Purpose:

Mortality due to acute myeloid leukemia (AML) remains high, and the management of relapsed or refractory AML continues to be therapeutically challenging. The reapproval of Mylotarg, an anti-CD33–calicheamicin antibody–drug conjugate (ADC), has provided a proof of concept for an ADC-based therapeutic for AML. Several other ADCs have since entered clinical development of AML, but have met with limited success. We sought to develop a next-generation ADC for AML with a wide therapeutic index (TI) that overcomes the shortcomings of previous generations of ADCs.

Experimental Design:

We compared the TI of our novel CD33-targeted ADC platform with other currently available CD33-targeted ADCs in preclinical models of AML. Next, using this next-generation ADC platform, we performed a head-to-head comparison of two attractive AML antigens, CD33 and CD123.

Results:

Our novel ADC platform offered improved safety and TI when compared with certain currently available ADC platforms in preclinical models of AML. Differentiation between the CD33- and CD123-targeted ADCs was observed in safety studies conducted in cynomolgus monkeys. The CD33-targeted ADC produced severe hematologic toxicity, whereas minimal hematologic toxicity was observed with the CD123-targeted ADC at the same doses and exposures. The improved toxicity profile of an ADC targeting CD123 over CD33 was consistent with the more restricted expression of CD123 in normal tissues.

Conclusions:

We optimized all components of ADC design (i.e., leukemia antigen, antibody, and linker-payload) to develop an ADC that has the potential to translate into an effective new therapy against AML.

Translational Relevance

The reapproval of gemtuzumab ozogamicin (Mylotarg) has demonstrated antibody–drug conjugates (ADCs) as a clinically validated therapeutic strategy for the treatment of acute myeloid leukemia (AML). However, improvements in ADC technology are warranted to enable a successful therapeutic strategy for the treatment of AML that would be potent against leukemic cells, while exhibiting a favorable safety profile. Here, we describe the development of a next-generation ADC platform that combines the elements of a novel DNA-damaging linker-payload with a site-specific conjugation methodology to provide a wider therapeutic index compared with certain existing platforms. In addition, we provide head-to-head efficacy and toxicity comparison of targeting two attractive AML antigens, CD33 and CD123, using an ADC approach. Overall, we optimized all components of ADC design (i.e., tumor antigen, antibody, and linker-payload) to develop a candidate that may translate into an effective new therapy against AML.

The current induction treatment of acute myeloid leukemia (AML) for fit patients is a combination of chemotherapeutics (cytarabine with anthracyclines), resulting in a complete remission rate of 60%–80% (1, 2). Unfortunately, more than 50% of the patients that initially respond eventually relapse, which is, in part, attributed to multidrug resistance (MDR) and AML reconstitution by leukemia-initiating cells (LSC; refs. 3–5). For unfit patients with AML who cannot tolerate an aggressive chemotherapy regimen, treatment options are limited, but recently a promising combination treatment of azacitidine (Vidaza) and venetoclax (Venclexta) has shown robust responses compared with azacytidine alone (6, 7). A successful novel therapeutic strategy for the treatment of AML would be highly potent, durable, and selectively killing leukemic cells, while exhibiting a favorable safety profile and enabling combination therapies.

Antibody–drug conjugates (ADCs) consist of three distinct components: (i) a mAb that binds to a tumor target cell surface antigen, (ii) a cytotoxic payload that kills the tumor cells, and (iii) a linker that stably joins the antibody and payload. This strategy allows targeted delivery and release of the cytotoxic agent in cancer cells with relatively higher level of target antigen compared with healthy cells. Gemtuzumab ozogamicin (Mylotarg) is the first FDA-approved CD33-targeted ADC for the treatment of AML. Gemtuzumab ozogamicin is a humanized anti-CD33 IgG4 mAb conjugated to the cytotoxic agent, calicheamicin, via an acid-labile hydrazone linker. Despite clinical efficacy, treatment with gemtuzumab ozogamicin is associated with serious adverse effects (8). Also, calicheamicin, the payload used in gemtuzumab ozogamicin, is susceptible to drug efflux pumps (9). Vadastuximab talirine, another ADC targeting CD33, is comprised of pyrrolobenzodiazepine (PBD) dimers and protease-cleavable maleimidocaproyl-valinealanine dipeptide linkers that are site specifically conjugated to an anti-CD33 mAb. However, vadastuximab talirine was withdrawn from clinical development due to a high rate of death, including fatal infections (10). The approval of gemtuzumab ozogamicin provides proof of concept for an ADC strategy in the treatment of AML. However, the lack of success of the ADCs that has followed has exacerbated the need for a next-generation ADC with a favorable pharmacokinetics profile, superior efficacy, and improved therapeutic index (TI).

CD33 and CD123 are highly expressed in a variety of hematologic malignancies, including AML, where >80% of patients harbor leukemias positive for both markers (11–13). Both antigens are expressed in leukemic blasts, as well as in LSCs. Thus, CD33 and CD123 are attractive oncology targets, and multiple types of therapeutics targeting these antigens, including ADCs, have been developed for hematologic cancers (3, 13–16). CD33 is a sialic acid–binding immunoglobulin-like lectin molecule that functions as a negative regulator of myeloid cell activation. CD123 is the alpha chain of the IL3 receptor that forms a heterodimer with the common beta chain CD131 to transmit the IL3 signal. Although very low or undetectable levels of CD33 and CD123 are expressed on hematopoietic stem cells (17–19), CD33 is expressed on myeloid progenitors and differentiated myeloid cells (20–22). In contrast, CD123 has a more restricted expression pattern in healthy human bone marrow (BM; ref. 23). This is particularly relevant for BM-derived diseases, such as AML, where identifying cancer-specific targets that are not expressed robustly in the normal BM is of paramount importance.

Here, we report, for the first time, the development of a novel DNA-damaging payload, cyclopropa [c] pyrrolo[3,2-e] indole-4-one dimer (CPI dimer), that is attached to an engineered Y296Q residue in the antibody heavy chain via transglutaminase-mediated conjugation (24, 25) to enable an ADC preparation consisting of a homogeneous load of drug-to-antibody ratio (DAR) of 2. Using this linker-payload, we show that our ADC platform circumvents the limitations observed with certain other ADC platforms and provides a wider TI. Next, we made a head-to-head comparison of the two AML surface proteins, CD33 and CD123, as ADC targets, which to our knowledge has not been reported before. We found that although both CD33- and CD123-ADCs showed comparable efficacy in several disseminated AML patient-derived xenograft (PDX) models, the CD123-ADC showed a more favorable safety profile in nonhuman primates, which we attribute to its more restricted expression pattern in normal BM. We believe that the CD123-ADC would have broad utility across the AML disease spectrum, including safety in the unfit patient population, which could potentially transform the clinical management of AML.

A complete Methods section is provided in the Supplementary Materials and Methods.

Generation of CD33- and CD123-CPI dimer ADC, and clinical CD33-ADCs

The Supplementary Materials and Methods provide details on the generation of CD33 and CD123 site-specific antibodies, synthesis of linker-CPI dimer payload, conjugation and purification of CPI dimer ADCs, and generation of clinical CD33-ADCs.

DNA cross-linking assay

Using methods adapted from previous reports (26), pBR322 plasmid was linearized with BamHI, dephosphorylated, and then purified by using QIAquick Kit (Qiagen). Purified DNA (25 ng/mL) was incubated with CPI analogue for 2 hours at 37°C in 15 mmol/L Tris, 0.75 mmol/L EDTA, and 75 mmol/L NaCl, pH 7.5. DNA strands were chemically separated by alkaline denaturation in 0.25 N NaOH for 30 minutes at 25°C, then fractionated by neutral 1% agarose/TAE electrophoresis, and stained with GelRed to differentiate ssDNA and double-stranded DNA.

Western blotting

Cells were treated at indicated timepoints and lysed on ice for 10 minutes followed by sonication for 3 minutes (Cell Lysis Buffer from Cell Signaling Technology, supplemented with PhosSTOP and Complete Protease Inhibitor; Roche Applied Science). Lysates were centrifuged for 10 minutes at 14,000 × g in a cold microfuge, and the supernatant was used for immunoblotting. The Supplementary Materials and Methods provide details on the antibodies used.

