Evaluating the Bioequivalence of Antibody–Drug Conjugates

Oncology treatment options have expanded considerably in the past decade, particularly to include biologic-based molecules. Classical small-molecule chemotherapeutic agents are generally nonspecific (i.e., they bind to and affect other physiological targets) and may cause significant side effects. In contrast, targeted monoclonal antibody (mAb) therapy generally is very selective and has a milder side-effect profile, although rare but serious adverse events can occur, including infusion reactions, thrombocytopenia, cardiac arrest, acute renal failure, immune toxicity, pulmonary toxicity, and susceptibility to latent viral infections (1).

Many clinically relevant mAbs were initially developed using mouse–human hybridoma technology, which resulted in mAbs that contained nonhuman proteins. Although therapeutically useful, these nonhuman proteins can cause cytokine-release syndrome or uncontrolled hypercytokinemia, thus resulting in multiple organ damage. Technologies to reduce the proportion of mouse protein in the mAb, including chimerization (i.e., partial replacement of mouse mAb sequences with human mAb) and humanization (i.e., essentially complete replacement with human sequences) reduce this toxicity. Humanized antibodies are approved to treat in cancer, inflammation, autoimmunity, and infectious diseases and to remediate transplant rejection. The pharmacokinetics (PK) and pharmacodynamics (PD) of mAbs are complex and depend upon the mAb isotype, subtype, half-life, and target indication (2).

INGRAM PUBLISHING/GETTY

Antibody–drug conjugates

Ideal therapeutics would combine the specificity of antibody-targeting with the potency and pharmacokinetics of small-molecule drugs, an idea that has spawned the field of antibody–drug conjugates (ADCs). These constructs combine an antibody against a target of interest with a compound that adds interesting pharmacology, such as cytotoxicity. However, the theory of using mAb targeting to deliver a lethal drug payload to tumor cells gives rise to complex technical challenges when put into practice. The mAb must be conjugated to a potent cytotoxin with a linker that is stable enough not to release large amounts of the drug into systemic circulation, but labile enough to release the cytotoxin upon internalization by the targeted cell. The manufacturing process to couple the cytotoxin onto the mAb must produce an ADC that maintains recognition of the target after conjugation. The number of conjugated cytotoxins per intact mAb, as well as the level of unconjugated mAb, must be controlled to reduce batch-to-batch variability.

Trastuzumab emtansine (T-DM1), an ADC developed by Genentech (now part of Roche), combines the mAb Herceptin (trastuzumab), which targets human epidermal growth factor receptor 2 (HER2) receptors in breast and stomach cancer, with maytansine, a small-molecule cytotoxin that binds to tubulin to prevent microtubule formation through a nonreducible bis-maleimido-trixyethylene glycol linker (see Figure 1). Trastuzumab is approved for use only in HER2-positive cancers, but not all HER2-positive cells have sufficient apoptotic capacity to be killed by trastuzumab binding alone (3). T-DM1 combines the ability of trastuzumab to selectively target the HER2 receptor with the potent cytotoxic agent, maytansine, to kill the cell regardless of the HER2-induced apoptotic response (4). T-DM1 binding to the HER2 receptor causes internalization into the cell where the maytansine is released from the antibody conjugate to kill the tumor cell (3–7). T-DM1 induces a potent antiproliferative effect in trastuzumab-resistant tumor cells, but does not affect normal human cells. Also, T-DM1, but not trastuzumab, induces cell lysis in breast tumor cells in vitro (4). Likewise, T-DM1-inhibited tumor growth caused tumor regression in animal models of HER2-positive breast cancer. T-DM1 showed overall better efficacy and PK than trastuzumabwith diminished toxicity in vivo and in vitro (5). T-DM1 also showed success in recent Phase II clinical trials (4–7).

Figure 1: Structure of trastuzumab emtansine (T-DM1). Reproduced with permission from Discovery Medicine, A. Beck et al., Discov. Med. 10 (53), 329–339 (2010). (FIGURE 1 IS COURTESY OF DISCOVERY MEDICINE AS CITED WITH CREDIT IN FIGURE 1.)

