Globally, about 1 in 6 deaths occurs due to cancer, and it is the second leading cause of death in the world. According to estimates from the International Agency for Research on Cancer (IARC),¹  there were 17 million new cancer cases and nearly 10 million cancer deaths worldwide in 2018, while by 2040, the global burden is expected to grow above 16 million cancer deaths, mostly due to the growth and aging of the population and the increasing prevalence of risk factors like smoking, unhealthy diet, physical inactivity, and fewer childbirths, in economically transitioning countries – a concerning development.

On the positive side, there are many ways to treat cancer, ranging from surgery to radiation therapy, chemotherapy, immunotherapy, targeted therapy, hormone therapy, and stem cell transplant. For solid tumors, the most effective therapy is usually surgical removal, often in combination with adjuvant chemotherapy. In the case of metastatic disease, chemotherapy is generally the only option, and it is rarely curative. For hematological tumors, surgery is not an option, and chemotherapy is typically the standard of care. However, chemotherapeutic drugs are limited by undesired toxicities, which leads to systemic side effects. In fact, all chemotherapeutic drugs are administered at doses that are barely tolerated by patients, but nevertheless well below desired therapeutically active levels (see Figure 1.1, left). A recent review concluded that most of the anti-cancer drugs that entered the market between 2009 and 2013 showed only a marginal gain in terms of overall survival, so there is an urgent need to increase therapeutic efficacy with more effective and safer anti-cancer drugs.²  Thus, improved anti-cancer drugs are highly desirable to offer an improved prognosis for cancer patients.

One class of drugs that has gained major interest in the biopharmaceutical industry is targeted cancer treatment with therapeutic antibodies. Since the advent of hybridoma technology,³  monoclonal antibodies can be mass produced, and antibody-based treatment of cancer has established itself as one of the most successful therapeutic strategies for both hematologic malignancies and solid tumors, with 30 monoclonal antibodies approved by the US Food and Drug Administration (FDA) for the treatment of cancer.⁴ 

In the past 30 years, huge efforts have been made to merge the positive features of biological therapy and chemotherapy in the form of an antibody–drug conjugate (ADC). The concept hinges on the idea of utilizing target-specific antibodies as vehicles for the delivery of highly cytotoxic drugs directly at the tumor site but not to healthy tissue, thus increasing efficacy and reducing toxic adverse events. The concept is inspired by the ‘magic bullet’-theory first proposed by Paul Ehrlich over a century ago.^5,6  Ehrlich coined the vision that it should be possible to administer a therapeutically effective dose of a cytotoxic compound targeted via an agent of selectivity to a specific subset of diseased cells in a living organism without evoking systemic toxicity, thereby effectively enhancing the therapeutic index of traditional chemotherapy (see Figure 1.1, right).

The challenge of realizing a true magic bullet concept employing an antibody as the agent of selectivity becomes clear from the fact that it took so many years until the first ADC was approved (Mylotarg^® in 2000). In 1958, the French oncologist Georges Mathé and his colleagues were the first to investigate the concept of curing xenografted mice through the administration of a mixture of antibodies, chemically modified with a cytotoxic drug.⁷  Specifically, antibodies were isolated from the blood of hamsters that had been immunized with leukemia cell line 1210. Separately, methopterin (N-demethyl methotrexate) was subjected to diazotization, then combined with the isolated pool of hamster immunoglobulins (IgGs), thereby providing the envisaged antibody conjugates through the reaction of lysine sidechain amino groups with the diazo-derivative of methopterin (see Scheme 1.1). The resulting mixture was administered at a dose of 125 mg kg⁻¹ to mice that had been grafted with leukemia 1210 on the day before dosing. It was found that ADC-treated mice survived longer (14–16 days), and their tumors shrank versus the appropriate controls (max. 10 days of survival with growing tumors).

It was clear to Mathé that a fundamentally new approach to cancer treatment had been revealed by this experimental result when he wrote: “Une nouvelle méthode générale de chimiothérapie semble ouverte: transporter électivement, par des -globulines de serum immune, des agents chimiques actifs jusqu’à l'agent pathogène ou aux cellules maladies, au moyen d'une function amine diazotable ou d'une autre combinaison chimique dont la recherche est en cours.” In English: “A novel general method for chemotherapy appears feasible: the effective transport of an active chemical substance, by means of gamma-globulins from immunized serum, to a pathogenic agent or cancerous cells, by means of amino group diazotization or another chemical modification”.

In the 1960s, Mathé's concept was expanded by a group at the Rand Development Corporation (Cleveland, Ohio), where methods were developed for the chemical coupling of a variety of chemotherapeutic agents to immunoglobulins.⁸  This new approach of cancer treatment was picked up by a group of surgeons and pathologists at the Victoria General Hospital in Halifax (Nova Scotia, Canada) in 1972, who likely were the first ever to treat a cancer patient with an ADC, albeit with the chemotherapeutic compound bound by adsorption to the antibody carrier, rather than by covalent coupling.⁹  In a case report, it was described how a patient with disseminated malignant melanoma was injected, first locally and then systemically, with anti-melanoma immunoglobulin to which the nitrogen mustard chlorambucil was bound. The single patient trial had a good outcome: all metastatic nodules regressed. Eli Lilly and Company were then the first pharmaceutical firm to expand upon the pioneering work of Mathé⁷  and Ghose,⁹  with a clinical feasibility study on the preparation of a conjugate of vindesine covalently linked to a sheep anti-CEA (carcinoembryonic antigen) immunoglobulin.¹⁰  The invention of monoclonal antibody technology by Köhler and Milstein³  in the mid-1970s enabled the true birth of the ADC field, moving beyond the coupling of chemotherapeutic agents to crude immunoglobulin serum fractions. In the years to follow, particularly in the mid- to late-1980s, multiple attempts were made to advance the concept into the clinic,¹¹  conjugating monoclonal antibodies having defined target specificity with a large range of cytotoxic agents (see Figure 1.2).

