Legionella pneumophila Carbonic Anhydrases: Underexplored Antibacterial Drug Targets

Carbonic anhydrases (CAs, EC 4.2.1.1) are metalloenzymes which catalyze the hydration of carbon dioxide to bicarbonate and protons. Many pathogenic bacteria encode such enzymes belonging to the α-, β-, and/or γ-CA families. In the last decade, enzymes from some of these pathogens, including Legionella pneumophila, have been cloned and characterized in detail. These enzymes were shown to be efficient catalysts for CO2 hydration, with kcat values in the range of (3.4–8.3) × 105 s−1 and kcat/KM values of (4.7–8.5) × 107 M−1·s−1. In vitro inhibition studies with various classes of inhibitors, such as anions, sulfonamides and sulfamates, were also reported for the two β-CAs from this pathogen, LpCA1 and LpCA2. Inorganic anions were millimolar inhibitors, whereas diethyldithiocarbamate, sulfamate, sulfamide, phenylboronic acid, and phenylarsonic acid were micromolar ones. The best LpCA1 inhibitors were aminobenzolamide and structurally similar sulfonylated aromatic sulfonamides, as well as acetazolamide and ethoxzolamide (KIs in the range of 40.3–90.5 nM). The best LpCA2 inhibitors belonged to the same class of sulfonylated sulfonamides, together with acetazolamide, methazolamide, and dichlorophenamide (KIs in the range of 25.2–88.5 nM). Considering such preliminary results, the two bacterial CAs from this pathogen represent promising yet underexplored targets for obtaining antibacterials devoid of the resistance problems common to most of the clinically used antibiotics, but further studies are needed to validate them in vivo as drug targets.

for at least 6 h, whereas vacuoles which did not contain the bacterium became highly acidic in 15 min after their formation [42]. One of the proteins involved in this process is a vacuolar V-ATPase [40], but as for other organisms [5,8,[18][19][20], the pH regulation is a complex process in which many other proteins are involved, among which the CAs (which generate protons and bicarbonate by the hydration of the highly available substrate CO 2 ) [43,44]. Thus, the investigation of L. pneumophila CAs started to interest researchers due to the potential of such proteins to develop alternative antibiotics [43,44]. Here I shall review the developments in the field of the cloning, characterization, and inhibition of the two such enzymes from this pathogen.

Cloning and Biochemical Properties of LpCA1 and LpCA2, the β-CAs from L. pneumophila
The genome of Legionella pneumophilia subsp. pneumophila strain Philadelphia-1 has been cloned some years ago [41], and inspection of the genome revealed the presence of two putative such enzymes belonging to the β-class, denominated LpCA1 and LpCA2 [43,44]. They have been cloned as glutathione S-transferase (GST)-fusion proteins [43,44]. LpCA1 contains 245 amino acid residues and has a calculated molecular weight of 26.994 kDa, whereas LpCA2 is shorter, comprising 208 amino acid residues, with a theoretical molecular weight of 22.997 kDa.
We have aligned the amino acid sequences of LpCA1 and LpCA2 with those of other such enzymes characterized earlier (kinetically, and in some cases also by means of X-ray crystallography), such as HpyCA from Helicobacter pylori [45], BsuCA213 and BsuCA219 (from Brucella suis [46,47]) and stCA1 and stCA2 from the bacterial pathogen Salmonella typhimurium [48,49] (Figure 1). Data of Figure 1 show that, as all other bacterial β-CAs investigated so far, the two Legionella proteins LpCA1 and LpCA2 possess the amino acid residues crucial in the catalytic cycle of CO 2 conversion to bicarbonate and protons: (i) the zinc coordinating amino acids, constituted by one His and two Cys residues, more precisely Cys90, His143, and Cys146 (LpCA1 numbering system); and (ii) the catalytic dyad constituted by residues Asp92 and Arg94, which is involved in the activation of the zinc-coordinated water molecule which leads to the formation of the nucleophilic, zinc-hydroxide species of the enzyme [43,44].
The measured CO 2 hydrase activity of the two enzymes, LpCA1 and LpCA2 (Table 1), where the activity of LpCA1 and LpCA2 was compared to that of other αand β-CAs from human (h), fungal (Cryptococcus neoformans [50], Candida albicans [51], Saccharomyces cerevisiae [52]), and bacterial (H. pylori [45] and B. suis [46,47]) sources, showed that both these enzymes possess a significant catalytic activity for the physiologic reaction catalyzed by these enzymes, i.e., hydration of carbon dioxide with formation of protons and bicarbonate. The first isoform, LpCA1 showed a moderate degree of activity with a k cat of 3.4ˆ10 5 s´1 and k cat /K m of 4.7ˆ10 7 M´1¨s´1. This activity is in fact comparable to that of other αand β-CAs, such as hCA I (a physiologically relevant human isoform) [18], Can2 from the fungal pathogen C. neoformans [50], and the bacterial pathogenic enzymes HpyCA and BsuCA219, from H. pylori and B. suis, respectively [45,46]. The second isoform, LpCA2, was on the other hand more active than LpCA1, with the folllowing kinetic parameters: k cat of 8.3ˆ10 5 s´1 and k cat /K m of 8.5ˆ10 7 M´1¨s´1 (Table 1). This is a level of activity similar to that of the C. albicans and S. cerevisiae enzymes or the BsuCA213 isoform (all belonging to the β-CA family) and is only slightly lower compared to hCA II, a highly effective catalyst [18] for the CO 2 hydration reaction (Table 1). It may also be observed that acetazolamide (5-acetamido-1,3,4-thiadiazole-2-sulfonamide) AAZ, a clinically used drug, inhibited efficiently the two Legionella enzymes, with K I s in the range of 36-95 nM. The moderately active enzyme (LpCA1) was less sensitive to this compound compared to LpCA2 which showed a higher catalytic activity and also higher affinity for the sulfonamide inhibitor (Table 1).