In contrast to the current success by T-DM1, the only approved ADC, Mylotarg (gemzutumab ozogamicin), was voluntarily removed from the market by Pfizer on June 21, 2010, after postapproval clinical trials in acute myeloid leukemia, a bone-marrow cancer, revealed safety and efficacy issues (8–12). Mylotarg is an ADC chemotherapy agent composed of a recombinant, humanized IgG4, kappa antibody against the CD33 antigen, conjugated to a cytotoxic antitumor antibiotic, N-acetyl gamma calicheamicin dimethyl hydrazide (DNA-disrupting agent), through the acid labile acetylphenoxy butanoic linker (12). Mylotarg is a heterogenous mixture of approximately 50% mAb with two to three cytotoxin moieties per mAb and 50% unconjugated mAb (10, 12). This ADC was never approved in Europe and did not extend survival rates, but showed a higher fatal toxicity rate than chemotherapy alone (8, 9, 11, 12).

Analytical methods for bioequivalence studies

Analytical-method development for biologics usually starts with establishing a suitable manufacturing process, followed by the development of analytical methods intended to confirm adequate structural and functional comparison with the target product profile (i.e., for ADCs, the starting mAb). The regulatory guidelines about this development require an understanding of the molecular structure, glycol microheterogeneity profile, impurities, degradation products, and bioavailability, as shown in Table I (13–16).

Table I. Analytical technologies for assessment of biologics.

The normal complexity in evaluating lot-to-lot variation and its implications is further compounded because each of the three major components (mAb, linker, and cytotoxin) must be fully characterized. According to EMA's biosimilar guidelines, "Every mAb is unique and small structural changes can have significant functional consequences since even the same expression system and similar culture conditions might lead to a distinct product profile (e.g., impurities and microheterogeneity)" (14). Manufacturing modifications or synthetic-process changes may require further nonclinical and clinical studies.

The structural integrity of ADCs is evaluated using techniques such as amino acid sequencing (N- or C- terminal), biochemical testing, carbohydrate mapping, high-performance liquid chromatography (HPLC), electrophoresis (i.e., sodium dodecyl sulfate polyacrylamide gel electrophoresis [SDS-PAGE], isoelectric focusing [IEF], and Western blotting), immunoassays (enzyme-linked immunosorbent assay [ELISA] and Gyros), macromolecular liquid chromatography–mass spectrometry (MS)/MS (LC–MS/MS), nuclear magnetic resonance (NMR) spectroscopy, optical spectroscopy, X-ray crystallography, peptide mapping, ultracentrifugation, and other physiochemical methods to show that the purified mAb is not fragmented, aggregated, or otherwise modified. Specificity characterization can be performed using in vitro cell-based assays or in vivo animal models and should provide evidence that the ADC is specific for its intended target and has low cross-reactivity with human tissues. Potency characterization is used for assessment of lot-to-lot consistency and stability and may involve techniques, such as ELISA, Gyros, flow cytometry, cytotoxicity, and animal models.

Most mAbs are produced in bioreactors, so DNA, RNA, viral burden, host-cell protein, endotoxins/pyrogenicity, and growth-media components all must be monitored and controlled in the final ADC product. mAbs are glycosylated proteins composed of four peptide chains connected by disulfide bridges. The oligosaccharide residues on mAbs may be involved in activity, and, therefore, mAbs should be characterized for protein, peptide sequence, glycoprotein, and oligosaccharide content. The binding activity, affinity, avidity, immunoreactivity, source of all materials, chemical structures and production processes, purity, viral load, and the average mAb-to-drug ratio all must be clearly defined. Contaminants and impurities must be carefully controlled, monitored, and minimized. Stability should be carefully monitored for intact ADC and bioactivity. Immunogenicity is a major concern for nonhumanized mAbs and must be thoroughly evaluated (16). Bioavailability, receptor binding, stability, and immunogenicity are major issues of pharmaceutical equivalence for ADCs.

Test article

ADCs are composed of three distinct portions: the mAb, linker, and cytotoxin, each of which has unique properties from a manufacturing and characterization perspective. For example, the mAb portion of Mylotarg was produced in a mammalian-cell suspension culture using a cell line and was purified using three independent steps (i.e., low pH treatment, diethylaminoethyl (DEAE) Sepharose chromatography, and viral filtration) to achieve antibody purification with acceptable retrovirus inactivation and removal (10). Mylotarg release testing included SDS–PAGE, IEF, HPLC, ELISA, aggregate and unconjugated calicheamicin ozogamicin, endotoxins by limulus amebocyte lysate and bioburden, protein content, peptide mapping, and oligosaccharide profiling (9). Potency was tested using a cytotoxicity assay and an immunoaffinity antigen-binding ELISA (9).

