The structural basis for substrate versatility of chloramphenicol acetyltransferase CAT_I

Abstract

Novel antibiotics are needed to overcome the challenge of continually evolving bacterial resistance. This has led to a renewed interest in mechanistic studies of once popular antibiotics like chloramphenicol (CAM). Chloramphenicol acetyltransferases (CATs) are enzymes that covalently modify CAM, rendering it inactive against its target, the ribosome, and thereby causing resistance to CAM. Of the three major types of CAT (CAT_I-III), the CAM-specific CAT_III has been studied extensively. Much less is known about another clinically important type, CAT_I. In addition to inactivating CAM and unlike CAT_III, CAT_I confers resistance to a structurally distinct antibiotic, fusidic acid. The origin of the broader substrate specificity of CAT_I has not been fully elucidated. To understand the substrate binding features of CAT_I, its crystal structures in the unbound (apo) and CAM-bound forms were determined. The analysis of these and previously determined CAT_I-FA and CAT_III-CAM structures revealed interactions responsible for CAT_I binding to its substrates and clarified the broader substrate preference of CAT_I compared to that of CAT_III.

Introduction

Chloramphenicol (CAM) [Fig. 1(A)] is a potent broad-spectrum antibacterial agent. Since its isolation from Streptomyces venezuelae in 1948,¹ CAM was one of the primary agents used to treat many infections in the decades that followed. To date, despite its relatively high toxicity,² CAM is used in many countries because of its affordability and its broad spectrum of activity. In the Western world, CAM is used in treatment of ophthalmic infections and as a last resort in cases of life-threatening brain infections, such as those caused by Neisseria meningitidis, which do not respond to other agents. CAM's ability to cross the blood-brain barrier makes it a potent therapeutic against brain infections. Because of the emergence of pathogens resistant to multiple drugs, CAM is now being reconsidered as a wider-spectrum therapeutic.³

Figure 1.

(A) Chemical structures of CAM and FA. (B) Acetylation of CAM by CATs.

CAM inhibits protein biosynthesis by binding to the 50S subunit of the bacterial ribosome. Recent crystal structures of the 50S subunit of the Escherichia coli and Thermus thermophilus ribosome in complex with CAM revealed that CAM binds to the A-site of the 50S subunit and occupies the binding site for the amino-acyl moiety of the A-site tRNA.^{4, 5} The 3-hydroxyl of CAM is buried in the interface with the ribosome through direct hydrogen bonding, potassium ion-mediated electrostatic interactions, as well as through van der Waals interactions with the RNA phosphosugar backbone.^{4, 5} The 1-hydroxyl of CAM forms hydrogen bonds with RNA bases. Therefore, any modification of the 1-hydroxyl or the 3-hydroxyl of CAM is predicted to be disruptive of CAM-ribosome binding.⁵ Bacterial resistance to CAM is caused by the chromosomally or plasmid-encoded enzyme chloramphenicol acetyltransferase (CAT) that catalyzes the transfer of an acetyl group from acetyl-coenzyme A (AcCoA) to the 3-hydroxyl group of CAM [Fig. 1(B)].⁶ A subsequent slow, non-enzymatic transfer of this acetyl group to the neighboring 1-hydroxyl group allows for a second CAT-catalyzed acetyl transfer from AcCoA onto the 3-hydroxyl group of the same CAM molecule, resulting in a di-acetylated CAM.^{7, 8} However, a single acetylation of CAM is sufficient to abolish its affinity for the ribosome⁹ as explained by the above-mentioned structural observations.^{4, 5}

