Ligand-Targeted Drug Delivery - ACS Publications - American

Sep 12, 2017 - receptor's natural ligand-binding site for drug targeting, in some cases design of a ligand to fit an allosteric or nonfunctional pocke...
18 downloads 15 Views 8MB Size
Review pubs.acs.org/CR

Ligand-Targeted Drug Delivery Madduri Srinivasarao and Philip S. Low* Purdue Institute for Drug Discovery, Purdue University, West Lafayette, Indiana 47907, United States ABSTRACT: Safety and efficacy constitute the major criteria governing regulatory approval of any new drug. The best method to maximize safety and efficacy is to deliver a proven therapeutic agent with a targeting ligand that exhibits little affinity for healthy cells but high affinity for pathologic cells. The probability of regulatory approval can conceivably be further enhanced by exploiting the same targeting ligand, conjugated to an imaging agent, to select patients whose diseased tissues display sufficient targeted receptors for therapeutic efficacy. The focus of this Review is to summarize criteria that must be met during design of ligand-targeted drugs (LTDs) to achieve the required therapeutic potency with minimal toxicity. Because most LTDs are composed of a targeting ligand (e.g., organic molecule, aptamer, protein scaffold, or antibody), spacer, cleavable linker, and therapeutic warhead, criteria for successful design of each component will be described. Moreover, because obstacles to successful drug design can differ among human pathologies, limitations to drug delivery imposed by the unique characteristics of different diseases will be considered. With the explosion of genomic and transcriptomic data providing an everexpanding selection of disease-specific targets, and with tools for high-throughput chemistry offering an escalating diversity of warheads, opportunities for innovating safe and effective LTDs has never been greater.

CONTENTS

References

1. Introduction 2. Elements in the Design of an Optimal LigandTargeted Drug 2.1. Selection of a Disease-Specific Receptor 2.1.1. Receptor Expression Profile 2.1.2. Receptor Location 2.1.3. Receptor Internalization 2.1.4. Receptor Topography 2.1.5. Receptor Competition 2.2. Selection of a Targeting Ligand 2.2.1. Ligand Size 2.2.2. Ligand Adaptability to Diverse Receptor Topographies 2.2.3. Ligand-Binding Affinity and Specificity 2.2.4. Ligand Conjugation Site Chemistries 2.2.5. Ligand Stability 2.2.6. Ligand Immunogenicity 2.2.7. Ease and Cost of Manufacturing 2.3. Selection of Spacers and Cleavable Linkers 2.3.1. Length of Spacer 2.3.2. Hydrophilicity of Spacer 2.3.3. Cleavability of Linker 2.4. Selection of an Optimal Payload 2.4.1. Biologic Properties of Optimal Payload 2.4.2. Chemical Properties of Optimal Payload 3. Ligand-Targeted Therapeutic Agents in Human Clinical Trials 4. Conclusions Author Information Corresponding Author ORCID Notes Biographies © 2017 American Chemical Society

12154

12133

1. INTRODUCTION The world is now entering the age of precision and personalized medicine, where drugs will be tailored for each patient’s needs and then delivered specifically to diseased but not healthy cells.1 One attractive strategy for achieving the desired pathologic cell specificity is to link an established therapeutic or imaging agent to a targeting ligand that can deliver the attached cargo selectively to the pathologic cell.2−4 As shown in Figure 1, ligand-targeted drugs (LTDs) of this type are constructed of a targeting ligand linked via a spacer and cleavable linker to a therapeutic or diagnostic payload. In this design, the efficacy/potency of the conjugate is primarily determined by the activity of the payload, while the safety of

12134 12134 12134 12136 12138 12139 12139 12140 12140 12142 12143 12143 12144 12145 12145 12146 12146 12146 12147 12151 12151 12152 12153 12153 12153 12153 12153 12153 12154

Figure 1. Design of an ideal ligand-targeted drug (LTD) will involve the tethering of a targeting ligand, spacer, cleavable linker, and therapeutic payload as shown above. Received: January 7, 2017 Published: September 12, 2017 12133