Unfortunately, this first generation of ADCs suffered from multiple shortcomings: (a) mouse monoclonal antibodies were used, leading to strong anti-drug immune responses in patients, (b) the cytotoxicity of the drugs was generally (far) too low given the quantitative limitation placed on cytotoxin delivery by the number of surface target antigens per cell, (c) unstable linkers were often used, and (d) only limited analytical tools were available to characterize the crude mixtures obtained by random conjugation technology.

The advent of chimeric and humanized antibodies drastically changed the potential of monoclonal antibody therapy, and hence also the field of ADCs, by solving the problem of immunogenicity, as nicely illustrated by the first two ADCs based on chimeric/humanized antibodies to enter the clinic in the mid-1990s: BR-96-DOX and CMA-676 (see below).

Beyond the abovementioned reasons (a–d) for the trail-blazing generation of ADCs to fail, it is important to be aware of a fundamental limitation in the conceptual idea of an ADC, that is, creating a cytotoxic agent that can enter a cancer cell and then kill it, while at the same time can remain benign if entering another cell, such as cells of liver, skin, or intestine. The dream that appending a cytotoxic molecule to a delivery vehicle would ameliorate its general toxicity is potentially rather naive: “it all goes somewhere!”. In fact, only a very small portion of the antibody (ADC) actually finds the cancer target cell (<0.003–0.08%injected dose per g of tumor).¹²  The remainder will be taken up in time by healthy tissues, either by macropinocytosis, or via the target receptor which may also be expressed on some normal tissue. After all, antibody molecules do not circulate forever! As a consequence, these healthy cells will generally also catabolize ADCs and release the active cytotoxic payload in their cytoplasm, thus potentially wreaking havoc both on the catabolizing cell as well as on neighboring cells. These considerations suggest that it may be difficult to achieve a therapeutic index of >1 using an ADC approach to targeted chemotherapy (see Figure 1.1, middle).

Although a therapeutic index >1 may be difficult to achieve, the toxicity profile of an ADC may be very different from that of systemic administration of the small molecule cytotoxic agent since, in an ADC, the distribution of the small molecule is tied to the distribution of the antibody. For example, early experience with BR-96-DOX (SGN-15/BMS-182284), an ADC targeting Lewis Y antigen with the chemotherapeutic doxorubicin (Adriamycin) as toxic payload,¹³  brought to light a great difference between the toxicity of the ADC (target-related gastrointestinal toxicity) and the systemic toxicity of doxorubicin (hematopoietic toxicity).^14–16  Besides the target-related toxicity of this particular ADC, the relative lack of potency of the payload, and the instability of the linker, ultimately led to discontinuation after a phase 2 study.¹⁷ 

The first ADC to receive market approval was CMA-676. Besides utilizing a humanized IgG4 anti-CD33 monoclonal antibody (overcoming the immunogenicity of early ADCs), this conjugate was the first to employ a payload with picomolar cytotoxic potency, the DNA-damaging agent calicheamicin (overcoming the insufficient potency of payload in early ADCs). It received accelerated approval in 2000 as Mylotarg^® for the treatment of acute myelogenous leukemia (AML). However, in 2010, it was voluntarily withdrawn by Pfizer due to lack of clinical benefit in a confirmatory phase 3 clinical trial. In fact, the development trajectory of ADCs is marked by hurdles and setbacks,^18,19  and ∼85 ADCs have been discontinued from clinical development over the years. It is not surprising, therefore, that many of the design elements of the second generation of ADCs – chimeric immunoglobulin and stochastic conjugation of toxic payloads via linkers that have poor stability in vivo – have now been mostly supplanted by new technologies.

The multiple clinical studies with ADCs, the failures, and the successes, have taught the field valuable lessons regarding the selection of target and antibody,²⁰  conjugation site(s), linker chemistry, payload optimization, and optimal drug loading.^21,22  In fact, as of April, 2021 a total of 10 ADCs have been approved by the US FDA, six of which were approved in the past two years alone (see Section 1.3). The excitement in the field is high, as is apparent from the hundreds of clinical trials initiated in recent years, and a total of ∼95 unique ADC assets currently in active clinical evaluation. The question remains, is it possible to achieve a therapeutic index >1 with an ADC?

In the ideal case, an optimized ADC is a unified partnership between the therapeutic target, monoclonal antibody (mAb), cytotoxic drug, and chemical linker. First of all, monoclonal antibodies provide an ideal delivery platform for selective targeting to tumors, based on the high selectivity for tumor-associated antigens (TAAs) (the target, see Section 1.4), long-circulating half-lives, and little to no immunogenicity (if human or humanized). The payload is typically a highly potent cytotoxic moiety (nanomolar to picomolar IC₅₀ values) for highly efficient cell killing once delivered to the tumor, with a mode of action (MOA) carefully tailored to the clinical indication (see Section 1.5). Thirdly, payloads must be attached to the antibody at preferred conjugation sites and preferably with a stable conjugation technology (see Section 1.6). Finally, the nature of the linker must be carefully chosen (non-cleavable or cleavable), besides other properties like stability, spacer length, and polarity. Considerations on linker design and properties will be briefly touched upon here (see Section 1.7) but will be thoroughly discussed in the ensuing chapters.