Cell culture

All human cell lines (MOLM13, MV4-11, NB4, HL60, OCI-AML3, JVM3, and Granta519) were grown in RPMI1640 media (Thermo Fisher Scientific) supplemented with 10% FBS, GlutaMAX, and 1% Penicillin/Streptomycin (Gibco). MOLM13, NB4, OCI-AML3, JVM3, and Granta519 cells were obtained from the Leibniz Institute DSMZ-German Collection of Microorganisms and Cell Cultures. MV4-11, HL60, and TF-1 cells were obtained from the ATCC.

Internalization assay

Cells were incubated with pHrodo (Thermo Fisher Scientific) intracellular indicator and with 20 μg/mL of anti–CD33-AF647 or anti–CD123-AF647 antibodies for 30 minutes on ice. Cells were then moved to Spinning Disk Microscopy (PerkinElmer) at 37°C to acquire images every 5 minutes for up to 3 hours. Images were analyzed by Volocity Software (Quorum Technologies).

In vitro cytotoxicity assay

For cytotoxicity assays, cells were plated in opaque 96-well plates (Corning) and treated with titrated compounds for 4 days. Viability was determined by CellTiter-Glo Luminescent Cell Viability Assay Kit (Promega) and luminescence was detected by using a Victor X3 Plate Reader (Perkin Elmer). The data were normalized to the control group (DMSO or PBS). IC50 values were calculated using nonlinear logistic regression, model No. 203 with XL fit v4.2 (IDBS). All experimental points were set up in two replicate wells and independently performed in duplicate. Recombinant IL3 was added to the cells together with α-CD123-ADC to determine the cytotoxicity of α-CD123-ADC in the presence of IL3.

Colony-forming unit assay

MethoCult was purchased from Stemcell Technologies and used according to the manufacturer's protocol. Briefly, CD34-enriched normal BM cells or AML patients' BM cells were treated with ADCs and then plated onto petri dishes. After 2 weeks, the colonies were enumerated.

Animal studies

All animal experiments were conducted in a facility accredited by the Association for Assessment and Accreditation of Laboratory Animal Care. All animal studies were approved by the Pfizer Institutional Animal Care and Use Committee in accordance with the guidelines described in “Guide for the Care and Use of Laboratory Animals” (NRC, 2011).

The Supplementary Materials and Methods provide details on in vivo efficacy and safety studies.

Flow cytometry analysis

The Supplementary Materials and Methods provide details on sample collection and antibodies used.

Toxicology studies

Clinical observations of cynomolgus monkeys consisted of twice daily general observations with additional sessions on dosing days; detailed clinical observations before initiation of dosing (PID), once weekly, and prior to necropsy; and body weight measurements PID, predose on each dosing day, weekly, and before scheduled necropsy. Blood samples were collected for routine hematologic evaluation until necropsy.

Next-generation linker-payload strategy robustly improves TI over other CD33-ADCs

To identify a potent payload that evades MDR, we screened more than 100 DNA-damaging agents across multiple classes in both MDR-positive HEL92.1.7 and -negative HL-60 AML cell lines (Fig. 1A). The payload used in Mylotarg, DMH-N-Ac-calicheamicin, was about 10-fold less active in HEL92.1.7 cells compared with HL-60 cells, which is consistent with previous findings that calicheamicin and its analogs are substrates of MDR (27–29). In contrast, the pyrrolobenzodiazepine dimer used in vadastuximab talirine was about 10- to 100-fold more potent than the payload used in Mylotarg and showed robust activity in MDR-positive cells (30). We observed that many payloads that belong to cyclopropa-X-indole (CXI) class exhibited femtomolar potency in both cell lines. After evaluating these further, we advanced with a new CPI dimer analogue, in which two CPI monomer motifs are joined via a bicyclopentyl spacer which provided the best balance of efficacy, stability, exposure, and safety (25, 31)

Figure 1.

A novel conjugation site and a linker-payload (Y296Q and AcLysValCit-CPI dimer) resulted in an ADC with a significantly wide TI. A, DNA-damaging payload structure–activity relationship plot. The x-axis shows the IC50 (nmol/L) of payloads in MDR-positive HEL92.1.7 cell line, and the y-axis shows the IC50 (nmol/L) of payloads in MDR-negative HL-60 cell line. The payload classes are color coded and each point in the plot represents a unique compound. The diagonal lines represent the ratio of activity in the two cell lines. The CXI dimer payloads in the blue parallelogram were further evaluated. The structures of the payloads used in Mylotarg (DMH-N-Ac-calicheamicin), vadastuximab talirine (pyrrolobenzodiazepines; PBD dimer), and CPI dimer, which is the focus of this work, are shown. B, Dose-dependent CPI dimer cross-linking of ssDNA. C, Western blot analysis of CPI dimer–induced DNA damage as measured by dose- and time-dependent increase in the phosphorylation of CHK1, CHK2, histone 2AX (gH2AX), FANCD2, and cleaved PARP in TF-1 cells that endogenously express MDR. Vinculin was used as loading control. Ctrl: control/no treatment. D, Structure of CD33-CPI dimer ADC with the linker (AcLysValCitPABC) and the payload (CPI dimer). E, Tumor volume measurements of AML xenograft models established with TF-1 MDR-positive (left) and HL60 cytarabine-resistant (right) cell lines. ADCs were given every 4 days, four cycles at the indicated doses and cytarabine (Ara-C) was dosed every day, seven cycles at 5 mg/kg. The arrows (↑) indicate dosing.

Figure 1.

A novel conjugation site and a linker-payload (Y296Q and AcLysValCit-CPI dimer) resulted in an ADC with a significantly wide TI. A, DNA-damaging payload structure–activity relationship plot. The x-axis shows the IC50 (nmol/L) of payloads in MDR-positive HEL92.1.7 cell line, and the y-axis shows the IC50 (nmol/L) of payloads in MDR-negative HL-60 cell line. The payload classes are color coded and each point in the plot represents a unique compound. The diagonal lines represent the ratio of activity in the two cell lines. The CXI dimer payloads in the blue parallelogram were further evaluated. The structures of the payloads used in Mylotarg (DMH-N-Ac-calicheamicin), vadastuximab talirine (pyrrolobenzodiazepines; PBD dimer), and CPI dimer, which is the focus of this work, are shown. B, Dose-dependent CPI dimer cross-linking of ssDNA. C, Western blot analysis of CPI dimer–induced DNA damage as measured by dose- and time-dependent increase in the phosphorylation of CHK1, CHK2, histone 2AX (gH2AX), FANCD2, and cleaved PARP in TF-1 cells that endogenously express MDR. Vinculin was used as loading control. Ctrl: control/no treatment. D, Structure of CD33-CPI dimer ADC with the linker (AcLysValCitPABC) and the payload (CPI dimer). E, Tumor volume measurements of AML xenograft models established with TF-1 MDR-positive (left) and HL60 cytarabine-resistant (right) cell lines. ADCs were given every 4 days, four cycles at the indicated doses and cytarabine (Ara-C) was dosed every day, seven cycles at 5 mg/kg. The arrows (↑) indicate dosing.

Close modal

Unlike calicheamicin that breaks double-stranded DNA and activates the ATM–CHK2 pathway, the CPI dimer alkylates and cross-links DNA, which leads to activation of both ATR and ATM pathways that effectively kill cancer cells (Fig. 1B). Furthermore, the CPI dimer can overcome MDR, as evidenced by the activation of checkpoint kinases downstream of the DNA damage response in a dose- and time-dependent manner in TF-1 cells that endogenously express high levels of the MDR1 gene (Fig. 1C). To determine whether CPI dimer can overcome resistance to standard-of-care chemotherapy agents, we generated cytarabine- and daunorubicin-resistant cell lines and tested the potency of CPI dimer in these cells. CPI dimer demonstrated a significantly higher potency in all cell lines, including the chemoresistant as well as MDR-positive cells, compared with the chemotherapy agents (Table 1). To test CPI dimer as the payload of an ADC, we chose to conjugate it to a human/cynomolgus cross-reactive CD33 antibody because CD33 is a clinically validated target for AML (Fig. 1D). Generation of a “well-behaved” ADC requires an optimal combination of a linker-payload, a specific site of linker-payload attachment on the antibody, and the conjugation method. We evaluated various combinations of linkers, conjugation sites, and chemistry for CD33-CPI dimer ADC, and discovered that a protease-cleavable linker-payload, AcLysValCitPABC-CPI dimer, site specifically conjugated by transglutaminase at an engineered residue, Y296Q, in the heavy chain constant region of the α-CD33 antibody yielded the best efficacy, stability, and lowest toxicity (refs. 25, 31; Supplementary Table S1). Next, we evaluated the CD33-CPI dimer ADC in cell line xenograft models established with MDR-positive TF-1 and cytarabine-resistant HL60 cells. In these experiments, the CD33-CPI dimer ADC showed superior efficacy at lower doses compared with gemtuzumab ozogamicin and cytarabine, respectively (Fig. 1E).

