Novel lead structures and activation mechanisms for CO-releasing molecules (CORMs)

Abstract

Carbon monoxide (CO) is an endogenous small signalling molecule in the human body, produced by the action of haem oxygenase on haem. Since it is very difficult to apply safely as a gas, solid storage and delivery forms for CO are now explored. Most of these CO-releasing molecules (CORMs) are based on the inactivation of the CO by coordinating it to a transition metal centre in a prodrug approach. After a brief look at the potential cellular target structures of CO, an overview of the design principles and activation mechanisms for CO release from a metal coordination sphere is given. Endogenous and exogenous triggers discussed include ligand exchange reactions with medium, enzymatically-induced CO release and photoactivated liberation of CO. Furthermore, the attachment of CORMs to hard and soft nanomaterials to confer additional target specificity to such systems is critically assessed. A survey of analytical methods for the study of the stoichiometry and kinetics of CO release, as well as the tracking of CO in living systems by using fluorescent probes, concludes this review. CORMs are very valuable tools for studying CO bioactivity and might lead to new drug candidates; however, in the design of future generations of CORMs, particular attention has to be paid to their drug-likeness and the tuning of the peripheral ‘drug sphere’ for specific biomedical applications. Further progress in this field will thus critically depend on a close interaction between synthetic chemists and researchers exploring the physiological effects and therapeutic applications of CO.

Linked Articles

This article is part of a themed section on Pharmacology of the Gasotransmitters. To view the other articles in this section visit http://dx.doi.org/10.1111/bph.2015.172.issue-6

Carbon monoxide: a small signalling molecule with specific target structures

Carbon monoxide (CO) is now well established as the third small signalling molecule in higher organisms, including humans (Wu and Wang, 2005; Kim et al., 2006; Piantadosi, 2008). It is endogenously generated by the action of haem oxygenase (HO) on haem (see Alexander et al., 2013c), which also leads to the formation of iron(II) and biliverdin. The latter is further converted to bilirubin by biliverdin reductase. The enzymatic mechanism has been elucidated at the molecular level by a combination of spectroscopic and theoretical approaches (Chen et al., 2008; Lai et al., 2010; Matsui et al., 2010) as well as X-ray crystal structure analysis (Friedman et al., 2003; Rahman et al., 2008; 2009). The two isoforms of HO, constitutively expressed as HO-2 and inducible HO-1, are involved in a large number of important physiological processes, in particular vasodilatation and responses to stress conditions (Morse and Choi, 2002; Ryter et al., 2002; 2006; Abraham and Kappas, 2008; Piantadosi, 2008; Rochette et al., 2013). CO signalling is intimately intertwined with that of the other two small signal-transducing molecules, NO and H₂S (Ignarro, 1999; Mustafa et al., 2009; Kajimura et al., 2010; 2012; Szabo, 2010; Li et al., 2011). However, there is a fundamental difference in the reactivity of CO compared to NO and H₂S (Fukuto et al., 2012), since at ambient pressure and temperature and in the absence of special catalysts, CO will only bind to transition metal centres, whereas the other two species react rapidly with both metal centres and many of the organic constituents of biological systems. This relative inertness and selective reactivity of CO has special implications for both the cellular target structures of CO and the development of delivery systems for CO in potential therapeutic applications (Motterlini and Otterbein, 2010; Wegiel et al., 2013). Thus, the present review will first summarize the most important properties of CO and its binding to metal centres for the non-chemist. Then, a short but critical view will be given on the currently discussed cellular target structures of CO. The main focus, however, is on artificial systems for the delivery of CO to cells and tissues and the activation mechanisms that trigger the release of CO from such carriers. The most important lead structures of CO-releasing molecules (CORMs) will be critically discussed, with particular emphasis on the latest development in this field during the past few years. Another important issue is the in vitro and ultimately in vivo detection and tracking of CO and CORMs, which will be discussed before this review concludes with an outlook on the challenges of turning CORMs into viable drug candidates.

