Quinoxaline 1,4-di-N-Oxides: Biological Activities and Mechanisms of Actions

Quinoxaline 1,4-di-N-oxides (QdNOs) have manifold biological properties, including antimicrobial, antitumoral, antitrypanosomal and antiinflammatory/antioxidant activities. These diverse activities endow them broad applications and prospects in human and veterinary medicines. As QdNOs arouse widespread interest, the evaluation of their medicinal chemistry is still in progress. In the meantime, adverse effects have been reported in some of the QdNO derivatives. For example, genotoxicity and bacterial resistance have been found in QdNO antibacterial growth promoters, conferring urgent need for discovery of new QdNO drugs. However, the modes of actions of QdNOs are not fully understood, hindering the development and innovation of these promising compounds. Here, QdNOs are categorized based on the activities and usages, among which the antimicrobial activities are consist of antibacterial, antimycobacterial and anticandida activities, and the antiprotozoal activities include antitrypanosomal, antimalarial, antitrichomonas, and antiamoebic activities. The structure-activity relationship and the mode of actions of each type of activity of QdNOs are summarized, and the toxicity and the underlying mechanisms are also discussed, providing insight for the future research and development of these fascinating compounds.


INTRODUCTION
Quinoxaline is a heterocyclic compound containing a benzene ring and a pyrazine ring. Oxidation of both nitrogens of the pyrazine ring to obtain QdNO offers it variety of biological properties, including antitumoral, antibacterial, anticandida, antitrypanosomal, antiinflammatory/ antioxidant, and mutagenic properties. The versatile activities of QdNOs have aroused worldwide interests and endowed them potential application in human and veterinary medicines. For example, the QdNO antibacterial growth promoters have been widely used as feed additives to prevent infectious disease and improve animal growth since 1970s (Carta et al., 2005), and the antitumoral drug, TPZ (3-amino-BTO 1,4 dioxide, SR4233), has been subjected to phase II clinical trial (Covens et al., 2006;Maluf et al., 2006;Cohen et al., 2007). Although QdNOs with other activities are still in the research stage, they have exhibited great application prospects.
With the wide use of QdNOs, the toxicity and drug resistance gradually have become the disadvantages for further application of these promising compounds. The molecular targets of quinoxaline derivatives, which is a step beyond simply looking at their activities, should be analyzed. Nevertheless, knowledge about the mode of actions of QdNOs is far from clear, hindering the development of this kind of drugs. Therefore, a detail study on the mode of actions of QdNOs will provide information about the drug targets and the drug action pathway(s) and be helpful to construct models to screen new drugs. Meanwhile, the study of SAR combined with the study of drug action will explain the reason for that different structures of QdNOs exhibit different activities. Furthermore, since most of the drugs have more than one mechanism of action, the deeper study of the drug actions may discover new drug target(s) and provide potential approaches to conquer the problem of drug resistance and to mitigate or avoid toxicity. In this review, the mode of actions of QdNOs and the SAR analysis are elucidated and updated in sort of the different biological properties, giving insight to the future development of these fascinating compounds.
Olaquindox exhibits wide spectrum of antibacterial effect, strongly inhibiting the growth of Escherichia coli, Proteus, Pasteurella, and dysentery spirochete (Drumev, 1981). Because of its genotoxic potential, OLA is not permitted to use in foodproducing animals in the European Union and Canada (WHO, 1994;European Union [EU], 1998; Standing Committee for Animal Nutrition [SCAN], 1998). CBX shows good therapeutic effect on swine dysentery caused by Brachyspira hyodysenteriae (Rainier et al., 1973a,b;Downing, 1974) and also has good antibacterial effects against Salmonella, E. coli, and other Gramnegative bacteria (Troutt et al., 1974;Das, 1984). CBX was once used as growth-promoting feed additives for piglets or pigs at growing phase (Yen and Pond, 1993) before it was prohibited by the European Union [EU] (1998) because of its mutagenic effects, developmental and reproductive toxicity and carcinogenicity (WHO, 1991;European Union [EU], 1998). In view of the significant effect of CBX on swine dysentery and bacterial enteritis, the United States, Canada, and other countries still allow it used as therapeutic agents. MEQ, which shows good antibacterial activity against Gram-negative bacteria, especially Salmonella, has been widely used in China as an animal feed additive and veterinary medication for diseases, such as swine dysentery and piglet white diarrhea (Liu et al., 2012). QCT is active against B. hyodysenteriae, and is also effective against Salmonella, E. coli, and other Gram-negative bacilli. In China, QCT is used as a growth promoting agent for pigs, poultry, and aquatic and has been approved as an animal growth promoter in China since 2003 (Ministry of Agriculture of P. R. China, 2003). CYA shows good antibacterial activity against Staphylococcus hyicus, Pasteurella multocida, and E. coli (Ding et al., 2006a,b), and also exhibits good growth-promoting effect on broilers and swine. With good clinical safety, CYA has been regarded as a potential replacement of OLA and CBX (Cihak and Srb, 1983).
Only a few studies have investigated the mechanism of antibacterial action of QdNOs. Suter et al. (1978) first found the synthesis of DNA (but not RNA and protein) was completely inhibited by QDX in the absence of oxygen. QdNOs also induced degradation of DNA in both proliferating and nonproliferating cells, while strains were more resistant in the presence of oxygen. QDX was reduced to quinoxaline-N-oxide by the intact E. coli cells or the cell-free extract. EPR measurements demonstrated the generation of free radicals during the reduction of QDX. Oxygen or deficiency of energy sources impaired the antibacterial activity and the reduction of QDX. In consistence with Suter's result, our group recently found that CYA also had anaerobe-selective activity, and losing one or two of the oxygen's of CYA exhibited no antibacterial activity (Cheng et al., 2015).
The ability of QDX to cleave DNA was explicitly characterized using in vitro assays by Ganley et al. (2001). The results evidenced that QDX was a hypoxia-selective, redox-activated DNA-cleaving agent. The action of QDX on DNA yielded direct strand breaks with almost no sequence specificity, consistent with the involvement of radical species. In the absence of oxygen, QDX (a) received an electron and a hydrogen ion to form a neutral radical (b) (Figure 2). They considered two possibilities about the DNA-cleaving radical resulting from redox activation of QDX (a). First, the radical b might directly abstract hydrogen FIGURE 2 | One-electron reduction of quinoxaline 1,4-dioxide by reductase (Ganley et al., 2001). atoms from the DNA backbone, followed by elimination of water to yield the quinoxaline monoxide (c). Alternatively, radical b might fragment to form the known DNA-cleaving agent • OH and monoxide c.
Early study showed that polA, recA, recB, recC, exrA, and uvrA E. coli mutants were more susceptible to QdNOs than the corresponding repair-proficient strains, while the QdNO reductase activity was demonstrated to be lower in QdNOresistant mutants than in the susceptible parent strain (Suter et al., 1978), indicating that QdNOs were reduced and caused extensive DNA damage in bacterial. Recently, our group investigated the transcriptomic and proteomic profiles of E. coli exposed to QdNOs and found that QdNOs mainly induced SOS response and oxidative stress (Cheng et al., 2015). We also confirmed that ROS were induced in the QdNO-treated bacteria and the free radical scavengers attenuated the antibacterial action of QdNOs and DNA damage, suggesting an oxidative-DNAdamage action of QdNOs. The QdNO radical intermediates, likely carbon-centered and aryl-type radicals, as identified by EPR, were the major radicals induced by QdNOs, and xanthine oxidase was identified as one of the QdNO-activating enzymes by using specific enzyme inhibitors (Cheng et al., 2015).
Tirapazamine is a prodrug undergoing clinical trials for various types of cancers (discussed in details in section "Antitumor activity of QdNOs"). Shah et al. (2013) showed that TPZ has antibacterial activity against E. coli, S. aureus, and Clostridium difficile, particularly at low oxygen levels. E. coli mutants deficient in HR were hypersusceptible to TPZ, suggesting that drug toxicity may be due to DNA damage. Moreover, E. coli strains deleted for genes encoding putative reductases were resistant to TPZ, implying that these enzymes are responsible for conversion of the prodrug to a toxic compound.
Though the two N-oxide groups are necessary for the antibacterial activity of QdNOs, some reduced form of quinoxaline compounds were still reported with antibacterial activity (Refaat et al., 2004;Singh et al., 2010Singh et al., , 2011. For example, when the C 2 chlorine of 2-Chloro-3-methylquinoxaline was replaced with a benzene ring harboring an aldehyde or a free amino group which can be further reacted with aromatic amines and aromatic aldehydes, this compounds also show antimicrobial activity (Singh et al., 2010). Therefore, in addition to the N-oxide group, the side chain of quinoxaline is another determinant of activity. Hansen et al. (2004) first reported the gene-encoded resistance mechanism to OLA. In their study, two genes of oqxA and oqxB, encoding for proteins homologous to resistance-nodulation-celldivision family efflux systems, were cloned from a conjugative plasmid isolated from E. coli. Plasmids containing the oqxAB genes yielded high resistance to OLA and chloramphenicol in E. coli, indicating oqxAB encodes a multidrug efflux pump. Later, they demonstrated the prevalence of the OqxAB efflux pump by horizontal transfer of OLA resistance from OLA-resistant isolates using an OLA-sensitive E. coli as recipient (Hansen et al., 2005). In addition to OLA and chloramphenicol, the OqxAB pump conferred antimicrobial resistance or reduced susceptibility toward a variety of substrates in E. coli, including animal growth promoters, antimicrobials, disinfectants and detergents (Hansen et al., 2007). Interestingly, oqxA gene was not detected in the CYA/OLA-resistant E. coli induced in vitro (Guo et al., 2012), suggesting there are other mechanisms conferring the bacterial resistance.