The characterization of a manufactured lot of an ADC usually reveals the amount of free and bound mAb protein as well as the mAb-to-drug ratio. Cytotoxic agents, such as DM1, have an absorption maximum at 252 nm, and mAbs absorb at 280 nm. Therefore, the amount of drug bound to mAb can be determined using differential-absorption measurements at 252 nm and 280 nm (17). Typical ADC drug-to-mAb ratios are between 2 and 4, as seen in hydrophilic interaction liquid chromatography (HILC) analysis of ADCs (see Figure 2) and reverse-phase chromatographic analysis of ADCs (see Figure 3) (18, 19).

Figure 2: Conjugation-related antibody drug conjugate variants using hydrophilic interaction liquid chromatography: MR0 is unbound free monoclonal antibody (mAb), MR1 is one cytotoxin, MR2 are two cytotoxins, MR3 are three cytotoxins, MR4 are four cytotoxins, MR6 are six cytotoxins, and MR 8 are eight cytotoxins. Reproduced with permission from Seattle Genetics from K. Anderson, "Physiochemical Characterization of Aurustatin-based Antibody Drug Conjugates," presented at the National Biotechnology Conference (San Francisco, 2011). (FIGURE 2 IS COURTESY OF SEATTLE GENETICS AS CITED, RESPECTIVELY, WITH CREDIT IN FIGURE 2.)

This gross characterization of each lot is problematic for several reasons. First, the nonconjugated antibody (i.e., free antibody) can bind with the same or even better avidity as the ADC, but the contribution of this component to product efficacy or toxicity is not systematically evaluated during ADC drug development. Preclinical efficacy studies should include a quantitative assessment of the impact of the free antibody in the mixture by using multiple manufacturing lots to capture the effect of that variability.

Figure 3: The drug load of antibody–drug conjugates can be determined by reverse-phase chromatography. Reproduced with permission from Seattle Genetics, from R.P. Lyon, et al., "Development of Parallel Conjugation and Assay Methodologies to Screen for Antibodies with Optimal Properties for use as Antibody-Drug Conjugates," presented at the 101st Annual American Association for Cancer Research Meeting (Washington DC, 2010). (FIGURE 3 IS COURTESY OF SEATTLE GENETICS AS CITED, RESPECTIVELY, WITH CREDIT IN FIGURES 3.)

Second, the average drug-load-per-antibody value does not ensure the efficacy of a particular lot. For example, an antibody ratio of four molecules of cytotoxin per antibody could result if all ADCs in the lot have four drug molecules attached, if 50% of the antibodies have eight drug molecules attached and 50% were free antibody, or if 25% of the antibodies have drug attached and 75% are drug-free antibodies. In the latter two examples, the free antibody would compete for binding with the conjugated antibody, thus limiting that particular lot's efficacy. Knowing the quantitative distribution of conjugation ratios is critical for batch-release testing. Similarly, the location of the conjugation reaction can affect activity, and in vivo liability should be evaluated. Finally, understanding the rate of in vivo cleavage at each conjugation site allows modeling of the outcome of a particular manufacturing lot if quantitative assessment of the species distribution is available.

The manufacturing process to link the small molecule to the antibody usually favors covalent binding to the hydrophilic domains of the antibody through linking to the sulfhydryl (SH) group on cysteine residues or coupling to the amine (NH₂) group on lysine molecules (18, 19). Thermodynamically, most conjugation reactions occur at the most accessible and lowest energy-barrier binding sites. While an in vitro efficacy assay can assess the potency of an individual manufacturing lot at the time of release, the dynamic loss of the linker-small-molecule complex that occurs upon dosing can cause different outcomes in vivo. In vivo exposure to the aqueous environment and endogenous enzymes could cause these easily accessible sites to be prone to degradation. Therefore, ADCs with only one or two small molecules attached may be more easily converted to the free mAb and block ADC interactions with the target molecule. Scientist still have insufficient data to predict this site-specific degradation or even the degradation dynamics in vivo.

Glycosylation and microheterogeneity

Humanized mAbs are produced in bioreactors that use cell lines cloned to express the desired product. While the fidelity of the amino acid backbone of the humanized mAbs is excellent, the control of posttranslational modifications, such as glycosylation, is not. Minor modifications of the manufacturing process, many unintentional, can produce dramatic differences in the microheterogeneity or distribution of glycoforms in the expressed protein. Different glycosylation patterns can dramatically affect mAb PK and PD because glycosylation appears to protect the protein from digestion by endogenous proteases, thus resulting in a different half-life and potentially different efficacy.

