Anti-CD45 PBD-based antibody-drug conjugates are effective targeted conditioning agents for gene therapy and stem cell transplant

Abstract

Stem cell gene therapy and hematopoietic stem cell transplantation (SCT) require conditioning to ablate the recipient’s hematopoietic stem cells (HSCs) and create a niche for gene-corrected/donor HSCs. Conventional conditioning agents are non-specific, leading to off-target toxicities and resulting in significant morbidity and mortality. We developed tissue-specific anti-human CD45 antibody-drug conjugates (ADCs), using rat IgG2b anti-human CD45 antibody clones YTH24.5 and YTH54.12, conjugated to cytotoxic pyrrolobenzodiazepine (PBD) dimer payloads with cleavable (SG3249) or non-cleavable (SG3376) linkers. In vitro, these ADCs internalized to lysosomes for drug release, resulting in potent and specific killing of human CD45⁺ cells. In humanized NSG mice, the ADCs completely ablated human HSCs without toxicity to non-hematopoietic tissues, enabling successful engraftment of gene-modified autologous and allogeneic human HSCs. The ADCs also delayed leukemia onset and improved survival in CD45⁺ tumor models. These data provide proof of concept that conditioning with anti-human CD45-PBD ADCs allows engraftment of donor/gene-corrected HSCs with minimal toxicity to non-hematopoietic tissues. Our anti-CD45-PBDs or similar agents could potentially shift the paradigm in transplantation medicine that intensive chemo/radiotherapy is required for HSC engraftment after gene therapy and allogeneic SCT. Targeted conditioning both improve the safety and minimize late effects of these procedures, which would greatly increase their applicability.

Graphical abstract

Amrolia and colleagues performed a pre-clinical assessment of clinically relevant anti-CD45 pyrrolobenzodiazepine (PDB) antibody-drug conjugates (ADCs) and demonstrated their potential utility as conditioning agents. Anti-CD45 PBD ADCs enabled human hematopoietic stem cell engraftment in humanized mouse models of gene therapy and stem cell transplantation and were also potent anti-leukemic agents in murine leukemia models.

Introduction

Allogeneic hematopoietic stem cell transplantation (SCT) is curative for patients with a range of malignancies, including acute myeloid/lymphoblastic leukemia (AML/ALL), genetic diseases such as hemoglobinopathies, primary immunodeficiency (PID), and metabolic diseases. For patients with genetic disorders of the hematopoietic system where no HLA-matched donor is available, gene therapy with autologous hematopoietic stem cells (HSCs) transduced with a corrected transgene is increasingly used.^1,2,3,4,5,6 However, for both SCT and gene therapy, intensive conditioning with chemotherapy and/or radiotherapy is required to eradicate the patient’s own HSCs and create a niche for the incoming graft. These conditioning regimens are associated with severe organ toxicities, e.g., veno-occlusive disease of the liver, gut mucositis, alopecia, and pneumonitis, as well as long-term toxicities, e.g., pubertal failure, infertility, cardiotoxicity, growth retardation, neurocognitive delay, and secondary malignancy. These off-target toxicities⁷ occur even with non-myeloablative conditioning, limiting the broader application of SCT and gene therapy. Many older patients are ineligible for SCT because they cannot tolerate conditioning, and late effects, e.g., infertility, are a major barrier to the greater uptake of gene therapy. For example, while recent data on gene therapy with CRISPR-Cas9-edited HSCs for hemoglobinopathies are compelling,³ uptake is likely to be limited because bulsulfan-based conditioning results in pubertal failure and infertility.

Conditioning toxicities arise because conventional chemo-/radiotherapy can kill any dividing cell. Monoclonal antibodies (mAbs) targeting molecules expressed on the surface of HSCs have the potential to provide targeted conditioning without non-hematologic toxicity. This approach has been developed with c-Kit (CD117), utilizing both naked antibodies^8,9,10 and antibody-drug conjugates (ADCs).^11,12,13 However, c-Kit is expressed at low levels on HSCs, potentially limiting its utility.

CD45 is an attractive candidate target for mAb-targeted conditioning, as it is highly expressed on nucleated hematopoietic cells at high antigen density but absent on non-hematopoietic tissues. Since CD45 is expressed on virtually all hematologic malignancies, anti-CD45 ADCs could also be used to cytoreduce refractory leukemia/lymphoma/myeloma prior to autologous or allogeneic SCT. In rodent models, lytic anti-murine CD45 mAbs induce aplasia¹⁴ and facilitate engraftment of allogeneic stem cells after non-myeloablative conditioning.¹⁵ Moreover, an anti-murine CD45-saporin immunotoxin facilitated durable multilineage engraftment of congenic murine HSCs in immunocompetent mice and engraftment of gene-corrected murine HSCs in mice with sickle cell disease, with minimal non-hematopoietic tissue toxicity.¹⁶ We have previously used the anti-human CD45 mAbs YTH24.5 and YTH54.12 for myelosuppression as part of a minimal-intensity conditioning regimen for allogeneic SCT in children with PID, and showed curative engraftment in 13 out of 16 patients.¹⁷ By virtue of their rat immunoglobulin G2b (IgG2b) isotype, the antibodies were rapidly cleared, allowing timely transplant. Importantly, apart from manageable allergic reactions, no significant toxicity to non-hematopoietic tissues was observed, and these patients had no reported late effects related to conditioning. However, myeloid engraftment was suboptimal in some patients, and this may preclude extension to patients without underlying immunodeficiency.

We sought to improve the efficacy of YTH24.5 and YTH54.12 as conditioning agents and for leukemia treatment. One potentially effective means of increasing mAb potency is to attach radionuclides.^18,19,20 However, radioimmunotherapy requires specialist facilities and is less attractive for patients with non-malignant disorders. A more broad approach is to conjugate the antibodies with potent cytotoxic drugs. We selected a pyrrolobenzodiazepine (PBD) dimer, which covalently crosslinks DNA in the minor groove.²¹

We demonstrate that the anti-CD45 PBD-based ADCs selectively kill human CD45⁺ cells in vitro, deplete human HSCs in vivo in humanized mouse models, and enable transplantation of allogeneic and gene-modified autologous human HSCs. Further, the anti-CD45 ADCs are highly effective anti-leukemic agents in xenogeneic mouse models of acute leukemia.