CAT proteins are historically divided into three types: CAT_I, CAT_II, and CAT_III, with all three types capable of catalyzing the acetyl transfer to CAM to generate 3-O-acetyl-CAM. Genomic analysis of different CAT sequences indicates that the boundaries between these CAT types are not sharp. Members of the CAT_I family are present in many important pathogens such as E. coli, Shigella flexneri, Serratia marcescens, and Salmonella enterica. CAT_I family enzymes display high sequence conservation among themselves (e.g. S. flexneri and S. marcescens CAT_I proteins are 98% and 99% identical to E. coli CAT_I, respectively); however, they display only a modest sequence identity to CAT_II (∼46%) and CAT_III (32–47%) (Fig. 2). The CAT_II family is not easily distinguishable from CAT_III and has been defined historically only through its extreme susceptibility to thiol-modifying agents compared with that of CAT_I and CAT_III.¹¹ There are no obvious additional Cys residues or other sequence features in CAT_II distinguishing it from the CAT_III variants. A slight variation in the pK_a of the Cys31 (in CAT_III nomenclature), the only Cys in vicinity of the substrate or the cosubstrate binding sites, was suggested to be responsible for the difference in reactivity with thiol-modifying agents,¹² although there is no evidence confirming this idea.

Figure 2.

Sequence alignment¹⁰ of CAT_I, CAT_II, and CAT_III enzymes from various bacteria (EC, Escherichia coli; YP, Yersinia pestis; KP, Klebsiella pneumoniae; SM, Serratia marcescens; SF Shigella flexneri; PM, Proteus mirabilis; HI, Haemophilus influenzae; SE, Salmonella enterica; VS, Vibrio sp.; BC, Bacillus cereus; BA, Bacillus anthracis; SP, Streptococcus pneumoniae; EF, Enterococcus faecium; LM, Listeria monocytogenes; SA, Staphylococcus aureus; CD, Clostridium difficile; CB, Clostridium botulinum; CT, Clostridium tetani). Important residues in the active site that are either conserved or non-conserved (vary) between CAT_I and CAT_III are indicated by red and blue circles, respectively. Residues involved in CAM and FA binding are marked by orange and yellow circles, respectively.

The sequence differences between CAT_I and CAT_III include several substitutions in the binding site (Fig. 2), potentially resulting in positional differences of CAM bound to these two proteins. A major consequence of this divergence is reflected in different substrate selectivities of CAT_I and CAT_III. In addition to binding and modifying CAM,⁶ CAT_I, unlike CAT_III, binds a much bulkier antibiotic, fusidic acid (FA).¹³ FA [Fig. 1(A)] is a steroidal antibacterial agent that is used topically or systemically, usually against infections caused by Gram-positive pathogens. CAT_I does not modify FA; rather, it sequesters it through binding by its CAM binding site. This type of mechanism of resistance through sequestration is not uncommon and has been observed for other antibiotics such as bleomycin and thiocoraline.^14–18 Kinetic studies have shown that FA competes with CAM for binding to CAT_I, but not the other CAT types.¹³ Various bile salts and some triphenylmethane dyes also exhibit similar competitive binding to CAT_I, but not to CAT_II/III.^13,19–21

CAT_I plays an important role in antibiotic resistance of many pathogenic bacteria. In addition, CAT_I has been used as a biochemical and proteomic tool in a number of systems^22–26 and as a common CAM-resistance marker encoded in laboratory plasmids. Despite its importance in drug resistance and biotechnology, CAT_I²⁷ has been much less investigated than CAT_III. Structural and biochemical studies of CAT_III^28–34 have been mostly used to understand general features of CAT_I proteins. Despite this progress, differences in the substrate selectivity between CAT_I and CAT_III remain unclear in absence of analysis of CAT_I-CAM and CAT_I-FA structures.

Herein we report crystal structures of CAT_I alone (apo) and in complex with CAM, which explain how CAT_I binds CAM despite differences in its binding site residues from those in CAT_III. Analysis of these structures along with that of the structure of CAT_I in complex with FA (deposited in the Protein Data Bank (PDB) by Roidis and Kokkinidis; accession code: 1Q23³⁵) provides an explanation of the differences in substrate preference among CAT types.

