OUP user menu

Antimicrobial strategies centered around reactive oxygen species – bactericidal antibiotics, photodynamic therapy, and beyond

Fatma Vatansever, Wanessa C.M.A. de Melo, Pinar Avci, Daniela Vecchio, Magesh Sadasivam, Asheesh Gupta, Rakkiyappan Chandran, Mahdi Karimi, Nivaldo A. Parizotto, Rui Yin, George P. Tegos, Michael R. Hamblin
DOI: http://dx.doi.org/10.1111/1574-6976.12026 955-989 First published online: 1 November 2013


Reactive oxygen species (ROS) can attack a diverse range of targets to exert antimicrobial activity, which accounts for their versatility in mediating host defense against a broad range of pathogens. Most ROS are formed by the partial reduction in molecular oxygen. Four major ROS are recognized comprising superoxide (Embedded Image), hydrogen peroxide (H2O2), hydroxyl radical (OH), and singlet oxygen (1O2), but they display very different kinetics and levels of activity. The effects of Embedded Image and H2O2 are less acute than those of •OH and 1O2, because the former are much less reactive and can be detoxified by endogenous antioxidants (both enzymatic and nonenzymatic) that are induced by oxidative stress. In contrast, no enzyme can detoxify OH or 1O2, making them extremely toxic and acutely lethal. The present review will highlight the various methods of ROS formation and their mechanism of action. Antioxidant defenses against ROS in microbial cells and the use of ROS by antimicrobial host defense systems are covered. Antimicrobial approaches primarily utilizing ROS comprise both bactericidal antibiotics and nonpharmacological methods such as photodynamic therapy, titanium dioxide photocatalysis, cold plasma, and medicinal honey. A brief final section covers reactive nitrogen species and related therapeutics, such as acidified nitrite and nitric oxide-releasing nanoparticles.

  • bacteria
  • hydroxyl radicals
  • singlet oxygen
  • photosensitizers
  • hydrogen peroxide
  • titanium dioxide


Since the evolution of oxygen-based life as we know it, all living organisms have had to cope with the possibility of their critical molecular components being damaged by reactive oxygen species (ROS). Low levels of ROS such as superoxide (Embedded Image) and hydrogen peroxide (H2O2) are produced as a consequence of aerobic respiration and metabolism, but sensing mechanisms combined with constitutive and inducible antioxidant defenses [superoxide dismutase (SOD) and catalase] have evolved to largely neutralize these ROS without damage occurring. However, larger concentrations of Embedded Image and H2O2 may overwhelm these defenses, and small amounts of ROS against which organisms do not have adequate defenses such as hydroxyl radical (̇OH) and singlet oxygen (1ΔgO2, commonly abbreviated as 1O2; an abbreviation that will be used throughout our manuscript) may rapidly prove fatal, especially to microorganisms.

Oxidizing agents have long been used as disinfectants and antiseptics, but have not found widespread applications as anti-infectives due to their perceived lack of specificity toward microbial cells when compared with host mammalian cells; in other words, they are considered likely to cause unacceptable damage to normal tissue. In recent times, however, new ways of generating ROS have been discovered that allow therapeutic application to be considered for actual infections, both systemically and topically. The discovery that microbicidal antibiotics produce their killing effect by inducing the formation of hydroxyl radicals within bacterial cells (Kohanski et al., 2007), together with the known use of ROS generation by host defense cells such as neutrophils and macrophages, has spurred new efforts to utilize ROS and oxidative stress as an active anti-infective strategy. Furthermore, nitrosative stress and reactive nitrogen species (RNS) are relative newcomers to the therapeutics arena, but may also have interesting roles to play.

Reactive oxygen species in living organisms

ROS is a collective term used to denote molecules and reactive intermediates with a highly positive redox potential (pE or Eh). The standard reduction potential is defined relative to a standard hydrogen electrode, reference electrode, which is arbitrarily given a potential of 0.00 volts. It is a measure of the tendency of a chemical species to acquire electrons and thereby be reduced (or in other words to act as an oxidizing agent). Reduction potential is measured in volts (V) or millivolts (mV). Each species has its own intrinsic reduction potential; the more positive the potential, the greater the species' affinity for electrons and its strength as an oxidizing agent. Table 1 provides a compiled list of ROS and their known redox potentials for the appropriate half-reactions; the list includes ROS, reduced oxygen molecules, secondary ROS radicals, RNS, high valence transition metals, and other reactive species. The table is organized in the order of decreasing redox potential to give what Buettner called a ‘pecking order’ (Buettner, 1993). In principle, each species can acquire an electron from a species below it in the list, but must donate an electron to each species above it. However, it must be stated that it is not only the electrode potential that governs whether a reaction can proceed or not, the activation energy is also important.

View this table:
Table 1

Standard reduction potentials for biologically relevant molecules and reactive species

Half reactionElectrode potential*
HO + e + H+ → H2O+2.31 V
O3 + 2e + 2H+ → H2O + O2+2.075 V
Embedded Image +1.9 V
Co(III) + e → Co(II)+1.82 V
Embedded Image +1.8 V
H2O2 + 2e + 2H+ → 2H2O+1.76 V
RO• + e + H+ → ROH (alkoxyl)+1.6 V
O• + e + 2H+ → H2O+1.46 V
Embedded Image +1.3 V
Embedded Image +1.27 V
Embedded Image +1.2 V
HOCl + H+ + 2e → H2O + Cl+1.08 V
Fe(III)(1,10-phen)3 + e → Fe(II)(1,10-phen)3+1.06 V
Embedded Image +1.06 V
Embedded Image +1.03 V
ROO• + e + H+ → ROOH (alkylperoxyl)+1.0 V
Embedded Image +0.99 V
Embedded Image +0.91 V
Embedded Image +0.81 V
Embedded Image +0.8 V
Embedded Image +0.64 V
Embedded Image +0.63 V
H2O2 + e + H+ → H2O + HO•+0.32 V
Embedded Image +0.36 V
ONOO• + e → ONOO+0.2 V
Cu(II) + e → Cu(I)+0.16 V
2H+ + 2e → H20.00 V (defined)
Embedded Image −0.18 V
Embedded Image −0.33 V
NAD+ + e + H+ → NADH+−1.58 V
Embedded Image −2.87 V
  • In aqueous solution (pH = 7), unless otherwise stated (g), at 25 °C and 1 M concentration.

In eukaryotic cells, mitochondria are the power house of the cell; they are the major consumers of oxygen and also the major sources of ROS (Droge, 2002). During aerobic energy generation, mitochondria reduce molecular oxygen (O2) to water (oxidative phosphorylation). In principle, the cytochrome c oxidase has to catalyze the transfer of four electrons to an oxygen molecule without any intermediates, but in reality, during this process, partially reduced oxygen species are produced as ‘by products’; thus, leakage of electrons from the electron transport chains (located on the inner membrane of mitochondria) is the source of Embedded Image, which then produces H2O2 by dismutation (Fridovich, 2004). Further reaction to give ˙OH and peroxynitrite (ONOO) can follow (Inoue et al., 2003; Buonocore et al., 2010). Figure 1 shows a schematic diagram illustrating that the generation and interactions of naturally occurring ROS produced from the mitochondria can damage host mammalian cells.

Figure 1

Production of ROS via leakage of Graphic from mitochondrial respiratory chain. Further ROS (H2O2 and •OH) are formed, and defense systems such as catalase and SOD can be induced to mitigate the resulting damage and prevent excessive oxidative stress.

Other potential endogenous sources include cytochrome P450 metabolism, peroxisomes, microsomes, inflammatory cell activation, monooxygenase system, nitric oxide synthase, and several other enzymes that are involved in inflammatory processes (Inoue et al., 2003; Authen & Davis, 2009). ROS are continuously generated through variety of pathways, involving both enzyme-catalyzed and nonenzymatic reactions. For instance, during the respiratory burst process, taking place in activated phagocytes, ROS damaging effects [as byproducts of photosynthesis and byproduct components of several cellular enzymes, including but not limited to flavins in NADPH oxidases NOX1-3 in smooth muscle and vascular endothelium, xanthine oxidase (XO), and uncoupled endothelial nitric oxide synthase], can be observed in the components of the cell membranes (Droge, 2002).

Besides mitochondria, there are other endogenous sources of cellular ROS, such as neutrophils, eosinophils, and macrophages (Conner & Grisham, 1996). Free radicals can also be produced by a host of exogenous processes such as environmental agents and xenobiotics (metal ions, radiation, barbiturates; Klaunig et al., 1997), and bacterial invasion. Oxidative stress and a higher ROS production can be also induced by stress factors such as tumor necrosis factor α (TNF-α), and the increase in ROS production can be recognized by various redox sensors; this recognition triggers a redox cascade, which leads to the activation of both prosurvival and pro-cell-death factors (Pourova et al., 2010). On that basis alone, one would expect these organelles to produce substantial amounts of the partially reduced oxygen products of Embedded Image and H2O2.

Furthermore, the products of reaction of ROS with organic molecules can themselves function as ROS, such as lipid hydroperoxides that can initiate free radical chain reactions. It is well known that ROS readily attack the polyunsaturated fatty acids of the fatty acid membrane, initiating a self-propagating chain reaction (Mylonas & Kouretas, 1999), and because lipid peroxidation is a self-propagating chain reaction, the initial oxidation of only a few lipid molecules can result in significant tissue damage (Mylonas & Kouretas, 1999). This pathway of lipid peroxidation is illustrated in Fig. 2.

Figure 2

Oxidative degradation of lipids. Lipid peroxidation starts with an ROS ‘stealing’ an electron from the lipid cell membrane and the process continuous via free radical chain reaction mechanism. The peroxidation reaction is especially affective in polyunsaturated fatty acids, because they contain multiple double bonds and in between methylene groups (–CH2–). This position of the methylene groups renders the hydrogen atoms especially reactive and susceptible to ROS attack. The degradation process constituting three major steps goes through initiation, propagation, and termination.

When we look closely into the energy generation process in mitochondria, we see stepwise reduction in ROS ‘byproducts’ [through electrons (e) addition]: Embedded Image 1-1 Embedded Image 1-2 Embedded Image 2 Embedded Image 3 Embedded Image 4

During this cascade of reactions, the hydroperoxyl radical (Embedded Image) further dissociates to generate Embedded Image (Eqn. 1-2).

Embedded Image is formed when an electron is captured by one of the π*2p orbitals of oxygen. Its biological effect stems from the fact that depending on the solution environment (i.e. pH and fluctuations), its reduction potential changes. In aqueous solution, it is a weak oxidizing agent and as such able to oxidize ascorbic acid and thiols, as well as act as a very strong reducing agent, and reduce iron complexes of cytochrome c and ferric EDTA. This anion radical transforms further via a well-known dismutation reaction.Embedded Image 5

Superoxide dismutase enzyme with copper–zinc center accelerates this reaction and produces H2O2. There are other enzymes (such as urate oxidase, D-amino acid oxidase, and glucose oxidase) that catalyze the same reaction as well.

By nature, H2O2 is a covalent and uncharged molecule that readily mixes with water, and is treated as such by the body; thus, it diffuses across cell membrane with ease. But at the same time, it has redox properties as well as capabilities to form highly reactive free radicals (in the presence of transition metal ions). As such, it is imperative that this molecule is ‘neutralized’ by the cells. The evolved body defense mechanisms operate via actions of selenium containing glutathione peroxidase, catalase, and certain other peroxidases.

Another major byproduct of high-energy ionization of water molecules is ˙OHEmbedded Image 6

This radical is an extremely aggressive oxidant, and as such, it can attack various biological molecules and impair their proper functioning. The multifaceted Fenton reaction cascade produces oxo-iron, ˙OH, and Embedded Image intermediatesEmbedded Image 7 Embedded Image 8 the overall reaction is as Embedded Image 9

One has to take into account the fact that in living organisms and at physiological pH, the ferrous ion (Fe2+; in the presence of oxygen and phosphate ions) has a transient lifetime, and it autoxidizes to ferric (Fe3+) state. During this process, an e is transferred from iron to the O2 and generates an Embedded Image. The unstable intermediates are perferryl ions (Embedded Image ), where iron is at V oxidation state. Embedded Image 10 Thus, it is clear that in aerobic life, oxygen has a dual effect: it is a crucial source of energy, its reduction through the electron transfer system of mitochondria is a lifeline of survival, and at the same time, the byproducts of the oxygen usage/metabolism generate the damaging molecules of Embedded Image, •OH, and H2O2. These three species together with unstable intermediates from the peroxidation of lipids (and referred to as ROS as well) have major life-effecting impact. Many diseases (such as Alzheimer's, Parkinson's, atherosclerosis, myocardial infarction, auto-immune diseases) are linked to the damaging effects of ROS and develop as a result of an imbalance between radical-generating and radical-scavenging systems, that is, oxidative stress condition. SOD activity of erythrocuprein together with the fact that almost all mammalian cells contain SOD suggests the physiological importance of Embedded Image as a ROS (McCord & Fridovich, 1969).

