YbtT is a low-specificity type II thioesterase that maintains production of the metallophore yersiniabactin in pathogenic enterobacteria

Clinical isolates of Yersinia, Klebsiella, and Escherichia coli frequently secrete the small molecule metallophore yersiniabactin (Ybt), which passivates and scavenges transition metals during human infections. YbtT is encoded within the Ybt biosynthetic operon and is critical for full Ybt production in bacteria. However, its biosynthetic function has been unclear because it is not essential for Ybt production by the in vitro reconstituted nonribosomal peptide synthetase/polyketide synthase (NRPS/PKS) pathway. Here, we report the structural and biochemical characterization of YbtT. YbtT structures at 1.4–1.9 Å resolution possess a serine hydrolase catalytic triad and an associated substrate chamber with features similar to those previously reported for low-specificity type II thioesterases (TEIIs). We found that YbtT interacts with the two major Ybt biosynthetic proteins, HMWP1 (high-molecular-weight protein 1) and HMWP2 (high-molecular-weight protein 2), and hydrolyzes a variety of aromatic and acyl groups from their phosphopantetheinylated carrier protein domains. In vivo YbtT titration in uropathogenic E. coli revealed a distinct optimum for Ybt production consistent with a tradeoff between clearing both stalled inhibitory intermediates and productive Ybt precursors from HMWP1 and HMWP2. These results are consistent with a model in which YbtT maintains cellular Ybt biosynthesis by removing nonproductive, inhibitory thioesters that form aberrantly at multiple sites on HMWP1 and HMWP2.


model in which YbtT maintains cellular Ybt biosynthesis by removing nonproductive, inhibitory thioesters that form aberrantly at multiple sites on HMWP1 and HMWP2.
The Yersinia High Pathogenicity Island (HPI) 6 is a virulenceassociated, mobile genetic element found in a number of Gramnegative pathogens, including Yersinia pestis, Klebsiella, and uropathogenic Escherichia coli (UPEC) (1,2). The HPI encodes a hybrid nonribosomal peptide synthetase/polyketide synthase (NRPS/PKS) pathway that produces yersiniabactin (Ybt), a small molecule metallophore initially recognized for its role in Fe(III) uptake and more recently found to engage in passivation and uptake of copper and nickel (3)(4)(5)(6)(7). Related NRPS/PKS pathways in Pseudomonas aeruginosa and Mycobacterium tuberculosis produce structurally related siderophores that also facilitate virulence, making these pathways possible antivirulence targets (8).
NRPSs and PKSs are large multidomain enzymes found in bacteria and fungi that produce an immense variety of chemically diverse secondary, or specialized, metabolites. Often likened to an assembly line, these pathways produce complex products from simple precursor molecules, such as amino acids, aryl acids, and acyl-CoAs. NRPS/PKSs consist of modules that incorporate one precursor into the elongating molecule by a series of condensation and modification reactions that are catalyzed by individual domains. Pathway intermediates are covalently attached to dedicated carrier protein domains (CPs) through a thioester bond to the domain's phosphopantetheine (ppant) post-translational modification. The finished products are released from the final CP by a terminal type I thioesterase domain (TEI) present in HMWP1 (9,10).
The authors declare that they have no conflicts of interest with the contents of this article. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. This article contains Figs. S1-S3. The atomic coordinates and structure factors ( The Ybt biosynthetic pathway has been reconstituted in vitro with four proteins that are encoded in a single operon on the Yersinia HPI: YbtE, HMWP2, HMWP1, and YbtU (3). Biosynthesis begins when the adenylation enzyme YbtE loads salicylic acid onto the first CP of a 230-kDa NRPS called HMWP2. HMWP2 then condenses and cyclizes this salicylic acid unit with 2 units of L-cysteine. The resulting intermediate is next transferred to HMWP1, a 350-kDa hybrid NRPS/PKS that adds a malonyl group from malonyl-CoA, another L-cysteine, and three methyl groups from S-adenosylmethionine (SAM). Next, the reductase YbtU acts on an HMWP1-bound intermediate before Ybt is finally released by the TEI (11). A fifth enzyme, YbtT, is encoded on the biosynthetic operon but is not required for in vitro reconstitution (3). Nevertheless, ybtT deletion mutants of Y. pestis and UPEC exhibit severe Ybt production defects (12)(13)(14). YbtT possesses a conserved Gly-His-Ser-Xaa-Gly thioesterase motif and is homologous to a class of NRPS/PKS-associated enzymes called type II thioesterases. To date, YbtT has not been biochemically characterized.
Type II thioesterases (TEIIs) are a family of NRPS/PKSassociated enzymes to which a variety of functions has been ascribed (15). The most common and well-understood function is to edit NRPSs and PKSs by removing nonelongatable precursors and intermediates from the CP domains. These erroneous molecules on CPs inhibit the assembly line, and regeneration of the free thiol on the ppant by a TEII salvages the NRPS or PKS for further rounds of biosynthesis. TEII-mediated editing is supported by in vivo reduced product levels in TEII deletion mutants and by in vitro biochemical characterizations in which TEIIs were shown to hydrolyze small molecules from thioester substrates (15)(16)(17). Substrate profiling results from these studies are largely consistent with a role in removing nonelongatable "mistakes," although in a few cases in vitro and in vivo data suggest another function, such as end product release (18,19), intermediate release (20), or provision of a key intermediate to a downstream biosynthetic protein (21)(22)(23).
Here, we report the first structural and biochemical characterization of YbtT. Crystal structures of E. coli YbtT reveal the ␣/␤ hydrolase fold and catalytic triad typical of TEIIs. Purified YbtT is an active thioesterase that interacts with HMWP1 and HMWP2. Using LC-MS-based approaches, we found that YbtT cleaves both cognate and noncognate Ybt precursors from acyl-CP domains. YbtT substrate specificity is consistent with an editing activity, the most commonly understood TEII function. The presence of a YbtT activity optimum during in vivo and in vitro Ybt biosynthesis is consistent with a low-specificity model of editing (17) in which futile hydrolysis of productive precursors competes with restorative hydrolysis of inhibitory thioesters. Optimal Ybt production may thus require a favorable balance between these opposing activities.