Antimycobacterial Activity
Tuberculosis is a common, and in many cases lethal, infectious disease caused by various strains of mycobacteria, usually Mycobacterium tuberculosis. TB has been a companion of mankind since the beginning of human history. In 2010, there where 8.8 million new cases, and 1.5 million deaths, mostly in developing countries (WHO, 2011), where more people contract TB because of compromised immunity due to high rates of AIDS (Lawn and Zumla, 2011). Drug-resistant TB is a public health issue in many developing countries, as treatment of it is longer and requires more expensive drugs. MDR TB is defined as resistance to the two most effective first-line TB drugs, rifampicin and isoniazid. Over the past years, QdNO derivatives as antitubercular drug candidates hold promise in the treatment of TB and its resistant form.
The first publication of QdNO derivatives as antimycobacterium agents dates back to the end of the 1990s by Monge's team (Montoya et al., 1998). Continued identification of novel antitubercular candidates based on their potency (MIC to H 37 Rv M. tb), selectivity [SI = IC 50 (to VERO cell)/MIC(to H 37 Rv M. tb)] and low cytotoxicity make them valid new leads for synthesizing additional analogs with improved antitubercular activity both in vitro and in vivo. A comprehensive review on the properties of QdNO derivatives developed as potential antitubercular agents by Monge's team was reported recently (Vicente et al., 2011). As shown in Figure 3, the QdNO antitubercular compounds can be divided into categories based on their structures, including amide derivatives (1-3), ketone derivatives (4-6), ester derivatives (7-10) and other derivatives (11-12). Compounds 8 and 10, proved efficacious in vivo in a murine model of low dose aerosol infection. Moreover, these two compounds also showed activity against non-replicating bacteria, indicating that QdNOs might lead to shortened therapy, because nonreplicating bacteria is believed to be a major factor responsible for the prolonged nature of antitubercular therapy. Compound 8 is also active on single-drug resistant and MDR clinical isolates.
Structure-activity relationship suggests that the 1,4-di-Noxide groups in quinoxaline ring are important to increase the antimycobacterial activity (Montoya et al., 1998;Ortega et al., 1999;Sainz et al., 1999). Quinoxaline-2-carbonitrile derivatives has good antimycobacterial activity but appeared to be quite toxic (Ortega et al., 2001(Ortega et al., , 2002; thus, the replacement of the carbonitrile group with a carboxamide , acetyl, benzoyl , or carboxylate groups  was proposed. Among the 6(7)substituted quinoxaline-2-carboxylate 1,4-dioxide derivatives, anti-TB activity principally depends on the substituents in the carboxylate group, improving in the following order: benzyl > ethyl > 2-methoxyethyl > allyl > tert-butyl, and the presence of a chloro, methyl, or methoxy group in position 7 of the benzene moiety reduces the MIC and IC(50) values . Recently, a research employed 3D-QSAR and docking analysis to identify molecular structural features required for effective antimycobacterial activity of quinoxaline-2-carboxamide 1,4-di-N-oxide derivatives inside the active site of the Mycobacterium DNA gyrase B subunit,  and the obtained binding mode was as same as that of the novobiocin X-ray structure (Radwan and Abdel-Mageed, 2014).