Carbon monoxide: properties and binding to metal centres

CO is a colourless and odourless flammable gas, which, in spite of its signalling function, is highly toxic to organisms at elevated concentrations. In that context, it should be noted that it is actually not so much an interference with oxygen transport in the blood due to tight binding of CO to the haem iron centres in haemoglobin, but rather the increase in tissue CO concentrations occurring upon inhalation, which disturb normal mitochondrial function, that is responsible for CO toxicity. As elegantly demonstrated by Goldbaum and co-workers, mammals actually have a rather large ‘built-in’ margin of safety in the oxygen-carrying capacity of the blood, and haemoglobin levels available for O₂ transport can be reduced to about 20–30% of the normal without detrimental effects, as long as high lung CO levels do not shift the equilibrium towards increased tissue accumulation (Goldbaum et al., 1975; Foresti and Motterlini, 2010). With a melting point of −205.1°C and a boiling temperature of −191.5°C, CO exists as a gas at ambient conditions, which is only sparingly soluble in water (0.35 L of CO per 1 L of H₂O or 14 mM). The short carbon-oxygen bond distance (1.1282 Å in the gas phase) and high dissociation energy (1070 kJ·mol⁻¹) are in line with the C≡O triple bond character. This is particularly evident from the molecular orbital (MO) diagram shown in Figure 1A. The overlap of the 2s atomic orbitals of carbon and oxygen leads to two σ and σ* MOs, whereas the 2p atomic orbitals form three bonding and three anti-bonding MOs, a pair of σ/σ* character and two each of π and π* character. Four of the bonding but only one of the anti-bonding MOs are doubly occupied, resulting in this exceptionally strong bonding. However, of particular importance for the reactivity and binding of CO to metals, is the presence of relatively low-lying empty MOs of π* character. As shown in Figure 1B, the energetically highest filled orbital of CO, which is of σ* character, can overlap with symmetry-adapted empty d orbitals on a transition metal centre along the M-C vector to form a bonding σ-donor interaction. At the same time, there is also a π-donor interaction between the next highest occupied π orbital on the CO with properly oriented empty metal d orbitals. In contrast, occupied metal d orbitals can also overlap with the empty low-lying MOs on the CO, forming an additional π-acceptor interaction, the so-called backbonding, which is among the characteristic features of transition metal carbonyl complexes and is responsible for the particular stability of this type of compound (Elschenbroich, 2006). Therefore, carbonyl complexes preferentially form with transition metal centres in relatively low oxidation states, since these provide filled d orbitals of proper energy to facilitate the backbonding. Furthermore, the strength of the M-(CO) bond can be modulated by the presence of electron-donating or withdrawing coligands on the metal-carbonyl fragment. A decrease in the metal d electron density will result in subsequent weakening of the backbonding and thus facilitate CO release from the metal coordination sphere. This can be achieved, for example, by chemical or electrochemical oxidation of a low-valent metal centre. Alternatively, electronic excitations due to absorption of light of a proper wavelength might also reduce the metal charge density in the excited state, particularly when metal-to-ligand charge transfer transitions are involved. These binding characteristics have important implications for the triggering of CO release from stable prodrug compounds, but also on the potential cellular target structures of CO. Another special feature of CO as a ligand is that it can coordinate terminally to a metal centre but also act as a μ₂- and even μ₃-ligand, bridging two or three metal centres respectively. Furthermore, in a suitable environment, it can even change its coordination mode from terminal to bridging and vice versa (Figure 1C).

Figure 1.

(A) Molecular orbital (MO) diagram of CO; (B) bonding and backbonding interactions in transition metal-CO complexes; and (C) CO coordination modes (Elschenbroich, 2006).