Pharmacokinetics: safety and effficacy

Humanized antibody drug development usually requires a PK analysis in conjunction with an antidrug-antibody assessment. However, there are immense technical challenges to determine the adsorption, distribution, metabolism, and excretion (ADME) of proteins, and human antibodies are a naturally occurring part of human biochemistry. Unlike classic small-molecule xenobiotic drug development, there has been little effort to determine ADME of these products. Because ADCs are a combination of humanized mAb, linker, and small molecule (see Figures 4 and 5), these assumptions of similarity to endogenous antibodies may not hold true.

Figures 4–5: Possible free small molecules can be derived in vivo from the following precursors: (1) monoclonal antibody (mAb)–linker–drug goes to mAb–linker + drug; (2) mAb–linker–drug goes to mAb + linker–drug then goes to linker + drug; (3) mAb–linker–drug gets peptide degraded to peptide–linker–drug, which can go to peptide–linker + drug or to peptide + linker–drug, which then goes to linker + drug. ADC is antibody–drug conjugate. (FIGURES 4 AND 5 ARE COURTESY OF THE AUTHORS)

The most complete assessment of plasma exposure during the development of an ADC includes the bioanalytical measurement of four critical attributes:

• The amount of free drug released from the ADC into circulation; this assay is relatively straightforward but requires method development and validation to ensure that no additional free drug is produced during execution of the assay.

• The amount of drug that is still attached to the humanized mAb, which is usually determined by either filtration or immune-precipitation of the ADC from a plasma sample, followed by chemical release of the small molecule from the protein, with subsequent LC–MS/MS analysis of the small molecule to yield the "bound-drug" amount.

• The amount of free antibody, which is determined using an immunoassay that is specific for the antibody but does not cross-react with the ADC. Often an assay is developed where all human IgGs are isolated by immune-precipitation specific for an intact Fc portion, followed by an ELISA specific for the idiotype.

• The ADC complex, which can be measured using an ELISA assay specific for the ADC that does not cross react with free antibody. The reagents for this assay are very difficult to develop because the assay requires an antibody that will recognize the linker-small-molecule complex as the epitope. Alternatively, these assays may involve digestion followed by LC–MS/MS detection, or, if the small molecule is attached to the Fc portion of the ADC, an immunoprecipitation of the Fc portion could separate the ADCs from the free antibody, followed by detection with an anti-idiotype ELISA assay.

Although this set of assays seems comprehensive, the lack of information regarding the degradation profile of an ADC may reveal some limitations. For example, these assays do not measure whether the linker portion of the ADC is released with the drug attached despite potential for drug release from this complex. Similarly, circulating peptide fragments containing the linker/small-molecule complexes might be capable of releasing the drug. Without mass balance and profiling information, unmeasured small molecules may be present in a dose and cause differences in safety and efficacy despite similar PK profiles for the two lots.

Safety

Premature release of cytotoxin into systemic circulation is the main safety concern for ADCs. A safety margin of free-circulating cytotoxin is established during drug development that accounts for freely circulated cytotoxin, but does not account for the possibility that the circulating free cytotoxin levels may change from lot to lot due to inherent variation during ADC production. Understanding the toxic potential of a particular ADC lot requires identifying all in vivo sources that could generate free cytotoxin, including intact ADCs, linker-bound cytotoxin, and partial linker-bound cytotoxin fragments or degraded peptide fragments containing intact linker and cytotoxin (see Figures 4 and 5). A comparison of the rate of cytotoxin released from each source with its respective PK elimination rate would enable assessment of the safety impact of each particular source of free cytotoxin. If the rate of elimination from a particular free cytotoxin source were slower than the release rate from that source, method development should include a measure of area under the curve (AUC) of that particular free-cytotoxin source. Safety studies also should compare the AUC of free cytotoxin and fast-release sources, as well as the safety margin for the free cytotoxin from toxicology studies. In addition to free cytotoxin analysis, the linker-bound cytotoxin also should be evaluated for toxicity to ensure that it is inactive. Obviously, any major, active metabolites of the free cytotoxin should be monitored in accordance with the Metabolites in Safety Testing guidelines (20).

Efficacy

The major determinants of efficacy are sufficient amounts of intact ADC to achieve the desired effect, as well as sufficiently low levels of free antibody so that there is no competition with the intact ADC to compromise its efficacy. If the ADC and free antibody have comparable affinity for the target, the intact ADC will compete equally with free antibody for the receptor. However, if the ADC has a lower affinity, then even small amounts of free antibody will compromise efficacy. If ADC affinity is higher than the free antibody, free-antibody levels will generally not significantly influence efficacy.