Results

Overall structure of CAT_I

E. coli CAT_I protein was initially co-crystallized with CAM in the P1 space group (Table I). Molecular replacement using either a monomer or trimer of apo-CAT_I (from the structure of a serendipitous complex of the nitric oxide synthase oxygenase domain with CAT_I; PDB code: 1NOC³⁷) as a search model did not yield a solution. This complication likely arose due to the presence of several copies of the protein molecules within a very large unit cell. Further crystallization trials yielded crystals of CAT_I alone in the P2₁ space group with a smaller asymmetric unit. These crystals grew under conditions similar to those of the CAT_I-CAM crystals. Molecular replacement with a CAT_I trimer from the 1NOC entry as a search model, yielded an apo-CAT_I structure with three CAT_I trimers in the asymmetric unit (Table I). This three-trimer structure was then successfully used as a molecular replacement search model to determine the structure of the CAT_I-CAM complex in the P1 crystal form. The asymmetric unit of the P1 crystal form contained six CAT_I-CAM trimers. The crystal structure of the apo-CAT_I and that of the CAT_I-CAM complex were refined to 3.2 Å and 2.9 Å resolution, respectively (Table I).

Table I.

X-ray Diffraction Data Collection and Refinement Statistics for apo-CAT_I and CAT_I-CAM Structures

	Apo-CATI	CATI-CAM
Data collection
Space group	P21	P1
Number of trimers per asymmetric unit	3	6
Unit cell dimensions
a, b, c (Å)	115.2, 102.7, 114.1	107.5, 114.5, 114.5
α, β, γ (°)	90, 119.9, 90	119.9, 97.8, 98.7
Resolution (Å)	50.0–3.2 (3.3–3.2)a	50.0–2.9 (3.0–2.9)a
I/σ	9.3 (2.1)	14.3 (2.3)
Completeness (%)	98.4 (86.1)	87.3 (85.0)
Redundancy	4.3 (3.5)	1.7 (1.7)
Rmerge	0.15 (0.476)	0.06 (0.38)
Number of unique reflections	35,818	82,522
Structure refinement statistics
Resolution (Å)	40.0–3.2	35.0–2.9
R (%)	23.8	24.0
Rfree (%)	30.1	30.9
Bond length deviation (RMSD) from ideal (Å)	0.009	0.006
Bond angle deviation (RMSD) from ideal (°)	1.08	0.907
Ramachadran plot statisticsb
% of residues in most allowed regions	84.9	88.7
% of residues in additional allowed regions	12.9	10.5
% of residues in generously allowed regions	2.3	0.8
% of residues in disallowed regions	0 (0 residues)	0 (0 residues)

^aNumbers in parentheses indicate the values in the highest-resolution shell.

^bIndicates Procheck statistics.³⁶

The structure of CAT_I protein in the apo form reported here is very similar to the structure of apo-CAT_I (PDB code: 1NOC) used for the molecular replacement (Cα RMSD = 0.7 Å) and to another previously deposited structure of apo-CAT_I (PDB code: 1PD5; Cα RMSD = 0.7 Å). Furthermore, the structures of apo-CAT_I are highly similar to the structure of CAT_I in complex with CAM (Cα RMSD ∼ 0.4 Å), suggesting that no major protein conformational changes occur upon CAM binding. Analogously, no major conformational differences were observed for CAT_III in the apo and the CAM-bound forms.²⁸ The overall fold and the oligomeric organization of CAT_I [Fig. 3(A)] resemble those of the previously characterized CAT_III²⁸ variant. Three identical monomers of CAT_I form a trimer with a 3-fold rotational symmetry. The overall trimeric scaffold is formed by three 7-stranded β-sheets, each of which is formed by six strands (β6, β5, β7, β9, β10, and β2) from one monomer and one strand (β8) from another monomer [Fig. 3(B)]. In each monomer, this β-sheet is flanked on the outside by five α-helices and a small three-stranded β-sheet. In the trimeric core, the aliphatic parts of buried Asp157 side-chains (in strand β8) of the three monomers come together to form intimate hydrophobic contacts with each other, while their carboxyl groups are engaged in intricate, asymmetric network of hydrogen bonding interactions with the side-chains of Ser155 and Asn159. The hydrophobic interactions between the Asp157 residues are likely critical for complex stability as this residue is either an Asp or an Asn in most CAT_I/CAT_III proteins. Ser155 could however be substituted by a Gly (Fig. 2). The side-chains of Asp157 residues are distorted so that the carboxyl groups form hydrogen bonds with their own backbone amide NH moieties justifying a weaker conservation of the Ser155. As the side-chains in these β-strands are generally buried away from the solvent, their identities are well conserved among CAT homologs.