1O2 is deemed to be one of the most damaging ROS. It is a highly reactive form of oxygen where the electron spin restriction that prevents ground-state triplet oxygen reacting with other singlet molecules is lifted, thereby increasing its oxidizing ability (Halliwell & Gutteridge, 2010). Formation of 1O2 is extremely important ‘natural’ process with enormous biological implications. There are few mechanisms through which living organisms generate the 1O2: (1) in photosynthetic cells, energy is transferred from a light-excited photosensitizer (i.e. a light-sensitive molecule) to ground-state oxygen. This process of light excitation of pigments or chlorophylls/bacteriochlorophylls (in photosynthetic microorganisms) is the most common source of 1O2 (Borland et al., 1988; Cogdell et al., 2000; Krieger-Liszkay, 2005; Telfer, 2005; Krieger-Liszkay et al., 2008; Ziegelhoffer & Donohue, 2009; Tomo et al., 2012); (2) in nonphotosynthetic cells, the main endogenous photosensitizers are flavins, porphyrins, rhodopsins, and quinones (Ryter & Tyrrell, 1998). Because these compounds are innately present in the cells, in events of simultaneous presence of oxygen and light, 1O2 is generated (proteo- and bacteriorhodopsins are considered to be light-driven proton pumps). A wide variety of heterocyclic aromatic compounds can generate 1O2 as well as naphthalene, anthracene, and their derivatives; dyes such as methylene blue (MB), rose bengal; herbicides; and various therapeutic agents; Paul et al., 2004; Sun et al., 2006; Grandbois et al., 2008); and (3) light-independent 1O2 generation can be due to respiratory burst of macrophages (where NADPH oxidases generate Embedded Image that spontaneously or enzymatically dismutates to H2O2, which then is reduced to hypochlorous acid (HOCl) via chloro-, lacto-, or myeloperoxidase catalyzes, and finally, the spontaneous reaction of H2O2 with HClO forms the 1O2; Ryter & Tyrrell, 1998; Tarr & Valenzeno, 2003; Davies, 2004). Other light-independent 1O2 sources are lipid peroxidation processes.

Antioxidant defense against ROS

As mentioned above, the reliance of most life on aerobic metabolism, together with the role of ROS in intracellular signaling, host defense against invading microorganisms, and in regulatory mechanisms, has meant that defense against excessive ROS has needed to evolve to deal with the harmful effects of ROS, such as lipid peroxidation (damaging cell membranes), oxidative damage to proteins (Davies, 2003), mutations to DNA, and activation of pro-cell-death factors.

When ROS produced overwhelm the cellular antioxidant defense system, either by increased ROS generation or by decreased cellular antioxidant capacity, oxidative stress occurs. Thus, oxidative stress is an imbalance occurring due to excessive ROS or overgenerated oxidants exceeding the capability of the cell to mount an effective antioxidant response. It is well known that oxidative stress results in macromolecular damage and is implicated in various diseases including and not limited to atherosclerosis (Paravicini & Touyz, 2006), diabetes (Paravicini & Touyz, 2006), carcinogenesis, and tumor metastasis through gene activation (Ishikawa et al., 2008; Trachootham et al., 2009), neurodegeneration (Andersen, 2004; Shukla et al., 2011), and aging (Haigis & Yankner, 2010). ROS destructive effect is so significant and defense against them such a priority that protective enzymes (such as SOD, catalase, and GSH peroxidase), substrates (such as GSH), and vitamin E, all, tend to be in higher concentration in locations where ROS damage is more likely to occur and be potentially more damaging (Moslen, 1994). Figure 3 shows how a wide range of physiologically relevant primary and secondary ROS and RNS are produced from superoxide as the principle initial species.

Figure 3

Generation and interconversion of physiologically relevant ROS. It is known that ROS sustain homeostasis but may also trigger cell death by apoptosis and/or necrosis (Pourova et al., 2010). Basal levels of ROS production in cells are beneficial for several physiological functions; however, excessive ROS production above basal levels can impair and oxidatively damage DNA, lipids, and proteins and consequently result in dysfunction of these molecules within cells and finally cell death (Winterbourn, 2008; Mohsenzadegan & Mirshafiey, 2012). Semiquinone-like radicals (SQ•) are generated by autoxidation of a range of compounds including adrenaline and DOPA or by enzymatic reduction in quinones such as ubiquinone or menadione. Flavonoids and other polyphenols can generate both semiquinone and phenoxyl radicals. Phenoxyl radicals (PhO•) are produced from tyrosine and other phenolic metabolites and xenobiotics. Aromatic amines and indoles are oxidized to radicals with similar properties. Glutathionyl radical (GS•) are generated from other thiols such as dihydrolipoic acid or cysteine residues. Only myeloperoxidases are capable of generating HOCl.

Complicating matters further, it is known that ROS can also serve in various critical cellular signaling pathways needed for optimum cellular function. There are good data to demonstrate that oxidative stress has effects on signaling pathways. But how about the direct interaction between ROS and elements of the signaling pathways? What happens at the ‘oxidative interface’, how is ROS initiating the signaling cascade(s)? By now, it is known that ROS regulates several signaling pathways via interaction with the signaling molecules affecting areas of cell proliferation, survival, differentiation, and metabolism through ASK1, PI3K, PTP and Shc pathways; anti-inflammatory/anti-oxidant responses via thioredoxin and Ref1-Nrf2 pathways; DNA damage via ATM pathway; and iron homeostasis via iron regulatory proteins pathway. To understand these complicated interactions and resulting functional alterations, the issue of ROS–protein interaction needs to be addressed. The oxidative interface functions mainly via redox regulation of redox-reactive cysteine (cys) residues of the proteins. Oxidation of these residues generates reactive sulfenyl moiety (–S–OH) that then forms disulfide bonds within the nearby cysteines (–S–S–) [it is well known that disulfide bonds can be reduced back to the free thiol moiety under physiological intracellular conditions (Paravicini & Touyz, 2006; Miki & Funato, 2012; Ray et al., 2012)] or undergoes further oxidation to form sulfinyl moiety (–SO–OH) and then via further oxidation sulfonyl moiety (–SOO–OH) acids. The sulfenyl moiety may react with the nitrogen atom in the peptide bond and form a sulfenyl amide bond, which can be reduced to a thiol moiety under physiological conditions). All these reversible oxidative changes/modifications end up changing the structure and because of that the function of the protein (Janssen-Heininger et al., 2008; Winterbourn & Hampton, 2008; Roos & Messens, 2011).

The action of ROS is implicated in wound healing as well (Miki & Funato, 2012); it was shown that upon local injury of the tail fin of zebrafish, H2O2 is produced extensively at the wounded area due to ROS-generated enzyme Duox, and H2O2 diffuses from the wound site and penetrates into neutrophils and oxidizes the Cys466 of Lyn; the oxidatively activated Lyn induces the migration of neutrophils to the wound site (Miki & Funato, 2012).

It is important to know not only how ROS signaling in diseases affects function, but also how ROS initiate/propagate cellular signaling in homeostatic conditions as well (Valko et al., 2007).

Cells have developed elaborate defense mechanisms to detoxify/neutralize the ROS and to preserve life (Droge, 2002). These defense mechanisms can operate at several different (even overlapping) levels within the cells: (1) preventing radical formation; (2) scavenging or quenching the radicals by reaction with antioxidants; (3) destroying the ROS using enzymatic action; (4) repairing the caused oxidative damage; (5) expediting the elimination of damaged molecules; and (6) not repairing the extensively damaged molecules to minimize the introduction of possible mutations.

Antioxidant vitamins and membrane defenses

Among the natural antioxidants that function effectively as ROS quenchers are ascorbic acid, carotenoids and tocopherols (Devasagayam & Kamat, 2002). Considering that the structure of membranes has a hydrophobic domain in the interior of the lipid membrane, which is separate from the aqueous milieu of the intracellular medium, the radicals that are formed in membranes are also lipophilic and are different from those formed in the aqueous cytoplasm. Therefore, the defense molecules (effective antioxidants) are also of different types. The α-tocopherol (commonly known as vitamin E) is a lipid-soluble (thus a poor antioxidant in aqueous milieu) chain-breaking molecule, and β-carotene is a lipid-soluble radical scavenger and 1O2 quencher, etc.

In photosynthetic microorganisms and in chloroplasts, antioxidants provide protection against 1O2 via either quenching the 1O2 directly or quenching the excited chlorophylls/bacteriochlorophylls (Cogdell et al., 2000; Trebst, 2003; Telfer, 2005; Krieger-Liszkay et al., 2008). Nonphotosynthetic microorganisms may also use the carotenoids as quenchers of 1O2 (Di Mascio et al., 1989, 1990), while some oxygenic phototrophs (such as cyanobacteria) can use tocopherol (Krieger-Liszkay & Trebst, 2006) and plastoquinone (Kruk & Trebst, 2008). 1O2 scavengers include the following: amino acids (such as l-histidine, tryptophan; Davies, 2004), polyamines (such as cadaverine, spermidine, spermine, and putrescine; Das & Misra, 2004), thiols (such as glutathione; Devasagayam et al., 1991), thioredoxin (Das & Das, 2000), mycosporine lysine (Suh et al., 2003).

There have been many clinical trials of antioxidant vitamins as supplements for a wide variety of diseases (particularly in the cardiovascular system) that were thought to involve oxidative stress (Jha et al., 1995; Lonn & Yusuf, 1997). By and large, these trials have mostly failed (Sugamura & Keaney, 2011). The reasons for this failure are usually attributed to the aforementioned beneficial role of low levels of ROS in cell signaling and homeostasis (Mishra, 2007). However, another possible reason can be appreciated when it is realized that free radicals react with antioxidant vitamins to give less reactive but longer-lived and more stable free radicals (Mortensen & Skibsted, 1997; El-Agamey et al., 2004).

Intracellular defenses

SOD (with Cu, Zn, or Mn active centers) catalytically removes Embedded Image; glutathione peroxidase (with Se active center) removes H2O2 when it is at low and steady-state concentrations; catalases (with 4 NADPH molecules and active Fe centers) remove H2O2 when it is in high concentrations; formation of cytochrome oxidase with Cu centers (in preventing the formation of H2O2, •OH, and Embedded Image) during which there is no release of oxygen molecules while the reduction in O2 to H2O. Enzymes also are employed to remove the intermediates/byproducts generated during oxygen metabolism speedily, specifically, and with high efficiency. Enzymes such as SOD rapidly trigger dismutation of Embedded Image to H2O2 and O2 with much faster rates than uncatalyzed one. The H2O2 byproduct can be destroyed by the two enzymes: catalase and glutathione peroxidase. During normal oxidation metabolism, these two enzymes are employed to eliminate the toxic intermediates of oxygen reduction inside the cells. The system is so finely tuned that this system at the same time allows a small amount of low molecular mass iron to exist safely, because its presence is imperative for signaling functions and synthesizing DNA and iron-containing proteins.

Extracellular defense systems

Molecules functioning at this level includes glutathione peroxidases and SOD employed as glycosylated extracellular proteins (transferrin to bind ferric ions, lactoferrin to bind ferric ions at lower pH, extracellular SOD (EC-SOD) removing Embedded Image catalytically, extracellular GSHPX (EC-GSHPX) removing H2O2 and H–O–O• catalytically). To provide both optimum function and protection, the system is finely tuned to allow limited survival of some ROS (such as Embedded Image, NO•, and H2O2) in the extracellular fluid milieu to use them as signaling, messenger, or triggering molecules, while the rest is safely removed/neutralized. There is another safety layer in this system, where Embedded Image and H2O2 are not allowed to come in contact with the reactive extracellular copper and iron (copper and iron are kept at poorly reactive or nonreactive states).

Antioxidant defense against ROS in bacteria

Due to their relative simplicity, bacteria have been widely used to study and explain the processes of oxidative stress. Bacteria offer several advantages due to the relatively easy ability to modify and manipulate them (and their external environmental conditions) to study the mechanisms that induce ROS formation and their impact on cells. The absence of intracellular organelles makes possible a more accurate quantization of oxidants generated in the reactions. Researchers have been able to generate knockout mutants of several enzymes that deal with Embedded Image and H2O2 stresses. The manipulation of bacterial growth conditions has allowed identifying the molecules more sensitive to radical species. Moreover, using Escherichia coli (E. coli) which can grow in an anaerobic environment, has been possible to investigate on the effect of oxygen introduction in mutants lacking oxidative defense (Imlay, 2008).