As the name suggests, an ADC consists of an antibody, covalently conjugated to a (cytotoxic) drug. It is of interest to note that, despite significant preclinical effort on other protein carriers, clinical ADCs are always built on a traditional IgG format, i.e. also containing an Fc-fragment (see Figure 1.3, left): to date, not a single small protein format like scFv, VHH (nanobody), Fab, F(ab)₂, diabody, nor any other small non-antibody target-binding protein (e.g. anticalins, affibodies, DARpins etc.) has entered the clinic in an ADC format. This lack of variability in antibody format is most likely due to the high clearance rate of small proteins versus full-length IgG, in combination with only modestly enhanced tumor uptake (also called the ‘valley of death’ in the relationship of protein size versus clearance).²³  In other words, to really benefit from the known enhanced permeability of tumors, a very small delivery vehicle with high affinity or avidity is mandatory, as, for example, is the case for peptide–drug conjugates and small molecule ligand–drug conjugates, new modalities for targeted delivery that are beginning to make their way into the clinic²⁴  (not discussed here). In an ADC, the antibody is covalently attached to a defined number of cytotoxic payloads (n), through a chemical linker (see Figure 1.3, right). The linker is attached to a specific number of sites on the antibody, the location of which is determined by the conjugation technology employed.

ADCs operate through a unique mechanism of action, a simplified description of which is depicted in Figure 1.4. After being introduced into plasma by intravenous administration, the antibody part of the ADC will recognize a TAA on the cell surface (step 1). The ADC–antigen complex then undergoes internalization by endocytosis (step 2). Once inside the tumor cell, the ADC–antigen complex is fused with the endosome, which may break up the complex for recycling of the antigen back to the cell surface, while the ADC is trafficked to the lysosome (not depicted). In other cases, the entire ADC–antigen complex is routed to the lysosome (step 3). In the lysosome, the antibody, and potentially the linker (if a peptide), undergoes proteolytic degradation to release the cytotoxic payload, which should then be transported by an active or passive process into the cytoplasm (step 4). Other linkers may be cleaved by different mechanisms and may not be cleaved exclusively within the lysosomal compartment, as detailed in the ensuing chapters. The cytotoxic payload is now ready to induce cell death or apoptosis, for example, through binding to tubulin to disrupt microtubule dynamics or, after entering the cell nucleus, by damaging DNA directly, or by inhibiting enzymes such as topoisomerases that maintain DNA integrity and/or participate in DNA replication (step 5).

In 2000, the first ADC to reach FDA approval, for the treatment of AML, was CMA-676/gemtuzumab ozogamicin (Mylotarg^®), an ADC designed and developed by Wyeth (later Pfizer) to destroy CD33-expressing myeloid cells. It received accelerated approval based on positive response rates in a single-arm clinical study.²⁵  This quick success was remarkable, given the fact that Mylotarg^® was perhaps the first of the second-generation ADCs (not based on a murine antibody) to enter the clinic in the mid-1990s. Not surprisingly, the approval of Mylotarg^® served as a great inspiration for others, which led to the emergence of a healthy pipeline of ADCs (14 different compounds) at various stages of clinical development by 2008.²⁰  Among the early believers (besides Wyeth) were Bristol Myers Squibb (BMS)/Medarex, Seattle Genetics (now Seagen), Genentech, Sanofi (now Sanofi-Aventis), AstraZeneca/MedImmune, and most notably ImmunoGen, whose technology was involved in >50% of all these early programs.

Despite the early success of Mylotarg^®, however, no other ADCs were approved for years to follow. Worse than that, Mylotarg^® was voluntarily withdrawn in 2010, when a confirmatory post-approval phase 3 clinical trial showed that it did not improve survival and was associated with significant toxicities, including fatal anaphylaxis, adult respiratory distress syndrome, and veno-occlusive disease,²⁶  which left the US and European markets with no approved ADCs (see Figure 1.5), although Mylotarg^® remained in the market in Japan. While this development was a great disappointment to those working in the ADC field, 2010 was also a year that saw the first peer-reviewed publications about phase 1 trials with ADCs utilizing highly potent tubulin-acting agents as payloads (maytansinoids and auristatins), which sparked renewed enthusiasm for ADCs.^27,28  Indeed, these reports mark the turning point in the perception of ADC technology, which began to receive widespread acceptance in oncology as a technology that had moved beyond an attractive idea that was deceptively difficult to implement, to one that was being reduced to practice.

Fortunately, the hiatus in any approved ADCs was short-lived (see Figure 1.5). In 2011, the FDA granted accelerated approval to brentuximab vedotin/SGN-35 (Adcetris^®), an ADC designed to bind to CD30-positive B-cells, for the treatment of Hodgkin's lymphoma as well as anaplastic large cell lymphoma (ALCL). Adcetris^® was developed by Seagen and was fundamentally new with regard to preceding technologies for three reasons: (a) use of a novel auristatin payload, (b) use of a protease-sensitive cleavable linker based on a dipeptide valine–citrulline attached to p-aminobenzyloxycarbonyl (Val-Cit-PABC), and (c) attachment of the linker–drug moiety to the antibody by means of reaction of a maleimido group of the linker–payload with sulfhydryl groups of cysteine residues generated by reduction of interchain disulfide bonds in the antibody. The approval of Adcetris^® was followed two years later by the first approval of an ADC for a solid tumor indication, Kadcyla^® (ado-trastuzumab emtansine, T-DM1). Kadcyla^® was developed for the treatment of HER-2-positive breast cancer and was the first ADC to receive full approval from the FDA based on the results of a randomized phase 3 clinical trial.²⁹  By that time, the auristatin-ADC technology developed by Seagen had become equally as popular as ImmunoGen's maytansinoid-based technology, with application in 13 of the 30 clinical-stage ADC programs versus the 14 clinical-stage maytansinoid-ADCs, thereby jointly clearly dominating the ADC field.⁵ 