Results

CD45 is an excellent target for ADCs

Human CD34⁺/CD38⁻/CD45RA⁻/CD90⁺ HSCs and other early progenitors expressed comparable CD45 antigen density (mean = 5.13 × 10⁴ molecules/cell) to bulk peripheral blood mononuclear cells (Figure S1A); c-Kit density on HSC/progenitor subpopulations was lower, at approximately 1.5 × 10³ molecules/cell (Figure S1B). The >30-fold higher cell-surface density of CD45 versus c-Kit indicated that CD45 would be a superior target for payload delivery.

Western blotting confirmed that CD45 expression was restricted to hematopoietic tissue (Figure S1C). Immunoassays with recombinant protein (data not shown) and species cross-reactivity studies confirmed that YTH24.5 and YTH54.12 bound to human CD45 but not Rhesus macaque or murine CD45⁺ cells (Figure S1D). Further, we demonstrated that both mAbs internalized upon antigen binding and were routed to the lysosomal pathway (Figure S1E), a prerequisite for effective ADC activity.

Anti-CD45 ADCs show potent activity against CD45⁺ cell lines and human CD34⁺ progenitors in vitro

ADCs were generated by conjugation of YTH24.5 and YTH54.12 to PBD payloads: (1) SG3249²¹ containing an enzyme cleavable valine-alanine dipeptide linker and (2) SG3376,²² the same PBD warhead with a non-cleavable linker (Figure S2A). Isotype ADCs were generated using a commercially available rat IgG2b isotype control mAb (with no known antigen) conjugated to the same two payloads. ADCs with enzymatically cleavable linkers release the cytotoxic warhead in lysosomes. The released drugs are typically membrane permeable and capable of bystander killing. ADCs with non-cleavable linkers require lysosomal proteolytic degradation to release an amino acid linker warhead which is not membrane permeable, resulting in less bystander killing. In cell viability assays, where ADCs were incubated with a 1:1 mixture of Nalm6 (human CD45⁻) and Jurkat (human CD45⁺) cells, Isotype- SG3249 resulted in a significant killing of both human CD45⁺ and CD45⁻ targets, most likely due to non-specific release of the payload due to the presence of the cleavable linker, whereas Isotype-SG3376 did not (Figure S2B). Furthermore, YTH24.5-SG3249 killed 80% of the CD45⁻ Nalm6 cells, but YTH24.5-SG3376 showed no significant bystander killing. These results demonstrate that whereas ADCs with a cleavable linker result in significant non-specific bystander killing in vitro, this is abrogated by the use of a non-cleavable linker.

We then assessed the cytotoxicity of the anti-CD45 ADCs against human CD45⁺ (hCD45⁺) or human CD45⁻ (hCD45⁻) cell lines. YTH24.5-SG3249 and YTH24.5-SG3376 killed hCD45⁺ leukemia cell lines (OCIM1 and Jurkat) with IC₅₀ values (determined as the concentration at which 50% cell viability was reached) of <0.1 pM for OCIM1 cells and <0.2 pM for Jurkat cells (Figure 1A and Table 1). Killing by Isotype-SG3249 and Isotype-SG3376 was only observed at levels 3–5 logs higher, and unconjugated mAbs did not kill. Isotype-SG3376 was less toxic than Isotype-SG3249, indicating that non-cleavable ADCs have a lesser bystander effect. Similar data were observed for YTH54.12-SG3249 and YTH54.12-SG3376 ADCs (Figure S3 and Table S1). These data indicated a wide (>4 log) therapeutic window between hCD45⁺ and hCD45⁻ (HEK293T) cells. Ten of 12 CD45⁺ cell lines tested were sensitive to killing by YTH24.5-SG3249 and 9 of 12 to YTH24.6-SG3376 (Table S2). TF1 cells may have been resistant to YTH24.5-SG3376 due to relatively low levels of SLC46A3, which is required for transport of non-cleavable payloads from the lysosome to the cytoplasm.²³

Figure 1.

Anti-CD45 PBD-based ADCs are potent and specifically kill human CD45⁺ cell lines and inhibit colony formation of human CD34⁺ cells

(A–C) Five-day cell viability assays were performed using serial dilutions of ADC or mAb or medium alone on (A) OCIM1 (CD45⁺ human AML cell line), (B) Jurkat CD45⁺ (human T-ALL cell line), and (C) HEK293T (CD45⁻ human embryonic kidney cell line). Data points are mean ± SD. Representative data shown. (D) Clonogenic assays were performed using purified healthy donor CD34⁺ cells. Pooled data (mean ± SD) from three donors are shown.

Table 1.

Summary of in vitro IC₅₀ values from cell-viability assays and clonogenic assays

ADC	Cell viability assays IC50 (pM)						CFU assays IC50 (pM)
	Jurkat (hCD45+)		OCIM1 (hCD45+)		293T (hCD45−)
	Average	n	Average	n	Average	n	Average	n
Isotype-SG3249	309.6	4	214.4	4	>6,000	3	17,131.92	6
Isotype-SG3376	3,399.8	4	2,288.3	4	>6,000	3	>100,000.00	3
YTH24.5-SG3249	0.213	4	0.030	4	>6,000	3	10.16	6
YTH24.5-SG3376	0.052	4	0.096	4	>6,000	3	4.92	3

Mean IC₅₀ values were determined from at least three independent experiments or from at least three normal CD34⁺ donors assayed in triplicate.

Both YTH24.5-SG3249 and YTH24.5-SG3376 potently inhibited colony formation in clonogenic assays using healthy donor CD34⁺ cells, with IC₅₀ values of <10 pM (Figure 1D and Table 1). Isotype-SG3249 inhibited colony formation at 3-log higher levels, and Isotype-SG3376 produced IC₅₀ values of >100,000 pM, confirming less non-specific toxicity with the non-cleavable payload (Figure 1D and Table 1).

