Correspondence: Lígia M. Saraiva, Av. da República (EAN), 2780-157 Oeiras, Portugal. Tel.: + 351 214469328; fax: + 351 214411277; e-mail: firstname.lastname@example.org
Carbon monoxide-releasing molecules (CO-RMs) are, in general, transition metal carbonyl complexes that liberate controlled amounts of CO. In animal models, CO-RMs have been shown to reduce myocardial ischaemia, inflammation and vascular dysfunction, and to provide a protective effect in organ transplantation. Moreover, CO-RMs are bactericides that kill both Gram-positive and Gram-negative bacteria such as Staphylococcus aureus and Pseudomonas aeruginosa. Herein are reviewed the microbial genetic and biochemical responses associated with CO-RM-mediated cell death. Particular emphasis is given to the data revealing that CO-RMs induce the generation of reactive oxygen species (ROS), which contribute to the antibacterial activity of these compounds.
Carbon monoxide (CO) is, at ambient temperature, a colourless and odourless gas that is generated by the incomplete combustion of fuels such as natural gas, coal, oil and wood, and is generally considered a highly poisonous gas. However, the current knowledge of the cytoprotective action of CO produced in the human body has established that CO has other effects in addition to being only a poisonous gas (Motterlini & Otterbein, 2010). To profit from the therapeutic properties of CO, and deliver it in specific and controlled ways, a large variety of CO-releasing molecules (CO-RMs) have been prepared (Romao et al., 2012). More recently, these prodrugs were also shown to act as bactericides (Nobre et al., 2007). This short review starts with a brief introduction to the biological role of CO and to the pharmacological use of CO-RMs, and focuses on the effect of CO-RMs on bacteria. It summarizes the mechanisms that underpin the bactericidal action of CO-RMs, which are associated with the production of deleterious reactive oxygen species (ROS).
Biological impact of CO in the human body
In the early 1950s, Sjostrand reported that the human body produces small quantities of CO that result from the degradation of haem catalysed by haem oxygenases (HO; Sjostrand, 1951; Tenhunen et al., 1970). Humans express two heme oxygenases, namely, the constitutive HO-2, and the inducible HO-1 that responds to cellular and systemic stress and pro-inflammatory conditions. HOs play an important physiological role in the turnover of haemoglobin, which is released upon degradation of senescent erythrocytes that takes place in the spleen, liver and kidney (Wagener et al., 2003). The breakdown products of haem catabolism are CO, biliverdin and iron. Endogenously produced CO has antioxidant and/or signalling functions that protect the cardiac, immune, respiratory and gastrointestinal mammalian systems (Wu & Wang, 2005; Kim et al., 2006; Ryter et al., 2006; Gullotta et al., 2012b). The role of CO in eukaryotes is not always beneficial and depends among several factors on the CO concentration produced and the type of cell where it acts (Gullotta et al., 2012b). Indeed, adverse CO-associated effects such as triggering of the inflammatory response and apoptosis are also observed (Gullotta et al., 2012b). Moreover, high levels of CO in the human blood correlate with the severity of health disorders such as asthma, cystic fibrosis, diabetes, cardiac disease and severe renal failure. Interestingly, the production of CO is reported to be higher in patients with bacterial infections (Zegdi et al., 2002; Foresti et al., 2008).
CO and bacteria
Several aerobic and anaerobic bacteria use CO as a source of carbon and energy for growth (Ragsdale, 2004; Oelgeschlager & Rother, 2008). In all CO-metabolizing bacteria, the CO dehydrogenase (CODH) enzyme plays a key role (Ragsdale, 2004; Oelgeschlager & Rother, 2008). This enzyme catalyzes oxidation of CO to CO2, which is then transformed into cellular carbon by reductive CO2 fixation pathways, such as the Calvin–Benson–Bassham cycle, the reverse tricarboxylic acid cycle, the 3-hydropropionate cycle or the Wood–Ljunddahl pathway (Ragsdale, 2004). The respiratory processes that can be coupled to CO oxidation are oxygen respiration, hydrogenogenesis, sulphate or sulphur respiration and carbonate respiration (Oelgeschlager & Rother, 2008).