ANTITUMOR ACTIVITY OF QdNOs
It has been recognized for more than 50 years that very low levels of oxygenation (hypoxia) protect cells from killing by X-irradiation. Hypoxic cells, which are resistant to radiotherapy, are present in solid tumors but not in normal tissues, therefore become a major problem for radiation therapy of tumors. Additionally, hypoxic cells are non-proliferating which do not respond to drugs active against only proliferating cells, and are distant from the blood vessels carrying the drug. Thus the development of drugs with selective toxicity toward hypoxic cells is a key objective in anticancer chemotherapy. The recent development of QdNOs that are non-toxic until they are activated in the hypoxic cell opens a new era.
Tirapazamine (Zeman et al., 1986) ( Figure 6A), which actually belongs to BTOs, is the first drug that has been shown to be an efficient and selective cytotoxin for hypoxic cells. TPZ has been subjected to phase II testing in patients with head, neck, and gynecological cancers (Covens et al., 2006;Maluf et al., 2006;Cohen et al., 2007).
In addition to TPZ, many research teams have been devoted to the synthesis and development of QdNO antitumoral drugs. Monge et al. (1995b) synthesized a series of 3-aminoquinoxaline-2-carbonitrile 1,4-di-N-oxides ( Figure 6B) with a range of electron-donating and -withdrawing substituents in the 6and/or 7-positions and evaluated for toxicity to hypoxic cells. As the electron-withdrawing nature of the 6(7)-substituent increases, the reduction potential becomes more positive and the compound is more readily reduced. The most potent cytotoxins were the 6,7-dichloro and 6,7-difluoro derivatives, which were 30-fold more potent than TPZ. The compound 7-(4-nitrophenyl)-2-quinoxalinecarbonitrile 1,4-di-N-oxide ( Figure 6C) is 150-fold more potent than TPZ, demonstrating that the amino group in three position is not necessary for activity (Monge et al., 1995a). Meanwhile, they developed two derivatives bearing a basic side chain. The 7-choloro- and 7-trifluoromethyl 3-(dimethylaminopropyl) amino-2quinoxalinecarbonitrile 1,4-dioxide ( Figure 6D) showed high potency and excellent selectivity (Monge et al., 1995a). They further studied the activity of the basic chain in 3 position of the 2-quinoxalinecarbonitrile 1,4-dioxides and showed that the replacement of (N,N-dialkyl amino) alkyl amino chain with aromatic rigid moieties (anilines and arylpiperazines) in three position held the potency but reduced the hypoxia-selectivity (Ortega et al., 2000). Continuingly, they synthesized a new series of 2-alkylcarbonyl and 2-benzoyl-3-trifluoromethylquinoxaline 1,4-di-N-oxide ( Figure 6E) and evaluated for in vitro antitumor activity against MCF7 (breast), NCI-H460 (lung), and SF-268 (central nervous system) cells. In general, anticancer activity depends on the substituents in the carbonyl group, improving in the order: ethyl < isopropyl < tert-butyl < phenyl-ones (Zarranz et al., 2004). By comparison of the 1 H NMR spectra, Solano et al. (2007) also showed that the best activity was observed in derivatives with electron-withdrawing groups in position 6 or 7 on the quinoxaline ring whereas loss of one or two oxygens reduced the cytotoxicity.
Although TPZ has attractive features of targeting hypoxic cells in tumors, it has limited clinical activity, in part because of poor extravascular penetration. Hicks et al. (2010) used a spatially resolved pharmacokinetic/pharmacodynamic model to guide the progression of 281 TPZ analogs through a hierarchical screen. SN29751 ( Figure 6I) and SN30000 ( Figure 6J) were identified as the most promising hypoxic cytotoxins, and SN30000, in particular, showed higher hypoxic potency and selectivity than TPZ in tumor cell cultures and faster diffusion through HT29 and SiHa multicellular layers.

QdNO-Induced Radicals
The mode of action of QdNO antitumoral drugs was most comprehensively studied in TPZ. As shown in Figure 7, it is known that TPZ (1) is an excellent substrate for a variety of intracellular reductases that add an electron to the drug to form an radical anion (2) (Brown, 1999), which then receives a proton to form a neutral radical (3). This neutral radical itself or produces other active radicals to abstract hydrogen from DNA, producing both SSBs and DSBs, resulting in chromosome breaks (Wang et al., 1992). Under aerobic conditions, oxygen can remove the additional electron from the radical anion (2), thereby back-oxidizing it to the non-toxic parent with a concomitant production of O •− 2 (Lloyd et al., 1991). Thus, the differential hypoxic cytotoxicity results from the fact that the TPZ radical is much more cytotoxic than the O •− 2 . There have been a number of suggestions about the identity of active radicals produced by TPZ. An early proposal was that the • OH radical was released from the protonated anion (3), and the monoxide of TPZ (4) was produced simultaneously (Laderoute and Rauth, 1986), but this was later rejected in favor of the protonated radical anion itself as being the active species (Laderoute et al., 1988). Release of an • OH radical has been favored by other research workers, based on the similarity in the spectrum of products arising from TPZ-mediated damage to both purine and pyrimidine bases and deoxyribose sugars when compared with damage arising from the • OH radical, although TPZ-mediated damage exhibited some preference for purine over pyrimidine damage (Kotandeniya et al., 2002;Birincioglu et al., 2003;Chowdhury et al., 2007). Additionally, an EPR study using DMPO as a spin trap reported a composite spectrum of both a Ccentered and an OH-adduct, DMPO-OH (Patterson and Taiwo, 2000), supporting evidence for the release of an • OH radical.
In contrast, Shinde et al. (2009Shinde et al. ( , 2010 could not find any evidence that • OH is eliminated from TPZ following enzymatic reduction. They presented that the protonated radical anion (3) undergoes dehydration to the BTZ (5) (Shinde et al., 2009), causing oxidative damage to DNA and to oxidize TPZ itself (Anderson et al., 2003a(Anderson et al., ,b, 2006. In situ EPR measurement confirmed that a multi N-centered radical, consistent with the BTZ radical (5), was formed (Shinde et al., 2009). Similarly, the one electron reduction potentials [E(1)R] of the BTZ radicals (5) also tracked the hypoxic cytotoxicity to human tumor cells in a series of BTO analogs of TPZ (3-amino BTO; Anderson et al., 2005). However, although the E(1)R value for the BTZ radical (5) of TPZ, 1.31 V (Anderson et al., 2003b), is sufficient to oxidize the purine bases of DNA, but not to directly oxidize the pyrimidine bases of DNA. Later, they trapped TPZ and a series of 3-substituted analogs by N-tert-butyl-R-phenylnitrone and observed by EPR the formation of their aryl (phenyl) radicals (6), which is stronger oxidants than the BTZ radical (5) (Shinde et al., 2010).
There are still some questions whether protonated radical anion (3) will undergo N-OH homolysis to its monoxide (4) and • OH, or alternatively, radical anion (2) may react by dehydration and form aminyl radicals (5) and H 2 O or phenyl radicals (6) and H 2 O. Recently, Yin et al. (2012) discovered that dehydration might be the result of a two-step sequence that involved N-OH homolysis and formation of • OH aggregates of the monoxide (4) followed by H-abstraction within the • OH aggregates to form hydrates of aminyl (5) or of phenyl (6) radicals (Figure 7, dashed arrows).
Electron paramagnetic resonance experiments indicate both an aryl-type radical and an oxidizing radical, trapped as a carbon-centered radical, are formed from the protonated radical anion of the bioreductive anticancer prodrug, SN30000 . The carbon-centered radical, produced upon the one-electron oxidation of the 2-electron reduced metabolite of SN30000, oxidizes 2deoxyribose, leads to double strand breaks. Hunter et al. (2014a) found hypoxia-activated prodrugs of DNA-damaging cytotoxins (TPZ, SN30000) might inhibit growth of triplenegative breast cancer (TNBC) by simultaneously addressing the two targets, tumor hypoxia and derangement of HR repair.
Recently, Yadav et al. (2014) found The • OH radical was released from 3-trifluoromethyl-quinoxaline 1,4-dioxides upon one-electron reduction by cytochrome P450 oxidoreductase. This process effectively competes with back oxidation of the intermediate radical anion by oxygen and underlies the increased aerobic cytotoxicity of such compounds compared to that seen for, TPZ.