Cellular target structures of carbon monoxide

Keeping in mind the special stability of the M-(CO) bond due to the interplay of bonding and backbonding interactions, it is very hard to imagine a recognition and binding of CO without the involvement of a transition metal centre, at least under physiological conditions. Therefore, it is not surprising that, in particular, haem proteins have been implicated as the primary cellular targets for CO binding. Soluble guanylate cyclase (sGC) has been the focus of many earlier works in this field. It is a haem protein in which the haem co-factor is covalently anchored to the protein via an axial histidine ligation to the iron to give rise to a five-coordinate metal centre, with the position trans to the histidine able to bind either NO or CO, but not O₂ (Poulos, 2006; Derbyshire and Marletta, 2012). This leads to an increase in cGMP production from GTP by about two orders of magnitude, at least in the case of NO. The cGMP then triggers further down-stream signalling events, including phosphodiesterase and protein kinase activation as well as modulation of gated ion channels. The study of sGC structure and function has, however, for a long time been hampered by problems with the overexpression of the full-length protein, and thus, usually only shortened constructs are investigated. Furthermore, sGC activation by CO alone only leads to a two- to fourfold increase in cGMP production. Additional synthetic sGC activators such as YC-1 were shown to potentiate these effects (Ibrahim et al., 2010), but the search for natural analogues thereof has so far remained elusive. Based on the very low sGC activation by CO alone, it is questionable whether this is indeed the main target of normal CO signalling in cells. Other haem proteins, where this activity seems to be better defined, have been identified more recently. Especially mentioned here is cystathionine-β-synthase, one of the enzymes involved in H₂S biosynthesis from sulphur-containing amino acids, and thus one of the sites of cross-talk between the different small-molecule signalling systems (Puranik et al., 2006; Weeks et al., 2009; Li et al., 2011). Other targets include nuclear hormone receptors (Gupta and Ragsdale, 2011), and particularly in the case of very simple organisms, different gas sensor proteins (Uchida and Kitagawa, 2005; Podust et al., 2008). Particularly intriguing, however, is the recent discovery that the activity of different ion channels, such as K_Ca1.1 (also know as BK_Ca), ENaC, K_2P2.1 (also known as TREK-1), K_v (see Alexander et al., 2013b), P2X₂ and P2X₄ (Alexander et al., 2013a), can be modulated by the application of CO (Wilkinson and Kemp, 2011). Taking the K_Ca1.1 channels as an example, it was demonstrated that they can be directly stimulated by CO and thus act as a gas sensor (Hou et al., 2009). However, the initial hypothesis that haem bound to the channel is the CO binding site is not in line with more recent results and the location of the haem binding site and the CO interaction seem to be spatially separated. Through site-directed mutagenesis, a number of histidine and asparagine residues in the RCK1 domain were shown to be essential for the CO sensitivity, particularly His³⁶⁵, His³⁹⁴ and Asp³⁶⁷ (Hou et al., 2008). However, the molecular mechanism of the interaction of CO with this part of the ion channel is still elusive. In the absence of an X-ray crystal structure of the channel or at least of the RCK1 domain, homology arguments have been drawn upon to suggest that this sensory domain does not contain a haem co-factor or any other metal. Since membrane proteins are notoriously difficult to handle and purify, it is possible that a relatively labile metal site has been lost during the channel isolation and preparation. Clearly, more careful biophysical characterization of the native ion channels is urgently required, including metal analysis of the different preparations with inductively coupled plasma mass spectrometry or atomic absorption spectroscopy, and electron paramagnetic resonance (EPR) as well as Mössbauer spectroscopy to try to detect a metal centre, with the latter method particularly well-suited for iron-containing biomolecules. An intriguing prospect is based on the observation that haem oxygenase 2 (HMOX2) seems to associate in the cellular membrane with K_Ca1.1 channels. Thus, it could be envisaged that both the ferrous iron as well as the CO generated by HMOX2's action on haem actually migrate within such a multi-enzyme cluster, without ever escaping to the cytosol, to form a very sensitive metal site in situ by iron coordination to the above-mentioned amino acid-derived ligands, which then binds the CO, but is easily lost upon handling during the purification procedure. A particularly well-studied example of such internal gas channels is found in dual-functional CO dehydrogenase/acetyl-coenzyme A synthase as part of the Wood-Ljungdahl pathway of anaerobic CO₂ and CO fixation in certain microorganisms (Ragsdale et al., 2012). Furthermore, since NO has been shown to interact with iron-sulphur clusters, leading to the formation of dinitrosyl iron complexes (Tinberg et al., 2010), one could speculate that CO might also target this very important class of metal centres, having functions in electron transfer and catalysis in a very wide range of proteins.

Delivery systems for carbon monoxide gas

One option to modulate the concentration of CO in an organism for biomedical applications is of course the pharmacological stimulation of haem oxygenase-1 activity (Li et al., 2007; Abraham and Kappas, 2008). However, many of the current approaches are more focused on an exogenous delivery of CO. The simplest way to achieve this is inhalation of a gas mixture containing CO. Due to the very high general toxicity of CO, this has to be carried out in a carefully controlled way, to prevent adverse effects on the patient and the medical personnel. In fact, there are devices now available that will adjust the quantity of CO gas delivered, depending on the breathing rate as well as other parameters, and immediately shut off the supply under abnormal conditions (Motterlini and Otterbein, 2010). Although this mitigates the safety issues associated with this mode of delivery, a fundamental drawback of the application of CO by inhalation remains. This is related to the fact that once the CO is taken up via the respiratory system, its further distribution is determined by the partition ratio between the different body fluids and tissues. Since this is a fixed value, it is very difficult to specifically address particular disease sites in the organism using this form of application, particularly when they are not part of the vascular system.

