Targeted in vivo delivery of genetic medicines utilizing an engineered lentiviral vector platform results in CAR T and NK cell generation

Abstract

The development of chimeric antigen receptor (CAR) T cell therapies has greatly impacted the treatment of B cell malignancies; however, manufacturing these patient-specific, autologous cell therapies is complex, costly, and requires preconditioning chemotherapy prior to infusion, limiting patient access. Here we describe the development of a lentiviral platform based on a novel, detargeted viral fusogen (Gen 2.1 Fusogen) and a membrane-bound targeting moiety to enable in vivo targeted delivery of stably integrating genetic medicines without the need for lymphodepletion. INT2104 employs a single chain variable fragment (scFv) targeting CD7 (“CD7 Binder”) to deliver a CAR20 transgene to CD7⁺ T and natural killer (NK) cells. Preclinical data generated in mouse and cynomolgus macaque models indicate INT2104 results in both CAR T cells (CD4⁺ and CD8⁺) and CAR NK cells with subsequent depletion of CD20⁺ B cells following a single intravenous administration. Thus, INT2104 could potentially provide a more accessible, off-the-shelf treatment option for patients who may benefit from CAR therapies.

Graphical abstract

In vivo generation of CAR cells enabled by the rational design of a lentiviral vector (INT2104) engineered to specifically transduce CD7⁺ T and NK cells. INT2104 specificity is confirmed by the detection of CAR20⁺ T and NK cells in both mouse and NHP models.

Introduction

Genetic medicines and the engineering of human cells has had a major impact on public health in the past decades, with over 30 Food and Drug Administration-approved cellular and gene therapy products to date.¹ The approval of autologous chimeric antigen receptor (CAR) T cell therapies has revolutionized the treatment of B cell malignancies, inducing prolonged remission in a subset of patients.² CAR T cell therapies have also shown promise outside of oncology for the treatment of patients with autoimmune disorders.^3,4 Current autologous CAR T cell therapy is limited by complex, costly, and cumbersome manufacturing with non-robust processing that can impact the quality and potency of the cell products.⁵ Furthermore, due to the requirement for lymphodepletion, autologous CAR T therapy can pose risks for cancer patients with advanced disease states or for patients with indications where chemotherapy is not standard of care (e.g., autoimmune disorders).⁶ Taken together, these shortcomings of autologous CAR T cell therapy have prompted clinical strategies that aim to support shortening CAR T cell manufacturing time, utilize alternative allogeneic or “universal donor” strategies, or use repeated dosing of CAR T cells to circumvent the need for patient lymphodepletion.^7,8,9,10 While the potential impact of these strategies on patients is recognized, significant challenges remain, including graft-vs.-host disease (GVHD), the need for selection of appropriate donor cells, or limited persistence of engrafted cells.^7,11

An alternative strategy to the ex vivo cell engineering utilized in all currently licensed CAR T cell therapies is in vivo genetic engineering. In this approach, the gene therapy delivery vehicle is administered directly to patients, allowing for the modification of the patients’ cells within the body (Figure 1A). These delivery vehicles range from viral vectors (e.g., lentivirus) to lipid or polymeric nanoparticles and deliver a variety of genetic cargos (e.g., RNA, DNA, CRISPR-Cas9 complexes). Regardless of the delivery vehicle, all in vivo genetic engineering approaches share a key challenge—the precision delivery of the genetic medicine to the intended target cell type. In recent years, in vivo genetic engineering, particularly for CAR T cell generation, has developed beyond the initial preclinical proof of concept^12,13 into a rapidly progressing field that is now being tested in the clinic.¹⁴

Figure 1.

Rationally designed lentiviral vector allows for targeted delivery

(A) Schematic comparing ex vivo with in vivo CAR cell generation. (B) INT2104 schematic highlighting key structural components. (C) Lentiviral titers (TU/mL) determined by titration on target SupT1 cells for VSV-G mutants. Each symbol represents an individual vector preparation. Open triangles denote vector with binder and open circles without binder. (D) Crystal structure of VSV-G (residues 1–410, black) interfacing with the CR3 domain of LDL-R (cyan), recolored from pdb:5OY9. Sites of the engineered mutations representing the Gen 2.1 Fusogen highlighted in red. VSV-G numbering based on the mature protein (i.e., after cleavage of the 16 amino acid signal peptide). (E) Mean fluorescence intensity (MFI) of PE-Cy7 anti-His binding to cells labeled with His-tagged sLDL-R. Points represent the mean of duplicate wells ± the standard deviation (SD). (F) Titration curve for vectors pseudotyped with WT VSV-G (black), the Gen 2.1 Fusogen alone (red), or the Gen 2.1 Fusogen + Binder (blue) on either CD7⁺ parental SupT1 cells (closed symbols) or engineered CD7⁻ SupT1 cells (open symbols). Points represent the mean of duplicate wells ±SD. (G and H) Representative flow plots showing specific transduction of targeted cells within activated human PBMCs 7 days after treatment with lentiviruses pseudotyped with the Gen 2.1 Fusogen and denoted binder moiety.

Here, we describe a clinical candidate developed with a platform enabling transduction of targeted cell types in vivo utilizing an engineered, third-generation lentiviral vector with rationally designed fusogen and binder moieties that enable the vector to target specific cells (Figure 1B). We report that a single point mutation in the fusion protein of vesicular stomatitis virus G protein (VSV-G) blinds VSV-G to its native receptor (low-density lipoprotein receptor [LDL-R]) without disrupting its fusogenic potential. This decoupling of VSV-G receptor binding and fusion enables retargeting of lentiviral vectors to specific cells via a high-affinity binder molecule provided as a second envelope protein. Model vectors encoding a green fluorescent protein (GFP) transgene were used to show the specificity of this system, leading to the development of clinical candidate, INT2104. INT2104 is a non-replicating, self-inactivating lentiviral vector delivering a transgene encoding a CAR specific for CD20 (CAR20) to CD7-expressing cells (Figure 1B). INT2104 was found to specifically transduce T (CD4⁺ and CD8⁺) and NK cells, generating functional CAR⁺ cells capable of killing CD20⁺ B cells. In vivo delivery of INT2104 to humanized mice via tail vein injection resulted in depletion of B cells, whether derived from engrafted hematopoietic stem cells or B cell tumor xenografts. Specific transduction of T and NK cells was also demonstrated in cynomolgus macaques following administration of an INT2104 surrogate vector with identical pseudotyping as INT2104 and differing only in a portion of the gag sequence to mitigate innate restriction factors present against HIV in these non-human primates (NHPs). Taken together, these results demonstrate that systemic delivery of INT2104 can specifically transduce CD7⁺ cells in vivo and effectively generate functional CAR⁺ cells, supporting the advancement of this platform in the clinic.

