Article pubs.acs.org/crt
Aortic Binding of AZD5248: Mechanistic Insight and Reactivity Assays To Support Lead Optimzation Ryan A. Bragg,†,§ Simon Brocklehurst,†,⊥ Frida Gustafsson,† James Goodman,†,▽ Kevin Hickling,† Philip A. MacFaul,‡ ,∥ Steve Swallow,*,†,○ and Jonathan Tugwood†,# †
Global Safety Assessment, AstraZeneca, Alderley Park, Macclesfield, Cheshire SK10 4TG, United Kingdom DMPK Innovative Medicines, AstraZeneca, Alderley Park, Macclesfield, Cheshire, SK10 4TG, United Kingdom
‡
ABSTRACT: The oral dipeptidyl peptidase 1 (DPP1) inhibitor AZD5248 showed aortic binding in a rat quantitative whole-body autoradiography (QWBA) study, and its development was terminated prior to human dosing. A mechanistic hypothesis for this finding was established invoking reactivity with aldehydes involved in the cross-linking of elastin, a major component of aortic tissue. This was tested by developing a simple aldehyde chemical reactivity assay and a novel in vitro competitive covalent binding assay. Results obtained with AZD5248, literature compounds, and close analogues of AZD5248 support the mechanistic hypothesis and provide validation for the use of these assays in a two tier screening approach to support lead optimization. The strengths and limitations of these assays are discussed.
■
rofecoxib,6−8 muzolimine,9 hydralazine,10 and methylimidazole,11−13 have been described in the literature to be retained selectively in the aorta in QWBA studies, leading to concomitant ultrastructural changes when examined by electron microscopy. Prior to the studies described in this publication, we were aware of ZD4407 (Chart 1), a compound that had been in development as a 5-lipoxegenase inhibitor,14 as the only other example within AstraZeneca of a compound being retained in the aorta in QWBA studies. Importantly, methylimidazole has been shown to have effects on aortic extensibility and tensile strength following 28 days of dosing in weanling rats.13 It has also been suggested that there is a link between aortic tissue binding of rofecoxib and cardiovascular effects observed with this compound.6 Some of these compounds have been shown to react with model aldehydes,7,12 either directly or in the case of methylimidazole via bioactivation, leading to the hypothesis that it is chemical reaction with the aldehyde functionality in allysine that leads to defective crosslinking and is responsible for the aortic retention, ultrastructural changes, and functional effects of these compounds. In the work presented here, we describe another class of compound, exemplified by AZD5248 (Chart 1),15 that shows a high level of aortic retention in rat QWBA studies with concomitant ultrastructural changes similar to those described in the literature when examined by electron microscopy. AZD5248 was being developed for treatment of chronic obstructive lung disease, and many patients with this disease have associated atherosclerosis, a condition in which remodelling
INTRODUCTION Elastin is a core constituent elastic protein of connective tissue whose mechanical and physical structure confers the property of elasticity within tissues. Additionally, elastin has been shown to regulate cell growth and behavior. Elastin is particularly abundant in large blood vessels such as the aorta, where it serves as the principal mediator of pressure wave propagation to smooth blood flow but is also important in other tissues such as the lungs, elastic ligaments, and elastic cartilage. It is secreted in the form of tropoelastin, a monomeric soluble precursor that consists of alternating highly hydrophobic and hydrophilic regions. The former are responsible for self-aggregation and tensile properties, while the latter are involved in cross-linking to form an insoluble and durable polymer (a process termed coacervation) that shows considerable variation in different organs. The organ specific process of elastogenesis is poorly understood, but, importantly, the cross-linking process involves the formation of a variety of cross-linking structures through the reaction of allysine. Allysine is an unusual amino acid containing a reactive aldehyde side chain which is generated by the oxidation of lysine side chains by a family of enzymes known as lysyl oxidases (Figure 1).1 Consistent with these findings is the observation that there is a significantly higher concentration of desmosine and isodesmosine cross-links (Figure 1) in aortic tissue relative to other elastin containing tissues.2 Various studies, including the generation of lysyl oxidase gene knockouts and the use of lysyl oxidase inhibitors, have implicated interference in this biochemical pathway with the generation of aortic aneurisms.3,4 Defects in the elastin gene have also been associated with a higher prevalence of aortic aneurisms in human subjects.5 Furthermore, a number of compounds (Chart 1), e.g., © 2015 American Chemical Society
Received: June 5, 2015 Published: September 9, 2015 1991
DOI: 10.1021/acs.chemrestox.5b00236 Chem. Res. Toxicol. 2015, 28, 1991−1999
Article
Chemical Research in Toxicology
Figure 1. Aortic tissue is rich in the highly cross-linked and elastically extensible protein elastin. Cross-linking of elastin chains provides its elastic properties and is mediated by specialized cross-linking groups, especially desmosine and isodesmosine that are particularly abundant in the aorta. The formation of desmosine and isodesmosine cross-links occurs through the uncatalyzed reaction of lysine with three equivalents of allysine. Isodesmosine, being a regioisomer of desmosine is also formed in this reaction. Allysine contains a reactive aldehyde (red), which is generated through the oxidation of lysine by lysyl oxidase. The figure showing the location of the aorta was created by J. Heuser and reproduced under the GNU Free Documentation License GFDL (http://www.gnu.org/copyleft/fdl.html; https://commons.wikimedia.org/wiki/File:Aorta_scheme.jpg).
