Disulfide-Linked Antibody−Maytansinoid Conjugates: Optimization of

22 Mar 2011 - ... is in advanced clinical evaluation for metastatic breast cancer, and ...... of Misfolded F508del-Cystic Fibrosis Transmembrane Condu...
0 downloads 0 Views 2MB Size
ARTICLE pubs.acs.org/bc

Disulfide-Linked Antibody-Maytansinoid Conjugates: Optimization of In Vivo Activity by Varying the Steric Hindrance at Carbon Atoms Adjacent to the Disulfide Linkage Brenda A. Kellogg, Lisa Garrett,† Yelena Kovtun, Katharine C. Lai, Barbara Leece, Michael Miller, Gillian Payne, Rita Steeves,‡ Kathleen R. Whiteman, Wayne Widdison, Hongsheng Xie,ζ Rajeeva Singh,* Ravi V. J. Chari, John M. Lambert, and Robert J. Lutz ImmunoGen, Inc., 830 Winter Street, Waltham, Massachusetts 02451, United States

bS Supporting Information ABSTRACT: In this report, we describe the synthesis of a panel of disulfide-linked huC242 (anti-CanAg) antibody maytansinoid conjugates (AMCs), which have varying levels of steric hindrance around the disulfide bond, in order to investigate the relationship between stability to reduction of the disulfide linker and antitumor activity of the conjugate in vivo. The conjugates were first tested for stability to reduction by dithiothreitol in vitro and for plasma stability in CD1 mice. It was found that the conjugates having the more sterically hindered disulfide linkages were more stable to reductive cleavage of the maytansinoid in both settings. When the panel of conjugates was tested for in vivo efficacy in two human colon cancer xenograft models in SCID mice, it was found that the conjugate with intermediate disulfide bond stability having two methyl groups on the maytansinoid side of the disulfide bond and no methyl groups on the linker side of the disulfide bond (huC242-SPDB-DM4) displayed the best efficacy. The ranking of in vivo efficacies of the conjugates was not predicted by their in vitro potencies, since all conjugates were highly active in vitro, including a huC242-SMCC-DM1 conjugate with a noncleavable linkage which showed only marginal activity in vivo. These data suggest that factors in addition to intrinsic conjugate potency and conjugate half-life in plasma influence the magnitude of antitumor activity observed for an AMC in vivo. We provide evidence that bystander killing of neighboring nontargeted tumor cells by diffusible cytotoxic metabolites produced from target cell processing of disulfide-linked antibody-maytansinoid conjugates may be one additional factor contributing to the activity of these conjugates in vivo.

’ INTRODUCTION Several antibody-maytansinoid conjugates (AMCs) are currently undergoing clinical evaluation for treatment of different cancers.1,2 Each AMC consists of an antibody linked at lysine residues via a cleavable disulfide linker or a noncleavable thioether linker to ∼3-4 cytotoxic maytansinoid molecules. Upon binding of the antibody component of the AMC to antigen on the surface of cancer cells, the AMC is internalized and cleaved proteolytically, releasing maytansinoids that bind tubulin and interfere with microtubule dynamics, leading to cell cycle arrest and ultimately to cell death.3 The linker type—noncleavable or disulfide—for each AMC in the clinic is based upon selection of the optimal linker from preclinical evaluation of different linkers in xenograft models. The HER2-targeting AMC, trastuzumab-DM1, is in advanced clinical evaluation for metastatic breast cancer, and consists of a noncleavable thioether linker.4,5 However, most AMCs that have entered the clinic have cleavable, disulfide linkers, for example, cantuzumab mertansine, IMGN901 (lorvotuzumab mertansine), IMGN388, and SAR3419.6-9 In an effort to understand the relationship of the lability of the disulfide linker to the activity of a conjugate, a study was initiated r 2011 American Chemical Society

with a panel of huC242 conjugates containing linkers designed with different degrees of steric hindrance around the disulfide bond in order to vary the sensitivity of the disulfide linkage to cleavage via thiol-disulfide exchange reactions. HuC242 is a humanized IgG1 antibody which recognizes CanAg, a tumorspecific antigen found on the surface of human colorectal carcinomas as well as a number of other human tumors.10,11 The goal of the study was to select the disulfide linker with optimal in vivo efficacy and tolerability so as to maximize the therapeutic window for huC242 antibody-maytansinoid conjugates. The effect of altering disulfide bond reactivity toward thiol-disulfide exchange on the efficacy and tolerability of immunoconjugates is not easily predicted since the rate of cleavage of the disulfide bond is expected to influence both conjugate half-life in the plasma and rate of intracellular release of the cytotoxic agent.12-16 Here, we evaluate a panel of disulfide-linked huC242 AMCs, having varying degrees of steric hindrance around the disulfide Received: November 2, 2010 Revised: January 25, 2011 Published: March 22, 2011 717

dx.doi.org/10.1021/bc100480a | Bioconjugate Chem. 2011, 22, 717–727

Bioconjugate Chemistry

ARTICLE

Table 1. Structures, Nomenclature, and in Vitro Cytotoxic Potency on COLO205 Cells of a Panel of AMCs Made with huC242 Antibody

a in vitro cytotoxic potency of conjugate on COLO205 cells measured using a clonogenic assay with continuous exposure. EC50 is expressed as concentration of antibody; the AMCs contained, on average, 3.3 ( 1.0 conjugated DMx molecules per antibody molecule.

bond, for sensitivity to cleavage in vitro and in vivo and for efficacy in mouse xenograft models of human colon cancer expressing the CanAg antigen. We show that a huC242 conjugate with intermediate disulfide bond reactivity to thiol-disulfide exchange, having two methyl groups on the drug side of the disulfide bond and no methyl groups on the linker side, demonstrates the best activity in vivo in both homogeneous and heterogeneous CanAg expressing xenograft models. Comparisons of in vitro potency and bystander activity provide insights into the factors contributing to the superior activity observed for this conjugate.