YTH24.5-SG3249 and YTH24.5-SG3376 have short half-life, are stable, and do not cause toxicity in non-humanized NSG mice

Conditioning agents for SCT or gene therapy require a short half-life (t_½) to avoid depleting the incoming HSCs. We performed pharmacokinetic (PK) studies in non-humanized NSG mice to determine the t_½ of ADCs and their stability in vivo. Non-compartmental analysis on serial serum samples revealed that YTH24.5-SG3249 had a t_½ of ∼1.5 days in non-humanized NSG mice (Figure 2A) and that the total and PBD-conjugated antibody levels were comparable, indicating no significant ADC deconjugation in vivo. Similar data were obtained for YTH24.5-SG3376 (Figure 2B). The non-humanized NSG mice tolerated 1 mg/kg YTH24.5-SG3249 and YTH24.5-SG3376, with stable weight throughout the experiment. Histopathological examination at 1 month post treatment showed no changes in bone marrow (BM) cellularity or tissue damage to the liver and other tissues (Figure 2C).

Figure 2.

Anti-CD45 PBD-based ADCs are stable in the serum with t_½ of 1.5 days and do not cause non-hematopoietic toxicity in non-humanized NSG mice

Mice were injected with 1 mg/kg YTH24.5-SG3249 intravenously, and serial serum samples were collected over the next 7 days. (A) Mean total antibody levels ± SD (unconjugated + PBD-conjugated) and mean PBD-conjugated antibody levels ± SD in three study mice dosed with 1 mg/kg YTH24.5-SG3249. (B) Mean total antibody levels ± SD (unconjugated + PBD-conjugated) and PBD-conjugated antibody levels ± SD in three study mice dosed with 1 mg/kg YTH24.5-SG3376. (C) Hematoxylin and eosin staining of tissues taken from representative mice at 1 month post treatment with PBS or 1 mg/kg YTH24.5-SG3249 or YTH24.5-SG3376 from mice in (A) or (B). Scale bars, 50 μm.

Anti-human CD45 ADCs efficiently deplete human CD45⁺ cells and human HSCs in humanized NSG mice

We then investigated whether the anti-CD45 ADCs could deplete hCD45⁺ cells and HSCs in vivo in humanized NSG mice that had been engrafted with human CD34⁺ cells from healthy donors (Figure 3A). Engraftment was allowed to proceed for 8 weeks with ∼20% human CD45⁺ cells detected in the blood (data not shown). The prior PK studies had shown that a dose of 1 mg/kg anti-CD45 ADCs was well tolerated, and this dose was used for all subsequent in vivo experiments. Higher doses of up to 3 mg/kg anti-CD45 ADC were assessed in humanized mice but resulted in significant weight loss after treatment (data not shown).

Figure 3.

Anti-CD45 PBD-based ADCs deplete human CD45⁺ cells in humanized NSG mice and ablate human HSCs in the bone marrow

(A) Experimental scheme. Sublethally irradiated mice received a transplant of 0.5 × 10⁶ CD34⁺ cells from a healthy donor. Engraftment was allowed to proceed for 8 weeks before mice were conditioned with PBS or a single dose of 1 mg/kg ADC. Mice were culled and analyzed after a further 8 days. (B) Percentage of hCD45⁺ in the BM of mice after 8 days of ADC treatment. (C) Percentage of hCD45⁺ in the blood of mice after 8 days of ADC treatment. (D) Absolute numbers of hCD45⁺/hCD34⁺ cells in the BM per leg of mice after 8 days of ADC treatment. (E) Absolute numbers of immunophenotypically defined HSCs in the BM per leg of mice after 8 days of ADC treatment. (F) Total MNCs in the BM of humanized mice at 8 days post-treatment. Unpaired t test was performed. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001; ns, not significant.

Humanized mice were treated with PBS or 1 mg/kg ADC for 8 days. We determined the percentage of hCD45⁺ cells in the BM of these mice (Figure 3B). Mean hCD45⁺ engraftment levels in the PBS, Isotype-SG3249, and Isotype-SG3376 cohorts were similar at 69.1%, 66.3%, and 75.4%, respectively. In contrast, hCD45⁺ cells were significantly reduced in the YTH24.5-SG3249 (mean = 14.4%; p < 0.001) and YTH24.5-SG3376 (mean = 3.85%; p < 0.0001) cohorts. There was no significant difference in the percentage of hCD45⁺ cells between YTH24.5-SG3249 and YTH24.5-SG3376 (p < 0.05). A similar pattern was seen in the blood (Figure 3C). Human CD34⁺ cells were significantly reduced in YTH24.5-SG3249 and YTH24.5-SG3376 (Figure 3D) compared to controls. Most importantly, human HSCs (Lin⁻/CD34⁺/CD38⁻/CD45RA⁻/CD90⁺) were completely eradicated in the YTH24.5 ADC-treated mice (Figure 3E) with no difference between YTH24.5-SG3249 and YTH24.5-SG3376 (p < 0.05), demonstrating that our anti-CD45 ADCs ablate human HSCs in vivo.

Since YTH24.5 does not cross-react with mouse CD45, bystander killing of murine cells in the BM was assessed by comparing total cellularity in mice treated with Isotype ADCs with the cleavable and non-cleavable payloads. In PBS-treated mice, the total mean number of mononuclear cells (MNCs) per leg was 6.5 × 10⁶ (Figure 3F), MNCs in the Isotype-SG3249-treated mice (mean = 2.5 × 10⁶) was significantly lower (p = 0.0071), whereas Isotype-SG3376-treated animals (mean = 10.01 × 10⁶) had cellularity similar to that of PBS-treated mice, suggesting that SG3249 had a bystander effect on both murine and human cells in the BM whereas SG3376 did not. Regarding BM cellularity of the mice treated with CD45-specific ADCs, mice treated with YTH24.5-SG3249 had significantly lower numbers of cells compared to the mice treated with YTH24.5-SG3376 (mean = 3.5 × 10⁵ vs. 4.65 × 10⁶, p = 0.0028), further indicating bystander activity of ADCs with SG3249 payload with cleavable linker, as these cohorts had similar levels of engraftment in the blood before treatment (data not shown).

Conditioning with anti-human CD45 ADCs enables engraftment of autologous GFP-marked CD34⁺ cells in a gene therapy model

The ability of anti-CD45 ADCs to generate a niche for engraftment of gene-modified human CD34⁺ cells was evaluated in a humanized murine model of gene therapy. Humanized NSG mice were conditioned with PBS or 1 mg/kg ADC or mAb (Figure 4A). After 2 weeks, to allow clearance of mAbs/ADCs, mice received a second transplant of autologous GFP-transduced human CD34⁺ cells. This 2-week washout period was selected because by this time point ADCs were undetectable in the blood by flow cytometry and/or ELISA (whereas at 7–9 days small amounts of ADCs could still be detected in some mice). The transduction efficiency by flow cytometry of hCD34⁺ cells prior to secondary transplant was 77%.

