ADCT-301, a Pyrrolobenzodiazepine (PBD) Dimer–Containing Antibody–Drug Conjugate (ADC) Targeting CD25-Expressing Hematological Malignancies

In 2015 in the United States, there were an estimated 80,900 and 54,270 new diagnoses of all-type lymphoma and leukemia, and 20,490 and 24,450 deaths, respectively (1). Both Hodgkin lymphoma and various T- and B-cell non-Hodgkin lymphoma tumor cells express IL2R-α (CD25) (2, 3). At least two leukemic diseases, adult T-cell leukemia lymphoma (ATLL; ref. 4) and hairy cell leukemia (5), show nearly 100% expression of CD25 on their circulating tumor cells. In the refractory setting, there is evidence of maintenance of CD25 expression, and in acute myeloid leukemia (AML), its upregulation (6). Although not unique to hematologic malignancy, increased CD25 expression can be detected via shedding of this receptor (7, 8). Indeed, high soluble IL2R-α serum levels have helped identify patients with poorer prognoses in diffuse large B-cell lymphoma, cutaneous lymphomas, follicular lymphoma, AML, acute lymphoblastic leukemia, Hodgkin lymphoma (9–14), and are used as a part of a prognostic index in ATLL (15).

A number of clinical trials have demonstrated the utility of targeting this receptor in oncology. Phase I clinical trials of three radioimmunoconjugates, ⁹⁰Yttrium-labeled anti-Tac (16), ¹³¹Iodine-labeled basiliximab (17), and ⁹⁰Yttrium-labeled daclizumab (18) have demonstrated complete responses in patients with CD25-expressing lymphomas. The delayed myelosuppression seen with each of these anti-CD25 radionuclides was thought to be due to each β-emitting radionuclide having an effect on red marrow rather than an anti-CD25 antibody–specific effect (16–18). Two anti-CD25 immunotoxin conjugates have been tested in a number of human trials. LMB-2, a pseudomonal exotoxin conjugated to an anti-CD25 single-chain variable fragment (19) has shown objective responses in CD25-expressing lymphomas including ATLL (20) and denileukin–diftitox (Ontak), a diphtheria exotoxin conjugated to an IL2 fragment (21), is licensed in CD25-positive cutaneous T-cell lymphoma. Despite immunogenicity and manufacturing obstacles (22), immunotoxins remain a promising field of targeted therapies (23). Anti-CD25 radioimmunoconjugates and immunotoxins have both provided proof of concept for the feasibility of targeting the CD25 receptor with few toxicity concerns and have provided a platform for the preclinical development of new anti-CD25 targeting drugs.

The development of antibody–drug conjugates (ADC), however, has arguably been the immunoconjugate class most propelled by the commercial successes of therapeutic mAb technologies (24). This is at least partly because the synthesis of purified ADCs can now be carried out in a controlled way using optimized linker technologies (24). The use of ADCs in oncology indications has risen to prominence in the past decade, highlighted by FDA licensing of Adcetris in 2011 and trastuzumab emtansine (Kadcyla) in 2013 (24). More recently, novel ADCs delivering pyrrolobenzodiazepine (PBD) dimers as the cytotoxic warhead have been developed. PBD dimers are a class of exquisitely potent DNA minor groove interstrand cross-linking agents, one of which, SG2000 (SJG-136), has shown activity against both solid tumors and hematologic malignancies (25, 26). Advantages over ADCs targeting other warheads including tubulin inhibitors (e.g., auristatins and maytansinoids), DNA damaging drugs (e.g., calicheamicin), and classical chemotherapeutics include the ability to target low copy number antigens, and to exploit low drug–antibody ratios (DAR). Because of their novel mechanism of action, PBD-containing ADCs are active in tumors inherently resistant to other warhead types and against multidrug-resistant tumors (27).

Here we report the characterization, mechanism of action, and preclinical evaluation of a novel anti-CD25 ADC, ADCT-301.

HuMax-TAC antibody was first buffer exchanged into a proprietary histidine buffer at pH 6 containing various carbohydrate-based stabilizing additives using tangential flow filtration (TFF) on Pellicon 30-kDa cassettes using 10 diavolumes at constant TMP of 1.0 bar. Antibody was concentrated to approximately 10 mg/mL and the pH was adjusted to 7.5 using a TRIS/EDTA pH 8.5 buffer. TCEP reductant (10 mmol/L) was added to the batch as a 1.25 excess with respect to HuMax-TAC mAb and the reduction reaction was allowed to proceed for 60 minutes at 20°C. Dimethylacetamide (DMA) and 10 mmol/L SG3249 (28) were added to the solution to a final 5% DMA v/v and 3-fold excess relative to antibody. The conjugation reaction was incubated at 20°C for 60 minutes, then quenched with 3-fold molar excess of N-acetyl cysteine and incubated at 20°C for 20 minutes. The pH was then decreased to 6.0 using histidine hydrochloride solution and the ADC purified by TFF using the same conditions as the first buffer exchange step, this time with 12 diavolumes of the histidine/carbohydrate additive buffer. The solution was filtered (0.22 μm) and stored at −70°C. Final yield was estimated from starting antibody by UV/Vis.