adjacent to the sulfur of the linker, N-hydroxy succinimide (NHS) ester-bearing linker-maytansinoid reagents (sulfo-NHS-SMPP-DM1 and sulfo-NHS-SMPP-DM4 respectively, Figure 1B) were chemically synthesized and used for direct coupling to the antibody. The synthesis of these reagents are described in the Supporting Information. The conjugation reactions for the 2:0 and 2:2 conjugates were performed by adding 6-10-fold molar excess of purified NHS-ester-linker-maytansinoid reagent to antibody at a final concentration of 2.5 mg/mL in pH 7.5 phosphate buffer, 20% DMA (v/v). The reaction was left for 3 h at room temperature and then purified from excess reagent using G25 chromatography. The maytansinoid load on the final conjugate was characterized as previously described.17 Kinetic Studies to Measure Rates of Reduction of Disulfide Linkage in huC242-Maytansinoid Conjugates by Dithiothreitol (DTT). Reactions were initiated by addition of DTT to huC242-maytansinoid conjugates (1 mg/mL) equilibrated at 37 C in 95% aqueous phosphate buffer, pH 6.5, (67 mM potassium phosphate, 67 mM NaCl, 2 mM EDTA) and 5% DMA (v/v). Maytansine (20 μM) was included as an internal standard. All reactions were carried out under pseudo first-order conditions with thiol (DTT) concentrations in excess to total accessible conjugate disulfide bonds. At several time points, aliquots of the reaction mixture were removed and quenched by addition to equal volumes of ice-cold ethanol (this does not cause protein precipitation) to promote dissociation of free drug from conjugate and then stored on dry ice prior to HPLC

’ EXPERIMENTAL PROCEDURES General. Chemicals were from Sigma Aldrich. Human tumor cell lines were from ATCC. COLO205 (human colon cancer) and Namalwa (Burkitts lymphoma) cells were grown in RPMI containing 10% heat-inactivated FBS and 2 mM glutamine. HT29 (human colon cancer) cells were grown in DMEM containing 10% heat-inactivated FBS. The humanized anti-CanAg monoclonal antibody, huC242, and isotype matched nonbinding huIgG1 antibody (hu control mAb) were isolated at ImmunoGen. Maytansinoids and maytansinoid conjugates of the humanized C242 antibody (collectively referred to as huC242-DMx), with the exception of the 2:0 and 2:2 conjugates (see Table 1 for description of notation), were synthesized at ImmunoGen according to published procedures.17,18 For the highly hindered 2:0 and 2:2 disulfidelinked conjugates with two methyl groups on the carbon atom 718

dx.doi.org/10.1021/bc100480a |Bioconjugate Chem. 2011, 22, 717–727

Bioconjugate Chemistry

ARTICLE

Figure 1. (A) Structures of linkers and maytansinoids utilized in antibody-maytansinoid conjugates (AMCs). In an AMC, a lysine residue on the antibody is connected via an amide bond to the linker moiety, and the maytansinoid thiol (DMx) is attached to the linker through a disulfide or thioether bond. Typically, AMCs contain about 3-4 maytansinoid molecules (average) per antibody molecule. (B) Structures of N-hydroxy succinimide (NHS) ester-bearing linker-maytansinoid reagents used for one step synthesis of the 2:0 and 2:2 conjugates.

nonlinear regression fitting of the ratio of peak area of released maytansinoid (PADMx) to peak area of maytansine internal standard (PAInternal std) vs time to a single exponential eq 1. Secondorder rate constants (k) for the reaction of DTT thiol with the disulfide were calculated by dividing kobs by the concentration of DTT thiol as determined by Ellman’s assay.19

Table 2. Rates of Reduction of Disulfide Bonds in AMCs by Dithiothreitol (DTT)a disulfide

rate of reduction by thiol k

relative stability to thiol-

conjugate

(M-1 min-1)b

disulfide exchangec

00 :0

14

1

1:0

2.0

7

0:2 2:0

0.98 0.83

14 17

0

0 :2

0.62

22

1:1

0.082

170

1:2 2:2

0.014 22000

a Rates of reduction of disulfide-linked conjugates by dithiothreitol (DTT) were measured under pseudo first-order conditions (DTT in excess to total accessible disulfide bonds in the conjugate) at 37 C in 95% aqueous phosphate buffer, pH 6.5, 2 mM EDTA/5% DMA (v/v). b k is the second-order rate constant obtained by dividing kobs (pseudo first-order rate constant for release of maytansinoid from conjugate) by the concentration of DTT thiol. c Relative stability toward thiol-disulfide exchange normalized to the 0':0 conjugate (huC242-SPDB-DM1).

analysis. The quenched aliquots were brought to 4 C and centrifuged at 14 000 rpm and then injected onto a Hisep HPLC column (Supelco) to separate the released maytansinoid thiol and the maytansine internal standard from the antibody species. A gradient of 25% CH3CN (v/v) in 60 mM NH4OAc aqueous buffer, pH 7.0, to 100% CH3CN over 30 min was used for separation. Pseudo first-order rate constants were calculated by 719