Table 1.

Summary of in vitro cytotoxicity of different payloads in the parental and standard-of-care resistant cells.

nmol/L IC50 ± SEMCytarabineDaunorubicinCalicheamicinCPI dimer
HL60 13.4 ± 8.2 6.1 ± 3.0 0.9 ± 0.3 0.005 ± 0.001 
HL60_Cyt_R >10,000a ND 0.6 ± 0.2 0.007 ± 0.001 
HEL 92.1.7 MDR+ 6.7 ± 6.1 6.9 ± 3.6 17.6 ± 7.1 0.003 ± 0.001 
HEL 92.1.7_Dau_R; MDR+ ND >1,000a >100a 0.052 ± 0.039 
TF-1 MDR+ 40.9 ± 8.2 ND 0.19 ± 0.015 0.006 ± 0.002 
nmol/L IC50 ± SEMCytarabineDaunorubicinCalicheamicinCPI dimer
HL60 13.4 ± 8.2 6.1 ± 3.0 0.9 ± 0.3 0.005 ± 0.001 
HL60_Cyt_R >10,000a ND 0.6 ± 0.2 0.007 ± 0.001 
HEL 92.1.7 MDR+ 6.7 ± 6.1 6.9 ± 3.6 17.6 ± 7.1 0.003 ± 0.001 
HEL 92.1.7_Dau_R; MDR+ ND >1,000a >100a 0.052 ± 0.039 
TF-1 MDR+ 40.9 ± 8.2 ND 0.19 ± 0.015 0.006 ± 0.002 

Note: Viability was measured by CellTiter-Glo (CTG) luminescent cell viability assay kit. IC50 values determined from n ≥ 2 experiments except for items in boldface.

Abbreviations: Cyt_R, cytarabine resistant; Dau_R, daunorubicin resistant; ND, not determined.

aSEM not available for >values.

We next asked how the tolerability of the CD33-CPI dimer ADC compares with two other CD33-ADCs that have undergone clinical evaluation (Supplementary Fig. S1). We performed a single-dose tolerability study with CD33-PBD dimer [an analogue of the clinical conjugate, vadastuximab talirine (generated in-house)], CD33-calich (clinical conjugate, gemtuzumab ozogamicin), and CD33-CPI dimer ADCs in rats (Table 2). As all three ADCs are not rodent cross-reactive, our evaluation excluded on-target toxicities and focused on assessing tolerability. Male rats dosed with CD33-PBD dimer or CD33-calich ADC at 3 or 2 mg/kg, respectively, did not tolerate these doses. In contrast, rats tolerated doses up to 10 mg/kg with CD33-CPI dimer ADC. Exposure (AUC) of the total antibody or total ADC in rat serum from the MTDs was measured. The CD33-CPI dimer ADC had a TI of >8.2×, much higher than the TIs for the CD33-PBD dimer or CD33-calich ADCs (0.8× or 2.2×, respectively). This suggests that our ADC platform has increased safety when combined with previous generations of CD33-targeted ADCs.

Table 2.

Summary of tolerability from single-dose exploratory toxicity studies in rats using three CD33 ADCs.

CD33-PBD analogue of clinical conjugateCD33-Calich clinical conjugateCD33-CPI preclinical conjugate
ADCα-CD33-S239C-mcValAla-PBDα-CD33-cG1-AcBut-calichα-CD33-Y296Q-AcLysValCit-CPI
Dose (mg/kg) 0.3 0.8 1.4 10 
Mortality          
MTD         
AUC (0–tau) (μg·h/mL) day 1  1,165a   1,085b    17,233a 
TI  0.8   2.2   — >8.2 
CD33-PBD analogue of clinical conjugateCD33-Calich clinical conjugateCD33-CPI preclinical conjugate
ADCα-CD33-S239C-mcValAla-PBDα-CD33-cG1-AcBut-calichα-CD33-Y296Q-AcLysValCit-CPI
Dose (mg/kg) 0.3 0.8 1.4 10 
Mortality          
MTD         
AUC (0–tau) (μg·h/mL) day 1  1,165a   1,085b    17,233a 
TI  0.8   2.2   — >8.2 

Note: CD33-PBD: α-CD33-S239C-mcvalAla- pyrrolobenzodiazepine dimer, in-house generated analogue of the clinical compound vadastuximab talirine. CD33-Calich: α-CD33(P67.6)-cG1-AcBut-calicheamicin, clinical compound, gemtuzumab ozogamicin. CD33-CPI: α-CD33(11A1)-Y296Q-AcLysValCit-CPI dimer. In this study, MTD was defined by the observation of unanticipated death or changes in body weight, food consumption, or clinical signs deemed unacceptable by Institutional Animal Care and Use Committee guidance at the testing facility and prompting immediate euthanasia. >, refers to predicted MTD above 10 mpk dose. Higher doses not tested. TI was calculated using AUC values reported from these single-dose rat studies divided by the calculated tumor stasis concentration from in vivo efficacy studies in mice (HL60 model for CD33-PBD and CD33-Calich and TF-1 model for CD33-CPI).

aRefers to an AUC from Ab.

bRefers to an AUC from ADC concentration profiles.

Generation of a CD123-targeted ADC using the novel linker-payload strategy

We asked whether targeting CD123 with the same strategy could further increase the TI providing the first head-to-head comparison of CD33 and CD123 as ADC targets in AML. We first assessed the expression of CD33 and CD123 on the surface of 41 AML patient BM samples by flow cytometry and found that both antigens were frequently and ubiquitously expressed in the leukemic cells (Supplementary Table S2), consistent with previous reports (3, 4, 11–13, 15, 17, 32). We then applied the CPI dimer–based ADC platform to generate a human/cynomolgus cross-reactive CD123-targeted ADC. Both ADCs have the same payload, the CPI dimer, and proteolytically cleavable linker, AcLysValCitPABC, that is attached by transglutaminase-mediated chemistry at Y296Q in the heavy chain constant domains of the anti-CD123 antibody. The only difference between both ADCs is their complementarity-determining regions (CDR), where these molecules bind to their specific antigen. Furthermore, both ADCs have the same DAR2 and the processing of linker-payload to active CPI dimer payload (Supplementary Fig. S2). From here on, we refer to them as CD33-ADC or CD123-ADC. We also generated a control ADC comprising a nonbinding mAb of the same isotype conjugated to the same linker-payload via the same transglutaminase approach. The negative control antibody was engineered using human germline heavy and light chain variable region sequences that include random CDR3 sequences obtained from public databases with no known corresponding antigens. Radiolabeled and/or fluorophore-labeled control antibody preparations were evaluated for binding to a panel of human, cynomolgus monkey, and mouse tissues, including, but not limited to, skeletal muscle, skin, heart, kidney, and liver, and no binding was detected (negative data not shown).

CD33- and CD123-ADCs exhibit potent and specific cytotoxicity in AML cells, and minimal toxicity in healthy BM cells at similar doses

Both CD33- and CD123-ADCs showed comparable dose-dependent cytotoxicity in CD33+/CD123+ AML cell lines (Fig. 2A). Importantly, the cytotoxicity of the CD123-ADC in these cell lines was not affected by the presence of recombinant human IL3, suggesting that IL3 does not readily outcompete the ADC for CD123/IL3Ra binding (Supplementary Fig. S3). Next, we tested other cell lines that express different levels of CD33 and CD123. Each ADC showed a higher potency (IC50 ≤ 1 ng/mL) against the cells that have higher antigen levels (+++) and did not induce cytotoxicity against CD33- or CD123-negative cell lines (Fig. 2B). These results suggest that both ADCs are specific and potent.

Figure 2.