CORMs: general design principles and activation mechanisms for CO release

Due to the above-mentioned inherent limitations in the use of CO gas, about 10 years ago, Motterlini and co-workers introduced the concept of using a chemically bound form of CO as a prodrug for physiological CO release, for which the term CO-releasing molecules (CORMs) was coined (Motterlini et al., 2002). Although a number of organic compounds was initially explored for their potential as CORMs, including haloalkanes, aldehydes, oxalates and silacarboxylic acids (Mann, 2010; Romao et al., 2012), it turned out that their release rate and toxicological profile were not favourable to justify further development. Consequently, due to the strong but modifiable binding of CO to transition metal centres, most of the research on CORMs has focused on metal carbonyl complexes. A notable exception, however, is a family of main group boranocarbonates introduced by Alberto's group (Motterlini et al., 2004; Pitchumony et al., 2008; 2010). Furthermore, very recently, two organic systems based on unsaturated cyclic 1,2-diketones and xanthene-9-carboxylic acid were introduced as the first really viable purely organic CORMs (Antony et al., 2013; Peng et al., 2013). Nevertheless, it is still the transition metal-based CORMs that offer the greatest flexibility in terms of molecular design and thus they will be the focus of the discussion herein. This is due to the fact that by proper selection of the metal centre, the number and spatial arrangement of CO ligands around it, and the nature of the coligands completing the preferred coordination sphere of the metal, it is possible to tune the CO-release properties in a wide range (Figure 2). Consequently, this ‘inner’ part of such a molecule has been termed the CORM sphere or coordination sphere by Romao et al. (2012). Key parameters are the number of CO molecules that can be released from the metal coordination sphere, the kinetics of the CO release process and the trigger mechanism required to initiate the liberation of the CO. Currently, there is no consensus on whether a high or a low number of CO ligands per metal complex unit would be more desirable and whether the CO release should be slow or fast. Monocarbonyl compounds of course have the simplest release kinetics, while in the case of more than one labile CO present per molecular unit, these might be liberated consecutively at a different speed, thus making kinetic analysis and identification of potential intermediates much more difficult. Furthermore, it is not clear whether a low but steady release of CO from CORMs would be most beneficial, or rather a quick burst of CO liberated. Very likely, there is no single answer to these questions and, instead, different sets of CORMs should be designed with these properties specifically optimized for a particular medical application in mind.

Figure 2.

The ligand environment around a metal-carbonyl complex can be divided in a ‘CORM sphere’ and a ‘drug sphere’. While the former determines the stoichiometry and kinetics of the CO release from the metal centre by proper choice of metal, metal oxidation state and coligands, the latter, as mainly defined by the periphery of the coligands L, can be modulated to enable tissue-specific targeting (Romao et al., 2012).

In addition to the CORM sphere, which can be tuned to control the CO release, it is also possible to modify the (co)ligand periphery pointing away from the metal centre. This is particularly important and the most significant advantage of CORMs over simple CO gas, since it allows the modulation of their partition ratio between the different body fluids and tissues and might also enable an active or passive targeting of specific cell subpopulations. Thus, the ‘outer part’ of CORMs has been termed the drug sphere (Figure 2), since it will dominate the pharmacological profile of these compounds (Romao et al., 2012). Unfortunately, much research has so far mainly focused on the CO-release properties and the proper tuning of the drug sphere has been more or less neglected, although this is absolutely vital if CORMs are indeed to be used beyond simple tools for fundamental biological studies on small-molecule signalling and turned into real drugs. However, the optimization of the drug sphere will require a clear focus on selecting and treating a particular medical condition, as nicely illustrated by liver targeting of CORMs in recent work from the Romao group (Marques et al., 2012). In the context of targeting specific cells and tissues, another currently unresolved question is whether it will actually be necessary for the CORM to exert its biological action to enter the cells by passive or active uptake, or whether it will be sufficient to liberate the CO close enough to the target site, which it will then reach by diffusion through cellular membranes. Since the diffusion behaviour and maximum free path length of CO in complex biological systems is not known at present, it is difficult to estimate how close will be close enough. However, some mathematical modelling on the ‘sphere of action’ of H₂S has already been carried out and it would be highly desirable if such methodology is also applied to the study of CO distribution (Cuevasanta et al., 2011).

Since CORMs serve as prodrugs for CO delivery, a very important question is also the trigger mechanism by which the CO liberation is actually induced, since this will be important to control site-specific delivery. In most of the metal-carbonyl complexes initially investigated for their suitability as CORMs, the release of CO from the metal coordination sphere is evoked by ligand exchange reactions with the medium (Figure 3). For example, while CORM-3, [RuCl(glycinato)(CO)₃], has a half-life of 98 h in distilled water at 37°C, this is reduced by three orders of magnitude to just 3.6 min in human plasma, probably due to an interaction with the thiol from glutathione (Johnson et al., 2007). Since upon i.v. delivery, CORMs will immediately be exposed to high concentrations of bio(macro)molecules as potential ligands, the tissue distribution and targeting capability of such ligand-exchange triggered ‘conventional’ CORMs will largely depend on the fine balance between their half-life in complex medium and the time required to reach and accumulate at a specific target site in the body. One way to modulate this is by incorporation of the CORM inro a local hydrophobic environment and such an approach has indeed been explored by incorporation of CORM-3 into a triblock co-polymer composed of a CO-releasing domain flanked by hydrophilic and hydrophobic sections, which self-assemble to a CO-releasing micelle with release kinetics slowed down compared with the free parent compound (Hasegawa et al., 2010).

Figure 3.

Common trigger mechanisms to initiate CO release from a metal coordination sphere: (A) ligand-exchange triggered, (B) enzyme-triggered and (C) photoinduced release.

In addition, researchers have started to explore alternative trigger mechanisms, in which a metal-carbonyl complex serves as a prodrug entirely stable towards ligand exchange even with highly abundant biological nucleophiles, and the CO release is only induced by a proper internal or external stimulus (Figure 3).