Reductase
Another major unresolved question is the identity of the enzyme(s) that activates a QdNO to cause cell death. Numerous enzymes, including xanthine oxidase (EC 1.2.3.1; Laderoute et al., 1988), cytochrome P450 (Wang et al., 1993), DT-diaphorase (EC 1.6.99.2; Patterson et al., 1994), and NADPH:cytochrome P450 reductase (EC 1.6.2.4; Patterson et al., 1995), are able to metabolize TPZ in vitro under hypoxic conditions. A correlation between NADPH:cytochrome P450 reductase level (Patterson et al., 1997) or NOS (Chinje et al., 2003) level and sensitivity to TPZ has been reported. For a long time, there was no agreement as to which enzyme(s) are involved in the DNA damage until Evans et al. (1998) showed that TPZ was metabolized to DNA-damaging radicals by intranuclear enzymes. TPZ radicals formed outside nuclei do not contribute to intranuclear DNA damage, and the 80% of the drug metabolism that occurs in the cytoplasm is probably irrelevant for hypoxic killing effect of this drug. Later, Delahoussaye et al. (2001) demonstrated that multiple reductases in the nuclear matrix metabolized TPZ under hypoxia. DNA SSBs were probably caused by the most abundant source of reductase in the nucleus, while DNA DSBs were formed by an unknown nuclear reductase requiring only NADPH for its activity. Hunter et al. (2014b) observed that forced expression of FAD-dependent oxidoreductase domain containing 2 (FOXRED2) increased activation of hypoxia-targeted prodrugs TPZ and SN30000. They also identified the flavoprotein P450 (cytochrome) oxidoreductase, which is responsible for prodrug activation during hypoxia, as the predominant determinant of sensitivity to SN30000 (Hunter et al., 2015).

Topoisomerase and Polymerase β
Topoisomerase II is essential in mammalian cells, because it resolves the unfavorable topological structures in DNA during replication and transcription. Drugs that stabilize Topo II with a DNA DSB to form cleavable complex are termed Topo II poisons (Pommier et al., 2010). Since TPZ produces a marked inhibition of DNA replication in the nuclear matrix (Peters et al., 2001), and TPZ-induced DNA DSBs are protein-associated (Olive, 1995;Siim et al., 1996), Peters and Brown (2002) considered TPZ might poison Topo II. They found under hypoxic conditions, the nuclear extracts from LXFL 529 human lung carcinoma cells treated with TPZ reduced the activity of Topo II. Inhibitors of the Topo II catalytic cycle abrogated TPZgenerated DNA DSBs and cytotoxicity, and TPZ stabilized DNA Topo II cleavable complexes (Peters and Brown, 2002). Using Saccharomyces cerevisiae as a model, overexpression of TOP2 (encoding Topo II) leads to hypersensitivity to TPZ, suggesting that Topo II is also a target of TPZ in yeast (Hellauer et al., 2005). XK469 (NSC 697887) and CQS (NSC 339004), two synthetic quinoxaline derivatives without two oxygens on the quinoxaline ring, also show solid tumor selectivity. XK469 and CQS have entered Phase I and Phase II clinical study, respectively (Miller et al., 1997;Bekaii-Saab et al., 2006;Alousi et al., 2007). Gao et al. (1999) reported the primary target of XK469 is Topo IIβ, and CQS was found to be both a Topo-IIα and a Topo-IIβ poison (Gao et al., 2000). The large aromatic side chains of XK469 and CQS can not be ignored to the drug action. Except for TopII, TPZ also induces other DNA-protein cross-links, including DNA-Topo I cleavable complexes (Evans et al., 2008) and DNA-Polβ cross-link with the lesion (Sung and Demple, 2006). Evans et al. (2008) further showed an overall model of TPZ damage in which DNA SSBs, base damage, and DNAprotein cross-links (including Topo I and II cleavable complexes) produced stalling and collapse of replication forks. The resolution of the complex required HR and XPF/ERCC1 protein for their repair. Cells defective in HR proteins were particularly sensitive to TPZ, and extensive sister chromatid exchanges occurred after treatment with TPZ. In addition, TPZ preferentially kills mutants both with defects in XPF/ERCC1 and base excision repair. H2AX, an indicator of DNA DSBs, is induced preferentially in cells in the S phase of the cell cycle.