Notwithstanding the healthy development of the ADC pipeline and total clinical candidates gradually increasing, it took another four years again until the next approval: that of CMC-544/inotuzumab ozogamicin, now marketed as Besponsa^®, in August 2017. Interestingly, just one month later, Mylotarg^® was reapproved by the FDA, based on new data in phase 3 clinical trials showing efficacy at a lower dose and a fractionated dosing schedule,³⁰  thus increasing the number of approved ADCs to four by the end of 2017. The difference in design between ADC technologies of the approved ADCs is remarkable (see Figure 1.6). Adcetris^® and Kadcyla^® share a common mechanism of action of the payload, i.e. tubulin inhibition, and have similar average drug-to-antibody ratios (DARs) (∼4), but they differ in the nature of linker (cleavable versus non-cleavable) and conjugation technology (cysteine versus lysine), respectively. Besponsa^® and Mylotarg^® are also both based on stochastic lysine conjugation like Kadcyla^®; however, they feature a calicheamicin payload with a distinctly different MOA, i.e. DNA damaging. Furthermore, the linker for these calicheamicin-ADCs is based on an acid-sensitive hydrazone motif (for a more detailed discussion of payloads and conjugation technologies, see Section 1.6 and Chapter 2, respectively).

It had become apparent by the end of 2017 that ADCs could provide a valuable treatment option for both solid tumors (Kadcyla^®) and hematological cancers (Adcetris^®, Besponsa^®, and Mylotarg^®). The total number of ADCs in the clinic had already risen to ∼75, and excitement in the field was now steadily increasing. Another factor driving increased enthusiasm for ADC technology, and investments in moving ADC compounds into clinical development, was the fact that four out of the 14 early-stage ADCs had now been approved by the FDA, a truly unprecedented success rate (29% from Investigational New Drug (IND) application to market approval) and comparing favorably with the success rate of small-molecule drugs in the field of oncology (∼5%)³¹  and even with that of monoclonal antibodies (∼22%).³²  Furthermore, the success of Kadcyla^® helped overcome some early skepticism about whether ADCs would be effective in treating solid tumors because of the supposed inadequate penetration of antibodies into solid tumor tissue.

The clinical success of ADCs is further underlined by the fact that by April 2021, 6 more FDA approvals were noted, bringing the total to 10 marketed ADCs in the United States, as well as four approvals by the European Medicines Agency (EMA), summarized in Table 1.1. Without going into more detail, the reader is referred to other reviews.^33,34  The growing panel of approved ADCs is a clear testament to the success of the technology developed by Seagen, with Polivy^® and Padcev^® also based on the payload monomethyl auristatin E (MMAE), and Blenrep based on monomethyl auristatin F (MMAF). In fact, Blenrep is the second example of success with a non-cleavable linker, after Kadcyla^®. More recently, ADCs containing payloads with a third MOA, namely topoisomerase I inhibitors, have also found their way into the market with the recent FDA approvals of Enhertu^® and Trodelvy^® (see Figure 1.7).

As a consequence of the clinical benefit for patients, ADCs are also becoming a commercial success, with the global market steadily growing from $0.9 billion in 2014 to more than $3 billion today³⁵  (see Figure 1.8). Moreover, the market is expected to reach $9.9 billion by 2025, with an annual growth rate of nearly 26%,³⁶  potentially exceeding $15 billion by 2030.³⁷ 

The success of the current marketed ADCs is a clear testament to the promise of targeted cancer treatment. Not surprisingly, the ADC pipeline has now risen to an additional ∼85 ADCs in the clinic, of which seven are in late-stage pivotal trials (see Table 1.2). Of the ADCs in the pivotal trial stage in mid-2021, tisotumab vedotin (Tivdak™) developed by Seagen and Genmab received accelerated approval from FDA in September, 2021, for the treatment of recurrent or metastatic cervical cancer, a fourth MMAE-based ADC. Likely to follow soon are ImmunoGen's IMGN853/mirvetuximab soravtansine and potentially others.

Based on the great clinical success of ADCs, it is tempting to conclude that the “magic bullet” concept of Ehrlich has now really been put to practice. However, one should be cautious about jumping to this conclusion. Without exception, all ADCs are still administered to patients at their respective maximum tolerated doses (MTDs), as determined in phase 1 dose-escalation clinical studies. Therefore, by definition, the therapeutic window of ADCs is still, at best, equal to 1 for the compounds where the clinical benefit allows for approval. However, in the majority of cases, the therapeutic window may be considered to be still <1, as one would like to be able to administer at a higher dose to reach a yet better clinical response. It must be acknowledged, therefore, that the dream of a truly “magic bullet” still awaits realization. With all these moving parts, it is extremely challenging to design the best possible ADC for a given indication – but the field is learning.