Results

Rational design of the Gen 2.1 Fusogen enables vector retargeting

To facilitate the specific in vivo delivery of genetic medicines, we took advantage of the independent binding and fusion functions of VSV-G, the standard glycoprotein used to pseudotype lentiviral vectors for ex vivo CAR T cell generation. Previous research identified amino acid changes (e.g., K47Q) that can ablate binding of VSV-G to its native receptor LDL-R, without affecting fusogenicity.¹⁵ To identify additional mutations that may blind VSV-G to LDL-R, in silico analysis was performed on the crystal structure of wild-type (WT) VSV-G bound to LDL-R¹⁵ searching for VSV-G residues in proximity to charged residues on LDL-R. In addition to the published blinding mutations,¹⁵ this approach identified residues Q10 and I182 as candidates: Q10 is oriented toward R103 in a cysteine-rich domain (CR3) of LDL-R while I182 is oriented toward D110 and D112, the negatively charged residues that drive much of the interaction with LDL-R (Figure S1A). We hypothesized that introducing opposing charges at these residues (e.g., Q10K or I182E) would destabilize the binding between VSV-G and LDL-R. To test this hypothesis, we generated lentiviral vectors encoding GFP and pseudotyped with the various VSV-G mutants and investigated their ability to transduce SupT1 cells. To confirm these mutations only impacted LDL-R binding without compromising the ability to undergo fusion, matched vectors were made that incorporated an additional proprietary binder molecule to “retarget” vector particles to cells expressing CD7. The ability of the resulting vectors to transduce CD7⁺ SupT1 cells was evaluated compared with control vectors pseudotyped with WT VSV-G or the published¹⁵ blinding mutations (K47Q + R354Q). Substitutions at Q10 were not effective at blinding VSV-G, with vectors transducing SupT1 cells independently of a binder molecule (Figures 1C and S1B). In contrast, both the charged substitutions at I182 (I182D and I182E) effectively detargeted VSV-G, resulting in lower transduction (Figures 1C and S1B). Importantly, incorporation of the CD7 binder restored dose-dependent transduction with similar infectivity to a WT VSV-G pseudotyped vector (Figures 1C and S1B). Consistent with the structure-based prediction of charge repulsion, substitution of the neutral side chain alanine did not abolish LDL-R binding (Figures 1C and S1B). The single amino acid substitution, I182E, was selected for detargeting as it resulted in less binder-independent transduction of SupT1 cells at high vector particle inputs compared with I182D.

A reported limitation of VSV-G pseudotyped vectors is their inactivation by human serum.^16,17 To address this issue, we incorporated amino acid substitutions previously reported to enhance serum and thermal stability (T214N and T352A).¹⁸ The optimized Fusogen, combining the I182E, T214N, and T352A substitutions, is designated the Generation 2.1 (Gen 2.1) Fusogen (Figure 1D). When tested in the presence of non-heat-inactivated, complement-containing medium, the Gen 2.1 Fusogen pseudotyped vector did not show a decrease in infectivity, in contrast to the WT VSV-G pseudotyped vector (Figure S1C).

Two orthogonal approaches were used to confirm that the Gen 2.1 Fusogen no longer binds to LDL-R. First, HEK293T cells were transiently transfected with plasmids expressing the Gen 2.1 Fusogen or control plasmids including WT VSV-G, I182E alone, or the K47Q detargeting mutation¹⁵ (Figure S1D) and screened for binding to soluble LDL-R (sLDL-R). While the serum-stabilizing mutations (T214N and T352A) reduced the measured binding relative to WT VSV-G, the I182E substitution alone or the three substitutions present in the Gen 2.1 Fusogen (I182E, T214N, and T352A) were effective in reducing binding of sLDL-R to assay background levels (Figure 1E). Next, a functional readout assessed the ability of sLDL-R to neutralize vector pseudotyped with the Gen 2.1 Fusogen and an additional binder molecule that targets the lentiviral vector to CD7-positive cells compared with vector pseudotyped with WT VSV-G. In this functional assay, sLDL-R did not neutralize vector pseudotyped with the Gen 2.1 Fusogen and CD7 Binder (half maximal inhibitory concentration [IC₅₀] > 50 μg/mL), despite potently neutralizing vector pseudotyped with WT VSV-G (IC₅₀ = 1.1 μg/mL) (Figure S1E).

Vector pseudotyped with Gen 2.1 Fusogen and CD7 binder enables specific transduction of target cells

After confirming the Gen 2.1 Fusogen no longer binds LDL-R but maintains its ability to drive membrane fusion, we further optimized the structure of the Binder molecule. The first iteration of the Binder was an anti-CD7 single chain variable fragment (scFv) supported on the surface of the vector by an immunoglobulin G subclass 1 (IgG1) Fc “stalk” anchored to the vector envelope by the transmembrane domain of human CD28 followed by the envelope incorporation motif of human immunodeficiency virus type 1 (HIV-1) gp41.¹⁹ Initial screening of a vector encoding GFP and pseudotyped with the Gen 2.1 Fusogen and this non-optimized CD7 Binder showed transduction of both CD7⁺ cells and some CD7⁻ B cell lines, suggesting interactions between the binder moiety and the target cells outside of the scFv element (data not shown). To reduce the risk of Fc receptor-mediated transduction, multiple mutations^20,21,22,23 were engineered into an optimized version of the Binder stalk to ablate Fc-mediated effector functions (Figure S1F) and were confirmed to limit transduction of a GFP-encoding vector to CD7⁺ cells (SupT1 cells and human peripheral blood mononuclear cells [PBMCs]) while having no detectable transduction of the evaluated B cell lines (Figure S1G). Further assessment of human PBMCs revealed that T cells, inclusive of both CD4⁺ and CD8⁺ subsets as well as NK cells were able to be transduced using the Gen 2.1 Fusogen and CD7 Binder (Figures S1H and S1I).

To evaluate the transduction specificity of this optimized CD7 Binder, CD7⁺ SupT1 cells were engineered to knock out CD7. When these CD7KO SupT1 cells were exposed to vector, the Gen 2.1 Fusogen in combination with the CD7 Binder did not mediate transduction while transduction mediated by vector pseudotyped with WT VSV-G was unaffected. These results confirm that transduction of this T cell line by Gen 2.1 pseudotyped vector is mediated by CD7 binding (Figure 1F).

As a platform, the Gen 2.1 Fusogen can be pseudotyped along with Binders targeting a variety of surface antigens via a number of different targeting moieties to direct binding to specific cells. To demonstrate this flexibility, vectors were pseudotyped with scFv-based Binders targeting CD7, CD4, or CD8. All three vectors were used to transduce human PBMCs activated with CD3/CD28 beads from three unique donors. Vectors made with the CD4 and CD8 Binders selectively transduced CD4⁺ and CD8⁺ cells, respectively, while the CD7 scFv-based binder transduced CD7⁺ cells, which includes both CD4⁺ and CD8⁺ T cells (Figures 1G, 1H, and S1J). Importantly no transduction occurred using vector pseudotyped with the Gen 2.1 Fusogen alone (Figure 1G). To demonstrate the flexibility of the Binder type used to direct vector transduction, CD4 Binders formatted as an scFv, Variable Heavy domain of Heavy chain (VHH), or designed ankyrin repeat protein (DARpin) were used to transduce PBMCs and functioned similarly to provide antigen-specific transduction (Figure 1H). This specificity was conserved across each of the three donors tested (Figure S1J).