Chart 1. Structures of Compounds Reported to Show Aortic Retention in QWBA Studies
of elastic tissues is known to occur.16 Concerns about the risk of toxicity to the aorta on chronic administration within this patient population contributed to the termination of AZD5248 from further development and prompted reoptimization of this series of compounds to remove this liability while retaining potency against its intended target. To this end, we have developed a two-tiered assay approach involving a simple aldehyde reactivity assay and an in vitro competitive covalent binding assay to assess compounds for their
propensity to bind irreversibly to rat aorta. In addition, we provide a mechanistic rationale for the binding of this compound to aorta.
■
EXPERIMENTAL PROCEDURES
Chemicals and Reagents. All chemicals were purchased from Sigma-Aldrich Company Ltd. (Poole, Dorset, UK) unless stated otherwise and were of the highest purity available. Test compounds (>95% purity) were obtained from Compound Management, AstraZeneca R&D, Macclesfield, UK, or were purchased from Sigma1992
DOI: 10.1021/acs.chemrestox.5b00236 Chem. Res. Toxicol. 2015, 28, 1991−1999
Article
Chemical Research in Toxicology Aldrich Company Ltd. or Sequoia Research Products Ltd. (Pangbourne, UK). [14C]AZD5248 was prepared by AstraZeneca Global Isotope Chemistry Group. Deschloromuzolimine was used in place of muzolimine, and the removal of the two chlorine substituents is expected to have minimal impact on its nucleophilic reactivity relative to muzolimine being remote from the putative reactive center. QWBA Studies. A single dose of 20 mg/kg (53 μmol/kg, 5 MBq/kg) of [14C]AZD5248 was given orally to Lister hooded rats. The rats were sacrificed at 0.5, 1, 4, 24, and 48 h. In addition, two male rats were sacrificed at 7 and 21 days after dosing. The animals were sacrificed with enflurane (Efrane, Abbott Laboratories, USA) and frozen in acetone, cooled to −70 °C with solid CO2. After the removal of limbs and tail, the carcass was embedded, with the left lateral side uppermost, in a 2.5% aqueous solution of natricum caramellosum and frozen for at least 10 min in acetone at −70 °C. The animal blocks were stored at −20 °C until sectioning. At sectioning, each block was mounted in a Leica CM3600 Cryomacrotome (Leica Microsystems GmbH, Germany) maintained at approximately −20 °C. Sagittal whole body sections (30 μm) were obtained at various levels through the carcass to include all major organs and tissues of interest. The sections were mounted on Invisible tape (type 810, 3M, USA) and numbered consecutively with radioactive ink. From the animals killed 2, 7, and 21 days after administration, sections were mounted on type 6890 tape (3M), which was extracted by treatment with solutions of trichloracetic acid (5%), ethanol (up to 99.5%), and heptane (100%). All sections were dried at −20 °C for at least 1 day prior to exposure on phosphor-imaging plates. Autoradiograms obtained from extracted tape sections are considered to represent the distribution of unextractable or firmly bound [14C]AZD5248-derived radioactivity. Electron Microscopy. Electron microscopy was performed on samples of aorta taken from an otherwise standard 1 month toxicity study in the rat intended for regulatory submission of AZD5248. Four groups of Wistar Hannover rats, each consisting of 10 males and 10 females, were given AZD5248 orally, once daily, for one month at dose levels of 0, 20, 60, and 200 mg/kg/day. Because of signs of toxicity, female high dose animals were only dosed for approximately 2 weeks. Multiple toxicological parameters were assessed, and target organs of the liver, kidney, and thyroid were examined by routine light microscopy. Following the results of the QWBA assay, samples of the aorta (fixed in 10% neutral buffered formalin) were selected from high dose and control animals. The samples were washed in 0.1 M sodium cacodylate prior to refixation in 2.5% glutaraldehyde in 0.1 M cacodylate buffer (pH 7.2). The tissue was processed using a standard processing regime, including secondary fixation in 1% osmium tetroxide, and embedded into epoxy resin for examination of toluidine blue-stained sections by light microscopy. Ultrathin sections (70−90 nM thick) were cut, stained using uranyl acetate and lead citrate, and examined using a Jeol 1400 transmission electron microscope at 80 kV (Jeol UK Ltd., Welwyn Garden City, UK). Aldehyde Reactivity Assay. Test compounds were added to a solution of aldehyde in pH 7.4 phosphate buffer. The reaction mixtures were incubated in sealed vials at 37 °C, samples taken at regular intervals, and analyzed by LC-UV. In order to quantify the reaction, the kinetics were monitored through the loss of the parent compound, as studied by UV spectroscopy. If the chromophore of the parent