YbtT and YbtT-S89A crystal structures
To provide a framework for understanding YbtT function, we crystallized WT YbtT and determined the structure at 1.9 Å resolution (Table 1). YbtT adopts an ␣/␤ hydrolase fold with core and lid subdomains (Fig. 1A). The core comprises a six-stranded parallel ␤-sheet encompassed by five ␣-helices, and the lid (residues 125-173) is formed by three ␣-helices inserted between ␤4 and ␤5 of the core. Analysis for structural neighbors using the DALI server (24) revealed that YbtT shares the highest structural homology with TEIIs from other NRPS pathways: RedJ from Streptomyces coelicolor (32% sequence identity) (22) and RifR from Amycolatopsis mediterranei (28% sequence identity) (25). It also shares homology with a human TEII associated with fatty acid biosynthesis (22% sequence identity) (Figs. 2, A and B, and 8A) (26). The active site is composed of the conserved serine hydrolase catalytic triad of Ser-89, His-225, and Asp-197 (Fig. 1B). The catalytic nucleophile, Ser-89, is located adjacent to the lid domain between strand ␤3 and helix ␣3 in the conserved thioesterase motif Gly-His-Ser-Xaa-Gly (Gly-87-His-88 -Ser-89 -Met-90 -Gly-91). The backbone amides of Met-90 and Ala-22 form the oxyanion hole, which serves to stabilize the negative charge that accumulates on the transition state during catalysis. A water molecule occupies the oxyanion hole in the crystal structure. In context with its genetic associations in the Ybt biosynthetic operon, this structure suggests that YbtT functions as a TEII and identifies residues that may be essential for enzymatic activity.
A previous study revealed that alanine substitution mutants in the analogous Y. pestis YbtT Ser and His active-site residues could not complement iron-restricted growth of a ybtT knockout strain (13). To determine whether this defect specifically reflects loss of catalytic residues or derives from more extensive structural distortion, we crystallized and determined the structure of YbtT-S89A to 1.4 Å resolution. The YbtT-S89A mutant adopts an overall structure very similar to WT, with a C␣ RMSD of 0.2 Å (Fig. 1A). A minor difference was noted in the orientation of the catalytic triad residues, with the His-225 YbtT structure and function imidazole ring in the YbtT-S89A structure flipped relative to its WT orientation (Fig. 1, B and C). Thus, the S89A mutation did not result in any gross structural changes to YbtT.

Structural comparison of YbtT active site with other NRPS/PKS TEIIs
Previous structural and enzymatic studies of NRPS TEIIs identified active-site features that may contribute to substrate specificity and enzyme function. Differences in substrate specificity between S. coelicolor RedJ and A. mediterranei RifR have been attributed to residues lining the substrate chamber, which emanate mostly from ␣L1 and ␣L3 of the lid domain. In RedJ, these residues are predominantly hydrophobic and at least partially explain its high specificity for substrates with long acyl chains, as opposed to short ones (22). In contrast, several residues lining this pocket in RifR are occupied by polar residues, a feature that likely contributes to the enzyme's low substrate specificity (25). Given these previous observations, we compared residues lining the substrate chamber of YbtT with those of RedJ and RifR for potential functional insights. We found that a number of the positions occupied by hydrophobic residues in RedJ were occupied by charged residues (Arg or Glu) in YbtT (Fig. 2, A and C, denoted by *). In addition, half of the positions occupied by polar residues in RifR were conserved in YbtT (Fig. 2, B and D, denoted by *). This analysis suggests that the substrate chamber of YbtT more closely resembles that of RifR, an enzyme that functions as a low-specificity editing TEII, suggesting that YbtT may display similar function and substrate promiscuity.

YbtT flexible lid domain is in an open conformation
Previous structural studies of NRPS TEIIs, both in crystals and in solution, suggest that the lid domain is conformationally flexible and thereby controls substrate access to the active site (22,25,27). Indeed, the crystal structures of both RifR and RedJ display multiple lid domain conformations. For RifR, all of the structures display a closed ppant entrance, whereas for RedJ, both open and closed conformations were observed. In YbtT, the ␣L2 helix is noticeably shifted away from the active site, leaving the lid domain in an open conformation (Fig. 3A). Atomic displacement parameter (B-factor) analysis suggests that the YbtT open conformation is likely stabilized by crystal contacts because the B-factors in the lid domain of YbtT are only slightly elevated relative to the core domain. In contrast, the RedJ open conformation displays much higher lid domain B-factors relative to the core (Fig. 3A). In the closed conformations of RedJ and RifR, the lid domains also display similar B-factors to the core, likely attributable to several contacts between the lid and core. Comparative cavity and tunnel analysis in CAVER (28,29) revealed that YbtT has an open ppant entrance that more closely resembles the RedJ open conformation than the blocked ppant entrances of the closed RedJ and  . ␣-Helices that compose the lid subdomain are denoted. The water molecule (red sphere) that occupies the oxyanion hole of the active site is shown for reference. C, comparison of residues that line the active-site pockets of YbtT and RedJ. Hydrophobic residues lie at these positions in RedJ, and in YbtT some are hydrophobic and others are charged (denoted by *). Hydrophobic residues are colored with carbon atoms magenta (YbtT) and cyan (RedJ). Carbon atoms of charged residues are colored dark gray. Other atoms colored by element are as in Fig. 1. D, comparison of polar residues that line the activesite pocket in YbtT and RifR. Polar residues lie at these positions in RifR, and in YbtT some are polar (denoted by *) and some are hydrophobic. Coloring is as in C.