DOI: 10.1021/acs.chemrev.7b00013 Chem. Rev. 2017, 117, 12133−12164

Chemical Reviews

Review

the conjugate is dictated by the specificity of the targeting ligand for the diseased cell type. By separating diseased cell selectivity and therapeutic activity into separate parts of the conjugate, the two most critical properties that determine FDA approval can be independently optimized without compromising the performance of the complementary part. Targeted drugs have several advantages over their nontargeted counterparts. First, targeted drugs can deliver their therapeutic payloads selectively into diseased cells, thereby avoiding nonspecific uptake and the associated toxicity to healthy cells (vide infra). Second, targeted drugs can enable the use of highly potent therapeutic warheads that exhibit no efficacy when administered in nontargeted form at the maximum tolerated dose. For example, the FDA-approved antibody−drug conjugate trastuzumab emtansine (T-DM1) has a reported maximum tolerated dose (MTD) of 40 mg/kg and 30 mg/kg in rats and monkeys, respectively, whereas the free drug, DM1, has an MTD of only 0.2 mg/kg.5 Similarly, the MTD of FDA-approved monomethylauristatin E−brentuximab is 100 mg/kg in mice, whereas the MTD of the free drug is only ∼1 mg/kg.6,7 In the case of the low-molecular-weight targeting ligand folate, its conjugate with desacetylvinblastine hydrazide (EC145, vintafolide) was tolerated up to ∼19 mg/kg in mice, whereas the MTD of nontargeted desacetylvinblastine hydrazide is only ∼0.8 mg/kg.8 Not surprisingly, folate-targeted desacetylvinblastine displayed acceptable safety9 (Table 1) and efficacy (Figure 2) in phase 1 and 2 human clinical trials even though the nontargeted form of desacetylvinblastine was found to be unsafe. A third advantage is that the targeting ligands can be simultaneously exploited to generate a companion diagnostic agent that can be used to select patients whose pathologic cells overexpress the targeted receptor (which can include any cellsurface protein, lipid, or carbohydrate for which a targeting ligand can be found).11 In the case of LTDs that target the folate receptor, folate-linked 99mTc and 68Ga radioimaging agents have been successfully employed to identify patients whose tumors express sufficient folate receptors to allow uptake of a therapeutic quantity of drug (Figure 3).12−17 Fourth, targeted drugs can often be administered at lower doses than their nontargeted counterparts, because receptor binding and internalization can concentrate the conjugates inside the receptor-positive pathologic cells. In this Review, we will summarize the overriding principles that govern the successful design of LTDs. In many cases these design features will be similar for drugs targeted with ligands as diverse as antibodies, aptamers, protein scaffolds, oligonucleotides, peptides, oligosaccharides, and small organic molecules. In other cases, differences in ligand size, shape, hydrophilicity, stability, etc. will require different linker and/or payload chemistries. These similarities and differences along with their consequences will be outlined later as we describe the functional and structural features of an optimal LTD.

Table 1. Safety and Efficacy of a Folate-Targeted Chemotherapeutic Agent in Human Clinical Trialsa,d maximum grade system organ classb/preferred termc blood and lymphatic system disorders anaemia gastrointestinal disorders abdominal pain constipation ileus nausea general disorders and administration site conditions asthenia fatigue metabolism and nutrition disorders decreased appetite musculoskeletal and connective tissue disorders back pain nervous system disorders coordination abnormal peripheral sensory neuropathy

grade 3 (N = 196)

grade 4 (N = 196)

5 (2.6%)

0 (0.0%)

4 5 3 2

0 0 0 0

(2.0%) (2.6%) (1.5%) (1.0%)

(0.0%) (0.0%) (0.0%) (0.0%)

2 (1.0%) 10 (5.1%)

0 (0.0%) 0 (0.0%)

2 (1.0%)

0 (0.0%)

2 (1.0%)

0 (0.0%)

2 (1.0%) 3 (1.5%)

0 (0.0%) 0 (0.0%)

a

A folate−desacetylvinblastine hydrazide conjugate (EC145, vintafolide) has shown minimal toxicity as a single agent in phase I and phase II studies of 196 cancer patients, even though the nontargeted desacetylvinblastine was found to be lethal. bMedical Dictionary for Regulatory Activities (MedDRA) system organ classes and preferred terms were used and are listed in alphabetical order. cThe number (%) of participants by grade represents the maximum intensity of events reported by the participants; the investigators used the Common Terminology Criteria for Adverse Events (CTCAE) to grade the severity of adverse events. dAdapted with permission from Endocyte, Inc.

