OR WAIT null SECS
This article discusses a number of factors that may influence the behaviour of conjugated biopharmaceuticals. Optimizing bioconjugation processes may be critical to achieve the desired drug performance.
Despite a 100-year lag between the recognition of their potential usefulness as pharmaceuticals and the first FDA-licensed product, conjugated biopharmaceuticals are rapidly entering the mainstream.1,2 The approval of the chemotherapy agent Mylotarg (Wyeth/UCB-Celltech) in 2000 was followed in the next few years by the radioimmunoconjugates Zevalin (Biogen IDEC/Schering) and Bexxar (Corixa/GSK), and around 30 conjugated biopharmaceuticals are now believed to be in clinical trials. All the drugs mentioned here are examples of 'armed antibodies', which combine the targeting properties of immunoglobulins with a toxic payload for therapeutic purposes,3 and, therefore, closely correspond to Paul Ehrlich's 'magic bullet' concept.
In other examples, the in vivo properties of a pharmaceutical may be improved by conjugation to a second component, which may reduce immunogenicity or otherwise enhance bioavailability. For example, interferon for hepatitis treatment has been conjugated to polyethylene glycol (PEG) in Pegasys (Roche, FDA-licensed), and methotrexate-human serum albumin (HSA) conjugates are in trial for cancer treatment by the German Association for Medical Oncology.
The agents discussed above include examples of protein–protein, protein–small molecule and protein–polymer conjugates. For the first of these categories, it is possible to perform the conjugation reaction chemically, but a preferred approach is often to create a recombinant fusion protein, allowing the conjugate to be expressed and cultured as a single molecule. The fusion protein method is beyond the scope of this article, but for all other categories of conjugation there are a number of factors that may influence the pharmaceutical properties of the product. This article aims to explore some of these factors.
Variables that affect conjugate behaviour include:
Regarding the first of these, it is clear that the choice of the conjugation partners is critical, and in the immunoconjugate field, for example, recent progress has been made in antibody structure and the efficacy of radionuclides and toxins.3 However, this fascinating area falls outside the scope of this article. The remaining factors are discussed though.
There are two main choices to be made under this heading that are closely linked. The first is whether to aim for site-specific coupling where the link(s) between A and B are fully defined and identical in each conjugate molecule, or for random coupling where conjugation occurs via multiple sites of similar reactivity, and the resulting conjugate is, consequently, certain to be heterogeneous in nature. The second choice concerns the actual coupling chemistry itself: which reactive groups on A and B are to be exploited in the conjugation, and are they to be linked directly or via a 'third party' linker molecule?
These choices are linked because the site-specific approach can typically only be employed where the reactive groups are present in very small numbers on each molecule, often ruling out coupling via widespread residues such as lysines. Thus, site-specific coupling frequently exploits rarer amino acids, such as cysteine. Some of the most commonly encountered reactive pairings for conjugation are listed in Table 1. The choice between these may also be influenced by in vivo stability requirements (which will be discussed in more detail later).
Table 1 Examples of reactive pairings for bioconjugation.
The chemistry options may be strictly limited if the structure of the conjugation partners is limited to their native state, however, recombinant techniques may allow the modification of proteins to enhance their suitability for site-specific coupling — by deleting all but one cysteine group, for example, so that only a single thiol is available for reaction.
Generally, this mutation approach relies on substituting one naturally occurring amino acid for another; however, in a more radical variation on this theme, pairings of mutually reactive unnatural amino acids can be introduced giving highly controllable chemistry for protein–protein conjugation.4,5
Where one of the conjugation partners is a polymer or small molecule rather than a protein, there is generally some leeway to generate functionally active analogues with a variety of different reactive groups or handles, again helping to facilitate site-specific coupling. A typical output of such molecular engineering (Figure 1) might be the conjugation of a protein engineered to contain a single thiol to a synthetic toxin containing a single amine group that has been quantitatively converted to a thiol-reactive maleimide moiety.
Both site-specific and random conjugation approaches are commonly encountered in biopharmaceutical conjugate development because there are advantages and disadvantages associated with each approach (Table 2).
Table 2 Advantages (Ã¢ÂÂ) and disadvantages (X) of site-specific and random coupling approaches.
In general, regulatory demands are likely to drive the balance increasingly in favour of site-specific coupling, but provided adequate analytical and clinical data can be supplied to support the safety and efficacy of heterogeneous conjugates, random coupling is unlikely to die out in the near future.