compounds was extremely weak, single-ion monitoring mass spectrometry was employed to quantify parent loss. The formation of any adducts was also assessed by LC-UV, and the identity was determined by mass spectrometry. A blank reaction in the absence of any aldehyde was also performed for each compound to assess for any hydrolytic instability under these conditions. Characterization of Compound 4 (R = ethyl). (S)-4-Amino-N(1-cyano-2-(4′-cyano-[1,1′-biphenyl]-4-yl)ethyl)tetrahydro-2H-pyran4-carboxamide (100 mg, 0.27 mmol) was taken up as a suspension in pH 7.4 phosphate buffer (5 mL) and then treated with propionaldehyde (54.3 mg, 0.93 mmol). The reaction mixture was allowed to stir at 40 °C for 18 h under closed atmosphere conditions (sealed microwave tube). The white precipitate was filtered, washed well with water, and dried under vacuum to afford 4′-((2S)-2-cyano-2-(2-ethyl-4-oxo-8-oxa-1,3-
diazaspiro[4.5]decan-3-yl)ethyl)-[1,1′-biphenyl]-4-carbonitrile (94 mg, 85%) as a white solid. 1H NMR (400 MHz, DMSO) 0.87 (t, J = 7.3 Hz, 4H), 0.92 (t, J = 7.3 Hz, 3H), 1.07−1.27 (m, 2H), 1.3−1.43 (m, 3H), 1.43−1.64 (m, 2H), 1.82 (ddt, J = 6.9, 13.1, 19.5 Hz, 3H), 1.98 (ddd, J = 2.7, 7.5, 13.4 Hz, 1H), 2.95 (d, J = 9.2 Hz, 1H), 3.05 (d, J = 8.6 Hz, 1H), 3.29 (s, 2H), 3.36−3.48 (m, 3H), 3.48−3.59 (m, 4H), 3.69 (dt, J = 4.0, 7.8 Hz, 1H), 3.77 (ddt, J = 4.0, 8.5, 12.7 Hz, 2H), 3.97 (td, J = 2.3, 8.3 Hz, 1H), 4.38 (td, J = 2.6, 9.0 Hz, 1H), 5.03 (dd, J = 7.4, 8.8 Hz, 1H), 5.35 (dd, J = 7.0, 9.8 Hz, 1H), 7.41−7.49 (m, 4H), 7.73 (dd, J = 8.3, 14.7 Hz, 4H), 7.82−7.95 (m, 8H). 13C NMR (176 MHz, DMSO) 7.25, 7.87, 25.60, 25.70, 30.93, 31.59, 32.81, 33.30, 34.40, 34.55, 41.70, 42.56, 57.15, 57.30, 62.00, 62.17, 70.94, 72.59, 109.46, 116.71, 117.76, 118.24, 126.46, 126.50, 126.83, 129.53, 129.66, 132.26, 135.09, 135.41, 136.57, 136.63, 143.49, 143.59, 175.86, 176.26. In Vitro Competitive Aortic Tissue Binding Assay. Aortic homogenate was prepared from the thoracic aortae of Han Wistar rats. Freshly isolated thoracic aortae were frozen and later thawed and stripped of nonelastic material. Stripped aortae were then weighed, cut into small pieces, and homogenized first with a rotor-stator homogenizer, and then with a loose-fit, and then a tight-fit, Dounce homogenizer in Puck’s saline (137 mM NaCl, 5.37 mM KCl, 4.17 mM NaHCO3, and 5.55 mM D-glucose). Homogenate concentration was adjusted to 30 mg/mL in Puck’s saline, and aliquots were stored at −80 °C until use. Positive and negative control compounds and test compounds were made up to 100 mM in dimethyl sulfoxide (DMSO) and added to 1 mL aliquots of aortic homogenate in Puck’s saline for a final concentration of 100 μM. Homogenate samples were preincubated with test compounds at 37 °C, rotating, overnight. [14C]AZD5248 was then added to all samples to a final concentration of 100 μM, and samples were incubated at 37 °C, rotating, for a further 2 h. Protein was precipitated from each sample by the addition of 10 mL of acetone, prechilled to −20 °C. Samples were left overnight at −20 °C to allow complete precipitation. Precipitate was pelleted by centrifugation at 4,500g at 4 °C for 20 min, an aliquot of supernatant was removed for analysis, and the remainder of the supernatant discarded. Precipitate was washed by resuspension in 10 mL of 80% methanol in distilled water and repelleted by centrifugation at 4,500g at 4 °C for 20 min. Washing was repeated for a total of 4 washes in 80% methanol, and 2 further washes in 100% methanol, an aliquot of supernatant being removed for analysis at each stage. After the final wash, the precipitate was air-dried and dissolved overnight in 1 mL of NCSII Tissue Solubiliser. One milliliter aliquots of supernatants were added to 5 mL of Ultima Gold scintillation fluid (PerkinElmer, MA, U.S.A.), and 1 mL of solubilized pellets were added to 5 mL of Hionic-Fluor scintillation fluid (PerkinElmer, MA, USA). Radioactivity of samples was determined on a Beckman LS6500 multipurpose scintillation counter (Beckman Coulter, IN, USA). On each occasion, AZD5248 was run as a positive control, and AZD5672 (Chart 2) and the DMSO vehicle were run as negative controls. Duplicate samples were tested for each compound on each experimental occasion, and at least 2 experiments were run for each test compound. Mean radioactivity of the samples preincubated with the DMSO vehicle control was taken to be 100% binding, and results for samples preincubated with other compounds were expressed as % difference from vehicle control. One-way ANOVA and Bonferroni’s multiple
Chart 2. Structure of Negative Control
1993
DOI: 10.1021/acs.chemrestox.5b00236 Chem. Res. Toxicol. 2015, 28, 1991−1999
Article
Chemical Research in Toxicology comparison tests were performed to calculate the significance of differences from the vehicle control.
