Targeted drug delivery for cancer therapy: the other side of antibodies

Rituxin, or Rituximab, was approved for the treatment of patients with Non-Hodgkin’s B-cell Lymphoma (NHL) in 1997 [15] and was actually the first TMA to be approved for cancer therapy. Rituxin is a chimeric antibody targeting CD20, a 33-35kDa cell-surface glycosylated phosphoprotein. Expression of CD20 is found on late-stage pro-B-cells (CD45R⁺/CD117⁺) and this expression increases with B-cell maturity, although it is absent from plasma cells [16]. CD20 was found overexpressed in several types of leukemias [17]. While no natural ligand has yet been described, its function may be related to effective B-cell responses to T-cell independent antigens, although other evidence suggests that CD20 acts as a calcium ion channel [16]. Early studies [18] demonstrated that Rituxin could induce B-cell depletion by several immune effector mechanisms including ADCC and complement-dependant cytotoxicity (CDC). Its ability to induce apoptosis has not been clearly demonstrated. Surprisingly, despite the pleiotrophic B-cell expression of CD20, B-cell lymphopenia does not seem to be related to increased rates of infection in long-term treated patients. This may be due to the ability of hematopoietic stem cells to regenerate the B-cell population relatively quickly. Rituxin does induce infusion reactions, possible resulting from release of inflammatory cytokines following administration [19]. Rituxin is currently employed in the management of several forms of NHL including Chronic Lymphocytic Leukemia (CLL), Follicular Lymphoma (FL) and Diffuse Large B-cell Lymphoma, most often in combination with chemotherapies [6]. Its use has had a profound effect on the management of patients, becoming the standard of care in NHL and FL, as well as in CD20⁺ CLL.

Ibritumumab is a murine anti-CD20 monoclonal antibody conjugated to the yttrium isotope (90Y-Ibritumumab tiuxetan). This intense β-radiation releasing immunoconjugate was approved in 2002 for use in patients with NHL but has also shown efficacy in Rituxin-refractory lymphoma [20]. Another immunoradioisotope, tositumomab-I¹³¹, was approved in 2003 for treatment of patients with CD20+ FL. Both drugs are efficacious but induce hemato-toxicity and have been the subject of several comparison clinical trials [21‐23].

Despite its proven clinical benefit, clinical response rates to Rituxin are still modest, at about 50% for NHL [24]. The even lower figures reported for CLL [25] likely reflect the variable expression of CD20 on the surface of these tumor cells as compared to some other forms of B-cell leukemia. Possibly because more promising B-cell leukemic markers have not yet been discovered, and also as a result of ongoing research into the biology of anti-CD20 antibodies, much effort has been invested recently in overcoming some of the limitations of these first generation antibodies. These include binding-induced modulation of CD20 expression [26], diverted binding to blood forms of free antigen [27], relatively low binding affinity [28, 29], development of resistance [28], genetic variability in the FcγRIIIa receptor gene among patients that affects ADCC activity [30] and uncertainty as to the mechanism of action. Studies on anti-CD20 antibodies has allowed their classification into Type 1 and Type 11 categories, based on their ability to induce changes in the membrane distribution and configuration of the antigen. These changes are thought to be related to the functional characteristics of the antibodies, such as their ability to form CD20/anti-CD20 lipid rafts, a function that may enhance their therapeutic efficacy [31]. In addition, while Type I antibodies effectively conscript CDC and ADCC, they are weak in inducing direct tumor cell death. They also modulate CD20 antigen expression, especially on CLL and mantle cell lymphoma cells [26]. On the other hand, Type II antibodies strongly induce direct tumor cell death , while inducing lower CDC but higher ADCC activity than Type I antibodies (reviewed in [16]). It should also be noted that while the tumor cell cytotoxicities of ADCC and CDC can be demonstrated, the technical validation of both the in vitro and in vivo methods is not-trivial [32‐34]. In addition, while NK cells, important effectors of ADCC, seem to infiltrate most types of solid tumors [35], it is unclear whether at least under some situations they also have a tumor promoting effect, [36]. Further understanding the molecular differences between Type 1 and Type 11 antibodies that might account for these differences in effector mechanisms should allow the engineering of more potent antibodies as recently discussed [37]. Several next-generation anti-CD20 TMAs are now discussed below.

Gemtuzumab is a recombinant, humanized IgG4 monoclonal antibody (mAb) targeting CD33, an antigen expressed on most leukemic blast cells but also on normal hematopoietic cells, although the intensity of expression diminishes with normal stem cell maturation. The antibody is linked to calicheamycin and was the first ADC approved by the FDA (in 2000), for use in patients with relapsed acute myelogenous leukemia [55]. Almost from the outset however, the use of Mylotag was associated with significant side effects [56, 57] and it was eventually withdrawn by its developers Pfizer in June 2010.