INT2104 specifically transduces T and NK cells, resulting in functional CAR⁺ cells

We next utilized the vector platform to generate CAR cells. Since CD7 expression is largely restricted to T cells and NK cells,^24,25 vector pseudotyped with a combination of the Gen 2.1 Fusogen and the CD7 Binder was evaluated to generate both CAR T and NK cells. INT2104 is a vector with a CAR20 transgene encoding a fully human anti-CD20 scFv (Figure 2A) packaged into a non-replicating, self-inactivating lentiviral vector pseudotyped with the Gen 2.1 Fusogen and CD7 Binder. Treatment of activated human PBMCs with INT2104 resulted in transduction of both CD4⁺ and CD8⁺ T cells and NK cells (Figures 2B and S2A). Notably, levels of transduction were comparable to those seen during manufacture of ex vivo CAR T cell products.²⁶ Human PBMCs transduced with INT2104 showed efficient killing of two B cell lines, Raji (Figure 2C) and Daudi (Figure 2D), in a dose-dependent manner over a range of Effector:Target (E:T) ratios. In a similar killing assay, INT2104 transduced PBMCs were cocultured with CD20⁺ Raji cells and an engineered Raji line with CD20 knocked out (CD20KO). Dose-dependent depletion of B cells was again evident in cocultures with CD20⁺ Raji cells while CD20KO Raji cells were not killed following coculture with INT2104-transduced, CAR20⁺ PBMCs, confirming antigen-dependent, CD20-mediated killing (Figure 2E). In parallel to B cell cytotoxicity, cytokine production was measured with two orthogonal approaches: (1) enzyme-linked immunosorbent assay (ELISA) to measure bulk cytokine secretion into coculture supernatants and (2) intracellular cytokine staining (ICS) to measure intracellular cytokine production. Both ELISA (Figures 2F–2H and S2B–S2D) and ICS analysis (Figures 2I, 2J, S2E, and S2F) revealed increased cytokine production (interferon [IFN]-γ, tumor necrosis factor [TNF]-α, and interleukin [IL]-2) in cocultures of CAR20⁺ PBMCs and CD20-expressing Raji cells compared with CD20KO Raji and non-transduced PBMC controls. Moreover, ICS results showed enhanced cytokine production in CAR20⁺ vs. CAR⁻ T cells within the same coculture (Figure S2E) with the CD20⁺ Raji cell line, further demonstrating CAR20-mediated biological activity. Control cells with CAR20⁺ PBMCs cultured without additional stimuli (e.g., Raji B cells or PMA/Ionomycin) were included and denoted as CAR20 non-stimulated.

Figure 2.

INT2104 transduction results in functional CAR⁺ cells

(A) Schematic representation of the CAR20 RNA genome delivered by INT2104. (B) Flow cytometry results displaying CAR20 transduction in T cell subsets and NK cells following INT2104 transduction of activated human PBMCs at an MOI of 3 based on SupT1 infectious titer assay. In vitro B cell killing assay results with CAR20 cells generated by INT2104 transduction against Raji (C) and Daudi (D) B cell tumor lines across a range of E:T ratios. Data in (B)–(D) represent eight individual donors displaying mean ± SD. (E) Summary plot of B cell depletion shows CD20-mediated killing when PBMCs transduced with INT2104 were incubated with CD20-expressing B cells (Raji WT) or B cells lacking CD20 (Raji CD20KO) at the indicated E:T ratios. Data representative of mean ± SD for three individual donors. Quantification of bulk cytokine production of IFN-γ (F), TNF-α (G), and IL-2 (H) by ELISA in the cell culture supernatants of cocultures containing CAR⁺ PBMCs and either Raji or Raji CD20KO cells at a 3:1 E:T ratio. Data in (F)–(H) representative of mean ± SD of three individual donors. (I) Representative flow plots depicting IFN-γ production following 18 h coculture of INT2104-transduced PBMCs with Raji WT or Raji CD20KO B cells at a 3:1 E:T ratio. (J) Quantification of cytokine production from CD3⁺ cells following 18 h coculture with target Raji cells. Data representative of mean ± SD for two individual donors.

Assessing the risk of transducing B cells with INT2104

The unintentional transduction of B cells is a recognized safety concern²⁷ regardless of whether CAR cells are manufactured for ex vivo CAR T cell therapies or generated in vivo as intended with INT2104. In the case of a leukemia patient receiving ex vivo CAR19 therapy,²⁷ CAR expression on malignant B cells was shown to render these cells resistant to killing by CAR T cells due to epitope masking.²⁴ To assess this potential safety concern, the transduction potential of INT2104 was tested in primary B cells isolated from healthy human donors, PBMCs isolated from donors with mantle cell lymphoma (MCL), follicular lymphoma (FL), or chronic lymphocytic leukemia (CLL), and B cell tumor-derived cell lines. The vector used for these studies delivered a bicistronic reporter transgene expressing both the CAR20 and GFP (CAR20-T2A-GFP) to maximize the ability to detect transduction of B cells. CD20 expression was confirmed in all primary B cell cultures and six of seven B cell lines, with Nalm6 lacking CD20 expression (Table S1). SupT1 cells were included as a CD20⁻ control. When primary B cells isolated from healthy human donors were exposed to vector at a wide range of multiplicities of infection (MOI), represented in this instance as the number of vector particles per cell, B cells were not transduced, even at high MOIs (Figure 3A). Similarly, the B cell lines evaluated including Nalm6, DB, VL51, JVM13, and Rec1 were inefficiently transduced across a wide range of MOIs tested, with GFP⁺ cells detected only at supraphysiological MOI (approximately 1 × 10⁵ particles/cell) (Figure 3B), mimicking data where B cell lines failed to be transduced with GFP-encoding vectors (Figure S1G). Similarly, B cell samples derived from patients with MCL, FL, or CLL also showed no signs of B cell transduction (Figure 3C).

Figure 3.

Assessing the risk of transducing B cells with INT2104

The frequency of GFP⁺ B cells 7 days after incubation with INT2104 isolated from primary healthy donors (A), immortalized B cell lines (B), or isolated from patients with B cell malignancies (C). Data representative of individual measures at day 7 post transduction. (D) Schematic depicting WT VSV-G pseudotyped vector used to generate DB cells expressing either CAR20&GFP or GFP alone. The transduced DB cells were then incubated for 48 h with PBMCs transduced with INT2104. (E) In vitro B cell killing assay results with INT2104-transduced PBMCs against GFP- or CAR20&GFP-expressing DB B cell tumor lines showing that CAR20 expressed on DB cells did not hinder the ability of transduced PBMCs to kill the cells. (F) Frequencies of transduced (GFP⁺) DB cells remaining in culture following coculture with CAR⁺ PBMCs. Data in (E) and (F) representative of four individual PBMC donors displaying mean ± SD.

INT2104-transduced PBMCs can kill B cells expressing CAR20

A further study was conducted to investigate whether unintentionally transduced B cells expressing CAR20 would remain susceptible to killing by INT2104 transduced PBMCs. Vector pseudotyped with WT VSV-G was used to transduce a malignant B cell tumor line (DB cells) to express either the CAR20 and GFP, or GFP alone. The resulting culture contained both transduced (CAR20 and/or GFP-expressing B cells), as well as non-transduced cells. INT2104-transduced PBMCs from four human donors were then cocultured with the DB cell lines and CAR-mediated cytotoxicity was evaluated (Figure 3D). INT2104-transduced PBMCs killed DB cells in a dose-dependent manner, regardless of whether they expressed CAR20 and GFP or GFP alone (Figure 3E). Across the range of E:T ratios tested, there was evidence to suggest an increase in the frequency of CAR⁺GFP⁺ cells (detected by GFP expression) among surviving DB cells in cocultures relative to transduced “DB alone” controls (Figure 3F). These results suggest that B cells expressing CAR20 remain susceptible to killing by CAR20 effector cells generated with INT2104.