Isozyme
We have also examined the phylogenetic relationship of the two new bacterial β-CAs, LpCA1 and LpCA2, with that of other such enzymes characterized earlier from various bacterial and fungal pathogens, as well as algae, plants, and Archaea (see Table 2 for the organisms included in the phylogenetic analysis). As seen from the dendrogram of Figure 2, the two Legionella enzymes are highly similar, clustering on their own branch at the upper part of the tree, with the next most similar enzyme being MinCA from the recently identified bacterial pathogen resistant to all known antibiotics Myroides injenesis (this β-CA was not yet cloned and characterized) [44]. On the other hand, the other evolutionarily most related enzyme to LpCA1 and LpCA2 is AbaCA, from Acinetobacter baumanii, another not yet cloned or characterized β-CA. The remaining β-CAs from Figure 2 were less related to LpCA1 and LpCA2 compared to the previously discussed ones, and clustered all together on the lower branches of the tree. This includes two main branches, one constituted only by the archaeal enzyme cab (from Methanobacterium thermoautotrophicum) [53], whereas the remaining,  Table 2. The tree was constructed using the program PhyML 3.0, phylogeny software based on the maximum-likelihood principle [44]. Branch support values are reported at each branch point.
We have also examined the phylogenetic relationship of the two new bacterial β-CAs, LpCA1 and LpCA2, with that of other such enzymes characterized earlier from various bacterial and fungal pathogens, as well as algae, plants, and Archaea (see Table 2 for the organisms included in the phylogenetic analysis). As seen from the dendrogram of Figure 2, the two Legionella enzymes are highly similar, clustering on their own branch at the upper part of the tree, with the next most similar enzyme being MinCA from the recently identified bacterial pathogen resistant to all known antibiotics Myroides injenesis (this β-CA was not yet cloned and characterized) [44]. On the other hand, the other evolutionarily most related enzyme to LpCA1 and LpCA2 is AbaCA, from Acinetobacter baumanii, another not yet cloned or characterized β-CA. The remaining β-CAs from Figure 2 were less related to LpCA1 and LpCA2 compared to the previously discussed ones, and clustered all together on the lower branches of the tree. This includes two main branches, one constituted only by the archaeal enzyme cab (from Methanobacterium thermoautotrophicum) [53], whereas the remaining, rather ramificated, branch includes yeast/fungal (SceCA, DbrCA and SpoCA), bacterial (HpyCA, BsuCA), algal (CspCA and CreCA), and plant enzymes (VraCA, FbiCA, ZmaCA and AthCA), which are all more distantly related to the bacterial CAs from Legionella examined here ( Figure 2).