dx.doi.org/10.1021/bc100480a |Bioconjugate Chem. 2011, 22, 717–727

Bioconjugate Chemistry To calculate the surviving fraction, the plating efficiency of the conjugate-treated plate was divided by the plating efficiency of the untreated (control) plates. Cytotoxicities of unconjugated maytansinoids were measured using a continuous exposure cell viability assay. COLO205 and HT29 cells were initially plated at 1000 and 2000 cells/well, respectively, in flat-bottom plates. After 5 day incubation with the maytansinoids, WST-8 reagent (Dojindo Technologies) was added and absorbance at 450 nm was measured after 2 h at 37 C. Bystander Killing Assay. To test the bystander potency of the conjugates, antigen-negative Namalwa cells and antigen-positive COLO205 cells were mixed together at different ratios (5000 Namalwa cells/well plus 0, 125, 500, 1250, or 5000 COLO205 cells/well), and plated in 96-well round-bottomed plates. Then, 1 nM of huC242-DMx was added to the cell mixtures. This concentration of conjugate had previously been shown to kill all of the COLO205 cells but none of the Namalwa cells when each cell line was plated alone. Plates were incubated for 4 days at 37 C, and viability of Namalwa cells in each well was determined using WST8 reagent. Pharmacokinetic Studies on huC242-DMx Conjugates in CD-1 Mice. CD-1 mice (9 mice/test substance) received 10 mg/kg of conjugate (antibody dose) as a single intravenous injection into a lateral tail vein. At 2 min, 30 min, and at 2, 4, and 8 h, and at 1, 2, 3, 5, 7, 10, 14, 21, and 28 days after administration of each test substance, approximately 70 μL of blood was collected from three mice per time point and centrifuged to separate the plasma. No mice were bled more than twice within 24 h. Plasma samples (∼30 μL) were stored at -80 C and assayed by ELISA for conjugate and antibody concentrations. ELISA of Conjugate and Antibody. The huC242-DMx conjugate in mouse plasma was measured using a sandwich ELISA method, captured with murine monoclonal antimaytansinoid antibody (ImmunoGen, Inc.), and detected with horseradish peroxidase (HRP)-conjugated donkey antihuman IgG (H and L) antibody (Jackson Immunoresearch, cat no. 709-035-149). This format of the conjugate ELISA measures the concentration in plasma of conjugate with one or more conjugated maytansinoids per antibody. The antibody component of conjugate in plasma samples was measured by capturing with goat antihuman IgG antibody and detection with HRP-conjugated donkey antihuman IgG antibody. Pharmacokinetic Analyses. Pharmacokinetic analyses were performed using the standard algorithm of the noncompartmental pharmacokinetic analysis program WinNonlin (Pharsight). First-order rate constants for the terminal elimination phase (t1/2,β) were evaluated using the concentration data from 8 to 672 h postadministration. Where concentrations in plasma are expressed as percentage of injected dose, the plasma volume was assumed to be 4% of the body weight of mice.21 In Vivo Efficacy Studies with Human Tumor Xenograft Models in SCID Mice. For the HT29 and COLO205 xenograft model of human colon adenocarcinoma, 2  106 and 3  106 cells, respectively, were injected subcutaneously into an area under the right shoulder. Dosing with 75 μg conjugated maytansinoid/kg/inj, qdx5, was initiated when the tumors had reached a size of approximately 100-200 mm3. Volumes of administration were calculated based on antibody concentration and maytansinoid to Ab ratio (D/A) for each conjugate. Tumor size was measured twice weekly in three dimensions using a caliper. The tumor volume was expressed in mm3 using the formula V = Length  Width  Height  1/2.

ARTICLE

Tolerability Studies in CD-1 Mice. CD-1 mice were randomly grouped according to their body weight with 12 mice per test substance. One to two intravenous injections of each conjugate were given through a tail vein. Animals were monitored daily for signs of toxicity including body weight loss, dehydration, hunched posture, ruffled fur, and loss of appetite or lethargy. If body weight loss exceeded 15% within the first 24 h after treatment or 20% at any other point in the study or if the animals were unable to reach food or water, they were sacrificed.

’ RESULTS Synthesis and Nomenclature Used for Antibody-Maytansinoid Conjugates. In order to evaluate the impact of

linker stability on the in vivo activity of disulfide linked AMCs, conjugates were synthesized using (1) a variety of linkers with different steric hindrance, 0, 1, or 2 methyl groups on the carbon atom next to the sulfur, and (2) three different maytansinoid thiols, DM1, DM3, and DM4 (collectively referred to as DMx), which differ in having 0, 1, or 2 methyl group substituents on the carbon next to the sulfur (Figure 1A). (Note: In these experiments, monosubstituted linker (SPP) and monosubstituted maytansinoid (DM3) were used, each being a mixture of stereoisomers at the monosubstituted carbon adjacent to the sulfur atom. As a result, conjugates made from these reagents are a mixture of diastereomers. The rates of reduction by a thiol such as cysteine or glutathione in vivo must then be considered to be a weighted average of the rates of reduction of each diastereomer in the mixture. It is possible that a diastereomerically pure conjugate might be reduced at a rate slightly different from that of the conjugate mixture used in these studies; however, we believe that the activities measured here, in vitro and in vivo, are likely reasonable estimates for the diastereomerically pure materials since it has been shown previously that for DM3 thiol the stereochemistry at the carbon adjacent to the thiol is not important for the activity of the free drug18.) In the AMCs described here, the antibody is connected via lysine residues to one end of the bifunctional linker through an amide bond while the maytansinoid agent is connected to the other end of the linker through either a disulfide or a thioether bond. The steric hindrance around the disulfide bond of an antibody-maytansinoid conjugate will be specified using the notation x:y, where x = the number of methyl groups on the carbon adjacent to the sulfur on the linker side of the disulfide bond, and y = the number of methyl groups on the carbon adjacent to the sulfur on the maytansinoid side of the disulfide bond (see Table 1). In addition, two different carbon chain lengths were evaluated for unhindered disulfide linkers (derived from SPDP and SPDB) and are referred to as 0 or 00 , respectively, in the disulfide notation (Table 1). A nonreducible thioether-linked conjugate, huC242SMCC-DM1, was also included in the analyses. All conjugates were characterized in terms of DMx/Ab ratio (3.3 ( 1.0), percent monomer (>95%), percent free drug associated with conjugate ( 1:1 > 1:2 = SMCC-DM1. Therefore, for conjugates with one methyl group on the linker side of the disulfide bond, increasing steric hindrance on the maytansinoid side leads to decreasing efficacy. These results suggest that, in this series of conjugates, the cleavability of the disulfide linkage is a more important determinant of antitumor activity than the increase in exposure (AUC) with increasing steric hindrance around the disulfide bond. The relative lack of activity for the noncleavable SMCC linked conjugate in these models supports this notion. In a second set of efficacy studies (Figure 4), we compared the activity of conjugates having the same degree of steric hindrance on the maytansinoid side while having different steric hindrance on the linker side (1:0 and 2:0). In this case, there was little difference in the efficacy of these conjugates in either model. This again may suggest that there is an important balance between improved exposure vs ease of cleavability for in vivo activity. A nonbinding humanized IgG1-SPP-DM1 conjugate had no activity in these models demonstrating that the activity observed for the huC242 conjugates is dependent on specific binding of conjugate to CanAg on the cell surface. In the same study, we also compared conjugates having the same degree of steric hindrance on the linker side vs the maytansinoid side of the disulfide bond