The selection of a TAA (the ‘target’) for an ADC requires some different considerations than for naked antibodies in cancer therapy. Antibodies without a cytotoxic payload have various mechanisms of action, including target receptor neutralization (antagonism), receptor downregulation, signaling disruption, antibody-dependent cell-mediated cytotoxicity (ADCC), antibody-dependent cell-mediated phagocytosis (ADCP), complement-dependent cellular cytotoxicity (CDC), and immune checkpoint inhibition. For ADCs, however, upon antibody binding to its target antigen, the target should internalize efficiently via receptor-mediated endocytosis, to allow the antigen–ADC complex to undergo appropriate intracellular trafficking and degradation to ultimately release the cytotoxic warhead. Target antigens on the cell surface can have variable rates for internalization; this can be an intrinsic feature of the antigen itself, or it can be induced by the binding of the antibody. Other factors that influence antigen–ADC internalization are the specific binding epitope and the affinity of the antibody for the target. Taken together, ADC internalization may be crucial for efficacy for most targets. Furthermore, target-mediated internalization of an ADC and intracellular payload release may reduce toxicity associated with extracellular release of the cytotoxic payload, perhaps in the conditions of the tumor microenvironment, leading to systemic toxicities due to free payload.

The overexpression of target receptors on tumor cells relative to all normal tissues is critical to assessing therapeutic potential: a target antigen with a high tumor-cell-to-healthy-cell ratio will lead to greater delivery of ADC to tumor cells versus normal healthy cells, thereby impacting efficacy. Furthermore, a low level of target expression on just a small amount of normal tissue will result in less target-mediated drug disposition (TMDD) of the ADC, thereby impacting safety, as well as allowing for greater delivery to tumor tissue than would be the case when the TMDD is high. For example, the expression of a candidate target on blood cells must be considered, as such expression on these highly accessible normal cells could significantly reduce ADC exposure, and expression on antigen-presenting cells may increase the risk of immunogenicity. The target antigen should also have minimal secretion in circulation from the cancer (“shedding”), because antibodies can bind to these shed antigens in the circulation, thus limiting the amount of antibody available for target cell binding.

The parallel optimization of all of these target attributes is a great challenge. As a consequence, the search for the ‘ideal’ ADC target remains largely empirical in the clinical setting: >80 different targets have been evaluated over the years. A plethora of TAAs have been investigated in preclinical models, and in clinical trials, including lineage-specific antigens expressed on B cells (e.g., CD20, CD22, CD40, CD79b) and T cells (CD25, CD30) and target antigens overexpressed on carcinoma cells (HER2, EGFR, EpCAM, EphA2, PSMA, Cripto), endothelial cells (endoglin), or stromal cells (fibroblast-activation protein), to name but a few (see Table 1.3).

Some targets clearly stand out in terms of their popularity as ADC targets (HER2, TROP-2, c-Met), while others have been unsuccessfully assessed multiple times (CD70, PSMA, MUC16). At the same time, the vast majority of targets have only been assessed a single time, or two at most (see Figure 1.9). Therefore, it would be inappropriate to judge the value of a target by the numbers only, because target selection is just one of the parameters to determine the success of an ADC in the clinic. For example, CD33-targeting Mylotarg^® is a marketed product, yet the development of three other ADCs targeting CD33 (AVE9633, SGN-CD33A, IMGN779) was discontinued following clinical evaluation.

Furthermore, there is a long list of targets that no longer have ADCs in active clinical evaluation: CD70, PSMA, Lewis Y, MUC16, SLITRK6, α_v-integrin, ASCT2, c-KIT, C4.4a, CA-IX, CD324, CD352, CD44v6, CD48a, CD56, CLL-1, Cripto, CS1, DLL3, DPEP3, ENPP3, Ephrin-A2, Ephrin-A4, ETBR, FGFR2, FGFR3, FLT3, GD3, Globo H, GPC3, gpNMB, LAMP-1, LRRC15, Ly6E, MFI2, NOTCH3, p-cadherin, PRLR, RNF43, STEAP1, and TIM-1. However, for these failed targets, one should not jump to conclusions quickly. For example, four ADCs targeting MUC-1 have failed in the past, but most recently, Merck Serono's ADC M1231, based on a MUC-1/EGFR bispecific antibody, entered clinical trials, and at least two other preclinical programs are currently re-assessing the potential of MUC-1 based on antibodies targeting other specific epitopes (for example, epitopes to domains of MUC-1 that are not shed from the cell surface). Similarly, Ephrin-A2 has failed as a target for an ADC, but it is the target of the peptide–drug conjugate BT5528, developed by Bicycle Therapeutics.³⁸ 

Obviously, an ideal target is highly expressed on cancer cells but has no (or very low) expression on healthy tissue, as the latter might directly translate into ADC toxicity. An early example of an ADC failure due to unforeseen target expression, beyond its very high expression in squamous carcinoma of head and neck, is a CD44v6-targeting ADC (bivatuzumab mertansine), which caused toxicity in the skin, consistent with the characterized expression of CD44v6 in skin keratinocytes.³⁹  Also in the case of glembatumumab vedotin (CDX-011), rash was one of the observed dose-limiting toxicities (DLTs),⁴⁰  which is probably due to membrane expression of its target gpNMB in epithelial cells of the skin. In contrast, SLC34A2 (NaPi2b) has notable expression in normal lung tissue, yet two ADCs targeting SLC34A2, namely DNIB0600A and XMT-1532, have shown good safety and pharmacokinetics in the clinic, with little evidence of clinically significant lung toxicity. HER2 is also expressed at relatively high levels in normal tissues compared to other ADC targets; nevertheless, DLTs of multiple HER2-targeting ADCs, including Kadcyla^® and Enhertu^®, are generally not attributed to target expression on normal tissues.