Under aerobic conditions, bacteria are more likely to experience oxidative stress and ROS production (Storz & Imlay, 1999; Imlay, 2003) that damage various cellular sites, including cysteine, methionine protein residues, DNA, and iron–sulfur clusters (Imlay & Linn, 1988; Storz & Imlay, 1999; Imlay, 2003). With bacteria, ROS is a natural byproduct of cellular activity – oxygen utilization and/or exposure. Flavin cofactors, found in the NADH dehydrogenase II of the respiratory chain, during their autoxidation and reaction with molecular oxygen produce H2O2 and Embedded Image (Eqn. 11; Messner & Imlay, 1999), and highly reactive •OH is formed when the H2O2 reacts with free ferrous ion via Fenton reaction (Eqn. 7; Imlay et al., 1988; Glaeser et al., 2011). Bacteria in their use of N2 during the denitrification process produce RNS as byproducts, in this case nitric oxide (Eqn. 12; Zumft, 1997).Embedded Image 11 Embedded Image 12

The gene regulatory response of E. coli to oxidative stresses (due to H2O2 and Embedded Image) is through induction of SOD and catalase. Embedded Image is dismutated to form H2O2 (Eqn. 13; McCord & Fridovich, 1969). In contrast to mammalian cells that have only two SOD isoforms, E. coli uses three types of SOD that have different metal cationic centers MnSOD (sodA), FeSOD (sodB), and CuZnSOD (sodC).Embedded Image 13

Catalases dismutate H2O2 to H2O and O2 (Eqn. 14; McCord & Fridovich, 1969), and E. coli has two catalases to do just that [hydroperoxidase I (HPI) and hydroperoxidase II (HPII)].Embedded Image 14

In E. coli, genes that regulate the SOD and catalases are members of two major oxidative stress regulons, the OxyR and SoxRS, in addition to the ‘general stress regulon’ RpoS, a fact in itself signifying the importance of ROS removal. Escherichia coli has been used as a model system to study the effects of these regulons functioning in a wide range of bacterial phyla, from Proteobacteria to Actinobacteria. Clearly, in bacteria, the oxidative stress response regulation (through specific regulators) is a key survival factor during episodic exposure to exogenous ROS as well as ROS from normal respiration. It is interesting that despite the huge genomic diversity of bacteria and the presence of other oxidative stress regulators (such as PerR, RpoS), these regulators are functionally conserved in wide range of bacterial species from diverse niches. Apparently, even in anaerobic species of Bacteroidetes that lack catalase homologs, OxyR is important for H2O2 tolerance (Diaz et al., 2006; Honma et al., 2009) In bacteria, OxyR is also implicated in protection against heat stress (Christman et al., 1985), singlet oxygen (Kim et al., 2002), lipid peroxidation–mediated cell damage (Yoon et al., 2002), neutrophil-mediated killing (Staudinger et al., 2002), and protection against near-UV damage (Kramer & Ames, 1987).

In bacteria, low molecular weight protein thiols (such as small peptide glutathione that reduces disulfides in conjunction with glutaredoxin is one class of small proteins that are doing just that, promoting the degradation of ROS, RNS, and intermediates to repair oxidative and nitrosative altered proteins.

Enzymes that catalyze the degradation of H2O2 are named ‘catalases’ and are widely spread in different bacterial species. In E. coli, these enzymes include thiol peroxidase, bacterioferritin comigratory protein, glutathione peroxidase, cytochrome c peroxidase, and rubrerythrins. However, their effects in in vivo conditions remain unclear (Mishra & Imlay, 2012). Also, enzymes such as SOD and alkyl hydroperoxidase are known to limit the accumulation of ROS and to provide antioxidant defense for bacterial cells. The catalytic mechanism of the alkyl hydroperoxidase requires two peroxidases: AhpC and AhpD, where AhpD is known to play the role of an essential adaptor protein (Clarke et al., 2011) and AhpC is the catalytic subunit responsible for alkyl peroxide metabolism (Mongkolsuk et al., 2000). On the other hand, AhpC from Mycobacterium tuberculosis (MtAhpC) is known to be a major component in the NADH-dependent peroxidase and peroxynitrite reductase system, where AhpC directly reduces peroxides and peroxynitrite, and at the same time, it is reduced by AhpD and other proteins. It has been shown that overexpression of MtAhpC in isoniazid-resistant strains of M. tuberculosis (that harbors mutations in the catalase/peroxidase katG gene) provides antioxidant protection, and it may even substitute for the lost enzyme activities (Guimaraes et al., 2005).

Interestingly, it has been shown that mutations in E. coli OxyR suppress sensitivity toward H2O2, which in turn cause rise in the levels of one of the three enzymes that are responsible for destroying organic ROS and H2O2, namely catalase–HPI (the katG gene product), catalase–HPII (controlled by katEF), or alkyl hydroperoxide reductase (specified by the Ahp genes). It appears that peroxides serve as mediators of toxicity for a variety of redox agents and as such are produced in sufficient quantities during normal metabolism to generate substantial elevations in the levels of ‘spontaneous’ mutations in cells that lack adequate antioxidant defenses (Greenberg & Demple, 1988).

It is expected that for pathogenic mycobacteria to be able to persist in macrophages and in granulomatous caseous lesions, they must be equipped to withstand the action of toxic metabolites/ROS. In Gram-negative bacteria, the OxyR protein is a critical component of the oxidative stress response where OxyR is not only a sensor for ROS but also a transcriptional activator, inducing expression of detoxifying enzymes such as catalase/hydroperoxidase and alkyl hydroperoxidase (Sherman et al., 1995). Sherman et al. (1995) in their study characterized the responses of various mycobacteria to H2O2, both phenotypically and at the levels of gene and protein expression, and found that only the saprophytic Mycobacterium smegmatis induced a protective oxidative stress response analogous to the OxyR response of Gram-negative bacteria. They also found that under similar conditions, the pathogenic mycobacteria exhibited a limited, nonprotective response, which in the case of M. tuberculosis was restricted to induction of a single protein, KatG. Their conclusion was that the response of pathogenic mycobacteria to oxidative stress differs significantly from the inducible OxyR response of other bacteria (Sherman et al., 1995).

ROS and RNS produce damage to iron–sulfur clusters, proteins, and DNA (Chiang & Schellhorn, 2012). In bacteria, proteins are major biological targets for oxidative damage within cells owing to their high abundance and rapid rates of reaction with ROS/RNS and numerous post-translational, reversible or irreversible modifications (Davies, 2004; Butterfield & Dalle-Donne, 2012). In these cases, oxidation can involve cleavage of the polypeptide chain, modification of amino acid side chains, and conversion of the protein to derivatives that are highly sensitive to proteolytic degradation (Stadtman, 2006). It has been shown that reaction of mainly aromatic and sulfur-containing amino acid residues (such as tryptophan, tyrosine, histidine, cysteine, and methionine), both free and on proteins, due to ROS effect, generates peroxides in high yield, which themselves may decompose into reactive intermediates (protein carbonyl derivatives) and induce damage to other targets, such as molecular oxidation of thiol moiety of cysteine (Butterfield & Dalle-Donne, 2012). Moreover, bacterial proteins can be modified by aldehydes and ketones produced during reactions of ROS with lipids and glycated proteins (Davies, 2003; Stadtman, 2006). This can result in the inactivation of cellular enzymes and the oxidation of other biological targets. Protein cross-linking and aggregation can also be induced by reactive species formed on oxidized proteins (Ray et al., 2012). The senescence-related oxidation targets are enzymes of the Krebs cycle, universal stress protein A, the Hsp70 chaperone DnaK, translation elongation factors, and histone-like proteins (Nystrom, 2002). Based on the identity of the oxidized proteins, it has been concluded that several different cell processes are targets for stasis-induced damage; these functions include peptide chain elongation, protein folding and reconstruction, large-scale DNA organization, gene expression, central carbon catabolism, and general stress protection (Dukan & Nystrom, 1999; Nystrom, 2002).

Oxidative stress responses coordinated by specific regulators ensure bacterial survival during ROS exposure. Recently, several ROS-sensing global regulators in both Gram-positive and Gram-negative pathogenic bacteria have been reported (Chen et al., 2011). Bacteria have several major regulators activated during oxidative stress, including bacterial transcription factors, viz. OxyR, SoxRS, and RpoS. OxyR and SoxRS undergo conformation changes when oxidized in the presence of H2O2 and Embedded Image, respectively, and subsequently control the expression of cognate genes. OxyR serves as a peroxide-sensitive thiol-based redox sensor and controls the expression of several genes involved in the antioxidant response. Among the genes controlled by OxyR are a peroxidase, AhpC, and its reductase, AhpF. AhpC belongs to the family of peroxiredoxin (Prx), which catalyzes the reduction in H2O2 and alkyl hydroperoxides through reversible disulfide formation (Lindahl et al., 2011). OxyR also has a role in protecting against heat stress, near-UV, 1O2, lipid peroxidation–mediated cell damage, and neutrophil-mediated killing. RpoS is a stationary phase and general stress response regulator (Dukan & Nystrom, 1999; Chiang & Schellhorn, 2012). Another example of redox regulation in bacteria is the oxidative activation of the molecular chaperone Hsp33, which contains four conserved cysteines prone to the formation of disulfide bridges (Kumsta & Jakob, 2009). The mechanisms that these redox-sensitive regulatory proteins employ to sense redox signals through covalent modification of redox active amino acid residues or associated metalloprotein centers will suggest further understanding of bacteria pathogenesis, antibiotic resistance, and host–pathogen interaction (Chen et al., 2011).

Genetic responses to ROS and oxidative stress in microorganisms

Microorganisms that regularly face photo-oxidative stress have evolved specific systems to sense 1O2 and tightly control the removal of 1O2 reaction products (Glaeser et al., 2011). A general distinction for the molecular machinery responsible for these responses can be made between photosynthetic and nonphotosynthetic microorganisms.

Cyanobacteria for example are exposed to photo-oxidative stress due to chlorophyll-mediated singlet oxygen production and by the UV-B-induced production of ROS (He & Hader, 2002). The key defense system implicated in response to photo-oxidation involves the five high-light-inducible (Rychlik) proteins from Synechocystis PCC6803. Hli proteins are essential for survival when the cells are absorbing excess excitation energy with a demonstrated crucial role in photoacclimation (He et al., 2001; Havaux et al., 2005). The homologs NblS (Synechococcus elongatus PCC7942) and DspA (Synechocystis PCC6803) control Hli protein expression synthesis and are required in survival in high light. In some cyanobacteria, iron starvation leads to the synthesis of proteins (IdiA, IsiA) and a regulatory system involving IdiB and Fur proteins as well as the isiAB genetic loci, which protect the photosystems against oxidative stress (Michel & Pistorius, 2004; Havaux et al., 2005; Dühring et al., 2006).

A complex network of different regulatory components controls the defense against photo-oxidative stress in anoxygenic photosynthetic bacteria. The photo-oxidation responses were first studied in anoxygenic prototrophs employing Rubrivivax gelatinosus and the characterization of mutants with impaired carotenoid synthesis (Ouchane et al., 1997). Apart from this first pilot study, Rhodobacter sphaeroides has been emerging as the model photosynthetic bacterium to dissect the photo-oxidative responses. Briefly, it involves regulation of genes by the alternative sigma factors RpoE, RpoHII, and RpoHI, by small noncoding RNAs (Berghoff et al., 2009) and the RNA chaperone Hfq (Berghoff et al., 2011). In addition, at least one-two-component system comprising an OmpR-like DNA-binding protein and a yet unknown sensor histidine kinase are involved (Nuss et al., 2010). Proteomic studies revealed altered synthesis rate in about 60 soluble proteins in response to 1O2 (Glaeser et al., 2007) with the majority of the corresponding genes controlled by the alternative sigma factors RpoHII and/or RpoHI, whereas a small number were directly depended on the extracytoplasmic function (ECF) sigma factor RpoE (Nuss et al., 2009, 2010). The comparison of the transcriptomic profiles between the wild type and a mutant lacking ChrR, the anti-sigma factor for RpoE revealed more than 180 genes clustered in 61 operons with at least threefold difference in expression levels between the two strains (Anthony et al., 2005).

