Perspective pubs.acs.org/jmc
Macrocyclic Drugs and Clinical Candidates: What Can Medicinal Chemists Learn from Their Properties? Fabrizio Giordanetto*,† and Jan Kihlberg*,‡ Cardiovascular and Metabolic Disorders Research Area, AstraZeneca R&D Mölndal, SE-431 83 Mölndal, Sweden
Downloaded via TUFTS UNIV on July 1, 2018 at 11:32:37 (UTC). See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.
S Supporting Information *
ABSTRACT: Macrocycles are ideal in efforts to tackle “difficult” targets, but our understanding of what makes them cell permeable and orally bioavailable is limited. Analysis of approximately 100 macrocyclic drugs and clinical candidates revealed that macrocycles are predominantly used for infectious disease and in oncology and that most belong to the macrolide or cyclic peptide class. A significant number (N = 34) of these macrocycles are administered orally, revealing that oral bioavailability can be obtained at molecular weights up to and above 1 kDa and polar surface areas ranging toward 250 Å2. Moreover, insight from a group of “de novo designed” oral macrocycles in clinical studies and understanding of how cyclosporin A and model cyclic hexapeptides cross cell membranes may unlock wider opportunities in drug discovery. However, the number of oral macrocycles is still low and it remains to be seen if they are outliers or if macrocycles will open up novel oral druggable space.
1. INTRODUCTION It has proven to be extremely difficult to develop small molecule probes and drugs against targets with extended binding sites, i.e., class B G-protein-coupled receptors (GPCRs), protein−protein interactions, and some enzymes.1 Currently “biologics” constitute the most common approach for modulation of these targets, with limitations such as high cost, reduced patient compliance, lack of cellular penetration, and low oral bioavailability. One reason for these difficulties could be that small molecule hits and leads that are compliant with Lipinski’s rule-of-five have been sought to increase the likelihood of obtaining favorable physicochemical and pharmacokinetic properties2 and in order to reduce risks of compound related toxicity.3,4 However, by stepping slightly outside the rule-of-five chemical space, compound diversity and opportunities to develop probes and drugs against “difficult” targets can be enhanced. Macrocycles have several features that make them interesting in efforts to tackle “difficult” targets with extended binding sites.5,6 Because of their size and complexity, they can engage targets through numerous and spatially distributed binding interactions, thereby increasing both binding affinity and selectivity. Furthermore, cyclization provides a degree of structural preorganization that may reduce the entropy cost of receptor binding compared to linear analogues.5,7 It should, however, be pointed out that macrocycles are not completely rigid. Instead they may offer a suitable compromise between preorganization and sufficient flexibility that could facilitate interactions with dynamic protein targets. In addition some reports suggest that cyclization has a favorable impact on other essential properties required for drugs, such as membrane permeability,8 metabolic stability, and overall pharmacokinetics.5,6 © 2013 American Chemical Society
A significant number of macrocyclic drugs are currently on the market, predominantly of natural product origin with complex structures. Concerns that synthetic tractability will limit opportunities for lead optimization and increase costs for scale up are reasons why the pharmaceutical industry has been cautious about development of macrocyclic drugs. These concerns have resulted in macrocycles being excluded from most academic and industry screening collections; for example, the AstraZeneca high throughput screening (HTS) collection contains 30 amino acids, as well as contrast agents and veterinary drugs were removed from further analysis. This curation resulted in a data set consisting of 68 macrocyclic drugs, which were further annotated with in silico molecular descriptors19 and information obtained from the literature on their route of administration,20 bioavailability, and dose in humans (cf. Table S1 in the Supporting Information for structures, indication, physical chemical properties, etc.). A data set of 35 macrocycles currently in clinical development was obtained in an analogous manner by mining of Adis R&D Insight21 and the literature, followed by curation and annotation just as for the set of marketed drugs (cf. Table S2 in the Supporting Information for structures, indication, physical chemical properties, etc.). In addition, compounds for which structures were not available were excluded from the latter data set. Through these efforts we have striven to make the two data sets of macrocycles as complete and accurate as possible. As information on launched drugs is readily available, we assume that the data set of marketed macrocyclic drugs is essentially complete. In contrast, the data set of macrocycles in clinical 279
