CD33 is expressed by acute myeloid leukemia (AML) cells in >80% of patients but not by normal hematopoietic stem cells, suggesting that elimination of CD33(+) cells may be therapeutically beneficial. A conjugate of a calicheamicin hydrazide derivative attached via hydrazone formation to the oxidized carbohydrates of the anti-CD33 murine antibody P67.6 had been chosen for use in AML prior to humanization of this antibody. However, the CDR-grafted humanized P67.6 could not be used to make the carbohydrate conjugate because of the unexpected sensitivity of this antibody to periodate oxidation. Exploration of a series of bifunctional linkers resulted in a new class of calicheamicin conjugates, termed the hybrid conjugates, that allows for the attachment of the calicheamicin to lysines but incorporates the site of hydrolytic release, a hydrazone, previously shown to be required for activity. The optimized conjugate chosen for clinical trials, gemtuzumab ozogamicin ("gem-ozo", Mylotarg, formerly designated CMA-676), was significantly more potent and selective than the carbohydrate conjugate it replaced. It was selectively cytotoxic to HL-60 leukemia cells in tissue culture with an IC(50) in the low to sub-pg cal/mL range (cal = calicheamicin equivalents). Doses of gem-ozo as low as 50 microg cal/kg given three times to mice bearing HL-60 xenografts routinely resulted in long-term, tumor-free survivors, while a nonbinding control conjugate was relatively inactive. Gem-ozo at a concentration of 2 to 10 ng cal/mL selectively inhibited leukemia colony formation by marrow cells from a significant proportion of AML patients. Gem-ozo has also shown significant activity against AML in Phase II trials and is the first antibody-targeted chemotherapeutic agent approved by the FDA.

The anti-CD33 antibody, P67.6, has been chosen to target the potently cytotoxic calicheamicin antitumor antibiotics to acute myeloid leukemia (AML) due to the presence of CD33 on >80% of patient samples and its lack of expression outside the myeloid cell lineages, especially its lack of expression on pluripotent stem cells. Previous calicheamicin conjugates relied on the attachment of a hydrazide derivative to the oxidized carbohydrates that occur naturally on antibodies. This results in a "carbohydrate conjugate" capable of releasing active drug by hydrolysis of a hydrazone bond in the lysozomes where the pH is low. Conjugates have now been made that are formed by reacting a calicheamicin derivative containing an activated ester with the lysines of antibodies. This results in an "amide conjugate" that is stable to hydrolysis, leaving the disulfide that is present in all calicheamicin conjugates as the likely site of drug release from the conjugate. In this article, these two classes of calicheamicin-antibody conjugates are compared for potential use in AML with the anti-CD33 antibody P67.6. Conjugates of P67.6 are shown to require the site of hydrolytic release afforded by the carbohydrate conjugates in order to retain good potency and selectivity in vitro, in vivo, and ex vivo. The P67.6 carbohydrate conjugate of calicheamicin is selectively cytotoxic at <0.006 ng/mL of calicheamicin equivalents (cal equiv) toward HL-60 promyelocytic leukemia cells in tissue culture. Long-term, tumor-free survivors are seen in xenograft models when mice bearing HL-60 subcutaneous tumors are treated with the P67.6 carbohydrate conjugate at a dose of 300 microg/kg cal equiv given three times. This conjugate also selectively inhibits the formation of colonies from AML marrow samples at 2 ng/mL cal equiv. The P67.6 carbohydrate conjugate of calicheamicin therefore appears to have promise as an antibody-targeted chemotherapeutic agent for CD33-positive diseases such as AML.

Heterogeneity in the underlying mechanisms of disease processes and inter-patient variability in drug responses are major challenges in drug development. To address these challenges, biomarker strategies based on a range of platforms, such as microarray gene-expression technologies, are increasingly being applied to elucidate these sources of variability and thereby potentially increase drug development success rates. With the aim of enhancing understanding of the regulatory significance of such biomarker data by regulators and sponsors, the US Food and Drug Administration initiated a programme in 2004 to allow sponsors to submit exploratory genomic data voluntarily, without immediate regulatory impact. In this article, a selection of case studies from the first 5 years of this programme - which is now known as the voluntary exploratory data submission programme, and also involves collaboration with the European Medicines Agency - are discussed, and general lessons are highlighted.

The chemically averaged molecular weights of a variety of native and conjugated monoclonal antibodies, approximately 150,000, were measured by matrix-assisted UV-laser desorption/ionization mass spectrometry. The average mass of the carbohydrate present in a monoclonal antibody was estimated from the difference between the measured mass of the monoclonal antibody and the mass of the protein present in the monoclonal antibody computed from the amino acid translation of the DNA sequence. The loading of chelators and anticancer drugs conjugated to a monoclonal antibody was quantitated from the difference in the measured masses for the conjugated and untreated monoclonal antibody relative to the expected mass change upon conjugation of 1 mol of chelator or drug. The loading results obtained by mass spectrometry were consistent in most cases with measurements obtained by radioactivity trace assay or UV spectrometry. Similar matrix-assisted UV-laser desorption/ionization mass spectrometric studies were also made after reducing untreated and conjugated monoclonal antibodies with dithiothreitol to determine the distribution of carbohydrate and chelator between the light and heavy chains of the molecules. Matrix-assisted UV-laser desorption/ionization mass spectra were used to compute loading values for covalently bound drugs and proteins, while the loading values obtained by use of gel-filtration HPLC and UV spectrometry cannot distinguish between covalently and noncovalently bound drugs and proteins.