Intravenous administration of INT2104 to human CD34-engrafted mice results in CAR⁺ T and NK cells and B cell depletion

The CD7 Binder on INT2104 was designed to enable the targeting of both T and NK cells in vivo following systemic administration. To assess whether both CAR⁺ T and NK cells were generated in vivo following intravenous administration, human CD34⁺ hematopoietic stem cell-engrafted NSG or NSG variant mice were used as a model system. CD34-engrafted NSG mice develop multi-lineage human immune cells, including human B, T, and NK cells, with T cell activation levels of ∼5%–10% in peripheral blood, and are a widely used platform for immunology or immuno-oncology studies.^28,29 In the first study, two donors of CD34-engrafted NSG mice were utilized and confirmed to have engraftment of human (CD45⁺) cells (Figure 4A). Donor-specific differences were seen in B and T cell distributions, with mice from one donor having comparable frequencies of B and T cells in blood while mice engrafted with a second donor had a higher frequency of B cells than T cells (Figure 4B). T cell activation levels were comparable between donors (Figure 4C) and each donor had similar low percentages of NK cells in blood (Figure 4D). On day 0, INT2104 was administered intravenously at either a low dose of 8.0 × 10⁶ transducing units (TU)/mouse or at a high dose of 2.5 × 10⁷ TU/mouse (n = 5/group) by tail vein injection (Figure S3A). Five mice not receiving vector were maintained as controls. Submandibular bleeds were performed weekly for the first 4 weeks after vector administration and once more on day 47. These samples were assessed for CAR⁺ cells and B cell depletion by flow cytometry (Figure S3B). The percentage of circulating B cells among human cells (hCD45⁺) decreased in the treated mice relative to untreated mice by day 7 post-injection, independent of INT2104 dose or donor (Figure 4E). By day 21, circulating CD20⁺ B cells had been eliminated in all vector-treated mice and remained below detection through day 47. This pattern held whether B cells were measured by CD20⁺ staining (Figure 4E) or by CD19⁺ (Figure S3C), indicating true depletion rather than an inability to detect CD20 on B cells due to downregulation of CD20 from the cell surface and/or epitope masking by CAR cell binding. The proportion of CAR20 cells in the circulation varied from mouse to mouse over time; however, most treated mice showed detectable CAR20 cells in the blood by day 21 (Figure 4F), inclusive of both CD4⁺ and CD8⁺ cells (Figure S3D). Due to low NK cell engraftment in this model, reliable quantitation of transduced NK cells was not feasible. Treatment with INT2104 was well tolerated by mice, with all mice in treated groups surviving for the duration of the study and no weight loss events attributed to vector treatment (Figure S3E).

Figure 4.

Intravenous delivery of INT2104 to humanized mice results in CAR⁺ T and NK cells and B cell depletion

Flow cytometry was used to assess the proportion of human cells in peripheral blood (A), B and T cells (B), activated T cells (C), and NK cells (D) in human CD34-engrafted mice prior to INT2104 administration. Each point in (A)–(D) indicates a single mouse with the color indicating the donor with data representing mean ± SD. (E) Human CD20⁺ B cells in mouse blood samples were tracked over time using flow cytometry. Data are shown as a percentage of the total human cell population representing mean ± SD of the indicated treatment group. (F) Frequency of CAR⁺ cells in blood among total human cell population with data representative of mean ± SD of the indicated treatment group. Flow cytometry was used to assess the proportion of human cells in blood (G), B and T cells (H), activated T cells (I), and NK cells (J) in Hu-IL15 NSG mice engrafted with CD34⁺ cells mice prior to INT2104 administration. Each point in (G)–(J) indicates a single mouse with the color indicating the donor with data representing mean ± SD. (K) Human CD20⁺ B cells in mouse blood samples were tracked over time using flow cytometry. Data are shown as a percentage of the total human cell population representing mean ± SD of the indicated treatment group. (L) Frequency of CAR⁺ T cells in blood among total human cell population with data representative of mean ± SD of the indicated treatment group. Assessment of bone marrow, spleen, and liver of INT2104-treated animals results in evidence of CAR⁺ T cells (M) and CAR⁺ NK cells (N) 14 days post vector administration. Data in (M) and (N) represent mean ± SD of the INT2104-treated animals with each symbol representing a different mouse and colors denoting different donors.

A follow-up study was performed in NSG-Tg (human interleukin 15 [Hu-IL15]) mice, a strain that produces human IL-15 to promote a higher level of NK cell engraftment than standard NSG mice.³⁰ Similar to the previous study, Hu-IL15 NSG mice were engrafted with CD34⁺ cells from either of two human donors. In this study, donors had similar levels of engraftment, with comparable B and T cell engraftment between donors, and T cell activation of ∼12% (Figures 4G–4I). The fraction of NK cells within the hCD45⁺ cell population was higher (Figure 4J) (average of 4.41% and 7.46% for the two donors) compared with the ∼1% observed in the standard CD34-engrafted NSG model (Figure 4D). On day 0, INT2104 was administered as a single intravenous dose of 1 × 10⁷ TU/mouse (n = 7/donor) by tail vein injection compared with no vector controls (n = 3/donor). The mice were monitored for up to 35 days to assess the effects of INT2104 intravenous infusion. Weekly blood samples were assessed for CAR⁺ cells and B cell depletion by flow cytometry. On study days 7, 14, and 35, a subset of mice from each group was euthanized and bone marrow, spleen, and liver were collected to assess CAR⁺ cells by flow cytometry (Figures S4A and S4B). Similar to what was seen with the non-transgenic CD34-engrafted NSG mouse model (Figure 4E), the percentage of circulating B cells decreased in the treated mice relative to untreated mice by day 7 post-injection regardless of human donor (Figure 4K) and remained below detection through the remainder of the study. CAR⁺ T cells were detected in blood of INT2104-treated mice, coincident with the observed B cell depletion, and remained detectable throughout the study (Figure 4L). CAR⁺ T cells were also detected in the spleen, bone marrow, and liver of INT2104-treated animals (Figures 4M and S4C). While CAR⁺ NK cells were detected but not quantifiable in blood, CAR⁺ NK cells were quantifiable in the spleen, bone marrow, and liver of vector-treated animals (Figures 4N and S4D). INT2104 administration was well tolerated in this model, with all mice surviving until their scheduled necropsy and no weight loss events attributed to vector treatment (Figure S4E).

Infusion with INT2104 results in clearance of established tumors

Last, the functionality of CAR cells generated in vivo was evaluated by testing their capacity to eliminate established B cell tumors in mice. For this assessment, NSG major histocompatibility complex (MHC) I/II knockout (KO) mice,³¹ which are severely immunodeficient and harbor additional mutations that allow for human cell engraftment with delayed graft-vs.-host disease (GVHD), were first engrafted with luciferase-expressing Raji cells to establish a B cell tumor burden. Five days later (day −1), 1 × 10⁷ activated human PBMCs from three individual human donors were engrafted, followed by INT2104 administration the next day (day 0) (Figure S5A). Raji tumor cells, PBMCs, and INT2104 were all administered intravenously by tail vein injections. INT2104 was infused at a high dose of 1 × 10⁷ TU/mouse into mice engrafted with PBMCs from donors 1, 2, and 3 and at a low dose of 6 × 10⁵ TU/mouse in mice engrafted with donor 1 PBMCs. INT2104-treated mice were compared with donor-matched control tumor-bearing mice that received no vector (denoted as “PBMCs Only”) and tumor-only controls (without PBMCs or vector, denoted as “Raji Only”) to confirm CAR-mediated B cell depletion.