This ADC consists of a chimeric antibody directed to CD30, a member of the tumor necrosis factor receptor family. CD30 is rarely expressed on T or B-cell lymphomas, but is a tumor marker for classic Hodgkin's Lymphoma, anaplastic large cell lymphoma and embryonal carcinomas. The antibody is conjugated to the antimitotic compound monomethyl auristatin E. Adcetris was granted accelerated marketing approval in August 2011 and is the first new Hodgkin’s Lymphoma drug in 30 years [58, 59].

A number of other ADCs are currently undergoing clinical assessment. Indeed a recent Biopulse survey of pharma and biotech companies clearly indicates that many drug manufacturers are actively involved in ADC development and expect that many more of their products will enter the clinic in the coming years (http://​www.​bptc.​com/​sites/​default/​files/​biopulse_​reports/​adc_​survey_​results_​2011-11-23.​pdf). Selected ADC candidates are discussed here.

This conjugate is based on the well studied antibody Trastuzumab (Herceptin) that targets the HER-2 cell surface protein and which was approved in 1998 for use in patients with Her-2+ metastatic breast cancer. The extensive clinical experience with this antibody has aided in both selection of an appropriate drug (maytansine derivative DM1) and the conjugation chemistry. Following the positive results of several preclinical [60] and recent clinical trials [61], even in woman whose disease had progressed during naked Trastuzumab therapy [62], the developers (ImmunoGen/Genentech/Roche) plan to apply for FDA approval before the end of 2012.

The conjugate contains an IgG4 monoclonal antibody targeting the CD22 antigen found on mature B cells. It is also being developed by Pfizer that is using the same hydrazone linker–calicheamicin drug combination as employed for Mylotag. It was shown to be more effective in vitro than Mylotag in killing primary pediatric acute lymphoblastic leukemia cells [63]. Despite the high potency of the drug moiety, initial clinical studies suggest relatively low tolerable doses of the CMC-544 ADC [64]. CMC-544 is currently the subject of 15 clinical trials, including two Phase III combination trials studies with Rituxin.

The Lorvotuzumab humanized monoclonal antibody targets CD56, an isoform of neural cell adhesion molecule expressed on NK cells, some T-cells and on the majority of Multiple Myeloma and Small Cell Lung Carcinoma cells. Different chemistry has been used to prepare this ADC in that the antibody is coupled to mytansioid (DM1) via a linker cleavable by disulfide reduction. In 2010 it was granted orphan drug status both in the USA and Europe for Mantle Cell Carcinoma, and is currently also under clinical investigation for Multiple Myeloma, Small Cell Lung Cancer and Ovarian Cancer.

CD19 is expressed on follicular dendritic cells. It is a validated marker of developing B cells which is lost on their maturation to plasma cells. SAR3419 consists of the humanized huB4 anti-CD19 antibody conjugated to a different derivative of mytansinoid (DM4), but as in IMGN901, the disulphide linker is cleaved by reduction. In vitro studies show that this ADC is internalized by lymphoid cell lines and Phase I studies indicate it produces low hematological toxicity [65]. It is being clinically assessed for use in several forms of B-cell NHL.

There are more than a dozen other ADCs at various stages of preclinical and early clinical development. Unfortunately, many of these still induce side effects often seen in conventional chemotherapy such as several manifestations of myelosupression and nephritis and further research is needed in this respect before ADC can fulfill their promise of inducing less morbidity than carrier-free chemotherapy. Further clinical details on the activity of these agents can be found in recent reviews [54, 66‐68].