YbtT structure and function
RifR lid subdomain structures (Fig. 3B). The open conformations of YbtT and RedJ displayed the largest active site cavities (756 and 705 Å 3 , respectively) compared with the closed conformations of RedJ and RifR (568 and 682 Å 3 , respectively). Tunnel analysis indicated that the active sites of YbtT and RedJ (open) are accessible from the putative ACP-binding surfaces, whereas RedJ (closed) and RifR are not (Fig. 3B). Together, this analysis indicates that our YbtT structure crystallized in an open conformation that displays external access to the active site.

YbtT is an active hydrolase
To determine whether YbtT possesses hydrolase activity, we incubated purified YbtT with the esterase substrates p-nitrophenyl (p-NP) acetate and p-NP butyrate (Fig. 4A) under steady-state conditions. p-NP esters have been widely used to investigate the in vitro activity and substrate specificity of TEIIs (15,17,30). Hydrolysis was measured continuously for 10 min, during which time less than 10% of the substrate was converted into product, and the appearance of product was linear as a function of time. YbtT hydrolyzed both esterase substrates, confirming its predicted activity as a hydrolase ( Fig. 4B and Table 2). The relative velocity of the reaction was proportional to YbtT concentration with both p-NP acetate and p-NP butyr-ate (Fig. S1). The poor solubility of p-NP butyrate limited the concentrations that could be tested such that only a k cat /K m value could be estimated. YbtT was inactive with the protease substrate glycine p-nitroanilide (Fig. 4, A and C). The catalytic efficiency of YbtT was low for both p-NP substrates (k cat /K m Ͻ50 M Ϫ1 s Ϫ1 ) compared with other TEIIs tested with the same substrates (15). These esterase substrates are structurally dissimilar from YbtT's hypothesized substrate, an acylated CP, which may contribute to the low measured catalytic efficiencies. These model substrate results demonstrate that YbtT is an active hydrolase.

YbtT serine hydrolase catalytic triad is necessary for activity
To test the hypothesized roles of the serine (Ser-89), histidine (His-225), and aspartic acid (Asp-197) catalytic triad residues in hydrolase activity, we constructed single Ala substitutions at each residue and tested them for hydrolase activity with p-NP acetate. YbtT-H225A and YbtT-D197A displayed elution profiles on a gel-filtration column identical to WT and YbtT-S89A, suggesting that these mutations did not grossly alter protein structure (Fig. S2). All three catalytic triad mutants were unable to hydrolyze p-NP acetate (Fig. 4C). Phenylmethylsulfonyl fluoride (PMSF), a general covalent inhibitor of serine hydrolases, inhibits YbtT activity, as has been shown for other

YbtT structure and function
TEIIs (17). These results are consistent with the hypothesized role for the serine hydrolase catalytic triad in YbtT hydrolase activity.

Catalytically active YbtT restores Ybt production in a ybtT deletion mutant
Previous work has demonstrated that ybtT deletion mutants in Y. pestis and UPEC are deficient in Ybt production (12)(13)(14). To further investigate the role of ybtT in UPEC, we used the model UPEC strain UTI89 and expressed plasmid-encoded YbtT in UTI89⌬ybtT and monitored Ybt production by LC-MS/MS. Expression of WT ybtT, but not catalytically inactive ybtT-S89A, partially restored Ybt production in UTI89⌬ybtT Figure 5. YbtT enzymatic activity is important for optimal Ybt production, and it forms a complex with Ybt biosynthetic proteins. A, Ybt levels, quantified by stable isotope dilution and LC-MS/MS, of UTI89, ⌬ybtT, and ⌬ybtT strains complemented with either WT ybtT or catalytically inactive ybtT-S89A cloned into the plasmid pTrcHis2. Strains were grown under iron-restricted conditions. Mean of n ϭ 3 with standard deviation plotted, one-way ANOVA compared with multiple comparison post-test to ⌬ybtT; ***, p Ͻ 0.0001; ns, p Ͼ 0.05. B, SYPRO Orange-stained SDS-polyacrylamide gel of in vitro pulldown experiment with YbtT loaded onto nickel beads and exposed to UTI89⌬fur cellular lysate (lane 4). Additional sets of beads were exposed to either YbtT only (lane 2) or lysate only (lane 3) as negative controls. Molecular mass markers (lane 1) are indicated on the left in kDa. YbtT specifically pulled down two proteins, which were identified as HMWP1 and HMWP2 by LC-MS/MS.   Table 2; mean of n ϭ 3 with standard error plotted. C, time courses of YbtTcatalyzed hydrolysis of the esterase substrate, p-NP acetate, and a protease substrate, glycine p-nitroanilide. All data have been normalized to account for spontaneous background hydrolysis of the substrates in the absence of enzyme. All reactions were performed with 0.5 M enzyme and 1 mM substrate. PMSF treatment was performed at 1 mM. Mean of n ϭ 3 for all conditions except YbtT with glycine p-nitroanilide and YbtT ϩ PMSF with p-NP acetate, where n ϭ 2. Standard deviation plotted.
YbtT structure and function ( Fig. 5A). Differences in Ybt expression were not attributable to growth defects, as all strains grew to the same extent (Fig. S3). These results suggest that although ybtT is not essential for bacterial growth in culture, its serine catalytic triad plays an important role in maximizing Ybt production.

YbtT interacts with HMWP1 and HMWP2
TEIIs hydrolyze thioester bonds that tether small molecules to phosphopantetheinylated CPs. In addition to Ybt biosynthetic enzymes, UTI89 can also express phosphopantetheinylated enzymes involved in enterobactin, fatty acid, and colibactin biosynthesis (31). To identify YbtT-interacting proteins, we performed pulldown assays with recombinantly purified YbtT. Because expression of YbtT and other phosphopantetheinylated enzymes are subject to transcriptional repression by the Fur transcription factor, we hypothesized that YbtT-binding partners would be enriched in prey lysates from a fur deletion mutant. YbtT was therefore exposed to UTI89⌬fur prey lysates, and the eluates were analyzed by SDS-PAGE. Two exceptionally high-molecular-weight proteins were specifically pulled down by YbtT (Fig. 5B). In-gel trypsin digest and mass spectrometric analysis of the peptides revealed these proteins as HMWP1 (350 kDa) and HMWP2 (230 kDa), the major Ybt NRPS/PKS biosynthetic proteins. No other phosphopantetheinylated enzymes were detected in the pulldown. These results are consistent with HMWP1 and HMWP2 as interacting partners of YbtT.