Figure 3.

(A) Overall fold and trimeric organization of CAM-bound form. The strands of the β-sheets comprising the central scaffold are marked. The CAM molecule and its molecular surface are shown in blue. (B) A close-up view of the active site at the interface of two monomers in CAT_I-CAM structure. Substrate binding residues of the binding monomer and catalytic monomer are colored in green and orange, respectively. A few highly conserved residues involved in catalysis are marked in red. (C) A representative view of one of the CAT_I active sites. The violet mesh clearly defining the CAM molecule (blue sticks) is Fo-Fc omit electron density generated without the CAM in the model and contoured at 3σ. (D) A zoomed in view of the active site shown in panel C, depicted in a slightly different orientation. (E) A schematic view of residues of the CAT_I active site and their interactions with CAM. Hydrogen bonds and hydrophobic contacts are marked by black dashed lines and the grey hashed lines, respectively. The color coding is consistent with that of panel B. An interactive view is available in the electronic version of the article.PRO2036 Figure 3

Chloramphenicol interactions in the active site

Upon trimerization, the active site is formed at the interface of two β-sheets predominantly with residues from strands β6, β5, β7, β9, and β8 of one monomer (termed as the binding monomer) and strands β2 and β10 of the other (the catalytic monomer). Each trimer contains three identical substrate binding sites [Fig. 3(A)]. The nature of this conserved trimeric assembly strongly suggests that CAT_I monomers either require trimeric assembly for proper folding or, if folded, CAT_I would be catalytically active only in the context of a trimer. Indeed, monomeric mutants of the CAT_III, whose overall fold is highly similar to that of CAT_I, were shown to be catalytically inactive.³⁸ Below, we discuss features of the active site of CAT_I and highlight its differences from that of CAT_III that specify the distinct substrate recognition properties of these two proteins.

In the structure of CAT_I-CAM complex, all three active site pockets of the CAT_I trimer are occupied with CAM molecules [Fig. 3(A)], whose positions are clearly defined in the electron density map [Fig. 3(C,D)]. One of the two monomers forming a binding site (called here the binding monomer) provides the majority of the residues (Cys91, Phe102, Ser104, Phe134, Phe144, Ser146, Leu158, and Val170) involved in binding of the CAM while the other one (called here the catalytic monomer) provides His193, which has been demonstrated to be one of the primary conserved catalytic residues^{7, 39, 40} [Fig. 3(B)]. A few other residues from the catalytic monomer (Phe25 and Cys31) also provide an important CAM-binding surface in the binding pocket. The disposition of the conserved catalytic residues [e.g. His193, Ser146, and Asp197; highlighted in red in Fig. 3(B)] in the CAT_I-CAM structure is highly similar to that observed previously in CAT_III-CAM complex.⁴¹ The position of His193, the likely general base, relative to the bound CAM is identical to its counterpart in CAT_III (His195). The Nε2 atom of His193 is located 2.7 Å away from the 3-hydroxyl of CAM [Fig. 3(E)]. The side-chain of His193 is in a distorted conformation (His193 χ₁ = −150.2° and χ₂ = −41.0° with CAM bound and χ₁ = −142.0° and χ₂ = −32.6° with FA bound). This conformation likely ensures that the imidazole ring is aligned appropriately for abstracting the 3-hydroxyl proton of CAM, promoting a nucleophilic attack by the oxygen on the acetyl group carbonyl of AcCoA,^{29, 30, 34} similarly to the proposed mechanism of the CAT_III variant.⁴⁰ The imidazole ring of His193 is positioned at a proper distance (approximately 3.6 Å) for a face-to-face π-π stacking contact with Phe25. This overall structural arrangement of the catalytic monomer for proper positioning of His193 at the subunit interface is stabilized by several interactions that include a chain of hydrogen bonds between His193, Asp197, Arg18, and the backbone carbonyl oxygen of Ala194 [Fig. 3(E)]. The conserved Ser146 (another catalytically important residue) of CAT_I is positioned similarly in the active site of all three CAT_I structures, and likely stabilizes the transition state oxyanion by donating a hydrogen bond (possibly water-mediated in CAT_I), as proposed for Ser148 of CAT_III.⁴²