■
RESULTS [ C]AZD5248 QWBA. Radioactivity was widely distributed in the body with peak concentrations being reached in most tissues after 1 h. Excretion occurred via bile and urine, with most of the tissue-related radioactivity eliminated at 21 days after administration with the exception of the aortic wall. Residual amounts were also detected in the outer ear, intervertebral ligament, eye, renal outer medulla, liver, and spleen. AZD5248 was concluded to have high affinity to elastic tissues, in particular the aorta (Figure 2). 14
Figure 3. Representative photomicrographs from rats administered AZD5248 or vehicle for 28 days (original magnification × 4000). Compared to the control (A and C), certain endothelial cells (EC) showed loss of attachment and morphological abnormality (B, arrows) and in D compared to the control, the tunica media appears disordered with the loss of attachment of smooth muscle cells (SM) and increase in intercellular substance (arrows).
reactivity of AZD5248 toward model aldehydes in vitro was therefore assessed. The aldehyde group in allysine is remote from other functionality in the amino acid that might affect its reactivity, so simple aldehydes such as propionaldehyde should be reasonable surrogates for allysine in reactivity assays. Benzaldehyde has been used in the literature as a model for rofecoxib allysine reactivity, and the literature method was initially employed where the 100 μM compound was incubated in the presence of 1 mM benzaldehyde.7 As rofecoxib was used in these original studies, this compound was also assessed alongside AZD5248. Reactions were initiated by the addition of the test compound to a solution of the aldehyde in pH 7.4 phosphate buffer and incubated in sealed vials for 18 h at 37 °C. Analysis by LC-UV-MS of the solutions after this time indicated that both compounds had reacted with benzaldehyde. In terms of a negative control, the assay was also repeated with AZD5672 which shows no evidence of retention in QWBA studies. In this case, no reactivity with benzaldehyde was detected. Compounds representing the different structural types known to show aortic retention were tested for benzaldehyde reactivity (capromorelin being structrurally similar to AZD5248 was not tested in this initial assessment). From these initial experiments, it was clear that reaction with benzaldehyde could be observed (Table 1); however, the use of benzaldehyde complicated the LC-UV analysis due to the high absorbance of this model aldehyde at a wavelength of 240 nM and above. Therefore, benzaldehyde was replaced with the UV transparent propionaldehyde,17 and the formation of adducts was similarly observed (Table 1). The concentrations were adjusted to give reasonable half-lives for the above reactive compounds employing an 18 h incubation period. The high reactivity of AZD5248 meant that pure samples could be obtained from preparative scale versions of the assay conditions allowing characterization of the diastereomeric imidazolidin-4-
Figure 2. Representative images from QWBA at 1 h and 2 and 21 days following a single dose of [14C]AZD5248. Dark areas represent the distribution of the radiolabel and indicate the strong retention of the label in aorta (arrows) even after 21 days post-administration.
Electron Microscopy. The tissue was originally fixed in 10% neutral buffered formalin and was therefore not optimal for electron microscopy. However, control samples revealed an ordered distribution of endothelium (EC) that had strong association with the elastic lamina. In the tunica media, there was an ordered and repetitive distribution of smooth muscle cells (SM) with minimal intercellular substance that also showed strong association to layers of elastin fibers (Figure 3). Certain test samples, however, showed a degree of change that, in-spite of suboptimal preparation, were viewed as similar to those previously described in mice deficient in lysyl oxidase.3 Principally, endothelial cells had reduced association with the elastic laminae and demonstrated degrees of morphological change and loss of their typical flattened structure. Smooth muscle cells also showed loss of organized contact, and there was an increase in the amount of intercellular substance. Reactivity of Compounds toward Model Aldehydes. Although QWBA studies are definitive in determining the risk associated with a compound in terms of its reactivity and subsequent retention in tissues, it was felt that a simple in vitro screen would be a useful addition in the early drug discovery phase. Such an assay could be employed to screen out compounds that were reactive toward aldehydes, or if the reactivity was unavoidable, the assay could be used to rank compounds and drive the design of less reactive analogues. The 1994
DOI: 10.1021/acs.chemrestox.5b00236 Chem. Res. Toxicol. 2015, 28, 1991−1999
Article
Chemical Research in Toxicology Table 1. Reactivity of Compounds toward Model Aldehydes in pH 7.4 Phosphate Buffer at 37 °C for 18 h compound
propionaldehyde reactivitya
mass changeb
benzaldehyde reactivityc
mass changed
hydrolysise
AZD5248 AZD5672 rofecoxib ZD4407 deschloro-muzolimine capromorelin 1 2 3
39 min ND 28.1 h 10.1 h 1.9 h 1.4 h 38.3 hf ND ND
+40
Y N Y Y Y NT N N N
+88
ND ND ND ND 10 h 5.6 h ND ND ND
+58 +40 +56 +40
+106 +88 +104
Half-life as measured by compound loss on incubation of 50 μM compound with 5 mM propionaldehyde. bMass of adduct formed in the presence of propionaldehyde relative to that of the parent. cFormation of adducts on incubation of 100 μM compound with 1 mM benzaldehyde. dMass of adduct formed in the presence of benzaldehyde relative to that of the parent. eHalf-life as measured by compound loss on incubation of 50 μM compound. ND = no loss of parent compound detected. fNo adduct detected. NT = not tested. a
Chart 3. Additional Structures Tested in the Aldehyde Reactivity Assay