Several parameters may determine the translation of target expression on normal tissue into DLT, such as the biodistribution of ADC and exposure in various tissues, the linker and drug used in the ADC, and the particular cell type that expresses the target. For example, non-proliferative cells are less susceptible to microtubule inhibitors because the drug's mechanism of action is dependent on cell cycle progression. The interplay of target-dependent and target-independent toxicities, which can vary based on these parameters, further complicates the predictions of toxicities based on non-clinical toxicology studies and underlines the empirical nature of ADC development. An interesting approach to mitigate on-target toxicity is the development of antibodies that show enhanced target binding at a lower pH, thereby taking advantage of the unique conditions in the tumor microenvironment. For example, BioAtla is developing several “conditionally active biologics” (CAB) such as BA3011 and BA3021, binding to Axl and ROR2, respectively. These CAB ADCs have reduced binding to the target antigen in healthy tissue (normal physiological conditions, ∼pH 7.4) but have stronger binding in the context of low pH of the tumor microenvironment (∼pH 6.5)due to the glycolytic metabolism of tumors (i.e. Warburg effect). Similarly, Chugai has developed the Switch-Ig™ technology⁴¹  around antibodies that only bind to targets in the presence of high concentrations of extracellular ATP found in cancer tissue. Another approach to enhance the uptake of ADCs in the tumor environment is the Probody™ technology⁴²  developed by CytomX, involving the transformation of an antibody into a prodrug that remains inert in healthy tissues and in systemic circulation, but is activated locally at the site of disease. A similar technology called SAFEbody™ is marketed by Adagene. These “proantibodies” consist of masking the antibody-binding site via a substrate peptide that is removed by proteases found in the tumor microenvironment. The Probody™ technology has demonstrated markedly improved safety and increased half-life in non-human primates and is now applied in various clinical programs (including the ADCs CX2009 and ABBV-CX-2029).

Besides target-related toxicity of ADCs, off-target toxicity also plays a key role, although the mechanisms behind it are more poorly understood. Indeed, “platform toxicity” – non-target-mediated toxicity – generally defines the DLT for the majority of ADCs evaluated in the clinic.⁴³  Neutropenia and thrombocytopenia could be explained by the cytotoxicity of the free payload after the processing of the linker–drug by the targeted cells (or in the tumor microenvironment). Alternatively, the processing of ADCs following the uptake by pinocytosis,⁴⁴  or by Fc-gamma receptor (FcγR)-mediated uptake,⁴⁵  has been proposed as a potential mechanism for non-target-mediated toxicity.

FcγRs are a family of glycoproteins, expressed on the membrane of immune cells (see Table 1.4), that are capable of binding to a region of the Fc portion of IgG antibody molecules that includes the glycan. These receptors can bind to the various IgG subclasses with different affinities,⁴⁶  and the affinities can be modulated by the glycan structure. When FcγRs are crosslinked by multivalent antigen–antibody complexes, they can induce different cellular responses; for example, the multivalent complex forms the synapse between immune cell and target cell, which can, for example, lead to FcγR-mediated cell-mediated cytotoxicity.

A clear example of ADC toxicity attributed to FcγR-mediated interactions is that of LOP628, a c-Kit-targeting ADC with DM1 payload developed by Novartis/ImmunoGen. Despite promising preclinical data, clinical trials were quickly terminated after the first three patients suffered from hypersensitivity due to LOP628-mediated mast cell degranulation, likely caused by the co-engagement of c-Kit with FcγRIa present on mast cells.⁴⁷  Another example of putative FcγR-mediated toxicity is thrombocytopenia induced by T-DM1 (Kadcyla^®) due to the binding of the antibody glycan to FcγRIIa Fc-gamma IIa receptors.⁴⁵  Megakaryocytes showed uptake of trastuzumab and Kadcyla^®, as well as an isotype control ADC, resulting in cytotoxicity that was not observed with unconjugated trastuzumab. Blocking T-DM1 interaction with FcγRIIa receptors significantly decreased internalization into megakaryocytes but did not completely block uptake and cytotoxicity, suggesting that additional mechanisms, such as uptake into endosomes via macropinocytosis, could also contribute to platelet loss. This is supported by a report showing that both T-DM1 and AGS-16C3F (ADC targeting ENPP3 conjugated via a non-cleavable linker to MMAF) induced thrombocytopenia using an in vitro megakaryocyte cell differentiation platform.⁴⁴  In this assay, there was no direct effect of AGS-16C3F on mature platelets, similar to T-DM1, and because ENPP3 is not expressed on circulating platelets or its precursor megakaryocytes, it was concluded that ADC-induced thrombocytopenia was rather the result of target-independent macropinocytosis of ADCs in megakaryocytes leading to the inhibition of their differentiation.⁴⁴ 

In view of the above hypothesis that thrombocytopenia (to cite one ADC-induced toxicity) may be a consequence of Fc/FcγR interactions, several ADCs are based on Fc-silent antibodies. The suitability of a given native or engineered Fc for an antibody or ADC product candidate is determined on a case-by-case basis,⁴⁸  and various clinical ADCs have been designed to display reduced or no binding to FcγR by protein engineering of the Fc-fragment of the antibody. For example, the biparatopic HER2-targeting ADC, MEDI4276, utilized an L234F mutation in combination with an S239C mutation to reduce binding to FcγR.⁴⁹  The S239C mutation also provides one of the two sites engineered into this bispecific-biparatopic antibody for site-specific conjugation via a maleimido-linker-payload.