The sequence alignment of CAT_I and CAT_III from E. coli demonstrates that of the 20 amino acid residues lining the CAM binding site, 9 are different between the two types (Fig. 2, blue circles). These differences [Ala24 (CAT_III, Phe24), Phe25 (CAT_III, Tyr25), Val28 (CAT_III, Arg28) Ala29 (CAT_III, Leu29), Cys91 (CAT_III, Gln92), Tyr133 (CAT_III, Leu134), Phe144 (CAT_III, Asn146), Phe166 (CAT_III, Tyr168), and Val170 (CAT_III, Ile172)] are significant as they include changes in the size and hydrophobicity of the residues. Remarkably, despite these differences, CAM binding affinities for CAT_I and CAT_III appear to be very similar.⁴³ Furthermore, the superposition of CAT_I-CAM and CAT_III-CAM structures demonstrates that the orientations of the CAM molecule in the active sites of the two proteins are nearly identical. The p-NO₂ group of CAM is solvent exposed when bound in the CAT_I active site and the aromatic ring rests on the hydrophobic surface provided primarily by Leu158 and Val170, as observed in the CAT_III-CAM structure. The side-chains of Leu158, Val160, and Phe166 that line the very bottom of the substrate binding pocket [Fig. 4(C)] are positioned through interactions of the trimeric assembly and show only minor alterations between CAT_I and CAT_III. The dichloroacetyl moiety of CAM closely interacts with Phe134 (Phe135 in CAT_III), likely indicating a strong hydrophobic interaction. A major difference between the CAT_I-CAM and CAT_III-CAM structures is that the residue analogous to Tyr133 of CAT_I is nonpolar (Leu134) in CAT_III. Tyr133 forms a strong hydrogen bond (2.9 Å) with the carbonyl group of CAM. Interestingly, this interaction occurs in place of the interaction of that between the hydroxyl of Tyr25 in CAT_III (Phe25 in CAT_I) and the carbonyl group of CAM, located at an O-O distance of 2.8 Å from each other.

Figure 4.

Close-up views of interactions of FA with active-site residues of (A) CAT_I and (B) CAT_III (quadruple mutant). Schematic views of interactions of FA with (C) CAT_I and (D) CAT_III (quadruple mutant). The color coding is consistent with that of Figure 3 (FA is shown in blue). An interactive view is available in the electronic version of the article.PRO2036 Figure 4