In vitro cytotoxicity of CD33- and CD123-ADCs in different tumor cell lines and BM cells from healthy donor and patient with AML. A,In vitro cytotoxicity of control-, CD33-, and CD123-ADCs in MV4-11 and MOLM13 cells. ADCs were titrated in growth media and cells were incubated with ADC for 96 hours. Viability was measured by CellTiter-Glo luminescent cell viability assay kit. B, A summary table of average IC50 values (ng/mL) ± SEM of control- (Ctrl), CD33-, and CD123-ADCs in AML, Non-Hodgkin lymphoma, and B-cell leukemia cell lines. For expression level, +++ indicates the highest based on mean fluorescence intensity. Control ADC was used as control. Number of independent experiments performed is shown in (n). C, Representative CFU assays of CD34+-enriched BM cells from a healthy donor (left) and an patient with AML (right; model No. 8). The normal BM samples were treated with 10 and 100 ng/mL of the ADCs and free payload CPI dimer was used as positive control at 100 pmol/L. The AML sample was treated with 1, 10, and 100 ng/mL of ADCs. After 14–16 days, the colonies were enumerated. *, P < 0.05 compared with mock at the high dose.

Figure 2.

In vitro cytotoxicity of CD33- and CD123-ADCs in different tumor cell lines and BM cells from healthy donor and patient with AML. A,In vitro cytotoxicity of control-, CD33-, and CD123-ADCs in MV4-11 and MOLM13 cells. ADCs were titrated in growth media and cells were incubated with ADC for 96 hours. Viability was measured by CellTiter-Glo luminescent cell viability assay kit. B, A summary table of average IC50 values (ng/mL) ± SEM of control- (Ctrl), CD33-, and CD123-ADCs in AML, Non-Hodgkin lymphoma, and B-cell leukemia cell lines. For expression level, +++ indicates the highest based on mean fluorescence intensity. Control ADC was used as control. Number of independent experiments performed is shown in (n). C, Representative CFU assays of CD34+-enriched BM cells from a healthy donor (left) and an patient with AML (right; model No. 8). The normal BM samples were treated with 10 and 100 ng/mL of the ADCs and free payload CPI dimer was used as positive control at 100 pmol/L. The AML sample was treated with 1, 10, and 100 ng/mL of ADCs. After 14–16 days, the colonies were enumerated. *, P < 0.05 compared with mock at the high dose.

Close modal

The cytotoxic effect of the ADCs on the progenitors of leukemic blasts was evaluated by a colony-forming unit (CFU) assay using BM samples from patients with AML treated with the ADCs. Free payload (CPI dimer) served as a positive control as it can freely enter the cells. CD33-ADC or CD123-ADC treatment already at 1 ng/mL yielded substantially fewer colonies compared with the control ADC. At higher concentrations of either ADC, no colonies were observed. This suggests that both treatments effectively killed all proliferative leukemic progenitors. The CD33-ADC at 100 ng/mL also minimally reduced the number of colonies established from CD34+-enriched BM cells from a healthy donor, whereas the CD123-ADC at the same concentration produced a similar number of colonies as the nontreated/mock control (Fig. 2C). The low ADC concentrations were selected in this assay to differentiate the effects of the two ADCs on normal BM stem and progenitor cells. Altogether, whereas both CD33- and CD123-ADCs efficiently eliminated leukemic progenitors, the CD33-ADC moderately reduced the number of colonies from healthy hematopoietic progenitors. These data were corroborated by flow cytometry analysis that displayed broader expression of CD33 in the healthy BM progenitors and stem cell populations than CD123 (Supplementary Fig. S4).

CD33- and CD123-ADCs demonstrate potent antitumor activity in disseminated AML PDX models

We evaluated the antileukemic activity of the CD33- and CD123-ADCs in MOLM13 and MV4-11 AML cell line xenograft models in immunocompromised mice (Fig. 3A). In both models, either ADC effectively controlled tumor growth, starting at a dose of 0.1 mg/kg. In contrast, the leukemias progressed in mice that received vehicle or control ADCs. Interestingly, although the internalization rates and cytotoxicity of the ADCs were comparable in vitro (Supplementary Fig. S5), and the relative receptor density of CD123 was slightly lower than CD33, the CD123-ADC was somewhat more efficacious in these models, indicating differences between in vitro and in vivo settings (Supplementary Table S3).

Figure 3.

Efficacy of CD33- and CD123-ADCs in cell line xenograft and disseminated AML PDX models. A, Tumor volume measurements of AML xenograft models established with MOLM13 and MV4-11 cell lines. ADCs were given every 4 days, four times total. () shows mg/kg; arrows ↑ indicate dosing. B, Immunocompromised mice were implanted with primary AML patient BM samples intravenously. When the PB showed engraftment of tumor cells (10%–20% hCD45+), mice received control, CD33-, and CD123-ADCs. PB and BM cells were collected to detect for any remaining tumor cells (hCD45+) by flow cytometry. Each dot represents one mouse (n = 9 per group). Showing AML PDX model No. 6 (poor risk, relapse, M2 subtype, and FLT-ITD). C, Summary of efficacy study in disseminated AML PDX models. If both ADCs have comparable efficacy and eliminate similar percentages of tumor cells at the same dose, and would be on the perpendicular line. However, if either ADC is more efficacious, it would be on either side of the line. ADCs were administered when the PB engraftment showed 10%–20% hCD45+ for all models, except for the high burden study, which was about 50% hCD45+. Tumor cells (hCD45+hCD33+hCD123+) are detected by flow cytometry. The x-axis shows the percentage of the remaining tumor cells in the CD33-ADC–treated mice, and the y-axis shows the percentage of the remaining tumor cells in the CD123-ADC–treated mice. Low (0.03 mg/kg) and high (0.1 mg/kg) doses. Suboptimal doses were used to better compare the efficacy of the ADCs. Each color represents a model.

Figure 3.

Efficacy of CD33- and CD123-ADCs in cell line xenograft and disseminated AML PDX models. A, Tumor volume measurements of AML xenograft models established with MOLM13 and MV4-11 cell lines. ADCs were given every 4 days, four times total. () shows mg/kg; arrows ↑ indicate dosing. B, Immunocompromised mice were implanted with primary AML patient BM samples intravenously. When the PB showed engraftment of tumor cells (10%–20% hCD45+), mice received control, CD33-, and CD123-ADCs. PB and BM cells were collected to detect for any remaining tumor cells (hCD45+) by flow cytometry. Each dot represents one mouse (n = 9 per group). Showing AML PDX model No. 6 (poor risk, relapse, M2 subtype, and FLT-ITD). C, Summary of efficacy study in disseminated AML PDX models. If both ADCs have comparable efficacy and eliminate similar percentages of tumor cells at the same dose, and would be on the perpendicular line. However, if either ADC is more efficacious, it would be on either side of the line. ADCs were administered when the PB engraftment showed 10%–20% hCD45+ for all models, except for the high burden study, which was about 50% hCD45+. Tumor cells (hCD45+hCD33+hCD123+) are detected by flow cytometry. The x-axis shows the percentage of the remaining tumor cells in the CD33-ADC–treated mice, and the y-axis shows the percentage of the remaining tumor cells in the CD123-ADC–treated mice. Low (0.03 mg/kg) and high (0.1 mg/kg) doses. Suboptimal doses were used to better compare the efficacy of the ADCs. Each color represents a model.

Close modal

Next, the efficacy of the ADC was evaluated in a disseminated AML PDX model established from the BM of a patient that underwent disease relapse, indicative of poor prognosis. When the peripheral blood (PB) sample of the immunocompromised mice showed engraftment of human CD45-positive cells (Supplementary Fig. S6), mice were treated intravenously with control-, CD33- or CD123-ADC. The percentage of leukemic load (hCD45+/hCD33+/hCD123+) within the total pool of white blood cells was evaluated from the PB and the BM by flow cytometry. Whereas 60% and 90% of white blood cells were leukemic cells in the control ADC–treated PB and BM, respectively, CD33- or CD123-ADC administration at 0.1 mg/kg eliminated almost all leukemic cells (Fig. 3B). To better compare the two ADCs, seven disseminated AML PDX models were generated using patient samples representing a wide range of cytogenetic profiles, molecular abnormalities, and disease stages observed in AML clinically (Supplementary Table S4). Subtherapeutic doses of the ADCs, 0.03 and 0.1 mg/kg, were used to differentiate the two ADCs, given that the higher dose would eliminate all leukemias and the efficacy of the ADCs could not be compared (Fig. 3C; Supplementary Fig. S7). Overall, CD33- and CD123-ADC treatments led to the comparable killing of leukemic cells. This result suggests that our linker-payload is efficacious across the wide spectrum of AML patient subsets.