to find, relative estimates of mRNA and protein levels in multiple pathologic and normal tissues is becoming increasingly available from an ever-expanding number of relevant databases. Some of our favorite databases for mining this type of information are the Human Protein Atlas and Oncomine.19−21 From our experience, a 3-fold or higher magnitude of receptor overexpression in pathologic tissues is usually sufficient to avoid unacceptable toxicity to healthy tissues.22−24 However, when the drug conjugate’s payload is largely nontoxic (e.g., as in the case of most imaging agents and some therapeutic drugs), a lower diseased-to-normal tissue ratio may be acceptable. Similarly, if the concentration dependence of payload toxicity is highly cooperative, i.e., such that a transition from nontoxic to unacceptable toxicity occurs over a narrow concentration range, a dose of targeted drug can sometimes be found that will modify a receptor-rich diseased cell without causing toxicity to a receptor-poor healthy cell. In the case of imatinib’s inhibition of Plasmodium falciparum malaria, for example, the transition from 0% to nearly 100% inhibition occurs over only a 5-fold change in drug concentration, allowing a nontoxic dose of imatinib to be identified.25 Because some receptors will obviously also be expressed in one or more healthy cell types, the question naturally arises regarding the impact of the LTD on these healthy cell types too. Briefly, one must anticipate that unless significant differences in LTD penetration into the healthy and diseased tissues exist, the distribution of the targeted drug must generally

2. ELEMENTS IN THE DESIGN OF AN OPTIMAL LIGAND-TARGETED DRUG 2.1. Selection of a Disease-Specific Receptor

2.1.1. Receptor Expression Profile. A major determinant of the safety profile of a targeted drug lies in the ratio of targeted receptor expression in diseased versus normal tissues. Although quantitative information on receptor expression in various healthy and diseased tissues typically has been difficult 12134

DOI: 10.1021/acs.chemrev.7b00013 Chem. Rev. 2017, 117, 12133−12164

Chemical Reviews

Review

Figure 2. Overall survival data from a phase II nonsmall-cell lung cancer clinical trial showed almost doubling of survival from 6.6 months (for docetaxel alone) to 12.5 months (for the combination of docetaxel plus vintafolide) in the adenocarcinoma subset.10 Adapted with permission from Endocyte, Inc.

administered in combination with one or more nontargeted drugs that have been approved for the same indication.10,27 As long as the toxicities of the component drugs in the combination therapy do not overlap and synergize, a therapeutically effective dose can often be achieved without decreasing the MTD.28−30 A third criterion to consider when selecting a receptor for ligand-targeted drug delivery depends on the investigator’s ability to identify a receptor isoform that is only overexpressed on the target cell. For example, although multiple α- and βadrenergic receptor isoforms bind epinephrine, only the β3adrenergic receptor is expressed on adipocytes.31 Thus, if the investigator’s objective was to deliver a drug specifically to adipocytes, identification of a receptor, such as the β3adrenergic receptor, together with a β3-specific ligand, should reduce uptake of the drug by the many cells that express other β-adrenergic receptor isoforms. Indeed, because many receptors exist in multiple isoforms, identification of a pathology-specific isoform and the design of an isoform-specific ligand can greatly improve diseased cell specificity.32−35 Receptors that are only induced upon cell activation or differentiation can also add specificity to drug targeting. For example, the beta isoform of the folate receptor (FRβ) is only expressed on activated macrophages (i.e., not on resting macrophages or any other major cell types), allowing selective drug delivery to the inflammatory subset of macrophages/ monocytes (Figure 4A).36 Because these FRβ-positive myeloid cells are almost exclusively found in inflamed tissues, FRβspecific drug targeting can avoid any drug-related toxicities to tissue resident and other noninflammatory macrophages in other tissues.36 Moreover, because FRβ-positive macrophages constitute major contributors to most autoimmune and inflammatory diseases (e.g., rheumatoid arthritis, Crohn’s disease, psoriasis, ulcerative colitis, sarcoidosis, Sjogren’s syndrome, idiopathic pulmonary fibrosis, osteoarthritis, multiple sclerosis, etc.), ligands that target FRβ can be exploited to concentrate imaging and therapeutic agents specifically in the most inflamed tissues (Figure 4B−D).37−44 Mutant forms of a receptor that are solely expressed on a pathologic cell type can also constitute excellent targets for ligand-mediated drug delivery.46,47 For example, a subset of