The anaplastic large cell lymphoma (ALCL) cell lines Karpas 299 (2012) and Su-DHL-1 (2012) and the Hodgkin lymphoma cell lines L-540 (2014) and HDLM-2 (2014) were obtained from the Deutsche Sammlung von Mikroorganismen und Zellkuturen (DSMZ). The Burkitt lymphoma cell lines Daudi (2012) and Ramos (2012) and the cutaneous T-cell lymphoma line HuT-78 (2013) were obtained from the ATCC supplier LGC Standards. The myeloid leukemia cell lines, EoL-1 and KG-1, were obtained from DSMZ and LGC Standards, respectively (2015). Characterization of these mycoplasma-negative cell lines was performed by the cell bank supplier and included the determination and the verification of tissue-specific antibodies, karyotyping, and expression of aberrantly expressed genes. As cell lines were passaged for fewer than 6 months after receipt or resuscitation, they were not authenticated or tested in our laboratory. Cells were cultured in RPMI1640 media (Sigma-Aldrich) or Iscove's modified Dulbecco's medium (LGC Standards) with added l-glutamine (Sigma-Aldrich) or in l-glutamine–containing media (Life Technologies) each of which was supplemented with heat-inactivated FBS as per the manufacturer's instructions.

Binding of HuMax-TAC or ADCT-301 to cell lines was detected with an Alexa Fluor 488 (AF488)-labeled anti-human IgG (Life Technologies). For internalization studies, Karpas 299 cells were exposed (1 hour at 4°C) to ADCT-301 in blocking buffer (PBS, 10% (v/v) normal goat serum, 0.1% sodium azide (w/v)). Cells were washed and incubated for various periods in medium over a time course at 37°C after which time they were centrifuged, washed, and fixed (neutral buffered formalin). Once all time points were collected, plates were centrifuged (400 × g), washed, and incubated with either a saturating antibody concentration of mouse IgG1 CD25 antibody (BD Biosciences) or a mouse nontargeted (isotype control) mAb MOPC (BD Biosciences), in blocking buffer for 1 hour at 4°C. The potential for competition of binding to the same IL2R-α epitope by ADCT-301 and mouse anti-human CD25 was measured in a competition assay by coincubation of cells with increasing concentrations of ADCT-301 and a saturating concentration of mouse anti-human CD25. After washing and resuspension, the Qifikit protocol (ref. 29; DAKO) was followed, labeling cells with an anti-mouse FITC-labeled antibody (DAKO). For immunofluorescent imaging of ADCT-301–treated CD25-positive cell lines, cells were fixed (4% paraformaldehyde) at each time point, resuspended in permeabilization/wash buffer (BD Biosciences), and then incubated with APC-labeled mouse anti-human LAMP-1 (BD Biosciences) and AF488-labeled anti-human antibody (Life Technologies), while nuclear counterstaining was performed with Hoechst 33342 (Life Technologies; 1 hour at 4°C). After cell cytospins were generated, slides were mounted with ProLong gold antifade (Life Technologies) and slide edges sealed with clear nail polish. Images from dried slides kept at 4°C were captured on a TCS SPE confocal system utilizing a DM2500 microscope (Leica; 40 or 63× oil immersion objective) using three solid-state excitation lasers (405, 488, and 635 nm). Alternatively, Karpas 299 cells were incubated with ADCT-301, washed, and then either kept at 4°C in PBS, 0.1% azide or resuspended in medium and incubated at 37°C over a time course to 4 hours. After each time point, cells were washed and incubated with an AF488-labeled anti-human IgG (1 hour at 4°C). After washing and resuspension of all labeled cells in PBS, 0.1% sodium azide, cells were analyzed on an Accuri C6 flow cytometer (BD Biosciences). Median fluorescent intensity (MFI) of the live cell population was determined using FlowJo 7.0 (Tree Star Inc.). GraphPad was used to calculate an EC₅₀ or a linear regression model for calculating CD25 molecule numbers per cell. The Qifikit protocol (29) was also used to determine CD30 copy number with a mouse IgG1 CD30 antibody (GeneTex).

Human soluble CD25-streptavidin fusion protein transiently produced in CHO cells (Evitria) was immobilized by amine coupling to a Biacore CM3 sensor chip (GE Healthcare). A concentration range (0.05–5 nmol/L) of HuMax-TAC and ADCT-301 was injected over the chip for 120 seconds as per the Biacore multi-cycle kinetics method and allowed to dissociate by running HBS-EP buffer (Hepes-buffered saline containing EDTA and surfactant P20) over the chip for 60 minutes. The equilibrium dissociation constant (K_D) was calculated with BIAevaluation software (GE Healthcare).

Cell viability was measured following either 96-hour incubation at 37°C to ADC, nonbinding ADC, or warhead dilution series, or a 2-hour exposure and wash, followed by further 94-hour incubation. For the measurement of bystander killing, Karpas 299 cells were incubated with serial dilutions of ADCT-301 or nonbinding ADC in 96-well round-bottom plates for 48 hours. Centrifuged Karpas 299 conditioned medium (50 μL) was added to 50 μL of Ramos cell culture in 96-well flat-bottom plates and incubated for 96 hours. For all experiments, cell survival was determined by absorbance measurements on a Varioskan Flash Multimode Reader (Thermo Fisher Scientific) according to a standard MTS protocol (30). Percentage growth inhibition was calculated from the mean absorbance in the triplicate ADC-treated wells, compared with the mean absorbance in the control wells (100%). Dose–response curves were generated from the mean data of each independent experiment. The GI₅₀ was determined by fitting a sigmoidal dose–response curve to the data using GraphPad.