Bacteria have several CO sensors that trigger the expression of CODH, the best known being the haem-containing transcriptional factor, CooA (Bonam et al., 1989; Roberts et al., 2001; Youn et al., 2004; Gullotta et al., 2012b). Whereas CooA seems to respond only to CO, other haem-based CO sensors such as FixLJ of Sinorhizobium meliloti, AxPDEA1 of Acetobacter xylinum, Dos of Escherichia coli and HemAT from Bacillus subtilis also bind oxygen (Table 1; Gilles-Gonzalez et al., 1994; Delgado-Nixon et al., 2000; Hou et al., 2000; Chang et al., 2001; Rodgers & Lukat-Rodgers, 2005). In Mycobacterium tuberculosis, the ligation of CO to the haem histidine kinases DosS and DosT induces the dormancy regulon, leading to a latent state that makes the bacterium unresponsive to drug therapy (Kumar et al., 2008).
Several bacteria express CO-producing HOs (Table 1), which when compared with the eukaryotic counterparts are smaller, soluble enzymes that lack the C-terminal membrane anchor domain of the mammalian enzymes (Frankenberg-Dinkel, 2004; Li & Stocker, 2009). Bacterial HOs promote degradation of the haem imported from the external environment, via the haem uptake system, to provide iron to the cell (Zhu et al., 2000b; Skaar et al., 2006; Reniere et al., 2007). However, the fate of the CO produced remains unclear.
Protein targets of CO-RMs
Many of the biological effects of CO are due to it binding to haemoproteins such as haemoglobin and myoglobin, soluble guanylyl cyclase (sGC), inducible nitric oxide synthetase, cytochrome P-450, cytochrome c oxidase, or phagocyte NADPH : oxidase. The interaction of CO with these haem proteins mediates a direct effect on protein function and eventually triggers a cascade of events, as described below.
The competition with oxygen for binding to haemoglobin (c. 240 times greater than oxygen) and the inhibition of the mitochondrial respiratory chain caused by the ligation of CO to the terminal cytochrome c oxidase are the basis of toxicity of CO to humans (Wikström et al., 1981).
The binding of CO to the haem-containing cystathionine β-synthase inhibits the protein, leading to an increase in the degree of intracellular protein methylation (Puranik et al., 2006; Yamamoto et al., 2011). CO has the ability to displace histidine, cysteine and tyrosine residues that are coordinated to metals. Indeed, this is the basis of several CO sensors where removal of the proximal histidine ligand of the haem iron by CO controls the protein's functional role (Tsai et al., 2012). CO has also been identified as a ligand to iron of the mixed metal Ni-Fe centre of hydrogenases. This is an unprecedented example of a native carbonyl complex in a biological system (Ogata et al., 2002). More recently, CO was reported to interact with proteins such as albumin, ferritin and lysozyme via a protein-Ru(II)-(CO)2 adduct. The formation of this complex accelerates the release of CO from CORM-3, suggesting that plasma proteins may control the pharmacokinetic properties of CO-RMs (Santos-Silva et al., 2011).
Although CO has affinity to other metal atoms such as cobalt, nickel and copper, so far only the direct binding of CO to iron in biological systems has been demonstrated (Bender et al., 2011). Hence, many intracellular targets for CO remain to be identified.
Therapeutic properties of CO-RMs
To overcome the limitations usually associated with gaseous drugs, a large variety of CO-RMs have been prepared. The majority of CO-RMs are composed of a transition metal (Fe, Co, Mn or Ru) bound to a variable number and type of ancillary ligands. Although non-metal CO-RMs (e.g. the boranocarbonate CORM-A1) are also available, the organometallic complexes seem to be the most suitable class of compounds to act as CO carriers. Apart from the nature of the transition metal, the members of the organometallic CO-RM family differ in the number and mode of liberation of the CO molecules. Examples of metals used in CO-RMs are: manganese (CORM-1, ALF021), ruthenium (CORM-2, CORM-3), iron (CORM-F3) and molybdenum (ALF062) (Fig. 1; Jaouen & Metzler-Nolte, 2010; Romao et al., 2012).