Sulfonamide/Sulfamate Inhibition Studies of LpCA1 and LpCA2
Sulfonamides and their isosteres (sulfamates, sulfamides, etc.) constitute the main and most investigated family of CAIs, with several drugs used clinically for the last 60 years [18]. Among them are first generation compounds such as acetazolamide AAZ, methazolamide MZA, ethoxzolamide EZA and dichlorophenamide DCP. Dorzolamide DZA and brinzolamide BRZ are topically-acting antiglaucoma agents belonging to the second generation of such drugs [54], whereas benzolamide BZA is an orphan drug belonging to this class of pharmacological agents [18][19][20][21][22]. Topiramate TPM (a sulfamate) and zonisamide ZNS (an aliphatic sulfonamide) are widely used antiepileptic drugs [55]. Other compounds incorporating primary/secondary sulfamoyl moieties such as sulpiride SLP and indisulam IND were also shown to belong to this class of pharmacological agents together with the COX2 selective inhibitors celecoxib CLX and valdecoxib VLX, the antiepileptic sulthiame SLT, the sweetener saccharin SAC, and hydrochlorothiazide HCT, a diuretic agent [56]-see Chart 1. Compound 1-24 are sulfonamides possessing a variety of scaffolds/substitution patterns, which were included in the study [44] in order to explore a wider chemical space for detecting effective inhibitors of these CAs, (Table 3).

Sulfonamide/Sulfamate Inhibition Studies of LpCA1 and LpCA2
Sulfonamides and their isosteres (sulfamates, sulfamides, etc.) constitute the main and most investigated family of CAIs, with several drugs used clinically for the last 60 years [18]. Among them are first generation compounds such as acetazolamide AAZ, methazolamide MZA, ethoxzolamide EZA and dichlorophenamide DCP. Dorzolamide DZA and brinzolamide BRZ are topically-acting antiglaucoma agents belonging to the second generation of such drugs [54], whereas benzolamide BZA is an orphan drug belonging to this class of pharmacological agents [18][19][20][21][22]. Topiramate TPM (a sulfamate) and zonisamide ZNS (an aliphatic sulfonamide) are widely used antiepileptic drugs [55]. Other compounds incorporating primary/secondary sulfamoyl moieties such as sulpiride SLP and indisulam IND were also shown to belong to this class of pharmacological agents together with the COX2 selective inhibitors celecoxib CLX and valdecoxib VLX, the antiepileptic sulthiame SLT, the sweetener saccharin SAC, and hydrochlorothiazide HCT, a diuretic agent [56]-see Chart 1. Compound 1-24 are sulfonamides possessing a variety of scaffolds/substitution patterns, which were included in the study [44] in order to explore a wider chemical space for detecting effective inhibitors of these CAs, (     Table 3 show the inhibition of the two Legionella enzymes, LpCA1 and LpCA2 with this set of sulfonamides/sulfamates. For comparison reasons, the inhibition of the two human (h) offtarget isoforms hCA I and II (belonging to the α-CA family) as well as the bacterial β-CA from the pathogen H. pylori (HpyCA) with this set of 40 compounds, are also presented in Table 3. The following structure-activity relationships (SAR) were observed for the inhibition of LpCA1 and LpCA2 with these compounds:  Table 3). In this case the SAR is very interesting. In addition to the heterocyclic derivatives AAZ and EZA, which are usually highly potent CAIs of most investigated Cas [18], acting thus as promiscuous inhibitors, the potent LpCA1 inhibitors possessed a rather similar structure of elongated, sulfonylated sulfonamide type. Aminobenzolamide 20 (and benzolamide BZA) are the prototype of such compounds, but interestingly, the aromatic compounds 23 and 24 were the best LpCA1 inhibitors. Furthermore, LpCA1 inhibition clearly increased with an increase in the spacer between the two aminobenzenesulfonyl fragments from the molecules. Indeed, 23 and 24 showed K I s of 59.8 and 40.3 nM, respectively (Table 3). (iv) The fast Legionella isoform, LpCA2 was also inhibited by all sulfonamides/sulfamates investigated so far ( Table 3). The inhibition range was however not as wide as for the previous isoform (LpCA1), as the best LpCA2 inhibitor showed a K I of 25.2 nM and the worst one of 933 nM. A small number of the investigated sulfonamides/sulfamates were quite weak LpCA2 inhibitors, among which compound 3, TPM, ZNS, VLX, and HCT (K I s in the range of 745-933 nM, Table 3).