Unlabeled Karpas 299 cells and PKH-26–labeled (Sigma-Aldrich) Ramos cells (20 μmol/L) were diluted (1.5 × 10⁵ cells/mL) in medium and dispensed into round-bottom, ultralow-attachment 96-well microplates (Corning) as follows: monoculture: 50 μL cells plus 50 μL growth medium; cocultures: 50 μL Karpas 299 plus 50 μL labeled Ramos. Serial dilutions of ADCT-301 and nonbinding ADC, containing 20 μmol/L TGF-β inhibitor, SB431542 (Sigma-Aldrich) were added (100 μL/well) to single and cocultures. Cells were mixed on an IKA Vibrax shaker (10 seconds) and incubated at 37°C for 96 hours. After washing and resuspension in PBS, Fixable Dead Cell Stain (Life Technologies) was added. Cells were resuspended and incubated in the dark for 30 minutes. Alternatively, in apoptosis experiments, Karpas 299 cells (1.5 × 10⁵/mL) were dispensed with ADCT-301 or nonbinding ADC as described above. Etoposide was used as a positive apoptosis control. Cells were incubated at 37°C over a time course. At each time point, cells were centrifuged, washed, resuspended in Fixable Dead Cell Stain, and FITC-labeled Annexin-V antibody (BD Biosciences) diluted in Annexin-V Calcium-free buffer and incubated for 30 minutes at room temperature. Annexin-V buffer (250 μL) was added to each well before analysis on an Accuri C6 flow cytometer. Data were analyzed in FlowJo. Total, live, and dead cell counts from three replicate experiments were obtained from quadrant plots after compensation was carried out as necessary. Counts were normalized to the total cell count per well for each condition, to reflect the percentage of cells that remained viable or were apoptosed.

Cell lines were incubated with ADCT-301 or nonbinding ADC (20 ng/mL) for 2 hours at 37°C, washed, resuspended in medium and incubated at 37°C over a further time course (2–36 hours). After each time point, cells were centrifuged, resuspended in freezing medium (FBS with 10% DMSO), and frozen at −80°C. An equimolar concentration of SG3199 was dispensed into cell suspension, incubated for 2 hours, and the same procedure for harvesting ADC-exposed cells followed. To determine percentage cross-linking of increasing doses of ADCT-301 or nonbinding ADC on expressing cell lines, each ADC dilution was added to cells for 2 hours at 37°C. Cells were washed, resuspended in culture medium, dispensed into 24-well plates, incubated for 24 hours at 37°C and harvested, as above. For analysis of cross-linking on Karpas 299 subcutaneously implanted into SCID mice, cell suspensions were prepared from the excised tumors removed 24 hours after dosing with 0.2 or 0.6 mg/kg ADCT-301 or left untreated and stored in freezing medium at −80°C. Peripheral blood mononuclear cells (PBMC) from each mouse were isolated from whole blood using a Ficoll gradient and stored in freezing medium at −80°C. Frozen cells were thawed on ice, resuspended in cold RPMI (Sigma), and counted. Cell suspensions (2–3 × 10⁴ cells/mL), except unirradiated controls, were irradiated (15 Gy) and kept on ice. The Comet assay was performed as described previously (31). Slides were reviewed under a 20× objective on an epifluorescence microscope equipped with Hg arc lamp, 580 nm dichroic mirror, and 535 nm excitation and 645 nm emission filters suitable for visualizing propidium iodide staining with a minimum of 50 Comet images acquired per treatment condition. The olive tail moment (OTM) was determined as the product of the tail length and the fraction of total DNA in the tail as recorded by Komet 6 software (Andor Technology) and the percentage reduction in OTM calculated according to the formula:

where TMdi = tail moment drug irradiated; TMcu = TM control unirradiated; TMci = TM control irradiated.

Analysis was performed on Microsoft Excel and GraphPad.

Karpas 299 cells treated with serial dilutions of ADCT-301 concentrations and frozen at −80°C were thawed, washed in PBS, fixed with 4% paraformaldehyde for 5 minutes at room temperature, and washed in PBS. Cell cytospins were generated for each treatment condition and cells were permeabilized with PBS/0.2% Triton-X100, washed in PBS, and blocked in 10% FBS, 5% BSA in PBS. After incubation overnight (4°C) with goat anti-mouse γ-H2AX (JBW301; Millipore) in 1% FBS in PBS, slides were washed and incubated (1 hour at room temperature) with AF488-labeled goat anti-mouse IgG antibody (Life Technologies) and then counterstained with Hoecsht 33342 (Life Technologies). Slides were washed in PBS and distilled water, mounted with ProLong gold antifade (Life Technologies), and the coverslip edges sealed with clear nail polish. Images captured on a confocal system as described above were compressed to a maximum intensity projection image of z-stack acquired images to allow maximal detection of foci and analyzed in ImageJ (NIH, Bethesda, MD). Cell Profiler was used to quantify γ-H2AX foci per cell in at least 20 cells per treatment.