Hypoxia-Inducible Factor 1
Hypoxia-inducible factor 1, a heterodimeric transcription factor that mediates the adaptation of tumor cells and tissues to the hypoxic environment, has attracted considerable interest as a potential therapeutic target. The subunit 1α of HIF plays an essential role in the transcriptional activation of genes involved in tumor angiogenesis invasiveness and metastasis with its downstream target, VEGF (Carmeliet et al., 1998).
Combinational use of TPZ with other inhibitors represents a novel mechanism for targeting tumor. In combination with TPZ, topoisomerase I inhibitors exhibited synergistic cytotoxicity and induced significant apoptosis in several hepatocellular carcinoma cell lines (Cai et al., 2014). The enhanced apoptosis induced by TPZ plus SN-38 (the active metabolite of irinotecan) was accompanied by increased mitochondrial depolarization and caspase pathway activation. The combination treatment dramatically inhibited the accumulation of HIF-1α protein, decreased the HIF-1α transcriptional activation, and impaired the phosphorylation of proteins involved in the HR repair pathway, ultimately resulting in the synergism of these two drugs. TPZ mediates central vascular dysfunction in tumors and is also a competitive inhibitor of NOS. Baker et al. (2013) further investigated the vascular-targeting activity of TPZ by combining it with NOS inhibitor L-NNA, or with low oxygen content gas breathing. Irreversible loss of perfusion and enhanced tumor cell death was observed when TPZ was combined with either low oxygen or a NOS inhibitor, illustrating a novel use of hypoxiaactivated cytotoxic prodrugs as vascular targeting agents.

ANTIPROTOZOAL ACTIVITIES OF QdNOs Antitrypanosomal Activity
Trypanosoma cruzi is the haemoflagellate protozoan that causes the Human American trypanosomiasis, or Chagas disease, representing a relevant health problem in Central and South America. The acute form of Chagas disease usually goes unnoticed and may present as a localized swelling at the site of entry. The chronic form may develop 10 to 20 years after infection, which affects internal organs (e.g., the heart, the esophagus, the colon, and the peripheral nervous system), sometimes causing death to affected people from heart failure. The first line of treatment is nifurtimox and benznidazole.
Cerecetto's team (Cerecetto et al., 1999) firstly noticed QdNO derivatives had antitrypanosomal activity against epimastigote forms of T. cruzi. They reported a vanadium complexe of QdNO, [V(IV)O(L)(2)], where L were 3-aminoquinoxaline-2carbonitrile 1,4-di-N-oxide derivatives (Figure 8A), was provide with excellent antitrypanosomal activity, similar to that of the reference drugs nifurtimox and benznidazole and higher than that of the corresponding free ligands. The antitrypanosomal activity of these vanadium complexes could be explained on the basis of their lipophilicity and the electronic characteristics of the quinoxaline substituents (Urquiola et al., 2006). They also reported two QdNO compound, R3 = methyl (SI > 53.3, selectivity index = ID 50 (macrophage)/ID 50 (T. cruzi)) and R3 = phenyl (SI > 33.3; Figure 8B), displayed excellent parasite/mammal selectivity indexes. Those compounds are able to accumulate squalene suggesting that anti-T. cruzi mechanism of action is not involved the inhibition of sterol biosynthesis (Gerpe et al., 2010). Monge's team also found that QdNO derivatives presented good inhibitor activity to the growth of T. cruzi in in vitro assays, among which a QdNO derivative (Figure 8C) were the most cytotoxic compounds against the protozoan, and the IC 50 were of the same order of that of nifurtimox (Aguirre et al., 2004). The authors reported some SAR deductions: the presence of a halo-substituent (electron-withdrawing) at benzene moiety of quinoxaline produces more active compounds; more hydrophilic compounds decrease the antitrypanosomal activity; a reductive metabolism could be implicated in the mechanism of action. Later, they prepared a series of heterocyclic-2-carboxylic acid (3-cyano-1,4-di-N-oxidequinoxalin-2-yl)amide derivatives (Figure 8D), among which compounds (R6 = F, R7 = F, X = O) and (R6 = F, R7 = F, X = S; Figure 8D) presented IC 50 values in the same order as nifurtimox (Ancizu et al., 2009). Recently, they synthesized a series of 3-trifluoromethyl QdNOs, among which derivatives possessing electron-withdrawing substituents in the 2-, 3-, 6-, and 7-positions were the most active compounds (Benitez et al., 2011). One 3-trifluoromethyl QdNO compound (Figure 8E), which was substituted with fluoro groups at the 6-and 7-positions of the quinoxaline ring, was the most active (IC 50 = 0.4 µM) and selective (SI = 10) in the cytotoxicity assay (Torres et al., 2013). It is demonstrated that inhibition of mitochondrial dehydrogenases are involved in the anti-T. cruzi activity of the most active derivatives (Benitez et al., 2011;Torres et al., 2013).
Leishmaniasis is a parasitic disease which appears in visceral, cutaneous and mucocutaneous forms affecting millions of people throughout the world. Monge's team tested pyrazole quinoxaline derivatives against Leishmania peruviana. 2,6-dimethyl-3-fquinoxaline 1,4-dioxide (Figure 8F) was the most active compound of this series against L. peruviana (IC 50 = 8.9 µM; Estevez et al., 2011). This compound resulted non-toxic for Vero and LLc-Mk2 cells, and was almost 6 to 13 times more active on Leishmania than on THP-1 or MPM. According to the study of Barea et al. (2012), piperazine linked QdNO (Figure 8G) emerged as the best leishmanicidal agent against L. infantum (IC 50 = 5.7 µM). Their later study indicated that the R = cyclohexyl derivative (Figure 8H) had the best antileishmanial activity against L. infantum (IC 50 = 2.5 µM) while the R = 3-chloropropyl derivative (Figure 8H) was the best against L. amazonensis (IC 50 = 0.7 µM; Barea et al., 2013). Villalobos-Rocha et al. (2014) evaluated the in vitro biological activity of 33 ethyl and methyl quinoxaline-7-carboxylate 1,4di-N-oxide derivatives on T. cruzi and L. mexicana, of which M7 ( Figure 8I) and E4 ( Figure 8J) displayed activity against both parasites. Compound M2 (Figure 8K) was predicted in the docking procedure as a potential T. cruzi trypanothione reductase inhibitor by its interaction with five residues close to the active site of the enzyme.