YbtT hydrolyzes ppant thioesters of the HMWP2 aryl carrier domain
We next sought to determine whether YbtT hydrolyzes ppant thioesters on HMWP2. We expressed the aryl carrier protein domain (ArCP), the first phosphopantetheinylated domain of HMWP2 (amino acids 1-100) and the site of salicylic acid loading, and we enzymatically generated both aromatic and aliphatic thioesters. Their identity was confirmed by full scan LC-MS and the ppant ejection method (LC-MS/MS) (32,33). Our panel of test substrates included the salicyl thioester of ArCP, which is a canonical intermediate of Ybt biosynthesis. After incubating each substrate with YbtT for 30 min, we used LC-MS to detect thioester bond hydrolysis. Complete conversion to the empty holo-ArCP product was observed for both aromatic and aliphatic thioesters except anthranilyl-ArCP, which was only partially hydrolyzed (Fig. 6B). Hydrolysis of this chemically diverse ppant thioester series, which includes the canonical salicyl thioester intermediate, is consistent with broad substrate specificity for YbtT.

YbtT hydrolyzes multiple aliphatic thioesters of the HMWP1 acyl carrier domain
To determine whether YbtT also hydrolyzes ppant thioesters on HMWP1, we prepared thioesters of phosphopantetheinylated ACP from the HMWP1 PKS module (amino acids 1813-1995). Here, we were limited to enzymatic synthesis of acetyl-, hexanoyl-, and malonyl-ACP thioesters. YbtT thioesterase activity with each ACP substrate was monitored by LC-MS, as described for ArCP above. YbtT hydrolyzed the short and medium length acyl chains but not malonyl-ACP, a carboxylated acyl chain and a canonical Ybt biosynthetic precursor (Fig.  6C). These results demonstrate that phosphopantetheinylated CPs necessary for both early (HMWP2) and late (HMWP1) Ybt biosynthesis may be YbtT substrates. The reduced activity on malonyl-ACP, relative to the other structurally related thioes-

YbtT structure and function
ters, suggests some degree of YbtT substrate selectivity that acts to spare the canonical precursor from hydrolysis.

YbtT increases Ybt biosynthesis in the presence of a noncanonical precursor
To determine how YbtT directly affects Ybt biosynthesis, we added it to the in vitro reconstituted system developed by Miller et al. (3), where no effect was observed with YbtT at 10-fold excess over the other enzymes. The four essential enzymes for Ybt biosynthesis, YbtE, HMWP2, HMWP1, and YbtU, were purified and combined with the canonical precursors: salicylic acid, L-cysteine, and malonyl-CoA. We found that L-serine, a physiologically plausible, noncanonical precursor that structurally mimics L-cysteine, inhibited Ybt production during in vitro biosynthesis (Fig. 7A). In the presence of competing L-serine, YbtT addition at 10-fold excess over the other enzymes restored the final Ybt yield to that observed in the absence of L-serine. These observations suggest that YbtT edits Ybt biosynthetic intermediates, preventing HMWP2 and HMWP1 inactivation by misincorporated L-serine thioesters that would otherwise irreversibly inhibit the biosynthetic pathway.

YbtT diminishes the rate of Ybt biosynthesis
YbtT slowed the initial rate of Ybt production in both the presence and absence of L-serine (Fig. 7A). Given the observed low substrate specificity of YbtT, we hypothesized that this is attributable to hydrolysis of productive, canonical thioester intermediates. Indeed, when only canonical Ybt precursors were made available to biosynthetic proteins, increasing YbtT concentrations progressively diminished Ybt production (Fig. 7B) to a degree (ϳ20% lower yield with 50 M YbtT) that may not have been evident in the previous analysis (3). These findings suggest that, whereas YbtT may maintain biosyn-thetic activity by removing nonproductive thioesters from HMWP2 and HMWP1, its low-specificity counterproductively can also reduce Ybt production by hydrolyzing canonical Ybt precursors.

High in vivo YbtT expression diminishes Ybt production
Given that increasing YbtT concentrations can diminish in vitro Ybt biosynthesis, we hypothesized that physiological YbtT expression must be optimized to balance hydrolysis of nonproductive thioesters with removal of productive, canonical Ybt precursors from HMWP1 and HMWP2. To test this hypothesis, we genetically complemented UTI89⌬ybtT with a ybtT expression vector under tight control by an arabinose-inducible promoter. YbtT expression in arabinose-supplemented medium was confirmed by Western blotting (data not shown).
In an iron-restricted medium with increasing arabinose concentrations, Ybt levels increased up to an arabinose concentration of 0.4% and decreased at concentrations exceeding 0.4% (Fig. 7C). In contrast, increasing expression of the inactive mutant ybtT S89A did not affect Ybt production (Fig. 7C). These data are consistent with the existence of a cellular YbtT expression level at which competing productive and nonproductive thioester hydrolysis rates lead to optimal Ybt production.

