Antibody-Drug Conjugates of Calicheamicin Derivative: Gemtuzumab Ozogamicin and Inotuzumab Ozogamicin

It is surprising that few monoclonal antibodies (mAb) have been approved for the treatment of acute myeloid leukemia (AML) and non-Hodgkin lymphoma (NHL) since the first approval of rituximab in 1997. These malignancies localize in areas that are readily accessible to circulating mAbs, such as the circulating blood, lymph nodes, and bone marrow. The disposition of intact mAbs is also favorable for their treatment: Distribution into other tissues is slow, and volumes of distribution are generally low (1). Antibody-drug conjugates (ADC) are an attractive approach for the treatment of these diseases, which in most cases are inherently sensitive to cytotoxic agents. This CCR Focus section describes a significant number of ADCs currently in clinical development (2–7).

The sialic-acid–binding immunoglobulin-like lectins (siglecs) include a family of receptors that are restrictedly expressed on one or a few immune cell types, making them attractive for targeted therapy (8, 9). CD33 (also known as siglec-3) and CD22 (siglec-2) were identified as markers of myeloid leukemias and B-cell lymphomas, respectively, almost 25 years ago. Siglecs are endocytic receptors and thus ideal for an ADC strategy because they can effectively carry the complex with its cytotoxic payload into the cell (8). This is a striking difference from other antibody targets, such as CD20, for which human cells are highly heterogeneous in their internalization ability (10).

Gemtuzumab ozogamicin (GO, Mylotarg; Fig. 1) consists of a semisynthetic derivative of calicheamicin (N-acetyl-γ calicheamicin 1,2-dimethyl hydrazine dichloride), a potent enediyne DNA-binding cytotoxic antibiotic, linked to an engineered humanized monoclonal IgG4 antibody (hP67.6) directed against the CD33 antigen present on leukemic myeloblasts in most patients with AML (∼80%). IgG4 has interesting properties for a carrier. It has the longest circulating half-life of all isotypes, with limited ability for complement fixation and antibody-dependent cellular toxicity. The unconjugated antibody hP67.6 is not known to be cytotoxic per se (11). Calicheamicin was isolated from the actinomycete Micromonospora echinospora calichensis (12). The cytotoxic drug is attached to the antibody through a covalent linkage (condensation) of a bifunctional linker, 4-(4-acetylphenoxy)butanoic acid (AcBut linker), which allows stability in physiologic buffers (pH 7.4) and efficient calicheamicin release inside lysosomes (pH ∼4; ref. 11). The average loading of calicheamicin on the antibody is 2.5 mol/mol (drug-loading range of 2–3 mol of calicheamicin per mole of antibody; ref. 11). Calicheamicin binds to the minor groove in the DNA and causes double-strand DNA breaks, resulting in cell death (please refer to Fig. 1 to follow this sequence; ref. 12). The custom-made, well-controlled, hydrolysable bond with the AcBut linker showed significantly more potent and selective calicheamicin conjugates of P67.6 against HL-60 cells in vitro. This effective intracellular hydrolytic release of the calicheamicin derivative was important for its intracellular trafficking and the subsequent access to its DNA target. This indicates the importance of designing linkers that are specific for the individual target cell type (11).

