintained in RPMI-1640 medium (Gibco BRL) supplemented with 10% heat-inactivated fetal calf serum without (FFRPMI) or with (RPMI) 2.4 μM folic acid (FA). KB (FRα-expressing human cervical carcinoma with HeLa markers) and CHO-β (Chinese hamster ovary cells transfected with human FRβ) were used as the sources of FRα and FRβ for radioligand binding assays, respectively (18). MDA-MB-231 represents a FRα+ subclone of human TNBC cell line. For AML studies, the green fluorescent protein (GFP)-expressing isogenic pairs of FRβ-positive (THP1-FRβ) and FR-negative (THP1-FG12) cell lines were kindly provided by Dr. Manohar Ratnam (The University of Toledo, Toledo, OH). Both were established from THP-1 (ATCC, TIB-202), a commonly used cell model for researching pediatric AML which was originally derived from a 1 year-old male infant with acute monocytic leukemia. For osteosarcoma studies, HOS-FRα was established by lentiviral transduction of FR-negative HOS-143b (ATCC, CRL8303) with FOLR1 gene encoding the human FRα. HOS-143b is originally established from a primary tumor of a 13 year-old Caucasian female and highly tumorigenic in NSG mice (35). The GFP-expressing bioluminescent pairs of FR+ HOS-FRα^fLuc and FR-negative HOS-143b^fLuc were transduced with lentiviral firefly luciferase and produced in the Jensen laboratory.

LEGENDplex™ human cytokine panels were purchased from BioLegend (San Diego, CA). The lactate dehydrogenase (LDH) based CytoTox 96® non-radioactive cytotoxicity assay kit was purchased from Promega (Madison, WI). Commercially available anti-human antibodies used for multicolor flow cytometry were: CD45RA (clone HI100), CD45RO (clone UCHL1), CD4 (clone SK3), and CD69 (clone FN50) from Thermo Fisher Scientific (Waltham, MA); CD3ε (clone SK7), CD8α (clone RPA-T8), CD137/4-1BB (clone 4B4-1), CD25 (clone M-A251), PD1 (clone EH12.1), LAG3 (clone T47-530), and TIM3 (clone 7D3) from BD Bioscience (San Jose, CA); biotinylated anti-human EGFR (Cetuximab, clone Hu1) from R&D systems (Minneapolis, MN); and FRα (clone LK26) from BioLegend (San Diego, CA). A fluorophore-conjugated anti-biotin was also purchased from BioLegend. APC-conjugated anti-FITC mouse IgG2a/kappa antibody (clone NAWESLEE), CountBright™ beads (Invitrogen), Annexin V staining buffer, and AlexaFluor-647-conjugated Annexin V were purchased from Thermo Fisher Scientific. For enzymatic digestion of tumor tissues, collagenase IV, hyaluronidase and DNase I were all purchased from Sigma-Aldrich (St. Louis, MO).

EC17 or folate-FITC [FA-(γ)-ethylenediamine-FITC] was synthesized at Endocyte. ³H-EC17 was either purchased from Moravek Biochemical (Brea, CA) at a specific activity of ~0.952 Ci/mmol or prepared at Endocyte by conjugating FITC with ³H-FA-(γ)-ethylenediamine made by ViTrax (Placentia, CA) at a specific activity of ~1.2 Ci/mmol. ³H-FA was also purchased from ViTrax at a specific activity of 59 Ci/mmol. For CRS rescue, sodium fluorescein dosing solution was diluted from AK-FLUOR® 25% (fluorescein injection, USP) which was purchased from Purdue Pharmacy (NDC 17478-250-25).

Developed by the Jensen team, a FITC-specific CAR construct was comprised of (1) a fully human anti-FITC scFv (clone E2, Kd = 0.75 nM), (2) an IgG4 hinge-CH2(L235D, N297Q)-CH3 spacer fused to a CD28-transmembrane domain (36), (3) a second-generation 4-1BB/CD3ζ-endodomain, and (4) a cell-surface human EGFRt tag, as published previously (37). To generate CAR-modified T cells, lentivirus was produced in 293T cells co-transfected with CAR-encoding epHIV7 lentiviral vector. Donor CD4+ and CD8+ T cells were purified by immunomagnetic selection and transduced separately or at a 50:50 ratio. In general, only one round of CD3/CD28 bead activation followed by one or two rounds of rapid in vitro expansion were carried out. For preclinical evaluations, several batches of EGFRt-sorted CD4, CD8 and unsorted CD4/CD8 CAR-T cells were used. All CAR-T cell preparations were analyzed prior to cryopreservation and after thawing to determine EGFRt expression and CD4/CD8 ratios by flow cytometry. Using combinations of surface markers, differentiation status of CD4+ and CD8+ CAR-T cell subsets on the day of infusion was analyzed and defined as T_N, CD45RA+ CD45RO– CD62L+ CD95– naïve T cells; T_SCM, CD45RA+ CD45RO– CD62L+ CD95+ stem cell memory T cells; T_CM, CD45RA– CD45RO+ CD62L+ CD95+ central memory T cells; T_EM, CD45RA– CD45RO+ CD62L– CD95+ effector memory cells; and T_EFF, CD45RA+ CD45RO– CD62L– CD95+ effector T cells. For preclinical testing described below, we employed two batches of EGFRt-sorted pure CD4 and CD8 subsets (after mixing at 1:1 ratios) and two batches of unsorted ~1:1 EGFRt+ CD4/CD8 admixture including a “clinical facsimile” preparation with low differentiation profiles.