Karpas 299 cells were treated with ADCT-301, SG3199, nonbinding ADC, or left untreated. Ethanol-fixed nuclei were stained with 0.5 mg/mL propidium iodide (PI; Sigma) in the presence of 100 μg/mL RNAase (Sigma; 1 hour at 37°C). Cells were analyzed by flow cytometry (FACSVerse, BD Biosciences). Histograms of the singlet live cell population in the PI emission spectrum were drawn using the cell-cycle function in FlowJo and analyses of cell-cycle percentages performed in GraphPad.

Karpas 299 xenografts were established in 6- to 8-week-old female Fox Chase SCID (C.B-17/Icr-Prkdcscid, Charles River Laboratories) by implanting 10⁷ cells subcutaneously into their flanks. Su-DHL-1 xenografts were established by injection of 10⁷ cells (50% Matrigel) into 10-week-old nude mice [Nu (NCr)-Fox1nu, Charles River Laboratories]. When tumors reached approximately 100–155 mm³ mice were randomly allocated into groups to receive ADCT-301, nonbinding ADC, PBS (vehicle), or Adcetris intravenously. Clinical grade Adcetris was purchased from Seattle Genetics. Mice were in the body weight range of 15.5 to 27.6 g on day 1 of the studies. Each animal was euthanized when its tumor reached the endpoint volume of 2,000 mm³ or at study end. The time-to-endpoint (TTE) for analysis was calculated for each mouse by the following equation:

where * of the line was obtained by linear regression of a log-transformed tumor growth dataset

The log-rank test was used to analyze the significance of differences between the TTE values of two groups. The setup of experimental procedures used in Karpas 299 and Ramos disseminated models are described in Supplementary Methods and Tables.

ADCT-301 was administered intravenously as a single dose (1–12 mg/kg) to non–tumor-bearing CB.17 SCID mice, following which mice were observed for weight changes and toxicities. Alternatively, to measure the pharmacokinetic profile of naked antibody, HuMax-Tac or ADCT-301, 1 mg/kg of each drug was administered intravenously as a single dose to non–tumor–bearing mice. Blood collection was performed at specified time points (25 μL of blood from mice tail veins diluted in 225-μL PBS in 6% EDTA). The samples were spun down at 400 × g for 5 minutes, the supernatant was aspirated and stored at −80°C, and measurement of samples' total antibody and ADC (DAR ≥ 1) concentrations performed by an optimized ELISA assay. To measure antibody in each sample, Immuno MaxiSorp plates (Thermo Fisher Scientific) were first coated with sCD25 ‐Strepavidin (Evitria) in PBS and incubated overnight (4°C). Alternatively, to measure conjugated ADC, DAR ≥ 1, identical plates were coated with anti‐PBD antibody in PBS and incubated overnight (4°C). Plates were then washed three times (0.05% Tween) and blocked (3% BSA/PBS) for 1 hour. Reference, QC, and test samples were prepared and added to polypropylene plates (Thermo Fisher Scientific; 2 hours at room temperature). They were then washed three times and mouse anti‐human IgG1‐-HRP conjugate (Sanquin) added and plates incubated for 1 hour at room temperature. Plates were washed three times, Ultra TMB (Thermo Fisher Scientific) added, and incubated for 10 minutes at room temperature. Stop solution of 1 mol/L HCL was added and absorbance was read at 450 nm on an Envision plate reader. To quantify absorbance, calibration curves were established in assay buffer (0.5× Blotto + 0.05% Tween in PBS) containing 2.5% or 0.5% pooled mouse serum. The pharmacokinetic analysis of ADCT-301 and HuMax-TAC concentrations was performed using Phoenix WinNonlin Version 6.2 with noncompartmental analysis.

The Institutional Animal Care and Use Program at Charles River Discovery Services are accredited by the Association for Assessment and Accreditation of Laboratory Animal Care International (AAALAC), which assures compliance with accepted standards for the care and use of laboratory animals.

Murine xenograft tumors untreated or treated with 0.2 mg/kg or 0.6 mg/kg ADCT-301 were fixed in buffered formalin and embedded in paraffin according to a conventional protocol. Two to 5-μm sections were cut and transferred on electrically charged slides to be subjected to IHC. Testing of an anti-SG3249 payload antibody (14B3-B7; ADC Therapeutics) and an antibody to detect phosphorylation of Histone H2A.X at position Ser139 (New England Biolabs) was carried out with staining conditions (i.e., antibody dilution, incubation time, and antigen retrieval method) optimized on ADCT-301–treated or irradiated Karpas 299 cytospins. Single immunostaining was performed using a BOND-III Autostainer (Leica Microsystems) according to protocols described elsewhere (32, 33). Stained slides were scanned with the NanoZoomer Digital Pathology System (Hamamatsu) and images reviewed with NDP viewer.

ADCT-301 is an ADC composed of HuMax-TAC, a human IgG1 anti-CD25 antibody, maleimide-conjugated at reduced interchain cysteines via a cathepsin-cleavable valine-alanine (val-ala) linker to PBD dimer warhead SG3199 (Fig. 1A). The ADC was 97% monodisperse as determined by size-exclusion chromatography (Supplementary Fig. S1A; Supplementary Table S1). Drug antibody ratio (DAR) was determined by reverse-phase liquid chromatography to be 2.25 (Supplementary Fig. S1B; Supplementary Table S2) and by hydrophobic interaction chromatography to be 2.19 (Supplementary Fig. S1C; Supplementary Table S3).