CD33- and CD123-ADCs have a comparable pharmacokinetics profile in mice, but CD33-ADC leads to target-mediated dose disposition in cynomolgus monkeys

The pharmacokinetics profiles of CD33- and CD123-ADCs were evaluated in female nu/nu mice after a single intravenous dose and in nonhuman primate cynomolgus monkeys after three intravenous doses 3 weeks apart. Plasma samples were collected at different timepoints up to 336 hours following the last dose in mice and in cynomolgus monkey. As both ADCs are human/cynomolgus monkey, but not mouse, cross-reactive, CD33- and CD123-ADCs showed an overall comparable mouse plasma pharmacokinetics profile with a slightly lower Cmax and a longer half-life in the case of CD123-ADC. In cynomolgus monkey, however, the CD33-ADC appeared to show target-mediated dose disposition at low doses (0.1 and 0.3 mg/kg), as evidenced by its rapid disappearance following administration (Fig. 4A). This suggests that the presence of CD33 expression in normal cells in cynomolgus monkey may mediate target-dependent clearance of the CD33-ADC at low doses. At a higher dose of 1 mg/kg, CD33- and CD123-ADCs demonstrated a comparable pharmacokinetics profile, which implies saturation target-mediated clearance of CD33-ADC.

Figure 4.

Pharmacokinetics of CD33- and CD123-ADCs in mice and cynomolgus monkeys, and hematologic toxicity profiles in cynomolgus monkeys. A, Plasma pharmacokinetics of mice (left) and cynomolgus monkeys (right) administered CD33- or CD123-ADC. Concentrations in plasma are based on the total antibody assay. In mice, the curves represent the mean (±) from three mice per timepoint using a staggered sampling design. In monkeys, the curves depict individual profiles (N = 2/group; one male and one female). Blue line with filled symbols, CD33-ADC; red line with open symbols, CD123-ADC. Because of toxicity at 1 mg/kg of CD33-ADC after the first dose on day 11, no plasma sample was available beyond 264 hours. B, Hematology and microscopic changes in the BM in cynomolgus monkeys administered control-, CD33-, or CD123-ADC each three times, 3 weeks apart. AUC was calculated on the basis of Ab or ADC concentration profiles. TI was calculated as Cav of ADC at highest nonseverely toxic dose in monkeys divided by tumor stasis concentration in mice. *, calculated from HL60 and TF1 models; average of 1.5 and 4.5. **, calculated from MOLM13 and MV4-11 models; average of 11 and 27. (Bottom) Ab, antibody; ANC, absolute neutrophil count; HGB, hemoglobin; ND, not determined; NEUT, neutrophil; PLT, platelet. C, Flow cytometry analysis of CD33 and CD123 expressions in total and stem cell populations (Lin, LinCD34+CD38) in cynomolgus monkey BM cells.

Figure 4.

Pharmacokinetics of CD33- and CD123-ADCs in mice and cynomolgus monkeys, and hematologic toxicity profiles in cynomolgus monkeys. A, Plasma pharmacokinetics of mice (left) and cynomolgus monkeys (right) administered CD33- or CD123-ADC. Concentrations in plasma are based on the total antibody assay. In mice, the curves represent the mean (±) from three mice per timepoint using a staggered sampling design. In monkeys, the curves depict individual profiles (N = 2/group; one male and one female). Blue line with filled symbols, CD33-ADC; red line with open symbols, CD123-ADC. Because of toxicity at 1 mg/kg of CD33-ADC after the first dose on day 11, no plasma sample was available beyond 264 hours. B, Hematology and microscopic changes in the BM in cynomolgus monkeys administered control-, CD33-, or CD123-ADC each three times, 3 weeks apart. AUC was calculated on the basis of Ab or ADC concentration profiles. TI was calculated as Cav of ADC at highest nonseverely toxic dose in monkeys divided by tumor stasis concentration in mice. *, calculated from HL60 and TF1 models; average of 1.5 and 4.5. **, calculated from MOLM13 and MV4-11 models; average of 11 and 27. (Bottom) Ab, antibody; ANC, absolute neutrophil count; HGB, hemoglobin; ND, not determined; NEUT, neutrophil; PLT, platelet. C, Flow cytometry analysis of CD33 and CD123 expressions in total and stem cell populations (Lin, LinCD34+CD38) in cynomolgus monkey BM cells.

Close modal

CD33-ADC leads to BM toxicity in cynomolgus monkeys

The cynomolgus monkey was used as a nonclinical species to assess the safety of CD33- and CD123-ADCs, given that both antibodies cross-react with monkey CD33 and CD123, respectively. First, we confirmed that binding of CD33 and CD123 mAb to human and cynomolgus monkey targets has similar EC50 values using monocytes and plasmacytoid dendritic cells (pDCs), respectively, as CD33 is highly expressed in CD14+ monocytes and CD123 is highly expressed in pDCs (refs. 33–36; Supplementary Table S5). One male and one female cynomolgus monkey were dosed at 0.1, 0.3, and 1 mg/kg every 3 weeks, three times total, and blood samples were collected before and after each dose to evaluate hematologic changes. Already at 0.1 mg/kg, the CD33-ADC produced severe neutropenia, but was tolerated up to a dose of 0.3 mg/kg. However, the animals administered with the 1 mg/kg dose had to be euthanized on day 11 because of clinical signs attributed to BM suppression, as evidenced by hematologic toxicity (Fig. 4B; Supplementary Fig. S8B). Although hematologic effects were seen in all lineages, effects on the myeloid lineage were most prominent with severe neutropenia observed at ≥0.1 mg/kg, consistent with the higher expression level of CD33 in myeloid progenitors (20–22).

In contrast, the CD123-ADC and the control ADC were tolerated up to 1 mg/kg without any clinical signs, so we increased the dose to as high as 3 mg/kg. One of two animals dosed with CD123-ADC at 3 mg/kg was euthanized on day 31 because of clinical signs attributed to esophageal ulceration; the other animal did not display such symptoms at the same dose (Fig. 4B). We are unable to determine whether this toxicity was related to on- or off-target toxicity or whether the animal had an unrelated condition. Monkeys administered CD123-ADC exhibited only moderate decreases in cellularity in other lineages at the high dose (Supplementary Fig. S8C).

The primary target organ of toxicity for CD33- and CD123-ADCs was the BM, and we did not observe any test article–related elevation in liver enzymes or liver histology changes. The most dramatic difference between the two ADCs was the magnitude of effects on myeloid lineages, of which neutrophils are the most limiting, given their long maturation period, short lifespan, and the risks associated with febrile neutropenia. Neutrophil counts lower than 100/μL were observed with the CD33-ADC at ≥0.1 mg/kg, whereas such profound effects were observed with CD123-ADC only at the maximal dose, 3 mg/kg. The expression levels for CD33 and CD123 were assessed in BM samples from normal human donors and cynomolgus monkeys (Fig. 4C; Supplementary Fig. S4), and, as supported by literature, CD123 expression was detected only on a very small percentage of PB and BM cells in both human and cynomolgus monkey. Meanwhile CD33 was expressed on 61.8% and 72.6% of total BM cells in human and cynomolgus monkey, respectively, predominantly in the myeloid lineage. Thus, the more severe BM toxicity observed with the CD33-ADC can be explained by CD33 expression in a significantly higher fraction of cynomolgus BM cells when compared with CD123 (Fig. 4C). The animals administered with the control ADC displayed overall similar or slightly less severe hematologic alterations than animals dosed with CD123-ADCs.

The cynomolgus monkey plasma exposures for the CD33 and CD123 antibodies and total ADCs demonstrated that CD123-ADC administration induced minimal toxicities at the same dose and exposure levels as CD33-ADC. Finally, the TI was calculated from the safety data gathered in monkeys and efficacy data in mice: CD123-ADC has a significantly higher TI of 19 than CD33-ADC TI of 3. This result suggests a much more favorable safety profile of CD123-ADC compared with CD33-ADC, and therefore, CD123 may be a more suitable target for anti-AML ADCs.