The bispecific affinities of EC17 CAM were assessed using ³H-EC17 in cell-based radioligand binding assays. For binding to FR+ targets, KB and CHO-β cells were pre-seeded overnight in 24-well tissue culture plates and incubated with 0.1, 0.5, 1, 5, 10, 20, and 40 nM of ³H-EC17 in FFRPMI for 2 h at 37°C. Afterwards, the cells were rinsed with phosphate-buffered saline (PBS, pH 7.4) and lysed with 1% sodium dodecyl sulfate. The whole cell lysates were quantitated for the level of radioactivity and cellular protein content by standard Pierce BCA protein assay. The number of ³H-EC17 molecules bound per cell was calculated to determine the dissociation constants (Kd) for FRα (KB) and FRβ (CHO-β), respectively. To determine the Kd value of ³H-EC17 to effector cells, two million E2-CAR-T cells (100% EGFRt+, ~1:1 CD8/CD4 ratio) were incubated (1.5 h, room temperature) with 2-fold serial dilutions of ³H-EC17 (0–1,000 nM) with or without 500x excess nonradiolabeled “cold” EC17 in 150 μL of T cell culture medium [TexMACS™ medium (Miltenyi biotec, Auburn, CA) supplemented with 2% heat-inactivated human AB serum]. Afterwards, the cells were washed 4x with PBS, lysed with 0.1% Triton X-100 plus 0.1 N NaOH and counted in a scintillation counter. The specific binding curve was generated by subtracting non-specific binding (³H-EC17 plus “cold” EC17) from total binding (³H-EC17) at each concentration of the radioligand.

To investigate EC17 CAM action in vitro, pure EGFRt-sorted CD4 and CD8 CAR-T cells were admixed in a 1:1 ratio and matching mock-transduced T cells were included per experimental design. Other than FR+/- isogenic pairs of AML and osteosarcoma cell lines, we added KB, MDA-MB-231 and OV90 to represent different histological tumor types and FR expression levels. All FR+ target cell lines were subjected to a direct ³H-FA binding assay to quantitate the number of FR molecules bound/cell (38). Prior to co-culture experiments, cryopreserved CAR-T cells were allowed to recover for 2–3 days in T cell culture medium. All effector and target cells were pre-washed with FFRPMI to remove any exogenous FA that could compete against EC17. Short-term co-culture studies of ≤ 3 days were carried out in FFRPMI without adding any exogenous cytokines.

For an EC17 dose response study, FR+ cancer cell lines (KB, MDA-MB-231, HOS-FRα, THP1-FRβ, OV90) were co-cultured with EGFRt-sorted E2-CAR-T cells at an effector-to-target (E/T) ratio of 1:1 in the presence of EC17 ranging from 0.1 pM to 100 μM in 10-fold increments. After 24 h of co-culture, supernatants were harvested to determine target cell killing by Promega's CytoTox 96® LDH assay kit and expressed as specific lysis (%) normalized to basal levels in the absence of EC17. To study the kinetics of T cell activation and FR correlation in vitro, co-cultures were carried out in FR+/- cell lines at varying E/T ratios (1:27, 1:9, 1:3, 1:1, 3:1) in the continuous presence of 10 nM EC17. The kinetics of target cell killing (i.e., % specific lysis) were quantified after 16 and 48 h of co-culture and normalized to basal levels in the absence of EC17 at defined E/T ratios. Meanwhile, a portion of the supernatants was taken after 44 h of co-culture for measurement of IFNγ, IL-2, and TNFα. The LEGENDplex™ human Th1 cytokine panel (BioLegend, San Diego, CA) was used to determine CAR-T cell derived cytokines.

To study CAR-T cell activation in the presence of FR+ tumor cell targets, co-culture experiments were initially carried out with MDA-MB-231, THP1-FRβ, and HOS-FRα. After a 30 min exposure to media containing 100 nM EC17, all target cells were washed and then incubated at 1:1 E/T ratio with EGFRt-sorted E2-CAR-T cells (1:1 CD4/CD8) or mock transduced control T cells. After 24 h of co-culture, the floating and adherent cells were harvested and pooled. The surface expression of T cell activation markers CD69, CD137 (4-1BB), and PD1 were analyzed by flow cytometry. For T cell exhaustion studies, FR+ (KB, MDA-MB-231, HOS-FRα^fLuc, THP1-FRβ) and FR-negative (HOS-143b^fLuc, THP1-FG12) tumor cell lines were used as targets. Here, target cells were incubated on day 0 without or with 0.1 and 10 nM of EC17 for 30 min at 37°C and the status of EC17 “pre-loading” was assessed by counterstaining cell-surface EC17 with an anti-FITC antibody. Three days later, the surface expression of PD1, LAG3, and TIM3 inhibitor receptors on co-cultured CAR-T cells were analyzed and higher frequencies of double- and triple-positive cells were considered approaching an exhausted phenotype (39). As tumor cells used in this study showed variability of doubling times, target cell apoptosis (%) was measured as Annexin V+ after 2 days of co-culture and normalized against background levels of apoptosis of target cells cultured under the same conditions in the absence of T cells.

All animal care and use were performed according to NIH guidelines and in compliance with protocols approved by the Purdue University Animal Use and Care Committee. Female 4 to 5-weeks-old NOD/SCID gamma (NSG™) mice (stock number: 005557) were purchased from The Jackson Laboratory (Bar Harbor, ME). Unless specifically indicated, all animals were maintained on a FA-deficient diet (TestDiet, St. Louis, MO) upon arrival and throughout the study. To establish subcutaneous xenografts, MDA-MB-231 and HOS-FRα were implanted in the right flank region at 2.5 × 10⁶ and 1 × 10⁶ cells per animal, respectively. For intravenous xenografts, THP1-FRβ cells were inoculated at 5 × 10⁶ cells per animal. Subcutaneous tumors were measured 2–3 times per week with a caliper and calculated using the ellipsoidal formula (length x width²)/2. Euthanasia was performed per study design or when (i) the animals had lost ≥20% of body weight or approached moribund conditions, (ii) subcutaneous tumors reached ≥1,500 mm³ in size, or (iii) animals displayed signs of swollen belly and severe distress (i.e., THP1-FRβ). All animal doses (CAR-T cells, EC17, sodium fluorescein) were given intravenously.