ones 4 wherein R is an ethyl group. 1H and 13C NMR characterization are consistent with the proposed structure, in particular the signals at δ 5.03 and 5.35 in the 1H NMR spectrum being indicative of the diastereomeric imidazolidin-4-one ring protons.18 A variety of test compounds (additional structures highlighted in Chart 3) were subsequently tested at a concentration of 50 μM in the presence of 5 mM propionaldehyde in pH 7.4 phosphate buffer at 37 °C. The results of these experiments are presented in Table 1. An additional 33 proprietary AZ compounds which had previously been assessed in a QWBA study and found to be devoid of aortic binding were also blind tested in the propionaldehyde reactivity assay. In each case, no reactivity of the compound was observed with propionaldehyde. In Vitro Competitive Aortic Tissue Binding Assay. Seeking a biologically more relevant second tier assay to further support lead optimization activities, we explored an in vitro covalent binding assay based on competition for reactive sites within the aortic tissue homogenate. While in vitro covalent binding assays using radiolabeled compounds have been used previously to characterize compounds with aortic retention and demonstrate the involvement of lysl oxidase, these have typically been very low throughput and have required the synthesis of radiolabeled test compounds.7,8 The use of an assay based on competition for available binding sites would obviate the need for the synthesis of radiolabeled test compounds, provided that a suitably reactive competitive probe ligand could be found. On the
basis of the high aldehyde reactivity and in vivo retention observed with AZD5248, we explored [14C]AZD5248 as a competitive probe ligand in our assay. Briefly, samples of aortic homogenate were incubated with test compounds at 37 °C overnight. A concentration of 100 μM was chosen as this had been used in previous experiments with rofecoxib and was expected to give usefully measurable responses.7 [14C]AZD5248 was then added to all samples to compete for remaining reactive binding sites, and samples were incubated for a further 2 h. The protein was then precipitated, washed, and then air-dried. The pellets were then solubilized and residual radioactivity determined. It should be noted that while activity in this assay is likely indicative of covalent binding to aortic tissue, interference with AZD5248 binding via effects on other steps in the cross-linking process, for example, inhibition of lysl oxidase, cannot be ruled out without additional experimentation. Results from the competitive binding assay for the compounds in Table 1 are highlighted in Figure 4.
■
DISCUSSION The observation of aortic retention of AZD5248 in a rat QWBA study was surprising and, in the absence of obvious pathology in the 1 month toxicology studies, prompted detailed evaluation of aortic sections by electron microscopy. Although the tissues were not optimally preserved for ultrastructural evaluation, clear changes were observed to both extracellular matrix and cellular components of the aorta in treated animals. Similar ultra1995
DOI: 10.1021/acs.chemrestox.5b00236 Chem. Res. Toxicol. 2015, 28, 1991−1999
Article
Chemical Research in Toxicology
were encouraged by the literature precedent for the formation of related structures 5 in vivo from the reaction of acetaldehyde with peptides (Scheme 1).18 The observation of adduct formation with benzaldehyde and more usefully with the UV transparent propionaldehyde (Table 1) supports this hypothesis. The observation of a range of halflives for known aortic binders alongside a negative control (AZD5672) further supports the utility of this assay in detecting aortic reactivity. It is noteworthy that in this assay AZD5248 shows very high reactivity to propionaldehyde relative to other known aortic binders with a half-life of 38 min compared to, for example, rofecoxib, which shows relatively low reactivity in our hands with a half-life of 28 h. The mass spectra of adducts are consistent with the addition of the aldehyde both with (e.g., ZD4407) and without dehydration (e.g., rofecoxib). More suprisingly, deschloromuzolimine appears to undergo an additional oxidation following its reaction with either propionaldehyde or benzaldehyde as evidenced by the mass change of the derived adducts. The adduct derived from muzolimine reaction is expected to be a secondary alcohol (via reaction at C-4),19 and we therefore assume this is an air based oxidation as oxygen was not excluded from the assay, and air based oxidations of secondary alcohols are precedented in the literature.20 The in vitro competitive binding assay shows similar results, with the rank order of aldehyde reactivities reflected in the in vitro competitive covalent binding assay (Figure 4 and Table 2). As expected, the negative control AZD5672 shows no effect on [14C]AZD5248 covalent binding and is indistinguishable from the vehicle control (Figure 4). In contrast, deschloromuzolimine shows activity comparable to that of AZD5248, while ZD4407 shows an intermediate level of activity over the time course of this experiment. Surprisingly, rofecoxib appears indistinguishable from the negative control in our hands; however, its reactivity with aldehyde is also relatively low suggesting the assay configuration may be insufficiently sensitive to detect less reactive compounds like rofecoxib. It is also conceivable that the reaction of some compounds with aldehydes could be reversible in the presence of [14C]ADZ5248, leading to false negatives in the competitive aortic binding assay. Other compounds such as methylimidazole are expected to be unreactive in this assay as
Figure 4. Combined data sets from several experimental sets showing relative aortic binding of AZD5248, compounds 1−3, literature compounds, and a negative control AZD5672. Each experiment set is indicated by a different color. Preincubations were taken out to 72 h for compounds 1 and 2. Mean radioactivity of aortic homogenate samples preincubated with DMSO vehicle control for each experimental set was taken to be 100% binding, and results for samples preincubated with other compounds are expressed as % difference from experimental setmatched vehicle controls. One-way ANOVA and Bonferroni’s multiple comparison tests were performed to calculate the significance of differences from experimental set-matched vehicle controls. Significance of difference from the vehicle is calculated at p < 0.0001 (***, highly significant), p < 0.01 (**, very significant), and p < 0.05 (*, significant).