RpoE and ChrR homologs exist in the heterotrophic proteobacterium Caulobacter crescentus and contribute in the regulation of photo-oxidative stress in a similar fashion with the anoxygenic prototrophs although some mechanistic elements of the RpoE/ChrR dissociation are still unknown (Lourenço & Gomes, 2009). A network of sigma factors, anti-sigma factors, repressor proteins, and antirepressors control the expression of carotenoid genes in response to light (Elias-Arnanz et al., 2010). The anti-sigma factor CarR together with the ECF sigma factor CarQ controls expression of the carQRS operon and the crtI gene in Myxococcus xanthus. CarS is an antirepressor of CarA and CarH, which repress the car operon in the dark. CarR is associated with the inner membrane and is unstable in illuminated stationary phase cells (Browning et al., 2003). The core of information for the photo-oxidative responses of the nonphotosynthetic bacteria comes mainly from the enteric pathogens E. coli and Salmonella typhimurium. The list includes the following:

  1. The OxyR gene, which function as a redox sensor and which is oxidized at elevated levels of H2O2, binds to DNA target sequences in its oxidized form and subsequently activates a small subset of genes (Storz et al., 1990ab). It was discovered in a screen for Salmonella mutants that were hyper-resistant to H2O2 (Christman et al., 1985). OxyR is the critical regulator in oxidative stress but also an important coordinator in pathogenesis and virulence in a variety of human and plant pathogens. With regard to the human pathogens, in Klebsiella pneumoniae, apart from the upregulation of defense mechanisms against oxidative stress, it enhances biofilm formation, fimbrial expression, and mucosal colonization (Hennequin & Forestier, 2009). In Shigella flexneri, the causative agent of bacterial dysentery mediates the regulation of the iron–sulfur cluster biosynthesis systems Suf and Isc (Runyen-Janecky et al., 2008). In the oral pathogens that encounter constant oxidative stress in the human oral cavity due to exposure to air and ROS from coexisting dental plaque bacteria as well as leukocytes, OxyR has been proven to be a critical regulator for both Tannerella forsythia (Honma et al., 2009) and Porphyromonas gingivalis (Meuric et al., 2008; Wu et al., 2008).

  2. The earlier but comprehensive reports for E. coli hemH or hemG mutants that accumulate noniron porphyrins to about 100 times higher level than the wild type and respond to blue light at an intensity 100-fold lower than that required for the wild type (Yang et al., 1995, 1996).

  3. The SoxRS regulon mobilizes diverse functions to scavenge free radicals and repair oxidative damage in macromolecules. The SoxRS system in particular responds to the Embedded Image stress (González-Flecha & Demple, 2000). SoxR contains a (2Fe–2S) cluster that is oxidized by Embedded Image and subsequently activates transcription of SoxS, an AraC family protein (Ding et al., 1996; Watanabe et al., 2008). SoxS binds to its target promoters and activates genes that encode SOD, DNA repair enzymes, and enzymes of the carbon metabolism (Pomposiello & Demple, 2002) It is likely that the SOD activated by SoxRS has a protective role against 1O2 as it was demonstrated for the iron SOD of Agrobacterium tumefaciens (Saenkham et al., 2008). A combinatorial approach employing DNA microarrays to evaluate the paraquat-induced global transcriptional response of Bacillus anthracis to endogenous superoxide stress and mutants lacking the SOD genes sodA1 and sodA2 was supportive for the protective role of the sod locus (Passalacqua et al., 2007). The exact stimuli and functionality of SoxRS as well as its partition in SOD activation are under investigation (Gu & Imlay, 2011; Fujikawa et al., 2012). The role of SoxRS in microbial virulence and pathogenesis is pivotal, and it has explored extensively in enteric bacteria (van der Straaten et al., 2004; Rychlik & Barrow, 2005; O'Regan et al., 2010). The E. coli-based SoxRS paradigm does not hold in Pseudomonas aeruginosa and the hypothesis fostered for a possible physiological role involves the SoxR-mediated transcriptional regulation of quorum-sensing genes and signaling through the phenazine pyocyanin (Palma et al., 2005; Dietrich et al., 2006). SoxR has been also proposed as the key regulator mediating oxidative stress and virulence capabilities of the prominent plant pathogenic bacteria Erwinia chrysanthemi (Nachin et al., 2001) Xanthomonas campestris pv. campestris (Mahavihakanont et al., 2012).

As previously mentioned, small RNAs (sRNAs) and the RNA chaperone Hfq play a pivotal role in the regulation of the photo-oxidative stress responses in the anoxygenic prototrophic bacterium R. sphaeroides. sRNAs not only interact with mRNAs in photo-oxidative stress, but possess broader functionality as small noncoding post-transcriptional gene regulators. The mRNA targets are influenced by means of stability and/or translation initiation (Repoila & Darfeuille, 2009; Vogel, 2009). As the knowledge of the regulated genes increases, it is possible to define categories of regulatory circuits, which are similar to those described for protein transcription factors (Beisel & Storz, 2010). Most sRNAs function as regulators in stress responses. For example, the OxyS sRNA regulates expression of the general stress sigma factor RpoS in E. coli (Repoila et al., 2003). They are implicated in many phenomena of microbial physiology including pathogenicity and virulence (Pichon & Felden, 2005; Bradley et al., 2011; Kröger et al., 2012; Pichon et al., 2012), catabolite repression, quorum sensing, nitrogen assimilation (De Lay & Gottesman, 2009). Hfq affects sRNA and mRNA stability as well as facilitates sRNA/mRNA interaction (Aiba, 2007). The role of Hfq in oxidative stress regulation has been explored in a variety of pathogens such as Salmonella sp (Bang et al., 2005), Vibrio parahaemolyticus (Su et al., 2010), Francisella novicida (Chambers & Bender, 2011), Francisella tularensis live vaccine strain (Honn et al., 2012).

Host defense systems against microorganisms utilizing ROS

The mammalian innate immune system has evolved sophisticated mechanisms to recognize and kill foreign invaders such as bacteria, fungi, and parasites. This is mediated mainly by the process of phagocytosis, by which macrophages and neutrophils engulf these foreign cells and kill them by a combination of ROS, RNS, and enzymatic digestion (Malle et al., 2007). The relatively sudden production of ROS when phagocytes have engulfed pathogens has become known as the ‘oxidative burst’ (Freitas et al., 2009) or ‘respiratory burst’ (Forman & Torres, 2001). Figure 4 shows how ROS produced in this manner form phagocytes, damage, and inactivate invading microbial pathogens.

Figure 4

Antimicrobial host defense relies on ROS generation by macrophages and neutrophils. The microbial cells are engulfed into a phagosome and this triggers NADPH oxidase generates Graphic, while myeloperoxidase generates HOCl and iNOS generates NO•. These ROS and RNS combine to kill the microorganisms.

NADPH oxidase

The phagocyte NADPH oxidase is the principal source of ROS generation in activated neutrophils and macrophages (Segal et al., 2012). In addition to the phagocyte NADPH oxidase isoform (NOX2), isoforms of NADPH oxidase exist in several cell types and mediate diverse biological functions. The phagocyte NADPH oxidase complex comprises a cytochrome component consisting of gp91phox (phagocyte oxidase) and p22phox embedded in membranes. The cytoplasmic subunits p47phox, p67phox, p40phox, and rac translocate to the membrane-bound cytochrome upon activation of the oxidase. NADPH is oxidized to NADP+, and electrons are transported down a reducing potential gradient that terminates when oxygen accepts an e and is converted to Embedded Image. The NADPH oxidase family is highly conserved being expressed in fungi, plants, and animals (Bedard et al., 2007).

Chronic granulomatous disease (CGD) is an inherited disorder of NADPH oxidase in which phagocytes are defective in generation of ROS. As a result of this defect, patients with CGD suffer from recurrent life-threatening bacterial and fungal infections (Segal et al., 2011). Among patients with CGD, the degree of impairment of NADPH oxidase in neutrophils correlates with clinical disease severity (Kuhns et al., 2010).


Myeloperoxidase (MPO) is an iron-containing heme protein localized in the azurophilic granules of neutrophil granulocytes and in the lysosomes of monocytes. MPO catalyzes the formation of HOCl from H2O2 and chloride ion, which has a strong microbicidal action (Winterbourn et al., 2006). It plays a major role in the killing of several microorganisms including bacteria, fungi, viruses, red cells, and malignant and nonmalignant nucleated cells (Stendahl et al., 1984; Nauseef, 1986; Lanza, 1998; Malle et al., 2003, 2007; van der Veen et al., 2009). Although HOCl dramatically enhances the microbicidal activity of H2O2, MPO appears to be nonessential for host defense (Prokopowicz et al., 2012), as MPO-deficient individuals do not have a high frequency of infections, with the exception of an increased susceptibility to fungal species like Candida (Lehrer & Cline, 1969). Despite the primary role of the oxygen-dependent MPO system in the destruction of certain phagocytosed microorganisms, the lack of an increased frequency of infections may be because other MPO-independent microbicidal pathways compensate for the lack of MPO (Kwakman, et al.).

There is also some evidence that the microbicidal action of MPO may involve •OH (Rosen, 1980) and/or 1O2 (Klebanoff & Rosen, 1978). Furthermore, activated human neutrophils were shown to use MPO for conversion of nitrite into the oxidants nitryl chloride (NO2Cl) and nitrogen dioxide (NO2) that can cause tyrosine nitration and chlorination of target molecules (Eiserich et al., 1998).

Xanthine oxidase

Xanthine oxidase is a major protein component of the milk fat globule membrane (MFGM) surrounding fat droplets in milk, and its enzymology is well characterized. The enzyme is widely distributed in mammalian tissues and is generally accepted to be a key enzyme of purine catabolism. It catalyzes the oxidation of a wide range of substrates and can pass e to O2 to form Embedded Image, similar reduction in nitrite yields RNS. Using the p47(phox−/−) mouse model of CGD, Segal et al. (2000) evaluated the residual antibacterial activity of XO. Clearance of Burkholderia cepacia, a major pathogen in CGD, was reduced in p47(phox/) mice compared with that in wild-type mice and was further inhibited in p47(phox/) mice by pretreatment with the specific XO inhibitor allopurinol. In CGD, XO may contribute to host defense against a subset of reactive oxidant-sensitive pathogens. Harrison (Harrison, 2004) also suggested that XO played a role in antimicrobial defense particularly in the gastrointestinal tract.

Inducible nitric oxide synthase

The discovery of nitric oxide (NO.) led to one of the most highly studied and important biological molecules (Ohshima et al., 2003). NO. plays an important role as a cell signaling molecule, anti-infective agent, and as most recently recognized an antioxidant. The metabolic fate of NO. gives rise to a further series of compounds, which are collectively known as RNS (Moncada et al., 1991; Rubbo et al., 1996; McAndrew et al., 1997). NO. is produced by macrophages as a cytotoxic agent in the immune or inflammatory response (Hibbs et al., 1988; Marletta et al., 1988; Moncada et al., 1991). NO. is also believed to be the key mediator of macrophage-induced cytotoxicity as NO. scavengers block the cytotoxic effect of macrophages (Hibbs et al., 1987; MacMicking et al., 1997). The innate immune systems relies the high-output isoform of inducible nitric oxide synthase [isoform of NOS (iNOS) or NOS2], and this high-output NO pathway has evolved to protect the host from infection (MacMicking et al., 1997). All NOS isoforms convert the amino acid l-arginine and O2 to l-citrulline and NO and require NADPH, FAD, FMN, tetrahydrobiopterin (BH4), and a thiol donor as cosubstrates and cofactors. Mice deficient in iNOS had an increased mortality in the cecal ligation and puncture model of sepsis (Cobb et al., 1999), which was considered surprising as NO has been proposed as the main cause of septic shock.

There is some debate on exactly how NO. exerts its antimicrobial effects (Vallance & Charles, 1998). Because the direct antimicrobial action of pure NO. is limited, it is thought that downstream reactions of NO. RNS such as ONOO, S-nitrosothiols (RSNO), nitrogen dioxide (Embedded Image), dinitrogen trioxide (N2O3), dinitrogen tetroxide (N2O4), and dinitrosyl–iron complexes. ONOO is formed from the rapid interaction of NO• and Embedded Image, from the combination of H2O2 and nitrous acid (HNO2), which would exist in equilibrium with nitrite (Embedded Image ) within an acidified phagolysosomal vacuole (Klebanoff, 1993), or from the interaction of nitroxyl anion (NO) and O2 (Kirsch & de Groot, 2002). S-nitrosothiols such as S-nitrosoglutathione can be formed from NO• and reduced thiols in the presence of an e acceptor (Gow et al., 1997). The potent oxidant Embedded Image can be formed by the autoxidation of NO• or possibly by the oxidation of Embedded Image by MPO and H2O2 (Eiserich et al., 1996).