Figure 4.

Anti-CD45 PBD-based ADCs enable the enhanced engraftment of GFP⁺ autologous CD34⁺ cells in humanized NSG mice

(A) Experimental scheme. NSG mice were sublethally irradiated with 2.5 Gy before transplant of 0.5 × 10⁶ human CD34⁺ cells on the following day, and engraftment was allowed to proceed for 8 weeks. The humanized NSG mice were treated with a single dose of 1 mg/kg isotype mAb, YTH24.5 mAb, YTH54.12 mAb, YTH24.5-SG3249, or YTH54.12-SG3249 at 2 weeks before transplant of GFP-transduced autologous CD34⁺ cells. A control group of NSG mice received sublethal irradiation followed by transplant of the GFP-transduced CD34⁺-only cells to determine the maximum possible GFP⁺ engraftment that could be achieved. (B) Percentage of total hCD45⁺ cells in the BM of mice culled at 9 weeks after the transplant of autologous GFP⁺ cells. (C) Percentage of GFP⁺/hCD45⁺ of all hCD45⁺ cells in the BM of mice culled at 9 weeks after the transplant of autologous GFP⁺ cells. (D) Percentage of GFP⁺/hCD34⁺ of all hCD34⁺ cells in the BM of mice culled at 9 weeks after the transplant of autologous GFP⁺ cells. (E) Percentage of GFP⁺/hCD45⁺ of all hCD45⁺ cells in the blood of mice culled at 9 weeks after the transplant of autologous GFP⁺ cells. (F) Percentage of GFP⁺/hCD45⁺ of all hCD45⁺ cells in the spleen of mice culled at 9 weeks after the transplant of autologous GFP⁺ cells. Unpaired t tests were performed. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001; ns, not significant.

At 9 weeks post-second transplant, hCD45⁺ engraftment in the BM of these mice was between 50% and 80% (Figure 4B). YTH24.5-SG3249 and YTH54.12-SG3249 resulted in significantly higher levels of GFP⁺/hCD45⁺ cells (i.e., cells derived from the second transplant), with mean hCD45⁺ levels of 73.75% (p < 0.05) to 74.95% (p < 0.01) compared to the Isotype-SG3249 cohort (mean = 41.50%) (Figure 4C). The levels of engraftment seen in YTH24.5-SG3249 and YTH54.12-SG3249 cohorts were similar to those in irradiated control mice receiving only the transduced GFP⁺/CD34⁺ cells. Since GFP⁺ transduction efficiency was <100% for the second transplant, it was not possible to determine whether GFP⁻/hCD45⁺ cells were derived from the first or second transplant. Enhanced engraftment of GFP⁺ cells in the anti-CD45 ADC cohorts compared to Isotype-SG3249 control was seen in both the B and myeloid lineages, but not in T cells (Figures S4A–S4I). Background engraftment of GFP⁺/hCD45⁺ cells was similar between PBS, YTH24.5, YTH54.12, and isotype mAbs and the Isotype-SG3249 controls (30%–40%) (Figure 4C). This was higher than expected, likely to due to NSG mice being more permissive to engraftment, particularly after prior irradiation for the primary transplant. Importantly, when the total human CD34⁺ cells were analyzed, the YTH24.5-SG3249 and YTH54.12-SG3249 treatment groups showed significantly higher levels of human GFP⁺/CD34⁺ cells (mean = 81.5% [p < 0.001] to 85.35% [p < 0.0001]) and similar to the irradiated control group when compared to Isotype-SG3249 (mean = 44.13%) (Figure 4D). Higher levels of engraftment of GFP⁺/hCD45⁺ cells, similar to the irradiation control cohort, were also observed in the blood (Figure 4E) and spleen (Figure 4F). These data show that anti-CD45 ADCs can create a sufficient niche in humanized mice to enhance the engraftment of autologous gene-modified human CD34⁺cells.

Anti-human CD45 ADCs in humanized NSG mice enables engraftment of HLA-mismatched allogeneic CD34⁺ cells in an allogeneic SCT model

We then tested whether conditioning with anti-CD45 ADCs could enable engraftment of allogeneic human cells. YTH24.5 was chosen for this and subsequent experiments, as it had superior manufacturability for future development. In an allogeneic SCT model (Figure 5A), mice were initially humanized with CD34⁺ cells from an HLA-A3⁺/HLA-B8⁻ healthy donor, and after conditioning received a second transplant of HLA-mismatched HLA-A3⁻/HLA-B8⁺ healthy donor CD34⁺ cells (Table S3). Human CD45⁺ engraftment in the blood before ADC conditioning was ∼4% (data not shown). Flow-cytometric analysis was performed to determine engraftment levels from the first and second transplant. The gating strategy and representative flow plots are shown in Figure S5.

Figure 5.

Anti-CD45 PBD-based ADCs can be used as conditioning agents to deplete human CD45⁺ cells in humanized NSG mice to enable engraftment of HLA-mismatched allogenic human CD34⁺ cells

(A) Experimental scheme. NSG mice were sublethally irradiated before transplant of 0.5 × 10⁶ CD34⁺ cells (HLA-A3⁺/HLA-B8⁻) and allowed to engraft for 4 weeks. Mice were then conditioned with PBS or 1 mg/kg ADC for 2 weeks before a second transplant of 1.0 × 10⁶ CD34⁺ cells from an HLA-mismatched allogeneic healthy donor (HLA-A3⁻/HLA-B8⁺). Mice were culled and analyzed after a further 7 weeks. Control NSG mice received sublethal irradiation before transplantation of CD34⁺ cells used for either the first transplant only (HLA-A3⁺/HLA-A8⁻) or the second transplant only (HLA-A3⁻/HLA-B8⁺). (B) Percentage of total human CD45⁺ cells in BM at 7 weeks after the second transplant. (C) Percentage of HLA-B8⁺ of all hCD45⁺ cells in the BM of mice at 7 weeks after the second transplant. (D) Percentage of HLA-B8⁺/hCD34⁺ of all human CD34⁺ cells in the BM. (E) Absolute number of human HLA-B8⁺ hCD34⁺ cells in the BM (per leg). (F) Absolute number of HLA-B8⁺ HSCs in the BM (per leg). Red data points are mice culled within 2 weeks prior to termination of experiment. Unpaired t tests were performed. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001; ns, not significant. The experiment was repeated on three donors, and representative data are shown.