To study CAR-T cell expansion in vivo, we started with 15 MDA-MB-231 tumor bearing mice having ~4.8 million EGFRt-sorted E2-CAR-T cells (1:1 CD4/CD8) injected on day 0. When the tumor sizes averaged at ~350 ± 60 mm³ on day 2, up to 6 weekly doses of EC17 at 500 nmol/kg were given on days 2, 9, 16, 23, and 30. For ex-vivo analysis, 3 animals were taken before the next EC17 dose on days 9, 16, 23, and 30 for a total of 4 collections. For comparison, we also gave the same EC17 dose to 3 CAR-T mice that had ~4.4 × larger tumor sizes on day 2 (~1555 ± 79 mm³). For each collection, fresh blood and tumor samples (if available) were analyzed for the presence of human CD3ε+ EGFRt+ CAR-T cells by flow cytometry. Moreover, we analyzed phenotypic changes of E2-CAR-T cells in the circulation (T_SCM, T_CM, T_EM, and T_EFF) and the activation status of tumor-infiltrating CAR-T cells (i.e., CD137/4-1BB, PD1).

For CAR-T cell analysis in the circulation, plasma was removed from a predetermined volume of whole blood collected into tubes containing ethylenediaminetetraacetic acid anticoagulant. After lysing red blood cells, leukocyte pellets were re-suspended in a flow cytometry staining solution comprising 1% bovine serum albumin, 50 mg/mL human IgG (Equitech Bio) and 0.09% sodium azide. The samples were stained for human leukocyte surface markers (CD3ε, CD4, CD8α, CD45RA, CD62L) and biotinylated anti-human EGFR (cetuximab) followed by a fluorophore-conjugated anti-biotin secondary antibody. For analysis of tumor-infiltrating CAR-T cells, pre-weighed fresh tumor fragments were finely minced and enzymatically digested with a digestion cocktail consisting of 0.5 mg/mL collagenase IV, 0.5 mg/mL hyaluronidase and 0.1 mg/mL DNase I in serum-free FFRPMI with vigorous shaking for 1 h at 37°C. Afterwards, tumor cell pellets underwent a red blood cell lysis step, washed with cold PBS and filtered through a 40 μm nylon cell strainer. The resulting single cell suspensions were stained for EGFRt and human leukocyte markers, CD137/4-1BB and PD1. A minimum of 20,000 propidium iodide negative live cell events were collected on the Gallios flow cytometer and analyzed with the Kaluza software, version 2.1 (Beckman Coulter, Brea, CA). CAR-positive T cells were identified in mouse blood as human CD3ε+ EGFRt+ events and absolute numbers per volume of blood were calculated using equal numbers of CountBright™ beads (Invitrogen, Carlsbad, CA) added to each sample. The number of tumor-infiltrating CD3ε+ EGFRt+ CAR-T cells was expressed as % total viable tumor cells analyzed.

In a therapeutic setting, EC17 CAM can theoretically be given before or after CAR-T cell injection (33). For all intended purposes here, we administered the first dose of EC17 2–3.5 days after CAR-T cells to allow for an observation period of human T cells in tumor-bearing mice. Two batches of unsorted E2-CAR-T cells (23% or 39% EGFRt+, 1:1 CD4/CD8) were used for in vivo studies. On the day of infusion for each experiment (day 0), frozen CAR-T cells were quickly thawed at 37°C, washed 2x with Dulbecco's 1X PBS (pH 7.4) and injected into the tail veil at desired EGFRt+ E2-CAR-T cell doses. In addition, a small aliquot of CAR-T cells was analyzed by flow cytometry for CD4 to CD8 ratio and differentiation status of CAR-T cells. On the first day of EC17 dose, tumor-bearing animals were randomly assigned into groups according to their tumor sizes or the same number of days post intravenous implantation (i.e., THP1-FRβ).

For MDA-MB-231 studies, mice received a high dose (~10 million) of a “clinical facsimile” of E2-CAR-T cells (~39% EGFRt+) with a low differentiation profile. Two days later, we started them with 3 different treatment regimens of EC17 at an average tumor size of ~293 ± 39 mm³. The EC17 dosing was given once-a-week (SIW) at 500 nmol/kg on Mondays, or as escalating doses of 5, 50, or 100, and 500 or 1,000 nmol/kg on Monday, Thursday, and Monday with a 6 day break in-between cycles. Control mice were left untreated (received CAR-T cells but no EC17). For comparison, a cohort of tumor-free littermates also received the same number of CAR-T cells without or with EC17 SIW at 500 nmol/kg. To study the effect of high dietary folate, two sets of NSG mice were placed on either a FA-replete diet (4 mg/kg, Envigo, Indianapolis, IN) or a FA-deficient diet (TestDiet, St. Louis, MO) upon arrival. On day 0, both sets of the mice received ~10 million of the same “clinical facsimile” CAR-T cells followed by no EC17 or EC17 SIW at 500 nmol/kg starting on day 3. On the first day of EC17 administration, MDA-MB-231 tumors size averaged around ~549 ± 184 mm³ (280–918 mm³) in FA-replete mice, and ~559 ± 165 mm³ (356–961 mm³) in FA-deficient mice.

