Article pubs.acs.org/journal/aidcbc
Toward Photopharmacological Antimicrobial Chemotherapy Using Photoswitchable Amidohydrolase Inhibitors Claire E. Weston,†,∥ Andreas Kram ̈ er,§,∥ Felix Colin,§ Ö zkan Yildiz,# Matthias G. J. Baud,† ,§ Franz-Josef Meyer-Almes,* and Matthew J. Fuchter*,† †
Department of Chemistry, Imperial College London, London SW7 2AZ, United Kingdom Department of Chemical Engineering and Biotechnology, University of Applied Sciences, Haardtring 100, 64295 Darmstadt, Germany # Department of Structural Biology, Max-Planck-Institute of Biophysics, Max von Laue Strasse 3, 60438 Frankfurt am Main, Germany §
S Supporting Information *
ABSTRACT: Photopharmacological agents exhibit lightdependent biological activity and may have potential in the development of new antimicrobial agents/modalities. Amidohydrolase enzymes homologous to the well-known human histone deacetylases (HDACs) are present in bacteria, including resistant organisms responsible for a significant number of hospital-acquired infections and deaths. We report photopharmacological inhibitors of these enzymes, using two classes of photoswitches embedded in the inhibitor pharmacophore: azobenzenes and arylazopyrazoles. Although both classes of inhibitor show excellent inhibitory activity (nM IC50 values) of the target enzymes and promising differential activity of the switchable E- and Z-isomeric forms, the arylazopyrazoles exhibit better intrinsic photoswitch performance (more complete switching, longer thermal lifetime of the Z-isomer). We also report protein−ligand crystal structures of the E-isomers of both an azobenzene and an arylazopyrazole inhibitor, bound to bacterial histone deacetylase-like amidohydrolases (HDAHs). These structures not only uncover interactions important for inhibitor binding but also reveal conformational differences between the two photoswitch inhibitor classes. As such, our data may pave the way for the design of improved photopharmacological agents targeting the HDAC superfamily. KEYWORDS: photopharmacology, histone deacetylase, Pseudomonas aeruginosa, arylazopyrazole, azobenzene
P
One potential disadvantage of PDT/PACT is that the photosensitizer is not selective for a particular target cell: upon light activation, it is indiscriminately cytotoxic and therefore kills normal human cells as well as the desired disease cells. In addition, there is an intrinsic requirement for oxygen for treatment efficacy. An alternative approach, which still capitalizes on the high temporal and spatial precision of light usage, but avoids the disadvantages above, is the recently established field of photopharmacology.15,16 Instead of a photosensitizer, photoswitchable ligands that are selective for a specific cellular target, such as a receptor or an enzyme, are employed as therapeutic entities. These ligands undergo a change in shape, flexibility, or electronic properties upon irradiation with light that leads to a change in the affinity for their cellular target; therefore, they exhibit a light-dependent therapeutic activity. Whereas this nascent field is showing significant promise for human therapeutic targets,15−22 far less
hotodynamic therapy (PDT) is a well-established treatment for certain cancers, for wet age-related macular degeneration, and for a variety of skin conditions, such as actinic keratosis and acne.1−5 PDT involves the administration of a photosensitizer that accumulates in a specific target cell or tissue, followed by the localized administration of a specific wavelength of light.6 The absorption of light by the photosensitizer ultimately leads to the formation of reactive oxygen species, which are highly cytotoxic. Thus, the targeted tissue is destroyed with high spatial selectivity, through controlled tissue irradiation. A functionally equivalent approachtermed photodynamic antimicrobial chemotherapy (PACT)continues to be studied both preclinically and clinically as an antimicrobial therapy7−11 and could be used to treat localized infections that are accessible by light, such as dermatological infections or respiratory infections (reached via optic fibers). Given the significant rise in antimicrobial resistance,12−14 and therefore the significant need for new approaches to treat infections from resistant organisms, alternative treatment regimens, such as PACT, are drastically needed. © XXXX American Chemical Society
Received: August 17, 2016 Published: October 18, 2016 A
DOI: 10.1021/acsinfecdis.6b00148 ACS Infect. Dis. XXXX, XXX, XXX−XXX
ACS Infectious Diseases
Article