structural changes have been previously reported with compounds that have been demonstrated to react with aldehyde mimics of allysine.7,13 Given the literature precedent for the involvement of direct chemical reactivity with aldehydes, we postulated that formation of stable imidazolidin-4-ones such as 4 could be responsible for the aortic retention of AZD5248. We Scheme 1. Proposed Mechanism of Aortic Binding for AZD5248
1996
DOI: 10.1021/acs.chemrestox.5b00236 Chem. Res. Toxicol. 2015, 28, 1991−1999
Article
Chemical Research in Toxicology Table 2 compound
propionaldehyde reactivitya (Tier 1)
competitive covalent binding activity (Tier 2)
QWBA evidence of retention in aorta
comment
AZD5248 deschloromuzolimine ZD4407 rofecoxib
39 min 1.9 h
Y Y
Y Yc
positive classification correct in both assays positive classification correct in both assays
10.1 h 28.1 h
Y ND
Y Yc
AZD5672 33 AZ compounds capromorelin
ND ND
ND NT
ND ND
positive classification correct in both assays positive classification correct in Tier 1. negative classification in Tier 2 incorrect (see text) negative classification correct in both assays negative classification correct in aldehyde assays
1.4 h
ND
NDc
1 2 3
38.3 hb ND ND
ND Y N
NT NT NT
a
positive classification in Tier 1 incorrect negative classification in Tier 2 correct (see text) No QWBA No QWBA No QWBA
Half-lives from Table 1; ND = not detected; NT = not tested. bNo adduct detected. cLiterature data; see text for references.
Scheme 2. Proposed Mechanism of Aortic Binding for ZD4407 and Explanation for Lack of Reactivity with Compound 3
observation of adduct formation in the aldehyde reactivity assay suggests reaction via an enolic form of the oxindole followed by dehydration. Such reactivity with oxindoles is well precedented in the chemistry literature,21 being driven by the acidification of the benzylic protons by the adjacent amide and aromatic stabilization of the derived enol (Scheme 2). Consistent with this hypothesis is the observation that no adduct is formed and that no aortic retention is observed with the ring expanded structure 3 in which deprotonation is expected to be greatly attenuated. Such structures typically require strong base and aprotic conditions to undergo condensation reactions. Care should be taken extrapolating the results from AZD5248 to other related structures. For example, it seems likely that the dialkyl substitution in the α-position of AZD5248 is at least partly responsible for the reactivity of AZD5248. We present no direct evidence for this in this article; however, the effect of such substitution is well known in the literature to significantly facilitate cyclization reactions and is commonly known as the Thorpe-Ingold or gem dimethyl effect.22 The results with
they need metabolic bioactivation to produce the aldehyde reactive species; thus, careful consideration of the reaction cascade and test compound properties are needed when interpreting assay results. The hypothesis that AZD5248 covalent binding is mediated by imidazolidin-4-one formation was further tested using 1 and 2. Both of these compounds are derivatized versions of the primary amine in AZD5248 that is involved in imidazolidin-4-one formation and should be unreactive toward aldehydes. Both are unreactive in the aldehyde reactivity assay, and 1 is unreactive in the competitive binding assay. Surprisingly, 2 appears to show some reactivity in the competitive binding assay; however, the results appear quite variable compared with those of other compounds. While difficult to reconcile, it is conceivable that hydrolysis or metabolism to a more reactive derivative, or another competing reactivity could be responsible for the retention of radioactivity in this case. The reactivity of ZD4407 is also interesting in that no hypothesis for its aortic retention had been proposed previously. Its chemical structure and the 1997
DOI: 10.1021/acs.chemrestox.5b00236 Chem. Res. Toxicol. 2015, 28, 1991−1999
Article
Chemical Research in Toxicology Present Addresses
capromorelin suggest that other subtle structure−activity relationships may also exist. Capromorelin23 appears to have all of the key features required for aldehyde reactivity and irreversible binding and shows reactivity in the aldehyde assay, alebit masked significantly by its hydrolytic instability, but no activity in the competitive binding assay. Interestingly, this latter result is consistent with the reported QWBA study for this compound which shows no evidence of aortic binding.24 While it is not clear what structural properties are responsible for these differences in reactivity, it highlights the need to be cautious in over-responding to a positive result in the aldehyde reactivity assay as this example illustrates that false positives can be observed and highlights the need for the second tier assay. The capromorelin example perhaps indicates the reversible nature of aldehyde adduct formation with some compounds. The configuration of the aldehyde reactivity assay does not make it easy to distinguish reversible from an irreversible reaction. However, the aortic binding assay, if it mimics the in vivo situation, involves irreversible consumption of aldehyde (allysine) in the cross-linking process, removing it from any reversible equilibrium such that any reversibly bound compound should appear as a negative in this assay. To explore the risk of identifying false positives in the aldehyde assay, we identified and assayed a further 33 internal compounds where the QWBA reports showed no evidence of aortic binding. In all cases, no aldehyde reactivity was observed. Table 2 summarizes the outcomes from the two tier assay approach relative to the ability of compounds to be retained in the aorta according to QWBA experiments.