ADC assessment using an average cytotoxin load per antibody does not provide information on free-antibody levels that can compete with the ADC. Similarly, after administration of an ADC there is significant cleavage of the cytotoxin that not only poses a safety concern, but potentially generates a free-circulating antibody that may compete with the ADC for binding, thus compromising its efficacy. Efficacy can be further complicated by the generation of ADC fragments in vivo (see Figures 4 and 5), whose binding and efficacy are unknown. Cleavage at the hinge region of the ADC could generate fragments that bind to the receptor and evade ELISA detection, yet compromise ADC efficacy. The glyco microheterogeneity in normal lot-to-lot variation can cause different degradation fragments between lots.

Conclusion

ADC technologies are rapidly expanding the use of these therapeutic modalities that combine the targeting potential of mAbs with the pharmacology of small molecules. However, these ADC combinations also present significant challenges in meeting the regulatory requirements for analytical method development. For example, establishing ADC equivalence requires characterization of the metabolism and catabolism and excretion of all parts of the ADC. Because antibody fragments and the in vivo generation of free antibody can compromise efficacy, the rate at which these metabolites and catabolites are formed in vivo, as well as their potential effects on safety and efficacy, must be assessed. Any metabolite or catabolite that induces cytotoxicity or the ability to bind and compete with the intact ADC for the receptor must be measured. These challenges require a thoughtful and thorough development plan that combines expertise from both small-molecule and biologics development.

Alan Breau*, PhD, is vice-president of bioanalytical and analytical services at MPI Research, 54943 North Main St. Mattawan, MI 49071, tel. 269.668.3336, ext. 3230; alan.breau@mpiresearch.com. Monica Lee Whitmire, MS, BS, MT (ASCP), is study director at MPI Research.

*To whom all correspondence should be addressed.

References

1. T.T. Hansel, et al., Nature Rev. Drug Discov. 9 (4), 325–338 (2010).

2. D.R. Mould and K.R.D. Sweeney, Curr. Opin. in Drug Discov. Devel. 10, (1), 84–96 (2007).

3. Genentech, "Herceptin (trastuzumab) Intravenous Infusion, Full Prescribing Information" (South San Francisco, CA), www.gene.com/gene/products/information/oncology/herceptin/, accessed Oct. 10, 2011.

4. G.D. Lewis Phillips et al., Cancer Res. 68 (22), 9280–9290 (2008).

5. A. Beck, Discov. Med. 10 (53), 329–339 (2010).

6. I.E. Krop et al., J. Clinical Oncology 28 (16), 2698–2704 (2010).

7. H.A. Buris et al., et al. J. Clinical Oncology 29 (4), 398–405 (2011).

8. FDA, "FDA: Pfizer Voluntarily Withdraws Cancer Treatment Mylotarg from US Market," Press Release (Rockville, MD, June 21, 2010, www.fda.gov/NewsEvents/Newsroom/PressAnnouncements/ucm216448.htm, accessed Oct. 10, 2011.

9. EMA, Procedure No. EMEA/H/C/000705, EMEA/HMP/5130/2008 "Refusal Assessment Report for Mylotarg"(London, 2008).

10. P. Van Arnum, Pharm. Technol. 32 (6), 54–58 (2008).

11. A. Beck, T. Wurch, J.M. Reichert, Landes Biosci. 3 (2), 111–132 (2011).

12. A. Beck, P. Senter, and R. Chari, Landes Biosci. 3 (4) 331–337 (2011).

13. FDA, "Points to Consider in the Manufacture and Testing of Monoclonal Antibody Products for Human Use" (Rockville, MD, Feb. 1997).

14. F. Ehmann, presentation at the GCC Workshop on Similar Biological Medicinal Products (Riyadh, Saudi Arabia, Apr. 19–20, 2011).

15. EMA, EMA/CHMP/BMWP/86289/2010, Guideline on Immunogenicity Assessment of Monoclonal Antibodies Intended for In Vivo Clinical Use (London, 2010).

16. Biophoenix, "Biosimilars, Biogenerics and Follow-on Biologics (Informa, UK 2007), www.scripintelligence.com/multimedia/archive/00000/BS1342_117a.pdf accessed Oct. 10, 2011.

17. A.G. Polson et al., Blood 110 (2), 616–623 (2007).

18. K. Anderson, presentation at the National Biotechnology Conference (San Francisco, 2011).

19. R.P. Lyon et al., presentation at the 101st Annual American Association for Cancer Research Meeting (Washington DC, 2010).

20. FDA, Guidance for Industry Safety Testing of Drug Metabolites (Rockville, MD, Feb. 2008).