HuMax-TAC and ADCT-301 showed strong binding affinity to CD25-positive human anaplastic large cell lymphoma–derived cell lines Karpas 299 and Su-DHL-1 (Fig. 1B). Estimated EC_50s were in the subnanomolar range and were marginally lower for HuMax-TAC than ADCT-301 in both lines. Binding saturated at around 1 μg/mL for both cell lines, but the MFI was approximately 3-fold higher in the Su-DHL-1 compared with the Karpas 299 cell line reflecting the difference in copy number (Table 1). In contrast, a nonbinding ADC containing the same drug linker payload showed no binding to either cell line at 30 μg/mL. Using soluble CD25 ectodomain, equilibrium dissociation constant (K_D) measurements of HuMax-TAC or ADCT-301 determined by surface plasmon resonance (SPR) were 15 pmol/L and 20 pmol/L, respectively. Representative SPR sensorgrams for HuMax-TAC and ADCT-301 are shown in Fig. 1C and D, respectively.

The in vitro cytotoxicity of ADCT-301 was determined in a panel of human lymphoma CD25-expressing (Su-DHL-1, HDLM-2, L540, Karpas 299) and nonexpressing (Ramos, HuT-78, Daudi) cell lines. These values were derived from graphs shown in Supplementary Fig. S2A. In addition, cytotoxicity in myeloid leukemia lines, EoL-1 (CD25-expressing) and KG-1 (CD25 non-expressing) was determined. As shown in Table 1, the GI₅₀ values for ADCT-301 in the CD25-positive lines ranged from 0.04–2.94 ng/mL. GI₅₀ values in the CD25-negative lines were >1,000 ng/mL. The nonbinding ADC gave GI₅₀ values >1,000 ng/mL in both antigen-expressing and nonexpressing lines (Supplementary Fig. S2B). No clear relationship between GI₅₀ value and mean CD25 copy number was observed for ADCT-301 in the five CD25-positive cell lines (r = −0.07; P = 0.91). In contrast to the ADC, the naked warhead, SG3199, was cytotoxic against all the lines tested and did not discriminate between CD25-expressing and nonexpressing cells (Table 1). HuMax-TAC antibody up to 1,000 ng/mL was not cytotoxic in CD25-positive or CD25-negative cell lines (Supplementary Fig. S2C).

After exposure to ADCT-301, reduction of CD25-negative Ramos cell viability was shown when cocultured (50:50) with CD25-expressing Karpas 299 cells, indicating a bystander killing effect (Fig. 2A, top). ADCT-301 at both 1 and 10 ng/mL induced a highly significant difference (P ≤ 0.001) in Ramos viability in monoculture compared with Ramos + Karpas 299 coculture. A significant difference (P ≤ 0.05) between GI₅₀ values for Ramos treated for 96 hours with the medium taken from 48-hour ADCT-301–treated Karpas 299 and Ramos treated for 96 hours with ADCT-301 in fresh medium was also demonstrated (Fig. 2B, top). In contrast, there was no significant effect on Ramos survival when treated with 1 or 10 ng/mL of nonbinding ADC (Fig. 2A, bottom) or conditioned media taken from nonbinding ADC-treated Karpas 299 (Fig. 2B, bottom).

CD25-expressing cells exposed to ADCT-301 (1 hour at 4°C) showed prominent cell surface binding (T = 0 hour; Fig. 2C). On incubation at 37°C, ADCT-301 was beginning to be internalized by 1 hour with cell membrane labeling still evident and colocalization with lysosomes observed by 4 hours. In contrast, no cell surface binding was observed with a nonbinding ADC at T = 0 hours. For Karpas 299, there was a reduction of CD25 molecules on the cell surface, as measured by a reduction in MFI of cells labeled with a FITC-conjugated noncompetitive anti-CD25 antibody, over the first three hours after treatment. The change in surface CD25 molecules after ADCT-301 exposure was significant (P ≤ 0.05) compared with that observed after nonbinding ADC exposure. A return to pretreatment CD25 MFI occurred by 16 hours (Fig. 2D). In addition to a change in CD25 receptor number, the MFI of cells incubated with ADCT-301 and stained with an AF488-labeled anti-human IgG antibody was reduced over a time course up to 4 hours after incubation at 37°C (Fig. 2E).

DNA interstrand cross-linking was measured using a modification of the single-cell gel electrophoresis (comet) assay. Representative comet images are shown in Fig. 3A. Cross-linking was expressed as the percentage decrease in the OTM compared with control irradiated cells. Following a 2-hour exposure of Karpas 299 cells to 20 ng/mL of ADCT-301, the time course of cross-link formation is shown in Fig. 3B. Cross-links formed after an initial delay, reaching a peak between 4 and 8 hours. In contrast, the free warhead at an equimolar equivalent warhead dose reached the peak of cross-linking during the 2-hour exposure. In both cases, cross-links persisted over at least 36 hours. The nonbinding ADC at 20 ng/mL did not show a significant reduction in OTM over the same time course. An equivalent experiment in Su-DHL-1 cells gave the same kinetics of cross-link formation resulting in a slightly higher peak (Fig. 3C), possibly reflecting the higher CD25 copy number and sensitivity of this line to ADCT-301.