On the other hand, it is intriguing to note that several ADCs with enhanced FcγR binding have also entered the clinic. Firstly, Blenrep (GSK2857916), an ADC targeting the B-cell maturation antigen (BCMA) that is selectively expressed on multiple myeloma (MM) cells, is based on a non-fucosylated antibody glycan expressed with POTELLIGENT™ technology. Similarly, BAT-8003, now discontinued, is based on a non-fucosylated glycan, which is potentially also the case for Enhertu^®. Due to the absence of fucose in the antibody glycan, affinity to FcγRIIIa is significantly enhanced, thereby contributing to enhanced immune cell eradication by ADCC. The relative role of enhanced ADCC versus payload delivery in the clinical efficacy of each of these ADCs is unclear.

Clearly, knowledge of FcγR biology, expression patterns in normal cells/tissues, and physiochemical factors of ADCs that contribute to FcγR binding is of great importance for understanding potential target-independent toxicity mechanisms. In particular, the role of FcγRs in mediating the off-target toxicity of ADCs, especially as a potential mechanism for hematotoxicity (toxicity to blood cells), needs to be considered when designing an ADC.

As a third non-specific mechanism of off-target toxicity (after the premature release of free payload, or antibody Fc-mediated cellular uptake), macropinocytosis may contribute to the in vivo uptake of ADC by antigen-negative hematologic cells, vascular endothelial cells, or epithelial cells in the vicinity of rich capillary beds (e.g., epithelial cells of skin, corneal epithelial cells) that are exposed to the high initial concentrations of ADC in blood plasma following infusion (see Figure 1.10). Other mechanisms of non-target-specific cellular uptake, not discussed in further detail here, that may also play a role in off-target toxicity are phagocytosis and clathrin-mediated and caveolin-mediated endocytosis.⁵⁰ 

The selection of an appropriate antibody for ADCs is essential, as it has a significant impact on efficacy, pharmacokinetic/pharmacodynamic profiles, and therapeutic index. The ideal mAb for an ADC should be target specific with a good target-binding affinity. In addition, it should have low immunogenicity, low cross-reactivity, efficient internalization, and a long plasma half-life. The IgG antibody subtype has been used exclusively to date in ADCs that have entered clinical evaluation due it its serum stability and slow clearance from plasma. An IgG antibody consists of two heavy chains and two light chains, which together form the antigen-binding domain (see Figure 1.3). In addition, an IgG contains a constant fragment (Fc), which mediates the binding of the antibody with effector cells of the immune system (via binding to FcγR as discussed above) and which regulates the half-life of the antibody in circulation through its interaction with the neonatal Fc receptor (FcRn). The vast majority of clinical ADCs are based on the IgG1 antibody subtype, with a few (including two approved ADCs) of the IgG4 subtype. A small minority of clinical-stage ADCs are based on the IgG2 subtype (e.g. TR1801). IgG4 is preferred for the development of ADCs when recruitment of host effector function is not required for anti-tumor activity and is desirable from the point of view of the potential FcγR-mediated off-target toxicity due to its reduced binding to FcγRs relative to IgG1.

Finally, a recent trend in the field is to use, besides traditional IgG, antibody formats targeting two different epitopes on a single TAA (biparatopic antibody) or targeting two different receptors on the same tumor (bispecific). For example, both MedImmune (MEDI4276, now discontinued) and Zymeworks (ZW49) have developed biparatopic ADCs that target two different epitopes on HER2 for the treatment of breast cancer, with potentially enhanced HER2 internalization and downregulation. A second-generation FRα-targeting ADC, IMGN151, is being developed by ImmunoGen, based on a biparatopic antibody design. Finally, the earlier mentioned ADC M1231, based on a bispecific antibody targeting MUC-1 and EGFR, recently entered clinical trials and is of particular interest in view of the fact that both receptors have been independently evaluated, but each has had several failed ADC programs. It will be interesting to see if a bispecific ADC can change the potential, and perspective, of these targets in the future.

The use of alternative formats to a traditional full-size IgG molecule in ADCs may well be the next major area of technology development, following the investments in the development of conjugation chemistries, the discovery of payloads suitable for ADCs (see Section 1.5), and the creation of a variety of linker technologies (subject of this volume), in order to design effective ADCs. Indeed, the emerging developments in the application of bispecific antibody technology, of conditionally active antibodies, of pro-antibody technology (Probodies™, SAFEBodies™), or of alternative bivalent formats such as diabodies may be fertile ground for further widening the therapeutic index of ADCs, pushing closer to the >1 goal of an index.

The payloads utilized in clinical ADCs are highly potent cytotoxic drugs, exerting their effects on critical cellular processes required for survival.⁵¹  Most ADCs that are marketed or in late-stage clinical development utilize either maytansine derivatives (DM1/DM4) or auristatins (MMAE/MMAF), both of which are microtubule inhibitors. ADCs based on microtubule inhibitors typically induce apoptosis in cells undergoing mitosis by causing cell cycle arrest at G2/M. Besides MMAE and MMAF, multiple other auristatin variants are being applied in clinical programs (monomethyl auristatin D, PF-06380101, duostatin5, AS269, Tap18Hr1, AGD-0182, HPA-Auristatin F), in addition to other tubulin-targeting payloads (hemiasterlin, tubulysin, eribulin). The second class of cytotoxic drugs used in ADCs are DNA-damaging payloads, which include enediynes (calicheamicin), duocarmycin derivatives, pyrrolobenzodiazepine dimers (PBD dimers), and indolinobenzodiazepine pseudo-dimers. Of these, calicheamicin is utilized in the marketed ADCs Besponsa^® and Mylotarg^®, while a PBD dimer is employed in ADCT-402 (loncastuximab tesirine), an ADC approved by FDA in April 2021, and marketed as Zynlonta™. A third, and rapidly emerging, class of payloads are the quinoline alkaloids (SN-38, DXd), which inhibit topoisomerase I. The latter payloads are used in two of the most recently approved ADCs, Trodelvy^® and Enhertu^®, respectively.

