- Top of page
- Materials and methods
- Supporting Information
Immunoconjugates are being explored as novel cancer therapies with the promise of target-specific drug delivery. The immunoconjugate, huN901–DM1, composed of the humanized monoclonal IgG1 antibody, huN901, and the maytansinoid drug, DM1, is being tested in clinical trials to treat small cell lung carcinoma (SCLC). huN901–DM1 contains an average of three to four DM1 drug molecules per huN901 antibody molecule. The drug molecules are linked to huN901 through random modification of huN901 at ε-amino groups of lysine residues, thus yielding a heterogeneous population of conjugate species. We studied the drug distribution profile of huN901–DM1 by electrospray time-of-flight mass spectrometry(ESI-TOFMS), which showed that one to six DM1 drug molecules were attached to an antibody molecule. Both light and heavy chains contained linked drugs. The conjugation sites in both chains were determined by peptide mapping using trypsin and Asp-N protease digestion. Trypsin digestion identified modified lysine residues, since these residues were no longer susceptible to enzymatic cleavage after conjugation with the drug. With respect to Asp-N digestion, modified peptides were identified by observing a mass increase corresponding to the modification. The two digestion methods provided consistent results, leading to the identification of 20 modified lysine residues in both light and heavy chains. Each lysine residue was only partially modified. No conjugation sites were found in complementarity determining regions (CDRs). Using structural models of human IgG1, it was found that modified lysine residues were on the surface in areas of structural flexibility and had large solvent accessibility.
Monoclonal antibody (MAb) therapy has emerged as an important therapeutic modality for the treatment of cancer. One approach is to use naked or unconjugated MAbs as therapeutic agents, such as Rituxan and Herceptin (Grillo-Lopez et al. 2001; Kim 2003; Ross et al. 2003). In another approach tumor-selective MAbs are used to target cytotoxic agents to diseased cells (Liu et al. 1996; Milenic 2002). Typically, several molecules of a potent cytotoxic agent are covalently linked to a MAb, thus forming an immunoconjugate (Blättler and Chari 2001; Milenic 2002; Lo-Coco et al. 2004; Lu et al. 2005). huN901–DM1 is an immunoconjugate composed of the humanized MAb, huN901, and the cytotoxic drug DM1. DM1 is a semisynthetic ansamacrolide, and belongs to the maytansinoid class of anti-microtubular cytotoxic agents originally defined by the natural compound Maytansine (Cassady et al. 2004). HuN901 binds to the CD56 antigen present on cells of small-cell lung carcinomas (SCLC), other neuroendocrine tumors, and a large proportion of multiple myelomas (Zeromski et al. 2001). HuN901–DM1 is currently being evaluated in clinical trials to treat SCLC (Fossella et al. 2002). In addition, it has shown significant anti-myeloma activity in a preclinical study (Tassone et al. 2004).
The procedure for the preparation of huN901–DM1 from huN901 and DM1 is illustrated in Figure 1. The antibody is first reacted with N-succinimidyl 4-(2′-pyridyldithio) pentanoate (SPP), which results in huN901 species with lysine residues that carry a 4-(2′-pyridyldithio)-1-oxopentyl substituent on their ε-amino group. The linker-modified antibody is then reacted with DM1, which contains a primary sulfhydryl group that rapidly undergoes a disulfide exchange reaction with the 2′-pyridyldithio moiety of the linker. This conjugation process is well controlled and introduces, on average, three to four DM1 molecules per huN901 molecule.
This method of conjugate preparation will result in a mixture of conjugate species that differ in the number of drugs linked as well as the sites of drug linkage, since huN901 has 86 lysine residues that potentially can be modified. Structural characterization is challenging due to the heterogeneous nature of the conjugate, the large molecular size of antibodies and possible low abundances of modified residues. There are only a few studies in which conjugated antibodies were analyzed by mass spectrometry to give the distribution and load of linked small molecules (Siegel et al. 1997; Adamczyk et al. 2000; Lu et al. 2005). In these cases, however, the exact locations of conjugation sites remained elusive. Determining the conjugation sites is important, since the affinity of MAb conjugates may be affected significantly by modification of residues in complementarity determining regions (CDRs) resulting in non-active conjugate species in the mixture.
Electrospray ionization mass spectrometry (ESIMS) has proven to be a powerful structural characterization technique for large proteins such as MAbs and MAb conjugates (Feng and Konishi 1992; Lewis et al. 1994; Siegel et al. 1997; Bongers et al. 2000). Recently, we published the structural characterization—primary sequence and posttranslational modifications—of huN901, which was performed with tryptic and Asp-N peptide mapping in combination with electrospray ionization time-of-flight mass spectrometry (ESI-TOFMS) (Wang et al. 2005). In the current study, ESI-TOFMS was used to both assess the drug distribution in huN901–DM1 and, for the first time, to determine the conjugation sites. Deglycosylation of huN901–DM1 and tryptic and Asp-N protease digestions were performed to help in the characterization.