correspond to the level of targeted receptor expression. This obviously argues that some toxicity will be experienced by the healthy tissues that express receptor (unless they are insensitive to the warhead). However, when it is appreciated that a nontargeted drug will usually distribute similarly into healthy and diseased tissues alike, any preferential uptake of the targeted drug afforded by overexpression of the targeted receptor in the pathologic tissues would seem to favor use of a targeted over a nontargeted drug. A second important criterion in receptor selection concerns the absolute level of receptor expression in the diseased cell type. Thus, if the desired therapeutic outcome can only be achieved when the intracellular concentration of the drug exceeds a minimum threshold, sufficient receptor numbers must be present on the targeted cell to enable receptormediated drug delivery to exceed this threshold. As an illustration, if the in vitro IC50 of a drug were found to be 10 nM and the pathologic cell was assumed to have a volume of 4 000 μm3 (i.e., a cell of diameter ∼20 μm), the cell would have to express >72 000 receptors in order for sufficient drug to be internalized to yield complete inhibition of the targeted enzyme’s activity, i.e., assuming 100% receptor occupancy and quantitative internalization of occupied receptors by the target cell.2 However, if the targeted receptor in the above example were to recycle at a rate of 3 times per day and the receptor were fully saturated each time it reentered the cell by endocytosis, the same nearly quantitative enzyme inhibition might be reached at minimum of 24 000 receptors/cell, i.e., assuming complete unloading of the drug during each endocytic event. In those not infrequent situations where the targeted receptor is expressed in insufficient numbers to deliver a therapeutic dose of drug to the pathologic cell type, two options still remain. First, a second disease-specific receptor can be targeted with the same (or a synergistic) warhead, with the hope that the two receptors can collectively deliver an effective dose of the drug.26 This dual targeting strategy, while economically challenging, can also add specificity to the therapy, because only cells that express a sufficient number of both disease-specific receptors will receive an effective dose of the drug. Second, and more commonly, the LTD can be 12135

DOI: 10.1021/acs.chemrev.7b00013 Chem. Rev. 2017, 117, 12133−12164

Chemical Reviews

Review

might be acceptable if the therapeutic warhead were to be released solely in diseased cells, currently available chemistries (see below) do not permit such diseased cell-specific drug release, i.e., assuring that some drug discharge will also occur in receptor-negative healthy cells. This off-target drug delivery will, of course, lead to off-target toxicity. Second, while ligands that naturally bind intracellular receptors must be intrinsically membrane-permeable, upon conjugation to their therapeutic payloads their membrane permeabilities can be significantly compromised, i.e., rendering the LTD much less active. Third, drug conjugates that can be directed to cell-surface exposed receptors can be engineered to be membrane-impermeable to receptor-negative cells (see below), thereby assuring that drug uptake must be facilitated by the disease-specific receptor. While the requirement that a disease-specific receptor be exposed on the cell surface reduces the repertoire of receptors from which one may choose, there are still many classes of receptors that can be exploited for such drug delivery.54−59 First, some proteins that are normally only expressed inside healthy cells can become exposed upon disease transformation, rendering them ideal targets for disease-specific drug delivery. For example, cell-surface expression of the intracellular chaperone HSP90 occurs primarily on certain cancers and inflamed cells, allowing HSP90-targeted drugs to concentrate in malignant and inflamed tissues without harming healthy tissues.60−63 Extracellular histones have also been exploited for drug targeting, because exposure of the nucleosome protein on the cell surface is limited to cells involved in sepsis, peritonitis, ischemia-reperfusion injury, pancreatitis, and stroke.64−74 Cell-surface translocation of TNFα converting enzyme (TACE) is similarly seen primarily in inflamed tissues, allowing TACE-specific ligands to be used for drug delivery to tissues involved in rheumatoid arthritis, idiopathic pulmonary fibrosis, cirrhosis of the liver, etc.75−78 Another class of cell-surface receptors that can constitute ideal targets for drug delivery are receptors whose accessibility in normal tissues is either limited or absent. For example, a number of receptors that are abundantly expressed in the healthy brain are overexpressed extracranially only in neuroendocrine tumors or sites of inflammation. Because a healthy blood−brain barrier will normally prevent LTD entry into the brain, the only accessible receptors of this type will be found in the pathologic tissue. One example of a receptor of this type is the neurokinin 1 receptor, which is present in both normal brain and most neuroendocrine cancers.79,80 A second is the cell-surface enzyme termed prostate specific membrane antigen (PSMA), which is prominently expressed in healthy brain tissue but best known as a specific marker for prostate cancer.81 Indeed, the best prostate cancer imaging agents available to date are LTDs that are targeted to PSMA and show no uptake in the healthy brain.82−84 A slightly modified version of the above scenario arises in the cases of receptors that are normally expressed solely on the apical surfaces of epithelial cells (Figure 5). Although such receptors are almost universally inaccessible to parenterally administered drugs (i.e., due to the barriers imposed by the tight junctions between epithelial cells that prevent leakage of interstitial fluids into apposing luminal compartments),85 they become readily accessible upon malignant transformation of the epithelial cells due to the universal loss of tight junctions during the transformation process. Thus, ligands that are designed to bind apically restricted proteins on epithelial cells will generally constitute ideal tumor-targeting ligands, because such receptors