Antimalarial Activity
Plasmodium falciparum is one of the species of Plasmodium that cause malaria in humans. Malaria caused by this species is the most dangerous form of malaria, with the highest rates of complications and mortality. For nearly half a century, chloroquine has been the primary therapy of choice. However, chloroquine-resistant P. falciparum is now observed in nearly all of the malaria-endemic regions and causes the most deadly form of malaria. Therefore, it is necessary to develop cheaper and more effective drugs against the parasite. Zarranz et al. (2005) synthesized new series of 3arylquinoxaline-carbonitrile derivatives ( Figure 9A) and tested for their in vitro and in vivo activity against the erythrocytic development of P. falciparum with different chloroquineresistance status. These series showed superior antimalarial activity in respect to reduced quinoxaline analogs. The best activity was observed with non-substituted QdNOs in positions 6 and 7 of the aromatic ring and with a hydrogen or chloro substituent in para position of the phenyl group ( Figure 9A). Vicente et al. (2008) identified new compounds structurally based on 3-phenyl-quinoxaline-2-carbonitrile 1,4-di-N-oxide derivatives ( Figure 9B) active against P. falciparum. Derivative 1 (Figure 9B Figure 9C) and evaluated for their capacity to inhibit the growth of chloroquine-resistant P. falciparum FCB1 strain in culture. Compound 7-chloro-2-(2-furylcarbonyl)-3-trifluoromethyl-1,4-quinoxaline di-N-oxide ( Figure 8C, Ar = 2-Furyl, R1/R2 = Cl/H) was the most active, being almost five times more active than chloroquine. It was also 50 times more active against P. falciparum than toxic toward MCF7 cells. SAR showed that bioisosteric modification of phenyl group by 2-thienyl or 2-furyl subunits, R2 position must be free or occupied by a methyl group and R1 position must be free or occupied by Cl, CH 3 , OCH 3 or CF 3 . Barea et al. (2011) synthesized new 3-amino-1,4-di-N-oxide quinoxaline-2-carbonitrile derivatives as acetoxybenzamides ( Figure 9D) and sulfonamides (Figure 9E), and evaluated for their in vitro antimalarial and antileishmanial activity. Compounds with one halogenous group substituted in position 6 and 7 provide an efficient approach for further development of antimalarial and antileishmanial agents. From a series of quinoxaline analogs of chalcones in another study, compounds 1a (IC 50 = 6.2 µM; Figure 9F) and 2a (IC 50 = 5.8 µM; Figure 9G) were the most active against FCR-3 P. falciparum (Gil et al., 2014). SAR demonstrated the importance of an enone moiety linked to the quinoxaline ring in the search for antimalarial ligands.
The intracellular mechanism of the antiparasitic action of QdNOs is barely elucidated. Recently, Brizuela et al. (2014) shows that one QdNO-derived compound, Conoidin A [2,3-bis(bromomethyl)-1,4-dioxide-quinoxaline] (Figure 9H), exhibites potent lytic activity against P. falciparum and constitutes an irreversible Peroxiredoxin-2 (Prx2) inhibitor. Conoidin A is first shown as an inhibitor of host cell invasion by the human pathogen T. gondii (Carey et al., 2004), and this compound is a covalent inhibitor of T. gondii peroxiredoxin II (Haraldsen et al., 2009). Conoidin A can also inactivate the peroxiredoxin-1 from the human hookworm Ancylostoma ceylanicum by alkylating or crosslinking the catalytic cysteines, while maintaining the enzyme in the "locally unfolded" conformation (Nguyen et al., 2013). Peroxiredoxin, a thioldependent peroxidase, serves a critical role in converting ROS signals into a cellular response. When Plasmodium sp. invades red cells, it imports Prx2 from the host cell to the parasite cytosol during intraerythrocytic development in an attempt to make up for degradation of peroxides generated during cell metabolism. Therefore, treatment of erythrocytes with Conoidin A produces an unviable growth of the parasite inside, and enhances parasite sensitivity to chloroquine (Brizuela et al., 2014).

Antitrichomonas Activity
Trichomoniasis is a protozoan infection of the human and bovine urogenital tracts. Metronidazole, Tinidazole and other nitroimidazoles are the most effective drugs, currently available for treatment. In the early 1980s, Glazer and Chappel (1982) reported the synthesis and the activity of a novel series of pyrido[2,3-b]quinoxaline 5-oxides against Trichomonas foetus. Carta et al. (2004) reported the synthesis and antitrichomonas activity of a series of 6,7-difluoro-3-methylquinoxaline 1,4dioxides ( Figure 10A). In particular, several 2-phenylthio derivatives resulted 20-to 30-fold more potent than the reference drug Metronidazole activity against T. vaginalis (SS22) in vitro, isolated in Italy from a case of acute vaginal trichomoniasis. For example, sulfoxide derivative ( Figure 10B) was reported to be more effective than the reference drug metronidazole against T. vaginalis while one compound (Figure 10C) inhibited the

ANTIINFLAMMATORY AND ANTIOXIDANT ACTIVITIES OF QdNOs
Arachidonic acid metabolism, mediated by the LOX enzyme family, leads to the generation of leukotrienes, a type of pro-inflammatory mediator involved in processes such as fever, asthma, or cardiovascular disease. Additionally, aberrant arachidonic acid metabolism is related to carcinogenesis. For example, increased LOX expression levels have been found in a wide range of cancers, including pancreatic, bladder, and breast cancer (Hofmann and Steinhilber, 2013). During the inflammation process, phagocytic leukocytes produce ROS. A number of commercially available non-steroidal antiinflammatory drugs, such as acetaminophen, salicylates, indomethacin, and nimesulide, have been shown to possess radical scavenging properties. Therefore, the development of new compounds having both antiinflammatory and antioxidant activities and being LOX inhibitors is an interesting approach for cancer prevention, treatment of chronic inflammation and other related pathological conditions. 3-phenyl-1-(1,4-di-N-oxide quinoxalin-2-yl)-2-propen-1one derivatives and their 4,5-dihydro-(1H)-pyrazole analogs were discovered to show very interesting antioxidant and antiinflammatory properties. Compound 2a (Figure 12, series 2) displayed an in vivo antiinflammatory effect (56.1%) higher than the reference drug, indomethacin, and promising in vitro inhibition values of LOX (IC 50 < 1 µM; Burguete et al., 2007). Compound 7b (Figure 12, series 7) showed significant protection against carrageenan-induced paw edema, in which the in vivo antiinflammatory effect (41%) was similar to that of indomethacin (47%; Burguete et al., 2011). Burguete et al. (2011) has studied the SAR which demonstrates that the radical scavenging ability needs a phenolic group (series 2) and/or a free amino pyrazoline ring (series 3, 8, and 10) in the structure (Figure 12). In the presence of the pyrazoline moiety, those with N-oxide groups in the quinoxaline ring (series 3 and 10) exhibited significantly increased reducing activity compared to their reduced analogs (series 8), resulting in increased scavenging activity. Compounds with an α, β-unsaturated ketone system (1b, 2e, 4f, 5a, 5b, 5g, and 6a-f) presented the best O •− 2 scavenging activities, indicating that the olefinic moiety might play an important role in the activity by trapping the O •− 2 . The derivatives are also good • OH scavengers. In the inhibition of LOX, the best IC 50 values were shown by compounds of series 7, 9, and compound 8f. Although compounds that displayed good activities of inhibition of lipid peroxidation also presented good values of inhibition of LOX, the best inhibitors of lipid peroxidation were those compounds without any substitution in the quinoxaline ring while the best inhibitors of LOX were obtained by fluoro-and methyl-substituted derivatives.