Figure 3. Comparison of antitumor activities of huC242 AMCs on human colon cancer xenograft models in SCID mice. Mice (5 per group) were dosed with 75 μg conjugated maytansinoid/kg/injection, daily for 5 days. AMCs with increasing steric hindrance on the maytansinoid side of the disulfide bond (1:0, 1:1, 1:2) were compared with the noncleavable SMCC-DM1 conjugate in COLO205 model (A) and HT29 model (B). The insets show CanAg expression levels determined by FACS on the two cell lines after staining with saturating huC242 antibody followed by FITC labeled goat antihuman IgG secondary reagent (M2 population) or secondary reagent only (M1 population). In cases where mice remained tumor free (TF) at the end of study, it is indicated in the figure.

hindered 1:0 or 0:0 conjugates (t1/2,βconj of 47 and 15 h, respectively). The 1:2 conjugate with hindrance on both sides of the disulfide linkage is even more stable with t1/2,βconj of 218 h. Conjugate clearance (CLconj) represents the total clearance of conjugated maytansinoid from plasma due to a combination of antibody clearance (CLAb) and loss of maytansinoid from antibody due to a bond cleavage event. The rate constant kcl for maytansinoid cleavage from antibody was calculated as described above for each conjugate and is given in Table 3. The relative stabilities of the conjugates to maytansinoid cleavage, normalized 722

dx.doi.org/10.1021/bc100480a |Bioconjugate Chem. 2011, 22, 717–727

Bioconjugate Chemistry

ARTICLE

Figure 4. Panels (A) and (B) show results from a second set of efficacy studies in SCID mice comparing conjugates having the same level of steric hindrance, either on the linker side or on the maytansinoid side, of the disulfide bond (1:0 vs 0:1, 2:0 vs 0:2) on COLO205 and HT29, respectively. All other conditions are identical to those described in Figure 3.

(0:1 vs 1:0; 0:2 vs 2:0) to determine if the location of the steric hindrance impacted the efficacy of the conjugate. In vivo pharmacokinetic studies had demonstrated that the location of the hindrance did not appreciably change conjugate half-life or exposure (Table 3). In both the HT29 and the COLO205 efficacy studies, however, it was observed that the conjugate with the steric hindrance on the maytansinoid side of the disulfide bond was significantly more active than the corresponding conjugate with the same level of steric hindrance on the linker side. The order of activity of conjugates in both models was 0:2 > 0:1 > 2:0 ≈ 1:0. These results suggest that additional factors beyond PK exposure and cleavability of the disulfide linkage may be involved in determining in vivo efficacy. In a further series of efficacy studies, we compared the activity of the 0:2 conjugate (SPDP-DM4) to the 00 :2 conjugate (SPDBDM4) that has a longer linker carbon chain. The two conjugates had similar efficacy in the models tested (data not shown). In Vitro Cytotoxicity of Unconjugated Maytansinoid Derivatives and AMCs. In an effort to understand what additional factors may be involved in determining the in vivo efficacy of an AMC, we evaluated the intrinsic cytotoxic potency of the unlinked maytansinoids and various conjugates in in vitro cytotoxicity assays. The in vitro cytotoxicity of unconjugated DM1 and DM4 maytansinoids were tested in COLO205 and HT29 cells as the stable S-methyl thioether species, since the potency of DM1 and DM4 measured in cell culture media can be variable due to formation of mixed disulfides by thiol/disulfide exchange with the cystine present in the media. In COLO205 cells, the IC50 values of the S-Me derivatives of DM1 and DM4 were 52 and 26 pM, while in HT29 cells, the IC50 values were 36 and 22 pM, respectively. These data suggest that introducing methyl groups

on carbon adjacent to the thiol group (as in DM3 and DM4) may slightly increase the potency of these unconjugated maytansinoids relative to DM1, likely due to increased hydrophobicity resulting in greater membrane permeability and/or tighter binding to microtubules. The cytotoxicity of huC242-DMx conjugates were tested in COLO205 cells in vitro using a clonogenic assay. All disulfidelinked conjugates were highly potent, with EC50 values ranging from 3.5 to 15 pM (antibody concentration; Table 1). There is no apparent correlation between the disulfide bond stability and the observed in vitro potency of the conjugates. Furthermore, the noncleavable thioether linked SMCC-DM1 conjugate showed similar in vitro cytotoxic potency to the disulfide-linked conjugates, suggesting that disulfide bond cleavage is not a critical factor for the in vitro cytotoxic potency of these conjugates, consistent with the demonstration of efficient lysosomal cleavage of AMCs in COLO205 cells.3 In Vitro Bystander Activity of Conjugates. Previously, we have shown that, under conditions where cells are grown at very high densities, as in a solid tumor grown in vivo, disulfide-linked conjugates can kill neighboring nontargeted bystander cells after being internalized and processed by the target cell to generate diffusible cytotoxic metabolites.23 This prompted us to investigate whether bystander killing potency of an AMC may vary with the stability of the disulfide linkage between antibody and maytansinoid and therefore be a factor that may contribute to in vivo efficacy. In Figure 5, we show the results of an experiment designed to measure the in vitro bystander killing activity of four maytansinoid conjugates: 1:0, 0:1, 0:2, and SMCC-DM1. In this experiment, COLO205 (antigen-positive) and Namalwa (antigen-negative) cells were mixed together in different ratios, and the huC242-DMx conjugates were added at a concentration 723

dx.doi.org/10.1021/bc100480a |Bioconjugate Chem. 2011, 22, 717–727

Bioconjugate Chemistry

ARTICLE

weight loss were similar for the two conjugates and the dose resulting in severe toxicity in 10% of the animals (STD10) was found to be similar for the two conjugates (approximately 1.2 mg DMx/kg). These results suggest that the balance of factors determining efficacy and tolerability are different.