The most attractive external stimulus is probably light, since highly focused and pulsed light sources will allow for a very precise spatial and temporal control of the biological activity of such photoactivatable CO-releasing molecules (PhotoCORMs) (Schatzschneider, 2011; Rimmer et al., 2012). In addition to topical applications, for example, in the treatment of skin diseases, it can generally also be envisioned to target sites deeper inside solid tissue due to modern developments in waveguide technology, but the direct correlation between excitation wavelength and tissue penetration depth remains a major problem (Lane, 2003; Szacilowski et al., 2005; Agostinis et al., 2011). Red light or even infrared photoactivation is the ultimate goal, both to ensure that structures deeper inside the body can be reached and to minimize photodamage to healthy tissue in the beam pathway.

A different way of elegantly combining external stimulation with ligand-exchange triggered CO-release from metal carbonyl complexes was recently reported by Kunz et al. (2013). In this work, a catecholate-modified CORM-3 was anchored to maghemite (Fe₂O₃) nanoparticles, which, upon exposure to an alternating magnetic field, locally heat up; this then accelerates the ligand exchange of CO with medium, leading to a twofold increase in CO release in some of the models studied.

More recently, work has also started to utilize differences in cellular microenvironments for a localized control of the CO release. While parameters like cellular pH and redox environment might also be exploited in this context, an interesting alternative approach was taken by the group of Schmalz, which is based on potential differences in cellular enzyme expression rates. These researchers prepared metal carbonyl complexes in which an organometallic ligand is ‘trapped’ in one of two possible tautomeric forms by ester formation or O-phosphorylation. Exposure of these CO prodrugs to esterases or phosphatases leads to a hapticity change of the metal-carbonyl unit, resulting in increased sensitivity to dioxygen and subsequent CO release (Romanski et al., 2011; 2012a,b; Botov et al., 2013). Although these enzyme-triggered CO-releasing molecules (ET-CORMs) are still at an exploratory stage, this approach also holds great promise for tissue-specific, internally-triggered, controlled CO bioactivity.

CORMs: important lead structures and novel developments

A significant number of different CORMs based on transition metal as well as main group elements have been reported during the past decade and a ‘periodic table of CORMs’ is shown in Figure 4. Since this extensive body of work has already been reviewed, for example, by Mann and co-workers (Mann, 2010; Romao et al., 2012; Zobi, 2013), no attempt will be made here to provide a comprehensive coverage of all CORMs published to date. Instead, the discussion will focus on some of the most important and most widely used CORMs with a particular focus on their benefits and limitations, and then particularly new developments and further prospects for the field.

Figure 4.

Periodic table of CO-releasing molecules (CORMs). Elements for which CORMs have been reported to date are highlighted in grey.

Initially, iron pentacarbonyl (Fe(CO)₅) and dimanganese decacarbonyl (Mn₂(CO)₁₀) were explored. However, these were found to be only poorly bioavailable due to their non-polar nature and also require photoactivation (Motterlini et al., 2002). Thus, most studies on the cellular delivery and biological activity of CO today utilizes either CORM-2 ([RuCl(μ-Cl)(CO)₃]₂) or CORM-3 ([RuCl(glycinato)(CO)₃]). Although many interesting results have been obtained employing these two CORMs (Figure 5), they are far from ideal for this purpose. In particular, CORM-3 shows a very complicated and solvent-dependent speciation in solution, which is also further influenced by the pH of the medium. A large number of different isomers and adducts can form, which are difficult to track down and might have variable CO release kinetics. This is particularly evident in the case of CORM-3, which shows a wide variation in half-life depending on the medium (Johnson et al., 2007). Furthermore, the Ru(CO)_n fragment was shown to bind to surface-accessible amino acid side chains in proteins such as lysozyme (Santos-Silva et al., 2011; Santos et al., 2012). Since ruthenium complexes are widely explored in the context of anti-cancer chemotherapy (Alessio et al., 2004; Melchart and Sadler, 2006; Bratsos et al., 2007; Levina et al., 2009; Süss-Fink, 2010), it is very important to keep in mind that the metal–coligand fragment remaining after liberation of CO from the coordination sphere might well possess biological activity on its own. Unfortunately, these reaction products, termed inactivated CORMs (iCORMs), are often not structurally characterized and, in many cases, not assayed for bioactivity separately. Thus, unless very careful controls are included in a study, using both CO gas as a positive and well-characterized iCORMs as a negative control, it remains doubtful whether the biological effects observed are actually indeed due to the CO only, or rather result from the metal–coligand fragment remaining, or at least a combination of both. Furthermore, since these CORMs have not been designed for a specific biomedical application, their drug sphere is probably not ideal in terms of cellular targeting, uptake and intracellular distribution.