In spite of the stable character of the CO molecule and of its binding ability being restricted to metals, CO-RMs exhibit vasodilatory, renoprotective, anti-inflammatory and anti-apoptotic properties. Moreover, CO-RM-based therapies for inflammation, sepsis, lung injury, cardiovascular diseases, cancer and organ transplantation and preservation have been supported by preclinical studies in animals (Johnson et al., 2003; Motterlini et al., 2005; Foresti et al., 2008; Motterlini & Otterbein, 2010; Gullotta et al., 2012a). So far, those studies indicate that upon treatment with CO-RMs only a small part of the CO released is found bound to haemoglobin, as judged by the low levels of COHb present in blood (Foresti et al., 2008). Clearly, other proteins need to be targeted to support the action of CO-RMs; however, until they have been identified, their pharmacological usefulness is severely hindered. Studies of the effects of CO-RMs on bacteria, which will be described in the following sections, might provide a significant contribution to the implementation of CO-RMs as therapeutic drugs. In particular, studies in bacteria have already revealed how the metal affects the properties of CO-RMs, a factor that cannot be neglected as it contributes to formation of ROS (to be discussed).
CO and CO-RMs are bactericidal agents
As in mammals, high concentrations of CO and CO-RMs cause the death of bacteria. These antimicrobial properties have been demonstrated for Gram-negative and Gram-positive bacteria such as Escherichia coli, Staphylococcus aureus and Pseudomonas aeruginosa (Nobre et al., 2007; Desmard et al., 2009). The bactericidal concentration depends on the microorganism, its growth requirements for oxygen, and the metal present in the CO-releasing molecule. For example, for P. aeruginosa the ruthenium-based carbonyls CORM-2 and CORM-3 are more bactericidal than the manganese-containing CORM-371 (Desmard et al., 2011).
A seminal study of the effect of CO on bacteria demonstrated that both CO and CO-RMs strongly decreased the cell viability of the Gram-negative E. coli and Gram-positive S. aureus (Nobre et al., 2007). The effect observed was confirmed to be bactericidal and not simply bacteriostatic. In particular, CORM-2 and CORM-3 were demonstrated to be very efficient bacterial killers, as after 30 min of treatment between 50% and 80% of the bacteria were not viable. Furthermore, even 4 h after addition of those CO-RMs, cells were not able to resume growth (Nobre et al., 2007). ALF021 and ALF062, which contain manganese and molybdenum, respectively, also proved to reduce the viability of the two pathogens. In all cases, supplementation with haemoglobin, a CO scavenger, abolished the bactericidal effect. The observation that the effectiveness of CO-RMs was greater under anaerobic conditions was explained by the preferential binding of CO to the ferrous state of proteins. More importantly, these results demonstrate that even when microorganisms are not using the aerobic respiratory chain for growth, they are still killed by CO-RMs. This indicates that proteins other than the respiratory cytochrome c oxidase are targeted by CO.
Davidge et al. (2009) reported that the growth of E. coli was impaired by CORM-3 but not by CO. In this study, exposure of aerobically grown E. coli cells to CORM-3 caused c. 50% inhibition of bacterial respiration due to the binding of CO to the terminal oxidases. Importantly, CORM-3 impaired growth of antibiotic-resistant strains of P. aeruginosa (Desmard et al., 2009).
Table 2 summarizes the microorganisms, and their conditions of growth, that have been shown to be killed by CO-RMs.
Table 2. Metal content and concentration of CO-RMs that kill bacteria, which were grown under the indicated oxygen growth conditions
CO source (metal, concentration, oxygen conditions)
The effects of CO on the genome-wide transcriptome profile has been analysed for cultures of E. coli grown aerobically and anaerobically in minimal medium salts with CORM-2, in glycerol (aerobically), and in glycerol/fumarate (anaerobically) with CORM-3 (Davidge et al., 2009; Nobre et al., 2009). In all cases, CO-RMs caused significant alteration of the mRNA abundance of a large number of genes (Fig. 2). Under aerobic conditions, CO-RM represses the transcription of E. coli genes involved in the citric acid cycle, respiration and iron homeostasis, whilst it up-regulates the expression of genes involved in general defence mechanisms, and in methionine, sulphur and cysteine metabolism. For E. coli grown anaerobically in the presence of CO-RM, the genes involved in iron homeostasis are down-regulated, whereas those involved in zinc homeostasis and biofilm formation are induced. Furthermore, genes participating in protein homeostasis, oxidative stress, zinc and methionine metabolism, and general defence mechanisms are up-regulated independently of the oxygen conditions in which E. coli is grown (Fig. 2).