Studies using classical biochemistry, gene and protein array technologies and now systems biology have demonstrated differences in the metabolism of tumor cells compared to normal, as well as normal proliferating cells [69‐71], one consequence of which is a modified “membrane-ome”, manifested either as altered expression of differentiation and constitutive mature cell proteins or re-expression of neonatal ones. This phenomenon led to the discovery of an array of tumor associated cell surface antigens (TAAs) [72‐74] Confidence in the ability to utilize antibodies to identify TAAs was based in part on the success of immunologists over the years at producing antibodies to a variety of different molecules, even when these were considered weak antigens [75]; indeed an impressive battery of antibodies to cell surface markers is now available. Nonetheless, one take-home message from this massive effort is that the vast majority of TMAs only target TAAs. This was also evident from studies using molecular cloning techniques such as SEREX (serological analysis of autologous tumor antigens by recombinant cDNA expression) in which a patient's own serum is analysed for immunoreactivity against proteins expressed by their own tumor cells [72, 76]. Only rarely have true tumor specific antigens (TSAs) so far been demonstrated, particular across different patients, the best examples being clonotypic antibodies expressed on antigen-educated B-cell leukemic cells such as those found in B-cell lymphoma [77] and Multiple Myeloma although the prevalence of these among different patient cohorts such is still inconclusive [78]. Aside from these, the most tumor cell “restricted” antigen identified so far, with regards to approved TMAs, is human epidermal growth factor receptor 2 (HER-2 or Erb-B2), which is targeted by Trastuzumab. While TAA-directed TMAs clearly have clinical benefit, their lack of specificity for true tumor cell targets is intrinsically associated with side effects that limit their efficacy.

It is noteworthy that currently approved TMAs and many of those in advanced stages of clinical development were generated several years ago, using technologies that may inherently limit the scope of molecular or structural targets identified. For example discovery platforms based on traditional hybridoma technologies coupled with high-throughput or FACS screening methods tend to skew towards selection of antibodies targeting dominant epitopes [79]. In addition, epitope dominance also depends of the genetic background of the rodent strain used for immunization, although this variable is not often studied due to the incurring development time and costs. Other limiting factors include the accessibility of potentially useful epitopes within a native multi-molecular complex to bulky antibody proteins and the lower antibody binding affinities towards weaker antigens. It remains to be seen whether more recent technical improvements aimed at overcoming these variables [80] will yield clinically successful TMAs to novel tumor cell markers.

While there are some exceptions [81], the majority of anti-cancer drugs act intracellularly. With regards to ADCs this means that the antibody must bind a cell surface component in such a way that stimulates the formers' internalization, usually by receptor mediated endocytosis and then delivery of the ADC to lysosomes. In many cases, internalization was not a criteria included in the original selection of TMAs currently in clinical use, although subsequent studies showed that some of them, such as Transtuzumab, do induce uptake [82]. The requirement for ADC internalization makes the selection of TMAs even more complex. For example CD19, a validated B-cell marker, forms dimers with CD21 [83]. Whereas some anti-CD19 antibodies are rapidly internalized and are being used to develop ADCs ([65] and see above), the uptake of others is inhibited by CD21 expression [84]. Similarly there are conflicting reports regarding anti-CD20 antibodies [82, 85]. In addition, studies with anti-HER-2 antibodies suggest that internalization is also related to affinity. For example Rudnick and colleagues recently reported that higher affinity antibodies were internalized and degraded faster than moderate affinity antibodies, thus limiting their tumor penetration [86]. These studies underline the importance of early stage selection of antibody clones with the appropriate functionality.

A number of studies have investigated the complex pharmacokinetics and pharmacodynamics of TMAs, which tend to be non-linear as compared to small molecule drugs [52, 87‐89]. The optimal values for these parameters can vary widely between antibodies as they depend on a wide variety of variables including heavy chain isotype, type and degree of glycosylation, route of administration, rate of absorption, antibody affinity, the location, cell surface density and turnover of antigen, charge, immunogenicity, ability to bind the FcRn receptor, effector function and degree of internalization.

One area that has received much attention in ADC development is the chemistry of drug attachment. Factors important here include selection of a linker attachment site that retains antibody activity, linker length and composition and the design of drug analogs for attachment to the linker [51, 53, 54, 82, 90]. Two methods are commonly used for conjugating drugs to antibodies: alkylation of reduced interchain cysteine disulfides through an enzymatically non-cleavable maleimido or simple and cleavable disulfide linker (Figure 1a) and acylation of lysines by cleavable linear amino acids (Figure 1b). Spacers are usually essential extensions of the drug linkage and are responsible for avoiding the shielding of the active site of the antibody as well as improving solubility properties of ADCs (for example in by the use of polyethylene glycol). Cathepsin-cleavable linkers are also utilized (for example Val-Cit, or Phe-Lys) bound to self-emulative moiety PABA (p-aminobenzyl alcohol), enabling selective drug release in cancer cells [54], Notably, the linkage technologies used in ADCs are also applicable in PDCs enriching their conjugation repertoire as will be discussed later due course There are 8 interchain cysteines and up to 100 lysines available for conjugation on IgG1 antibodies and conjugation to these sites results in heterogeneous mixtures. Cysteine conjugates provide a greater degree of uniformity than lysine-based conjugates [91, 92] while recombinant methods in which cysteines are introduced into the antibody backbone at specific sites result in still more uniform conjugation [93, 94]. In some instances, it has been observed that the location of the conjugated drug is not as important as the stoichiometry of drug attachment, although this is difficult to ensure [91, 93]. ADCs with two to four drugs per antibody are generally superior to more heavily loaded conjugates that tend to be cleared very rapidly from the circulation [95]. Nonetheless, it has been proved to be difficult to use chemical methods to prepare ADCs with a predefined (2, 3 or 4) number of drug molecules per antibody. Because the drugs are often conjugated to the side chains of reactive lysine or cysteine residues of the antibody, a distribution of products results in variable numbers of drug moieties attached to various sites. Thus, the formation of reproducible ADCs is difficult. Protein engineering has been used to circumvent these issues, such as inserting additional cysteine residues [94], or replacing solvent accessible cysteines with serines [93], although the drug loading stoichiometry still varies markedly so it is still difficult to control the number and regioselectivity of conjugated drug molecules.