Several novel agents have emerged in clinical development for the treatment of AML, such as ADCs, bispecific antibodies, chimeric antigen receptor T cells (CAR T), epigenetic modifiers, and tyrosine kinase inhibitors (3, 15). Multivalent antibodies, such as dual-affinity retargeting agents and bispecific T-cell engagers, utilize T cells to kill tumor cells, which have demonstrated efficacy in acute lymphoblastic leukemia (ALL), but several clinical trials for AML utilizing similar strategies to target CD33 or CD123 were suspended because of toxicity, including cytokine release syndrome (37). CAR T therapy targeting CD19 has also proven successful for the treatment of ALL and B-cell lymphoma, and early clinical studies targeting AML antigens as CAR T candidates, including CD33 and CD123, reported feasibility and safety (38, 39). The therapies described above rely on the presence and function of T cells in the tumor microenvironment, which in AML is immunosuppressive and thus, may require combination with immune checkpoint inhibitors. ADCs have been developed to widen the therapeutic window of nontumor-targeted cytotoxic drugs in cancer therapy. The approval of gemtuzumab ozogamicin sparked a clinical interest in this modality, and brentuximab vedotin (anti-CD30), inotuzumab ozogamicin (anti-CD22), and polatuzumab vedotin-piiq (anti-CD79b) have been approved for Hodgkin lymphoma, ALL, and diffuse large B-cell lymphoma, respectively. For solid cancers, trastuzumab emtansine targeting HER2 was approved already in 2013, and today there are ADCs in phase II and III clinical trials targeting EGFR, mesothelin, PSMA, and DLL3. To date, more than 80 ADCs have been investigated in clinical trials. Our understanding of ADCs has improved substantially over the past decade, including the identification of multiple critical factors required for their successful development.

Here, we report the development of a next-generation anti-AML ADC with a novel linker-payload, conjugation site, and chemistry that exhibits an improved TI compared with certain currently available ADC technologies. Mechanistically, both the CPI dimer and pyrrolobenzodiazepine dimer payloads differentiate from the calicheamicin payload because they both act as DNA bis-alkylators, whereas calicheamicin is a DNA double-strand breaker. The CPI dimer payload, however, preferentially bis-alkylates AT-rich regions, which along with its chemically more robust core motif, further differentiates it from the pyrrolobenzodiazepine dimer payload class that cross-links guanines (-GATC-; refs. 40). Furthermore, the CD123 conjugate, IMGN632, uses the indolino-benzodiazepine (IGN) payload that preferentially mono-alkylates DNA at guanine (-GATC-), which is the same as for the pyrrolobenzodiazepine dimers, but inherently less cytotoxic (41), including IMGN632, a CD123-targeting ADC with a novel IGN payload. Like our CD123-ADC, IMGN632 has a cleavable linker, a DAR of 2, and uses site-specific conjugation. However, unlike the CPI dimers, the IGN payloads alkylate DNA without cross-linking. In addition, a cysteine residue was engineered in the constant region for site-specific conjugation of IGN payload in IMGN632, whereas transglutaminase-mediated site-specific conjugation was used for our CD123-ADC (42). IMGN632 is currently being investigated in phase I clinical trials for the treatment of AML (NCT03386513).

We engineered the antibody to support site-specific conjugation at Y296Q, and also removed the N-linked glycosylation site at N297. Complete removal of the glycan at N297 greatly reduced binding to effector FcγRs expressed by B cells, natural killer cells, granulocytes, and monocytes, as well as to the complement protein C1q, which significantly diminished binding to FcγRI and reduced toxicity, as reported previously (43–45). In addition, we found that transglutaminase-mediated Y296Q site–specific conjugation of the linker-payload yielded a “well-behaved” ADC compared with cysteine conjugation at other sites of the antibody, which led to lower off-target toxicities and improved safety. The conjugation site has a significant impact on the stability of ADCs, as observed in previous reports (46, 47). Taken together, we demonstrated that our ADC platform provided the optimal balance of ADC stability, efficacy, and TI, and was much improved over other conventional and other site-specific ADCs, such as gemtuzumab ozogamicin and vadastuximab talirine. Thus, this linker-payload, coupled with optimized site-specific conjugation technology, has considerable potential in other ADC applications beyond CD33 and CD123.

We performed a head-to-head comparison of CD33- and CD123-ADCs, which has not been reported before. Both ADCs have specific and potent activity against several disseminated PDX models of AML established from patients with relapsed/refractory disease, and diverse risk profiles, cytogenetics, and molecular abnormalities. This suggests that these ADCs have the potential to treat a broad spectrum of the AML population, unlike other targeted drugs, such as tyrosine kinase and signaling pathway inhibitors, or epigenetic modulators that target alterations found in only subsets of patients with this heterogeneous disease.

A clear differentiation was observed in the safety studies in cynomolgus monkeys, with the BM being the primary organ of toxicity in the case of both ADCs. The differences in toxicity profiles between CD33- and CD123-ADCs were likely driven by the more restricted expression of CD123 in the BM of cynomolgus monkeys as both CD33 and CD123 mAbs bind to human and monkey cells with equivalent binding affinity and the plasma exposure profiles in monkeys were comparable between the two ADCs at equivalent doses. The in vitro CFU data also corroborated with the more severe BM toxicity observed with the CD33-ADC in monkeys as compared with the CD123-ADC. In the CFU assay, lower concentrations of ADCs were tested (up to 100 ng/mL) to differentiate the CD33- and CD123-ADC, which were still much lower when compared with the in vivo exposure in the cynomolgus monkey safety studies (e.g., Cmax for the CD123-ADC was around 90 μg/mL at 3 mg/kg dose). This may explain the overall higher BM toxicity observed in vivo compared with in vitro. In addition, the CFU assay measures the proliferation and differentiation ability of hematopoietic stem and progenitor cells, whereas in the monkey study, the overall response from the BM was evaluated via BM microscopic evaluation and hematology parameters.

Evaluation of overall expression pattern of CD33 compared with CD123 in healthy human donors also supports the more restricted expression level of CD123, which suggests that the differences in safety profiles of CD33- versus CD123-ADCs in monkeys would likely translate to humans in terms of on-target off-tumor toxicity. Consistently, others have shown that CD33- or CD123-targeted CAR T demonstrates efficient antileukemic activity in vivo, but CD123-targeted CAR T cells had significantly less activity against normal hematopoietic stem/progenitor cells than CD33-targeted CAR T cells (48). This suggests that CD123-ADC could potentially be a treatment option for older and unfit patients who cannot tolerate intensive induction chemotherapy. Our results constitute a significant conceptual advance in AML therapy that motivates the clinical development of CD123-ADCs.

J. Kahler reports personal fees from Pfizer outside the submitted work. N. Piché-Nicholas reports personal fees from Pfizer Inc. outside the submitted work, as well as has a patent for WO 2019/082020 A1 issued to Pfizer Inc. S. Thibault reports full-time employment with Pfizer Inc. F. Jiang reports personal fees from Pfizer Inc. outside the submitted work. M. Katragadda reports a patent for WO2019/082020 pending. R. Dushin reports personal fees and other from Pfizer Inc. outside the submitted work, as well as has a patent for WO2015/015448 pending. M. Charati reports personal fees from Pfizer Inc. outside the submitted work, as well as has a patent for WO2019/082020 issued to Pfizer Inc. T. Clark reports personal fees from Pfizer Inc. outside the submitted work. E. Rosfjord reports personal fees from Pfizer Inc. outside the submitted work. H.-P. Gerber reports personal fees from PFE outside the submitted work, as well as has a patent for WO 2019/082020 A1 issued to PFE Inc. F. Loganzo reports personal fees from Pfizer Inc. outside the submitted work. C.J. O'Donnell reports a patent for WO 2019/082020 issued to Pfizer Inc., WO2015/110935 issued to Pfizer, WO2012/059882 issued to Pfizer, and WO2015/015448 issued to Pfizer. P. Sapra reports personal fees from Pfizer Inc. outside the submitted work. Y. Han reports personal fees from Pfizer Inc. outside the submitted work, as well as has a patent for WO2019/082020 issued to Pfizer Inc. No disclosures were reported by the other authors.