All mice developed established Raji B cell tumors prior to treatment with INT2104 as measured by evaluation of total flux in images obtained by IVIS (Figures 5A and 5B). The established tumors were eliminated by day 22 post treatment in all 15 mice treated with a high dose of INT2104 (1 × 10⁷ TU/mouse) across all three donors, but not in the control mice engrafted with activated PBMCs (Figures 5A and 5B). All five mice treated with the low dose INT2104 (6 × 10⁵ TU/mouse) also eliminated the established tumors, but at a slower rate than the high-dose groups, with all tumors eliminated by day 26. Weekly blood samples were assessed for CAR⁺ cells by flow cytometry (Figure S5B). CAR⁺ cells could be measured in the peripheral blood starting at day 14 and coincident with B cell depletion for the high-dose groups (Figures 5C–5E). Quantifiable circulating CAR⁺ cells were observed in the blood of only two of five mice in the low-dose group. Treatment with INT2104 was well tolerated by mice, with all mice in the INT2104-treated groups surviving for the duration of the study and no weight loss events attributed to the infused vector (Figure S5C). These data suggest that INT2104 can result in the clearance of established tumors following a single intravenous administration to humanized mice.

Figure 5.

In vivo delivery of INT2104 to mice results in killing of established B cell tumors

(A) Raji-fLuc-eGFP xenografted mice were imaged at the indicated time points by IVIS. Tumor burden is indicated as radiance (photons/second/cm²/steradian) and visualized as a log-scale heatmap superimposed over a light image of the mice scanned. Scale bar on the right indicates the correspondence between color and radiance value. (B) Tumor burden is indicated as total flux (photons/s) measured by IVIS at each time point presented as mean flux ±SD for each treatment group. The No Tumor group represents five mice used as assay controls that had not been engrafted with Raji-fLuc-eGFP cells, PBMCs, or INT2104 to inform the limit of detection. The PBMCs Only and 1.0 × 10⁷ TU groups represent the mean across each donor. (C–E) CAR20 cells in mouse blood samples were tracked over time using flow cytometry with data shown as a percentage of the total human CD45⁺ cell population and each line representative of individual mice. (C)–(E) represent mice engrafted with a single donor. INT2104-treated mice were denoted with treatment (“1.0 × 10⁷ TU” or “6.0 × 10⁵ TU”) while donor-matched control tumor-bearing mice that received no vector were denoted as “PBMCs Only” and tumor-only controls (without PBMCs or vector) were denoted as “Raji Only.”

Intravenous administration of INT2104 to cynomolgus macaques results in specific transduction of T and NK cells

After showing intravenous administration of INT2104 to humanized mice generates functional CAR⁺ cells, the cynomolgus macaque model was utilized to further assess targeting specificity in a large animal model. This study was enabled by the cross reactivities of both the CD7 Binder on INT2104 and the encoded CAR20 with the homologous proteins expressed by cynomolgus NHPs. However, INT2104, the clinical candidate, is manufactured with the WT human immunodeficiency virus type 1 (HIV-1) gag-pol plasmid and cynomolgus macaques have host cell restrictions against HIV (e.g., TRIM5α^32,33). Previous studies have utilized a chimeric gag-pol consisting of HIV and simian immunodeficiency virus (SIV) sequences to mitigate these innate host cell restrictions and enhance cell transduction in this model.^34,35,36,37 Based on this strategy, a surrogate vector was manufactured using a chimeric HIV/SIV gag-pol (xHIV) for this NHP study (Figure 6A). This INT2104 Surrogate Vector differs from INT2104 only within the region of gag encoding the capsid but maintains the same Gen 2.1 Fusogen, CD7 Binder, and CAR20 transgene as INT2104.

Figure 6.

Intravenous administration of INT2104 to cynomolgus macaques results in specific transduction of T and NK cells

(A) Schematic representation of the INT2104 Surrogate Vector. (B) Timeline representing the dosing of animals and subsequent necropsy for biodistribution. (C) Depletion of CD20⁺ B cells was noted in peripheral blood of vector-treated NHPs at the time of necropsy. Representative immunofluorescence images of stained formalin-fixed paraffin-embedded (FFPE) samples from no vector control (left panels) and INT Surrogate Vector-treated animals (right panels) showing CAR expression on splenic T cells (D), splenic NK cells (E), and liver T cells (F). No NK cells were identified in the liver. (G) Representative immunofluorescence images of stained FFPE liver samples from no vector control (left panels) and INT Surrogate Vector-treated animals (right panels) showing no CAR-expressing hepatocytes in the liver. T cells were identified as CD3⁺, NK cells were identified as NKG2A⁺ cells, and hepatocytes were identified as HNF4α⁺. CAR⁺ cells were identified using an anti-idiotype monoclonal antibody specific for the OFA-CAR. Images shown were collected using a 40× objective with the scale bar representing 20 μm. Single color panels are digital enlargements of the delineated areas.

To assess the potential for targeted transduction of T and NK cells in a large animal model, two adult cynomolgus macaques were infused with 2 × 10⁹ TU of the INT2104 Surrogate Vector (∼3.5 × 10⁸ TU/kg) on day 0. The treated animals were monitored for 4 days post treatment before their scheduled necropsies on day 4 post vector administration (Figure 6B) and compared with an untreated, male control animal. This early necropsy was scheduled to allow identification of transduced cell types before any potential adaptive immune response against the fully human CAR led to cell elimination. The onset of B cell depletion was noted at the time of the scheduled necropsy (day 4 post vector administration), with both treated NHPs showing greater than 50% reduction in CD20⁺ B cells in peripheral blood (84% reduction in animal V690 and 53% reduction in animal V745) (Figure 6C).

Biodistribution was assessed by measuring proviral DNA genomes on day 4 by droplet digital PCR (ddPCR) in PBMCs, bone marrow mononuclear cells (BMMCs), and additional tissues including spleen, liver, and lung. Biodistribution analysis revealed highest levels of provirus DNA in tissues known to harbor CD7⁺ cells including spleen, liver, PBMCs, bone marrow, and lung on day 4 at the scheduled necropsy (Table S2). Based on this biodistribution analysis, immunofluorescent histological methods were used to investigate which cells were expressing CAR protein within the liver, spleen, and lung.

Consistent with vector targeting to CD7⁺ cells, CAR20 protein was only detected on T and NK cells (Figures 6D–6G; Table 1). T cells (defined as CD3⁺) expressing CAR were detected in both the spleen and liver of both vector-treated NHPs (Figures 6D and 6F). NK cells (defined as NKG2A⁺) expressing CAR protein were also detected in the spleens of both vector-treated animals (Figure 6E). No NK cells were identified in liver sections. Importantly, no hepatocytes (defined as HNF4α⁺) in the liver were found to be expressing CAR protein (Figure 6G). CAR⁺ cells were also not detected among the evaluated B cells (CD19⁺ or CD20⁺) in the spleen. Macrophages are known scavengers that can phagocytose particulate materials and have been previously reported to take up lentiviruses.^38,39,40 In this study, no CAR20 protein was detected in macrophages in either the spleen or liver of treated animals (Table 1). Consistent with previous reports, evidence of uptake of lentiviral particles was detected in liver macrophages of treated animals using RNAscope (Figure S6; Table S3). No cells expressing the CAR protein were detected in sections of lungs from either of the two vector-treated NHPs (Table 1). Prior to their necropsies on day 4 post vector administration, no vector-associated toxicity or safety signals were noted in either of the vector-treated animals. Consistent with the data generated in mice, these biodistribution data suggest that the pseudotyping of INT2104 with the combinations of the CD7 Binder and Gen 2.1 Fusogen enables specific targeting of T and NK cells following intravenous administration.