Saturation and internalization were analyzed in a phase II study (13). Within 3 to 6 hours after a 9 mg/m² infusion of GO, nearly complete saturation of CD33 antigens was reached for leukemic and normal myeloid cells in blood. Although the mean maximal binding of GO to myeloid blast cells was comparable to monocytes, the maximal binding to granulocytes was significantly lower, with no binding to lymphocytes. These data are in accordance with the known expression levels of the CD33 antigen on these cells. On the other hand, this target is not (considerably) expressed on normal CD34⁺ pluripotent hematopoietic stem cells or on nonhematopoietic tissues (see “Safety” section for gemtuzumab ozogamicin below; ref. 14). Saturation levels before the second infusion were increased compared with baseline levels (before the first infusion) in that phase II study, apparently due to remaining GO from the first infusion. Upon binding to CD33, GO was rapidly internalized, as determined by the decrease in maximal membrane binding (13). In vitro studies using pulse labeling with GO showed a continuous renewal of CD33, which could increase the internalization and thereby the intracellular accumulation of the cytotoxic agent (13). Using a lentivirus-mediated gene transfer to manipulate CD33 expression in cell lines that normally lack CD33 or have very low levels of CD33, Walter and colleagues (15) showed a quantitative relationship between CD33 expression and GO-induced cytotoxicity. The CD33 cytoplasmic immunoreceptor tyrosine-based inhibitory motifs controlled the internalization process, and point mutations that disrupted these motifs not only prevented effective internalization but also significantly reduced GO-induced cytotoxicity. However, preliminary multivariate analyses showed that CD33 expression levels were not correlated with clinical response or overall survival (13, 16). This observation raises the possibility that the effects of GO in the clinic might be influenced by other features, such as cytokinetics, cytogenetics, multidrug resistance (MDR), and other prognostic factors. It is well known that AML is a heterogeneous disease with a complex biology (17). Cytotoxicity by clinically relevant GO concentrations is partially mediated by CD33 expression, and efficient GO uptake can also occur via non-CD33–mediated endocytosis. GO-mediated death of human CD33-negative leukemic lymphoblasts was shown both in vitro and in vivo in a NOD/SCID mouse model (18). Moreover, sensitivity to GO was at least partially determined by the activation status of leukemic cells, with the cells in activated phases of the cell cycle being most effective in CD33-specific internalization, renewed expression of CD33, and non-CD33–mediated GO uptake via endocytosis (18). The impossibility of quantitating CD33 expression per cell in the multicenter clinical setting may also contribute to the lack of correlation. CD33 expression classified as positive (>20% of blasts positive) versus negative (<20% of blasts positive) was not predictive, and there was no relationship with the degree of blast positivity in the AML15 trial (the largest GO randomized phase III study; ref. 17). Because cytotoxicity is dependent on the calicheamicin component, tumor cells exhibiting P-glycoprotein (Pgp)-mediated MDR may be able to escape the effect of GO. This is suggested by the correlation between clinical response and low Pgp activity, or low levels of dye efflux by leukemic blast cells (16, 19, 20). In addition, in vitro drug-induced apoptosis could be increased by Pgp antagonists (e.g., cyclosporine; ref. 20).

The pharmacokinetics of GO were characterized in patients with AML (21). The recommended phase II dose (RP2D) was determined to be 9 mg/m² infused i.v. over 2 hours, repeated on day 14 (16). Increased serum concentrations were observed after the second dose and were believed to be due to a decrease in clearance by CD33-positive blast cells, a result of the lower peripheral leukemic burden after the first dose (21, 22). The concentration profiles of calicheamicin derivative followed the same time course as hP67.6, evidence that the payload remained conjugated to the antibody and was delivered to the leukemic cells. No relationship was found between plasma concentration and response at the RP2D (21). Despite high interpatient variability, the pharmacokinetic parameters of the components (hP67.6, total and unconjugated calicheamicin derivative) were not different based on gender or age, including a comparison between pediatric patients and adults (22, 23). Weight and body surface area did not affect the pharmacokinetics of GO (23). Although most mAbs in oncology are administered on the basis of body weight or surface area, the issue of how to dose an antibody like GO, with a significant target-mediated disposition, is still controversial.

In phase I studies, the major toxicity of GO was reversible myelosuppression, especially neutropenia and thrombocytopenia, which was thought to result from the expression of CD33 on myeloid progenitor cells (24). Although patients treated with GO in phase II studies had relatively high incidences of myelosuppression, grade 3 or 4 hyperbilirubinemia (23%), and elevated hepatic transaminase levels (17%), the incidences of severe mucositis (4%), severe nausea and vomiting (11%), and infections (28%) were relatively low. No treatment-related cardiotoxicity, cerebellar toxicity, or alopecia was observed (16). Infusion-related symptoms were usually transient and occasionally required hospitalization. GO was easy to administer compared with conventional chemotherapy regimens for relapsed AML. A significant proportion of patients received GO on an outpatient basis (38% and 41% for the first and second doses, respectively), and the median duration in hospital was 24 days (16). Postmarketing reports of fatal anaphylaxis, adult respiratory distress syndrome (ARDS), and hepatotoxicity [especially veno-occlusive disease (VOD), also referred to as sinusoidal obstruction syndrome], with and without associated hematopoietic stem cell transplantation (HSCT), required labeling revisions and the initiation of a registration surveillance program. Patients on GO treatment need to be monitored for delayed hepatotoxicity, hypoxia, and infusion reactions (25). Hepatotoxicity was originally believed to reflect nonspecific hepatocellular endocytosis rather than antigen-mediated effects. However, an alternative hypothesis considers that GO targets CD33-positive cells in hepatic sinusoids (including Kupffer cells; refs. 26 and 27). It is now clear that when administered as a single agent at the recommended dose and schedule, GO has a relatively small association with VOD (1–2%), but patients who undergo HSCT within a short interval after GO administration (≤3.5 months) are at increased risk (28–30). Tumor lysis and ARDS have been reported in patients with leukocytes >30,000/μL, resulting in the recommendation that the leukocyte count be reduced to <30,000/μL prior to GO administration (25). Another potential benefit from reducing the leukemic cell burden prior to GO administration is that high CD-33 antigen loads in blood may consume GO and limit its distribution into bone marrow, which would decrease leukemic cell killing (31).