At maximum cross-linking (24 hours) in both Karpas 299 (Fig. 3D) and Su-DHL-1 (Fig. 3E), a correlation between loss of viability and cross-link formation (r = 0.9745 and r = 0.9485, respectively; P ≤ 0.05) was established. No evidence of cross-linking after treatment with 1 μg/mL ADCT-301 could be demonstrated in CD25-negative Daudi cells (Supplementary Fig. S3). However, cross-linking by naked warhead in CD25-negative Daudi was similar to that in CD25-positive cell lines.

We have previously shown that PBD-induced DNA interstrand cross-links induce a robust, but delayed γ-H2AX response (34). In Karpas 299 cells, γ-H2AX foci were observed 24 hours after a 2-hour exposure to sub-GI₅₀ concentrations of ADCT-301 (Fig. 4A). After treatment with 60 pg/mL or 1,000 pg/mL ADCT-301, γ-H2AX foci increased by 17- or 34-fold relative to untreated cells, respectively (P ≤ 0.05 and P ≤ 0.01). In these cells, continuous exposure to 3 or 10 ng/mL ADCT-301 resulted in a dose-dependent G₂–M arrest reaching a maximum at 48 hours as evidenced by a decreased percentage of cells in G₀–G₁ and an increased percentage in G₂–M compared with untreated control (Fig. 4B). A GI₅₀ warhead concentration (15 pmol/L) induced G₂–M cell-cycle arrest in a similar fashion to a GI₅₀ concentration of ADCT-301 (3 ng/mL, containing the equivalent of 46 pmol/L of payload in this cell line), although this occurred earlier peaking at 24 hours (Fig. 4C). After continuous exposure to 10 ng/mL ADCT-301, the peak of the early apoptosis marker Annexin-V on the cell surface of Karpas 299 cells was observed between 60 and 72 hours (Fig. 4D). Maximal loss of viability was at 96 hours. Nonbinding ADC (10 ng/mL) did not induce cell-cycle G₂–M arrest (Fig. 4B) or increased Annexin-V and loss of viability (Fig. 4D).

In vivo, ADCT-301 demonstrated dose-dependent antitumor activity against both subcutaneous and disseminated lymphoma models. In the subcutaneous CD25-expressing Karpas 299 xenograft, a single dose of ADCT-301 delayed tumor growth in a dose-dependent fashion with the 0.6 mg/kg cohort demonstrating 10 of 10 tumor-free survivors at day 60 (Fig. 5A). TTE survival analyses compared with vehicle control were superior at 0.1 mg/kg (log-rank test P ≤ 0.05) and at 0.2, 0.4, and 0.6 mg/kg (P ≤ 0.001; Supplementary Fig. S4A). Similarly a single dose of 0.5 mg/kg ADCT-301 produced 10 of 10 tumor-free survivors maintained at day 75, whereas a single dose of 0.5 mg/kg Adcetris (targeting CD30) only delayed tumor growth and even four 0.5 mg/kg doses (q4d × 4) failed to cure mice (Fig. 5B). CD30 expression on Karpas 299 is two to three times higher than CD25 (inset to Fig. 5B). Kaplan–Meier analyses showed ADCT-301 significantly superior (P ≤ 0.01) to Adcetris single-dose administration (Supplementary Fig. S4B). In a disseminated Karpas 299 model, single-dose ADCT-301 also showed significant dose-dependent extensions of survival in comparison with vehicle control at all four doses tested (P ≤ 0.001; Supplementary Fig. S5A). Single-dose ADCT-301 at 0.6 mg/kg showed significant (P ≤ 0.05) improvement in survival compared with single-dose Adcetris administration (Supplementary Fig. S5B). In both subcutaneous and disseminated models (0.5 mg/kg and 0.6 mg/kg, respectively), nonbinding ADC failed to show significant tumor regression or survival benefit compared with vehicle controls (Fig. 5B and Supplementary Figs. S4B and S5A). In the subcutaneous Su-DHL-1 xenograft, a single dose of ADCT-301 delayed tumor growth in a dose-dependent fashion with the 0.6 mg/kg cohort demonstrating 6 of 8 tumor-free survivors at day 74 (Fig. 5C). In contrast, single-dose (0.5 mg/kg) ADCT-301 in disseminated CD25-negative Ramos xenografts failed to show a survival benefit (Supplementary Fig. S6). In non–tumor-bearing SCID mice the MTD, as defined by less than 10% weight loss, was 9 mg/kg ADCT-301. The half-life of naked antibody, HuMax-Tac, was 7.3 days. Similarly, the half-life of ADCT-301 by analysis of the conjugated antibody with DAR component ≥ 1 was 7.3 days.

In SCID mice with Karpas 299 subcutaneous tumors, a single dose of ADCT-301 was administered at 0.2 or 0.6 mg/kg. Twenty-four hours after treatment, excised tumors showed a dose-related increase in intensity of staining by an anti-PBD payload antibody and a γ-H2AX antibody (Fig. 5D). A magnified section of the ADCT-301 0.6 mg/kg anti-PBD payload antibody panel indicates that the tumor cells demonstrate both membrane and cytosolic staining. Reduction in OTM was 23% (0.2 mg/kg) versus 49% (0.6 mg/kg; P ≤ 0.01) using the comet assay (Fig. 5E). No cross-linking was observed in peripheral blood lymphocyte samples from the same mice (Supplementary Fig. S7). The selective targeting of ADCT-301 to CD25-expressing tumor cells and resulting formation of DNA interstrand cross-links and associated DNA damage response are therefore confirmed in vivo.