Structural comparison of YbtT to human TEII
Our results indicate that YbtT is required for optimal Ybt production; hence, YbtT inhibition represents a candidate therapeutic strategy for enterobacterial infections. To investigate whether selective inhibition of this bacterial thioesterase is feasible, we carried out initial structural analysis comparing YbtT to human TEII (hTE2) (Fig. 8). The core hydrolase folds for YbtT and hTE2 are very similar, with the major difference being in the conformation of the structurally dynamic lid sub- YbtT structure and function domains (Fig. 8A). The YbtT lid subdomain is in the open configuration, whereas the hTE2 structure is in a closed conformation, with the lid subdomain collapsed onto the active site (26). This difference is indicated by the active-site cavity volumes (YbtT ϭ 756 Å 3 ; hTE2 ϭ 128 Å 3 ) (Fig. 8C). Comparisons of the active-site cavities also indicate several differences in side chain size, chemical nature, and charge between YbtT and hTEII in the putative binding region for competitive inhibitors (Fig. 8B). These observations suggest that YbtT-specific inhibitors with decreased or little potency against hTE2 could be designed.

Discussion
In this study, we report the first structural and biochemical characterization of YbtT, a TEII required for maximal in vivo Ybt production, but unnecessary for maximal in vitro reconstitution. Structural, biochemical, and bacterial culture studies suggest that YbtT helps maintain HMWP1 and HMWP2 activity through a relatively nonspecific thioesterase activity that removes nonproductive, inhibitory thioesters aberrantly incorporated into the Ybt biosynthetic pathway during bacterial growth. YbtT slowly hydrolyzes a variety of productive and erroneous molecules from HMWP1 and HMWP2 CP domains. The resulting free NRPS/PKS ppant thiols can then resume Ybt biosynthesis. This is exemplified by YbtT-mediated restoration of in vitro biosynthesis in the presence of a nonproductive precursor molecule.
Two models have been proposed for TEII editing of NRPS/ PKS pathways, specifically pertaining to the enzyme's substrate specificity at the level of the hydrolyzed small molecule (15,17,25). In the high-specificity model, the TEII specifically recognizes noncognate precursors attached to CPs and hydrolyzes these with high-catalytic efficiency. In the low-specificity model, TEII hydrolyzes both cognate and noncognate precursors from CPs with low-catalytic efficiency. This low-specificity model can be tolerated because flux through the pathway greatly exceeds nonspecific hydrolysis or because cognate precursors are more strongly bound to the NRPS/PKS, shielding them from hydrolysis.
The substrate profiling, in vitro reconstitution, and in vivo overexpression results for YbtT are consistent with the lowspecificity editing model for Ybt biosynthesis. In cultured E. coli, this is suggested by diminished Ybt production when YbtT expression increases beyond an optimal level. During in vitro reconstituted biosynthesis, the low-specificity model is supported by diminished Ybt production upon YbtT addition at Ͼ10-fold excess (100 M) and at longer time points in the presence of only cognate precursors. The catalytic efficiency (k cat /K m ) of YbtT on p-NP acetate and butyrate was quite low, although it is possible that it would be more active on other p-NP ester substrates or other thioester substrates, such as acyl-CPs. YbtT hydrolyzed both productive and nonproductive precursors to Ybt at the ArCP domain, including aromatic and short acyl groups, with only a few exceptions. It was active on salicylic acid, a correct building block of Ybt but not anthranilic acid. It may be that replacing the 2-hydroxyl of salicylic acid with an amino group, as in anthranilic acid, sterically hinders substrate binding by YbtT. The ability of YbtT to hydrolyze 2,3-dihydroxybenzoyl-ArCP (a misincorporated enterobactin biosynthetic intermediate) and to reverse L-serine-mediated inhibition of Ybt biosynthesis is consistent with a role for YbtT in removing physiologically relevant misincorporation errors (34,35). YbtT can thus hydrolyze both productive and nonproductive intermediates at NRPS CP domains.
In contrast to its mode of action at the pathway level, YbtT activity at the PKS CP domain was more consistent with the high-specificity editing model, avoiding malonyl thioester hydrolysis but readily hydrolyzing nonproductive acetyl and hexanoyl thioesters. Similar PKS CP domain specificity was observed in RifR and tylosin TEII, where substrate selectivity was attributed to the absence of a terminal carboxyl group (17,25). Acetyl thioesters may arise by spontaneous decarboxylation of malonyl thioesters. They may also arise during post-

