Answer by Tirumalai Kamala:
Antibody-drug-conjugate (ADC) technology has several limitations that become clear by looking at its development history and component parts.
What are ADCs?
An ADC is a monoclonal antibody (mAb; more commonly mouse, recently some human) + a linker (to attach[conjugate]) + a small molecule cytotoxic drug (SMD; older active drug, newer prodrug).
This web-site that maintains a comprehensive and updated list of therapeutic antibodies and ADCs:
What’s the idea behind the ADC? A century back Paul Ehrlich coined the phrase magic bullet in the following passage (2):
“If we picture an organism as infected by a certain species of bacterium, it will obviously be easy to effect a cure if substances have been discovered which have an exclusive affinity for these bacteria and act deleteriously or lethally on these alone, while at the same time they possess no affinity for the normal constituents of the body and can therefore have the least harmful, or other, effect on that body. Such substances would then be able to exert their full action exclusively on the parasite harboured within the organism and would represent, so to speak, magic bullets, which seek their target of their own accord”.
Substitute tumor cell for Ehrlich’s bacterium and the ADC becomes our synthetic oncological exemplar of his magic bullet, in the hope that the ADC’s mAb specifically delivers its deadly prodrug/drug partner to a tumor which then chemically and/or enzymatically converts it to toxicity, killing itself in the process.
A viable ADC takes a lot of work (3, 4). Important considerations include
- Is the ADC stable in blood? If it isn’t, then the toxic drug could be released in the bloodstream, causing systemic toxicity.
- Does the ADC’s mAb bind its target antigen in the same manner and with the same affinity? In other words, the linker should neither structurally nor functionally modify the antigen-binding capacity of the mAb.
- Does linkage compromise the cytotoxic drug’s mechanism of action? There are two main classes of ADC drugs (1, 3, 5), the auristatines and maytansines that bind to microtubules leading to G2/M-phase cell cycle arrest and apoptosis, and the calicheamicins that target DNA.
- Is sufficient amount of ADC taken up by the target cells? This requires sufficient number of antibody targets (antigens) expressed on the cell surface. A sufficient number will differ for each antigen and cell type.
- Once taken up by the tumor, is sufficient amount of SMD released from the ADC to kill it?
- How well does the mAb penetrate solid tissue? Antibodies are notoriously poor at doing so. Add to that the counter-intuitive results that antibody affinity and tumor mass penetration are inversely related (6, 7), and we have a pretty uphill task since most mAbs in the commercial porfolio were originally developed for high, not low, affinity.
Antibodies have a wide variety of effector functions that make them a versatile tool for treating cancers and autoimmune diseases.
- They neutralize toxins (8, 9, 10).
- They neutralize cytokines (11).
- They block receptors (12).
- They bind cell-surface receptors to drive specific immunity (13) such as Antibody-Dependent Cellular Cytotoxicity (ADCC).
- Combination of above (14).
Regardless these multiple effector functions, in an ADC, the primary job of the mAb is to simply specifically target a cell-specific antigen, i.e. function as a Targeted Drug Delivery vehicle for the toxic drug, and little else. Some drawbacks of ADC mAbs in that regard?
- Most candidate mAbs were originally developed for one or more of the effector functions listed above but are now being co-opted as mere delivery vehicles. Maybe some of those effector functions get in the way of simple delivery (15)? In this context getting in the way implies they engage in their effector functions and compromise patient safety. Choice of mAb isotype is thus key. In hindsight, not surprising that most common mAb format is IgG1, an isotype very good at engaging ADCC. After all, mAbs were developed for their potent effector functions, not their prowess as vehicles for Targeted Drug Delivery.
- Most of these mAbs have high or very high affinity but decades later, research suggests mAbs with lower affinities are counter-intuitively better at penetrating solid tissues such as tumors (6, 7).
- Most of the current portfolio of more than 160 therapeutic mAb candidates in clinical trials for cancer alone (16) were developed decades back using outdated approaches, primarily mouse models, and then modified for human use by a process called humanization. Obviously considerable modification is necessary to prevent human immune responses against such mAbs.
There are two general processes used to modify mAbs for human use.
Replace as much as possible of the mouse sequence of the antibody with its human counterpart while retaining its original binding property.
Identify and replace the T cell epitopes of the antibody.
Both methods are fallible, laborious, and require painstaking effort.
What does mAb humanization do?
- Reduces human anti-mouse antibody responses. Obviously good for an ADC mAb.
- Increases blood half-life of the mAb.Obviously good for an ADC mAb.
- Improves ability of mAb to elicit immunity. For an ADC mAb that should serve as nothing more than a reliable Targeted Drug Delivery vehicle, not so good. For example, mouse mAbs converted to human IgGs are retained longer in blood circulation through binding to the human neonatal Fc receptor (FcRn), which also enhances their ability to elicit ADCC (4).