Fusidic acid interactions in the active site

In the CAT_I-FA complex [Fig. 4(A)], FA occupies the same binding site as CAM, which explains its observed behavior as a competitive inhibitor of CAM acetylation.¹³ The differences between the active site residues of CAT_I (as described above) and those of CAT_III, while having little effect on CAM binding,^{41, 44} create a unique surface suitable for binding to FA in CAT_I. In particular, the placement of Ala24 and Ala29 of CAT_I shapes the substrate binding cavity such that the ring D and the 2-methylhex-2-ene “tail” of FA can be accommodated. The hydrophobic steroid ring system of FA makes numerous hydrophobic contacts with active site residues, including Thr93, Phe102, Phe134, Phe144, Ser146, Phe156, Leu158, Val160, Phe166, and Val170 of the binding subunit, as well as Ala24, Phe25, Val28, and Ala29 of the catalytic subunit [Fig. 4(A,C)]. The hydroxyl moiety of ring A of FA closely aligns with the 3-hydroxyl of CAM and forms a very strong hydrogen bond with the Nε2 atom of His193 at a distance of 2.9 Å [Fig. 4(C)]. The hydroxyl group of Tyr133 points inward towards the binding pocket forming a hydrogen bond with the hydroxyl moiety of ring C, the atoms being separated by a distance of about 2.6 Å [Fig. 4(C)], similarly to the interaction between the 3-hydroxyl and Tyr133 in the CAT_I-CAM structure [Fig. 3(E)]. Residues from both the binding monomer (Phe134) and the catalytic monomer (Ala24 and Val28) form a hydrophobic zone near the entrance of the binding pocket in CAT_I that cradle the “tail” section of FA and dictate its conformation [Fig. 4(A,C)].

Valuable insight can be gained by comparing this structure with the previously reported structures of CAT_III in complex with CAM²⁹ and a quadruple mutant of CAT_III in complex with FA.⁴³ In the quadruple mutant of CAT_III, four catalytic pocket residues were mutated (Gln92Cys/Asn146Phe/Tyr168Phe/Ile172Val) to mimic those of CAT_I. This comparison [Fig. 4(B)] indicates a disruption of the FA tail-interacting hydrophobic zone in the quadruple mutant of CAT_III, in particular due to the Ala24Phe and Val28Arg substitutions. The carboxylic acid and acetoxy moieties of ring D are highly solvent exposed when bound to both CAT_I and the CAT_III mutant. The acetoxy group of FA makes a hydrophobic contact with Phe166 in CAT_I (Phe168 in CAT_III) [Fig. 4(C,D)]. Despite the same general protein backbone scaffold of CAT_I and the mutant CAT_III structures, there are several differences in the FA-protein contacts for the two enzymes. Most strikingly, several bulkier residues of CAT_III: Phe24 (Ala24 in CAT_I), Tyr25 (Phe25 in CAT_I), Arg28 (Val28 in CAT_I), and Leu29 (Ala29 in CAT_I) prevent the FA molecule from binding in a position similar to that in CAT_I. A required shift of the FA molecule must not be accommodated due to structural rigidity of wild-type CAT_III resulting in the lack of binding to FA. The mutations of the CAT_III quadruple mutant apparently relax this rigidity and surprisingly accommodate the FA molecule in a very different position from that seen in the CAT_I-FA structure. The hydrophobic “tail” of FA now adopts a very different conformation and gets buried in the disordered loop region (residues 138–141) of the CAT_III mutant. This disorder is very likely due to both the Asn146Phe and the Gln92Cys substitutions in the CAT_III mutant, which cause displacement of the His144 and Thr140 side-chains, respectively, thereby distorting the local backbone. This displacement allows the FA-tail to occupy its altered position in the CAT_III mutant.⁴³ Notably, Tyr25 in the CAT_III quadruple mutant structure (positionally analogous to Phe25 of CAT_I) forms hydrogen bond with FA, at a distance of 2.8 Å to the hydroxyl on the A-ring [Fig. 4(D)], and stabilizes the altered FA orientation. Phe168 and Val172 residues in the CAT_III quadruple mutant make direct hydrophobic contacts with the FA molecule, which explains the contribution of these substitutions to the change in binding affinity to FA.⁴³

We observe no major differences in the backbone conformations near the active site of CAT_I in the structures of apo-CAT_I (PDB code: 3U9B), CAT_I bound to CAM (PDB code: 3U9F), and CAT_I bound to FA (PDB code: 1Q23³⁵). This strongly suggests that CAT_I has evolved to bind multiple ligands, even as large as an FA molecule, without any major protein conformational changes in its backbone.