The transcription data acquired for E. coli grown aerobically with CO-RMs suggests that the respiratory chain may be hindered (Fig. 2). In accordance, P. aeruginosa treated with CORM-3 reduced oxygen less rapidly (Desmard et al., 2009). As blockage of the electron transport chain enhances the generation of ROS, the gene expression profile of E. coli in the presence of CO-RMs is expected to share similarities with its transcriptional response to hydrogen peroxide (Zheng et al., 2001; Zuckerbraun et al., 2007; Wang et al., 2009). In fact, the expression of a number of genes is affected similarly in cells treated with either of the two chemicals. They include the E. coli spy, encoding a periplasmic protein that is induced by envelope stress, the ibpA and ibpB genes, encoding two heat-shock proteins that are related to protein stability, hptX, coding for a heat shock protein, dnaK, dnaJ and hslO genes, encoding chaperones, and genes encoding proteins involved in sulphur metabolism such as sbp and cysWA (Zheng et al., 2001; Davidge et al., 2009; Nobre et al., 2009; Wang et al., 2009; Table 3).
Table 3. Escherichia coli genes associated with the generation of intracellular oxidative stress that are induced by CO-RMs
In general, the two major transcription regulators, SoxRS and OxyR, control the bacterial response to oxidative stress (Storz & Imlay, 1999; Chiang & Schellhorn, 2012). Data from DNA microarray experiments revealed that CORM-2 increases expression of the soxS gene and of members of the SoxRS regulon, such as the marAB operon, encoding a multiple antibiotic resistance protein, and micF coding for a major outer membrane porin (Nobre et al., 2009). This is consistent with the observation that E. coli single mutants with deletions in soxS and sodAB are less resistant to CORM-2 than the parental strain (Nobre et al., 2009; Tavares et al., 2011).
Studies in E. coli demonstrated that the OxyR-regulated genes dps, katG, grxA, ahpCF and trxC are up-regulated in cells exposed to sublethal concentrations of H2O2 (Zheng et al., 2001; Wang et al., 2009). Interestingly, real-time RT-PCR analysis of cells treated with a sublethal 150-μM dose of CORM-2 also caused up-regulation of katG and ahpC (our unpublished data). Furthermore, oxyR and katEG mutant strains are more susceptible to CORM-2 (Nobre et al., 2009; Tavares et al., 2011).
The microarray data revealed that the expression of several genes that are transcriptionally altered by CORM-2 is also modified in E. coli biofilm-forming cells (e.g. ibpAB, soxS and tqsA; Ren et al., 2004; Nobre et al., 2009). Consistent with these results, the biofilm content of E. coli exposed to CORM-2 increased by c. two-fold (Nobre et al., 2009). Furthermore, deletion of tqsA, a putative transport protein of the quorum-sensing signal autoinducer-2 involved in biofilm formation, yields a strain with higher resistance to CORM-2 (Nobre et al., 2009). Increased biofilm formation constitutes a defensive response of bacteria, which is triggered by several other stress agents such as hydrogen peroxide, acid and heavy metals and is associated with increased bacterial resistance (Zhang et al., 2007; Weber et al., 2010).
The yqhD gene, encoding an alcohol dehydrogenase proposed to protect cells against lipid oxidation, and yeeD, a redox protein that regulates the formation of disulphide bonds, were also induced by CORM-2 and H2O2 (Zheng et al., 2001; Perez et al., 2008; Nobre et al., 2009; Wang et al., 2009). Moreover, CO-RMs interfere with the metabolism of methionine, as judged by the alterations observed in the expression of methionine biosynthesis-related genes metF, metNI, metBL and metR (Davidge et al., 2009; Nobre et al., 2009). Consistent with these data, deletion of metR, metI and metN enhanced the sensitivity of E. coli to CORM-2, whereas supplementation with methionine abolished its bactericidal activity (Nobre et al., 2009; Tavares et al., 2011). It has been demonstrated that oxidative stress is associated with methionine auxotrophy (Hondorp & Matthews, 2004). The interference of CO-RMs with the methionine pathway might therefore result, at least partially, from the generation of intracellular oxidative stress conditions.
Altogether, these data suggest that CO-RMs trigger an oxidative stress-like response in E. coli cells (Table 3).
CO-RMs generate intracellular oxidative stress
It is well established that haem-containing proteins are preferential targets for CO. Accordingly, CORM-3 was shown to decrease the respiratory rates in E. coli, P. aeruginosa and Campylobacter jejuni due to the binding of CO to terminal oxidases to form carbon-monoxy adducts (Davidge et al., 2009; Desmard et al., 2009; Smith et al., 2011). As expected, in all but one case, impairment of the respiratory chain was reported to be linked to the decrease of cell viability. For reasons that remain unclear, the exception is C. jejuni (Smith et al., 2011).
