On target with antibody drug conjugates

With the recent approvals of Adcetris (brentuximab vedotin) and Kadcyla (ado-trastuzumab emtansine), antibody drug conjugates (ADCs) represent a growing area in oncology with great potential. With the success of Kadcyla against solid tumours comes the real promise of ADCs in tackling the most aggressive and resistant tumours with the worst clinical outcomes. Current targets include difficult-to treat cancers such as triple-negative breast cancer, gastric tumours, small-cell lung cancer, melanoma and pancreatic cancer. More than 30 other ADC candidates are currently in trials for about 25 different targets, with many more programs in development. The number of ADC candidates in the clinic has doubled since 2010 and further growth is anticipated.

Now that the therapeutic potential of the class is finally being realised, ADCs have become a hot field in research and development, with most major pharmaceutical companies either already established in the area or looking to get involved. Many collaborative deals have been announced, including the recent deal between Pfizer and CytomX, worth a potential $635M.

The rapid growth in ADC technology is being driven by the flexibility of the platform. Conjugation is applicable to all antibody technology types, including full-length monoclonals, antibody fragments and bispecifics. Both antibodies and cytotoxics that have previously failed in the clinic can even be given a new lease of life as ADCs. In addition, their potential is not just limited to oncology– they are also being investigated for indications such as inflammation.

Market potential
The long term future of ADC programs is difficult to predict, but their potential to extend the on-patent life of existing products and maintain market position is attractive to pharma companies. Kadcyla is a good example – it extends the potential of Herceptin as the monoclonal antibody itself approaches the end of its patent protection. New ADC products like Kadcyla, with their increased complexity, will offer oncology companies new ways to discourage biosimilar competition in the future.

Most analysts predict that ADCs will become a major product grouping within the oncology market, but the main obstacle to widespread uptake is their extremely high price – Kadcyla costs more than twice as much as Herceptin, and both the currently available ADC products are priced at around $100,000 per patient treatment regimen. Although Kadcyla does show patient benefits over the current best treatment option – extending median life expectancy by six months while reducing serious adverse events by about a third – at a time of squeezed health budgets, achieving high patient uptake is likely to be challenging. Nonetheless, analysts are predicting a market size anywhere between $2 billion and $5 billion for Kadcyla, compared to the total annual sales of $6 billion for Herceptin. It is likely to be the largest ADC product for the foreseeable future, on account of the large existing market for Herceptin.

The majority of the ADCs that are set to reach the market in the next few years are for lymphoma indication. They are likely to be treatment options for smaller patient groups, and there is greater opportunity for competition. The sales figures for Adcetris thus far give a greater insight into the potential future sales performance of ADCs. It targets a much smaller patient population than Kadcyla, but the benefits are much more significant. However, the very high per-patient cost has led to slower uptake than had been predicted, and it has failed to live up to market expectations. Recent quarterly sales of $34.5 million, Jan-Mar 2012, were somewhat short of the anticipated $39 million. Regardless, Adcetris is still predicted to be a big seller, and sales of $1 billion a year might even be achieved. The main driver for this growth is an expanding pool of patients who might benefit from the drug, in the light of ongoing clinical trials that are investigating earlier use, and its potential in a wider range of cancers.

The full potential of the ADC market will become clearer in the next few years, as additional products progress through late-stage clinical trials. ADCs are expected to make up more than 10% of revenues for the total antibody market, approximately $10 billion by the end of 2023, according to a recent market analysis report from ADC Review: Journal of Antibody Drug Conjugates. The number of IND submissions for ADC products is accelerating- more than 30 from 2008 to mid-2012, compared to 15 between 1993 and 2007. The attrition rate remains relatively low and, given the current pipeline, multiple ADC products might be expected to reach the market by 2020. Key programmes to watch will be those in difficult-to-treat solid tumours, and success here would likely generate multi-billion dollar drugs.

Targeted Therapies
The success of these ADC therapies comes from a combination of a highly potent toxin with the ability of a monoclonal antibody to target malignant cells. The benefit of targeted delivery of therapeutics was predicted long before it was possible to achieve. Early experience in the oncology field demonstrated the benefit of chemotherapy in treating metastatic disease, but the lack of specificity of the treatment severely limits therapeutic success. Greater discrimination between malignant and healthy tissue is required but, to date, this was difficult as the therapies used were, essentially, non-targeted.