Data of
(v) Many of the investigated sulfonamides were medium potency LpCA2 inhibitors, with K I s in the range of 103-721 nM.  1, 2, and 3-19. (vi) The best LpCA2 inhibitors were 20-24, AAZ, MZA, and DCP (K I s in the range of 25.2-88.5 nM, Table 3). SAR is highly interesting here. Except the clinically used drugs (AAZ, MZA, and DCP) which have not much in common, all the other effective LpCA2 inhibitors possess the same scaffold, of the arylsulfonylated aminosulfonamide type. Thus, aminobenzolamide 20 was 3.3 times more effective as LpCA2 inhibitor compared to benzolamide BZA, whereas for the aromatic componds 22-24, as for LpCA1, the activity increases with the increase of the molecule length, the best LpCA2 inhibitor being 24 (this was also the best LpCA1 inhibitor detected so far). (vii) Except for some of the effective CAIs detected here, which showed a good activity against both LpCA1 and LpCA2 (e.g., 22-24 and AAZ), generally the two isoforms had a rather different affinity for these inhibitors. For example SAC was a very weak LpCA1 inhibitor but a medium potency LpCA2 inhibitor. The same behavior was observed for DCP. Most of the time, these compounds showed an enhanced inhibition of LpCA2 over LpCA1, although several compounds with the reverse profile (e.g., 4, EZA, VLX) were also detected. (viii) The inhibition profiles of the two Legionella enzymes is very different compared to that of other bacterial β-CAs (e.g., HypCA) or the off-target, human isoforms hCA I and II (Table 3). This is of interest in case some of these compounds should be used for targeting the bacterial over the human isofoms in experimental or clinical settings.

Inorganic Anions and Other Small Molecule LpCA1/LpCA2 Inhibitors
Anions and other small molecules (e.g., sulfamide, sulfamate, phenylboronic acid, or phenylarsonic acid) represent a class of well-established CAIs, as they bind to the metal ion from the enzyme active site and impair catalysis [18,34,35]. An inhibition study of the two Legionella enzymes with a range of such inhibitors, including simple and complex inorganic anions as well as sulfamide, sulfamate, phenylboronic acid, and phenylarsonic acid was thus reported [43] (Table 4). The following has been noted regarding inhibition of the two enzymes by the anions and small molecules shown in Table 4. Table 4. Inhibition constants of anionic inhibitors against α-CA isozymes derived from human (hCA II) and the β-CAs from bacteria (H. pylori, HpyCA), and the two enzymes LpCA1 and LpCA2 from L. pneumophila [43].  (i) Perchlorate and tetrafluoroborate did not inhibit the two new β-CAs reported here (K I > 200 mM). Similar results have been observed in most of the CAs examined to date: only HpyCA was effectively inhibited by perchlorate, with a K I of 6.5 mM [45]. Sulfate was also an ineffective LpCA1/LpCA2 inhibitor, with K I values between 77.9-96.5 mM (  Table 4). The best inhibitor in this subseries was tellurate, with K I values of 0.24 and 0.29 mM against LpCA1 and LpCA2, respectively.