The first characteristic encountered with respect to conjugate stoichiometry is often the mean incorporation; simplistically the mean A:B ratio in the conjugate. The optimum incorporation may be influenced by a number of factors, as exemplified for an antibody–toxin conjugate in Table 3.
Table 3 Advantages (Ã¢ÂÂ) and disadvantages (X) of high and low toxin incorporation in an antibodyâtoxin conjugate.
For site-specific conjugation, the mean incorporation may well adequately characterize the composition because of the homogeneous nature of the product. For random conjugation via multiple lysines, for example, this mean value must be interpreted in terms of the likely distribution of the conjugate subpopulations. Using fairly simple statistics it is possible to model such distributions assuming large numbers of independently and equally reactive groups (not an unreasonable assumption for large proteins such as antibodies containing many lysine residues).
Using the Poisson distribution, the populations shown in Figure 2a are predicted for such a reaction where the mean A:B incorporation is 1.0;6 note the significant quantities of unreacted A and of conjugates with B:A ratios up to 4:1. Studies have been performed on the stoichiometric distribution of real conjugates, and these can yield distributions similar to theory (Figure 2b).6,7 However, this is not always the case: the assumption of equal reactivity may not always be justified. The first FDA-licensed immunotherapy conjugate Mylotarg has a highly non-Poisson distribution,8 the mean incorporation of 2–3 disguises two distinct populations with incorporations of 0 and 4–6.
In summary, therefore, it should be noted that unless site-specific chemistry assures a homogeneous stoichiometric distribution, mean incorporation values give only a partial understanding of the true situation.
The factors discussed so far address the key in vitro properties of the conjugate, but a critical design issue for pharmaceutical application is how the conjugate behaves in vivo. In the simplest scenario, the conjugate may only be required to remain intact for as long as possible in the body — this may be the case for untargeted examples such as some of the PEG- and HSA-conjugated drugs discussed previously. However, the rate of clearance from the body (typically via the liver and kidneys) is enhanced among other factors by low molecular weight, so once the drug has achieved its purpose, dissociation into smaller fragments may be advantageous.9
For an antibody–toxin conjugate for cancer treatment, the optimum behaviour is very different again. The toxicity of many payload compounds is often conveniently diminished by chemical derivatization, opening up the elegant possibility of delivering the drug to the tumour site in a nontoxic (conjugated) form, then triggering the release of the active drug in situ by cleavage of the antibody–toxin linkage. Thus, immunoconjugate development often encompasses the following design goals:
Targeted drug therapies from an inactive drug derivative are illustrated in Figure 3. The result of such a strategy is to yield a drug that is only activated at the desired site of action in the body. The success of this approach depends principally on how much the activity of the drug is reduced by derivatization (to form a prodrug), and how selectively the derivatization can be reversed at the target site compared with the rest of the body.
In one form of this approach, the prodrug itself is not targeted at the site of interest and is distributed evenly throughout the body, but its cleavage is accomplished by a second, target-specific agent. Thus, a derivatized drug that can be reactivated by a nonnative enzyme can be employed combined with a conjugate of the enzyme with a tumour-targeting antibody — such as the antibody-directed enzyme prodrug therapy (ADEPT) — or with gene therapy that causes the enzyme to be expressed selectively at the tumour site, such as gene-directed enzyme prodrug therapy (GDEPT).10,11
Although these approaches are particularly elegant, they require composite treatments, and while promising trials are under way, they have yet to make an impact on the market place. A simpler adaptation of the approach is to rely on inherent characteristics of the target site, which do not require a second, activating agent. It is difficult to find such characteristics that are 100% selective for the target site, but it may be enough to rely on such a factor just being enhanced in this location. Thus, for example, it is known that β-glucuronidase enzymes are selectively found at high concentrations in tumours, therefore, prodrugs formed by synthesizing the inactive β-glucuronide of a toxin will have a greater tendency to be activated at the target site than elsewhere in the body.12 Other examples rely on chemical characteristics of the target site, such as a lower pH or more reducing environment than the body in general, and some of these are noted in Table 1.
Without belittling the excellent work that has been put into developing these self-activating magic bullets, it is worthwhile sounding a note of caution. The in vivo processing of conjugate molecules can be complex and unpredictable, and the understanding of such effects is really only in its infancy. Thus, active toxins can be released successfully from conjugates that do not on paper appear to offer the potential for in situ cleavability: in a recent example13 an antibody–prodrug conjugate linked via a succinimidyl thioether linkage gave unexpectedly good in vivo results — the authors hypothesizing that the antibody itself had been broken down at the target site in such a way as to free the active drug in a cysteine-linked form, despite the absence of a deliberately cleavable linker. It should be stressed that such surprises do not always happen — in the same study other 'noncleavable' conjugates stayed true to form in giving poor results — but it is worth bearing in mind that in vivo behaviour does not always respect the design intentions of the developers!