Figure 5. In vitro bystander killing of antigen-negative Namalwa cells by CanAg antigen-positive COLO 205 cells treated with huC242 AMCs. Namalwa and COLO205 cells were mixed in the ratios indicated and treated with one of C242-SPP-DM1 (1:0), C242-SPDP-DM3 (0:1), C242-SPDP-DM4 (0:2), or C242-SMCC-DM1 at 1 nM conjugate concentration. All conjugates had ∼4 DMx/Ab. The absorbance at 450 nm is a measure of the viability of the Namalwa cells after a 4 day incubation with conjugate.

of 1 nM, sufficient to kill all the COLO205 cells but none of the Namalwa cells. In the absence of COLO205 cells, the Namalwa cells were not killed by any of the conjugates, as expected. In contrast, when COLO205 cells were present in the cell mixture, all of the disulfide-linked conjugates were able to kill surrounding antigen-negative Namalwa cells to some extent. However, the noncleavable SMCC-DM1 conjugate showed no bystander killing even at the highest ratio of COLO205:Namalwa cells used (1:1). These results are consistent with previously reported results showing that the cytotoxic metabolites produced upon target cell internalization of disulfide-linked conjugates include the membrane-permeable DMx thiol and S-methyl DMx species, while the sole metabolite produced from internalization of the SMCC-DM1 conjugate is the poorly membrane-permeable lysine-SMCC-DM1 species.3,24 Overall, the trends in bystander killing potency in vitro agree well with the trends in antitumor activity of these conjugates in vivo, with 0:2 ≈ 0:1 > 1:0 > SMCC-DM1, suggesting that bystander killing may play a role in antitumor activity in vivo. Tolerability of huC242-SPP-DM1 (1:0) and huC242-SPDBDM4 (00 :2) Conjugates in CD-1 Mice. Interestingly, while the in vivo efficacy of the huC242-SPDB-DM4 (00 :2) conjugate is significantly greater than the huC242-SPP-DM1 (1:0) conjugate in the models tested, the acute toxicity of the two conjugates in mice is similar. CD-1 mice (12 mice/group) were dosed intravenously with either huC242-SPP-DM1 (1:0) or huC242-SPDBDM4 (00 :2) conjugate. Measurements of toxicity based on body

’ DISCUSSION Cantuzumab mertansine (huC242-SPP-DM1), a disulfidelinked maytansinoid conjugate with sterically hindered (1:0) disulfide linkage, has been studied in several phase I human clinical trials, predominantly in subjects with colon cancer.25-27 Because of its specificity and potency toward CanAg antigenexpressing tumors, the huC242-maytansinoid conjugate is able to selectively target malignant tumor cells for destruction while sparing healthy normal tissue. However, in humans25 and in preclinical models,6 the half-life of the conjugate in plasma was found to be significantly shorter than the half-life of the antibody component, likely due to reductive cleavage of DM1 from antibody as the conjugate circulates in the plasma compartment (t1/2,β conjugate = 41 h vs t1/2,β antibody = 230 h in humans). This observation prompted us to investigate whether huC242maytansinoid conjugate activity might be improved by increasing the stability of the disulfide linker connecting the maytansinoid to the antibody toward thiol-disulfide exchange reactions, thereby increasing conjugate half-life and exposure of tumor to conjugated maytansinoid. Toward this end, a panel of disulfide-linked huC242-DMx conjugates was generated with different degrees of steric hindrance around the disulfide bond, by introducing methyl groups on carbon atoms adjacent to the disulfide linkage (Table 1). The SMCC-DM1 conjugate with DM1 linked via a thioether bond was also included in this investigation to represent a conjugate with a nonreducible linker. Evaluation of this set of conjugates for their sensitivity to reductive cleavage in vitro demonstrated that as the degree of steric hindrance around the disulfide bond increased the rate of cleavage of maytansinoid from antibody decreased. In vivo, these conjugates showed an identical trend, with increasing steric hindrance around the disulfide bond leading to increased conjugate half-life in plasma, suggesting that maytansinoid is being lost from disulfide linked conjugates in plasma by a reductive cleavage event. The location of the steric hindrance (maytansinoid vs linker side) in the disulfide bond did not significantly influence the rate of cleavage of maytansinoid from conjugate, which is consistent with a chemical reduction occurring in vivo via a small molecule reductant like cysteine.28 The rate of loss of maytansinoid from the AMCs in vivo, inferred from the pharmacokinetic measurements, is consistent with that predicted from the second-order rate constants for reduction measured in vitro (the second-order rate constant measured in vitro at pH 6.5 has to be adjusted upward by a factor of ∼8 to correct for the higher pH found in plasma (7.4), which will lead to an 8-fold increase in the concentration of reactive thiolate nucleophile and a corresponding increase in rate of reduction) and the known concentrations of free thiol in plasma (∼5 μM) in the form of cysteine.28 While the greater stability of the more sterically hindered disulfide-linked conjugates might be predicted to lead to increased tumor exposure and potentially greater efficacy for the more hindered conjugates, evaluation of the in vivo efficacy resulted in an unexpected ranking, with conjugates of intermediate disulfide bond stability having the greatest antitumor activity. 724