■
CONCLUSIONS
■
AUTHOR INFORMATION
§
(R.A.B.) Drug Safety and Metabolism, AstraZeneca, Unit 310, Cambridge Science Park, Milton Road, Cambridge CB4 0WG, U.K. ∥ (P.A.M.) Redx Pharma Plc, Mereside, Alderley Park, Macclesfield, Cheshire, SK10 4TF, U.K. ⊥ (S.B.) VWR International, AstraZeneca, Silk Road Business Park, Macclesfield, Cheshire, SK10 2NA, U.K. # (J.T.) Clinical & Experimental Pharmacology, Cancer Research UK Manchester Institute, The University of Manchester, Wilmslow Road, Manchester M20 4BX, U.K. ▽ (J.G.) Department of Religions and Theology, The University of Manchester, Oxford Road, Manchester M13 9PL, U.K. ○ (S.S.) Pharmaceutical Development, AstraZeneca, Silk Road Business Park, Macclesfield, SK10 2NA, U.K. Notes
The authors declare no competing financial interest.
■
ACKNOWLEDGMENTS Special thanks to Steve Glossop for NMR characterization of compound 4.
■
REFERENCES
(1) Wise, S. G., and Weiss, A. S. (2009) Tropoelastin. Int. J. Biochem. Cell Biol. 41, 494−497. (2) Yamaguchi, Y., Haginaka, J., Kunitomo, M., Yasuda, H., and Bandô, Y. (1987) High-performance liquid chromatographic determination of desmosine and isodesmosine in tissues and its application to studies of alteration of elastin induced by atherosclerosis. J. Chromatogr., Biomed. Appl. 422, 53−59. (3) Mäki, J. M., Räsänen, J., Tikkanen, H., Sormunen, R., Mäkikallio, K., Kivirikko, K. I., and Soininen, R. (2002) Inactivation of the Lysyl Oxidase Gene Lox Leads to Aortic Aneurysms, Cardiovascular Dysfunction, and Perinatal Death in Mice. Circulation 106, 2503−2509. (4) Kanematsu, Y., Kanematsu, M., Kurihara, C., Tsou, T., Nuki, Y., Liang, E. I., Makino, H., and Hashimoto, T. (2010) Pharmacologically Induced Thoracic and Abdominal Aortic Aneurysms in Mice. Hypertension 55, 1267−1274. (5) Szabo, Z., Crepeau, M. W., Mitchell, A. L., Stephan, M. J., Puntel, R. A., Yin Loke, K., Kirk, R. C., and Urban, Z. (2006) Aortic aneurysmal disease and cutis laxa caused by defects in the elastin gene. Journal of Medical Genetics 43, 255−258. (6) Oitate, M., Hirota, T., Murai, T., Miura, S., and Ikeda, T. (2007) Covalent Binding of Rofecoxib, but Not Other Cyclooxygenase-2 Inhibitors, to Allysine Aldehyde in Elastin of Human Aorta. Drug Metab. Dispos. 35, 1846−1852. (7) Oitate, M., Hirota, T., Takahashi, M., Murai, T., Miura, S., Senoo, A., Hosokawa, T., Oonishi, T., and Ikeda, T. (2007) Mechanism for Covalent Binding of Rofecoxib to Elastin of Rat Aorta. J. Pharmacol. Exp. Ther. 320, 1195−1203. (8) Oitate, M., Hirota, T., Koyama, K., Inoue, S., Kawai, K., and Ikeda, T. (2006) Covalent binding of radioactivity from [14C]rofecoxib, but not [14C]celecoxib or [14C]CS-706, to the arterial elastin of rats. Drug Metab. Dispos. 34, 1417−1422. (9) Schmidt, A., Busse, W., Garthoff, B., Gau, W., Ritter, W., Wünsche, C., and Buddecke, E. (1984) Influence of muzolimine on arterial wall elastin. Biochem. Pharmacol. 33, 1915−1921. (10) Moore-Jones, D., and Perry, H., Jr (1966) Radioautographic localization of hydralazine-1-C-14 in arterial walls. Exp. Biol. Med. 122 (2), 576. (11) Ohta, K., Yamaguchi, J. I., Akimoto, M., Fukushima, K., Suwa, T., and Awazu, S. (1996) Retention mechanism of imidazoles in connective tissue. I. Binding to elastin. Drug Metab. Dispos. 24, 1291−1297. (12) Ohta, K., Fukasawa, Y., Yamaguchi, J., Kohno, Y., Fukushima, K., Suwa, T., and Awazu, S. (1998) Retention mechanism of imidazoles in
In summary, we have shown that AZD5248 is retained in aortic tissue and put forward a mechanistic hypothesis based on aldehyde reactivty to explain this finding. We have proposed a related mechanism for the aortic retention of another AstraZeneca compound ZD4407 for which no previous rationale had been provided. We have provided supporting evidence in the form of an aldehyde reactivity assay and simple structure− activity relationships that are consistent with the proposed mechanism. Our results show that the combination of simple aldehyde reactivity assessment combined with a more biologically relevant second tier competitive covalent binding assay provides a powerful, yet simple, cascade to detect and optimize away from irreversible aortic binding. The importance of the second tier assay should not be underestimated as one can anticipate certain reversible aldehyde adducts being formed leading to false positives, and care should also be taken extrapolating the results from AZD5248 to related structures. Without this kind of screening, the nucleophilic reactivity responsible for the aortic binding described in this, and other papers, would not normally be assessed during the routine characterization of candidate drugs. While not anticipated to be highly prevalent, this reactivity is of potentially high impact if detected late in the drug development process.