Progress in the treatment of AML has been less evident in patients 60 years of age or older than it has in younger patients. Standard chemotherapy produces dismal results in the older population, usually with poor tolerance and sometimes with a high incidence of toxic events, including death. A total of 40 patients were treated at 8 dose levels from 0.25 to 9 mg/m² in the original phase I study of GO. Morphologic evidence of leukemia within the bone marrow (>5% blast cells) was seen in all patients before treatment. After treatment with 1 to 3 doses of GO, 8 of 40 patients (20%) had <5% leukemic blast cells upon morphologic examination of bone marrow aspirate and biopsy specimens (24). Three open-label, multicenter trials were conducted to evaluate the efficacy and safety of single-agent GO. The study population for the initial report from these 3 studies comprised 142 patients with AML in first relapse with a median age of 61 years. Thirty percent of the patients obtained remission as characterized by ≤5% blasts in the marrow, recovery of neutrophils to at least 1,500/μL, and transfusion independence (16). Based on the efficacy and safety data of these pivotal studies, the U.S. Food and Drug Administration granted marketing approval under the accelerated approval regulations in 2000 (25). GO was indicated for the treatment of patients with CD33-positive AML in first relapse who were 60 years of age or older and were not considered candidates for cytotoxic chemotherapy, and later it was also approved in Europe and Japan. The initial publication (and submission; ref. 16) was followed by a final report covering the total of 277 patients enrolled in these 3 studies (29). A number of phase II studies in relapsed, refractory, and untreated AML have been published, including feasibility trials integrating GO with chemotherapy (Tables 1 and 2). In general, the results suggested clinical activity in relapsed AML with an acceptable safety profile, opportunity for first-line therapy, and remarkable activity in relapsed acute promyelocytic leukemia (APL; Tables 1 and 2). This type of AML combines a high and homogeneous expression of CD33 with low levels or absence of Pgp (32–34). GO has shown prolonged molecular remissions in APL, both as a single agent and in combination, without reports of VOD (35–40).

However, the required post-approval study (SWOG S0106), combining GO with daunorubicin and cytarabine (Ara-C) in first-line AML patients under the age of 61, failed to confirm clinical benefit. This randomized study was stopped early based on interim results showing no evidence of improved rates of complete remission (CR) or survival, with a significantly higher rate of fatal induction toxicity (5.8% in the GO-chemotherapy arm vs. 0.8% in the chemotherapy-alone arm; P = 0.002; ref. 41). The most common fatal adverse events that were at least possibly attributable to treatment were hemorrhage, infection, and ARDS. One death in the GO-chemotherapy arm was attributable to VOD (41). A second phase III study (AML15) evaluated the addition of GO to induction and/or consolidation therapy in the first-line treatment of patients up to 70 years of age with AML. This study also failed to show improvement in clinical benefit in the intent-to-treat population with the addition of GO, but there was no significant additional toxicity. Of note, a subset analysis by cytogenetics showed a highly significant interaction with induction GO (P = 0.001), and a significant survival benefit for patients with favorable cytogenetics (17, 42). The sponsor, Pfizer Inc., voluntarily withdrew the new drug application for GO (Mylotarg) in the United States in October 2010. GO was withdrawn from the U.S. and European markets; however, it continues to be commercially available in Japan, where it has received full regulatory approval. GO is still available for selected investigator-initiated research, including clinical trials in APL.