The mean hCD45⁺ engraftment (reflecting both primary and secondary grafts) in the BM of ADC treatment groups was similar that of the PBS (mean = 59.44%) except for YTH24.5-SG3376 (mean = 24.33%) (Figure 5B). However, mice receiving YTH24.5-SG3249 (mean = 98.07%) and YTH24.5-SG3376 (mean = 99.0%) showed significantly higher levels of HLA-B8⁺ cells compared to isotype ADC controls (Isotype-SG3249: mean = 68.19%, p = 0.0334; Isotype-SG3376: mean = 36.77%, p < 0.0001), indicating that the anti-CD45 ADCs enabled engraftment of the HLA-mismatched allogeneic second transplant (Figure 5C). Similar results were observed in the blood and spleen (Figures S6A–S6D). Moreover, mice conditioned with YTH24.5-SG3249 and YTH24.5-SG3376 demonstrated multilineage engraftment of B cells, myeloid cells, and T cells from the second transplant (Figures S7A–S7C). Isotype-SG3249-treated mice showed higher levels of engraftment from the second transplant compared to PBS, suggesting non-specific killing that was not observed with Isotype-SG3376. Increased levels of engraftment of HLA-A3⁻/HLA-B8⁺ cells from the second graft following conditioning with YTH24.5-SG3249 and YTH24.5-SG3376 were also detected in the human CD34⁺ compartment (Figure 5D).

When absolute numbers of HLA-A3⁻/HLA-B8⁺ human CD34⁺ cells were determined (Figure 5E), there were no significant differences between mice treated with PBS (mean = 1.04 × 10⁵), Isotype-SG3249 (mean = 1.07 × 10⁵), and Isotype-SG3376 (mean = 2.05 × 10⁴). In contrast, significantly higher numbers of engrafted CD34⁺ cells from the secondary donor were seen in the corresponding YTH24.5-SG3249 and YTH24.5-SG3376 groups (means = 8.85 × 10⁵, p < 0.0001 and 1.81 × 10⁵, p < 0.05, respectively). Similar results were obtained in the HSC compartment (Figure 5F). The number of HLA-A3⁻/HLA-B8⁺ HSCs in the YTH24.5-SG3249 cohort (mean = 2,099) was significantly higher (p < 0.0001) than that of the Isotype-SG3249 group (mean = 92; p < 0.0001). This pattern was repeated with the cleavable ADCs, where the mean number of HLA-A3⁻/HLA-B8⁺ HSCs for YTH24.5-SG3376 was 440, vs. 48 for Isotype-SG3376 (p = 0.0391). The higher number of HLA-A3⁻/HLA-B8⁺ HSCs and CD34⁺ cells in the YTH24.5-SG3249 cohort compared to the YTH24.5-SG3376 cohort is likely due to the slightly higher numbers of MNCs in the YTH24.5-SG3249 group compared to the YTH24.5-SG3376 group (Table S5), though not statistically significant combined with a reduction in total hCD45⁺ cells in the YTH24.5-SG3376 group.

A few mice in the PBS, Isotype-SG3249, and Isotype-SG3376 groups required culling in the 2-week period prior to the planned end of the experiment, due to cytopenia in the BM. This was associated with very high levels (>80%) of HLA-A3⁻/HLA-B8⁺ CD3⁺ cells from the secondary transplant in the blood, BM, and spleen (data not shown). While these could potentially have been primed by either mouse antigen-presenting cells (APCs), i.e., xenogeneic graft-versus-host disease or primary human graft-derived APCs (secondary graft vs. primary graft), the fact that this was observed only in the PBS, Isotype-SG3249, and Isotype-SG3376 cohorts (where 60%–80% of the marrow was derived from the primary graft) suggests it is more likely that human APCs from the primary graft stimulated an alloreactive T cell response from the small numbers of HLA-A3⁻/HLA-B8⁺ T cells transferred with the CD34⁺ selected cells in the second graft. One mouse from the YTH24.5-SG3376 group developed cytopenia and was culled before the end of the experiment, but no expansion of human CD3⁺ cells was detected.

Overall, these data demonstrate that the anti-CD45 ADCs as single agents enable high-level multilineage and stem cell engraftment in the allogeneic transplant setting.

Anti-human CD45 ADCs show significant anti-leukemic activity in mouse tumor models

To determine whether anti-CD45 ADCs could be used as cytoreductive agents, NSG mice transplanted with firefly luciferase⁺ (fLuc⁺) OCIM1 cells were treated on the following day with PBS or 1 mg/kg mAb or ADC (Figure 6A). Isotype-mAb and isotype ADC-treated mice developed progressive leukemia (Figures 6B and 6C), and all mice were culled by day 32 (Figure 6D). Mice treated with YTH24.5 mAb showed slightly delayed tumor growth, but all mice were culled by day 32. In contrast, YTH24.5-SG3249 and YTH24.5-SG3376 effectively suppressed tumor growth so that all mice in these cohorts were tumor-free and alive at the end of the experiment (Figure 6D). Both YTH24.5 ADCs resulted in significantly increased survival compared to their corresponding isotype ADC-treated mice (p = 0.0021 and p = 0.0021, respectively).

Figure 6.

A single dose of anti-human CD45 ADC can prevent and treat established AML and prolong survival in a human AML model in NSG mice

(A) Experimental scheme. NSG mice were transplanted with fLuc⁺ OCIM1 AML cells before a single dose of 1 mg/kg ADC or mAb on the following day. Regular monitoring and IVIS imaging was carried out. (B) Bioluminescence images of mice dosed with PBS, or 1 mg/kg YTH24.5 mAb, Isotype-SG3249, Isotype-SG3376, YTH24.5-SG3249, or YTH24.5-SG3376 from day 1 to day 32. (C) Average total bioluminescence signal ± SD for each cohort. (D) Survival curves for experimental mice up to 60 days after transplant of fLuc⁺ OCIM1 cells. (E) Experimental scheme for established AML model. NSG mice were transplanted with fLuc⁺ OCIM1 AML cells and treated on day 10 with 1 mg/kg ADC or mAb or PBS. Regular monitoring and IVIS imaging were carried out. (F) Bioluminescence images of treated mice from day 10 to day 57. (G) Average total bioluminescence signal ± SD for each cohort. (H) Survival curves for experimental mice up to 60 days after transplant of fLuc⁺ OCIM1 cells. Log-rank tests were performed.