Corresponding Author
*Tel: +44 1625 515478. E-mail: Steve.Swallow@astrazeneca. com. 1998
DOI: 10.1021/acs.chemrestox.5b00236 Chem. Res. Toxicol. 2015, 28, 1991−1999
Article
Chemical Research in Toxicology connective tissue. IV. Identification of a nucleophilic imidazolone metabolite in rats. Biol. Pharm. Bull. 21, 1334−1337. (13) Ohta, K., Kohno, Y., Fukushima, K., Suwa, T., and Awazu, S. (1999) Effect of 2-Methylimidazole on the Mechanical, Biochemical, and Morphological Properties of the Rat Aorta by Its Chronic Treatment along Maturation. Yakubutsu Dotai 14, 16. (14) Hutton, J., Jones, A. D., Lee, S. A., Martin, D. M. G., Meyrick, B. R., Patel, I., Peardon, R. F., and Powell, L. (1997) Use of a Titanium Thienyl Anion and a Simple Procedure for Introducing a Thiol Group into Thiophene in the Development of a Manufacturing Route to the 5Lipoxygenase Inhibitor ZD4407. Org. Process Res. Dev. 1 (1), 61. (15) Furber, M., et al. (2014) Cathepsin C Inhibitors: Property Optimization and Identification of a Clinical Candidate. J. Med. Chem. 57, 2357−2367. (16) Mithieux, S. M., and Weiss, A. S. (2005) In Elastin (Parry, D. A. D., and John M. Squire, J. M., Eds.) Advances in Protein Chemistry, Vol. 70, pp 437−461, Academic Press, New York. (17) Akagawa, M., Yamazaki, K., and Suyama, K. (1999) Cyclopentenosine, Major Trifunctional Crosslinking Amino Acid Isolated from Acid Hydrolysate of Elastin. Arch. Biochem. Biophys. 372, 112. (18) Fowles, L. F., Beck, E., Worrall, S., Shanley, B. C., and de Jersey, J. (1996) The formation and stability of imidazolidinone adducts from acetaldehyde and model peptides: A kinetic study with implications for protein modification in alcohol abuse. Biochem. Pharmacol. 51, 1259− 1267. (19) Shermolovich, Y. G., and Yemets, S. V. (2000) Alkylation of 1phenyl-3,5-disubstituted pyrazoles with polyfluorinated aliphatic aldehydes: Properties of 1-phenyl-4-(1-hydroxypolyfluoroalkyl)pyrazole derivatives. J. Fluorine Chem. 101, 111−116. (20) D’Auria, M., Mauriello, G., and Racioppi, R. (1999) An unusual oxidation of thiazol-2-ylmethanol in hydrolytic conditions. J. Chem. Soc., Perkin Trans. 1 1, 37. (21) Sun, L., Tran, N., Tang, F., App, H., Hirth, P., McMahon, G., and Tang, C. (1998) Synthesis and Biological Evaluations of 3-Substituted Indolin-2-ones: A Novel Class of Tyrosine Kinase Inhibitors That Exhibit Selectivity toward Particular Receptor Tyrosine Kinases. J. Med. Chem. 41, 2588−2603. (22) Chen, N., Huang, Z., Zhou, C., and Xu, J. (2011) Thorpe−Ingold effect in the reaction of vicinal amino primary alcohol hydrogen sulfates and carbon disulfide. Tetrahedron 67, 7971−7976. (23) Adunsky, A., Chandler, J., Heyden, N., Lutkiewicz, J., Scott, B. B., Berd, Y., Liu, N., and Papanicolaou, D. A. (2011) MK-0677 (ibutamoren mesylate) for the treatment of patients recovering from hip fracture: A multicenter, randomized, placebo-controlled phase IIb study. Arch. Gerontol. Geriatr. 53, 183−189. (24) Khojasteh-Bakht, S. C., O’Donnell, J. P., Fouda, H. G., and Potchoiba, M. J. (2005) Metabolism, Pharmacokinetics, Tissue Distribution, and Excretion of [14]CP-424391 in Rats. Drug Metab. Dispos. 33, 190−199.
1999
DOI: 10.1021/acs.chemrestox.5b00236 Chem. Res. Toxicol. 2015, 28, 1991−1999