- Top of page
- Materials and methods
- Supporting Information
The antibody–drug conjugate, HuN901–DM1, is prepared by first random modification of the antibody with a cross-linking reagent to which the drug is covalently linked in a second step. This method yields a conjugate preparation that is heterogeneous at the molecular level. The current study of the structural composition of the conjugate shows that the heterogeneity is twofold. First, the preparation contains several conjugate species that differ in the number of linked drug molecules per antibody molecule: Species with one to six linked drugs were identified, as well as small amounts of unconjugated (naked) antibody. Second, each species with a certain number of linked drug molecules can have the linked drugs distributed to different sites of modification: Forty different sites of modification were identified considering that there are two copies of light and heavy chains in each antibody molecule. The identification of 40 modification sites in the antibody indicates that the drug molecules distribute over ∼47% of the 86 lysine residues in the antibody. This number of conjugation sites is large enough so that each site is only partially modified, which leads to low abundances for many modified peptides in the enzymatic digests. Consequently, such peptides could not be detected in the chromatographic digestion maps. Only careful analysis of the MS data through searching for expected mass peaks allowed the identification of most of the modification sites. Since such an analysis can detect modified peptides present at an amount of as low as a few percent of the total expected amount, one cannot exclude that more than 47% of the lysine residues are modified, albeit at lower levels they are not detectable with the applied technique. It is also worth mentioning that a recent study suggested that tyrosine and histidine residues can also be modified by succinimide esters (Leavell et al. 2004). However, extensive searches for expected masses indicated that modification of tyrosine and histidine residues were not present in huN901–DM1.
The determination of the conjugation sites in huN901–DM1 also showed that the lysine residues in the CDRs of huN901 were not modified (Fig. 8). This is consistent with the observation that huN901–DM1 had retained the full binding affinity of the naked antibody. Although in this study peptide mapping combined with LC/MS analysis was only performed for one batch of huN901–DM1, the modification sites in different batches of the conjugate are expected to be largely reproducible, assuming that the protein conformation has no change under the same reaction conditions. UV and MS analysis of intact conjugate did show consistent drug load and drug distribution profile from batch to batch (data not shown).
Our results show that not all lysine residues of huN901–DM1 are modified to a degree detectable by ESI-TOFMS. Under the same reaction conditions, whether or not a lysine residue is modified and to what extent it is modified is expected to depend on the location of the residues in the antibody, which determines the local environment. For example, one of the modified Asp-N peptides that were detected in the HPLC peptide mapping, HD8 (heavy-chain residues 222–249), is located in the hinge region. Furthermore, the trypsin digestion analysis (Table 1) showed that both lysine residues in this peptide, K223 and K247, which are the only two lysine residues in the hinge region, were modified. This may be due to the large solvent exposure and the structural flexibility of the hinge region of the antibody.
To understand the structural features of the modification sites in huN901–DM1, the crystal structural models of human IgG1 (huIgG1) Fab (PDB code 1UCB; Sheriff et al. 1996), and Fc (PDB code 1H3X; Krapp et al. 2003) were used to mimic the structure of the constant regions of huN901, since huN901 and huIgG1 share identical amino acid sequences in constant regions (Roguska et al. 1994). The solvent accessibility and average depth for the lysine residues in huIgG1 were calculated, and the B-factors of the side-chain nitrogen of lysine residues were also obtained (Fig. 9). The modified lysine residues in huN901–DM1 were mapped to the corresponding lysine residues in huIgG1. As shown in Figure 9, most of the modified lysine residues have relatively large solvent- accessible areas, large B-factors, but small values of residue depth, indicating that these residues either are located on the protein surface or have relatively large structural flexibilities. Although this correlation is generally true for most modified lysine residues, exceptions do exist. For example, residues K126 in the light chain, and K213 and K290 in the heavy chain of huIgG1 are surface residues with relatively large solvent-exposure areas or B-factors (Fig. 9). However, the three lysine residues at similar locations in huN901–DM1 were not modified. These exceptions suggest that other factors, such as effects from neighboring residues or hydrogen bonds, may also contribute to the reactivity of individual lysine residues.
The crystal structures of huIgG1 Fab and Fc were also used to examine the locations of the modified lysine residues in the secondary structures of huN901 constant regions. As shown in Figure 10, most modified lysine residues are located in the surface loops of the antibody, with only two modified lysine residues in the heavy chain (K323 and K393) located in β-sheets. Although they are not located in loop regions, these two residues still have relatively large solvent accessibilities and large B-factors (Fig. 9).
In summary, the structural heterogeneity caused by random conjugation of lysine residues in huN901– DM1 was characterized by mass spectrometry and peptide mapping. Twofold heterogeneity was revealed: First, the antibody contains species linked with different number of drug molecules; second, antibody with the same number of drug molecules may have different sites of linkage. The conjugation sites were determined by both trypsin and Asp-N protease digestion. Although trypsin digestion led to determination of specific sites of modification, chromatographic differences between maps of naked antibody and conjugate appeared more evident upon Asp-N protease digestion. Modification sites detected by both methods are generally consistent, therefore strengthening the results. All modifications are partial, with about seven lysine residues in the light chain and 13 lysine residues in the heavy chain modified. The detected conjugation sites in huN901–DM1 were mapped in the crystal structural models of huIgG1, and the modified residues were generally found in areas with large solvent accessibility and structural flexibility. Results of this study have enhanced the understanding of the structure of MAb–maytansinoid conjugates and other types of immunoconjugates. Additionally, correlation of residue location and modification is expected to provide useful information to improve conjugation selectivity and to design conjugates with better structural homogeneity and biological activities. The methods used in this study should be applicable to elucidate structural details of other immunoconjugates.