dx.doi.org/10.1021/bc100480a |Bioconjugate Chem. 2011, 22, 717–727

Bioconjugate Chemistry Overall, the order of efficacy observed was as follows: 00 :2 g 0:2 > 0:1 > 1:0 ≈ 2:0 > 1:1 > 1:2 ≈ SMCC-DM1. The huC242maytansinoid conjugates with a highly hindered disulfide linkage (1:2) or noncleavable thioether linkage (SMCC-DM1) were much less active suggesting that the reductive release of a maytansinoid thiol is required for maximal antitumor activity in these xenograft models. We also observed that the configuration of steric hindrance at the disulfide bond for the conjugates with intermediate disulfide bond stability (the “sidedness”) is important in determining the magnitude of antitumor activity. We found that having one or two methyl groups on the maytansinoid side of the disulfide bond confers an advantage relative to having the same degree of steric hindrance on the linker side, despite the fact that these conjugates have approximately equivalent in vivo plasma stability (Figure 4). This observation suggests that perhaps the nature of the maytansinoid metabolite released by reductive cleavage of the conjugate at the tumor cell is also important in determining conjugate activity (i.e., DM4 released from the 0:2 conjugate is more effective at inhibiting tumor growth than DM1 released from the 2:0 conjugate). Although the 0:2 and 0:1 conjugates do have slightly higher in vitro potencies than the corresponding 2:0 and 1:0 conjugates in vitro (Table 1), which is consistent with their greater activities in vivo, the order of intrinsic in vitro potency of the panel of conjugates taken as a whole does not adequately predict the trend in in vivo activities observed, suggesting additional factors contribute to activity in vivo. This is most clearly demonstrated by the fact that although the 0:2 and SMCC-DM1 conjugates have very similar potencies in vitro, the 0:2 conjugate is strikingly more active in vivo, despite having a shorter conjugate half-life in plasma and about half the plasma exposure (AUC0-¥). One property differentiating tumor xenografts growing in vivo from tumor cells growing on a cell culture plate may be the presence of inaccessible (i.e., nontargeted) tumor cells in the tightly packed three-dimensional tumor spheroid in vivo. Such nontargeted cells may be tumor cells that lack expression of the target antigen (see, for example, the heterogeneous expression of the CanAg antigen in the HT29 xenograft model) or cells that are inaccessible to conjugate due to barriers to macromolecule delivery.29,30 There is evidence in the literature from our group and others to support bystander killing of nontargeted tumor cells by immunoconjugates in vivo.23,31 Bystander killing requires processing of the conjugate by target cells followed by release of a membrane-permeable cytotoxic metabolite which then diffuses into neighboring tumor cells. Disulfide-linked AMCs have been shown to be effective at bystander killing because target cell processing of these conjugates yields membrane-permeable DMx metabolites including the thiol and the S-methylated derivative of the thiol.3,24 For bystander killing by disulfide-linked AMCs, release of the more hindered DM3 or DM4 thiols from a 0:1 or 0:2 conjugate might be expected to be more effective than release of the less hindered DM1 thiol from a 1:0 or 2:0 conjugate, since DM1, having a more reactive thiol, would likely be inactivated more quickly in the extracellular space by disulfide interchange with, for example, cystine. Additionally, it has been shown that the DM4 metabolite is efficiently S-methylated by the target cell in vivo (DM3 has also been shown to be efficiently S-methylated by COLO205 cells in vitro; unpublished data), while the DM1 metabolite is not.3 This S-methylation prevents any possibility of thiol-disulfide exchange reactions of the DM4 metabolite, which may contribute to greater bystander activity.

ARTICLE

Finally, the increased hydrophobicity of the more substituted DM3 and DM4 thiols relative to DM1 may also promote more efficient passage across cell membranes and tighter binding to microtubules leading to better bystander killing. The data presented in this report showing that the 0:2 and 0:1 huC242 conjugates have superior in vitro bystander activity as well as superior in vivo efficacy supports a role for bystander killing in the in vivo activities of these conjugates. The superior in vivo antitumor activity and improved therapeutic window of the 00 :2 disulfide-linked conjugate reported here for the huC242 maytansinoid conjugates has also been observed for other AMCs in development including antibody maytansinoid conjugates targeting Rv integrin, CD19, CD33, and CD138. 8,9,32,33 However, the relationships reported here may not be true for all AMC targets. For example, one notable exception is trastuzumab-SMCC-DM1 targeting HER2, where the noncleavable thioether-linked conjugate has comparable activity to the disulfide-linked conjugates in xenograft models. 4 The accumulated preclinical data therefore suggests that the optimal linker for each AMC may depend on multiple factors and needs to be explored for each target and antibody. For the CanAg-targeting huC242 antibody discussed in this report, the disulfide-linked huC242-SPDB-DM4 (00 :2) maytansinoid conjugate showed the best antitumor activity in the preclinical models. On the basis of the combined efficacy and tolerability results reported here, we conclude that this conjugate has an improved therapeutic window relative to the 1:0 conjugate, cantuzumab mertansine, which had previously been studied in clinical trials25-27 and provides a rationale for its clinical development.

’ ASSOCIATED CONTENT

bS

Supporting Information. Synthesis of sulfo-NHS-SMPPDM1 and sulfo-NHS-SMPP-DM4 reagents is described. This material is available free of charge via the Internet at http:// pubs.acs.org.

’ AUTHOR INFORMATION Corresponding Author

*Corresponding author. Phone: 781-895-0600. Fax: 781-8950611. E-mail: [email protected]. Present Addresses †

Norfolk Agricultural School, 400 Main Street, Walpole, MA 02081-3709. ‡ Acceleron Pharma, 128 Sidney St., Cambridge MA 02139. § Shire HGT, 125 Spring Street, Lexington, MA 02421.

’ ABBREVIATIONS: AMC, antibody-maytansinoid conjugate; AUC, area under the curve (plasma exposure); CL, plasma clearance;0 Cmax, maximal 0 2 (3-mercapto-1plasma concentration; DM1, N2 -deacetyl-N 20 20 -(4-mercapto-1oxopropyl)-maytansine; DM3, N -deacetyl-N 0 0 oxopentyl)-maytansine; DM4, N2 -deacetyl-N2 -(4-methyl-4mercapto-1-oxopentyl)-maytansine; DMx, any of DM1, DM3, or DM4; DTT, dithiothreitol; mAb, monoclonal antibody; SMCC, N-succinimidyl 4-(N-maleimidomethyl)cyclohexane-1-carboxylate; SMPP, N-succinimidyl 4-methyl-4-(2-pyridyldithio)pentanoate; SPDB, N-succinimidyl 4-(2-pyridyldithio)butanoate; SPDP, 725

dx.doi.org/10.1021/bc100480a |Bioconjugate Chem. 2011, 22, 717–727

Bioconjugate Chemistry

ARTICLE

N-succinimidyl 3-(2-pyridyldithio)propanoate; SPP, N-succinimidyl 4-(2-pyridyldithio)pentanoate; sulfo-NHS-SMPP-DM1, 0 0 N-sulfosuccinimidyl ester of N2 -deacetyl-N2 -[3-(3-carboxy-1, 1-dimethyl-propyl-dithio)-1-oxopropyl]-maytansine; sulfo-NHS0 0 SMPP-DM4, N-sulfosuccinimidyl ester of N2 -deacetyl-N2 -(4methyl-4-(2-methyl-5-carboxy)pentan-2-yl)disulfanyl)pentanoyl)maytansine; t1/2,β, β phase half-life for elimination; Vss, volume of distribution.