For AML studies, mice were intravenously infused with THP1-FRβ tumor cells 1 day prior to receiving a low dose of ~6 million E2-CAR-T cells (~23% EGFRt+). At ~3.5 days post CAR-T cell infusion, EC17 was dosed in 3 different ways including (i) SIW at 500 nmol/kg, (ii) thrice at 5/50/500 nmol/kg on Monday/Wednesday/Friday followed by a 9 day break in between cycles (TIW On/Off), and iii) as escalating doses of 5/10/100 nmol/kg in Cycle 1, 5/30/300 nmol/kg in Cycle 2, and 5/50/500 nmol/kg in Cycle 3, all on Monday/Thursday/Monday with a 6 day break in between cycles (M/Th/M, On/Off). On day 31, satellite animals were harvested for quantification of CAR-positive T cells identified in the mouse blood as human CD3ε+ EGFRt+ events and calculated as absolute numbers per 100 μL of whole blood. Upon euthanasia or at the end of study on day 38, total tumor load in THP1-FRβ tumor-bearing mice was assessed by measuring GFP+ tumor cells in the blood by flow cytometry and collecting liver weight (with metastatic lesions) and total weight of nonliver macrometastases found in the body cavities. Tumor fragments of liver metastases were enzymatically digested into single-cell suspensions and viable cell populations were analyzed for the status of CAR-T cell exhaustion using anti-human PD1, LAG3, and TIM3 (clones EH12.1, T47-530, and 7D3, respectively).

For osteosarcoma study, two cohorts of mice were subcutaneously implanted with HOS-FRα tumor cells 3 days prior to receiving ~6 million of the same CAR-T cell preparation used in the THP1-FRβ study. At ~3.5 days post CAR-T cell infusion, one cohort of mice was given up to 3 cycles of EC17 at 5/10/100 nmol/kg in Cycle 1, 5/30/300 nmol/kg in Cycle 2, and 5/50/500 nmol/kg in Cycle 3, all following the Monday/Thursday/Monday regimen with a 6 day break in-between cycles. At the end of study, circulating CD3ε+ EGFRt+ CAR-T cells per 100 μL of mouse blood were enumerated. HOS-FRα tumors (+/- EC17 treatment) were also harvested and digested for flow cytometric analysis of tumor-infiltrating CAR-T cells.

Depending on the CAR-T cell dose and how EC17 is administered, tumor-bearing mice receiving FITC-specific CAR-T cells can experience severe CRS and varying degrees of body weight loss (27, 33, 34). Therefore, we developed a CRS grading system (0–5 scale) to empirically assess CRS toxicity from animals' gross morphology and social behavior to allow for better timing of CRS rescue. While grades 0 and 5 indicated normal (no CRS) or death (due to severe CRS), respectively, grades 1, 2, and 3–4 were considered mild, light-to-moderate, and severe CRS. In addition, all EC17 doses were intentionally given toward the end of any given day to allow potential CRS symptoms to develop overnight. On the days immediately following each EC17 dose, animals were scored and rescue agents such as sodium fluorescein, folic acid and leucovorin were used to mitigate CRS toxicity (33). In this particular instance, a batch of E2-CAR-T cells with a high EGFRt+ CD4/CD8 ratio (~93:7) was given to HOS-FRα tumor-bearing mice at a mixture of ~2.5 million CD8 and ~33 million CD4 on day 0. When EC17 was dosed, respectively, at 500, 100, and 500 nmol/kg on days 3, 10, and 12, low CRS (grades 1–2) was noted on days 4 and 11 but high CRS (grade 3) was noted on day 13. To mitigate the severe CRS toxicity, sodium fluorescein was intravenously administrated at ~96 mg/kg in the morning of day 13 and mouse plasma samples were collected 6 h later for analysis of CRS-associated cytokines.

Statistical analyses were performed using the computer program GraphPad Prism (GraphPad Software Inc., San Diego, CA). Data were analyzed using Student's t-test or one-way ANOVA. If applicable, data were further analyzed across treatment groups using appropriate multiple comparison post-test. ^*p < 0.05 was considered statistically significant in all tests.

The basic components of our CAR-T cell therapy involve FITC as the pseudo tumor antigen, the high affinity EC17 CAM, and a rationally designed anti-FITC CAR and CAR-modified T cells (Figure 1). To our advantage, the bispecific engager, EC17, has already been tested in the clinic for immunotherapy and optical imaging purposes (28, 29, 32, 40). To directly quantify its bispecific binding affinities, ³H-EC17 was synthesized and radioligand binding assays were carried out on KB and CHO-β cell lines representing FRα+ and FRβ+ target cells, respectively, and on E2-CAR-T cells representing the effector cells. When binding to its targets, EC17 demonstrated similar affinities toward both FRα and FRβ with low Kd values of 1.7 and 0.8 nM, respectively (Figure 2A). Upon binding to unsorted E2-CAR-T cells (~100% EGFRt+, 1:1 CD8/CD4 ratio), the Kd value was estimated at ~142 nM (Figure 2B).

Amid a series of different CAR constructs synthesized and evaluated by the Jensen team, a fully human anti-FITC scFv (FITC-E2) CAR was chosen for preclinical development (Figure 1D). This second-generation fully human CAR consisted of anti-FITC scFv (clone E2), an IgG4-Fc spacer/hinge with double mutations in the CH2 region (L235D and N297Q) to reduce binding to FcγR (36), a CD28 transmembrane domain, and 4-1BB/CD3ζ signaling domains appended to a cell-surface EGFRt tag by a T2A ribosomal skip sequence (i.e., FITC-E2-scFv-IgG4hinge-CD28tm-4-1BB/CD3ζ-T2A-EGFRt). For preclinical studies, both EGFRt-sorted and unsorted E2-CAR-T cells were prepared at ~1:1 CD4/CD8 ratios, and T cell subtype phenotyping was routinely performed by flow cytometry at the time of CAR T cell infusion (day 0) for each in vivo experiment. A typical expression pattern of EGFRt-sorted CAR-T cells was comprised of both CD4 and CD8 subsets at ~42% T_SCM, 10% T_CM, 12% T_EM, and 34% T_EFF (Figure 1D, pie charts on the left). Only EGFRt-sorted CAR-T cells were used for co-culture and pharmacokinetic studies. For tumor therapy, we used a “clinical facsimile” batch with a low differentiation profile (Figure 1D, pie charts on the right) for MDA-MB-231 studies and a research batch for THP1-FRβ and HOS-FRα studies. The “clinical facsimile” batch (~39% EGFRt+) was comprised of CD4+ subsets at ~66% T_SCM and ~32% T_CM and CD8 subsets at ~95% T_SCM and 3% T_CM. The research batch (~23% EGFRt+) was more differentiated and comprised of CD4 subsets at 32% T_SCM, 53% T_CM, 11% T_EM, and 3.7% T_EFF and CD8 subsets at 44% T_SCM, 0.28% T_CM, 3.4% T_EM, and 52% T_EFF.