shown an interesting inhibitory effect on the growth of P. aeruginosa upon inhibition of APAHs PA1409 and PA0321 under glycose starvation conditions.38 It was also found that one of the putative APAHs from P. aeruginosa, PA3774 (referred to as PA-HDAH herein), has much closer activity to that of B/A-HDAH, suggesting an alternate role in the biological regulation of P. aeruginosa.38 Perhaps notably, P. aeruginosa is responsible for 10% of all hospital-acquired infections, exhibits high intrinsic resistance to many antibiotics, and has a particularly high mortality rate in hospital-based infections.39−41 P. aeruginosa infections have also been a key focus in research into PACT.42−47 While we were conducting this work, two papers were published disclosing photoswitchable ligands for human HDACs19,20 (one of which20 was focused on hydroxamate inhibitors, similar to this study), further advocating the choice of HDACs (and homologues) as suitable photopharmacological targets. Herein we report our initial results on the development of photoswitchable ligands of bacterial HDAC homologues. Not only do we report the identification of ligands containing the ubiquitous photochromic azobenzene chromophore, similar to the recent studies,19,20 but we also report advanced derivatives based on our arylazopyrazole photoswitches,48 which exhibit significantly improved photoswitch performance. Furthermore, we report the first protein−ligand crystal structures of these photoswitchable ligands, which give important structural insight into the binding mode of photopharmacological ligands of the HDAC family.
work has been performed on the development of photopharmacological antimicrobial chemotherapeutic agents.23−25 Previous attempts to generate photopharmacological antimicrobial agents have involved the synthesis of quinolone antibiotics conjugated to photoswitchable scaffolds23,24 or a photoswitchable analogue of the antimicrobial peptide gramicidin S, where the photoswitch alters the peptide conformation.25 We considered an approach to develop a molecular targeted antibacterial agent, where the photochromic functionality would be intrinsic to the pharmacophore of the agent in question. Such an approach, however, requires detailed knowledge of the pharmacophoric features of a given ligand and ideally knowledge of the target binding site. Given our prior work on histone deacetylase (HDAC) medicinal chemistry26−29 and biochemistry,30,31 the very well-known pharmacophore of HDAC inhibitors32−34 (see Figure 1), and the fact that bacterial
■
RESULTS AND DISCUSSION The enzyme active site of the HDAC superfamily contains a key zinc ion that sits at the base of a narrow hydrophobic channel.37,38,49 As is well-known, the HDAC pharmacophore consists of a zinc chelating group, such as a hydroxamic acid or benzamide, a capping group that interacts with amino acids on the surface of the channel, and a connecting hydrophobic linker that lies in the channel, suitably positioning the zinc chelating group and cap (Figure 1).32−34 Using this knowledge and the established HDAC inhibitors panobinostat and belinostat as inspiration, an initial photoswitchable inhibitor was designed (Figure 1), consisting of a phenyl capping unit, an azobenzene linker, and a hydroxamic acid zinc binding group. Azobenzene photoswitches exist as E- or Z-isomers, which can be interconverted by irradiation with UVA (E-Z) or blue/green (Z-E) light.50 Due to the narrow diameter of the active site channel, our initial hypothesis was that the twisted Z-isomer would induce more unfavorable steric interactions with amino acids lining the channel than the planar E-isomer, leading to reduced inhibitory activity. To increase the proposed steric effect, several analogues with bulky para-substituents on the terminal aromatic ring were chosen for synthesis (Figure 2). (While completing this work, compound 1c was published in another study.20) When considering biological assessment of the designed photoswitchable inhibitors, we were mindful of two common issues in the usage of azobenzene photoswitches: (1) overlapping absorbance spectra between the two isomers, leading to less than quantitative photoswitching at a given wavelength;50 (2) thermal isomerization of the metastable Zisomer back to the more stable E-isomer, with a half-life of milliseconds to several days, depending on the system in question.50 Due to these two factors, assessment of the inhibitory activity of isomerically pure Z-azobenzene isomers is
Figure 1. (A) Pharmacophore for HDAC inhibition; (B) clinically approved HDAC inihibitors; (C) initial photoswitchable HDAC inhibitor design (1a), incorporating the cinnamyl hydroxamic acid moiety seen in several HDAC inhibitors, such as panobinostat and belinostat.