Inhibitor
(iv) The best anionic LpCA1 inhibitors were cyanide, azide, hydrogen sulfide, diethyldithiocarbamate, sulfamide, sulfamate, phenylboronic acid, and phenylarsonic acid (K I of 6-94 µM). N,N-diethyldithiocarbamate had a much higher affinity for LpCA1, with a low micromolar value for K I of 6 µM. However, all of the small molecules were low micromolar inhibitors of LpCA2, with K I values ranging between 2 and 13 µM (Table 4). (v) There are net differences in the behavior of the two Legionella enzymes towards the anionic inhibitors investigated here. LpCA1 showed higher affinity for some poisonous metal anions such as cyanide, azide, and hydrogen sulfide, which are around one order of magnitude more potent inhibitors against LpCA1 than LpCA2. However, LpCA2 showed higher affinity for N,N-diethyldithiocarbamate, sulfamide, sulfamate, phenylboronic acid, and phenylarsonic acid compared to LpCA1.
A tentative catalytic/inhibition mechanism of the two LpCA1/2 enzymes is presented in Figure 3, based on crystallographic and kinetic data published for similar β-CAs [5,8,14,15,23,30], as the X-ray crystal structures of the two Legionella enzymes is not known at the moment.  . Proposed catalytic/inhibition mechanisms of LpCAs (exemplified by using LpCA1, and its amino acid residues numbering system) [43]. The catalytic cycle for the CO2 hydration is shown in the clockwise direction, the bicarbonate dehydration in the counterclockwise one. See text for details.
A tentative catalytic/inhibition mechanism of the two LpCA1/2 enzymes is presented in Figure 3, based on crystallographic and kinetic data published for similar β-CAs [5,8,14,15,23,30], as the X-ray crystal structures of the two Legionella enzymes is not known at the moment.
The catalytically active species of the enzyme (A in Figure 3) has a hydroxide ion coordinated to the Zn(II), in addition to the protein ligands (which in Figure 3 are those from LpCA1, i.e., Cys90, His143 and Cys146). This is valid for all CAs belonging to the six different genetic families mentioned in the introduction. In the case of the α-CAs, the zinc hydroxide species is generated through a proton transfer process from the zinc coordinated water to the environment, assisted by the active site . Proposed catalytic/inhibition mechanisms of LpCAs (exemplified by using LpCA1, and its amino acid residues numbering system) [43]. The catalytic cycle for the CO 2 hydration is shown in the clockwise direction, the bicarbonate dehydration in the counterclockwise one. See text for details.
The catalytically active species of the enzyme (A in Figure 3) has a hydroxide ion coordinated to the Zn(II), in addition to the protein ligands (which in Figure 3 are those from LpCA1, i.e., Cys90, His143 and Cys146). This is valid for all CAs belonging to the six different genetic families mentioned in the introduction. In the case of the α-CAs, the zinc hydroxide species is generated through a proton transfer process from the zinc coordinated water to the environment, assisted by the active site residue His64, acting as a proton shuttle [18]. In the β-CAs the activation of the zinc-coordinated water molecule (shown in D in Figure 3) is more complex than for the α-CAs, as the catalytic dyad mentioned earlier (Asp-Arg) takes part in the process. Furthermore, the nature of the amino acid(s) acting as proton shuttles in the β-class enzymes is not completely elucidated [23]. After the nucleophilic attack of the zinc hydroxide species of the enzyme on the bound CO 2 (shown in B) bicarbonate is formed (C in Figure 3) which might be bidentately coordinated to the zinc ion (as in the α-CAs) [59]. Inorganic anions are usually weak ligands of the zinc ion in metalloenzymes, so that species C is probably easily converted to species D, by a substitution reaction of the bicarbonate by means of a water molecule. This leads to liberation of the reaction product (bicarbonate) into solution, with generation of the acidic form of the enzyme, D, which is catalytically inactive. To generate the zinc hydroxide, catalytically effective species A, a proton transfer reaction must occur (Figure 3), probably assisted by an active site residue. In the β-CA from the alga Coccomyxa, it has been hypothesized that there are two residues participating in this process, Tyr88 and His92 [60]. However, neither LpCA1 nor LpCA2 possess a sequence of Tyr and His residues close to each other as the algal enzyme, which leads the nature of the proton shuttling residue in these novel β-CAs unknown at this moment. The inhibition mechanism of LpCA1 and LpCA2 (shown in steps E and F in Figure 3) involves the inhibitor substituting the fourth, non-protein zinc ligand, leading to tetrahedral species, as those shown in E. Anions such as acetate bound to Can2 [50], as well as thiocyanate, azide, and phosphate bound to the Coccomyxa β-CA [60], were shown to possess this inhibition mechanism by means of high resolution X-ray crystallography. An alternative inhibition mechanism has been also reported again for the Coccomyxa β-CA for iodide [60], which was found anchored to the zinc-coordinated water molecule/hydroxide ion, as shown schematically in Figure 3, F. This mechanism was observed for phenols [61], polyamines [62], and sulfocoumarins [63] for the inhibition of α-CAs and may be more general than originally considered [23].

Conclusions
Legionella pneumophila is a Gram-negative bacterium causing Legionnaires' disease and infections caused by it are increasing all over the world. Two β-CAs were recently identified, cloned and purified in this pathogen (LpCA1 and LpCA2) [43,44] and considering the successful inhibition of growth of other bacteria when their CA was inhibited [57,58], the two enzymes were proposed as possible drug targets. LpCA1/LpCA2 efficiently catalyzed CO 2 hydration to bicarbonate and protons, with k cat in the range of (3.4-8.3)ˆ10 5 s´1 and k cat /K M of (4.7-8.5)ˆ10 7 M´1¨s´1, being thus among the very efficient bacterial CAs discovered and characterized so far. They were inhibited by many sulfonamides and sulfamates, the main class of CAIs, with several nanomolar inhibitors detected, although few studies are available in the literature to date [44]. The best LpCA1 inhibitors were aminobenzolamide and structurally similar sulfonylated aromatic sulfonamides, as well as acetazolamide and ethoxzolamide (K I s in the range of 40.3-90.5 nM). The best LpCA2 inhibitors belonged to the same class of sulfonylated sulfonamides, together with acetazolamide, methazolamide, and dichlorophenamide (K I s in the range of 25.2-88.5 nM). Inorganic anions were millimolar inhibitors, whereas diethyldithiocarbamate, sulfamate, sulfamide, phenylboronic acid, and phenylarsonic acid were micromolar ones against both bacterial enzymes. As these enzymes may be involved in pH regulation in the phagosome during Legionella infection, their inhibition may lead to antibacterials with a novel mechanism of action, but in vivo studies are necessary in order to validate them definitively as drug targets.