Similar data were seen with a T cell ALL (T-ALL) model, in which NSG mice were transplanted with fLuc⁺ Jurkat cells, followed by treatment the following day with 1 mg/kg of mAb, ADC, or PBS (Figure S8A). Mice treated with PBS, isotype mAb, and Isotype-SG3249 developed progressive leukemia by day 13 post transplant and were all culled by day 31 due to morbidity (Figures S8B and S8C). YTH24.5 mAb-treated mice had delayed tumor progression but progressed and were all culled by day 42. In contrast, mice receiving YTH24.5-SG3249 showed no significant tumor progression compared to day 0, with three of five mice surviving at day 70 (Figure S8D). YTH24.5-SG3249 significantly increased survival compared to the YTH24.5 mAb (p = 0.0027) and Isotype-SG3249 (p < 0.0001).

For more stringent assessment, anti-CD45 ADCs were tested in an established leukemia model (Figure 6E). NSG mice were engrafted with fLuc⁺ OCIM1 cells and allowed to establish for 10 days before treatment (Figure 6F), and a mean in vitro imaging system (IVIS) signal of 6.64 × 10⁷ p/s ± 2.45 × 10⁷ across the transplanted mice was determined (Figure 6G). From day 10, the signal for all the control groups (Isotype-SG3249, Isotype-SG3376, YTH24.5 mAb) continued to increase (Figure 6G) until mice were moribund and culled by day 40 (Figures 6G and 6H). However, in the YTH24.5-SG3249- and YTH24.5-SG3376-treated cohorts, a marked decrease in tumor signal was observed in the first 12 days after ADC injection (mean = 1.45 × 10⁶ p/s ± 0.58 × 10⁶ and 1.22 × 10⁶ p/s ± 0.23 × 10⁶, respectively). A subsequent slow re-emergence of the tumor occurred such that it returned to pre-ADC levels after ∼3.5 weeks. By day 58 after tumor injection, four of five mice from each of the YTH24.5 ADC cohorts were alive with significantly increased survival compared to the corresponding isotype ADC-treated mice (p = 0.0016 and p = 0.0023, respectively). These data demonstrate that the anti-CD45 ADCs mediated potent anti-leukemic effects in vivo, even in an established tumor model.

Discussion

To overcome the limitations associated with the toxicities of conventional conditioning agents for SCT and gene therapy, we developed potent ADCs selectively targeting the hematopoietic system through recognition of human CD45. We demonstrate that anti-CD45 PBD-based ADCs potently and specifically target human CD45⁺ cell lines and primary human progenitors. These ADCs were stable in the circulation, rapidly cleared, and did not result in non-hematopoietic toxicity in mouse models. In vivo treatment of humanized mice with the ADCs resulted in complete eradication of human HSCs and, as a single conditioning agent, resulted in enhanced engraftment of gene-modified autologous human HSCs and enabled multilineage engraftment from an allogeneic human donor. Additionally, these agents showed potent anti-leukemic effects and prolonged survival in xenogeneic models of AML and T-ALL.

We focused on CD45 as a target because of its specificity to hematopoietic cells and its higher expression on HSCs than other HSC-specific targets, such as c-Kit. An anti-human c-Kit monoclonal antibody, JSP191, has entered clinical studies as a conditioning agent in patients with SCID (NCT02963064), but to date when used alone the levels of myeloid engraftment achieved have been suboptimal.¹⁰ Additionally, based on pre-clinical work with an anti-murine saporin immunotoxin,¹⁵ Magenta Therapeutics developed MGTA-117, a rapidly cleared anti-human c-Kit-amanitin drug conjugate that enabled engraftment of gene-corrected HSCs in a non-human primate model,¹⁶ and were testing this in a phase 1/2 dose-escalation study in patients with relapsed/refractory AML (NCT05223699).¹³ However, this trial was paused due to two severe adverse pulmonary events and a patient’s death. We found that c-Kit was expressed at very low levels in human HSCs, and pre-treatment of human CD34⁺ progenitors with a c-Kit-specific ADC with a monomethyl auristatin E payload did not prevent engraftment in an NSG mouse model (data not shown), suggesting that c-Kit may not be an optimal target. ADCs targeting CD45 would have additional benefits of deleting T cells, B cells, and natural killer cells, resulting in immunosuppression to prevent graft rejection as well as myelosuppression pre-transplant and showing potent anti-leukemic effects, potentially reducing relapse risk and enabling refractory patients previously ineligible for transplant to become candidates for SCT. Wellhausen et al.²⁴ recently used a novel alternative strategy to target CD45. They mapped the epitope on CD45, which was recognized by their chimeric antigen receptor (CAR), and used CRISPR adenine base editing to introduce a mutation to prevent CAR-T cell recognition and confer resistance to fratricide. The CD45 on normal HSCs were also base edited to be resistant to CAR-T cell recognition to allow potentially safer and effective use of CD45-directed CAR-T cells, which target the non-edited CD45⁺ cells for the universal treatment of hematologic malignancies. However, because of the cost and complexity of generating patient-specific base-edited CD45 CAR-T cells as well as the need to base edit transplanted HSCs if they persist, this approach is likely to be less feasible than anti-CD45 ADCs.