dx.doi.org/10.1021/jm400887j | J. Med. Chem. 2014, 57, 278−295
Journal of Medicinal Chemistry
Perspective
Table 1. Calculated Physical Chemical Propertiesa for Oral Small Molecule Drugs, Different Classes of Registered Macrocycle Drugs, and Macrocycles in Clinical Development status registered
class small molecule macrocycle cyclic peptide macrolide
clinical development
macrocycle cyclic peptide macrolide “de novo designed”
route of administration
N
HBDa,b
PSAa,c
cLogP a,d
MWa,e
oral oral, MW > 500 oral parenteral oral parenteral oral parenteral
1589 89 18 45 1 26 14 9
2 (2−2) 2 (2−3) 4 (3−5) 14 (11−17) 5 19 (15−22) 4 (3−4) 6 (3−9)
74 (71−76) 139 (127−152) 212 (200−225) 417 (347−488) 290 529 (444−614) 205 (194−215) 240 (194−286)
2.2 (2.1−2.4) 4.2 (3.7−4.9) 4.4 (3.0−5.8) 0.6 (−1.2−2.3) 14.4 −0.8 (−3.0−1.4) 3.7 (2.7−4.6) 2.1 (−1.2−5.5)
322 (317−328) 602 (582−624) 852 (805−901) 1126 (996−1257) 1203 1294 (1128−1461) 831 (798−864) 867 (771−963)
oral parenteral oral parenteral oral parenteral oral parenteral
15 20 3 8 2 3 9 1
3 (2−4) 8 (5−11) 5 (4−6) 15 (10−20) 3 (2−3) 3 (2−4) 2 (1−2) 4
171 (130−214) 270 (189−351) 277 (227−327) 450 (339−561) 203 (190−215) 173 (87−260) 121 (78−165) 173
6.9 (5.3−8.5) 2.4 (0.5−4.2) 11.3 (6.3−16.2) −0.1 (−2.7−2.7) 5.2 (1.2−9.1) 4.8 (0.9−8.8) 5.7 (4.9−6.4) 6.1
775 (634−916) 908 (751−1065) 1162 (946−1379) 1176 (936−1416) 911 (755−1067) 733 (402−1064) 598 (494−701) 937
a Mean value (lower and upper margins of 95% confidence interval). bHydrogen bond donors. cPolar surface area. dCalculated log P. eMolecular weight.
explanation for the route of administration (Table 1). Parenteral macrocycles as a class are significantly more polar than either the oral macrocycles or small molecule drugs. Thus, the parenterals have more hydrogen bond donors, and consequently larger polar surface areas, than the two classes of orals. The calculated lipophilicity (cLogP) is also significantly lower for the parenteral macrocycles than for either class of oral drugs. In addition, the mean molecular weight of the parenteral macrocycles is higher than for the oral macrocycles. Comparing the oral macrocycles with the oral small molecule drugs reveals that molecular weight, cLogP, polar surface area, and number of hydrogen bond donors are on average ≥2 times higher for the oral macrocycles. A small number (N = 89, 5.6%) of the oral small molecule drugs have a molecular weight of >500 and are not macrocyclic. A comparison of this subset with the oral macrocyclic drugs reveals that the polarity is lower (fewer hydrogen bond donors and smaller polar surface area) for the nonmacrocyclic drug subset with molecular weight of >500, that the lipophilicity is similar, and that the macrocyclic drugs have a higher molecular weight. In conclusion, parenteral administration of macrocycles appears to be driven by the higher number of hydrogen bond donors displayed by this class than for macrocycles and small molecule drugs that are administered orally. Interestingly, even though relatively few oral macrocyclic drugs exist, their molecular weight is indeed significantly higher than for oral small molecules, as well as the subset of nonmacrocyclic drugs with molecular weight of >500 (Table 1). Most likely, oral availability among macrocyclic drugs and nonmacrocyclic drugs with molecular weight of >500 is facilitated by their relatively high lipophilicity compared to the “average” oral small molecule drug. 3.2. Properties of Macrocyclic Drugs in the Cyclic Peptide Class. Exactly 30 of the macrocyclic drugs are cyclic peptides, which can be divided further into several subclasses (Figure 2, Table S1). The disulfide-bridged cyclic peptide subclass is the largest of these, with members such as argipressin and lanreotide. The glycopeptide subclass (e.g., vancomycin hydrochloride), the lipopeptide class of echinocandins (e.g., anidulafungin), and cyclic peptides with chemical modifications
Figure 2. Distribution of oral (green bar) and parenteral (red bars, N = 29) macrocyclic peptide drugs across different subclasses.
on amino acid side chains (e.g., pasireotide) each have three members, whereas the few remaining cyclic peptide drugs belong to six different subclasses. All macrocyclic peptide drugs except cyclosporin A, an N-methylated undecapeptide, are administered through various parenteral routes, in line with the general observation that insufficient oral bioavailability is a key limitation for peptide therapeutics. Inspection of the physical chemical properties of this set of parenteral cyclic peptide drugs shows that they are highly polar with a large number of hydrogen bond donors, a resulting large polar surface area, and low cLogP (Table 1). The mean molecular weight is also high. Taken together, the physical chemical properties of nonoral cyclic peptide drugs provide a clear rationale for why parenteral administration is required. As mentioned above, cyclosporin A is the only one of the marketed macrocyclic peptide drugs that is administered orally. However, absorption in the gastrointestinal tract is highly variable, resulting in bioavailabilities ranging from