Besides pediatric cancer cell lines transfected with FRα (HOS-FRα) and FRβ (THP1-FRβ), we included cancer cell lines of different histology and FR expression levels (Figure 2C). As estimated by a radioligand binding assay (100 nM ³H-FA, 1 h at 37°C), the ranking order of total available FRs on these cell lines was: 9 × 10⁴ (OV90, a low-FR expressing ovarian cancer cell line), 1.9 × 10⁵ (THP1-FRβ), 2.4 × 10⁵ (HOS-FRα^fLuc), 7 × 10⁵ (HOS-FRα), 2.1 × 10⁶ (MDA-MB-231), and 4.8 × 10⁶ (KB) FA molecules/cell. Also included as FR-negative controls were HOS-143b^(fLuc) and THP1-FG12 parent cell lines. Thus, the general ranking of functional FR expression on co-cultured FR+ cancer cell lines was: KB > MDA-MB-231 > HOS-FRα > HOS-FRα^fLuc > THP1-FRβ (AML) > OV90.

Monovalent EC17 in its free form (i.e., one FITC per FA ligand) does not automatically activate anti-FITC CAR-T cells (27). To study the robustness of CAR-T cell activation, 5 FR+ cell lines (KB, MDA-MB-231, HOS-FRα, THP1-FRβ, OV90) were used along with a wide range of EC17 concentration (0.1 pM to 100 μM) at an E/T ratio of 1:1 (Figures 2D,E). Specific lysis (%) was quantified by Promega's LDH cytotoxicity assay kit after 24 h of co-culture. In all cell lines, EC17-dependent cytolysis followed a bell-shaped dose response with a somewhat broad concentration plateau (~0.1 nM to 1 μM) (Figure 2D). When the dose-response curves were fitted to the summit (up to 100 nM), half-maximal effective concentrations (EC₅₀) were obtained at 4.5 pM (MDA-MB-231), 5.3 pM (KB), 15 pM (THP1-FRβ), 30 pM (HOS-FRα), and 418 pM (OV90), respectively. However, maximal lysis obtained for each cell line ranked differently at 45% (MDA-MB-231), 27% (THP1-FRβ), 18% (HOS-FRα), 15% (OV90), and 11% (KB) (Figure 2E). The low EC50 values in general suggested that EC17 was highly potent in triggering FR-specific target cell lysis by activated E2-CAR-T cells.

To expand our investigation into FR dependency, 5 FR+ (MDA-MB-231, KB, HOS-FRα, THP1-FRβ, OV90) and 1 FR-negative (HOS-143b) cell lines were co-cultured in the continuous presence of 10 nM EC17 with E2-CAR-T cells at varying E/T ratios (1:27, 1:9, 1:3, 1:1, and 3:1). Specific lysis (%) was quantified after 16 and 48 h of co-culture, and Th1 cytokines (IFNγ, IL-2, TNFα) in culture media were measured after 44 h of co-culture. While the onset of specific lysis (~16 h) varied among different tumors, an increase in specific cytolysis was observed at 48 h for all FR+ cancer cell lines (Figure 3A). It generally took a longer time (48 h) and a higher E/T ratio (≥1:1) to see significant activity in the lowest FR-expressing OV90 cells (Figure 3A). Interestingly, the high FR-expressing KB cells responded very slowly and also required a higher E/T ratio (≥1:3) and longer exposure time to cause significant cell death (Figure 3B, left plot). When enough effector cells were present (≥1:1 E/T ratios), all FR+ cancer cells responded after 48 h of co-culture. Notably, only the FR-negative HOS-143b cells were essentially unharmed (Figure 3B, right plot).

To establish a correlation of CAR-T cell activity with FR expression in vitro, we plotted the values of 16 h specific lysis at an E/T ratio of 1:1 against the functional FR levels found on tumor cells (Figure 2C). Excluding KB cells that showed an unusual resistance at the beginning, there was a strong semi-log correlation between FR expression and the onset of cytolysis with an R-value of 0.98 (Figure 3C). Similarly, all but the FR-negative HOS-143b tumor cells triggered high levels of IFNγ, IL-2, and TNFα production after 44 h of co-culture at an E/T ratio of 1:1 with E2-CAR-T cells (Figure 3D). Interestingly for Th1 cytokine production, a semi-log correlation was also observed with all cell lines included at R-values of 0.88 (IFNγ), 0.89 (IL-2), and 0.89 (TNFα), respectively (Figure 3E). In short-term co-cultures, MDA-MB-231 was found most sensitive to killing by CAR-T cells, but HOS-FRα osteosarcoma and THP1-FRα AML also responded quickly in a FR level-dependent manner (Figure 3C). Together, these data suggested that (i) CAR-T cell is activated only when EC17 forms the intercellular bridge between CAR-T cell and tumor cell, (ii) other factors such as E/T ratio, exposure time and tumor sensitivity confer the overall activity of CAR-T cell against a particular FR+ tumor target, and (iii) too much EC17 would interfere with CAR-T cell engagement leading to decreased target cell death.