Discussion

Chloramphenicol acetyltransferase (CAT) is found in many pathogenic bacteria and is often the cause of resistance against chloramphenicol (CAM), once a widely used antibiotic. Of many known CATs, the type-I appears to be the most prevalent. Recent studies have found CAT_I in many pathogenic bacteria. CAT_I has a preference for binding to a variety of substrates; not only does it inactivate CAM but it also binds and sequesters other antibiotics such as FA. A clear understanding of the mechanism of substrate binding by CAT_I is important to address the intriguing question of how CAT proteins from different classes with similar overall structures display different substrate selectivity profiles. In comparison to CAT_III that has been studied almost exclusively, there are only few mechanistic studies that have been performed on CAT_I.

The general fold and the trimeric organization of CAT proteins have been observed in enzymes of primary metabolic pathways in bacteria and eukaryotes, such as pyruvate dehydrogenases^{45, 46} and α-keto acid dehydrogenases.^{47, 48} Therefore, CAT appears to be a product of an ancient gene duplication event, which underwent subsequent specialization through evolution to serve a protective role against toxic compounds such as CAM. The general catalytic mechanism proposed for CAT proteins is based on studies of many such proteins. The residue primarily responsible for catalysis of CAT_I appears to be His193 (His195 in CAT_III).²⁸ This role was proposed based on a previous study in which a mutant CAT_III (His195Tyr) was shown to be devoid of catalytic activity.⁴⁹ Another conserved residue, Ser146, likely stabilizes the oxyanion formed upon an attack on the AcCoA carbonyl carbon by the 3-hydroxyl of CAM. Mutagenesis studies with CAT_III confirmed that Ser148 (Ser146 in CAT_I) is crucial for efficient catalysis.⁴²

The CAT_I protein structure is similar in the apo form and in the CAM- and the FA-bound states, indicating that no major changes in the backbone conformations or in positions of the side-chains occur upon ligand binding. It is quite remarkable that such nearly rigid scaffold is evolutionarily conserved and yet CAT_I can bind chemically diverse substrates. Analysis of the aligned sequences shows that several residues of CAT_I are different than corresponding residues in CAT_III. Our investigation of the CAT_I structures indicates that many of these differences are in residues lining the substrate binding pocket (Fig. 2, blue circles). The most striking differences are concentrated around a small patch of residues (Ala24-Cys31, contributed by the catalytic monomer) that enable the FA molecule to be accommodated only in the pocket of CAT_I. The bulkier residues of CAT_III in this patch would push the FA towards the opposite side of the pocket and consequently disrupt the structure. Interestingly, the flexibility (apparently resulting in the reduced rigidity and increased disorder of the backbone) of the quadruple mutant of CAT_III helps it accommodate the pushed out FA in a different conformation. The “tail” of FA now finds a different hydrophobic pocket to rest in and in turn provides stability to the FA in this altered binding pocket. The mutant CAT_III shows a 200-fold higher affinity to FA than the wild-type CAT_III. However, the quadruple mutant of CAT_III binds FA with a much (4-fold) weaker affinity⁴³ than CAT_I. In addition, the K_m for CAM acetylation by the CAT_III quadruple mutant was somewhat compromised (with respect to either CAT_I or CAT_III) and the value of k_cat was between those for CAT_III and CAT_I. With the direct structural evidence, it is now clear how the tail of FA nests in a hydrophobic pocket and renders CAT_I more energetically favorable to bind to FA. In CAT_III, a similarly positioned FA “tail” would be sterically blocked by Phe24 and Arg28, and it is not surprising that CAT_III does not show affinity towards FA.