There is usually at least a 3 log₁₀ difference in molecular weight between antibody and the linker-drug moiety. Therefore the conjugation of only 2–4 drug molecules produces a situation in which it is not a trivial task to separate the conjugated ADC from the unconjugated antibody. For example is was reported that in the formulation of the now discontinued Gemtuzumab Ozogamicin ADC (Mylotarg), only 50% of the anti-CD33 antibody was bound to 4–6 drug molecules. The remaining antibody was unconjugated [55]. This study underscores two problems. First, the presence of unconjugated antibody would presumably undermine the efficacy of the ADC. Second, the molar ratio of antibody:drug is an average value, indicating that the drug conjugation process produces a heterogeneous mixture of products with variable efficacy (discussed further below). This mixture would also not be exactly reproducible between batches.

The limited penetration of full length antibodies into solid tumors is a recognized factor restricting their efficacy. Their large molecular size effects the rate of mass transfer due to higher interstitial pressure and hypoxia caused by leaky vasculature in the interior of the tumor mass, factors that can also effect levels of antigen expression [96‐98]. It has however been possible to select antibody clones according to their degree of tumor penetration [86, 99]. Interestingly, in those studies truncated (single chain variable fragment, scFv) and full length antibody penetration into the tumor mass was inversely related to the antibody affinity and internalization rate of the target antigen. These results are counterintuitive to the traditional strategy of developing high affinity antibodies and again highlight the need for a deep understanding of the biology of immunotherapy when selective TMA candidates. These issues seem less pertinent for micrometastases and even more so for hematological tumors such as CLL where the malignant cells are blood borne.

The current limited levels of TMA efficacy require the use of significant quantities of product (150-350mg/m² single dose) over multiple treatments. This situation has stimulated an impressive expansion and improvement in the whole gamut of upstream and downstream processes involved in antibody manufacture (reviewed in [14]), including animal cell recombinant expression systems, cell culture media, cell growth conditions and product yield/cell. Several important challenges remain in the area of product quality control such as the production of aggregates [100] and variation in glycosylation [101]. Moreover, further improvements are urgently needed in bioreactor design and protein purification processes if overall manufacturing costs is to be lowered to enable TMAs and ADCs to be affordable for patients who need them [14, 102].

TMAs have already proven their clinical effectiveness and it is predicted that the coming two years will see 1–2 ADCs receiving FDA approval, at least one of which will be for hematological cancers. Nonetheless, the points raised above indicate that both TMAs and ADCs have some inherent limitations that are not easily overcome by technology. In order to broaden the scope of effective TDD therapies it is therefore prudent to search in parallel for additional strategies that would complement the use of ADCs for cancer therapy.

Peptide display technologies do not rely on the antigenicity of the target molecule. Therefore it is not necessary to develop specialized immunization protocols to generate binding ligands to weak antigens. Also, depending of the technology used, the displayed peptides can be structurally extended from the display platform, allowing them for freedom to bind locations on single molecule or complex target binding sites that might be sterically hindered from more bulky antibodies. This has already generated the discovery of novel peptide ligands of cancer cell surface targets [121].