The innate specificity of antibodies was recognised early on and, in combination with the identification of tumour specific antigens, this opened the door to the possibility that tumour cells might be selectively targeted and destroyed. The concept is simple, but the results obtained were, at first, disappointing. Big advances were required for the industrial production of antibodies, and to overcome the first wave of setbacks from issues such as immunogenicity. However, the success rate remained low with the first monoclonal antibody therapies, especially in solid tumours. There was huge excitement surrounding these early targeted antibody therapies in cancer, and there were a few notable successes. But it did not transform the outlook for patients which, in all too many cases, remained poor.

The lack of success of monoclonal antibodies in cancer treatment did not result from lack of specificity, but more from a lack of activity. They could be shown to reach and bind to receptors on the tumour cells, but they did not seem to recruit the immune system to destroy the tumour as had been anticipated. In most cases, the antibody had little or no effect, and the treatments only showed significant value as a component of a multiagent chemotherapy regime. Outcomes improved a little, but side-effects still limited treatment because of the non-specific activity of the other components of the chemotherapy regimen.

The components required for a potential ‘magic bullet’ therapy were now coming together. Advances in chemotherapy, radiotherapy and monoclonal antibodies sparked a further round of enthusiasm, as many groups looked at combining cytotoxic chemotherapy agents with monoclonal antibodies. Two main approaches were attempted. First, existing chemotherapy agents such as anthracyclins and vinca alkaloids, with known clinical benefits, were conjugated to monoclonal antibodies. The second looked to direct radioisotopes to the tumour via the conjugation of radioisotopes, or chelating groups that bind them, directly to the antibody.

As before, the early optimism faded as the first antibody conjugates hit the clinic. Although exciting data had been seen in preclinical models, all too often the results in patients did not meet expectations. Some even showed a lower clinical efficacy than the untargeted cytotoxic. Radioisotope conjugation showed some success but, again, the results were far from the breakthrough that had been hoped for. Early successes were limited to haematological cancers, with solid tumours proving remarkably resistant. Three programmes had some success in the clinic: one antibody drug conjugate (Mylotarg – gentuzumab ozogamicin) and two radioconjugates (Bexxar- – tositumomab and iodine 1131 tositumomab, and Zevalin – ibritumomab tiuxetan conjugated with yttrium 90). Mylotarg’s s ozogamacin payload had never been used as an unconjugated cytotoxic because of its extremely high potency. The drug was voluntarily to its ozogamicin payload. This drug was never used as a single agent due to its extremely high potency. Although Mylotarg was approved in 2000, it was voluntarily withdrawn from the market in 2010 after little evidence of clinical benefit in the acute myloid leukaemia it was licensed to treat.

Significant side-effects were also seen with all three of these products as, despite the targeting activity of the antibody, off-target damage was still occurring at doselimiting levels. Mylotarg has serious liver toxicity, while the radioimmunoconjugates irradiate healthy tissue around the cancers – and elsewhere in body as they circulate in the bloodstream. The beta and gamma emitting isotopes travel far from the point of origin, leading to wide-ranging irradiation of healthy tissue.

Research into the early failed conjugates determined that access to the tumour was a significant factor limiting the therapeutic effect. A study suggested that only about 0.01% the conjugate actually reached the tumour and was internalised. This explains the failure of the early conjugates using existing chemotherapy agents; the cytotoxic agents were simply not active enough to kill the target cell with such a small amount of payload entering the cell.

Improving drug properties
The early experience with ADCs quickly confirmed that the technology was going to be significantly more complicated than first expected.

Further progress would need to be made in antibody selection, the payload and the linking chemistry to join the two elements together. Target selection would also be key for a successful conjugate. A suitable antigen target would be significantly overexpressed on a cancer cell compared to healthy cells. It also has to be expressed on the cell surface at points that are accessible to the antibody. For an ADC, the antigen needs to be internalised, although this is not a requirement for a radioimmunoconjugate. Not many antigens meet these criteria, and this is a key area for research.

One antigen that does meet these requirements is HER2. It is substantially overexpressed on many breast cancers to a level significantly higher than healthy tissue. It is internalised on binding by an antibody and trafficked to the lysosome for degradation allowing payload release. Experience with Herceptin (trastuzumab), which targets HER2, highlights the potential of this antigen as a target for ADCs.