Figure 5.

Selected examples of CO-releasing molecules, by activation mechanism: (A) ligand-exchange triggered CORMs, (B) enzyme-triggered CORMs (ET-CORM), and (C) photoactivated CORMs (PhotoCORMs).

A notably different approach has been taken by Romao et al., who very carefully modified and explored a large family of molybdenum(0) tricarbonyl complexes of the general formula [Mo(CO)₃(L)₃], in which L is a neutral monodentate ligand, for the treatment of acute livery injury resulting from poisoning with acetaminophen (Marques et al., 2012). It turned out that isonitriles were the ligands of choice, and the correct selection of peripheral functional groups indeed led to the desired tissue selectivity. Furthermore, for a related molybdenum(0) compound, Na[Mo(histidinate)(CO)₃], it was shown that this is finally metabolized to a polyoxomolybdate cluster [PMo₁₂O₄₀]³⁻, as demonstrated by X-ray crystallography of this species bound to lysozyme (Seixas et al., 2013). Thus, instead of somewhat arbitrarily preparing and screening further metal carbonyl complexes, there is a very clear need for compounds that are designed for a very particular biomedical application and thoroughly examined not only for their CO release behaviour, but also the follow-up products.

Since the compounds discussed previously all start to undergo CO release as soon as they are dissolved in the medium, it is largely their half-lives in the circulation after i.v. injection that determine the target structures that can be addressed in the body. Recent work is therefore aimed at CORM prodrugs that are stable in serum and only get triggered in the targeted tissue by a very specific stimulus. A very interesting example of this class of compounds is the enzyme-triggered CORMs (ET-CORMs) from Schmalz's group (Romanski et al., 2011; 2012a,b; Botov et al., 2013). This concept is based on the trapping of an α,β-unsaturated ketone, such as cyclohexenone, in the enolate form by ester formation or O-phosphorylation. The resulting diene, with the acyloxy or phosphoryloxy group in either the 1 or 2 position, then acts as a η⁴-ligand to the Fe(CO)₃ fragment (Figure 5), which is introduced by reaction with diiron nonacarbonyl. Cleavage of the C–O or P–O bond by suitable esterases or phosphatases regenerates the dienol, which undergoes a hapticity change from η⁴ to η². These intermediates are much more prone to oxidative decomposition, giving rise to ferric iron and the α,β-unsaturated ketone, as well as the release of all three equivalents of carbon monoxide. Differences in tissue expression rates of esterases or phosphatases between organs as well as healthy and diseased sites might allow for a locally restricted CO release and thus bioactivity. Although the potency of the iron(III) as well as the α,β-unsaturated ketone needs to be further explored in separate controls, this system also gives rise to well-defined follow-up products, which can be independently prepared and assessed. An alternative, external trigger is based on the use of light to induce CO release from a transition metal carrier system (Schatzschneider, 2010; 2011; Rimmer et al., 2012). In addition to the problems with independent evaluation of the iCORM products, which is inherent to all CORMs (vide supra), these PhotoCORMs face an additional difficulty due to the inverse correlation between the tissue penetration depth of light and the incident wavelength. Thus, the further the PhotoCORM absorption can be shifted to the red or even infrared part of the electromagnetic spectrum, the more likely the deeper structures can be addressed inside the tissue (Lane, 2003; Szacilowski et al., 2005). In addition to tuning the PhotoCORM absorption itself, alternative strategies might be based on its conjugation to established photosensitizers, the exploration of two-photon excitation and the use of up-converting nanoparticles. A significant number of such compounds have been explored in the last few years, in particular based on iron and manganese compounds (Niesel et al., 2008; Gonzalez et al., 2011; 2012a,b; Ward et al., 2012). A particularly interesting PhotoCORM has been reported by Pierri and Ford. The rhenium(I) compound [Re(bpy)(CO)₃(PR₃)]⁺, with R = CH₂OH, upon photolysis at 405 nm, undergoes specific liberation of only one of the three carbonyl ligands, the one trans to the phosphane group (Pierri et al., 2012). The most interesting feature of this system is due to the fact that both the starting material as well as the resulting aqua compound are luminescent and can be tracked inside cells by confocal fluorescence microscopy with excitation at 405 nm and detection at 465–495 nm versus 660 nm, respectively, for the PhotoCORM and the follow-up product (Figure 5). Also, very recently some of the first non-metal-based CORMs with photoactivation have been reported (Antony et al., 2013; Peng et al., 2013).

An alternative external stimulus to trigger CO-release locally from a prodrug system is the application of an alternating magnetic field to CORM-loaded magnetic nanoparticles, which leads to local heating and thus acceleration of the ligand exchange reaction of CO with medium, as demonstrated by Kunz et al. (2013).