YbtT structure and function
translational modification of NRPS/PKSs through utilization of an acetyl-CoA, rather than a free CoA, by the ppant transferase (34). Nevertheless, low-specificity hydrolysis at other CP domains likely dominates the overall picture during Ybt biosynthesis, and YbtT acts as a low-specificity editor. It is possible that YbtT inactivity on malonyl-CP is a retained, vestigial function of an ancestral TEII that subsequently evolved with the biosynthetic pathway to edit NRPS domains. It remains to be determined whether the ability of YbtT to spare malonyl thioesters is due to the identity of the small molecule, the identity of the CP domain, or a combination of both. Overall, our data suggest that low-specificity editing of a multi-CP domain biosynthetic pathway may conceal high-specificity editing at specific CP domains. It also suggests that the low-specificity model of editing may be a useful strategy for TEIIs that edit hybrid NRPS/PKS pathways with many structurally unique precursors, such as the Ybt biosynthetic pathway.
The crystal structures of YbtT provide a structural explanation for the observed substrate specificity. Previous structural and functional characterizations of RedJ (22) and RifR (25) have noted that the chemical nature of residues in the lid subdomain lining the active-site substrate pocket correlate with substrate specificities. Specifically, this region of RedJ is rich in hydrophobic side chains, which would favor binding to the longchain acyl thioesters that are the preferred substrates for this enzyme (22). In contrast, the same region of RifR contains more polar side chains, and correspondingly, RifR exhibits lower substrate specificity (25). Our YbtT structure, like that of RifR, also contains several polar and charged side chains in this region, consistent with the low substrate specificity we observe. Our structures also reveal that YbtT crystallized with the lid subdomain in an open conformation that may be stabilized by crystal contacts, based on low relative B-factors. Such a conformation in solution would suggest ready access to the active site and, hence, fast enzyme kinetics. However, YbtT displayed slow enzyme kinetics. Previous crystallographic studies have noted multiple conformations for the lid subdomain of TEIIs and TEIs with this fold (22,36). In addition, solution studies by NMR indicate dynamics between the open and closed lid subdomain conformations (27,37). Our kinetics results are supportive of these previous observations, and access to the YbtT active site may be controlled by the conformational dynamics. Additionally, structural comparison to hTE2 shows that the YbtT active-site pocket contains some vastly contrasting features, suggesting that specific inhibition of YbtT might be achieved without potently inhibiting hTE2.
The results of the pulldown with purified YbtT suggest that this TEII interacts with HMWP1 and HMWP2 but not other phosphopantetheinylated proteins in E. coli. There are few reports on the protein-protein interactions between TEIIs and the NRPSs and PKSs on which they act. EntH, the TEII from enterobactin biosynthesis, was shown to interact with an ArCP on one of the pathway's NRPSs in a bacterial two-hybrid experiment (38). The TEII involved with surfactin biosynthesis was shown to bind a CP by NMR, and binding was enhanced when the CP was phosphopantetheinylated and loaded with a small molecule, although it should be noted that the CP that was used did not come from the same biosynthetic pathway (27). Our pulldown does not tell us whether YbtT is primarily interacting with either HMWP1 or HMWP2, and it does not tell us precisely how these proteins interact, so this is an area for future research. Our acyl-CP activity data do suggest that YbtT may bind the ArCP and ACP domains because YbtT could hydrolyze small molecules from both. If TEIIs are scanning their cognate NRPSs and PKSs, then these protein-protein interactions may be very low affinity and difficult to quantify with standard protein-binding techniques (39). The application of MS to study TEII activity on CPs, as applied in this study, is very revealing and further investigation of protein-protein interactions may build upon this approach to understand the molecular mechanisms behind TEII substrate specificity for CP domains.

Bacterial strains, culture conditions, and plasmid construction
The cystitis-derived UPEC model strain UTI89 was used in this study (40,41). The creation of the isogenic mutant UTI89⌬ybtT was previously described (14). UTI89⌬fur was generated using the Lambda Red recombinase method, and the antibiotic cassette was subsequently removed (42). For growth in iron-restricted conditions, strains were first grown for ϳ18 h in Luria-Bertani (LB) broth (BD Biosciences) at 37°C under rotating conditions. These cultures were then inoculated 1:1000 into M63 minimal media supplemented with 0.2% (v/v) glycerol and 10 mg/ml niacin (Sigma) and grown at 37°C under rotating conditions for ϳ18 h. Ampicillin (100 g/ml, Goldbio) was used for plasmid selection, where appropriate. Arabinose was added at inoculation to induce YbtT and YbtT-S89A expression during the 18-h M63 growth in the YbtT overexpression study.
Plasmids for YbtT expression studies and protein purification were constructed using the vector backbones pTrcHis2 (Invitrogen) and pBAD/Myc-His (Invitrogen) and standard molecular biology techniques, including PCR and the restrictions enzymes NcoI and HindIII (New England Biolabs). The S89A and H225A mutants were generated with a standard sitedirected mutagenesis kit (Quikchange, Agilent). pET22b-ArCP and pET22b-YbtU were constructed as described previously (3,43). YbtE was cloned into a pET28a vector with an N-terminal His 6 tag using NdeI and XhoI restriction sites. YbtT, ArCP, YbtU, and YbtE were all cloned from the UTI89 genome. All other plasmids for ACP, Sfp, HMWP1, and HMWP2 expression were kind gifts from Dr. Tim Wencewicz and the Walsh strain collection.

Recombinant protein expression and purification
pTrcHis2 expression plasmids of YbtT, YbtT-S89A, YbtT-D197A, and YbtT-H225A with a C-terminal Myc (EQKLISEEDL)-His 6 tag were expressed in the BL21(DE3) E. coli strain. The expression strains were grown overnight in LB at 37°C, and then back-diluted 1:50 into 3 liters of LB supplemented with 100 g/ml ampicillin. The cultures were grown at 37°C to OD 0.4 -0.6 and then induced with 1 mM isopropyl ␤-D-thiogalactopyranoside (IPTG, Goldbio) for 4 h at 37°C. The cells were resuspended in lysis buffer (20 mM Tris (pH 8.0), 300 mM NaCl, YbtT structure and function 5 mM ␤-mercaptoethanol, 5 mM imidazole) at a ratio of 10 ml of buffer to 1 liter of culture and then stored at Ϫ80°C.
For YbtE purification, the expression vector was transformed into Rosetta D3 E. coli cells. The cells were grown at 37°C to an OD ϳ1.0 and then induced with 250 mM IPTG for 5 h at 30°C. Cells were pelleted and resuspended in 150 mM NaCl, 10 mM Tris (pH 8.5), 50 mM K 2 HPO 4 , 5 mM imidazole, and 10 mM ␤-mercaptoethanol supplemented with lysozyme and DNase. They were lysed by sonication, and the debris was pelleted by centrifugation for 30 min at 16,000 ϫ g. The supernatants were passed over Ni-NTA super flow resin (Qiagen); washed with 300 mM NaCl, 20 mM imidazole, and 50 mM K 2 HPO 4 (pH 8.0); and then eluted in 300 mM NaCl, 250 mM imidazole, and 50 mM K 2 HPO 4 (pH 8.0). Eluted protein was further purified by sizeexclusion chromatography on a Superdex 16/600 S75 column (GE Bio-Sciences) in 20 mM HEPES and 150 mM NaCl with 1 mM dithiothreitol (DTT).
Purification of ArCP and ACP in their apo and holo states was performed as described previously (43,44). Purification of Sfp was also performed as described previously (45). HMWP2, HMWP1, and YbtU were expressed and purified as described previously, with minor modifications (3,46,47).
For HMWP2 purification, the BL21(DE3) E. coli strain expressing pHMWP2.CH8 was grown in 2ϫ YT broth containing 2 mM MgCl 2 at 37°C to OD 0.4, then moved to 16°C, and at OD 0.8 induced with 50 M IPTG for 16 h at 16°C. For HMWP1 purification, the BL21(DE3) E. coli strain expressing pET28b-HMWP1 was grown in LB at 37°C to OD 0.6, and then induced with 75 M IPTG for 16 h at 16°C. For both proteins, the cells were resuspended in lysis buffer (20 mM Tris (pH 8.0), 500 mM NaCl, 5 mM ␤-mercaptoethanol, 1 mM phenylmethylsulfonyl fluoride (PMSF, Sigma), EDTA-free protease inhibitor mixture tablet (SIGMAFAST TM ), 5 mM imidazole) at a ratio of 15 ml of buffer to 1 liter of culture and then stored at Ϫ80°C. The lysates were obtained as described above and incubated in batch with Ni-NTA-agarose beads overnight at 4°C. The beads were then packed into a column and washed with buffer (20 mM Tris (pH 8.0), 500 mM NaCl, 5 mM ␤-mercaptoethanol, 1 mM PMSF) containing 5 mM imidazole and 20 mM imidazole, sequentially; and the protein was eluted with buffer and a gradient from 30 to 100 mM imidazole. All steps were performed at 4°C. The elutions containing HMWP2 or HMWP1 were pooled, concentrated in a 100,000 concentrator (Amicon), and further purified on a Mono Q HR 5/5 anion-exchange column (GE Healthcare) with running buffer A (20 mM Tris (pH 8.0), 5 mM ␤-mercaptoethanol) and running buffer B (20 mM Tris (pH 8.0), 1 M NaCl, 5 mM ␤-mercaptoethanol). Fractions containing the target protein were analyzed, concentrated, dialyzed into storage buffer (25 mM Tris (pH 8.0), 2 mM TCEP, 10% glycerol), flash-frozen, and stored at Ϫ80°C, as described above.