The majority of payloads utilized in ADCs are highly potent, often with cytotoxicity in the picomolar range, which is a requirement because only a very small amount of the injected dose of an antibody/ADC localizes to the tumors (<0.003–0.08%injected dose of ADC/g of tumor).¹²  However, it is this potency that also drives the toxicity of ADCs, resulting in the majority of toxicities being characterized by the class of payload (see Table 1.5).

In the selection of payload for a given clinical indication, it is essential to bear in mind that not all mitotically active cells will be sensitive to a particular cytotoxic, even if effectively delivered by the ADC. This factor may be a mitigating factor for minimizing the potential for unwanted on-target toxicity – if the normal tissue expression of the TAA is found only on terminally differentiated non-dividing cells, then delivery of a tubulin agent payload (for example) via an ADC might not cause toxicity. Another factor in payload selection is the consideration of the sensitivity of the disease to the MOA of the payload. More than 50 years of clinical experience with cytotoxic therapy has taught us important lessons on the ‘intrinsic sensitivity’ of specific cancer types to particular agents based on thousands of clinical trials. Based on this history, doubt should be cast upon the relevance and clinical application of ADCs armed with a given anti-microtubule-based payload that demonstrate activity in preclinical models of tumor types, where such tubulin-acting agents have never demonstrated clinical activity.⁵² 

In view of the above considerations, it is clear that it is essential to match the payload to the clinical indication. Fortunately, the toolbox of new payloads, with alternative MOA, is growing beyond the current widely utilized molecules, as exemplified by payloads like amanitin, dmDNA31, BCL-XL, KSP inhibitors, and TLR agonists. A previous volume in this series provides a comprehensive review of payloads that are, or may have been, utilized in creating ADCs.⁵¹  Given the broad spectrum of cytotoxic, biologic, and targeted agents, this should propel creative thinking in targeted payload selection and delivery while reinforcing the concept of target delivery of payloads when systemic delivery is not feasible.

The conjugation technology employed to attach a cytotoxic payload to an antibody can fundamentally alter the pharmacokinetics and therapeutic index of ADCs. Conventional drug conjugation typically occurs via alkylation of cysteine or acylation of lysine sidechains, both of which are used in all marketed ADCs. These conventional drug conjugation strategies are stochastic and produce a heterogeneous mixture of ADC molecules having different DARs and with partial modification at multiple potential modification sites, leading to variable pharmacokinetics, efficacy, and safety profiles. In particular, ADCs with high drug loading may suffer from destabilization, aggregation, increased off-target toxicity, and enhanced clearance from systemic circulation. Hence, there is a clear trend toward site-specific conjugation technologies that produce more homogenous, and often more stable, ADCs.⁵³  These strategies include the insertion of engineered cysteine residues, insertion of unnatural amino acids in the antibody sequence, or enzymatic conjugation methods, some of which require the introduction of enzyme recognition sequences (see Figure 1.11). An elaborate discussion on antibody conjugation technologies for ADCs can be found in Chapter 2 of this book.

The chemical linker connecting the antibody and the cytotoxic payload plays an essential role in ADC design. Various factors such as linker chemistry, mode, and the site of conjugation play a crucial role in the pharmacokinetic and pharmacodynamic properties of ADC. Linkers must maintain the stability of the ADC in the blood circulation to enable it to reach the cancer cell intact but also must be readily cleaved when internalized (cleavable linker) so that the payload can be released, or else form an innocuous element if it is not cleaved from the payload (non-cleavable linker) so that the payload can still exert its killing MOA. Cleavable linkers can be further subcategorized into acid-sensitive, protease-sensitive, and glutathione-sensitive linkers. Illustrative examples of the above are formed by Kadcyla^® (non-cleavable linker), Mylotarg^® (acid-sensitive linker), and Adcetris^® (protease-sensitive linker) (see Figures 1.6 and 1.7). An example of a glutathione-sensitive linker is mirvetuximab soravtansine, an ADC currently in pivotal late-stage clinical trials (see Figure 1.12).

Linker technology is the topic of this book. The linker is often overlooked in its importance, just regarded as “the piece in the middle” between the antibody (the target-binding biological delivery vehicle) and the payload (the cell-killing effector moiety) of the ADC. However, the chemical and biophysical properties of the linker play a crucial role in the design of an ADC. The biophysical and biochemical stability of an ADC can be markedly influenced by the chemical properties of the linker, which in turn can have a profound effect on the pharmacokinetic and pharmacodynamic behavior of ADCs in vivo. At the cellular level, the mechanism of release of the cytotoxic payload, the rate of release, the site of release (which intracellular subcellular compartment), whether extracellular release occurs due to local conditions in the tumor microenvironment, and perhaps the efficiency of transmembrane transport of the released payload metabolite, all are key properties of linker chemistry that enable an active and effective ADC to be made from a particular target/antibody pair with a particular payload type. There is certainly no “one size fits all” with regard to the linker in an ADC. The diversity in ADCs approved to date, including two with non-cleavable linkers, three with acid-sensitive linkers utilizing two different chemistries (hydrazone and carbonate), and five with protease-cleavable peptide linkers (three different peptide structures), provide testimony to the key role of the linker in ADC design. The remaining chapters of this book provide a detailed review of the variety of linker chemistries used in ADCs and a thorough analysis of how the properties of different linkers can influence the efficacy and toxicity profiles of ADCs, as researchers in the field strive to make safe and effective medicines for cancer treatment.