Inotuzumab ozogamicin (Fig. 1) is an ADC composed of G544, an IgG4 isotype that specifically recognizes human CD22, and the derivative of calicheamicin. The majority (>90%) of NHLs are of B-cell origin, with CD22 being expressed in ∼60% to >90% of B-lymphoid malignancies (43–45). CD22 has many of the ideal properties for an ADC target (Table 3). Unconjugated G544, having no effector function, has no antitumor activity; instead, conjugation with the cytotoxic payload confers potent dose-dependent cytotoxicity in in vitro and in vivo animal tumor models (46–48). Inotuzumab ozogamicin displayed greater single-agent therapeutic benefit than either CVP (cyclophosphamide, vincristine, and prednisone) or CHOP (cyclophosphamide, doxorubicin, vincristine, and prednisone) in xenograft models, and it induced superior antitumor activity when coadministered with standard chemotherapeutic regimens (49). Nevertheless, as with GO, inotuzumab ozogamicin was not effective in Pgp-positive sublines (Daudi/MDR and Raji/MDR cells), and MDR modifiers (PSC833 and MS209) restored the cytotoxic effect (48). In clinical samples, the cytotoxic effect was inversely related to the amount of Pgp (P = 0.003) and to intracellular rhodamine-123 accumulation (P < 0.001). Conversely, the effect positively correlated with the amount of CD22 (P = 0.010; ref. 48).

Moreover, inotuzumab ozogamicin was effective against human B-cell lymphomas resistant to rituximab in murine models, and its combination with rituximab showed additive efficacy (50). Downregulation of CD20 and upregulation of complement inhibitory factors play a central role in acquired resistance to rituximab. CD55 is a regulator of complement-dependent cytotoxicity in malignant B cells, and its expression correlates with resistance to complement-dependent cytotoxicity, which is one of the mechanisms of action of rituximab. In Daudi and Raji cells, the effect of rituximab significantly increased within 12 hours after incubation with inotuzumab ozogamicin. The levels of CD22 and CD55 were significantly reduced (P < 0.001), but the CD20 level remained constant or increased for 12 hours. Similar results were obtained in cells from patients. The antiproliferative and apoptotic effects of inotuzumab ozogamicin were greater than those of rituximab (51).

In humans, decreases in mean CD22⁺ fluorescence intensity by flow cytometry were observed immediately after the first dose of inotuzumab ozogamicin, and its intensity seemed even more attenuated after dose 2 (Fig. 2A; ref. 52). Finally, inotuzumab ozogamicin showed significant activity against CD22⁺ leukemia, and it is now being evaluated in patients with CD22⁺ acute lymphoblastic leukemia (53). The relationship between CD22 expression and inotuzumab ozogamicin cytotoxicity has not been fully evaluated yet.

The first-in-human study determined the maximum-tolerated dose (MTD), RP2D, and safety profile of inotuzumab ozogamicin in 79 patients with CD22-positive B-cell NHL (52). The MTD was declared to be 1.8 mg/m². No dose-limiting toxicity occurred at this dose given once every 4 weeks in 6 patients, and thus this regimen was selected for an expansion cohort. Pharmacokinetic samples were available for inotuzumab ozogamicin, G544, and total and free calicheamicin derivatives. The data indicated that drug disposition was nonlinear with dose or number of doses. For the ADC and total calicheamicin derivative, increases in area under the curve extrapolated over dosing interval τ (AUC_τ) with dose and increases in AUC_τ with period were observed. The increase in AUC_τ with period is thought to be secondary to the reduction in target (CD22) after the initial dose. G544 and total calicheamicin derivative exhibited similar trends of elimination, and the free calicheamicin derivative concentration remained below 1 ng/mL over time, suggesting that the linker is considerably stable in plasma (Fig. 2B; ref. 52). The pharmacokinetic profiles of Japanese patients seemed to be similar to those of non-Japanese patients (54). The pharmacokinetic profile of inotuzumab ozogamicin when combined with rituximab also seemed to be similar to that reported for its use as monotherapy (55).