ADCs delivering PBD dimer payloads represent a novel mode of action in the ADC arena. They exploit a completely different cellular mechanism to the widely used auristatin and maytansinoid tubulin inhibitors and a different mode of action to other DNA-interacting warheads such as calicheamicin (27). Several PBD-containing ADCs are now in phase I clinical trials including SGN-33A (27), SGN-70A (35), SC16LD6.5 (36), and ADCT-301. In the synthesis of ADCT-301, the warhead SG3199 is connected via its N10 position to a maleimidocaproyl val-ala dipeptide linker (as in SC16LD6.5). The maleimide then covalently and stochastically reacts with a limited number of reduced inter-chain disulphide bonds in HuMax-TAC. In contrast, SGN-33A and SGN-70A are conjugated via a val-ala linker to the C2 position on the PBD dimer SGD-1882. As the N10 position imine is involved in covalent binding to DNA, linker attachment at this point produces a prodrug and an ADC payload as well as potentially adding a further level of safety. As a result, attachment at the N10 imine requires a self-immolative linker that becomes completely traceless following cleavage.

Binding of ADCT-301 and HuMax-TAC to CD25-positive cell lines (flow cytometry) and to soluble CD25 (SPR) was comparable, indicating that conjugation to PBD linker did not affect antigen binding. Our in vitro models show that HuMax-TAC is not itself directly cytotoxic. All cell lines studied were sensitive to SG3199 warhead. The CD25-positive cell line least sensitive to ADCT-301 was also least sensitive to SG3199 warhead. In all four CD25-positive lymphoma cell lines tested, picomolar potency to ADCT-301 is demonstrated but there was no clear relation between sensitivity to ADCT-301 and CD25 expression level. All these cells lines, however, had high levels of CD25 expression (77,000–310,000) and cell lines with much lower CD25 expression may be required to determine the threshold for cytotoxic sensitivity. An eosinophilic leukemia cell line, EoL-1, with a surface CD25 expression of 17,000 demonstrated femtomolar in vitro sensitivity to ADCT-301 and free SG3199, suggesting that tumor type sensitivity to free warhead varies considerably between tumor types and may predict targeted ADC cytotoxicity independently of CD25 expression level above a certain threshold expression level.

If ADCT-301 is to be efficacious in lymphomas with heterogeneous CD25 expression (2), the existence of bystander toxicity on target-negative tumor cells is important. PBD warhead released from the target-positive cell would be expected to be diffusible in contrast to less permeable payloads such as MMAF (37). Initial coculture experiments of CD25-negative Ramos with CD25-positive Karpas 299 resulted in loss of Ramos cells, possibly due to the Treg-suppressive phenotype of Karpas 299 (38). Therefore, Ramos and Karpas 299 were cocultured in the presence of a TGF-β signaling antagonist to ablate the Treg-suppressive effect (39). In both medium transfer and coculture experiments, Ramos cells were killed by a soluble factor released from ADCT-301–treated Karpas 299 cells into the medium. We hypothesize that this soluble factor is SG3199 warhead, released by lysosomal cleavage of the val-ala dipeptide linker in ADCT-301 and the self-immolative cleavage of the residual linker stub on the PBD N10 imine, as we have shown that any transfer of intact ADC would be inactive against the CD25-negative Ramos cells.

Evidence for internalization of ADCT-301 is provided both by the decline of detected ADC on the cell surface of Karpas 299 and the reduction in intensity of membrane immunofluorescent staining on L-540 cells. Furthermore, a reduction of CD25 surface expression over the first 3 hours after ADCT-301 treatment is evident, after which expression returns to pretreatment levels by 16 hours. There is no binding competition between the two anti-CD25 antibodies used in this latter assay suggesting that the decline in the CD25 copy number represents internalization and downmodulation. Immunofluorescent colocalization images show that processing of ADCT-301 is at least, in part, lysosomal.

An important feature of the highly cytotoxic interstrand cross-linking produced by PBD dimer warheads is the minimal DNA distortion which may be responsible for a lack of repair of these cross-links and their persistence (40, 41). The time lag between the peak of cross-linking by ADCT-301 and by equimolar warhead likely reflects the slower time taken for ADC internalization and cellular processing compared with the diffusible active warhead. We detected no cross-link repair over 36 hours. The strong correlation between the reduction in OTM and the loss of cell survival indicates that DNA cross-linking is an important mode of action of ADCT-301.

A DNA damage response (increased γ-H2AX foci) is evident in CD25-positive cells 24 hours after a 2-hour exposure to ADCT-301 at concentrations as low as 60 pg/mL. This supports previous observations that γ-H2AX is highly sensitive to cellular events induced by DNA cross-linking agents including PBD dimers (34, 41). DNA cross-linking leads to G₂–M cell-cycle arrest and apoptotic cell death (40, 42). We have shown that ADCT-301 caused G₂–M arrest peaking at 48 hours and apoptotic cell death. Thus, we confirm that DNA cross-linking peaked by 4 to 8 hours and was followed by a γ-H2AX response by 24 hours. G₂–M arrest ensued at 48 hours, phosphatidylserine and phosphatidyltyramine exposure at 72 hours, and loss of cell viability at 96 hours.