Figure 3. Ability of a folate-targeted radioimaging agent to predict response to a folate-targeted chemotherapy in ovarian cancer patients. (A) 99mTc-etarfolatide (a folate-targeted molecular imaging agent) accumulates in FR-positive (top row, panel A) but not FR-negative lesions (bottom row, panel A), allowing 99mTc-etarfolatide to be used to select patients for therapy with a folate-targeted therapeutic agent (i.e., vintafolide). (Top) CT image (left) shows a cancer lesion. The corresponding 99mTc-etarfolatide SPECT image (right) shows uptake in the target lesion (yellow circle). (Bottom) CT image (left) shows a cancer lesion. However, the corresponding 99mTc-etarfolatide SPECT image (right) shows no uptake in the target lesion. CT = computed tomography; SPECT = single-photon emission computed tomography; FR = folate receptor. Data in (A) adapted with permission from Endocyte, Inc. (B) Best change in tumor size of FR-positive and FRnegative lesions. One negative lesion had an increase of 450% but was truncated to 150% for display purposes. Folate receptor expression in a patient’s cancer correlates with their tumor’s response to vintafolide in a phase II ovarian cancer trial. These data suggest that the FR-targeted imaging agent can serve as a companion diagnostic agent to select patients for an FR-targeted therapy.18 (B) Adapted with permission from ref 18. Copyright 2014 Oxford University Press.

glioblastoma multiforme cells overexpress a mutated epidermal growth factor receptor, termed variant III, that is likely involved in promoting the malignant cell’s proliferation. Any ligand (e.g., an antibody) that can selectively target this cancer cell variant should enable selective drug delivery to glioblastoma multiforme cells.47,48 2.1.2. Receptor Location. Although both intracellular and cell-surface exposed receptors can be overexpressed in pathologic cells,49−52 we prefer to exploit primarily exoplasmically exposed receptors for drug targeting for three reasons. First, LTDs that target intracellular receptors (e.g., steroid hormone, retinoic acid, vitamin D receptors, etc.) must be intrinsically membrane-permeable in order to reach their intracellular targets, i.e., requiring that they will also be able to enter healthy cells.53 While such nonspecific cell penetration 12136

DOI: 10.1021/acs.chemrev.7b00013 Chem. Rev. 2017, 117, 12133−12164

Chemical Reviews

Review

Figure 4. Receptors that are only expressed upon target cell stimulation constitute good disease-specific receptors for drug targeting. (A) As an example, macrophages only express FRβ upon activation to either a pro-inflammatory (M1) or an anti-inflammatory (M2) phenotype, i.e., allowing folate ligands to be used for drug delivery to autoimmune and inflammatory diseases. (B) Synovial cells from 4 patients with rheumatoid arthritis were labeled with anti-CD11b antibody to stain human macrophages and then incubated with folate-FITC (fol-FITC) in the absence (left panel) and presence (right panel) of 100-fold excess free folic acid. The flow cytometry data clearly show FRβ expression in a subset of the synovial macrophages of rheumatoid arthritis patients. (C) Radioimages of the hands (left panel) and feet (right panel) of a patient with active rheumatoid arthritis, showing sites of uptake of a 99mTc-etarfolatide, in multiple joints of the hands and feet.45 (B, C) Adapted with permission from author’s own work, ref 45. Copyright 2009 American Society of Hematology. (D) Representative images of immunohistochemistry staining of FRβ in sections of Crohn’s disease and ulcerative colitis. FITC = fluorescein isothiocyanate; IHC = immunohistochemistry.

Figure 5. Receptors that are normally expressed exclusively on the apical surface of polarized epithelial cells constitute good receptors for cancerspecific drug targeting. Thus, receptors that are normally expressed only on the apical side of a healthy polarized epithelium are inaccessible to parenterally administered drugs. However, upon malignant transformation, the tight junctions between epithelial cells collapse, allowing the apically restricted receptors to redistribute to all surfaces of the cancer cell. As a consequence, those receptors that are normally confined to the apical surfaces of healthy epithelial cells can only be targeted with LTDs when the epithelial cell undergoes malignant transformation. Because 80% of human cancers derive from epithelial cells, apically restricted proteins represent good targets for LTD-mediated drug delivery.