To examine antigen-dependent CAR-T cell activation, we initially pre-loaded MDA-MB-231, THP1-FRβ, and HOS-FRα cells with 100 nM EC17 over 30 min. The cells were washed and then incubated at 1:1 E/T ratio with either EGFRt-sorted E2-CAR-T cells (1:1 CD4/CD8), or mock transduced control T cells. After 24 h of co-culture, we measured the surface expression of T cell activation markers CD69, CD137 (4-1BB), and PD1 (Figure 4A). EC17-preloaded MDA-MB-231 and HOS-FRα were strong activators of our CAR-T cells. THP1-FRβ cells, which express 11-fold lower FR compared to the MDA-MB-231 cells, did trigger CAR-T cell activation, but at lower levels. Importantly, the mock control T cells were not activated under these conditions. As some FR+ tumor types (e.g., KB) display a natural resistance to killing by EC17-directed CAR T cells (Figures 2, 3), we studied T cell activation vs. exhaustion as a mechanism of EC17 CAM action. Both FR+ (KB, MDA-MB-231, HOS-FRα^fLuc, THP1-FRβ) and FR-negative (HOS-143b^fLuc, THP1-FG12) tumor cell lines were chosen for this purpose. HOS-FRα^fLuc expressed ~5.6x lower level of functional FR than its parent HOS-FRα, and HOS-143b^fLuc was FR-negative similar to HOS-143b (Figure 2C). Here, FR+ and FR-negative tumor cells without or with EC17 pre-loading (0.1 or 10 nM, 30 min pulse at 37°C) were incubated with EGFRt-sorted CAR-T cells (1:1 CD4/CD8) at an E/T ratio of 1:2 (Figures 4B,D). EC17 dose-dependent pre-loading was confirmed on day 0 by counter-staining of membrane bound EC17 with an APC-conjugated anti-FITC antibody (clone NAWESLEE) (Figure 4C). Three days later, co-cultured CAR-T cells were measured for upregulation and co-expression of T cell activation/exhaustion markers PD1, LAG3, and TIM3. In the absence of target cells, E2-CAR-T cells underwent a low-grade differentiation in culture with low frequencies of double or triple-positive cells. Interestingly, the presence of tumor cells alone (without EC17 pre-loading) appeared to “relax” the CAR-T cells to various degrees. But when encountering EC17-preloaded FR+ (but not FR-negative) tumor cell targets, E2-CAR-T cells underwent significant differentiation with increased co-expression of exhaustion markers. Based on increased frequencies of triple- and double-positive cells, a more exhausted phenotype appeared on co-cultured CAR-T cells in the order of KB > HOS-FRα^fLuc > THP1-FRβ > MDA-MD-231. While EC17 dose-dependent apoptosis (Annexin V+) was seen in all FR+ tumor cell lines on day 2, high FR-expressing KB cells again showed disproportionally low apoptosis (Figure 4D).

The dynamic interplay between EC17 and CAR-T cells in vivo was further interrogated using MDA-MB-231 tumor model, which was previously found to be among the highest FR expressing tumors at 103 ± 15 pmol/mg membrane protein using a ³H-FA-based radioligand binding assay (41). In our study, we evaluated the kinetics of CAR-T cell expansion in tumor-bearing mice that received ~4.8 million of EGFRt-sorted E2-CAR-T cells (Figure 5A). Seven days following a single EC17 dose (i.e., study day 9), human CD3ε+ EGFRt+ CAR-T cells were detected in the blood of mice, with elevated numbers circulating in mice bearing larger tumors (Figure 5B). In addition to trending toward a higher expansion in vivo, these blood borne CAR-T cells also exhibited a more differentiated profile (Figure 5B, 2504 ± 441 mm³ vs. 422 ± 16 mm³) in mice possessing larger tumors. Mice with the smaller sized tumors continued on study to receive a total of 5 weekly EC17 doses at 500 nmol/kg. As shown in Figure 5C, the MDA-MB-231 tumors started to respond after the second EC17 dose. Mice did experience mild body weight loss after EC17 administration, especially after the third dose. As circulating CAR-T cells peaked around day 15 and persisted for up to 30 days (last analysis by flow cytometry), we observed a steady increase in tumor-infiltrating CAR-T cells (Figure 5D, top plot, dashed line). In addition, CAR-T cells underwent dynamic phenotypic changes in the blood from a predominantly T_SCM phenotype on day 9, to more differentiated profiles on later days with continued EC17 treatment (Figure 5D, bottom bar graph). Accordingly, those tumor-infiltrating CAR-T cells were activated and expressed early (CD137/4-1BB) as well as late (PD1) T cell activation markers (Figure 5E). CD137/4-1BB expression peaked around day 9 while PD1 expression remained high from day 9 and forward. Thus, EC17 CAM dosing drives CAR-T cell activation, expansion and tumor uptake in vivo resulting a robust antitumor immunity.

The initial EC17 dose finding studies were conducted in MDA-MB-231 tumor bearing mice as shown by the schematic diagram of the experimental layout (Figure 6A). NSG mice without or with MDA-MB-231 tumors (~211 ± 65 mm³) were engrafted on day 0 with ~10 million of a “clinical facsimile” batch of E2-CAR-T cells (~39% EGFRt+, 51:49 CD4/CD8). This batch of CAR-T cells consisted of mostly T_SCM and T_CM phenotypes (see Figure 1D). Two days after the CAR-T cell injection, one cohort of tumor-bearing mice were left untreated while three cohorts were treated with different regimens of EC17, including single injection per week (SIW) at 500 nmol/kg, or as escalating EC17 dose levels of 5/50/500 (Escalation-1) or 5/100/1000 nmol/kg (Escalation-2) given on a Monday/Thursday/Monday schedule with 1-week drug-free intervals. For comparison, two tumor-free cohorts were either untreated or treated with EC17 SIW at 500 nmol/kg.