Our understanding of CAM's mechanism of action as well as the mechanisms of resistance to it were largely based on biochemical and structural information available on CAM binding to CAT_III and to the bacterial ribosome.⁴ The present structural study augments this knowledge by filling in the gap in our understanding of the recognition of both CAM and FA by CAT_I. CAM has been largely removed from the clinic in the Western world due to its safety concerns, even though cases of extreme toxicity are exceedingly rare. CAM has remained a popular drug in underdeveloped areas due to its low cost and effectiveness against a variety of pathogens. However, as with other antibiotics, development of resistance against CAM is a major obstacle to its power to save lives. The detailed picture of the CAT_I structure is expected to aid in design of inhibitors of CAT enzymes that could re-sensitize CAM-resistant strains. In addition, structure-guided design of CAM analogs could lead to new antibiotics of this class that would be less toxic and more refractory to inactivation by CAT.

Materials and Methods

Expression and purification of CAT_I

CAT_I was expressed in BL21 (DE3)/RIL cells (Stratagene), which harbor a plasmid containing a constitutively expressed CAM resistance gene camR encoding untagged CAT_I protein. The cells were grown in LB medium (200 rpm, 37°C) containing CAM (25 μg/mL) until the culture reached an attenuance of 0.4 at 600 nm. The cells were harvested after an additional 3 h growth. Pelleted cells (centrifugation at 5,000 g, 10 min, 4°C) were resuspended in the lysis buffer [MES pH 6.5 (40 mM), NaCl (200 mM), glycerol (5%), β-mercaptoethanol (2 mM), and EDTA (0.1 mM)] and lysed by sonication. The lysate was clarified by centrifugation at 35,000 × g for 45 min at 4°C. We took advantage of the thermostability of CAT proteins⁵⁰ in purifying CAT_I without an affinity tag. The clarified lysate was heated (75°C, 20 min) and subsequently centrifuged (35,000 × g, 45 min, 4°C) to remove unfolded precipitated proteins. The CAT_I in the soluble fraction was further purified by size-exclusion chromatography on an S-200 column (GE Healthcare) equilibrated with buffer [Tris pH 8.0 (40 mM) and NaCl (100 mM)]. The fractions containing pure CAT_I, as determined by SDS-PAGE, were concentrated to 5 mg/mL using an Amicon Ultra centrifugal filter device (Millipore) and used for crystallization.

Crystallization of CAT_I alone and in complex with CAM

Crystals of CAT_I alone and a complex of CAT_I with CAM (CAT_I-CAM) were grown by vapor diffusion in hanging drops containing 1 μL of protein and 1 μL of the reservoir solution [HEPES (100 mM) pH 7.5 (pH of 1 M stock of HEPES acid was adjusted by adding NaOH), PEG 4000 (20% w/v), isopropanol (10% v/v)] or 1 μL of the reservoir solution containing CAM (1 mM), respectively. Irregularly shaped crystals, 40–60 μm in each of the three dimensions were formed in 7–10 days when incubated at 22°C against the respective reservoir solutions. The crystals were gradually transferred into the reservoir solution containing glycerol (15% v/v) and flash frozen in liquid nitrogen.

Data collection and structure determination

X-ray diffraction data were collected at 100 K at the ×25 beamline of the National Synchrotron Light Source at the Brookhaven National Laboratory. The data were processed with HKL2000.⁵¹ The crystals of apo-CAT_I and CAT_I-CAM complex were in the P2₁ and P1 space groups, respectively. The structures of both apo-CAT_I and CAT_I-CAM complex were determined by molecular replacement with MOLREP⁵² as described in Results. The locations of the CAM molecules in the active sites of CAT_I were clearly identified and positioned in the omit Fo-Fc density and then refined. The structures were iteratively manually built and refined using programs Coot⁵³ and REFMAC,⁵⁴ respectively. The data collection and refinement statistics are given in Table I. The structures of apo-CAT_I and CAT_I-CAM complex were deposited in the Protein Data Bank with accession codes 3U9B and 3U9F, respectively.

Glossary

Abbreviations

AcCoA

acetyl-coenzyme A

CAM

chloramphenicol

CAT

chloramphenicol acetyltransferase

CoA

coenzyme A

FA

fusidic acid.