The important question of appropriate carrier-linker-drug design and synthesis is easily approached with PDCs. As with full length antibodies, the site of conjugation between peptide and drug molecule can have a profound effect on maintaining peptide binding affinity, drug activity and conjugate stability. However exact knowledge of the peptide sequence and the amino acids responsible for maintaining high binding affinity often allows a higher degree of flexibility in the design of linker length, its composition and conjugation chemistry to the drug. Compared to antibody-linker conjugation, peptide conjugation can then include a wide range of chemistries encompassing amides, carboxylic acid esters, hydrazones, thioethers and carbamates (for more details see [120]. Aside from small molecule drugs, peptides can also be easily conjugated to transition metals. Mononuclear and dinuclear platinum complexes tethered to an α9β1-integrin targeting peptide and a nuclear-localisation peptide have been synthesized using solid-phase synthesis [122] (see Figure 2). The cellular uptake, DNA binding and cytotoxicity of the complexes was monitored in a number of different cell lines. Some of these conjugates exerted remarkable targeted cytotoxicity. However, linkage of platinum complex to antibody is very problematic due to the number of competing Lys amino groups situated across the antibody molecule.

An additional advantage with random peptide libraries is the ability to select combinatorial cyclic peptides which may demonstrate higher selectivity for the target and longer stability than linear peptides because of conformational restriction [123, 124]. There is also is flexibility in cyclization strategy [120] in which an amide bond formed between N- and C-termini or the disulphide bond formed between two Cys residues can be engineered outside of the receptor recognition sequence and adjacent to a linker moiety.

For the most part the molecular targets of anti-cancer drugs are intracellular, a trend likely to continue with the explosion of new information in areas such as signaling pathways, epigenetics, DNA repair mechanisms, cell cycle regulation and mitochondrial metabolism. Therefore it is imperative that carrier-drug complexes induce efficient cellular uptake, preferably through receptor-mediated endocytosis (RME). This will initially traffic the conjugate to endosomes/lysosomes, induce enzymatic or chemical degradation of linker and release active drug into the cytoplasm [125, 126]. This process has been extensively studied, with the case of transferrin being only one example [127‐130]. It should be noted that peptide carriers can also be taken up non-specifically by pinocytosis [130], however these peptides can be selected out by testing for uptake on appropriate receptor negative control cells. Not only does the RME pathway ensure that the drug payload is delivered intracellularly, but this form of uptake has been shown to one pathway to bypassing the development of drug resistance, a strategy that should be further exploited to enhance the clinical efficacy of drugs already in the clinic [131], (Gellerman et al., submitted).

There are many aspects of carrier-drug uptake/processing pathway to be considered, the first of which is the ability of the carrier itself to bind and activate the receptor in such a way as to induce efficient RME. Peptide display libraries afford a higher statistical chance of presenting at least some candidates in the (novel?) structural orientation to engage the receptor binding pocket correctly and the library platforms more easily lend themselves to high throughput screening for this specification. Furthermore internalizing carrier peptides can be selected from these libraries without prior knowledge of receptor structure or biology. This is important as cell surface receptors vary widely in their cell surface density, turnover time, RME kinetics and recycling and trafficking routes.

Peptide carriers are in the order of a hundredth the mass of antibodies [132] and this is clearly an advantage to overcoming the interstitial tumor pressure which can limit antibody transport into the tumor interior [133, 134]. Two examples that emphasize this point are the identification of peptides containing a CendR motif that enhance penetration into tumor tissue and internalization into cells [135] and the use of tumor-vasculature homing peptides containing the CD13 receptor binding motif Asn-Gly-Arg [136]. Finally, while members of the cell-penetrating peptide family can [137], but usually do not demonstrate tumor tissue specificity [138], they can be grafted to a variety of targeting peptides to enhance the efficacy of TDD [139‐142].

There are now several dozen peptide therapeutics approved for clinical use and the market for therapeutics peptides is estimated to reach $11.5billion by 2013 (BioNest). Over the last decade or so, processes in peptide manufacture have considerably evolved, partly due to the experience of scaling up production from milligram to multi-ton levels [143, 144]). These requirements have generated important improvements in both technical and cost efficiency aspects of process design, manufacturing capacity and peptide synthesis [145]. With regards to synthesis, both recombinant and chemical synthetic pathways are available and while traditionally the latter approach had the advantage of being able to insert unnatural amino acids that can increase stability and flexibility in conjugation to drugs, this division is now becoming blurred [146].

As mentioned briefly above, natural peptide drugs are limited by their sensitivity to enzymatic degradation, extensive renal filtration, and nonspecific uptake into tissues such as the liver, all of which resulted in reduced bioavailability and reduced half life in the circulation (minutes) as compared to antibodies (reviewed in [147]. Many of the shortcomings have been overcome by different synthetic chemistry strategies, resulting in the development of peptidomimetics. These include incorporation D-enantiomeric amino acids and replacement of alanine residues to increase resistance to gut and serum proteases and coupled the peptide to polyethylene glycol (PEGylation) to reduce liver and kidney elimination [105, 147].