Human T cell-derived IFNγ levels in mouse blood were measured in all cohorts using satellite animals on days 11 and 12 (~20 and 42 h after the previous EC17 dose). Compared to tumor-bearing mice that received CAR-T cells only, all EC17 treated tumor cohorts had ~30x (day 11) and ~10x (day 12) higher IFNγ production in mouse plasma, and the levels of this cytokine decreased naturally from 20 to 42 h later (Figure 6B). In accordance with IFNγ release, EC17 also triggered CAR-T cell expansion identified as human CD3ε+ EGFRt+ events in mouse blood, and cells persisted up to 54 days in tumor-bearing animals (last measurement) (Figure 6C). In tumor-free cohorts, no CAR-T cell expansion was detected by flow cytometry, but low levels of IFNγ were detected on days 11 and 12 in animals that received the same number of CAR-T cells plus EC17 (Figures 6B,C). Moreover, no CRS symptoms (grade 0 out of a 0–5 scale) or body weight loss was observed in tumor-free mice that received 2 weekly doses of EC17 at 500 nmol/kg.

Moderate-to-severe CRS symptoms (grades 2-3) and significant body weight losses (Figure 6D) were observed in tumor-bearing cohorts with continued EC17 dosing independent of the regimen. While EC17 SIW at 500 nmol/kg caused the earliest onset of CRS and body weight loss, the aggressive EC17 Escalation-2 regimen (up to 1,000 nmol/kg) caused persistent body weight loss with animal recovery occurring after EC17 dose cessation (Figure 6D, top row). Notably, symptoms of graft-vs.-host disease (GVHD) included red itching skin and hairlessness became obvious at ~1 month after CAR-T cell engraftment. Although animals receiving only CAR-T cells showed signs of nonspecific CAR-T cell/tumor alloreactivity, only EC17-treated cohorts produced 100% cures (Figure 6D, bottom row). Therefore, EC17 administration in the presence of FR+ tumors was the key to drive (i) CAR-T cell activation, (ii) cytokine production, and (iii) in vivo CAR-T cell expansion and persistence. Under specific conditions, however, severe CRS (grade ≥ 3) was triggered by a high CAR-T cell dose in combination with EC17 doses equal or >500 nmol/kg (from herein referred to as 100% full dose).

Previously, it was shown that administration of FA or a FA ligand helped reduce the severity of CRS in MDA-MB-231 tumor-bearing mice (27, 33). To test the long-term effect of dietary folate, mice were maintained on defined diets for ~73 days and used when their tumors reached ~549 ± 184 mm³ (280–918 mm³) in FA-replete mice, and ~559 ± 165 mm³ (356–961 mm³) in FA-deficient mice (Figure 7). All mice received ~10 million of the same “clinical facsimile” of E2-CAR-T cells on day 0, and EC17-treated cohorts received 8 weekly doses of 500 nmol/kg starting on day 2 (FA-replete) or day 3 (FA-deficient). As seen previously, the first full dose of EC17 was generally safe and did not cause any CRS or body weight loss in mice on either diet. Actually, FA-replete animals did not show symptoms of CRS throughout EC17 treatment, whereas FA-deficient animals experienced grades 2–3 CRS and body weight loss (up to ~11.4%) with each of the subsequent EC17 doses (Figure 7A, bottom panels). In the same duration of EC17 treatment, a much higher level of CD3ε+ EGFRt+ CAR-T cells was found in mice on FA-deficient diet (1.6–20.4 × 10⁴ per 100 μL blood on day 52) when compared to mice on FA-replete diet (0.027–4.3 × 10³ per 100 μL blood on day 59) (Figures 7B,C). Importantly, there was no CAR-T cell expansion detected in FA-deficient animals that did not receive EC17 treatment (Figure 7C).

Since we started MDA-MB-231 tumors at larger sizes than the previous study (see Figure 6), tumor regression due to alloreactivity at this 10-million “clinical facsimile” CAR-T cell dose was not apparent (Figure 7A, top rows). Control tumors (no EC17) in FA-replete mice appeared to have faster growth kinetics compared to control tumors in FA-deficient mice; but overall, there were 2/5 cures, 2/5 complete responses, and 1/5 partial response in FA-replete mice that did receive EC17 (Figure 7A, top left figure). A more robust in vivo T cell expansion was seen in mice on FA-deficient diet and resulted in 5/5 cures, although these animals' health deteriorated with each CRS episode (Figure 7A, top right figure).

Using anti-human EGFR (clone Hu1) to identify MDA-MB-231 cancer cells within mouse tumor masses after an enzymatic digestion, we analyzed FRα protein expression (clone LK26) in three tumors recovered from mice on FA-replete diet that experienced a relapse despite continuing EC17 treatment. By flow cytometric detection, EGFR+ cancer cells isolated from the smallest tumor (21 mg) expressed FRα, while EGFR+ cancer cells isolated from the two larger tumors (90 and 363 mg) had completely lost their FRα expression (Figure 7D). Using the quantitative radioligand binding assay with ³H-FA, we further confirmed the loss of functional FR levels on tissue homogenates obtained from these tumors (~2.1 pmol/mg binding potential vs. ~103 pmol/mg in tumors from untreated animals). Therefore, FA-deficiency may lead to enhanced activity and CRS toxicity; while non-physiological FA intake prevents CRS and continuous consumption may lead to reduced activity (see Discussion).