In addition, targeted receptors whose ectodomains are cleaved or shed into extracellular fluid may not constitute ideal receptors because these cleavage fragments can serve as decoy receptors that inhibit LTD binding to their intact counterparts on the targeted cells. For example, while fulllength Mucin 1 (Muc 1) is highly upregulated in many solid tumors, its ectodomain is cleaved into a membrane-bound Muc1-C domain and free-floating Muc1-N fragment.88,89 While most of the antibodies and aptamers have been designed to target the cancer cell bound Muc 1-C fragment, those that bind to a released epitope will suffer from reduced cancer cell

are inaccessible in healthy tissues but readily accessible in cancer tissues. Certain mucins, such as MUC1 and MUC16 (i.e., CA125), for example, have proven to be excellent tumorspecific targets, because they are almost completely inaccessible in their apical locations on healthy epithelial cells but invariably accessible on many cancers of epithelial origin.85−87 Because such cell polarity is absent from cells/tissues of nonepithelial origin, receptors to be targeted on derived nonepithelial cancers such as sarcomas, leukemias, and lymphomas will be uniformly accessible in both healthy and malignant tissues alike. 12137

DOI: 10.1021/acs.chemrev.7b00013 Chem. Rev. 2017, 117, 12133−12164

Chemical Reviews

Review

Figure 6. Trafficking of internalized LTDs through intracellular compartments. Upon internalization, some LTDs can traffic through intracellular organelles such as early endosomes and compartments for uncoupling of receptor and ligand (CURLs), where conjugates are dissociated from their receptors prior to recycling of the empty receptors back to the cell surface. If the therapeutic/diagnostic payload is membrane-permeable and attached to the targeting ligand via a cleavable linker, it can release from the linker and diffuse out of the endosome into the cytoplasm where it can perform its function. In other cases, LTDs can traffic to lysosomes, where they are degraded at low pH by lysosomal enzymes, via late endosomes, allowing the released drugs to enter the cytoplasm as noted earlier. In cases where the therapeutic cargo is not membrane-permeable, additional mechanisms must be introduced to enable the released drug to escape the organelle and enter the cytoplasm.

targeting in patients where many of the Muc 1 copies have been cleaved.90,91 2.1.3. Receptor Internalization. Most cell-surface receptors are internalized into endosomes either during normal membrane recycling or in response to stimuli,92,93 enabling such receptors to ferry bound ligands into various intracellular compartments (Figure 6). In some cases (e.g., folate receptor alpha), the receptors are internalized constitutively, regardless of receptor occupancy or aggregation.94 While in most cases endocytosis of the targeted drug should improve its potency, in a few cases retention of the drug on the cell surface might be preferred. Such a situation arises when the drug’s site of action is directly on the cell surface. For example, folate-linked immunogenic haptens have been found to bind folate receptorpositive cancer cells and thereby “paint” the malignant cell’s surface with an immunogenic molecule.95−97 If the immunogenic hapten remains on the tumor cell surface for a prolonged period of time, it can trigger an attack by the immune system if it has been sensitized to the foreign hapten. While such strategies to label diseased cells as “foreign” have proven effective in mediating elimination of receptor-positive pathologic cells,97 their efficacies are significantly compromised if the ligand-linked hapten is rapidly internalized by the target cell.98 In the majority of cases, i.e., when receptor internalization is desired, selection of the optimal frequency for conjugate administration can be estimated from the recycling rate of the targeted receptor into intracellular endosomes and back to the cell surface.99−101 Thus, administration of an LTD more often than the rate of reemergence of unoccupied receptors at the cell surface cannot increase conjugate uptake but instead will aggravate off-target toxicity due to rejection of the LTD by the saturated target tissue and the consequent nonspecific drug release in healthy tissues. In contrast, administration of the conjugate less frequently than receptor reappearance on the target cell surface will not exploit the full capacity of the receptor delivery system. Thus, in cases where drug potency is

marginal, maximization of drug delivery by synchronization of LTD administration with the return of empty receptors to the cell surface can improve net uptake. To determine this rate of receptor recycling in vivo, a simple analysis of the rate of reappearance of unoccupied receptors at the cell surface can be performed as described previously.100,102 In the case of FRtargeted drug delivery to cancer cells, an optimal tumor-tobackground ratio was observed when the folate−drug conjugate was dosed every ∼13 h in L1210A tumors that recycle their receptors every ∼6 h and every ∼40 h in M109 tumors that recycled their receptors every ∼20 h.100 Although unanticipated, we have also observed that a noninternalizing receptor can sometimes be successfully exploited for intracellular drug delivery if certain conditions are met.103 First, the therapeutic warhead must be releasable from its conjugate by an extracellular process/reaction unique to the diseased tissue microenvironment.104−106 Second, the receptor must retain its bound conjugate for sufficient time to allow release of the therapeutic warhead within the diseased tissue. Third, the therapeutic warhead must be capable of passively diffusing across the diseased cell plasma membrane and into its cytoplasm. Finally, the size of the disease lesion must be sufficiently large to ensure that any released warhead will have to pass through many diseased cells before it can reenter general circulation. For example, some tumors secrete large quantities of reducing agents that can cleave a disulfide bond in a receptor-bound LTD, thereby releasing the attached drug at the cancer cell’s surface.107,108 If the tumor mass is large and the path for drug diffusion/convection out of the tumor and into normal tissues is long and slow, the probability that the released drug will enter and kill a cancer cell will be greater than its probability of damaging a normal cell. A relevant example of this tumoricidal mechanism involves an LTD targeted to the cholecystokinin 2 receptor (CCK2R) that is overexpressed in gastrointestinal stromal tumors. Although the CCK2 receptor-targeted LTD is not internalized, significant 12138