Bactericidal antibiotics

Antibiotics can be divided into two broad classes: bacteriostatic drugs that inhibit bacterial growth (Pankey & Sabath, 2004) and bactericidal drugs that kill bacteria. Induction of ROS following treatment of bacteria with bactericidal antibiotics has been well documented using different types of antibiotic such as aminoglycosides (Kohanski et al., 2008), quinolines (Goswami et al., 2006; Dwyer et al., 2007), rifampicin (Kolodkin-Gal et al., 2008), and chloramphenicol (Kolodkin-Gal & Engelberg-Kulka, 2008). These antibiotics induce oxidative stress regardless of their specific targets. The general mechanism induced from bactericidal antibiotics involves tricarboxylic acid cycle metabolism that can cause a transient depletion of NADH, destabilization of iron–sulfur clusters, and iron misregulation (Kohanski et al., 2007). In a follow-up study, it was observed that using aminoglycoside antibiotics, the production of •OH was generated by mistranslation and misfolding of membrane associated proteins, producing stress response (Kohanski et al., 2008). Furthermore, it was more recently reported that oxidation of the guanine nucleotide pool to 8-oxo-guanine underlies cell death of bactericidal antibiotics (Foti et al., 2012). In another study, the role of gyrase was investigated and mechanisms were proposed that can contribute to gyrase inhibitor-mediated cell killing. Gyrase is involved in bacterial chromosomal supercoiling of DNA. Gyrase inhibitors are known to induce cell death by stimulating DNA damage formation, impeding lesion repair, and blocking replication processes (Drlica & Zhao, 1997; Couturier et al., 1998). Dwyer et al. (2007) showed that oxidative damage by Embedded Image and •OH contributes to bacterial cell death following gyrase inhibition. Indole is a bacterial signaling molecule that is synthesized from tryptophan via the action of tryptophanase (TnaA). A study (Kuczynska-Wisnik et al., 2010) demonstrated that the treatment of E. coli biofilm with bactericidal antibiotics that generate ROS could inhibit biofilm formation via indole signaling. The increase in the expression of indole is a consequence of overexpression of TnaA. The inhibition or lack TnaA restored the generation of E. coli biofilm. However, although it has been demonstrated that ROS generated from treatment with bactericidal antibiotics play a key role in killing bacterial, several studies have demonstrated that ROS are also directly involved in the induction of mutagenesis and resistance. In a recent study, Kohanski and colleagues showed that sublethal levels of ROS have an important role in mutagenesis inducted by bactericidal antibiotics. The increase in ROS was correlated to mutagenesis, and also, mutagenesis was prevented by inhibiting ROS formation (Kohanski et al., 2008).

Antimicrobial photodynamic inactivation

Photodynamic therapy (PDT) involves the use of photosensitizers (PS) in combination with visible light of the correct wavelength to excite the PS. In the presence of molecular ground (triplet)-state oxygen (3O2), the excited-state PS transfer energy or electrons to producing ROS that are able to kill cells (Dai et al., 2009ab). PDT has mostly been developed as a cancer therapy (Agostinis et al., 2011), but recently has been proposed as an antimicrobial therapy (Hamblin & Hasan, 2004). PDT is an approach that has several favorable features for the treatment of infections caused by microbial pathogens, including a broad spectrum of action, the efficient inactivation of multi-antibiotic-resistant strains, the low mutagenic potential, and the lack of selection of photo-resistant microbial cells (Jori et al., 2006).

The PS molecule has a stable electronic configuration in the singlet state with the most energetic electrons in the highest occupied molecular orbital (HOMO; St. Denis et al., 2011). Following absorption of a photon of light of the specific wavelength according to its absorption spectrum (Fig. 5 shows a Jablonski diagram illustrating the process), an electron in the HOMO is excited to the lowest unoccupied molecular orbital (LUMO), causing the PS to reach the unstable and short-lived excited singlet state. In this state, several processes may rapidly occur such as fluorescence and internal conversion to heat, but the most critical of these to PDT is the reversal of the spin of the excited electron (known as intersystem crossing) to the triplet state of the PS. This excited triplet state is less energetic than the excited singlet state, but has a considerably longer lifetime, as the excited electron, now with a spin parallel to its former paired electron, may not immediately fall back down (as it would then have identical quantum numbers to that of its paired electron, thus violating the Pauli exclusion principle). This much longer lifetime (many microseconds as compared to a few nanoseconds) means the triplet PS can survive long enough to carry out chemical reactions, which would not have been possible with the excited singlet PS. Dyes without a significant triplet yield may be highly absorbent or fluorescent but are not good PS.

Figure 5

Jablonski diagram illustrating the mechanisms of PDT including type I and type II photoreactions. The PS absorb photons from light and causes excitation to the singlet excited state (PS*). The singlet excited PS* can decay back to the ground state with release of energy in the form of fluorescence. It is possible for the singlet to be converted into the long-lived triplet excited state (PS*), which is able to transfer energy to another triplet (ground-state oxygen) or alternatively carry out electron transfer to oxygen producing a range of ROS via superoxide.

The photochemical reactions of the triplet state can be divided into two different pathways, either the type I mechanism involving e or hydrogen atom transfer from one molecule to another or the type II mechanisms involving energy transfer to molecular oxygen. It should be noted that both these mechanisms can occur at the same time, but the relative proportions may depend on the PS structure and also on the microenvironment.

The type I pathway can involve an electron transfer reaction from the PS to O2 in the triplet state, which results in the formation of toxic oxygen species such as Embedded Image that can further transfer to form ROS, such as H2O2 and •OH, which are formed by the Fenton reaction in the presence of divalent metal ions as Fe2+ (Glaeser et al., 2011; Eqns 79). Another possible mechanism has been proposed that may operate in cases where the triplet-state PS are a good e donor. Here, H2O2 (formed from Embedded Image) can undergo a one e reduction to form •OH + OH (the redox potential is only +0.32 V, see Table 1).

The two most prevalent damaging ROS (•OH and 1O2) are able to react with many biomolecules in microbial cells. The exact targets and reaction mechanisms involved depend on the following considerations. First, the localization of the PS generation is critical because most of the ROS are highly reactive and cannot travel far from their site of production before disappearing. Second, the relative abundance of the target biomolecule is important. Davies (2003) calculated the following percentage reactions of 1O2 in leukocytes: protein 68.5%, ascorbate 16.5%, RNA 6.9%, DNA 5.5%, beta-carotene 0.9%, NADH/NADPH 0.6%, tocopherols 0.5%, reduced glutathione 0.4%, lipids 0.2%, cholesterol 0.1%. It should be noted that the distribution of 1O2 may be different in bacterial cells from that found in leukocytes. Third, we have the question of whether type 1 or type 2 mechanism produces the ROS in question. In Table 2, we have tried to point out differences in products that are formed depending on whether •OH and 1O2 are the predominant ROS involved.

View this table:
Table 2

Differences in reaction products between type I reactions (HO˙) and type II reactions (1O2)

TargetType 1 reactionType 2 reaction
Amino acidsTyr (phenol coupling), Phe (hydroxylation), TrpHis (oxygen addition), Trp (oxygen addition) Cys (disulfide), Met (sulfoxide)
ProteinsCross-linking, carbonylationCross-linking, Proteolysis
LipidsRadical chain reactionsLipid hydroperoxides
Nucleic acids (2′-deoxyguanosine)5-carboxamido-5-formamido-2-iminohydantoin8-oxo-7,8-dihydroguanosine
Cholesterol5- & 6-hydroperoxides7-hydroperoxides

The amino acids that are susceptible to oxidation and the products formed are listed in Table 3. Lipid peroxidation starts with a ROS ‘stealing’ an electron from the lipid in the cell membrane, and the process continues via a free radical chain reaction mechanism proceeding through initiation, propagation, and termination steps as shown in Fig. 2. The peroxidation reaction is especially affective in polyunsaturated fatty acids containing nonconjugated double bonds, because the methylene groups (–CH2–) between double bonds are especially susceptible to ROS attack.

View this table:
Table 3

Biological oxidation products of amino acids

Amino acidPhysiological oxidation products
CysteineDisulfides, mixed disulfides (e.g. glutathiolation), HNE-Cys
MethionineMethionine sulfoxide
TyrosineDityrosine, nitrotyrosine, chlorotyrosines, dopa
TryptophanHydroxy- and nitro-tryptophans, kynurenines
Valine, LeucineHydro(pero)xides
Histidine2-Oxohistidine, asparagine, aspartate, HNE-His
GlutamylOxalic acid, pyruvic acid
ProlineHydroxyproline, pyrrolidone, glutamic semialdehyde
Threonine2-Amino-3-ketobutyric acid
ArginineGlutamic semialdehyde, chloramines
Lysinea-Aminoadipic semialdehyde, chloramines, MDA-Lys, HNE-Lys, acrolein-Lys, carboxymethyllysine, pHA-Lys
  • HNE, 5-hydroxynonenal; MDA, malondialdehyde; pHA, p-hydroxyphenylacetaldehyde.

PS are usually organic delocalized aromatic molecules consisting of a central chromophore with auxiliary branches (auxochromes) that are responsible for further electron delocalization of the PS, thus altering the absorption spectra of the PS. Due to extensive electron delocalization, PS tend to be deeply colored. This means that the energy required to excite the e in the HOMO to the LUMO is low compared with less delocalized molecules, and therefore, the absorption bands are in the longer wavelength (red) spectral region and are large, reflecting the high probability of excitation. Acridine orange was the first photodynamic agent used for microbial killing (Von Tappenier, 1900), but since then, an astonishing array of compounds have been tested (Sharma et al., 2011ab). Many of the PS that have been employed for the treatment infections (and indeed of cancer as well) are based on the tetrapyrrole nucleus, with emphasis on the use of porphyrins, chlorins, bacteriochlorins, and phthalocyanines. Broadly speaking, the PS based on the tetrapyrrole architecture tend to largely undergo type 2 photochemical mechanisms, that is, generate 1O2 (Maisch et al., 2007; Regensburger et al., 2010). The one exception to this rule appears to be bacteriochlorins, which have been reported (Vakrat-Haglili et al., 2005; Silva et al., 2010; Dabrowski et al., 2012) to generate •OH and other ROS/RNS generated from Embedded Image (Ashur et al., 2009). It is thought that PS with different molecular frameworks (nontetrapyrrole based) are more likely to undergo type I photochemistry as well as type 2, and examples of these PS include phenothiazinium salts, such as toluidine blue O (TBO; Martin & Logsdon, 1987ab) and MB (Sabbahi et al., 2008), perylenequinones such as hypericin (Lopez-Chicon et al., 2012), functionalized buckminsterfullerenes (e.g. C60; Mroz et al., 2007), halogenated xanthenes such as Rose Bengal (Lambert & Kochevar, 1997), etc.

Various methods have been employed to distinguish between the different types of ROS produced during PDT. Some of these methods, however, frequently attract criticisms based on the perceived lack of selectivity for different ROS (Chaudiere & Ferrari-Iliou, 1999). The most popular method has been the use of various quenchers or scavengers. These are molecules that can be added to the illuminated microbial suspension containing the PS and will then either consume the ROS or deactivate it, thus decreasing the extent of microbial killing. Azide has been used as a selective deactivator of 1O2 by physically removing the energy from the excited state and has been used to reduce microbial killing in several PDT studies (Maisch et al., 2005; Spesia et al., 2010; Tavares et al., 2011; Huang et al., 2012ab) and to explore to what extent type 1 and type 2 mechanisms operate. Interestingly, Huang et al. (2012ab) recently reported that azide anion could paradoxically potentiate the killing of both Gram-positive and Gram-negative bacteria by MB and light via the generation of azidyl radicals. Other quenchers that have been used for 1O2 include histidine (Banks et al., 1985) and beta-carotene (Bohm et al., 2012). The increase in 1O2 lifetime in deuterium oxide giving greater microbial killing has been used as evidence of type 2 photochemistry (Ragas et al., 2010abc). Quenchers that have been used for •OH include mannitol (Rywkin et al., 1992), thiourea (Martin & Logsdon, 1987ab), and dimethyl sulfoxide (Martin & Logsdon, 1987ab). Another popular approach is to use various compounds whose reaction with ROS can be followed by either absorption or fluorescence spectrometry. These techniques have often been used to determine relative 1O2 quantum yields (ΦΔ; Redmond & Gamlin, 1999) by comparison of reaction rates with reference compounds whose ΦΔ is known (RB and MB were often used for this purpose in the past, but now, it is known that these PS are not exclusively type 2). Examples of these compounds are 1,3-diphenylisobenzofuran (Ma et al., 1994), 2-amino-3-hydroxypyridine (Ma et al., 1994), 4-nitrosodimethylaniline (Matthews & Cui, 1990), and various anthracene derivatives (Kuznetsova et al., 2001). In recent years, the use of fluorescent ROS probes has become popular, and 1O2 sensor green (SOSG; Lin et al., 2013) and hydroxyphenyl fluorescein (HPF), specific for •OH, are becomingly popular (Price et al., 2009; Price & Kessel, 2010).