Crystallization, data collection, and structure determination
For crystallization, YbtT and YbtT-S89A were purified as described above and concentrated to 15 or 17 mg/ml for WT and S89A, respectively, in 20 mM HEPES (pH 7.4), 150 mM NaCl, and 1 mM DTT. Crystals of WT and S89A YbtT were obtained by mixing the concentrated protein solution with reservoir solution (either 200 mM tri-sodium citrate, 21% PEG 3350, 0.5% n-dodecyl-␤-D-maltoside (S89A) or 200 mM di-ammonium hydrogen citrate, 20% PEG 3350 (WT)) in a 1:1 ratio. Crystals routinely formed over the course of 2-4 days. Crystals were cryoprotected in reservoir solution supplemented with 20% ethylene glycol and stream-frozen at 100 K. X-ray diffraction data were collected at the Advanced Light Source beamline 4.2.2 at Lawrence Berkeley National Laboratory. Crystals contained one molecule in the asymmetric unit with 43% solvent content. Data were processed using XDS (48). Data collection and refinement statistics are given in Table 1.
The phase problem was solved by molecular replacement in the program PHASER (49) using a polyalanine model of the S. coelicolor thioesterase RedJ (PDB code 3QMW (22)). Following rigid body refinement, the model was built by a combination of auto-rebuilding with PHENIX (50) using RESOLVE (51) and manual building in COOT (52). WT and S89A YbtT were refined against data to 1.9 and 1.4 Å, respectively, using PHENIX. Hydrogens were generated in the final stages of refinement and included as a riding model. The final model for both WT and S89A consisted of residues 1-248. All residues are within the most favored (98.0/97.6%) and additional allowed regions (2.0/2.4%) of a Ramachandran plot. Molprobity scores were 1.08/1.00, and clash scores were 2.86/2.08 for the YbtT structure and function final models of WT/S89A YbtT, respectively. Molecular graphics figures were created using PyMOL. All crystallographic and analysis software used were compiled and distributed by the SBGrid resource (53), and diffraction images were archived with the SBGrid Data Bank (54). The coordinates and experimental structure factors of WT and S89A have been deposited in the RCSB Protein Data Bank and assigned the PDB codes 6BA8 and 6BA9, respectively.

p-NP acetate, p-NP butyrate, and glycine p-nitroanilide activity assays
Enzymatic assays were performed as described previously (17,30). All substrates were from Sigma. Reactions were performed in 1-ml cuvettes at 30°C in 50 mM potassium phosphate (pH 7.4). All reactions contained 5% DMSO (v/v), as it was the vehicle for solubilizing the substrates. YbtT and mutants were used at 0.5 M, and substrates were used at variable concentrations in excess of enzyme. All reaction components except substrate were pre-incubated at 30°C for 5 min to reach temperature, and then the reactions were initiated by adding substrate and mixing. Time courses with p-NP acetateandp-NPbutyratewerecollectedbymonitoringabsorbance at 405 nm, and the product concentration was calculated with ⑀ ϭ 18,000 M Ϫ1 cm Ϫ1 (manufacturer's protocol). Time courses with glycine p-nitroanilide were collected by monitoring absorbance at 405 nm, and the product concentration was calculated with ⑀ ϭ 9500 M Ϫ1 cm Ϫ1 (55). Data were corrected for background spontaneous hydrolysis in the absence of enzyme. Analysis was performed with GraphPad Prism 4 (GraphPad Software). Because p-NP-butyrate had limited solubility and saturation could not be reached, a full Michaelis-Menten analysis could not be performed. The k cat /K m value reported was estimated from the reaction progress curve, as described by Copeland (56).