The following factors also affect conjugate behaviour:
In vitro stability. The in vitro stability of the A–B link has to be considered in tandem with the in vivo properties just discussed. Some linkages such as disulfides and hydrazones have a poorer reputation for stability than others, such as amides or thioethers. However, it should be noted that even in the latter category complications can occur. While thioethers are not prone to cleavage of the A–B bond, ring-opening of the maleimide-derived examples commonly encountered can sometimes yield a heterogeneous mixture where the linkage contains both succinimide and succinamic acid linkages (Figure 4).14
Immunogenicity. The immunogenicity of the linker group can be significant, and among heterobifunctional reagents for maleimide derivatization of amines, those containing simple aliphatic structures are generally less immunogenic than examples containing bulky groups such as cyclohexyl rings.
Length. The length of the linker group is often interpreted in terms of a long span providing a desirable 'spacer' effect between A and B. However, if the linker is flexible this is unlikely to result in a real separation of the two molecules, particularly where there is a tendency for the two partners to aggregate. Thus, linkers containing bulky groups that confer some 'rigidity' to the bridge may be more successful.
Hydrophobicity. Hydrophobic linker groups might be expected to exacerbate hydrophobic interactions between conjugation partners, fuelling the tendency for aggregation, while more hydrophilic examples may have the opposite effect.12
Incomplete reaction. Even if site-specific chemistry is chosen (as described earlier) an incomplete reaction can lead to the presence of monomeric A or B or linker molecules in the product, which can be challenging for subsequent purification and characterization steps. Conjugation by-products can also be problematic: it has been noted above that ring-opened succinimidyl-thioethers can lead to heterogeneity in thiol-maleimide products, while succinimide ester coupling targeted at lysine residues has the potential to form less stable conjugates via tyrosine or histidine residues.15 It is always simpler to optimize the conjugation chemistry to avoid such by-products than it is to purify them out after the reaction.
There are many factors that must be considered to optimize conjugation procedures for biopharmaceutical applications. While the choice of the molecules to be conjugated often receives the greatest attention, it is essential that due consideration is also given to the site of conjugation on each molecule, and to the likely in vivo and in vitro properties of the linkage between them.
Alastair H. Dent is managing director at Fleet Bioprocessing Ltd (UK). He has spent over 20 years in the development and optimization of protein conjugation processes for pharmaceutical and diagnostic applications, including senior technical roles for Amersham International, Kodak and Johnson & Johnson. Alastair has authored and reviewed numerous publications in this field, and is a member of the UK Royal Society of Chemistry's Analytical Biosciences Committee.
1. A.K. Pavlou and M.J. Belsey, Eur. J. Pharm. Biopharm., 59(3), 389–396 (2005).
2. A.H. Dent, Contract Services Europe, September, 29–32 (2006).
3. A.M. Wu and P.D. Senter, Nature Biotechnol., 23(9), 1137–1146 (2005).
4. Anon., Science, 308(5718), 44 (2005).
6. M. Aslam and A. Dent, Eds., Bioconjugation: Protein Coupling Techniques for the Biomedical Sciences (Macmillan Reference Ltd, London, UK, 1998) pp 92–93.
7. E. åkerblom et al., Bioconjugate Chem., 4, 455–466 (1993).
8. P.F. Bross et al., Clin. Cancer Res., 7(6), 1490–1496 (2001).
9. D.L. Kukis et al., Cancer Biother. Radiopharm., 16(6), 457–467 (2001).
10. K.D. Bagshawe et al., Expert Opin. Biol. Ther., 4(11), 1777–1789 (2004).
11. H.K. Han and G.L. Amidon, AAPS Pharmsci., 2(1), 1–11 (2000).
12. S.C. Jeffrey et al., Bioconjugate Chem., 17, 831–840 (2006).
13. S.O. Doronina et al., Bioconjugate Chem., 17, 114–124 (2006).
14. Nektar Therapeutics AL, Corp., World Patent WO 2004/06095 (2004).
15. M. Aslam and A. Dent, Eds. Bioconjugation: Protein Coupling Techniques for the Biomedical Sciences (Macmillan Reference Ltd, London, 1998) pp 74–76.