Some methods have exquisite selectivity for different specific ROS and are therefore preferred as gold standards, but do require specialized equipment that may be considered expensive. More laboratories are now using the 1270-nm luminescence emission from 1O2 now that photomultiplier tubes from Hamamatsu that are sensitive to photons at that wavelength are commercially available (Niedre et al., 2002). Perhaps, the most accepted gold standard method is the use of electron paramagnetic resonance (EPR) spectroscopy combined with spin trap probes (spin-trapping; Augusto & Muntz Vaz, 2007). If the correct spin trap is chosen, then products can be generated with hyperfine splitting patterns that can be unequivocally attributed to 1O2, •OH, and Embedded Image, not to mention other possible free radicals that may be involved (Villamena & Zweier, 2004). The azidyl radical referred to previously was identified by EPR spin-trapping using DMPO (Huang et al., 2012ab). Girotti's laboratory (Geiger et al., 1997) developed an HPLC system based upon the formation of different cholesterol hydroperoxides. Cholesterol 5-alpha-, 6-alpha-, and 6-beta-OOH were characterized as 1O2 adducts, while 7-alpha- and 7-beta-OOH were products of free radical oxidation.

Various studies have shown a fundamental difference in susceptibility to PDT between different classes of microorganisms (fungi, Gram-negative bacteria, Gram-positive bacteria and protozoa). This difference is explained by their physiology and morphology. Jori et al., (2006) distinguished two broad groups of microbial cells that interact differently with PS group 1 involves Gram-positive bacteria and protozoa in the trophozoic stage, and a direct translocation of the PS through the relatively permeable outer cell wall to the plasma membrane occurs. Group 2 consists of Gram-negative bacteria, fungi, and protozoa in the cystic stage, and an initial increase in the permeability of the outer cell wall is required to allow the PS to penetrate to sensitive cellular locations such as the plasma membrane. Figure 6 shows how the structures of the cell wall and their associated permeability barriers vary between Gram-negative bacteria, Gram-positive bacteria and fungi. Although the first reports used chemical methods of increasing this Gram-negative outer membrane permeability such as treating bacteria with polymyxin B nonapeptide (Nitzan et al., 1992) with Tris–EDTA to remove lipopolysaccharide (Coratza & Molina, 1978), with high calcium to produce competence (Bertoloni et al., 1984), or with toluene (Villamena & Zweier, 2004), it has now become apparent that the most simple and effective method of achieving effectiveness against these classes of microorganism is to choose or synthesize PS with pronounced cationic charges (Costa et al., 2012). This desirable goal can be achieved in two different ways. First, PS used for other diseases such as cancer and bearing neutral or anionic charges can be chemically conjugated to carriers that have cationic charges at physiological pH such as poly-l-lysine (Hamblin et al., 2002) or polyethylenimine (Tegos et al., 2006) or encapsulated in liposomes with cationic charges on the outside (Bombelli et al., 2008). The second approach is simply to have constitutive cationic charges on the PS molecule itself using quaternary ammonium or phosphonium groups (Sharma et al., 2011ab). Furthermore, the class of dyes typified by phenothiazinium salts (MB, TBO, etc.) have constitutive cationic charges as part of their chemical structure (Wainwright et al., 1997). Although there is general agreement that cationic charges on the PS are important for broad spectrum antimicrobial activity, it may be the case that different amounts of cationic charges can be optimal for different classes of microbial cell. For instance when studying a set of four bacteriochlorins, Huang et al. (2010) found that six quaternary groups were optimal for Gram-negative bacteria, while two quaternary groups were best for Gram-positive bacteria, and two basic groups (no constitutive cationic charges) were best for Candida fungal cells.

Figure 6

Structures of the cell walls of different classes of microbial cells. (a) Gram-positive bacteria have a relatively porous outer cell wall composed of peptidoglycan, teichuronic acids and lipoteichoic acids. (b) Gram-negative bacteria have a thin layer of peptidoglycan and then a second lipid bilayer incorporating lipopolysaccharide and providing a permability barrier. (c) Yeast have a relatively impermeable cell wall composed of beta-gucan and chitin.

It is quite clear that although microbial cells that have formed biofilms can still be killed by PDT, they are significantly more resistant than planktonic cells (Mantareva et al., 2011). Studies that have dissociated bacterial biofilms into planktonic cell suspensions before subjecting them to PDT and comparing the results to those obtained with logarithmic phase planktonic cells have shown that the resistance of biofilms mainly consists of difficulties faced by PS in penetrating through the exopolysaccharide matrix, rather than intrinsic changes in gene expression found in biofilm cells affecting susceptibility to ROS. Furthermore, it may be the case that the highly effective polycationic PS find it more difficult to penetrate the anionic biofilm matrix because they bind too strongly to it compared with less effective PS with fewer cationic charges.

The use of cationic PS also has a second highly beneficial property besides the property of broad spectrum activity (because they are taken up by all classes of microbial cells). This second property allows selectivity for microbial cells over host mammalian cells that will be present in an infected lesion (Mantareva et al., 2011). This is because the binding and uptake of cationic compounds by microbial cells are rapid while the uptake of cationic compounds by mammalian cells is a slow process. Cationic compounds are taken up into mammalian cells by the slow process of absorptive endocytosis as contrasted with the rapid process of diffusion through the plasma membrane that occurs with neutral lipophilic compounds. Therefore, if the drug–light interval is short (few minutes) after the cationic PS have been topically applied to an infected lesion, microbial cells will be selectively killed, while if the drug-light interval is long (hours), the possibility of harming surrounding host tissue is much higher. Furthermore, it is generally accepted that the appropriate manner to deliver antimicrobial PS to infected lesions is by topical or local delivery into the infected lesion rather than the intravenous, oral, or intraperitoneal systemic delivery routes that are mostly used for cancer (both in animal models and in patients).

This selectivity of cationic PS for microbial cells has been demonstrated many times both in vitro and in vivo. Our laboratory has devised mouse models of localized infections caused by bacteria and fungi that have been genetically engineered to emit bioluminescence allowing monitoring of the infection by real-time noninvasive optical imaging (Demidova et al., 2005). These models have allowed us to demonstrate the ability of PDT to effectively treat wound, burn, and soft tissue infections using a variety of PS including pL-ce6 (Gad et al., 2004), PEI-ce6 (Dai et al., 2009ab), cationic fullerenes (Lu et al., 2010), phenothiazinium salts (Ragas et al., 2010abc), and a cationic porphycene (Ragas et al., 2010abc).

Several clinical applications of PDT for localized infections have emerged or are in the process of being tested (Kharkwal et al., 2011). The most successful is ‘Periowave’ for periodontitis (Berakdar et al., 2012) in which MB and red light are introduced into the dental pocket (Atieh, 2010). MRSAaid is a PDT nasal decontamination system (http://mrsaid.com/), and applications of PDT for chronic sinusitis (Biel et al.,2011ab), endotracheal tube disinfection (Biel et al.,2011ab), and infected leg ulcers (Clayton & Harrison, 2007) are also under investigation.

Photocatalytic disinfection

The ability of titanium dioxide (TiO2) to act as a photocatalyst has been known for 90 years (Hashimoto et al., 2005). In 1972, Fujishima & Honda first reported the photoelectrolysis of water at a semiconductor electrode, and this property was then utilized to catalyze the oxidation of pollutants (Carey et al., 1976; Frank & Bard, 1977). Photocatalytic surfaces can be manufactured into construction and building materials (Chen & Poon, 2009), and commercial uses include self-cleaning windows and self-cleaning glass covers for road lights (Hashimoto et al., 2005).

The mechanism of photocatalysis relies upon the fact that TiO2 is a semiconductor. The adsorption of a photon with sufficient energy promotes an electron from the valence band to the conduction band leaving a positively charged hole in the valence band. The band gap energy (energy required to promote an electron) of anatase TiO2 is c. 3.2 eV, which means that photocatalysis can be activated by photons with a wavelength shorter than 385 nm (i.e. UVA). The hole may be filled by migration of an electron from an adjacent molecule, leaving that molecule with a hole, and so on. A recent paper (Schrauben et al., 2012) suggested that the mechanism would be better characterized as ‘proton-coupled electron transfer’. Electrons and holes may recombine and neutralize each other (bulk recombination), or when the electrons reach the surface, they can react O2 to give Embedded Image. As discussed previously, Embedded Image can react further to form H2O2 and •OH. It is possible to apply an electric field to enhance charge separation of the e and holes and reduce the likelihood of bulk recombination termed ‘photoelectrocatalysis’ (Harper et al., 2000; Zhang et al., 2008).

There are three main polymorphs of TiO2: anatase, rutile, and brookite. The majority of studies show that anatase was the most effective photocatalyst and that rutile was less active; the differences are probably due to differences in the extent of recombination of e and hole between the two forms (Lilja et al., 2012). However, studies have shown that mixtures of anatase and rutile were more effective photocatalysts than 100% anatase and were more efficient for inactivating viruses (Lilja et al., 2012).

The utility of photocatalytic disinfection is limited by the requirement for UVA light for activation. Catalysts can be modified by doping with metals such as Sn, Pd, and Cu to reduce the band gap and extend the useful spectrum into the visible region (Rehman et al., 2009; Lilja et al., 2012). Zinc oxide can also be used as a photocatalyst (Giraldi et al., 2011), but even though it has a higher band gap (3.37 eV) than TiO2, it was found to perform better in sunlight (Sakthivel et al., 2003). Zhang et al. (2010) prepared mixed nanoparticles containing both TiO2 and ZnO.

Because the efficiency of generation of ROS by photocatalysis depends on the ratio of surface area to mass, it is not surprising that a wide range on nanoparticles have been tested for this application (Di Paola et al., 2012). These have included nanotubes (Roy et al., 2011; Xu et al., 2012), nanorods (Ma et al., 2011), nanowires (Wang et al., 2011), nanopyramids (Li et al., 2011), and core–shell nanocomposites (Zhang et al., 2012), and Matsunaga et al. (1985, 1988) were the first to use TiO2 photocatalysis to kill microorganisms. This subject area has recently been comprehensively reviewed (McCullagh et al., 2007; Foster et al., 2011), and the effect of key variables on the effectiveness has been studied (Cushnie et al., 2009). The applications that have been proposed include sterilization and decontamination of wastewater (Praveena & Swaminathan, 2003), manufacture of antibacterial fabrics (Kangwansupamonkon et al., 2009), self-sterilizing urinary catheters (Sekiguchi et al., 2007), self-sterilizing lancet for blood glucose determination (Nakamura et al., 2007), antibacterial food packaging film (Chawengkijwanich & Hayata, 2008), antibacterial dental implants (Suketa et al., 2005), antibacterial surgical implants (Lilja et al., 2012) and many more.

Cold atmospheric plasma

Plasma is considered the fourth state of matter and in fact is the commonest form of matter in the universe by both mass and volume. It forms the material that stars are composed of and (at very low density) fills the interstellar space. Plasma contains ionized atoms and e and responds to magnetic fields and conducts electricity. The degree of ionization reflects its temperature; plasmas with temperatures of millions of degrees are completely ionized, while cold or nonthermal plasmas may only have 1% of the gas molecules ionized. Cold plasma is generated by a pulsed high-voltage electric discharge between electrodes between which a gas is flowing (often a mixture of helium and oxygen). The primary chemical process is ionization of oxygen molecules, which can then react with nitrogen in the air to produce a cocktail of reactive species including 1O2, •OH, Embedded Image, O3, NO•, Embedded Image, etc.; charged particles (e, ions); and UV photons (4 peaks between 300 and 400 nm derived from emission from excited nitrogen molecules; Maisch et al., 2012). All of these species are well known to have antimicrobial activity (Laroussi, 2005). Wu et al. (2012) used electron spin resonance to demonstrate the presence of 1O2 and •OH in cold plasma and, by using quenchers, suggested that 1O2 contributed most to the killing. There have been efforts to introduce the new concept of ‘plasma medicine’, and international conferences have been held on this topic (http://www3.interscience.wiley.com/journal/121376857/issue). A new journal entitled ‘Plasma Medicine’ has also been launched in 2011.