In the first-in-human study, dose-limiting toxicity consisted of grade 4 thrombocytopenia and grade 4 neutropenia. Common adverse events at the MTD were thrombocytopenia, asthenia, nausea, and neutropenia (52). As a single agent, in patients with indolent relapsed/refractory NHL, the most common treatment-related adverse events were thrombocytopenia (65%), neutropenia (53%), elevated aspartate aminotransferase (48%), leukopenia (40%), nausea (40%), fatigue (35%), lymphopenia (35%), decreased appetite (33%), and elevated alkaline phosphatase (25%). Common grade 3/4 adverse events were all hematologic: thrombocytopenia (48%), neutropenia (28%), and lymphopenia (12.5%; ref. 56). Three of the ∼390 patients who have been exposed to inotuzumab ozogamicin have developed VOD (incidence < 1%; data on file, Pfizer).

Clinical activity was observed in previously treated patients in the first-in-human study, with an objective response rate of 39% [68% in patients with follicular lymphoma (FL) treated at the MTD; ref. 52]. Thus, inotuzumab ozogamicin monotherapy is being further investigated in patients with indolent NHL (56). Preliminary data for 43 patients with CD22⁺ indolent NHL (including 35 with FL) who had progressed to ≥2 therapies and exhibited refractory disease to anti-CD20 therapy were recently presented. An objective response rate of 53% (19% CR; 66% in FL patients) was observed (56). Potential clinical and molecular determinants of response to inotuzumab ozogamicin are being evaluated in phase II studies.

The combination of inotuzumab ozogamicin plus rituximab was evaluated in phase I–II studies, and investigators obtained encouraging evidence of clinical activity and a safety profile similar to that previously reported for inotuzumab ozogamicin as a single agent, with hematologic adverse events being the most frequent toxicities. During the dose-escalation phase, the MTD was 375 mg/m² of rituximab on day 1, followed by 1.8 mg/m² of inotuzumab ozogamicin on day 2, repeated once every 4 weeks (57). With this schedule, the preliminary objective response rate is 84% (median progression-free survival not reached) in patients with relapsed FL (n = 38; 1–2 prior therapies) and 80% in patients with relapsed diffuse large B-cell lymphoma with a median progression-free survival of 15.1 months (n = 40; 1–2 prior therapies; ref. 55). Response to previous therapy seems to be a major prognostic factor: The objective response rate was just 18% in patients with rituximab-refractory disease, with a median progression-free survival of only 1.7 months (Fig. 2C; ref. 55). Table 4 presents a summary of the clinical studies on inotuzumab ozogamicin conducted to date. A randomized, open-label, phase III trial is currently evaluating rituximab plus inotuzumab ozogamicin versus a defined investigator's choice of rituximab plus bendamustine or rituximab plus gemcitabine in patients with relapsed/refractory aggressive B-cell NHLs who are not candidates for high-dose chemotherapy.

AML and NHL are generally effectively treated with cytotoxic agents in the first-line setting. Despite the chemosensitivity of these diseases and advances in treatment, the curability rates remain low. Although it has been shown that a higher dose of daunorubicin can have a beneficial effect on AML, recent studies have suggested a plateau in the dose-response relationship of conventional cytotoxics, likely due to a higher incidence of toxic effects (58–61). Efficient drug delivery to malignant cells through ADCs minimizes drug exposure in normal tissues, increasing the therapeutic index of the attached cytotoxic drug. GO was the first drug to prove the ADC concept in the clinic, specifically in phase II studies that included substantial proportions of older patients with relapsed AML. In contrast, it has thus far failed to show clinical benefit in first-line treatment in combination with standard chemotherapy. ADCs may have a narrow therapeutic window when combined with a highly toxic, multiagent induction regimen in an unselected population. Inotuzumab ozogamicin has shown remarkable clinical activity in relapsed/refractory B-cell NHL, and phase III evaluation has begun. Two important questions remain unanswered: Are there better ways to dose these drugs, which are subjected to target-mediated disposition, and how should they be integrated with conventional chemotherapy to provide the most effective and best-tolerated treatment options? GO and inotuzumab ozogamicin are molecularly targeted agents that clearly suggest a new paradigm. As we steadily move toward the goal of personalized medicine, these kinds of agents will show a benefit in selected subpopulations based on the expression of specific targets (42).

Alejandro D. Ricart is employed by and has an ownership interest in Pfizer Inc.

The author thanks Mark Shapiro (global medical affairs lead for hematology programs at Pfizer) and Erik Vandendries (clinical lead for inotuzumab ozogamicin at Pfizer) for their critical reading of the article and helpful suggestions. The author thanks Leonard A. Mattano for editorial suggestions.