enzymes with high sequence homology to the HDACs are known,35,36 such as histone deacetylase-like amidohydrolase (HDAH) from Bordetella/Alcaligenes (B/A-HDAH)37 and acetylpolyamine amidohydrolases (APAHs) from Pseudomonas aeruginosa,38 we sought to develop photoswitchable amidohydrolase ligands as proof of concept photopharmacological agents against bacterial targets. Although there is still much to be learned about the role of these enzymes in the pathophysiology of a given infectious disease, we have recently B
DOI: 10.1021/acsinfecdis.6b00148 ACS Infect. Dis. XXXX, XXX, XXX−XXX
ACS Infectious Diseases
Article
Figure 2. E- and Z-azobenzene HDAC inhibitors and thermally stable E- and Z-stilbene analogues.
Figure 3. UV/vis spectra recorded in DMSO at 1 × 10−5 M concentration; E-isomers, UVA (λmax ∼ 365 nm) and 532 nm PSSs, and extracted pure Z-isomer spectra.
not straightforward. To combat this, stilbene isosteres 2a−c (Figure 2) were also prepared, with the view that these could be used as thermally stable (and synthetically accessible) models of the isomeric azobenzenes. Comparison of the IC50 values for E-azobenzenes and analogous E-stilbenes would allow for assessment of the validity of alkenes as azo mimics. Details of the synthesis of the planned azo compounds and stilbene isosteres can be found in the Supporting Information. Photoisomerization and thermal isomerization of the azo compounds was determined. Photoswitching was carried out using UVA (λmax ∼ 365 nm) or green (532 nm) light and monitored by UV/vis spectroscopy (Figure 3). A hand-held UVA lamp was used for 365 nm irradiation, whereas 532 nm irradiation was carried out using the third harmonic of a Nd:YAG laser. In our previous work,48 we found that irradiation of arylazopyrazoles with 532 nm at the very tail end of the n−π* absorbance increased the percentage of Zisomer present in the PSS compared with irradiation at wavelengths closer to the absorption maxima. For consistency between different switches, this monochromatic wavelength was used for all Z-E photoswitching in this study. Photostationary state (PSS) compositions were quantified by NMR spectroscopy, which allowed the spectra of the pure Z-isomers
to be extracted (Figure 3). Upon irradiation with UVA light, the three analogues were switched to PSSs containing 85−90% Z-isomer. Irradiating with 532 nm monochromatic light led to PSSs containing 74−82% E-isomer. This level of switching is consistent with typical PSSs obtained for azobenzenes.50 Clearly, such incomplete photoswitching is an issue for such an inhibitor design, where >10% of the opposite isomer would always be present. It has been suggested that such mixtures may not be suitable to study differential enzymatic activity in a cellular context.51 The thermal isomerization rates of the azobenzenes were initially measured in aqueous phosphate buffer at 30 °C by UV/ vis spectroscopy (Table 1), mimicking the conditions of the biological assay (vide infra). However, azobenzene 1b was found to be insufficiently soluble in aqueous solutions, with 100-fold or 10-fold dilutions from DMSO, and therefore the kinetics were measured in pure DMSO. Kinetic data in DMSO were therefore additionally obtained for the other analogues to allow for comparison (Table 1). Azobenzenes 1b,c both had faster thermal isomerization rates than 1a, due to the combination of a para-electron donating group (tert-butyl or methoxy) and a para-electron withdrawing N-hydroxyacrylamide, leading to a “push−pull” system. Such “push−pull” C
DOI: 10.1021/acsinfecdis.6b00148 ACS Infect. Dis. XXXX, XXX, XXX−XXX
ACS Infectious Diseases
Article
Table 1. Thermal Isomerization Kinetics of 1a−c at 30 °C
a
Table 2. IC50 Values with Standard Errors for EAzobenzenes 1a−c and the UVA PSSs and E- and ZStilbenes 2a−c against Bacterial HDACs, with SAHA as Positive Controla
Not measured due to insolubility in pH 7.2 phosphate buffer.
azobenzenes are well-known to have shorter thermal half-lives for the Z-isomer.50 Given the time scale of the assay (