TOXICOLOGICAL EFFECTS OF QdNOs
The toxicology effects, including short-and long-term toxicities Ihsan et al., 2010Ihsan et al., , 2011Wang et al., 2010Wang et al., , 2011aWang et al., ,b, 2012, genotoxicity (Ihsan et al., 2013a,b), and photoallergic toxicity , have been extensively studied in the veterinary-used QdNOs. The liver is the main target organ of QdNOs, and OLA and MEQ show obvious toxic effect on the kidney and the adrenal gland. High dose of OLA, MEQ, CBX, and QCT can significantly inhibit the growth and development, reproductive function and embryonic development of rats, whereas CYA has minor effect. OLA, CBX, and MEQ have genotoxicity, and CBX shows obvious carcinogenic effects. QCT has certain genetic toxicity, and no genotoxicity has been detected in CYA.
Quinoxaline 1, 4-di-N-oxides show mutagenic and DNAdamaging effects on various organisms. Voogd et al. (1980) reported that QDX, CAX, and OLA were mutagenic on Saccharomyces cerevisiae, Klebsiella pneumoniae, S. typhimurium and E. coli, causing base-pair substitutions and frame-shift mutations. Beutin et al. (1981) found that mutagenicity of QdNOs was dependent on the presence of N-oxide groups, since quinoxaline, a completely reduced derivative of QDX, was not mutagenic, whereas the partially reduced quinoxaline-N-oxide exhibited a lower mutagenic activity than QDX. It was suggested that the mutagenicity of QdNOs resulted from the error-prone repair involved in SOS responses (Nunoshiba and Nishioka, 1989). Ihsan et al. (2013a,b) compared the genotoxic potential of QdNOs in Ames test, HGPRT gene mutation test in V79 cells, unscheduled DNA synthesis assay in human peripheral lymphocytes, chromosome aberration test, and micronucleus test in mice bone, and found that OLA, CBX, and MEQ were more genotoxic than QCT and CYA. Using alkalic comet assay, pronounced increase of DNA fragmentation were observed in Vero cells treated with CBX, OLA, and QCT . In contrast, DNA damage was significantly decreased after incubation with S9 mix, suggesting that the intermediate metabolites of these compounds exerted lower genotoxicity than their parent drugs. Jin et al. (2009) also found that QCT caused significant DNA fragment migration in a dose-dependent manner in human hepatoma (HepG2).
The mutagenicity of QdNOs with antiparasitic activity has also been studied. Introduction of electron-withdrawing substituents at C-6 and/or C-7 of the quinoxaline ring enhanced in vitro biological activity against T. cruzi. In addition, this led to the obtainment of non-mutagenic derivatives in both the Ames assays that were performed or their mutagenicity disappeared when performing the assay using metabolic activation (Torres et al., 2013). Using Ames test, Gabay et al. (2014) investigated the mutagenicity of an in-house chemical library of eighty five N-oxide containing heterocycles, and they found that in some cases, a relationship was found between the presence of N-oxide and mutagenicity, and in other cases, such as quinoxaline dioxides with antiparasitic activity, mutagenicity was substituent dependent.
Quinoxaline 1, 4-di-N-oxides can produce ROS and cause oxidative cell damage. It was found that OLA induced the increased levels of ROS and 8-OHdG in HepG2 cells, inferring that OLA exerts genotoxic effects in HepG2 cells probably through the ROS-induced oxidative DNA damage . Using γ-H2AX as a surrogate marker for DNA damage, Liu et al. (2012) found that MEQ treatment induced cellular DNA damage, which paralleled the chemicalinduced elevation of ROS levels, and expression of the antioxidant enzyme catalase partially alleviated these MEQassociated effects. Huang et al. (2010b) found that OLA irritated a persistent and utmost release of ROS while MEQ made a similar but weaker reaction. CYA, however, had a short and unstable release of intracellular ROS. On the other hand, quinoxalinine-2-carboxylie acid, one of the metabolites of OLA and MEQ, did not cause any significant production of ROS and showed relatively lower toxicity than its parents. Zhang et al. (2014) reported that QCT damaged the antioxidant defense abilities of HepG2 cells by reducing the activities of endogenous antioxidant enzymes, lowering glutathione concentration, and elevating malondialdehyde level. One hundred and sixty QCT-responsive genes were found to be associated with cell proliferation, glucose metabolism, oxidative stress, and apoptosis, such as NAD(P)H dehydrogenase. However, QCT metabolites (1,4-bisdesoxyquinocetone and 3methylquinoxaline-2-carboxylic acid) showed little effects on HepG2 cells. Wang et al. (2015b) demonstrated that the rank orders of the desoxy and bidesoxy rates in rat and pig liver microsomes were QCT < CBX < MEQ < OLA < CYA and QCT < MEQ < CBX < OLA < CYA, respectively. In rats, porcine primary hepatocytes, and HepG2 cells, oxidative stress indices and DNA damage showed inverse relationships with the deoxidation rate, indicating that faster deoxidation of QdNOs results in lower DNA-damage-induced toxicity.
The molecular mechanism of cell cycle arrest and apoptosis induced by QdNOs were explored. Liu et al. (2012) report that MEQ inhibited cell proliferation by arresting cells at the G2/M phase of the cell cycle. Zou et al. (2011) demonstrated that OLA induced cell cycle arrest to the S phase and dose-dependent apoptotic cell death in HepG2 cells through a caspase-9 and -3 dependent mitochondrial pathway. Zhang et al. (2013) found that QCT induced apoptosis in HepG2 cells via activation of caspase, interaction of TNF-α and TNFR1 and modulation of the protein levels of Bid, Bax, and Bcl-2, involving the participation of p53, p38, and JNK. c-MYC-dependent activation of the mitochondrial apoptotic pathway may be associated with QCTinduced toxicity . Dysregulated or excessive autophagy may lead to 'type II programmed cell death, ' which is closely associated with apoptosis. Zhao et al. (2015) showed that OLA-induced autophagy in HepG2 cells is upregulated by Beclin 1 but downregulated by ROS-dependent c-Jun N-terminal protein kinase (JNK).
The organ toxicities of QdNOs have also been studied. Huang et al. (2009) reported the dose-dependent long-term toxicity of MEQ on adrenal gland in male rats, and its mechanism may involve in oxidative stress and steroid hormone biosynthesis pathway. Then, they found the complex interactions of MEQ metabolism, renin-angiotensin-aldosterone system, NADPH oxidase and oxidative stress in response to MEQ-induced tissue toxicity and aldosterone secretion (Huang et al., 2010a). Wang et al. (2015a) estimated the adrenal cell damage induced by QCT and its bidesoxy-QCT (B-QCT) metabolite, and suggested that its toxic effects resulted from N-oxide groups, and its toxic mechanism might involve the interference of the steroid hormone biosynthesis pathway. They also found ROS, the Janus kinase-signal transducer and activator of transcription (JAK/STAT) pathway, suppressors of cytokine signaling and inflammatory cytokines (TNF-α and IL-6) were involved in the liver and spleen toxicities of MEQ (Wang et al., 2011c). Ihsan et al. (2011) verified that MEQ exerted testicular toxicity by causing oxidative stress and steroidal gene expression profiles.