We have utilized PBD payloads that induce highly toxic DNA interstrand crosslinks,²⁵ resulting in apoptosis of both dividing and non-dividing cells.²¹ While the PBD dimer warhead is genotoxic, we believe the efficient and specific targeting by the antibody moiety will restrict this genotoxicity to the hematopoietic lineage, particularly with the non-cleavable linker. ADCs with other payloads, e.g., amanitin (an RNA polymerase inhibitor) in MGTA-117, are also being investigated. While we did not directly compare our ADCs with MGTA-117, the anti-CD45 ADCs have IC₅₀ values on cell lines, and CD34⁺ progenitors are equivalent to those reported for MGTA-117. Moreover, PBD payloads have the advantage of being clinically validated.^26,27

Comparison of anti-CD45 ADCs bearing cleavable (SG3249) and non-cleavable (SG3376) payloads showed selective CD45⁺ cell killing with similar potency in vitro, and both were able to completely eradicate human HSCs in vivo. However, Isotype-SG3249 ADCs elicited non-specific killing of CD45⁺ and CD45⁻ targets in vitro and in vivo, which was sufficient to induce higher levels of secondary donor engraftment than PBS in our allogeneic SCT model. These effects were not seen with Isotype-SG3376, which has a non-cleavable linker. These data indicate that YTH24.5-SG3376 would be a better candidate to take forward clinically, as it is likely to result in less non-hematopoietic toxicity than the YTH24.5-SG3249 payload, enabling the use of a higher dose without dose-limiting toxicity.

ADC conditioning agents must have a short t_½ to avoid toxicity to the incoming gene-corrected or allogeneic cells. The t_½ of the anti-CD45 ADCs in non-humanized NSG mice was determined to be 1.5 days, and they were undetectable in the serum at 2 weeks. While slightly longer than the t_½ for the naked YTH antibodies in humans,^15,17,28 perhaps reflecting differences in both affinity of rat IgG2bs to mouse neonatal Fc receptor (FcRn) compared to human FcR and the absence of a human CD45⁺ sink in NSG mice, this would certainly be compatible with use as a conditioning agent. Rat IgGs lack binding/low-affinity binding to human FcRn and therefore have reduced t_½ in human serum.^29,30 Moreover, the t_½ of the YTH24.5 ADCs would be expected to be short in human patients, as CD45 is very highly expressed on human hematopoietic cells, which will lead to clearance of antibody-bound cells. Immunogenicity caused by the administration of rat IgG2b ADC into patients is unlikely, as the ability of the patient to mount an immune response should be compromised by depletion of recipient immune cells. Furthermore, our data suggest that anti-CD45 ADCs could be administered as a single cycle of therapy prior to SCT rather than repeat dosing.

We chose to develop the YTH24.5 and YTH54.12 anti-CD45 clones as ADCs because they were well tolerated, rapidly cleared, and together with low-dose chemotherapy were sufficient to enable engraftment of allogeneic HSCs in our previous clinical study of patients with primary immunodeficiency.¹⁷ However, the lack of cross-reactivity of these clones against murine CD45 necessitated humanization of immunodeficient NSG mice in SCT and gene therapy models in order to assess their efficacy. We acknowledge that this does not fully recapitulate the transplant setting. Moreover, as the YTH clones are selective for human CD45, a pharmacologically relevant non-clinical model to evaluate the efficacy and safety of ADCs containing these antibodies does not exist. Given that the CD45 mAbs used for these ADCs have already been shown to be safe in our previous clinical study, we would argue that meaningful further data on the toxicity, PK, and utility of the anti-CD45 ADCs as conditioning agents can only be obtained in clinical studies, for example as a cytoreductive bridge to allogeneic SCT in patients with refractory AML.

In summary, we have, for the first time, developed potent, highly specific, rapidly cleared anti-human CD45 ADCs and have demonstrated that these enable engraftment of both gene-modified autologous and allogeneic human HSCs as well as mediating potent anti-leukemic effects in murine models without significant toxicity to the non-hematologic tissues. Taken together, our data indicate that anti-CD45 ADCs show great promise as conditioning agents for gene therapy, while SCT and may also have utility as a bridge to allogeneic SCT in patients with refractory hematologic malignancies. If successful, targeted conditioning with these or similar anti-CD45 ADCs would be transformational for SCT and gene therapy by eliminating the non-hematologic toxicity and late effects of current conditioning regimens. This in turn could greatly increase the applicability and uptake of these approaches, particularly for hemoglobinopathies.

Materials and methods

ADC generation

YTH24.5 and YTH54.12 VH and VL sequences were provided by Cambridge Enterprises and used to generate recombinant full-length rat IgG2b and kappa antibodies by transient transfection in CHO cells, followed by purification with Protein A (Evitra, Basel, Switzerland). Conjugation of the anti-CD45 antibodies or a rat IgG2b, kappa isotype control (clone RTK4530) (BioLegend, San Diego, CA, USA) was carried out by Sterling Pharma Solutions (Dundee, UK). Initial pH adjustment of the formulated antibody with a Tris-EDTA buffer to a pH of ∼7.4 was performed, followed by partial reduction with Tris(2-carboxyethyl)phosphine hydrochloride. The reduced antibody was then conjugated with an excess of the maleimide-activated PBD payload in dimethyl acetamide followed by quenching of unreacted payload with N-acetylcysteine. After conjugation, the material was buffer swapped by tangential flow filtration into the final formulation buffer, 3% cyclodextrin in PBS, and filter sterilized using a 0.22-μm filter.

Cell line culture

Jurkat (American Type Culture Collection [ATCC], Manassas, VA, USA) and Nalm6 (ATCC) cells were cultured in RPMI 1640 (Thermo Fisher Scientific, Paisley, UK) containing 10% fetal bovine serum (FBS) (Thermo Fisher Scientific) (R10). OCIM1 (Leibniz Institute DSMZ, Braunschweig, Germany) cells were cultured in Iscove’s modified Dulbecco’s medium (Thermo Fisher Scientific) containing 10% FBS (I10). 293T (ATCC) cells were cultured in DMEM containing 10% FBS (D10). Cells were incubated in a humidified incubator at 37°C with 5% CO₂.

Cellviability assays

Cells at 0.5 × 10⁵/mL were cultured for 5 days in mAb or ADC or medium (in triplicate wells) in wells of flat-bottom 96-well plates. PrestoBlue Cell Viability reagent (Thermo Fisher Scientific) was added and incubated for 1.5 h. Fluorescence intensity from each well was detected using a FLUOstar OPTIMA microplate reader (BMG Labtech, Aylesbury, UK) using excitation filter of 560-10 and emission filter of 590-10, and gain at 1,500. Data were plotted using GraphPad Prism software, and sigmoid dose-response non-linear regression was carried out to determine IC₅₀ values.