deglycosylated Ricin A-chain and hindered disulfide linkages. Cancer Res. 48, 6396–6403. (14) Flavell, D. J., Boehm, D. A., Okayama, K., Kohler, J. A., and Flavell, S. U. (1994) Therapy of human T-cell acute lymphoblastic leukemia in severe combined immunodeficient mice with two different anti-CD7-saporin immunotoxins containing hindered or non-hindered disulphide cross linkers. Int. J. Cancer 58, 407–414. (15) Arpicco, S., Dosio, F., Brusa, P., Crosasso, P., and Cattel, L. (1997) New coupling reagents for the preparation of disulfide crosslinked conjugates with increased stability. Bioconjugate Chem. 8, 327– 337. (16) Dosio, F., Arpicco, S., Adobati, E., Canevari, S., Brusa, P., De Santis, R., Parente, D., Pignanelli, P., Negri, D. R., Colnaghi, M. I., and Cattel, L. (1998) Role of cross-linking agents in determining the biochemical and pharmacokinetic properties of Mgr6-Clavin immunotoxins. Bioconjugate Chem. 9, 372–381. (17) Chari, R. V., Martell, B. A., Gross, J. L., Cook, S. B., Shah, S. A., Bl€attler, W. A., McKenzie, S. J., and Goldmacher, V. S. (1992) Immunoconjugates containing novel maytansinoids: promising anti-cancer drugs. Cancer Res. 52, 127–131. (18) Widdison, W. C., Wilhelm, S. D., Cavanagh, E. E., Whiteman, K. R., Leece, B. A., Kovtun, Y., Goldmacher, V. S., Xie, H., Steeves, R. M., Lutz, R. J., Zhao, R., Wang, L., Bl€attler, W. A., and Chari, R. V. (2006) Semisynthetic maytansine analogues for the targeted treatment of cancer. J. Med. Chem. 49, 4392–4408. (19) Hermanson, G. T. (1996) Bioconjugate Techniques, Part I, Academic Press, San Diego. (20) Furth, E. E., Thilly, W. G., Pennan, B. W., Liber, H. L., and Rand, W. M. (1981) Quantitative assay for mutation in diploid human lymphoblasts using microtiter plates. Anal. Biochem. 110, 1–8. (21) Morton, D. B., Abbot, D., and Barclay, R. (1993) Removal of blood from laboratory mammals and birds. Lab. Anim. 27, 1–22. (22) Worrell, N. R., Cumber, A. J., Parnell, G. D., Mirza, A., Forrester, J. A., and Ross, W. C. (1986) Effect of linkage variation on pharmacokinetics of ricin A chain-antibody conjugates in normal rats. Anti-Cancer Drug Des. 1, 179–188. (23) Kovtun, Y. V., Audette, C. A., Ye, Y., Xie, H., Ruberti, M. F., Phinney, S. J., Leece, B. A., Chittenden, T., Bl€attler, W. A., and Goldmacher, V. S. (2006) Antibody-drug conjugates designed to eradicate tumors with homogeneous and heterogeneous expression of the target antigen. Cancer Res. 66, 3214–3221. (24) Erickson, H. K., Widdison, W. C., Mayo, M. F., Whiteman, K., Audette, C., Wilhelm, S. D., and Singh, R. (2010) Tumor delivery and in vivo processing of disulfide linked antibody-maytansinoid conjugates. Bioconjugate Chem. 21, 84–92. (25) Tolcher, A. W., Ochoa, L., Hammond, L. A., Patnaik, A., Edwards, T., Takimoto, C., Smith, L., de Bono, J., Schwartz, G., Mays, T., Jonak, Z. L., Johnson, R., DeWitte, M., Martino, H., Audette, C., Maes, K., Chari, R. V., Lambert, J. M., and Rowinsky, E. K. (2003) Cantuzumab Mertansine, a maytansinoid immunoconjugate directed to the CanAg antigen: A phase I, pharmacokinetic and biologic correlative study. J. Clin. Oncol. 21, 211–222. (26) Helft, P. R., Schilsky, R. L., Hoke, F. J., Williams, D., Kindler, H. L., Sprague, E., DeWitte, M., Martino, H. K., Erickson, J., Pandite, L., Russo, M., Lambert, J. M., Howard, M., and Ratain, M. J. (2004) A phase I study of Cantuzumab Mertansine administered as a single intravenous infusion once weekly in patients with advanced solid tumors. Clin. Cancer Res. 10, 4363–4368. (27) Rodon, J., Garrison, M., Hammond, L. A., de Bono, J., Smith, L., Forero, L., Hao, D., Takimoto, C., Lambert, J. M., Pandite, L., Howard, M., Xie, H., and Tolcher, A. W. (2008) Cantuzumab mertansine in a three-times a week schedule: a phase I and pharmacokinetic study. Cancer Chemother. Pharmacol. 62, 911–919. (28) Mills, B. J., and Lang, C. A. (1996) Differential distribution of free and bound glutathione and cysteine in human blood. Biochem. Pharmacol. 52, 401–406. (29) Jain, R. K. (1994) Barriers to drug delivery in solid tumours. Sci. Am. 271, 59–65.