Table 1.

Summary of CAR protein expression at day 4 in spleen, liver, and lung by histology

Cell type	Marker	No vector control		INT surrogate vector		INT surrogate vector
		V735		V745		V690
		Total cellsa	CAR+ cellsb	Total cellsa	CAR+ cellsb	Total cellsa	CAR+ cellsb
Spleen
T cells	CD3	3,568	0	4,685	21	2,821	6
NK cells	NKG2A	644	0	622	9	718	9
B cells	CD19	2,017	0	5,035	0	4,379	0
	CD20	4,335	0	5,266	0	4,542	0
Macrophages	CD68	1,012	0	694	0	3,037	0
Liver
T cells	CD3	189	0	152	42	187	9
Macrophages	CD68	909	0	921	0	1,630	0
Hepatocytes	HNF4α	6,435	0	8,685	0	9,914	0
Lung
All cells	DAPI	196,848	0	438,815	0	192,925	0

^aTotal cells indicates the number of Marker⁺ cells (e.g., CD3⁺) observed.

^bCAR⁺ cells indicates the number of Marker⁺ cells colocalized with CAR signal.

Discussion

Although autologous CAR T cell products have become the standard of care in the treatment of B cell malignancies and multiple myeloma,^{41,42,43,44,45,46,47} there are significant challenges to broad adoption including the complexity of manufacturing a patient-specific cell product, the requirement for lymphodepleting chemotherapy prior to product administration, and the cost and patient access of such a personalized medicine. With advancements in targeted delivery systems and vector engineering, the potential for in vivo CAR T cell therapy is now being realized, with a variety of techniques utilized to attempt specific delivery to target cells.^48,49,50,51 Here, we describe a lentivirus platform where a detargeted fusogen based on VSV-G was utilized in combination with a Binder to pseudotype lentiviral particles that specifically bind and transduce target cells. Since the binding and fusion functions are biologically decoupled, this platform is modular in nature, allowing for any number of binder types or cell targets to be incorporated (Figures 1G and 1H). Since the Gen 2.1 Fusogen no longer binds its native receptor, LDL-R, there is a reduced risk of off-target transduction in unintended cells, evidenced by histological assessment in the NHP model (Figure 6). Incorporation of two additional amino acid substitutions in the Gen 2.1 Fusogen supports intravenous administration by improving serum stability and complement resistance of the vector. Therefore, utilization of the Gen 2.1 Fusogen allows for systemic administration of the engineered INT2104 product whereas other in vivo delivery systems (e.g., lipid nanoparticles, lentiviruses with fusogens that are not detargeted) may suffer from off-target uptake, binding, or transduction and thus be limited to localized injection routes, diminishing access to target cells and complicating clinical use.

Recent data have also shown that CAR T cell product efficacy is correlated with reduced ex vivo manipulation of cells,⁵² prompting efforts toward rapid manufacturing to limit cell differentiation.^8,10 These data suggest that CAR expression on more naive or stem-like T cell phenotypes would be beneficial for improved product performance. Additionally, ex vivo manufactured CAR⁺ NK cells have shown clinical efficacy.⁵³ Thus, CD7 was strategically selected for INT2104 as the target for the binder moiety to enable the transduction of both T cells and NK cells in vivo. As CD7 is highly expressed on less-differentiated T cells,⁵⁴ vector binding to more naive and stem-like T cells may be preferentially favored. This selection of CD7 as a vector binding target may be beneficial over other cell surface markers that limit access to T cell subsets (e.g., CD4 or CD8) or a pan T cell marker that induces T cell activation (e.g., CD3), increasing the risk that adsorption of particles present in lentiviral products (both non-infectious and infectious alike) lead to uncontrolled T cell activation and potentially safety signals (e.g., cytokine release syndrome [CRS]).⁵⁰

Off-target transduction of B cells via binding of the CAR moiety that may be present on vector particles is a recognized safety risk,²⁷ whether that transduction occurs with ex vivo (e.g., commercially available CAR T cell therapies) or in vivo (INT2104) generated products. While transduction of primary B cells may be inefficient with VSV-G-based lentiviruses⁵⁵ and a rare event, INT2104 was nonetheless evaluated using several approaches to assess the magnitude of this risk. In vitro evaluation of B cells, derived from a variety of sources including primary healthy human donors, human donors with a subset of B cell malignancies, or a wide range of immortalized B cell tumor lines (Figure 3A–3C), showed B cells were not transduced across a wide range of MOIs tested with INT2104. We further confirmed that B cells engineered to express the CAR20 using a standard VSV-G lentivirus could still be recognized and killed by INT2104-transduced PBMCs (Figures 3D–3F). Since CAR/antigen interactions on transduced B cells could in theory mask the transduced B cell, protecting it from CAR cell killing, the lack of an increase in GFP⁺ cells following coculture suggests that this phenomenon did not occur in the tested setting. While the risk of off-target transduction cannot be completely eliminated, these studies suggest that the risk of B cell transduction by INT2104 is low.

Intravenous administration of INT2104 is enabled by the novel lentiviral platform providing a means to transduce target effector cells that can be found in the circulation as well as in tissues (e.g., lymph nodes, spleen). An added benefit of intravenous delivery is the simplicity by which this administration route can be adopted in the clinical setting, providing a path forward for in vivo genetic medicine to be translated to a broad patient population. The vector engineering and design decisions reported here have enabled intravenous administration, which was demonstrated using in vivo models. In three humanized mouse models, INT2104 was delivered as a single injection via the tail vein to mice and was well tolerated. Compared with another study that reported CRS in mice following in vivo CAR19 generation,¹³ there were no vector-related adverse events or safety signals observed following INT2104 administration. Importantly, transduction of target human cells was seen across all mouse models, with the detection of CAR cells (both T and NK cells) approximately 2 weeks post INT2104 administration (Figures 4 and 5). B cell depletion, either the engrafted normal human B cells (Figures 4E and 4K) or established B cell tumors (Figure 5B), was evident in all models coinciding with the detection of CAR⁺ cells in the circulation.

The NHP model served as a representative model to assess vector targeting following intravenous administration of vector. Intravenous administration of the INT2104 Surrogate Vector at a dose of 2 × 10⁹ TU/animal was well tolerated with no resulting safety signals, consistent with unpublished results from the GLP Toxicology study for INT2104. Day 4 post treatment was selected to identify transduced cell types prior to any potential adaptive immune response against cells expressing the fully human CAR. At this early time point, data for both vector-treated animals suggested evidence of reduced B cell counts although circulating CAR⁺ cells were not detected by flow cytometry (Figure 6C). This cynomolgus macaque model also provided a means to assess the specificity by which INT2104 can transduce target CD7⁺ cells in a model system that more closely mimics the human environment. In this setting, CAR⁺ cells were detectable in liver and spleen in target T lymphocytes and NK cells.