Several devices have been described; one of which was termed the ‘plasma pencil’ (Laroussi & Lu, 2005; Peplow, 2005). The device (illustrated in Fig. 7) emits a cold plasma plume several centimeters in length. The plasma plume can be touched by bare hands and can be directed manually by a user to come in contact with delicate objects and materials including living tissue without causing any heating or painful sensation. It was reported to be highly effective at killing bacteria (Laroussi et al., 2006). Another device termed the ‘FlatPlaSter’ used a 6.75-kHz pulsed voltage with a peak to peak of c. 7 kV to excite ambient air (Maisch et al., 2012). Other devices use radiofrequency voltages (for instance 13.56 MHz) to generate cold plasma in N2 or N2/O2 mixtures (Sureshkumar et al., 2010). Cold argon plasma is widely used for coagulation in endoscopic GI surgery (Manner, 2008), but can also be used as an antimicrobial therapy (Fricke et al., 2012).

Figure 7

Plasma pencil. The 5-cm-long plasma plume is generated when a stream of helium gas containing a trace of oxygen passes between two high-voltage copper electrodes. Helium is very difficult to ionize, but the plume's oxygen molecules break into two highly reactive oxygen atoms, which then attack the bacteria. The key to keeping the plasma pencil cool is its kilovolt electric field, which switches on and off thousands of times a second.

One of the primary applications of cold plasma is to kill microorganisms for sterilization of foods (Niemira, 2012) or for dermatological (Heinlin et al., 2011) and dental (Yang et al., 2011) uses. This microbial killing occurs via several mechanisms: (1) permeablization of the cell wall/membrane, leading to leakage of potassium; (2) intracellular damage due to oxidative or nitrosative damage to critical proteins; and (3) direct chemical damage to DNA. There have been several studies directed toward using cold plasma as a specific approach for inactivating microbial biofilms (Joaquin et al., 2009). Biofilm inactivation has been shown with P. aeruginosa (Alkawareek et al., 2012), Enterococcus faecalis (Jiang et al., 2012), and Candida albicans (Sun et al., 2012). Cold plasma has been used to inactivate bacterial endospores from Bacillus and Clostridium species (Tseng et al., 2012).

One of the advantages claimed for cold plasma as an antimicrobial treatment is its nontoxicity toward host tissue. For instance, Maisch et al. (2012) demonstrated a reduction of 5 log10 in CFU of Staphylococcus aureus (both methicillin-sensitive and methicillin-resistant) and E. coli on ex vivo pig skin without morphological changes or histological signs of necrosis or apoptosis as determined by the TUNEL assay.

Isbary et al. (2010) conducted a prospective randomized controlled trial to decrease bacterial load using cold atmospheric argon plasma on chronic wounds in 36 patients. Patients received 5 min daily applications of cold atmospheric argon plasma in addition to standard wound care. Analysis of 291 treatments in 38 wounds found a highly significant (34%, P < 10−6) reduction in bacterial load in treated wounds, regardless of the species of bacteria. No side effects occurred, and the treatment was well tolerated.

Medicinal honey

Honey has been used in treatment of wounds since ancient times (Zumla & Lulat, 1989; Kwakman et al., 2011). However, with the emergence of antibiotics, the clinical application of honey was abandoned in modern Western medicine. In recent times, the development of antibiotic resistance has increased interest in alternative antimicrobial agents, and honey is regaining its popularity again. In several studies, honey has been shown to be bacteriocidal or bacteriostatic against organisms such as Bacillus subtilis, S. aureus, methicillin-resistant S. aureus (MRSA), Staphylococcus epidermidis, extended-spectrum β-lactamase producing E. coli, Enterococcus faecium, Enterobacter cloacae, Klebsiella oxytoca, ciprofloxacin-resistant P. aeruginosa, and vancomycin-resistant E. faecium (VRE), C. albicans, Acinetobacter baumannii, and Leishmania parasites (Zeina et al., 1997; Kwakman et al., 2008, 2010; Tan et al., 2009; Alzahrani et al., 2012). Manuka and Revamil® are the two major types of medical-grade honeys currently used for clinical application. The high sugar concentration, action of glucose peroxidase to produce H2O2 and •OH, low pH, lysozyme, antibacterial aromatic acids, trans-10-hydroxy-decene-(2)-acid-(1) (10-HAD, the main royal jelly acid), methylglyoxal (MGO), and the antimicrobial peptide bee defensin-1 have been identified as important antibacterial components in honey (Russel et al., 1988; Isidorov et al., 2011; Chen et al., 2012; Kwakman & Zaat, 2012).

Different microbial species have varying susceptibility to different types of honey because H2O2, MGO, and bee defensin-1 differentially contribute to each type of honey and in turn to their antimicrobial activity (Kwakman et al., 2010, 2011). For instance, it has been shown that E. coli and P. aeruginosa are substantially less susceptible to manuka honey than S. aureus and B. subtilis (Kwakman et al., 2010). Bee defensin-1 and H2O2 were shown to be the major factors involved in rapid bactericidal activity of Revamil source honey; however, these factors were absent in manuka honey, and instead, manuka honey contained 44-fold higher concentrations of MGO than Revamil source honey (Kwakman et al., 2010).

Brudzynski et al. (2011) reported that the damaging effects of honey H2O2 were strongly influenced by the microbial sensitivity, defense mechanisms to oxidative stress, the growth phase (exponential vs. stationary), and survival strategy (non-spore forming vs. spore forming). Glucose oxidase, a carbohydrate-metabolizing enzyme secreted from the hypopharyngeal glands of honey bees and added to nectar, converts glucose into H2O2 and gluconic acid under aerobic conditions (White & Subers, 1963; Bang et al., 2003). The function of H2O2 is assumed to be prevention of spoilage of unripe honey when the sugar concentration has not yet reached sufficient levels to prevent microbial growth (Kwakman & Zaat, 2012). During the ripening of honey, glucose oxidase is inactivated, due to lack of free water; however, its activity is restored when honey is diluted with water, providing a slow, sustained release of H2O2 (Chen et al., 2012; Kwakman & Zaat, 2012). H2O2 accumulation is highest in the range of 30–50% honey in water, and below 30%, it starts declining rapidly (Schepartz & Subers, 1964). Furthermore, accumulation of H2O2 varies among different honeys (Kwakman et al., 2011; Chen et al., 2012; Kwakman & Zaat, 2012). Even though a functional relationship between H2O2 produced in honey and antibacterial activity strongly suggests that H2O2 is the main contributor to this activity (White et al., 1963; Bang et al., 2003; Brudzynski, 2006; Chen et al., 2012), recent new findings question this view. One reason is that molecular H2O2 is a relatively weak oxidant and it requires high concentrations to exert its cytotoxic effect (Finnegan et al., 2010); however, as shown by Brudzynski et al., the average H2O2 content in honey is more than 900-fold lower than that observed in H2O2 disinfectant, and yet even at these low concentrations, honey effectively inhibited bacterial growth and caused DNA strand breaks (Molan, 1992; Brudzynski et al., 2011). Second, the oxidative stress of endogenous H2O2 was clearly augmented by the action of unknown honey components (Brudzynski et al., 2011). Third, even though honey samples with little or no H2O2 have a correspondingly low ability to inhibit bacteria and fungi, when present, the level of H2O2 and the degree to which the honey is antimicrobial do not necessarily correlate (Chen et al., 2012), and lastly, Brudzynski et al. (2012) demonstrated for the first time that honeys of high bacteriostatic activity possessed significantly higher levels of phenolic compounds with radical-scavenging activities than honey with average bacteriostatic activity, suggesting that a coupling chemistry between H2O2 and polyphenols, rather than H2O2 alone, may exert the oxidative effect causing bacterial growth arrest and DNA degradation.

Although high oxygen radical quenching capacity of polyphenols is usually indicative of their efficient antioxidant abilities (Cao et al., 1997; Price et al., 2006), the same polyphenols have the potential to turn into powerful pro-oxidants when oxidized. Oxidized polyphenols further generate H2O2, and in the presence of transition metals such as Cu(I) or Fe(II), they drive the generation of OH from H2O2 via the Fenton reaction (Puppo, 1992; Hanasaki et al., 1994; Cao et al., 1997; Brudzynski & Lannigan, 2012). Honey possesses all necessary substrates for the Fenton reaction: H2O2, polyphenols as well as transition metal ions (Bogdanov et al., 2007).

When two bacterial strains of MRSA and VRE were incubated with dilutions of buckwheat honey, there was a steady accumulation of •OH up to the 16-fold dilution, and the peak of •OH generation corresponded to the MIC of honey (Brudzynski & Lannigan, 2012). However, further honey dilutions resulted in a decrease in •OH levels and loss of growth inhibition (Brudzynski & Lannigan, 2012). These findings support a causal relationship between •OH generation and bacterial growth inhibition (Brudzynski & Lannigan, 2012) and suggest that the oxidative stress caused by honey action on bacterial cells may result from •OH rather than from the action of molecular H2O2 itself, and the redox capacity of honey polyphenols plays a role (Brudzynski et al., 2012). However, in a subsequent study, •OH radicals could not be identified as inducers of bacterial cell death (Brudzynski et al., 2011), which suggests that other free radicals generated from the initial Fenton reaction may be effector molecules as well (Brudzynski et al., 2012). Lastly, it is important to note that the •OH-based mechanism of honey action is equally effective against antibiotic-sensitive and antibiotic-resistant bacteria (Brudzynski et al., 2011; Brudzynski & Lannigan, 2012).

A synthetic honey containing glucose and glucose oxidase has been developed with antimicrobial activity. It was based on a technology that allows for the enzyme-catalyzed synthesis of H2O2 to be in a stable sustained release format. This product has demonstrated better antibacterial efficacy than manuka honey in vitro. The organisms that have been shown to be susceptible include MRSA, Campylobacter spp., Streptococcus pyogenes, P. aeruginosa, S. aureus, E. coli, Bacillus cereus, and Propionibacterium acnes. The technology has also demonstrated significant antifungal activity against C. albicans, Saccharomyces cerevisiae, Botrytis cinerea, and a number of unidentified mold isolates.

In terms of clinical applications, best evidence regarding the efficacy of honey has been obtained for treatment of mild-to-moderate superficial and partial thickness burn wounds, where honey was found to be more effective than conventional treatment for reduction in microbial colonization and improved healing (Moore et al., 2001; Jull et al., 2008). Johnson et al. (2005) also demonstrated that application of honey resulted in bacteremia-free period of hemodialysis catheters compared with that obtained with mupirocin treatment. Surgical wounds (Chirife et al., 1982; Greenwood, 1993), decubitus ulcers (Van der Weyden, 2003), medium for storing skin grafts (Subrahmanyam, 1993) are among its other potential applications.

Miscellaneous ROS

There are a set of miscellaneous techniques that exert an antimicrobial effect (among other activities) mediated by ROS or oxygen and its derivatives. These are sometimes referred to as ‘oxygen therapies’ (Altman, 2007), but use of this term carries a connotation of alternative and unproven approaches (Cassileth, 2009).

Hyperbaric oxygen

Hyperbaric oxygen therapy (HBOT) involves exposure to and inhalation of 100% oxygen at supra-atmospheric ambient pressure (2–3 atmospheres). HBOT has been used for many different medical applications throughout the years including infections (Cimsit et al., 2009). HBOT is well established to be effective when used as either a primary or adjunctive treatment in the management of infections such as gas gangrene (Slack et al., 1969), necrotizing fasciitis (Wilkinson & Doolette, 2004), diabetic foot infections (Chen et al., 2010), refractory osteomyelitis (Andel et al., 1998), neurosurgical infections (Larsson et al., 2008), and fungal infections (Bitterman, 2007). HBOT has been proposed to act as a bactericidal/bacteriostatic agent against both aerobic, and particularly anaerobic, bacteria by increasing the formation of free oxygen radicals (Cimsit et al., 2009). HBOT also restores the bacterial killing capacity of leukocytes in hypoxic wounds by increasing tissue oxygen tensions (Clark & Moon, 1999). At present, it is uncertain whether the direct ROS-mediated killing of microbial cells or the potentiation of antimicrobial host defense by HBOT is most important for its success in lowering the infectious burden in the tissue.