’ REFERENCES (1) Lambert, J. M. (2010) Antibody-maytansinoid conjugates: A new strategy for the treatment of cancer. Drugs Future 35, 471–480. (2) Chari, R. V. (2008) Targeted cancer therapy: Conferring specificity to cytotoxic drugs. Acc. Chem. Res. 41, 98–107. (3) Erickson, H. E., Park, P. U., Widdison, W. C., Kovtun, Y. V., Garrett, L. M., Hoffman, K., Lutz, R. J., Goldmacher, V. S., and Bl€attler, W. A. (2006) Antibody-maytansinoid conjugates are activated in targeted cancer cells by lysosomal degradation and linker dependent intracellular processing. Cancer Res. 66, 4426–4433. (4) Lewis Phillips, G. D., Li, G., Dugger, D. L., Crocker, L. M., Parsons, K. L., Mai, E., Bl€attler, W. A., Lambert, J. M., Chari, R. V., Lutz, R. J., Wong, W. L., Jacobson, F. S., Koeppen, H., Schwall, R. H., KenkareMitra, S. R., Spencer, S. D., and Sliwkowski, M. X. (2008) Targeting HER2-positive breast cancer with Trastuzumab-DM1, an antibodycytotoxic drug conjugate. Cancer Res. 68, 9280–9290. (5) Krop, I. E., Beeram, M., Modi, S., Jones, S. F., Holden, S. N., Yu, W., Girish, S., Tibbitts, J., Yi, J. H., Sliwkowski, M. X., Jacobson, F., Lutzker, S. G., and Burris, H. A. (2010) Phase I study of trastuzumabDM1, an HER2 antibody-drug conjugate, given every 3 weeks to patients with HER2-positive metastatic breast cancer. J. Clin. Oncol. 28, 2698–2704. (6) Xie, H., Audette, C., Hoffee, M., Lambert, J., and Bl€attler, W. (2004) Pharmacokinetics and biodistribution of the antitumor immunoconjugate, cantuzumab mertansine (huC242-DM1), and its two components in mice. J. Pharmacol. Exp. Ther. 308, 1073–1082. (7) Tassone, P., Gozzini, A., Goldmacher, V., Shammas, M. A., Whiteman, K. R., Carrasco, D. R., Li, C., Allam, C. K., Venuta, S., Anderson, K. C., and Munshi, N. C. (2004) In vitro0 and in vivo activity of 0 the maytansinoid immunoconjugate huN901-N2 -deacetyl-N2 -(3-mercapto-1-oxopropyl)-maytansine against CD56þ multiple myeloma cells. Cancer Res. 64, 4629–4636. (8) Chen, Q., Millar, H. J., McCabe, F. L., Manning, C. D., Steeves, R., Lai, K., Kellogg, B., Lutz, R. J., Trikha, M., Nakada, M. T., and Anderson, G. M. (2007) Rv Integrin-targeted immunoconjugates regress established human tumors in xenograft models. Clin. Cancer Res. 13, 3689–3695. (9) Al-Katib, A. M., Aboukameel, A., Mohammad, R., Bissery, M. C., and Zuany-Amorim, C. (2009) Superior antitumor activity of SAR3419 to Rituximab in xenograft models for non-hodgkin’s lymphoma. Clin. Cancer Res. 15, 4038–4045. (10) Johansson, C., Nilsson, O., Baeckstr€om, D., Jansson, E. L., and Lindholm, L. (1991) Novel epitopes on the CA50-carrying antigen: chemical and immunochemical studies. Tumour Biol. 12, 159–170. (11) Baeckstr€om, D., Hansson, G. C., Nilsson, O., Johansson, C., Gendler, S. J., and Lindholm, L. (1991) Purification and characterization of a membrane-bound and secreted mucin-type glycoprotein carrying the carcinoma-associated sialyl-Lea epitope on distinct core proteins. J. Biol. Chem. 266, 21537–21547. (12) Thorpe, P. E., Wallace, P. M., Knowles, P. P., Relf, M. G., Brown, A. N., Watson, G. J., Knyba, R. E., Wawrzynczak, E. J., and Blakey, D. C. (1987) New coupling agents for the synthesis of immunotoxins containing a hindered disulfide bond with improved stability in vivo. Cancer Res. 47, 5924–5931. (13) Thorpe, P. E., Wallace, P. M., Knowles, P. P., Relf, M. G., Brown, A. N., Watson, G. J., Blakey, D. C., and Newell, D. R. (1988) Improved antitumor effects of immunotoxins prepared with 726

dx.doi.org/10.1021/bc100480a |Bioconjugate Chem. 2011, 22, 717–727

Bioconjugate Chemistry

ARTICLE

(30) Minchinton, A. I., and Tannock, I. F. (2006) Drug penetration in solid tumours. Nat. Rev. Cancer 6, 583–592. (31) Okeley, N. M., Miyamoto, J. B., Zhang, X., Sanderson, R. J., Benjamin, D. R., Sievers, E. L., Senter, P. D., and Alley, S. C. (2010) Intracellular activation of SGN-35, a potent anti-CD30 antibody-drug conjugate. Clin. Cancer Res. 16, 888–897. (32) Giles, F., Morariu-Zamfir, R., Lambert, J.,Verstovsek, S., Thomas, D., Ravandi, F., and Deangelo, D. (2006) Phase I study of AVE9633, an antiCD33 maytansinoid immunoconjugate, administered as an intravenous infusion in patients with refractory/relapsed CD33 positive acute myeloid leukemia (AML). Blood (ASH Annual Meeting Abstracts) 106, Abstract 4548. (33) Ikeda, H., Hideshima, T., Fulciniti, M., Lutz, R. J., Yasui, H., Okawa, Y., Kiziltepe, T., Vallet, S., Pozzi, S., Santo, L., Perrone, G., Tai, Y. T., Cirstea, D., Raje, N. S., Uherek, C., D€alken, B., Aigner, S., Osterroth, F., Munshi, N., Richardson, P., and Anderson, K. C. (2009) The monoclonal antibody nBT062 conjugated to cytotoxic maytansinoids has selective cytotoxicity against CD138-positive multiple myeloma cells in vitro and in vivo. Clin. Cancer Res. 15, 4028–4037.

727

dx.doi.org/10.1021/bc100480a |Bioconjugate Chem. 2011, 22, 717–727