Importantly, no evidence of CAR protein expression was seen by other cells in the spleen, liver, and lung, including B cells and hepatocytes, indicating specific transduction of targeted cells was achieved following intravenous administration (Figures 6D–6G). Consistent with previous reports,^38,39,40 some amount of particle uptake by macrophages is expected and was detectable in liver macrophages using RNAscope (Figure S6; Table S3). However, no CAR protein was detectable on macrophages in either the liver or spleen of treated NHPs (Table 1). This may be due in part to macrophages having high levels of SAMHD1, a known host restriction factor that limits the infectivity of phagocytic cells.^56,57,58

Here we show evidence that lentiviral vectors, pseudotyped with the Gen 2.1 Fusogen and a CD7 Binder, can transduce T and NK cells in vivo, suggesting the potential for this lentiviral platform to be utilized for targeted in vivo transduction of other cells of interest. A modular lentiviral platform holds the potential to treat a wide variety of diseases beyond hematological cancers, including solid tumors, autoimmune disorders, and genetic diseases. Application of the lentiviral platform into other disease areas could require the development of novel binders specific to cell surface molecules on the target cells and the optimization of the genetic cargo to accomplish the required biological effect with an acceptable safety profile. In many settings, the persistence of transduced cells will be important and require the use of multi-transgene constructs that enhance the expansion and/or persistence of the transduced cells. Clinical data for INT2104 are expected to demonstrate the potential of this novel lentiviral platform providing a foundation for expanding its application beyond B cell malignancies.

Materials and methods

Primary cells and cell lines

Primary healthy human PBMCs were procured from AllCells (Discovery Life Sciences). HEK293T, Raji, Ramos, Daudi, GA10, DB, HT, and JVM-13 cells were purchased from American Type Culture Collection (ATCC). Raji cells expressing GFP and firefly luciferase (Raji-fLuc-EmGFP cells) were purchased from Imanis Life Sciences. SLVL (VL51) cells were purchased from JCRB (Japanese Collection of Research Bioresources)/FUJIFILM. SupT1 and Raji cells were engineered using CRISPR-Cas9 editing to generate SupT1 CD7KO and Raji CD20KO cells, respectively. Primary PBMCs isolated from patients with MCL, CLL, and FL were purchased from StemCell, AllCells (Discovery Life Sciences), or Precision for Medicine. Cryopreserved human PBMCs were thawed and activated with CD3/CD28 DynaBeads. Forty-eight hours later, DynaBeads were magnetically removed from cells and cells were used for transductions. To generate CAR20 and/or GFP-expressing DB cells, WT VSV-G pseudotyped vector carrying the respective transgenes was utilized and both GFP and CAR20 expression were confirmed prior to use.

Vector design and production

Lentiviral vectors were prepared using a five-plasmid transient transfection protocol. Briefly, 24 h prior to transfection, HEK293T cells were seeded and the next day, transfection using PEIpro was conducted. Forty-eight hours later, supernatant was collected, centrifuged to remove cellular debris, and then centrifuged over sucrose overnight prior to resuspension of vector particles in formulation buffer before use in downstream assays. Depending on the experiment, transgene plasmids encoded either GFP, CAR20, or CAR20-T2A-GFP. In some indicated instances, vectors were produced without the CD7 Binder or with alternative binders (e.g., CD4 scFV, CD4 VHH, CD4 DARPin, CD8 scFv) or by replacing the Gen 2.1 Fusogen with WT VSV-G. Vector particles were quantified using p24 ELISA kits (Takara or ABL) to determine p24 protein content per volume then converted to particle numbers with the conversion factor where 1 ng of p24 is equal to 1.25 × 10⁷ lentiviral particles.

Infectious titer assay

Vector potency was assessed via serial dilution of vector on target cells. Briefly, vector dilutions were added to SupT1 cells for 4–6 h prior to a full media exchange. At designated time points after vector addition (3–7 days depending on the assay), cells were collected and analyzed via flow cytometry to quantify the percentage of transduced cells expressing transgene. Functional titer (TU/mL) at each dilution is calculated as the (% of transgene⁺ cells multiplied by the number of cells plated)/volume of vector added to cells. TU/mL was calculated from wells with between 2% and 20% transgene positivity to ensure linearity; for titrations with multiple wells in this range, an average was calculated.

Flow cytometry and cytokine analysis

Flow cytometry was performed to evaluate the percentage of transduced cells among an indicated population. Quantibrite PE beads (BD) were used per the manufacturer’s instructions to enumerate CD20 surface expression on select cell types. For ICS experiments, on day 7 post transduction, four conditions were plated: PBMC-only control, PBMCs cocultured with Raji CD20KO cells, PBMCs cocultured with WT Raji cells, and PBMCs treated with a cell stimulation reagent of a cocktail of 50 ng/mL phorbol 12-myristate 13-acetate (PMA) and 1 μg/mL Ionomycin. Cocultures were plated at an effector-to-target (E:T) ratio of 3:1, with 120,000 effector cells to 40,000 target cells. Two hours post plating, brefeldin A (BFA) was added at a final concentration of 1 μg/mL. Eighteen hours post plating, samples were subjected to flow cytometric analysis. Surface staining was performed followed by fixation and permeabilization. Intracellular staining for CAR20 and cytokines of interest (e.g., TNF-α, IFN-γ, and IL-2) was then performed. A CytoFLEX S cytometer was used for all flow cytometry analyses.

To assess bulk cytokines secreted into coculture supernatants, human TNF-α, IFN-γ, or IL-2 ELISA kits (Invitrogen) were used per the manufacturer’s instructions with diluted supernatant and absorbance (OD 450 nm) was measured using a Varioskan LUX microplate reader.

In vitro B cell killing assay

Human PBMCs transduced with INT2104 were cocultured with B cell tumor cell targets. Briefly, transduced PBMCs from multiple human donors were evaluated for CAR expression via flow cytometry as outlined above. The percentage of CAR⁺ cells was normalized across each experiment by adding non-transduced, donor-matched PBMCs to transduced cells so that all donors had the same percentage of CAR⁺ cells among the same total cell numbers. PBMCs containing CAR⁺ cells were then added to cultures containing B cells at specified E:T ratios. Forty-eight hours later, B cell depletion was determined via flow cytometry by quantifying the remaining B cells in culture normalized to a non-transduced PBMC control where B cell cytotoxicity was equal to (1 − (% B cells in transduced PBMC cultures)/(% B cells in donor-matched non-transduced PBMC cultures)).

In vivo administration of INT2104 to humanized mice

All animal studies were approved by the Saint Joseph’s University (formerly University of the Sciences) Institutional Animal Care and Use Committee (IACUC) under protocol #21-002 and conducted in accordance with local, state, and federal guidelines. For CD34-engrafted NSG mouse studies, 30 female mice at 20 weeks old were assigned to the study. Mice were engrafted with CD34⁺ cells from two donors by the vendor (Jackson Labs, Maine, USA). Two days prior (day −2) to INT2104 treatment, blood samples were taken from all animals to characterize the baseline levels of human cell engraftment. On day 0, INT2104 was administered as a single intravenous dose at either a low dose of 8.0 × 10⁶ transducing units (TU)/mouse or at a high dose of 2.5 × 10⁷ TU/mouse (n = 5/group) by tail vein injection. Five mice not receiving vector were maintained as controls. The mice were monitored for up to 47 days. Submandibular bleeds were performed weekly for 4 weeks and on day 47 and these samples were assessed for CAR⁺ cells and B cell depletion by flow cytometry.