Clonogenic assays

Cryopreserved CD34⁺ cells purified from the peripheral blood of granulocyte colony-stimulating factor mobilized healthy donors were thawed, washed, and cultured at a final cell density of 0.5 × 10⁵/mL in R10 medium in the presence of mAb, ADC, or medium alone for 2 h in 96-well U-bottom plates. Cells were gently mixed, and 1,000 cells were transferred to 2 mL of StemMACS HSC-CFU complete with Epo (Miltenyi Biotech, Bergisch Gladbach, Germany) and mixed well. 0.5 mL of this mixture was plated in triplicate into wells of a 24-well plate (Corning) using an 18-gauge blunt-ended needle and 1-mL syringe. Colonies were enumerated between 10 and 14 days. Use of human cells used in this study was approved by NHS Health Research Authority Research Ethics Committee, reference 16/LO/1290.

Transplantation of NSG mice

Animal research was carried out under Home Office license PP2527748 and internally reviewed by the biological services unit manager, appointed Home Office veternarian and project license holder. Adult NSG mice (Charles River Laboratories, UK) were irradiated with 2.5 Gy, and the next day 0.5 × 10⁶ thawed healthy donor human CD34⁺ cells were injected intravenously into the tail vein. For conditioning experiments, humanized NSG mice were treated with ADC or controls and culled after 8 days for analysis of BM, spleen, and blood by flow cytometry. For autologous/gene therapy and allogeneic SCT experiments, ADC- or control-treated humanized mice were transplanted with 0.5 × 10⁶ GFP-transduced autologous human CD34⁺ cells or 1 × 10⁶ human CD34⁺ cells from an HLA-mismatched allogeneic donor. Blood, spleen, and BM were taken for analysis by flow cytometry at the end of the experiment.

Transduction of human CD34⁺ cells

Transduction of human CD34⁺ cells was carried out as previously described³¹ using a multiplicity of infection of 12.5 for 18–24 h using concentrated viral supernatant from 293T cells transfected with pRRLSIN cPPT.PGK.GFP.WPRE (a gift from Didier Trono [Addgene plasmid #12252; http://n2t.net/addgene:12252; RRID:Addgene_12252]), pMD.G2 (a gift from Didier Trono [Addgene plasmid #12259; http://n2t.net/addgene:12259; RRID:Addgene_12259]), and pCMVR8.74 (a gift from Didier Trono [Addgene plasmid #22036; http://n2t.net/addgene:22036; RRID:Addgene_22036]). Between 18 and 24 h after transduction, cells were harvested, washed twice in PBS, and resuspended in PBS prior to injection into mice.

Leukemia models

NSG mice were transplanted with fLuc⁺ Jurkat (T-ALL) or OCIM1 (AML) cells with or without 2.5 Gy irradiation, respectively, on the preceding day. Mice were treated with ADC or controls on the following day or at 10 days. For imaging, mice were injected intraperitoneally with 3 mg of D-luciferin (Regis Technologies, Morton Grove, IL, USA), and imaged in an IVIS Lumina III (PerkinElmer, Beaconsfield, UK) under anesthesia. Quantification was performed using Living Image software (v4.5).

Flow cytometry

Single-cell suspensions from BM and spleen were prepared and blood samples taken into heparin solution. Red cell lysis was performed, and cells were blocked with mouse and human Fc blocking reagents before staining with antibodies detailed in Table S4. Samples were analyzed on a BD LSRII or Beckman Coulter CytoFLEX flow cytometer. Post-acquisition analysis was carried out using FlowJo software (BD Biosciences).

PK studies

Three mice were dosed with 1 mg/kg for each ADC, and serial serum samples were taken to determine total antibody levels and PBD-conjugated antibody levels. For detection of total antibody levels, a standard Meso Scale Discovery (MSD) plate (Meso Scale Diagnostics, Rockville, MD, USA) was coated with recombinant human CD45 (R&D Systems, Abingdon, UK) at room temperature for 1 h in PBS pH 7.4 (Life Technologies, Paisley, UK) and blocked with 3% bovine serum albumin (BSA) (Merck Life Sciences UK, Gillingham, UK) in PBS for 1 h at room temperature. Standard curve and quality control (QC) samples were prepared in mouse serum (Sera Laboratories International, Burgess Hill, UK) and diluted in assay buffer (PBS/1% BSA/0.05% Tween 20; Merck Chemicals, Nottingham, UK). Samples were incubated on the blocked plate for 1 h at room temperature. A biotinylated mouse-anti-rat antibody (BioLegend) for YTH24.5-SG3249 or biotin goat anti-rat IgG (H+L) (cross-adsorbed) (Thermo Fisher Scientific) for YTH54.12-SG3249 was added for detection and incubated for 1 h before a 30-min incubation with sulfoTAG streptavidin (Meso Scale Diagnostics). The plate was washed before adding 2× Read Buffer with Surfactant (Meso Scale Diagnostics). Electrochemiluminescence (ECL) signal data were acquired on the Quickplex MSD Plate Reader.

For detection of the PBD-conjugated antibody levels, a standard MSD plate was coated with anti-PBD 14B3-B7 (ADC Therapeutics, London, UK) at room temperature for 1 h in PBS (pH 7.4) and blocked with 3% BSA in PBS for 1 h at room temperature. Standard curve and QC samples were prepared in mouse serum and diluted in assay buffer (1% BSA/0.05% Tween 20). Samples were incubated on the blocked plate for 1 h at room temperature. A biotinylated mouse-anti-rat antibody was added for detection and was incubated for 1 h before a 30-min incubation with sulfoTAG streptavidin. The plate was washed before adding 2× Read Buffer with Surfactant. ECL signal data were acquired on the Quickplex MSD Plate Reader.

The half-life of the PBD-conjugated or total antibody was determined by non-compartmental analysis (NCA) using Phoenix 64 WinNonLin version 8.3.3.33 (Pharsight, Sunnyville, CA, USA). NCA allows determination of PK parameters of a drug from the time course of measured drug concentrations and does not require the assumption of a specific compartmental model.

At 1 month, organs were fixed in 10% neutral buffered formalin and prepared for histological analysis at Great Ormond Street Hospital (GOSH) Pathology Department.

Statistical analysis

Statistical tests were performed using GraphPad Prism as indicated in the text.

Data and code availability

All data are available in the main text or the supplemental information. For original data, please contact persis.amrolia@gosh.nhs.uk.