Y.-C. Han: Conceptualization, software, formal analysis, supervision, validation, writing-original draft, project administration, writing-review and editing. J. Kahler: Formal analysis, investigation, writing-original draft, writing-review and editing. N. Piché-Nicholas: Conceptualization, software, formal analysis, validation, investigation, writing-original draft, writing-review and editing. W. Hu: Conceptualization, data curation, software, formal analysis, validation, investigation, writing-original draft, writing-review and editing. S. Thibault: Conceptualization, software, formal analysis, validation, investigation, writing-original draft, writing-review and editing. F. Jiang: Software, formal analysis, validation, investigation. M. Leal: Conceptualization, software, formal analysis, validation, investigation, writing-original draft, writing-review and editing. M. Katragadda: Conceptualization, formal analysis, supervision, validation, investigation, writing-review and editing. A. Maderna: Conceptualization, data curation, formal analysis, investigation, writing-review and editing. R. Dushin: Conceptualization, data curation, software, formal analysis, supervision, investigation, methodology, writing-review and editing. N. Prashad: Data curation, software, validation, investigation, methodology, writing-review and editing. M. Charati: Conceptualization, formal analysis, validation, investigation, methodology, writing-review and editing. T. Clark: Formal analysis, validation, investigation, methodology, writing-original draft, writing-review and editing. L.N. Tumey: Conceptualization, formal analysis, validation, investigation, methodology, writing-review and editing. X. Tan: Investigation. A. Giannakou: Conceptualization, investigation. E. Rosfjord: Conceptualization, formal analysis, investigation, writing-original draft, writing-review and editing. H.-P. Gerber: Conceptualization, supervision. L. Tchistiakova: Conceptualization, supervision. F. Loganzo: Conceptualization, supervision, visualization, writing-original draft, writing-review and editing. C.J. O'Donnell: Conceptualization, supervision, writing-original draft, writing-review and editing. P. Sapra: Conceptualization, supervision, writing-review and editing.

The authors thank Marc Damelin for helpful suggestions, Kenny Kim, LuAnna Lemon, Christine Hosselet, Lori Horton, and Judy Lucas for establishing disseminated AML PDX models, Cynthia Drupa for the cynomolgus monkey flow cytometry analysis, and Maureen Dougher, Jia Liu, Tasneem Kausar, and Sylvia Musto for early discovery work on the payloads.

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.