The MTD, pharmacokinetics, and stability of ADCT-301 were explored in non–tumor-bearing SCID mice. These mice tolerated ADCT-301 up to 9 mg/kg. Measured half-lives of naked antibody and ADCT-301 in mice serum were identical. CD25 is constitutively expressed on Treg and on activated cytotoxic T lymphocytes, B lymphocytes (43), macrophages (44, 45), eosinophils (46), and basophils (47). Anti-CD25 immunoconjugates may contribute to the depletion of intratumoral CD25-expressing T cells including Tregs as has been demonstrated for ⁹⁰Yttrium-labeled daclizumab (18) and LMB-2 (48). Although the safety and efficacy of administering anti-CD25 radiommunoconjugates (16–18) and immunotoxins (19, 20) has been established in previous phase I trials, the consequences of prolonged ablation of Tregs remains to be determined.

Mice xenografted with Karpas 299 and treated with ADCT-301 show remarkable dose–response improvements in survival. Single-dose ADCT-301 shows superiority to single-dose Adcetris administration in TTE statistical analyses despite the 3-fold higher cell surface target expression for Adcetris and the higher DAR of Adcetris versus ADCT-301. Another CD25-expressing xenografted cell line, Su-DHL-1, showed significant dose-dependent delays in tumor growth and tumor-free survival after treatment with ADCT-301. The targeted in vivo efficacy of ADCT-301 is illustrated in the lack of efficacy in both nontargeted ADC treatment of Karpas 299 xenografts and in ADCT-301 treatment of a CD25-negative Ramos model.

Using a monoclonal anti-PBD antibody against a PBD payload and anti-γ-H2AX staining has confirmed the dose-dependent delivery of ADC to the tumor xenograft. The comet assay and γ-H2AX immunofluorescence have been used as pharmacodynamic assays in previous clinical trials of PBD dimer, SG2000 (25, 26) and the comet assay was used to quantitatively assess DNA damage in excised tumor samples in the current study. Extrapolating from in vivo data, the mean reduction in OTM in 0.6 mg/kg-treated tumor cells compared with the 0.2 mg/kg cohort, suggests that a threshold level of DNA cross-linking may be required for in vivo efficacy. The absence of significant DNA cross-linking in SCID mouse PBMCs (CD25-negative) supports the CD25 specificity of ADCT-301 in vivo. These comet data thus provide a relevant pharmacodynamic assay for use in the current clinical development of PBD-based ADCs including ADCT-301.

P.C. Tyrer has ownership interest (including patents) in Medimmune. F. D'Hooge is the Head of the Bioconjugation Unit at Novasep. P.W.H.I. Parren has ownership interest (including patents) in Genmab. P.W. Howard has ownership interest (including patents) in ADC Therapeutics. P.H.C. van Berkel has ownership interest (including patents) in ADC Therapeutics. J.A. Hartley reports receiving a commercial research grant from, has ownership interest (including patents) in, and is a consultant/advisory board member for ADC Therapeutics. No potential conflicts of interest were disclosed by the other authors.

Conception and design: M. Flynn, C.E.G. Havenith, J.-N. Levy, A. Tiberghien, L.A. Masterson, P.W.H.I. Parren, D.G. Williams, P.W. Howard, P.H.C. van Berkel, J.A. Hartley

Development of methodology: M. Flynn, P.C. Tyrer, C.E. Britten, C.E.G. Havenith, L.A. Masterson, T. Marafioti, P.W. Howard, J.A. Hartley

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Flynn, P.C. Tyrer, N. Janghra, C.E. Britten, J.-N. Levy, A. Tiberghien, C. Barry, F. D'Hooge

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Flynn, F. Zammarchi, N. Janghra, C.E. Britten, C.E.G. Havenith, J.-N. Levy, A. Tiberghien, L.A. Masterson, C. Barry, F. D'Hooge, T. Marafioti, P.H.C. van Berkel, J.A. Hartley

Writing, review, and/or revision of the manuscript: M. Flynn, F. Zammarchi, C.E.G. Havenith, F. D'Hooge, P.W.H.I. Parren, D.G. Williams, P.W. Howard, J.A. Hartley

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Flynn, F. Zammarchi, A.U. Akarca, C.E. Britten, C.E.G. Havenith, F. D'Hooge

Study supervision: F. Zammarchi, P.H.C. van Berkel, J.A. Hartley

Other (performed IHC-related experiments and provided images via scanning): A.U. Akarca

Other (development of pbd chemistry): L.A. Masterson

The authors thank Charles River Laboratory Services for conducting the majority of the in vivo efficacy studies. Much appreciation is extended to members of other departments at UCL including GCLP staff, Janet Hartley, and Victoria Spanswick for sharing their expertise (comet assay and γ-H2AX immunofluorescence), staff of the Biological Services Unit, Mathew Robson and Barbara Pedley for their assistance and guidance in conducting additional in vivo experiments, and Vishvesh Shende and Joseph Linares from the Pathology Department for their help in setting up IHC assays. Other laboratory staff at Spirogen and ADCT in both chemistry and biology departments are acknowledged for their guidance in setting up assays used in this project.

J.A. Hartley acknowledges Programme Grant support from Cancer Research UK (C2559/A16569).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.