DOI: 10.1021/acs.chemrev.7b00013 Chem. Rev. 2017, 117, 12133−12164

Chemical Reviews

Review

Figure 7. Commonly used targeting ligands are shown in increasing order of size. Molecular weights and the Stokes−Einstein radii are shown below the ligand’s label.165 Stokes−Einstein radius estimates were made from molecular weight using the relationship MW (in kDa) = 1.32 × R3 (in nm3). MW = molecular weight; R = Stokes−Einstein radius.166,167

always be preferred over the traditional small organic molecule binders.122,124−131 Moreover, while most investigators may choose to use the receptor’s natural ligand-binding site for drug targeting, in some cases design of a ligand to fit an allosteric or nonfunctional pocket on the receptor may actually be preferred.132−141 Thus, the allosteric/nonfunctional site will not generally be as frequently occupied by endogenous ligands as the primary ligand-binding site, thereby decreasing the chances of competition with the natural ligand.142 Moreover, because such allosteric/nonfunctional sites will be less-highly conserved among isoforms of a receptor family (e.g., adrenergic receptors), opportunities for the design of an isoform-specific targeting ligand will be greatly increased.132,143 Finally, in cases where the targeted receptor is physically obstructed by a dense glycocalyx or extracellular collagen network, combining the LTD with a reagent that can break down these extracellular barriers may offer improved access to the desired receptor.144−155 For example, Diop-Frimpong et al.145 found that intratumoral distribution of nanotherapeutics of ∼100 nm size was significantly improved in multiple murine tumor models when combined with losartan, an antifibrotic agent that inhibits collagen synthesis by tumor-associated fibroblast cells. 2.1.5. Receptor Competition. During the receptorselection process, the potential for competition with the receptor’s natural ligand should not be overlooked. For example, while tumor-enriched receptors such as the folate receptor and the glucose transporter may constitute ideal tumor-specific targets, the investigator must still evaluate the impact that the natural ligand (folate and glucose, respectively, in the above examples) might have on LTD binding. Interestingly, although serum concentrations of folic acid (∼5−50 nM) and glucose (∼4−6 mM) lie within the concentration ranges of the drugs used to target these receptors, competition for the natural ligand has not yet constituted a significant problem for LTD delivery. Thus, patients preparing to receive a folate-targeted drug are usually asked to refrain from taking a vitamin pill on the day of therapy and patients scheduled to undergo glufosfamide therapy are

tumor reduction is nevertheless seen when the cytotoxic warhead is linked to the CCK2 receptor targeting ligand via a cleavable disulfide bond.109 Finally, when evaluating receptor internalization and trafficking, the investigator should always consider the impact of receptor occupancy on its intracellular trafficking pattern. Thus, in some cases (e.g., the beta adrenergic receptor), binding of a ligand to its receptor will enhance and/or accelerate receptor internalization.110−114 In other cases, the fraction/percent of receptors that become occupied will determine its intracellular trafficking itinerary, i.e., with high receptor occupancy promoting receptor translocation to lysosomes and low receptor saturation stimulating receptor recycling to the cell surface.112 In still other cases (e.g., the EGF receptor), ligand-induced receptor cross-linking will influence receptor fate, with cross-linked receptor often trafficking to lysosomes and unclustered receptors returning to the cell surface27,115−119 Finally, in a few cases (e.g., the folate receptor), receptors internalization always occurs at the same constant rate, regardless of receptor saturation, but receptor cross-linking forces it to traffic to lysosomes, thereby preventing its normal recycling to the cell surface.94 Taken together, one can appreciate that the nature of a ligand−receptor interaction can significantly influence the potency and metabolism of an LTD.120,121 2.1.4. Receptor Topography. The molecular topography of the targeted receptor’s surface should also be carefully considered when selecting a receptor for diseased-cell drug delivery. Because the binding of small molecules to proteins is driven by weak interactions, including hydrophobic, H-bond, electrostatic, π-stacking, van der Waals, and dipole−dipole interactions, high affinity/specificity binding can only be achieved if multiple weak interactions can be simultaneously established.122,123 Thus, when small-molecular-weight ligands are preferred for drug targeting, receptors with highly functionalized, deeply contoured binding cavity(s) must be sought to provide the weak-bonding potential required for high-affinity interactions. Indeed, when receptors with lesscontoured topographies must be targeted, use of larger ligands such as antibodies, aptamers, or protein scaffolds will almost 12139