1.
Fernandez
HF
,
Sun
Z
,
Yao
X
,
Litzow
MR
,
Luger
SM
,
Paietta
EM
, et al
Anthracycline dose intensification in acute myeloid leukemia
.
N Engl J Med
2009
;
361
:
1249
59
.
2.
Schaich
M
,
Röllig
C
,
Soucek
S
,
Kramer
M
,
Thiede
C
,
Mohr
B
, et al
Cytarabine dose of 36 g/m(2) compared with 12 g/m(2) within first consolidation in acute myeloid leukemia: results of patients enrolled onto the prospective randomized AML96 study
.
J Clin Oncol
2011
;
29
:
2696
702
.
3.
Dohner
H
,
Weisdorf
DJ
,
Bloomfield
CD
. 
Acute myeloid leukemia
.
N Engl J Med
2015
;
373
:
1136
52
.
4.
Hanekamp
D
,
Cloos
J
,
Schuurhuis
GJ
. 
Leukemic stem cells: identification and clinical application
.
Int J Hematol
2017
;
105
:
549
57
.
5.
Kuykendall
A
,
Duployez
N
,
Boissel
N
,
Lancet
JE
,
Welch
JS
. 
Acute myeloid leukemia: the good, the bad, and the ugly
.
Ame Soc Clin Oncol Educ Book
2018
;
38
:
555
73
.
6.
Stone
A
,
Zukerman
T
,
Flaishon
L
,
Yakar
RB
,
Rowe
JM
. 
Efficacy outcomes in the treatment of older or medically unfit patients with acute myeloid leukaemia: a systematic review and meta-analysis
.
Leuk Res
2019
;
82
:
36
42
.
7.
DiNardo
CD
,
Pratz
K
,
Pullarkat
V
,
Jonas
BA
,
Arellano
M
,
Becker
PS
, et al
Venetoclax combined with decitabine or azacitidine in treatment-naive, elderly patients with acute myeloid leukemia
.
Blood
2019
;
133
:
7
17
.
8.
Pfizer Inc.
Mylotarg (gemtuzumab ozogamicin)
.
Available from:
www.accessdata.fda.gov/drugsatfda_docs/label/2017/761060lbl.pdf.
9.
Naito
K
,
Takeshita
A
,
Shigeno
K
,
Nakamura
S
,
Fujisawa
S
,
Shinjo
K
, et al
Calicheamicin-conjugated humanized anti-CD33 monoclonal antibody (gemtuzumab zogamicin, CMA-676) shows cytocidal effect on CD33-positive leukemia cell lines, but is inactive on P-glycoprotein-expressing sublines
.
Leukemia
2000
;
14
:
1436
43
.
10.
Taksin
A-L
,
Legrand
O
,
Raffoux
E
,
de Revel
T
,
Thomas
X
,
Contentin
N
, et al
High efficacy and safety profile of fractionated doses of Mylotarg as induction therapy in patients with relapsed acute myeloblastic leukemia: a prospective study of the alfa group
.
Leukemia
2007
;
21
:
66
71
.
11.
Ehninger
A
,
Kramer
M
,
Röllig
C
,
Thiede
C
,
Bornhäuser
M
,
von Bonin
M
, et al
Distribution and levels of cell surface expression of CD33 and CD123 in acute myeloid leukemia
.
Blood Cancer J
2014
;
4
:
e218
.
12.
Florian
S
,
Sonneck
K
,
Hauswirth
AW
,
Krauth
M-T
,
Schernthaner
G-H
,
Sperr
WR
, et al
Detection of molecular targets on the surface of CD34+/CD38- stem cells in various myeloid malignancies
.
Leuk Lymphoma
2006
;
47
:
207
22
.
13.
Taussig
DC
,
Pearce
DJ
,
Simpson
C
,
Rohatiner
AZ
,
Lister
TA
,
Kelly
G
, et al
Hematopoietic stem cells express multiple myeloid markers: implications for the origin and targeted therapy of acute myeloid leukemia
.
Blood
2005
;
106
:
4086
92
.
14.
Li
Fu
,
Sutherland
MK
,
Yu
C
,
Walter
RB
,
Westendorf
L
,
Valliere-Douglass
J
, et al
Characterization of SGN-CD123A, a potent CD123-directed antibody-drug conjugate for acute myeloid leukemia
.
Mol Cancer Ther
2018
;
17
:
554
64
.
15.
Robak
T
,
Wierzbowska
A
. 
Current and emerging therapies for acute myeloid leukemia
.
Clin Ther
2009
;
31
:
2349
70
.
16.
Sutherland
MK
,
Yu
C
,
Lewis
TS
,
Miyamoto
JB
,
Morris-Tilden
CA
,
Jonas
M
, et al
Anti-leukemic activity of lintuzumab (SGN-33) in preclinical models of acute myeloid leukemia
.
MAbs
2009
;
1
:
481
90
.
17.
Arndt
C
,
von Bonin
M
,
Cartellieri
M
,
Feldmann
A
,
Koristka
S
,
Michalk
I
, et al
Redirection of T cells with a first fully humanized bispecific CD33-CD3 antibody efficiently eliminates AML blasts without harming hematopoietic stem cells
.
Leukemia
2013
;
27
:
964
7
.
18.
Hauswirth
AW
,
Florian
S
,
Printz
D
,
Sotlar
K
,
Krauth
M-T
,
Fritsch
G
, et al
Expression of the target receptor CD33 in CD34+/CD38-/CD123+ AML stem cells
.
Eur J Clin Invest
2007
;
37
:
73
82
.
19.
Jordan
Ct
,
Upchurch
D
,
Szilvassy
Sj
,
Guzman
Ml
,
Howard
Ds
,
Pettigrew
Al
, et al
The interleukin-3 receptor alpha chain is a unique marker for human acute myelogenous leukemia stem cells
.
Leukemia
2000
;
14
:
1777
84
.
20.
Dinndorf
PA
,
Andrews
RG
,
Benjamin
D
,
Ridgway
D
,
Wolff
L
,
Bernstein
ID
. 
Expression of normal myeloid-associated antigens by acute leukemia cells
.
Blood
1986
;
67
:
1048
53
.
21.
Gorczyca
W
,
Sun
ZY
,
Cronin
W
,
Li
X
,
Mau
S
,
Tugulea
S
. 
Immunophenotypic pattern of myeloid populations by flow cytometry analysis
.
Methods Cell Biol
2011
;
103
:
221
66
.
22.
Griffin
JD
,
Linch
D
,
Sabbath
K
,
Larcom
P
,
Schlossman
SF
. 
A monoclonal antibody reactive with normal and leukemic human myeloid progenitor cells
.
Leuk Res
1984
;
8
:
521
34
.
23.
Ayehunie
S
,
Snell
M
,
Child
M
,
Klausner
M
. 
A plasmacytoid dendritic cell (CD123+/CD11c-) based assay system to predict contact allergenicity of chemicals
.
Toxicology
2009
;
264
:
1
9
.
24.
Dorywalska
M
,
Dushin
R
,
Moine
L
,
Farias
SE
,
Zhou
D
,
Navaratnam
T
, et al
Molecular basis of valine-citrulline-PABC linker instability in site-specific ADCs and its mitigation by linker design
.
Mol Cancer Ther
2016
;
15
:
958
70
.
25.
Maderna
A
,
Doroski
MD
,
Chen
Z
,
Risley
HL
,
Casavant
JM
,
O'Donnell
CJ
, et al
Bifunctional cytotoxic agents
.
WO2015110935, Pfizer, Inc. 2015
.
26.
Gregson
SJ
,
Howard
PW
,
Hartley
JA
,
Brooks
NA
,
Adams
LJ
,
Jenkins
TC
, et al
Design, synthesis, and evaluation of a novel pyrrolobenzodiazepine DNA-interactive agent with highly efficient cross-linking ability and potent cytotoxicity
.
J Med Chem
2001
;
44
:
737
48
.
27.
Matsui
H
,
Takeshita
A
,
Naito
K
,
Shinjo
K
,
Shigeno
K
,
Maekawa
M
, et al
Reduced effect of gemtuzumab ozogamicin (CMA-676) on P-glycoprotein and/or CD34-positive leukemia cells and its restoration by multidrug resistance modifiers
.
Leukemia
2002
;
16
:
813
9
.
28.
Walter
RB
,
Gooley
TA
,
van der Velden
VHJ
,
Loken
MR
,
van Dongen
JJM
,
Flowers
DA
, et al
CD33 expression and P-glycoprotein-mediated drug efflux inversely correlate and predict clinical outcome in patients with acute myeloid leukemia treated with gemtuzumab ozogamicin monotherapy
.
Blood
2007
;
109
:
4168
70
.
29.
Walter
RB
,
Raden
BW
,
Hong
TC
,
Flowers
DA
,
Bernstein
ID
,
Linenberger
ML
. 
Multidrug resistance protein attenuates gemtuzumab ozogamicin-induced cytotoxicity in acute myeloid leukemia cells
.
Blood
2003
;
102
:
1466
73
.
30.
Kung Sutherland
MS
,
Walter
RB
,
Jeffrey
SC
,
Burke
PJ
,
Yu
C
,
Kostner
H
, et al
SGN-CD33A: a novel CD33-targeting antibody-drug conjugate using a pyrrolobenzodiazepine dimer is active in models of drug-resistant AML
.
Blood
2013
;
122
:
1455
63
.
31.
Procopiou
G
,
O'Donnell
CJ
. 
Chapter 10 CXI dimers as antibody–drug conjugate (ADC) payloads
. In:
Cytotoxic payloads for antibody–drug conjugates
.
United Kingdom:
The Royal Society of Chemistry
; 
2019
. p.
209
40
.
32.
Arndt
C
,
Feldmann
A
,
Koristka
S
,
Cartellieri
M
,
von Bonin
M
,
Ehninger
A
, et al
Improved killing of AML blasts by dual-targeting of CD123 and CD33 via unitarg a novel antibody-based modular T cell retargeting system
.
Blood
2015
;
126
:
2565
-.
33.
Lambert
C
,
Sack
U
. 
Monocytes and macrophages in flow cytometry
.
Cytometry B Clin Cytom
2017
;
92
:
178
9
.
34.
Musumeci
A
,
Lutz
K
,
Winheim
E
,
Krug
AB
. 
What makes a pDC: recent advances in understanding plasmacytoid DC development and heterogeneity
.
Front Immunol
2019
;
10
:
1222
.
35.
Zuniga
EI
,
McGavern
DB
,
Pruneda-Paz
JL
,
Teng
C
,
Oldstone
MB
. 
Bone marrow plasmacytoid dendritic cells can differentiate into myeloid dendritic cells upon virus infection
.
Nat Immunol
2004
;
5
:
1227
34
.
36.
Marti
LC
,
Bacal
NS
,
Bento
LC
,
Rocha
FA
. 
Phenotypic markers and functional regulators of myelomonocytic cells
. In:
Marti
LC
,
Bacal
NS
,
Bento
LC
,
Rocha
FA
, eds.
Biology of Myelomonocytic Cells. InTech Open
; 
2017
:
10.5772/67324
.
37.
Zhao
Z
,
Chen
Y
,
Francisco
NM
,
Zhang
Y
,
Wu
M
. 
The application of CAR-T cell therapy in hematological malignancies: advantages and challenges
.
Acta Pharm Sin B
2018
;
8
:
539
51
.
38.
Hofmann
S
,
Schubert
M-L
,
Wang
L
,
He
B
,
Neuber
B
,
Dreger
P
, et al
Chimeric antigen receptor (CAR) T cell therapy in acute myeloid leukemia (AML)
.
J Clin Med
2019
;
8
:
200
.
39.
Wang
Q-S
,
Wang
Y
,
Lv
H-Y
,
Han
Q-W
,
Fan
H
,
Guo
Bo
, et al
Treatment of CD33-directed chimeric antigen receptor-modified T cells in one patient with relapsed and refractory acute myeloid leukemia
.
Mol Ther
2015
;
23
:
184
91
.
40.
Smellie
M
,
Bose
DS
,
Thompson
AS
,
Jenkins
TC
,
Hartley
JA
,
Thurston
DE
. 
Sequence-selective recognition of duplex DNA through covalent interstrand cross-linking: kinetic and molecular modeling studies with pyrrolobenzodiazepine dimers
.
Biochemistry
2003
;
42
:
8232
9
.
41.
Miller
ML
,
Fishkin
NE
,
Li
W
,
Whiteman
KR
,
Kovtun
Y
,
Reid
EE
, et al
A new class of antibody-drug conjugates with potent DNA alkylating activity
.
Mol Cancer Ther
2016
;
15
:
1870
8
.
42.
Adams
S
,
Wilhelm
A
,
Harvey
L
,
Harvey
L
,
Audette
CA
,
Wilhelm
A
, et al
IMGN632: A CD123-targeting antibody-drug conjugate (ADC) with a novel DNA-alkylating payload, is highly active and prolongs survival in acute myeloid leukemia (AML) xenograft models
.
Blood
2016
;
128
:
2832
-.
43.
Tao
MH
,
Morrison
SL
. 
Studies of aglycosylated chimeric mouse-human IgG. Role of carbohydrate in the structure and effector functions mediated by the human IgG constant region
.
J Immunol
1989
;
143
:
2595
601
.
44.
Vivier
D
,
Fung
K
,
Rodriguez
C
,
Adumeau
P
,
Ulaner
GA
,
Lewis
JS
, et al
The influence of glycans-specific bioconjugation on the FcgammaRI Binding and in vivo performance of (89)Zr-DFO-pertuzumab
.
Theranostics
2020
;
10
:
1746
57
.
45.
Walker
MR
,
Lund
J
,
Thompson
KM
,
Jefferis
R
. 
Aglycosylation of human IgG1 and IgG3 monoclonal antibodies can eliminate recognition by human cells expressing Fc gamma RI and/or Fc gamma RII receptors
.
Biochem J
1989
;
259
:
347
53
.
46.
Puthenveetil
S
,
He
H
,
Loganzo
F
,
Musto
S
,
Teske
J
,
Green
M
, et al
Multivalent peptidic linker enables identification of preferred sites of conjugation for a potent thialanstatin antibody drug conjugate
.
PLoS One
2017
;
12
:
e0178452
.
47.
Tumey
LN
,
Li
F
,
Rago
B
,
Han
X
,
Loganzo
F
,
Musto
S
, et al
Site selection: a case study in the identification of optimal cysteine engineered antibody drug conjugates
.
AAPS J
2017
;
19
:
1123
35
.
48.
Pizzitola
I
,
Anjos-Afonso
F
,
Rouault-Pierre
K
,
Lassailly
F
,
Tettamanti
S
,
Spinelli
O
, et al
Chimeric antigen receptors against CD33/CD123 antigens efficiently target primary acute myeloid leukemia cells in vivo
.
Leukemia
2014
;
28
:
1596
605
.