The choice of antibody has to complement the antigen. All antibodies against an antigen are not equal, and some will have much more preferable characteristics as therapeutic agents. The binding affinity, epitope bound on the antigen and the antibody class all have an impact on the suitability of the antibody for use as an ADC. Currently IgG1, IgG2 and IgG4 isotypes are used, as well as antibody fragments and antibody fusion proteins. There are advantages for each approach. Full length monoclonal antibodies are still the most common. They have long circulating half-lives, and can retain full antibody functionality, which may enhance potency. The size of the fully intact antibody may count against it, however, when it comes to penetrating into a solid tumour. Small molecules are more likely to enter the tumour, which may favour smaller antibody fragments. However, this must be balanced against renal clearance, which will shorten circulating half life dramatically. Bifunctional and other engineered antibodies potentially offer greater functionality and specificity. The science behind selecting the best antibody is still in its early days, but this is an active area of research, and is sure to become better understood as more successful ADCs are identified. Research into the early failed conjugates determined that access to the tumour was a significant factor limiting the therapeutic effect. A study suggested that only about 0.01% the conjugate actually reached the tumour and was internalised. This explains the failure of the early conjugates using existing chemotherapy agents; the cytotoxic agents were simply not active enough to kill the target cell with such a small amount of payload entering the cell.

Linker choices
The linker is an important element of the ADC, and its importance was underestimated in the early products. Initially, most researchers assumed the linker would have to be cleavable to release the active free drug. The challenge thus became finding a linker that could be cleaved inside the cancer cell, but was stable in the circulation to prevent the highly potent drug from being released in the circulation. Three main approaches came from the lysosome, which would be the final destination of the ADC. First, the endosome has lower pH, and acid labile linkers could be useful. Hydrazone chemistry was used as this is relatively stable at physiological pH, but unstable at the lower pH of the endosome. Secondly, the lysosome has proteases such as cathepsins, which could digest peptide linkers. Peptide linkers that contain a target cut sequence for cathepsins were tested. Finally, the lysosome was thought to be a reducing in nature, so disulfide bond-containing linkers were investigated. A variety of cleavable linkers was identified that exhibited good stability and activity.

Surprisingly, non-cleavable linkers also showed promise, as they were active for some drugs when the full antibody was digested, liberating the cysteine/lysine conjugate of the drug linker that retained activity. Both cleavable and non-cleavable drug linkers are active in cancer cells, but show considerable differences in efficacy. This is often specific to the individual cancer cell line. One observation was that cleavable drug linkers could release an active free drug that killed the target cancer cells, but also antigen negative cells nearby. This effect was termed bystander killing, and while it is useful for tumours with heterogeneous target antigen expression, it comes at the cost of more off-target toxicity. Non-cleavable drug linkers tend to generate free drug with a charged lysine or cysteine amino acid attached. This prevents passive uptake into neighbouring cells, and prevents bystander and other off-target toxicity.

The linker can also provide other benefits. For some drugs, aggregation is a big problem. Many highly potent drugs used in ADCs are very hydrophobic and, consequently, can promote denaturation of the antibody during conjugation. Early products used un-optimised linkers that did little to space the drug the optimal distance away from the antibody, or introduce a more polar structure to the overall drug linker to add solubility. Recent developments have used linkers based on polyethylene glycol and other biologically compatible polymers to improve solubility and drug loading. Increasing the polarity of the drug linker also has a secondary effect, in that it may make the drug linker less of a suitable substrate for multidrug resistant proteins that pump the toxins out of the cell. This is now a key area of research.

Forming the links
In terms of the linker chemistry, two main conjugation routes have been used. Most antibodies have no free cysteine groups, but these can be generated by the partial reduction of inter-chain disulfide bonds. This provides reactive thiol groups that are easy to conjugate using a variety of different chemistries. This method creates pairs of free thiols, as each disulfide bond is reduced. Different combinations of disulfide bonds are reduced in the reaction, and a variety of different species is created. For an IgG1, conjugates containing 0, 2, 4, 6 or 8 highly potent drug molecules are produced, and the average drug load is adjusted by titrating the molar ratio of reducing agent to the antibody. The resulting conjugate is stable despite the reduction and conjugation of the interchain disulphide bonds, but stability remains a concern after these key structural groups have been broken. Traditionally, maleimides have been reacted with the thiol groups. Exchange reactions have been reported with other thiols, but these may result in off-target toxicity.