Intravenously implanted GFP-expressing THP1-FRβ tumor cells developed disseminated diseases in NSG mice with tumor cells in the circulation and liver/non-liver metastases throughout the body. We found that THP1-FRβ tumor cells could also localize to the mouse ovary which appeared inflamed during the early stage of tumor progression. Therefore, total tumor burden in each animal in study cohorts was assessed by quantitating circulating GFP+ tumor cells in the blood, liver weights, and all-inclusive non-liver macrometastases visible to the naked eye. Although THP1-FRβ expressed a low level of FR in vitro, THP1-FRβ tumor metastases were found to express a higher than expected functional FRs level at ~8.9 ± 2.8 pmol/mg membrane protein. Thus, THP1-FRβ tumor-bearing mice were engrafted with a research batch of EGFRt-unsorted E2-CAR-T cells (~23% EGFRt+, 1:1 CD4:CD8) at ~6 million/animal and then treated with 3 different EC17 dosing regimens (Figure 8). Starting 3 days after CAR-T cell injection, EC17 dosing regimens began as SIW at 500 nmol/kg continuously, thrice a week (TIW) at 5/50/500 nmol/kg on Monday/Wednesday/Friday followed by a 9 day break, or by an accelerated dose escalation regimen at 5/10/100 nmol/kg in Cycle 1, 5/30/300 nmol/kg in Cycle 2, and 5/50/500 nmol/kg in Cycle 3, on Monday/Thursday/Monday (M/Th/M) followed by a 6 day break (Figure 8A). While some body weight loss and grade 1–2 CRS in Cycles 2 and 3 were observed in animals received EC17 SIW treatment, animals that received EC17 TIW had the least body weight loss with grade 0–1 CRS only (Figure 8B). Amongst the animals that received the 3 cycles of EC17 M/Th/M dose escalation, grade 0–1 CRS and very mild body weight loss were observed in Cycle 2 after the last dose of EC17 of 300 nmol/kg. Using satellite animals on day 31, we enumerated CAR-T cells in the blood and demonstrated EC17-dependent CAR-T cell expansion and persistence in all treated cohorts (Figure 8C). Compared to control animals that received tumor cells only or tumor cells plus CAR-T cells without EC17, EC17 dosed with any of the 3 “intra-patient” escalation formats effectively reduced circulating THP1-FRβ tumor cells in the blood and showed similar activities against liver tumor metastases (Figure 8D, left and middle bar graphs). Only minor allogeneic reactivity was seen against THP1-FRβ liver metastases in mice that received CAR-T cells only. While EC17 SIW and TIW at 10-fold dose escalation (on/off) successfully controlled non-liver macrometastases, EC17 M/Th/M dose escalation at a slow pace per cycle (Figure 8A) failed to control the non-liver macrometastases (Figure 8D, far right panel). At the end of study (i.e., 39 days post CAR-T cell injection), CAR-T cells isolated from liver THP1-FRβ tumor metastases of EC17 TIW-treated animals appeared to have the least expression of double- and triple-positive T cell inhibitory receptors, PD1, LAG3, and TIM3 (Figure 8E). Overall, no severe CRS (i.e., grades ≥ 3) was observed in any EC17 treated cohorts. Nevertheless, the more aggressive EC17 TIW dose escalation (on/off) trended as the best regimen for reducing overall THP1-FRβ tumor burden in these mice.

For our intended purpose, HOS-FRα is a low FR-expressing but most aggressive tumor model with a functional FR level of ~5.82 ± 1.45 pmol/mg protein. In parallel to the THP1-FRβ study described above, two cohorts of mice with 3 day-old HOS-FRα tumors were given the same E2-CAR-T cells at the same dose (~6 million). One cohort was treated with the same accelerated EC17 dose escalation regimen at 5/10/100 nmol/kg in Cycle 1, 5/30/300 nmol/kg in Cycle 2, and 5/50/500 nmol/kg in Cycle 3, on Monday/Thursday/Monday schedule followed by a 6 day break (Figure 9A). As HOS-FRα tumors grew aggressively without EC17 treatment, accelerating EC17 dose escalation at this CAR-T cell dose was safe (no CRS or body weight loss) and resulted in a significant delay in tumor growth within the first two cycles of treatment (Figure 9B). Upon protocol-mandated euthanasia due only to tumor size (≥1,500 mm³), flow cytometric analyses performed on whole blood showed an EC17-dependent CAR-T cell expansion up to day 47 but higher on day 33 (Figure 9C, left bar graph). Ex vivo tumor analysis on day 33 indicated a low but significant intratumoral CD3ε+ EGFRt+ CAR-T cell population in EC17-treated animals which amounted to ~1% total viable digested tumor-derived cells (Figure 9C, right bar graph). As large HOS-FRα tumors stopped responding to treatment in Cycle 3 (Figure 9B), intratumoral CAR-T cells also diminished on day 47 (Figure 9C). Notably, HOS-FRα tumors analyzed by ³H-FA radioligand assay upon disease progression showed similar FRα levels with and without EC17 treatment. Thus, it appeared that HOS-FRα tumors in NSG mice quickly outgrew the tumor infiltrating capability of CAR-T cells and may have decreased cytolytic activity due to T cell exhaustion as suggested by in vitro co-culture studies (Figure 4B).

To scrutinize our CRS rescue capabilities with sodium fluorescein, we developed a CRS scoring system and was able to capture the onset of severe CRS (grades 3/4) following EC17 administration (Figure 9D). As shown by a challenging case of rapid CRS onset in HOS-FRα tumor-bearing mice due to a very high dose of 35 million total CAR-T cells (93:7 CD4/CD8), intravenous sodium fluorescein at a human equivalent dose (~95 mg/kg in mice) acted very quickly to reduce CAR-T cell derived human cytokines (Figure 9E).

All authors listed as employees of Endocyte Inc. (YL, HC, LW, MN, EW, NP, MV, L-CX, EZW, PK, PL, CL) were stock holders at the time this work was completed. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.