DOI: 10.1021/acs.chemrev.7b00013 Chem. Rev. 2017, 117, 12133−12164

Chemical Reviews

Review

either placed on a low-carbohydrate diet or injected with insulin several hours before treatment.156−158 Similarly, imaging with [18F]fluorodeoxyglucose positron-emission tomography (PET) is usually only performed after the patient has been on a lowcarbohydrate diet for 24 h. While similar measures are required to prepare mice for experiments with LTDs derived from natural ligands, in the case of folate-targeted drugs the mice must be maintained on a folate-deficient chow for ∼2 weeks prior to therapy, because most animal chows are supplemented with megadoses of folic acid, causing their serum folate levels to increase well beyond their natural concentrations.159 2.2. Selection of a Targeting Ligand

With recognition that a targeting ligand can improve both drug potency and specificity, an explosion in the diversity of potential targeting ligands has emerged over the last few years. Thus, targeting ligands under current clinical development range from small molecules to peptides, to aptamers, to antibody fragments, to novel protein scaffolds, and to intact antibodies (Figure 7).2,124,160−164 While each of these ligand classes has its uniquely valuable attributes, each also is encumbered by its own set of disadvantages. In the following section, a critical analysis of the pros and cons of each ligand type for use in drug targeting will be outlined. 2.2.1. Ligand Size. Ligand size may exert its greatest impact on the pharmacokinetics of a ligand-targeted drug. Thus, molecules of Mr < ∼40 000 are readily filtered through the glomerulus of the kidney into the urine within 15−25 min of injection, whereas larger molecules and their drug conjugates can circulate in the blood for several days before they are cleared.168 For example, the FDA-approved antibody−drug conjugate Adcetris (an anti-CD30 antibody conjugated to monomethyl auristatin E of Mr ≈ 150 000) has a half-life of ∼4−6 days in human circulation, whereas the low-molecularweight cancer agent etarfolatide (folate-99mTc conjugate, Mr ≈ 856) has a half-life of only 27 min.169,170 Indeed, as shown in Figure 8, filtration rates through the kidneys are related to the size and charge of the circulating solute.171 For some applications, the longer circulation times of larger LTDs can be very beneficial, because they increase the proverbial “area under the curve” (AUC) that provides sustained LTD exposure to diseased tissues. Although larger LTD size has also been reported to increase passive accumulation of a conjugate in a tumor mass via the enhanced permeability and retention (EPR) effect, EPRmediated accumulation in malignant tissues is now thought to be substantially less pronounced in humans than in animal models.175−177 Moreover, a recent survey178 of the accumulated literature on the biodistribution of nanoparticle drugs (which rely on EPR effect) in tumor-bearing animals has found that only 0.7% of an administered nanoparticle dose is commonly delivered to solid tumors (for a more detailed discussion of the limitations of animal tumor models, see refs 179−186).179−186 Thus, without further improvements in methods for tumorspecific nanoparticle delivery, one should exercise caution before relying on a nanoparticle formulation for treatment of most solid tumors. Nevertheless, in those occasional tumors where an EPR effect is prominent, poorly formed junctions between endothelial cells allow extravasation of LTDs that would normally be too large to escape the vasculature into healthy tissues. Subsequent drainage of these passively accumulating LTDs from a tumor mass may also be retarded due to an inadequate lymphatic

Figure 8. Effect of molecular size on kidney permeability: fractional clearance or sieving coefficient of a solute plotted against its Stokes− Einstein radius. From the top, the curves represent slightly positive solutes (with charge densities of 0.002 Coul/m2), neutral solutes, negatively charged solutes (with charge densities of −0.006 Coul/m2), and highly negatively charged solutes similar to that of albumin (−0.022 Coul/m2). The approximate location on the plot of a small molecule and a typical antibody are shown in red color. For reference to known molecules, the following Stokes−Einstein radii and molecular weights apply: 5-fluorouracil (