Finally, a number of carrier systems have been used to confer some target specificity to CORMs. This includes conjugation to biomolecules and macromolecules as well as soft and hard nanomaterials. For example, Zobi and co-workers have attached rhenium-based CO-releasing groups to cobalamin as a potential naturally occurring carrier system (Zobi et al., 2012), whereas the Schatzschneider group as well as others use peptide conjugates for CORM conjugation (Pfeiffer et al., 2009; 2013; Matson et al., 2012). Other hard and soft nanomaterials used for the attachment of CORMs include silica nanoparticles as well as carbon nanomaterials (Dördelmann et al., 2011; 2012), dendrimers (Govender et al., 2013), and micelles (Hasegawa et al., 2010). Alternative approaches do not depend on the attachment of molecular metal-carbonyl units to nanoscale carrier systems, but are instead based on the direct loading of porous nanomaterials like metal-organic frameworks (MOFs) with CO gas, which is liberated again upon controlled degradation of the material (Ma et al., 2013). Of course, good biocompatibility of both the organic and the inorganic parts of the MOFs has to be ensured for potential therapeutic applications.

In summary, a wide range of different trigger mechanisms to induce the CO release from the metal coordination sphere and a significant number of core structures now exist, and these have been explored for use as CORMs. Future generations of CORMs to be added to this arsenal certainly have to be designed with very clear therapeutic applications in mind, in order to properly tune the release kinetics and stoichiometry to the desired range, and with a specific emphasis on the drug-like properties of any new compound. Thus, much more emphasis has to be placed on the choice and variation of the ‘drug sphere’ in the future and the biomedical community has to provide information on the required amount of CO at a target site to elicit a physiological response (high or low) and the time scale over which a certain concentration level needs to be kept (fast or slow CO release).

Detecting and tracking CO and CORMs

The quantification of the number of CO ligands released from a CORM as well as the release kinetics and the nature of potential intermediates, if the CO ligands are released in a stepwise instead of a concerted process, is a very important parameter in the characterization of CORMs.

The simplest method for this purpose is the so-called ‘myoglobin assay’, in which the conversion of dexoymyoglobin (Mb) to carboxymyoglobin (MbCO) is followed spectrophotometrically, by the decrease in intensity of the band of Mb at 557 nm and the increase of the MbCO absorption at 540 and 577 nm. The amount of CO released per molecular unit of carbonyl compound is then determined from the plateau value at the end of the measurement, which is reached once no more spectral changes can be observed at prolonged reaction time, and the kinetics can be obtained from a proper fitting of the increase in MbCO absorption with time. For this type of experiment, one always has to ensure that the molar amount of Mb exceeds the maximum number of moles of CO that a CORM may potentially release, in order to provide at least one free Mb for each CO liberated and thus prevent premature signal saturation. Although the myoglobin assay will probably remain the procedure of choice for initial screening for CORM activity since it is very easy to carry out, there are some severe limitations with this method. First of all, especially for highly coloured metal-carbonyl compounds, which is usually the case with PhotoCORMs, there might be an overlap in the absorption bands of the myoglobin, the CORM, and potentially also the follow-up iCORM (inactivated CORM) products (Figure 6), which will require complicated spectral deconvolution techniques (Atkin et al., 2011). Furthermore, the assay has to be carried out under an atmosphere of dinitrogen or argon protective gas in order to keep the iron centre in the ferrous (+II) state and to prevent formation of oxymyoglobin from air. However, some CORMs, particularly the ET-CORMs from the Schmalz group, require an oxidative process to complete the CO release from the metal coordination sphere (Romanski et al., 2011). The reduction of the myoglobin is usually carried out by the addition of a large excess of aqueous sodium dithionite, which causes additional problems due to sulphite species generated from the reductant, which might react with either the CORM or the myoglobin (McLean et al., 2012). Since commercially available dithionite is often not of the purity required for such experiments, it needs to be carefully checked and, if necessary, recrystallized before use in the myoglobin assay (McKenna et al., 1991). Another currently unresolved question is whether the release of the CO from the CORM and its binding to the myoglobin are two totally uncoupled processes, with free CO intermediately present in solution, or whether there might be a more intimate association between the CORM and the myoglobin, with a more or less direct transfer of CO between the two metal centres, at least in some special cases.

Figure 6.

Application of a trigger to a CO-releasing molecule (CORM) prodrug leads to the release of CO from the metal coordination sphere. In addition to the CO thus generated, inevitably a metal–coligand fragment, termed inactivated CORM (iCORM), is also formed, which might have biological activity on its own in addition to that of the CO. Therefore, iCORMs should always be independently prepared, fully characterized and included in bioassays as a control.

Therefore, a number of alternative methods have been explored to study the release kinetics of CO and to quantify the amount of CO liberated per mole of CORM. They are usually based on a separation of the CORM solution from the sensor unit (Mann, 2010; Rimmer et al., 2012) and also circumvent the use of reducing or anoxic conditions. The most accurate but also most expensive technique is GC, which is to be considered the gold standard in the field (Vreman et al., 2005; 2011; Bernardi et al., 2008).