Ozone and polyatomic oxygen

It is well known that ozone is poisonous, that it is an important factor in air pollution particularly affecting children, and that inhalation can damage the lungs with possible serious consequences (Mathieu-Nolf, 2002). However, (in common with many other therapies that induce ROS) at very low doses (Bocci, 2004), the outcome of treatment with ozone can be beneficial rather than damaging (Valacchi et al., 2005; Bocci et al., 2009). Therapeutic applications of ozone to prevent or control infection in animal models have comprised repetitive pneumoperitoneum with ozonized oxygen to reduce the severity of sepsis developing after cecal ligation and puncture in rats (Schulz et al., 2003; Silva et al., 2009). Treatment with perfluorocarbons loaded with ozonized oxygen was applied to infected wounds in rats (Dmitrieva et al., 2009). Intraperitoneal ozone administration reduced parasite burden in a mouse model of Schistosomiasis mansoni infection (Thabet et al., 2007).

Clinical applications have included intra-operative delivery during surgery for total hip plasty to prevent infection (Bialoszewski, 2003) and the combination of ozone with ultrasound in the treatment of purulent wounds (Lipatov et al., 2002).

There was an alternative medicine treatment approach called ‘polyatomic oxygen and polyatomic apheresis’ that was championed by Basil Earle Wainwright and which involved treatment of blood with oxygen containing both O3 and O4 (Viebahn-Haensler, 1999). This approach was largely discredited after Wainwright was prosecuted for claiming his system could cure AIDS (Farley, 1994).

Hydrogen peroxide

Hydrogen peroxide (3% v/v) has long been used as a topical antiseptic to be applied to wounds and as a disinfectant for various applications. In many dental applications such as endodontics (Poggio et al., 2012) and periodontitis (Wennstrom et al., 1987), H2O2 is employed as a topical antimicrobial. In endodontics, H2O2 can be combined with sodium hypochlorite as an irrigant for the infected root canal (Grossman, 1943); this reaction is known to produce 1O2 (Tsukagoshi et al., 2006). A Japanese group reported a disinfection system that relied upon the photolysis of H2O2 to form •OH (as shown by electron spin resonance spectroscopy spin-trapping), which was catalyzed by violet light (405 nm) either from a laser (Ikai et al., 2010) or from a LED array (Hayashi et al., 2012). The system was designed to be used in dental infections and was tested in the oral mucosa and wounded skin of rats with no evidence of injury and with some evidence of improved wound healing (Yamada et al., 2012). Another paper from this group showed a synergistic effect between photolysis of H2O2 and elevated temperature (up to 55 °C; Shirato et al., 2012). Again, it should be mentioned that there is a substantial movement in alternative medicine to use systemic H2O2 therapy (30% food grade) for treating cancer, AIDS, and infections (Douglass, 2003). Although conventional medical opinion remains that H2O2 is poisonous and can kill by causing oxygen gas embolisms (Watt et al., 2004).

Superoxidized water and atomic oxygen radical anion

Superoxidized water is a solution of sodium chloride, which has been electrolyzed by passage over titanium electrodes at 9 amps. The product that is produced has a pH of 5.0–6.5 and an oxidation potential of > 950 mV and is mainly HClO at a concentration of c. 144 mg L−1. Depending on the manufacturer, superoxidized water can be called ‘Sterilox’ (Selkon et al., 1999), ‘Microcyn’ (Landa-Solis et al., 2005), ‘Optident’ (Martin & Gallagher, 2005), or ‘Oxum’ (Abhyankar et al., 2009). These solutions have been used in wound sterilization where Oxum was found to be superior to povidone iodine (Kapur & Marwaha, 2011) for care of the diabetic foot (Peters et al., 2012). One study (Wang et al., 2007) reported that ‘atomic oxygen radical anion’ or O• was present in superoxidized water and killed E. coli. This chemistry of this species (O•) has been comprehensively reviewed (Lee & Grabowski, 1992), and it was also reported (Lee & Grabowski, 1992) to be released from the surface of a crystal composed of 12CaO : 7Al2O3.

RNS-based approaches

Nitric oxide (NO•) serves multiple physiological functions in the cardiovascular, respiratory, gastrointestinal, genitourinary, and central and peripheral nervous systems. However, synthesis of NO• also contributes to host defense and has cytostatic and cytotoxic effects against certain pathogens. NO• is a naturally occurring short-lived free radical that itself is not a ROS. The sphere of influence of NO• is around 100 μm from its origin, and it is likely that it only affects a few cells (Miller & Megson, 2007). The dual action effect so characteristic of ROS/RNS is observed in NO• as well: in nanomolar concentrations (generated by constitutive NO synthase (NOS) isoforms), NO• acts as a cell signaling molecule and interacts preferentially with its physiological target enzymes – the most significant of which seem to be soluble guanylyl cyclase and possibly cytochrome C oxidase, whereas at higher concentrations (micromolar and higher), when the cytokine-induced iNOS is expressed in cells, NO• is cytotoxic. Thus, it is necessary to find practical delivery systems for NO• to exert its therapeutic antimicrobial effect in infections.

Acidified nitrite

NO• apparently plays an important role in adult stomachs as well. It is found that human breast milk contains high concentrations of nitrate and nitrite in the early postpartum period. By acidification, the nitrite leads to the generation of NO• in the gastric lumen. Breast milk thus regulates the mucosal blood flow and gastric motility and achieves bacteriostasis via induction of NO• generation in the neonatal stomach (Iizuka et al., 1999). Xu et al. studied the bactericidal effects of NO• on E. coli and lactobacilli. They demonstrated the bactericidal effect and chemical reactions of acidified nitrite under conditions simulating the stomach, suggesting its role in the ecology of the gastrointestinal tract and in host physiology (Xu et al., 2001).

Major et al. (2010) demonstrated that acidified nitrite could kill the cystic fibrosis pathogens P. aeruginosa, S. aureus, and B. cepacia under anaerobic planktonic and biofilm conditions. Schlag et al. (2007) showed that staphylococcal biofilm formation and polysaccharide intercellular adhesion synthesis can be inhibited by nitrite and preformed biofilm removal. Dave et al. (2012) developed a hydrogel-based dressing containing citric acid cross-linked cotton gauze and sodium nitrite loaded gelatin. The dressing showed significant antimicrobial activity against both planktonic and biofilm forms and was effective during consecutive reuses.

Ormerod et al. (2011) carried out an observational prospective clinical trial of topical acidified nitrite cream (4.5% citric acid and 3% sodium nitrite) for killing MRSA in contaminated wounds. Nine of 15 wounds (60%) and three of eight patients (37%) were cleared of infection. A similar preparation was used to treat onychomycosis (Finnen et al., 2007) in 13 patients with positive mycological cultures for Trichophyton or Fusarium species for 16 weeks. The formation of S-nitrosocysteine was shown throughout the treated nail, and > 90% of the nails became culture negative.

NO•-releasing nanoparticles

There have been several published studies describing the use of NO•-releasing nanoparticles (NO-np) for eradication of soft tissue and skin infections.

NO-np was first introduced by Freidman et al., and this nanoparticle has many of the critical characteristics that an ideal NO• carrier is supposed to have (Englander & Friedman, 2010). They developed a new biocompatible platform based on hydrogel/glass composites to prepare NO-np. The sucrose and trehalose were doped with glucose or tagatose for preparation of these composites, and they act as the thermal electron source during the thermal conversion of nitrite to NO•. The data demonstrated that therapeutic levels of NO• in a controlled and sustained manner were released following exposure of these nanoparticles to moisture. The release profiles for the NO• from these new platforms can be easily tuned by adding of specific additives in the structure of these composites.

Hetrick et al. (2009) investigated the ability of NO-releasing silica nanoparticles to eradicate biofilms of P. aeruginosa, E. coli, S. aureus, S. epidermidis, and C. albicans. The results showed that this approach was more effective, and more than 99% of cells from each type of biofilm were eliminated. Furthermore, investigators observed that Gram-negative bacteria were more susceptible than Gram-positive species. Based on these results, they concluded that using nanoparticles for delivery of antimicrobial agents such as NO• to microbial biofilms could be a promising approach.

Recently, NO-np was applied to treat wounds infected with A. baumannii (Ab; Mihu et al., 2010). The results demonstrated that NO-np significantly promoted healing of infected wounds in murine model. Additionally, in the treated model where nanoparticles were used, less inflammation, microbial burden, and degradation of collagen were observed. These findings suggested that NO-np was a promising candidate for the treatment of Ab wound infections. Han et al. (2012) investigated the effects of NO•-releasing nanoparticles on wound healing in mice. Their results showed that wound healing was accelerated through application of NO-np due to its promoting effects on fibroblast migration and collagen deposition. Moreover, NO-np modified leukocyte infiltration and increased angiogenesis and TGF-beta production in the wound area.

In the study by Friedman et al. (2011), the efficacy of a NO-np against certain Gram-negative and Gram-positive bacteria isolated from clinical samples was investigated. Samples were tested against both antibiotic-sensitive and antibiotic-resistant strains. The results showed that NO-np was uniformly effective against all clinical isolates obtained that is, E. faecalis, S. pyogenes, K. pneumonia, P. aeruginosa, and E. coli, and concluded that NO-np was a very powerful antimicrobial agent with a wide spectrum antimicrobial activity.

An innovative therapeutic method based on NO-np for C. albicans burn infection was developed by Macherla et al. (2012). Candida albicans growth and morphogenesis were arrested using NO-np; thus, its antifungal activity was demonstrated in a murine burn model. The results revealed that hydrogel/glass nanoparticles containing NO• enabled acceleration of wound healing in cutaneous burn infections, through modification of leukocyte migration, minimizing the fungal burden, and reducing collagen destruction. Based on these results, it can be suggested that NO-np can be emerged as a novel antifungal approach for the treatment of cutaneous burn infections and wounds (Macherla et al., 2012).

Antimicrobial activity of NO-np against MRSA abscesses in a mouse model was demonstrated by Han et al. (2009). The involved area and bacterial burden were reduced following topical or intradermal administrations of NO-np in abscesses, and skin architecture was preserved through inhibition of collagen degradation. Furthermore, it was observed that NO-np stimulated cytokine response in the mice model as quantities of TGF-β, MCP-1, IL-12, IL-1β, TNF-α, and IFN-γ increased following the treatment of infected mice with NO•-releasing nanoparticles.

Schairer et al. (2012) compared the effectiveness of NO-np and vancomycin for the treatment of intramuscular abscesses infected with MRSA. The results showed that treatment with topical (TP) and intralesional (IL) NO-np administration in combination with vancomycin was able to accelerate resolution of the abscesses, and both TP and IL administration of NO-np were more effective than systemic treatment with vancomycin alone.

Recently, a simple and reproducible method was developed for synthesis of NO–ruthenium nanoparticles. This water-soluble derivative of a novel NO-np could deliver NO• to reduced myoglobin following light irradiation in aqueous media. It is supposed that this new kind of NO-np may find some applications in biomedicine (Ho et al., 2011). To generate NO• in a controlled and sustained manner, a new platform based on hydrogel/glass hybrid for preparation of NO-np was synthesized. The requirement for stimulation of release of NO• from these nanoparticles is moisture environment, so when the prepared NO-np is exposed to moisture, the therapeutic levels of NO• start being released in a slow controlled manner. Later on, the NO• was trapped within the dry nanoparticle following thermal reduction in nitrite to NO•. The advantages of NO-np relative to other NO• releasing compounds are the sustained release of NO• controlled by the rate of hydration and no necessity for chemical decomposition nor enzymatic catalysis (Cabrales et al., 2010; Muzykantov, 2010).


ROS and RNS are attractive weapons to kill pathogenic microbial cells. They can be produced by a wide variety of mechanisms ranging from the use of standard antibiotics and disinfectants, to naturally occurring honey, to more hi-tech routes such as PDT, photocatalysis, cold plasma, and NO•-releasing nanoparticles. The selectivity for microbial cells over host mammalian cells appears to be favorable for many of these delivery routes. Although microbial cells can develop resistance by upregulating defenses to some types of ROS such as Embedded Image and H2O2, they are unable to defend themselves against others such as •OH and 1O2. The involvement of RNS in host defense and gastric pathogen destruction has been known for some time, but only recently has it been proposed as a therapy with NO•-releasing nanoparticles. The relative future development of these diverse ROS-/RNS-based strategies as clinical therapies remains difficult to predict, but at least one of them (and possible more) should make it to the clinic.


Research in the Hamblin laboratory is supported by the US NIH (R01AI050875). A.G. was supported by Department of Science and Technology, Government of India (BOYSCAST Fellowship 2010–11). R.Y. was supported by the National Natural Science Foundation of China (Grant No. 81172495). N.A.P. was supported by National Council for Scientific and Technological Development of Brazil (CNPq 200824/20111-2] and Fundacao de Amparo a Pesquisa do Estado de Sao Paulo (FAPESP 2010/07194-7, 2011/06240-8 and 2012/05919-0]. G.P.T. was supported by US NIH (5U54MH084690-02).


  • Editor: Antoine Danchin