The blockage of the respiratory chain usually translates into the formation of ROS. Indeed, in eukaryotes, the binding of CO to proteins of the mitochondrial electron transfer chain led to an increase in the intracellular ROS content (Taille et al., 2005; Zuckerbraun et al., 2007). Likewise, cells of E. coli exposed to CO-RMs such as CORM-2 and ALF062 contained higher levels of intracellular ROS (Tavares et al., 2011). The same study revealed that the free iron content originating from the dismantling of Fe-S clusters increases in CORM-treated cells. Further evidence linking the action of CO-RMs to the deleterious formation of intracellular ROS has been presented. In particular, E. coli cells treated with CORM-2 exhibited higher levels of DNA damage and lower DNA-replication ability. Deletion of E. coli recA, a gene involved in double-strand break repair, rendered the strain less viable in the presence of CORM-2 when compared with the parental strain. CORM-2 was also shown to oxidize free thiol groups (Tavares et al., 2011). An E. coli catalase mutant was more sensitive to CORM-2 and the killing of E. coli by CO-RMs was abrogated upon addition of antioxidants, such as reduced glutathione, cysteine and N-acetylcysteine, further confirming that CO-RMs generate an intracellular oxidative stress (Desmard et al., 2011; Tavares et al., 2011). Similarly, the lethal effect on E. coli of the ruthenium-based carbonyl ALF492, which was used as co-adjuvant for treatment of cerebral malaria (Pena et al., 2012), was abolished upon supplementation of cells with reduced glutathione (our unpublished results). More recently, treatment of P. aeruginosa with CORM-2 was shown to increase the production of ROS in biofilms (Murray et al., 2012).
Moreover, release of H2O2 was detected when C. jejuni was exposed to CORM-3 (Smith et al., 2011). Additionally, an EPR (Electron Paramagnetic Resonance) study revealed that CO-RMs are able to produce hydroxyl radicals per se in a CO-dependent mode, as addition of haemoglobin prevented their formation (Seixas, 2010; Tavares et al., 2011). The generation of hydroxyl radicals from CORM-2 is proposed to result from the reduction of oxygen by reduced ruthenium species, which mediate the water–gas shift reaction that is initiated with the attack of water on one of the CO ligands of the RuII(CO)3 moieties of CORM-2. During the water–gas shift reaction, H2 or two electrons are formed (Eqn (1)) (Greenwood & Earnshaw, 1997).
ROS formation is not exclusively linked to the presence of ruthenium in the CO-RM, as ALF062, a Mo-containing CO-RM, also induces the formation of hydroxyl radicals. In this case, it is plausible that hydroxyl radicals originate from the reaction of the electron-rich metal in the [Mo(CO)5Br]-[Net4] complex with water oxygen (Tavares et al., 2011).
Hence, the mechanisms that underlie the killing effect of CO-RMs on bacteria include the production of ROS (Fig. 3).
Evidence that CO-RMs are able to eradicate pathogens suggests a previously unforeseen role of the CO that is endogenously produced by the human body. This might help explain earlier observations that exposure of macrophages to CO increases their ability to engulf bacteria and enhances the rate of bacterial phagocytosis (Otterbein et al., 2005). However, CO gas is less bactericidal for E. coli, S. aureus and P. aeruginosa than the organometallic CO-RMs (Nobre et al., 2007; Desmard et al., 2009). In fact, the currently available data indicate that the metal influences the function of CO-RMs, as ruthenium- and molybdenum-based CO-RMs induce the formation of ROS. The results compiled in this review, including those demonstrating that the ROS generated by CO-RMs contribute to their killing properties demonstrate conclusively that the formation of ROS needs to be considered when using this class of compounds. Hence, and independently of their pharmacological applications, CO-RMs no longer should be seen as simple CO delivery systems. To which point in animal cells the cytoprotective and potent anti-inflammatory properties of CO-RMs are linked to ROS formation is an open question that requires investigation to fully understand the mode of action of this novel class of compounds that exhibit a wide range of therapeutic properties.
This work was supported by Project Grants PEst-OE/EQB/LA0004/2011 and PTDC/BIA-PRO/098224/2008 (LMS) from Fundação para a Ciência e Tecnologia (FCT). A.F.T. and L.S.N. are recipients of FCT grants, SFRH/BD/38457/2007 and SFRH/BPD/69325/2010, respectively.