An even more challenging field is the in vitro/in vivo detection of either CO endogenously generated by HO activity or exogenously delivered from CORMs. Only very recently, two fluorescent sensor systems have been described, which are initial proof of principle for this concept (Yuan et al., 2013). Thus, He and co-workers developed a genetically encoded fluorescent probe, which is based on the CO binding affinity of the CooA protein, which has been fused to yellow fluorescent protein (YFP) (see Figure 7A). Upon CO binding to the CooA domain, the whole construct, called ‘COSer’ (CO Sensor), undergoes a conformational change, which leads to the YFP emission lighting up (Wang et al., 2012). Significant signal response was found down to 1–2 μM CO and little interference was observed from imidazole, cyanide, NO and dioxygen. However, a significant drawback of this system is the need to transfect cells to be studied with a COSer-containing expression vector, which will very likely severely limit its broader application. In contrast, Chang et al followed an approach based on a synthetic small molecule, which is composed of a BODIPY (boron dipyrromethene difluoride) unit and an attached cyclometalated dimeric palladium(II) moiety, which has been termed ‘COP-1’ (CO Probe 1) (see Figure 7B). While this compound is non-luminescent, addition of CO leads to CO insertion in the carbon-palladium bond with subsequent carboxylation of the pendant substituent and the probe then lights up (Michel et al., 2012). Challenge experiments with a number of small molecules were carried out and a detection limit of about 1 μM was estimated. Although this small-molecule probe is much more suitable for general use, both in solution and in cell cultures, important questions remain. Firstly, what is the fate of the palladium metal liberated during the CO binding process by the probe, and secondly, what is the membrane permeability of COP-1 and whether a uniform distribution inside cells can generally be reached. In addition, the concentration ratio of COP-1 to CORM as well as the proper order and timing of the addition of both the fluorescent probe and the CORM to cell cultures seems to play a very important role in obtaining meaningful results. Thus, although these two systems finally provide some initial tools to study in vitro/in vivo CO generation and delivery, there is clearly an urgent need for further improvement in fluorescent CO detection in complex biological environments.

Figure 7.

(A) ‘CO Sensor’ (COSer) and (B) ‘CO Probe 1’ (COP-1) as fluorescent switch-on probes for CO.

Summary, conclusion and outlook: challenges to turn CORMs into drugs

CO is now well established as a small signalling molecule in higher organisms. While its endogenous production by the action of HO enzymes on haem and the regulation of this process is quite well understood now at the molecular level, this is not the case with the direct down-stream effects of the CO generated. Although the chemical properties of CO strongly suggest that binding to a transition metal centre will be required for its recognition, surprisingly little is known about its primary targets in the cell. Although implicated for a long time, sGC is rather questionable since its activation by CO is very low compared to NO in the absence of additional synthetic effectors, for which no natural analogue has been found so far. However, a fascinating prospect is the activation of ion channels by CO, although further light needs to be shed on the molecular details of this process. It is quite possible that a rather labile metal site has so far been lost during isolation and purification of these delicate to handle membrane proteins and better application of current biophysical techniques such as Mössbauer and EPR spectroscopy as well as metal analysis will certainly be required to solve this riddle. To modulate these signalling pathways by exogenous application of CO sources, for either fundamental biological studies or potential therapeutic applications, metal carbonyl complexes as CORMs have emerged as very valuable tools and a wide variety of such compounds with different trigger mechanisms (ligand exchange, enzymatic activation, photoactivation) is now available with modifiable release stoichiometry and kinetics. Recent efforts are also under way to modify the outer periphery of such compounds, the so-called ‘drug sphere’, to modulate the target specificity of CORMs, either by careful choice of peripheral functional groups or attachment to carrier systems such as bio(macro)molecules or hard and soft nanomaterials. Further progress in the field will depend on reliable methods to study the CO release from these carrier systems and in particular on the tracking of CORMs as well as CO in complex biological environments, both in vitro and in vivo. Although some recent progress has been made in the latter field with the introduction of the first fluorescent probes for CO, further development is urgently needed. The major challenge is now to design systems suitable for real clinical applications and take them beyond animal models of disease.

Glossary

BODIPY

boron dipyrromethene difluoride

CORM

CO-releasing molecule

EPR

electron paramagnetic resonance

ET-CORM

enzyme-triggered CORM

HO

haem oxygenase

Mb

myoglobin

MbCO

carboxymyoglobin

MO

molecular orbital

MOF

metal-organic framework

PhotoCORM

photoactivatable CORM

sGC

soluble guanylate cyclase

YFP

yellow fluorescent protein

Conflict of interest

The author has no conflict of interest to declare.