Though newer therapeutic mAbs are generated using newer technologies like phage display libraries or humanized mice (transgenic mice genetically engineered to produce human, not mouse, antibodies) such as Cambridge Antibody Technology Ltd’s Mapatumumab for refractory colorectal cancer (17), humanization alone is not sufficient to ensure optimal ADC function nor does it eliminate unwanted immune responses against the ADC itself. An ADC is a therapeutic biopharmaceutical designed not to drive an immune response so much as to deliver a death cargo to a specific target. Thus, inducing immune responses against itself negates the ADC’s usefulness. ADCs could acquire unwanted and undesirable immunogenic attributes, i.e. the capacity to induce an immune response, during culture and manufacturing processes. These include attributes such as glycosylation and aggregation, which only come to light empirically, i.e. through trial and error.
A common and complicated post-translational modification, glycosylation enzymatically adds glycans to proteins.
- Glycosylation influences both the physico-chemical (electrical charge, mass, stability, structure, size, solubility) and biological (activity, half-life, receptor function) properties of a protein such as a mAb.a) Proper glycosylation is critical for the antibody’s effector function. For example, engineered IgG antibodies lacking fucosylated oligosaccharides had enhanced ADCC both in vitro and in vivo (18), an attribute desirable for a therapeutic mAb, not so much for an ADC mAb.
b) Glycosylation is species-specific.
- Humans make immune responses to non-human glycosylation such as galactose- α1, 3-galactose (α-Gal), N-glycolylneuraminic acid (Neu5Gc) [common in pigs and mice], and beta1, 2-xylose (core-xylose), and α1, 3-fucose (core-α1, 3-fucose) [common in plants]. This includes circulating IgG antibodies implying full-fledged immunity involving T and B cell activation followed by T cell help for B cells driving antibody class-switch from IgM to IgG. Cell lines and culture conditions used for protein synthesis influence its glycosylation pattern (19). For example, cetuximab is a chimeric mouse-human IgG1 mAb approved for colorectal and squamous cell head and neck cancers. The mouse cell line SP2/0 used for producing cetuximab expresses the gene encoding α1,3-galactosyltransferase, the enzyme responsible for the synthesis of the α-Gal epitope. Controversially, Chinese Hamster ovary (CHO) cells might not synthesize the α-Gal epitope (20, 21). Apart from mammalian cell lines used to manufacture mAbs, culture medium itself could provide the immunogenic sugars if it contained animal-derived materials since the cell could metabolically incorporate such glycoepitopes into the secreted mAb (22).
Modified from 19.
Conformational changes leading to aggregation is a key process that could confer immunogenicity on biopharmaceuticals like ADCs.
While glycosylation and aggregation are drawbacks common to other biopharmaceuticals as well, additional drawbacks of ADCs include lack of stringently defined tumor-specific antigens, poor capacity for being internalized by tumors, and extremely complicated pharmacology including PK (Pharmacokinetics)/PD (Pharmacodynamics) and ADME (Absorption, Distribution, Metabolism, Excretion).
What’s the target of the mAb in the ADC? Best-case scenario is where the target is expressed by a tumor cell alone, and not by any other normal cell in the body. Such tumor-specific expression is extremely rare, especially across patients. Moreover, of the >160 therapeutic mAbs either being tested or approved, less than a handful are stringently tumor-specific. The rare examples are B-cell lymphoma (anti-CD20; 23), human epidermal growth factor receptor 2 (HER-2 or Erb-B2) targeted by Trastuzumab (24), CD30 for Hodgkin’s lymphoma and MUC16 for ovarian cancer (1). This highlights one of the most important drawbacks of the current portfolio of mAbs used in ADCs. With few exceptions, most of these were generated years or even decades back, when target identification methods were much more primitive, vastly less sensitive and overly reliant on mouse and other animal models, leaving open-ended whether a mAb target is strictly tumor-specific (24). Even now, how is antigen expression assessed and on what? In situ patient sample immunohistochemistry would be quite optimal but in practice, even now most commonly, assessment involves cell lines and xenograft tumor models (1). With cell lines, culture conditions introduce yet another huge artifact. Such lack of specificity about the target antigen opens the door to a questionable length of therapeutic window (25), i.e. when there is maximal differential expression of the target antigen between tumor and normal tissue (25, 26, 27) to maximize the therapeutic index (maximum tolerated dose/minimum efficacious dose).
How efficiently are the mAbs used in ADCs taken up by tumors? After all, these mAbs were developed for immune effector functions, and not for being internalized. It is more by happenstance rather than by design that some such as Trastuzumab induce uptake (28).
Extremely complicated to say the least. Why? The type of mAb heavy chain (isotype), glycosylation, route of administration, affinity, location and number of surface antigens (mAb targets), immunogenicity of the ADC, circulating ADC stability endowed by the linker are just a few of the factors that influence the ADME and PK/PD of a typical ADC (29, 30, 31, 32).
How easy is it to modify an existing antibody to attach a conjugate payload?