PERSPECTIVES
The versatile activities of QdNOs are close related to their chemical structures. The two N-oxide groups in the quinoxaline ring are necessary for the antibacterial (Montoya et al., 1998;Ortega et al., 1999;Sainz et al., 1999;Zhang, 2012), antitumor , anti-Trypanosoma (Aguirre et al., 2004), and antimalaria  properties of QdNOs. Some reduced forms of QdNOs, in which the two N-oxide groups are absent, still possess biological activities. We assume the large side chains of these reduced quinoxaline compounds contribute more to the activities, based on the fact that in the 2-quinoxalinecarbonitrile 1,4-dioxides, the replacement of (N,Ndialkyl amino) alkyl amino chain with aromatic rigid moieties (anilines and arylpiperazines) in three position held the potency but reduced the hypoxia-selectivity (Ortega et al., 2000). The different side chains probably affect the desoxy and bidesoxy rates of QdNOs (Wang et al., 2015b), which may result in different levels of activities and toxicities.
For QdNO antitumoral compounds, when the electronwithdrawing nature of the 6(7)-substituent increases, the compound is more readily reduced and show the best activity (Monge et al., 1995b;Solano et al., 2007). This also is the case for antimycobacterial , antitrypanosomal (Aguirre et al., 2004;Benitez et al., 2011), and antimalarial  activities. Interestingly, among 2-alkylcarbonyl-3-trifluoromethylquinoxaline 1,4di-N-oxide derivatives, anticancer activity depends on the substituents in the carbonyl group, improving in the order: ethyl < isopropyl < tert-butyl < phenyl-ones (Zarranz et al., 2004), while among 6(7)-substituted quinoxaline-2-carboxylate 1,4-dioxide derivatives, anti-TB activity depends on the substituents in the carboxylate group, improving in the following order: tert-butyl < allyl < 2-methoxyethyl < ethyl < benzyl . We speculate that the activity of alkyl group substitution must be mainly affected by the substituted carbonyl or carboxylate group, but the large benzene ring itself may contribute more to the anticancer or anti-TB activities.
In summary, SAR studies for QdNOs suggest that the position C(2) of quinoxaline should be substituted by electronwithdrawing moieties, preferably by a nitrile or an aroyl, an ester, an N-substituted amide group or a short brominated alkyl chain (Jampilek, 2014). The position C(3) is most often substituted by short alkyl chain (e.g., CH3, CF3) or phenyl. Substitution of C(6) or C(7) of QdNOs by Cl, F, CF 3 , or OCH 3 is also advantageous. When the nitrile moiety is C(2) substituent, substituents in other positions should not be so much electron-withdrawing. It can be concluded that generally lipophilic and mostly electronwithdrawing substituents are preferred.
Although it has been generally accepted that QdNOs are prodrugs, transport forms of agents that are reduced to active metabolites in the body, the modes of actions of QdNOs are still indistinct. The research of the mechanism of actions of QdNOs is in progress, pushing the development of new compounds with more efficient potentials and less harmful effects. Though modes of antitumoral and antibacterial actions of QdNOs are comparatively clear, there are still some puzzles need to be explained. First, the enzyme(s), reducing and activating QdNOs to produce the free radicals, need to be identified. Since free radicals are short-lived, the location of enzyme(s) might indicate the subcellular or molecular target(s) of the drugs. Second, the type of free radical(s) and how they attack DNA or other target(s) should be identified and elucidated. Third, the network of the drug action and the interactions between drugs and organisms should be fully studied. Fortunately, the genomic and proteomic techniques have become effective ways to investigate drug action pathway and discover new drug targets. Furthermore, the research of the mode of action of QdNOs will provide clues to the study of drug resistant and toxicological mechanism, and guide more effective medication of QdNO drugs, e.g., drug combination therapy. For those modes of actions of QdNOs which are still vague, such as antiprotozoal and anticandida activity, researchers are suggested to investigate them according to the already known mechanisms and the SAR of QdNOs combined with new research tools, such as omics technology and 3D-QSAR. A deeper knowledge of the molecular targets of QdNOs is required for the development of new and more specific drugs through a rational design strategy to avoid undesirable side effects.

AUTHOR CONTRIBUTIONS
GC contributed to the design of the review and wrote the review. WS, CC, LG, HH, and ZL revised the review. XW and ZY contributed to the conception of the review.