For Hu-IL15 CD34-engrafted mouse studies, 20 female 19- to 20-week-old Hu-IL15 mice were engrafted with cells from either of two human CD34⁺ cell donors (10 mice/donor) prior to shipment. On day −2, blood samples were collected to ensure baseline engraftment of human lymphocytes. On day 0, INT2104 was administered as a single intravenous dose of 1 × 10⁷ TU/mouse (n = 7/donor) by tail vein injection compared with no vector controls (n = 3/donor). The mice were monitored for up to 35 days to assess the effects of INT2104 treatment. Weekly blood samples were assessed for CAR⁺ cells and B cell depletion by flow cytometry. On study days 7, 14, and 35, a subset of mice from each group was euthanized and bone marrow, spleen, and liver were collected to assess CAR⁺ cells by flow cytometry.

For tumor studies, 40 NSG MHC I/II KO mice (five females/group) were first engrafted with 1 × 10⁶ luciferase-expressing B cells (Raji-Fluc-eGFP [enhanced green fluorescent protein]) to establish a tumor burden. Five days later (day −1), 1 × 10⁷ activated human PBMCs from three individual human donors were engrafted, followed by INT2104 administration the next day (day 0). Raji tumor cells, PBMCs, and INT2104 were administered intravenously by tail vein injection. Weekly blood samples were assessed for CAR⁺ cells by flow cytometry. Tumor burden was monitored by measuring the total flux in images obtained by IVIS.

For all mouse studies, mice were monitored for morbidity, mortality, and other clinical signs (hunched posture, hindlimb paralysis, body weight change) throughout the study. Body weights were measured weekly until study termination.

Tissue preparation and flow cytometry

At designated time points, cell suspensions, blood, or other tissue were processed for analysis. For mouse blood collection and processing, blood was collected via submandibular bleeding into an EDTA-coated tube. For tissue, samples were processed to single cell suspensions via mechanical dissociation through a cell strainer. Red blood cells were lysed with ACK lysis buffer prior to staining with viability dye and designated antibodies. Flow cytometry was performed using a Beckman Coulter CytoFLEX S system.

In vivo administration to NHP and tissue collection

All study procedures were conducted with regard to animal welfare and procedures in accordance with BIOQUAL Inc. IACUC protocol number 22-016P, which is compliant with the US Department of Agriculture regulations and the Animal Welfare Act. Three male cynomolgus macaques were included in this study with two animals (V690, V745) treated with the INT2104 Surrogate Vector at a dose of 2 × 10⁹ TU/animal and one animal (V735) was left untreated as a control. Vector administration was conducted as intravenous injection while animals were sedated with 10 mg/kg of Ketamine HCl. Blood was collected from anesthetized animals using femoral venipuncture. Bone marrow aspirate was collected from sedated animals from the sternum, the iliac crest, and the intertrochanteric fossa of the femur by a qualified veterinarian. At necropsy, tissue samples were collected using RNAse-free techniques then fixed in 4% PFA for 24 h, transferred to 70% ethanol, and processed for paraffin embedding. For flow cytometry, tissues were processed to single cell suspensions and analyzed as above.

ddPCR analysis of NHP tissue

DNA was isolated from identified tissues and normalized amounts of DNA were loaded into ddPCR reactions. Vector sequence was amplified by targeting the woodchuck hepatitis virus post-transcriptional regulatory element (WPRE). The WPRE assay was multiplexed with an assay targeting exon 8 of the cynomolgus albumin gene. Data were collected with a Bio-Rad QX200 droplet reader. Vector copies are expressed as copies per cell (copies/cell) and copies of vector per microgram of genomic DNA (copy/μg), with the mass of DNA calculated from the number of albumin copies.

Immunofluorescent analysis of NHP tissue

Tissue sections taken from formalin-fixed paraffin blocks were mounted on glass slides. Briefly, the slides were deparaffinized in xylene and rehydrated through a gradient of ethanol solutions followed by pressure cooker antigen retrieval in Borg Decloaker RTU. After washing with water and TBST (TBS-0.1% tween 20), the slides were blocked with 5% donkey serum in CAS-Block, followed by incubation with primary antibodies at 4°C overnight in a wet chamber. The slides were then washed and incubated with secondary antibodies for 2 h at room temperature. After washing with TBST, the slides were treated with ReadyProbes Tissue Autofluorescence Quenching Kit for 5 min then washed in PBS. For nuclear staining, the slides were incubated in DAPI for 5 min at room temperature and then washed. Stained sections were then mounted with ProLong Diamond Antifade Mountant prior to imaging with either an Agilent BioTek Cytation 7 or Leica Stellaris.

For select tissue, immunofluorescent analysis was combined with in situ hybridization utilizing RNAscope. Briefly, fresh-frozen tissue sections in 10-μm thickness were fixed in freshly made 4% PFA for 60 min followed by dehydration in an ethanol series. Tissue sections were then treated with RNAscope Protease IV (ACD) for 30 min at room temperature prior to hybridization with a proprietary RNAscope probe specific to WPRE for 2 h at 40°C in the HybEZ hybridization oven (ACD). Signal was amplified using RNAscope HiPlex Amp reagents then stained with RNAscope HiPlex Fluor T1–T4 reagents. Immediately after RNAscope HiPlex Fluor T1–T4 staining, slides were blocked with 5% donkey serum in CAS-Block for 30 min at room temperature, followed by incubation with the primary antibodies at 4°C overnight in a wet chamber. The slides were then washed and incubated with secondary antibodies for 2 h at room temperature. After washing with TBST, the slides were treated with ReadyProbes Tissue Autofluorescence Quenching Kit for 5 min then washed in PBS. For nuclear staining, the slides were incubated in DAPI for 5 min at room temperature and then washed followed by mounting with a ProLong Diamond Antifade Mountant.

Data availability

Requests for raw data will be considered by the corresponding author.

Acknowledgments

The authors acknowledge the Interius Discovery, Process Development, Drug Product Development, and Analytical Development teams in support of these studies. The authors acknowledge Dr. Saar Gill for his support and scientific guidance. The authors acknowledge St. Joseph’s University (formerly University of the Sciences) and BIOQUAL Inc. for their support in conducting animal studies. Some figures in this manuscript were generated with BioRender. Crystal structures were generated using the PyMOL Molecular Graphics System, Version 3.0 Schrödinger, LLC.

Author contributions

J.I.A., R.M.R., B.C.S., A.W., and L.J.C. designed and executed the in vitro experiments. J.I.A., D.G., and K.F.M. designed and executed the in vivo mouse experiments. J.I.A., K.F.M., T.O., and L.K.-C. designed, coordinated, and analyzed data relating to the NHP experiment. J.I.A., T.D.C., and P.R.J. designed the research and drafted the manuscript with input from all authors.

Declaration of interests

All authors of this paper are paid employees of Interius BioTherapeutics, Inc. and hold equity in the company. J.I.A., R.M.R., B.C.S., T.D.C., and P.R.J. are listed as inventors on patents issued and/or pending related to this work.