Ybt detection in bacterial culture
Ybt was detected in spent supernatants by stable isotope dilution after growth under iron-restricted conditions. A 13 C-Ybt internal standard was prepared by growing UTI89 in M63 minimal media supplemented with 0.2% (v/v) 13 C 3 -glycerol (Sigma) and 10 mg/ml niacin, as described previously (2). After bacterial growth, 25 l of 13 C-Ybt internal standard (4) and 6 l of 0.5 M FeCl 3 were added to 1 ml of culture supernatant. The samples were extracted on C18 silica columns (United Chemical Technologies) with 80% methanol (Sigma). Ybt was detected using a multiple reaction monitoring method on a Shimadzu Prominence UFLC-coupled AB Sciex QTrap 4000 mass spectrometer with a Turbo V electrospray ionization source. The LC separation was performed with an Ascentis Express phenyl-hexyl column (100 ϫ 2.1 mm, 2.7 M; Sigma) and the mobile phases: A ϭ HPLC-grade water ϩ 0.1% (v/v) formic acid (Sigma) and B ϭ 90% acetonitrile (EMD Millipore) ϩ 0.1% (v/v) formic acid. All LC and MS settings were the same as described previously (57). Ybt levels were assessed by taking the peak area ratio of the native isotope Fe-Ybt to the internal standard Fe-13 C-Ybt.

Pulldown and protein identification by LC-MS/MS
The UTI89⌬fur lysate was obtained by growing the strain at 37°C overnight; back-diluting it into 1 liter of fresh LB and growing at 37°C shaking for 18 h; pelleting the cells at 8000 rpm for 10 min, 4°C (Beckman Coulter); discarding the supernatant; resuspending the cell pellet in 10 ml of loading buffer (20 mM Tris (pH 8.0), 300 mM NaCl, 5 mM ␤-mercaptoethanol, 10 mM imidazole); and freezing the resuspension at Ϫ80°C. After thawing the cell pellet, it was lysed and clarified as described for protein purification, with the exception that an additional 30 ml of lysis buffer was added after treatment with lysozyme and DNase I but prior to sonication. Three mini-spin columns were prepared with a 50-l bed volume of Ni-NTA beads. Two of these columns were loaded with 0.5 mg of purified YbtT-Myc-His, and one was mock-treated with YbtT storage buffer (described above) to serve as a prey-only control. Loading of YbtT (bait) was performed by rotating the columns for 15 min at 4°C. Excess bait was washed away with 500 l of loading buffer. 1500 l of UTI89⌬fur lysate (prey) was added to one YbtT-loaded column and the mock-treated column. The remaining YbtT-loaded column was treated with 1500 l of loading buffer as a bait-only control. The columns were rotated at 4°C overnight and washed twice with 500 l of YbtT wash buffer (see above), and protein complexes were eluted with 100 l of YbtT elution buffer (see above). The elutions from the three columns were analyzed by SDS-PAGE and SYPRO Orange staining according to the manufacturer's protocol (Sigma).
To identify the proteins pulled down by YbtT, the gel bands at 250 and Ͼ250 kDa were excised and digested with trypsin overnight. The extracted peptides were dried down and redissolved in 10 l of 5% acetonitrile, 0.1% formic acid. 5 l were injected and run on an LTQ-Orbitrap Velos Pro using a 1-h gradient on a Dionex RSLCnano HPLC with a 0.075 ϫ 250-mm C18 Waters CSH130 1.7-m, 130 Å column. All MS/MS samples were analyzed using Mascot (Matrix Science, London, UK; version 2.5.0). Mascot was set up to search the NCBInr database (selected for E. coli, 1,210,071 entries) assuming the digestion enzyme trypsin. Mascot was searched with a fragment ion mass tolerance of 0.60 Da and a parent ion tolerance of 10.0 ppm. Deamidation of asparagine and glutamine, oxidation of methionine, carbamidomethyl of cysteine, and phosphopantetheinylation of serine were specified in Mascot as variable modifications. Scaffold (version Scaffold_4.4.0, Proteome Software Inc., Portland, OR) was used to validate MS/MS-based peptide and protein identifications. Peptide identifications were accepted if they could be established at greater than 80.0% probability by the Scaffold local false discovery rate algorithm. Protein identifications were accepted if they could be established at greater than 99.0% probability and contained at least two identified peptides.

LC-MS-based YbtT assay on ArCP and ACP
Activity assays were performed in a 100-l reaction volume with 20 M acyl-CP substrate and 1 M YbtT (or YbtT storage buffer as a negative control) in 50 mM potassium phosphate (pH 7.4) ϩ 100 mM KCl ϩ 2 mM TCEP. The reactions were incubated for 30 min at 37°C. The samples were diluted 1:2 with 100 mM ammonium bicarbonate and immediately analyzed by LC-MS (instrument details above) in a positive ion mode with an enhanced resolution scan centered on the ϩ12 charge state of the protein.

Ybt in vitro reconstitution
A solution of 5 M YbtE, 5 M HMWP2, 5 M HMWP1, 5 M YbtU, 0.5 M Sfp, 100 mM Tris (pH 8.0), 10 mM MgCl 2 , 1 mM TCEP, 100 M CoASH, 2 mM L-cysteine, 0.75 mM SAM, 0.75 mM NADPH, 1 mM malonyl-CoA, and 2 mM salicylic acid was incubated at 30°C for 0.5 h to allow for phosphopantetheinylation of HMWP2 and HMWP1. In the case of YbtT rescue experiments, 1.5 mM L-cysteine and 30 mM L-serine were used. Then, 50 M YbtT, YbtT-S89A, or YbtT storage buffer (described above) was added. 10 mM ATP was immediately added to initiate the reaction, and reaction mixtures were incubated at 30°C. 20-l aliquots of the reactions were quenched at the designated time point with 10 l of 10% formic acid. To extract Ybt for quantification, 100 l of acetonitrile and 5 l of 13 C-Ybt internal standard were added, and the mixtures were vortexed rigorously for 30 s; centrifuged at 14,000 ϫ g for 15 min at 4°C; and the supernatant was analyzed by LC-MS/MS (see above).