The second route is via reaction with lysine residues on the antibody. In this process, the antibody’s disulfide structure is retained, but it can alter the overall charge of the antibody, depending on the chemistry used. The main issue with lysine chemistry is the large number (>25 for IgG1) of conjugatable lysine residues within the constant and variable regions of the antibody. For most antibodies with an average load of 3 to 4, there will be species ranging from 0 to 8 drug loading, with the drugs distributed between any of the conjugatable lysines. As a result, the product may be very heterogeneous product, posing a huge analytical challenge.

For both cysteine- and lysine-linked conjugates, it is likely that higher loaded species, while more potent, are more likely to be cleared from the circulation and cause off-target toxicity. Unconjugated antibodies may act as a competitive inhibitor, while conjugates with a low loading may not be sufficiently potent. Both cysteine- and lysine-linked conjugates have recently been commercialised, proving that either system can result in successful ADCs.

The drawbacks associated with the traditional chemistries have resulted in major research programmes looking at site-specific conjugation chemistries, and improved chemical conjugation techniques. Groups researching improved chemical conjugation techniques are developing chemistries that are more robust than those using maleimides, and which are capable of re-bridging the disulfide bond after conjugation. This can also generate more controlled drug-to-antibody ratios, and give a maximum drug load of 4 for an IgG1. This type of technology has an advantage that the antibody does not need to be reengineered.

However, other site-specific technologies do require the antibody to be engineered. The first approach was to engineer additional cysteines into the antibody. As with many aspects of ADC development, this looked to be a simple approach, with the introduction of a specific residue for conjugation, but it proved much more difficult than anticipated. The engineered cysteine residues were found to be capped, and very little free thiol was present. While the free thiols could be regenerated by reduction, a selective reoxidation step was necessary to re-form the disulfide bonds while leaving the engineered thiols free for conjugation.

Although this is possible, it requires the engineered cysteine groups to be carefully positioned in the antibody to prevent crosslinking and disulfide scrambling. A different approach involves using engineered cell lines to insert non-standard amino acids into the antibody. These can provide the basis for more specific conjugation chemistries, such as copper-free click chemistry. Peptide sequences recognised by particular enzymes can also be used for enzymatic conjugations. A wide range of site-specific conjugations is being developed, and it is likely that this will become the next standard. Site-specific conjugates are now in clinical trials, and the first successful site-specific conjugate will likely set a new standard for the development of conjugates.

Toxin selection
The final element is the choice of toxin. The relative success of Mylotarg over other early ADC conjugates demonstrated the requirement for a highly potent payload. Microtubule disrupting drugs (vinca alkaloids) and DNA damaging compounds (anthracyclins) were first developed as payloads for ADC use, but lacked potency. Screening for novel cancer agents had identified many very highly potent cytotoxics, but most of these had failed to progress clinically as a result of their very narrow therapeutic window for these compounds. Dolastatins and maytansinoids had been identified as highly potent microtubule disrupting agents, but failed in Phase I trials as unconjugated drugs because of their dose-limiting toxicity. However, they are currently the leading payloads in ADC therapy, with Adcetris using the dolastatin derivative auristatin E, and Kadcyla using the maytansine derivative DM1. The calicheamicin payload used by Mylotarg is still in use following improvements to the linker chemistry, and the anti-CD22 ADC inotuzumab ozogamicin is currently in Phase III trials.

Despite this success, new payloads are required, especially those with new mechanisms of action. Other payloads under development include the DNA-damaging duocarmycins and pyrrilobenzodiazapenes, which look to be more potent than the microtubule disrupting drugs, with greater activity against slowly dividing cancer cells. Other novel payloads include the highly potent amatin RNA polymerase II inhibitors derivatives isolated from the death cap mushroom.

Manufacturing challenges
The development and manufacture of ADCs can be a challenge, as it requires the ability to handle both biologics and highly potent small molecule toxins. This combination of capabilities is not often found in the same organisation, and only a relatively small number of CMOs cover the ADC manufacturing space. As the clinical trial results of ADCs continue to be largely positive, with few failures, this will put a strain on ADC manufacturing capacity. In-house manufacture will be difficult for many pharma companies, particularly those with a small number of programs, and experienced ADC CMO capacity is limited.

ADCs offer the potential to revolutionise cancer treatment. After many false starts, the potential of this technology is now becoming clear. The true potential of this technology may be realised in coming years, when it will become evident whether the success of the recent commercialised products can be repeated, especially in the treatment of solid tumours.