As the history of chimeric and humanized mAbs shows, existing mAbs can be modified considerably. Rather, even though advances in ADC Linker technology has greatly improved stability, the crux of the ADC synthesis problem is ensuring a reproducibly homogeneous mix of mAb-drug, i.e. same number of drug molecules per mAb.
Early ADCs had unstable linkers with half-lives as short as 1 to 2 days, examples being disulfides (34, 35, 36) and hydrazones (37, 38). New linkers with better stability and longer half-lives include peptide linkers (39, 40) and glucuronides (41). Broadly, linkers are either cleavable or non-cleavable (42).
Cleavable linkers depend on three types of intracellular processes for release, namely, cleavage of linker dipeptide bond by lysosomal proteases such as cathepsin B (e.g., Adcetris), hydrolysis of linker acid labile group such as hydrazone within the low acidic pH of the lysosome (e.g., Mylotarg) or reduction of linker disulfide bonds by the higher intracellular concentration of glutathione (e.g., SAR3419, IMGN901, AVE9633). With non-cleavable linkers, the antibody is proteolytically cleaved to release the toxin and its attached linker (e.g., Kadcyla) (43, 44). Non-cleavable linkers are more stable in circulation compared to cleavable ones (1).
How to decide which linker to use? Preclinical in vitro and in vivo (mostly mouse model) studies show how an ADC is taken up and degraded. That information plus information about the target cell drives the choice of linker. While the field has moved from the problematic random to the more controlled site-specific conjugation, there is no one-size-fits-all. Each mAb, each tumor requires optimization of a specific delivery system (3). This adds to overall cost.
How many drug molecules attached per mAb makes for an optimal ADC? More attached, faster the ADC is cleared by the body. Two to four drugs per mAb work better (45). Most linkers bind via cysteine or lysine residues on the mAb. A typical antibody has 80 lysines and 20 cysteines. That’s a lot of potential conjugation sites! Notwithstanding tweaks such as adding additional cysteine residues (46), replacing solvent accessible cysteines with serines (47) or adding non-natural amino acids (nnAA) as specific conjugation sites (48), reproducibly loading same number of drug per mAb has proven elusive (45). Which site on a mAb yields the most homogeneously conjugated ADC? This too remains unanswered thus far for any ADC.
Another puzzling aspect of mAbs is their poor efficacy (16). For example, the recently approved Kadcyla requires 160mg per dose (33). Why such high amounts? Because only 1 to 2% of the administrated dose of a given ADC is estimated to reach the tumor site (49)! Obviously, rather than epitomize more bang for the buck, ADCs do the opposite. Are routes (typically subcutaneous or intravenous) of administration to blame or is it something else entirely? In the case of solid tumors, should we try to inject them directly into the tumor, for example?
Other manufacturing issues
Approved ADCs require mammalian expression and bioreactor systems with very expensive upstream and downstream production processes. Open cell-free protein expression systems would go a long way in mitigating the many drawbacks of mammalian cell expression systems for mAb manufacture (50, 51, 52) though this is very much virgin territory requiring much optimization to reduce immunogenicity, and likely years away from commercialization and regulatory approval.
What about the SMD part of the ADC? Cytotoxic drugs used in ADCs are 100 to 1000 times more cytotoxic than traditional anti-cancer drugs and were in fact too toxic as stand-alones but became workable as ADC payloads (1, 3, 5). This greatly enhanced cytotoxicity hugely adds to the burden of designing appropriate manufacturing processes and systems to specifically mitigate and minimize occupational exposure during manufacture.
As an immunologist, what’s my bottom-line about ADCs and similar biopharmaceuticals? Engineering one’s way out of unwanted immune responses is easier said than done.
- Ehrlich, Paul. Address delivered at the Dedication of the Georg-Speyer-Haus. In: Himmelweit, Fred, ed. The collected papers of Paul Ehrlich. London: Pergamon Press; 1960 (1906), p. 53-63 (59).
- Goldmacher, Victor S., and Yelena V. Kovtun. “Antibody-drug conjugates: using monoclonal antibodies for delivery of cytotoxic payloads to cancer cells.” Therapeutic delivery 2.3 (2011): 397-416.
- van Beers, Miranda, and Muriel Bardor. “Minimizing immunogenicity of biopharmaceuticals by controlling critical quality attributes of proteins.” Biotechnology journal 7.12 (2012): 1473-1484.
- Jefferis, Royston. “Recombinant antibody therapeutics: the impact of glycosylation on mechanisms of action.” Trends in pharmacological sciences 30.7 (2009): 356-362.
- Miller, Richard A., et al. “Treatment of B-cell lymphoma with monoclonal anti-idiotype antibody.” The New England journal of medicine 306.9 (1982): 517-522.
- Junutula, Jagath R., et al. “Site-specific conjugation of a cytotoxic drug to an antibody improves the therapeutic index.” Nature biotechnology 26.8 (2008): 925-932.
- Swartz, James R. “Universal cell-free protein synthesis.” Nature biotechnology 27.8 (2009): 731-732