Molecular Basis of BioJ, a Unique Gatekeeper in Bacterial Biotin Synthesis

Summary

Biotin is an indispensable cofactor in the three domains of life. The unusual virulence factor BioJ of Francisella catalyzes the formation of pimeloyl-ACP, an intermediate in biotin synthesis. Here, we report the 1.58 Å crystal structure of BioJ, the enzymatic activity of which is determined with the in vitro reconstituted reaction and biotin bioassay in vivo. Unlike the paradigm BioH, BioJ displays an atypical α/β-hydrolase fold. A structurally conserved catalytic triad (S151, D248, and H278) of BioJ is functionally defined. A proposed model for BioJ catalysis involves two basic residues-rich cavities, of which cavity-1, rather than cavity-2, binds to the ACP moiety of its physiological substrate, pimeloyl-ACP methyl ester. In summary, this finding provides molecular insights into the BioJ gatekeeper of biotin synthesis.

Graphical Abstract

Highlights

• •BioJ is a distinct gatekeeper of biotin synthesis
• •We report structural and functional definition of BioJ
• •We propose a working model for BioJ-substrate binding in addition to catalytic triad
• •It furthers our understanding how the substrate is gatekept by BioJ in biotin synthesis

Biological Sciences; Microbiology; Structural Biology

Introduction

Biotin (vitamin B7 or H), is an indispensable micronutrient required in three domains of life (Beckett, 2007, Beckett, 2009, Rodionov et al., 2002). As the carboxyl group carrier, this prosthetic cofactor biotin participates into the enzymatic reactions of carboxylation, decarboxylation, and trans-carboxylation in the context of central metabolism, such as type II fatty acid synthesis (Beckett, 2007, Beckett, 2009). Plants and certain microorganisms possess the ability of de novo biotin synthesis, whereas mammals and birds do not (Cronan, 2014). Therefore, it is reasonable that a scavenging/uptake pathway of biotin from food is present in animals or from the inhabiting niche in the biotin auxotrophic microorganisms (Hebbeln et al., 2007) (like Lactococcus [Zhang et al., 2016] and Streptococcus [Ye et al., 2016]). The most of knowledge on biotin synthesis is from studies with the model bacterium Escherichia coli (Cronan, 2014). In general, the de novo pathway involves two critical steps: (1) the synthesis of pimelate, a seven-carbon α, ω-dicarboxylate intermediate (Cronan and Lin, 2010, Lin and Cronan, 2010, Lin et al., 2010); and (2) the assembly of the fused heterocyclic rings of biotin (Rodionov et al., 2002). The latter step of the biotin synthesis route is well known for years (Lin and Cronan, 2010), which is extremely conserved and successively catalyzed by four enzymes, namely, BioF, BioA, BioD, and BioB (Figure 1A). In contrast, the earlier steps by which the precursor, pimeloyl moiety (pimeloyl-CoA or pimeloyl-ACP), is synthesized remained a mystery for around 70 years (Cronan, 2014), until a recent discovery by Lin et al. (Lin and Cronan, 2010, Lin and Cronan, 2012, Lin et al., 2010) that biotin synthesis begins by hijacking a modified type II fatty acid synthesis pathway (FAS II). Unlike acetyl-CoA, a normal primer of FAS II synthesis, the molecule of methyl malonyl-CoA is recruited and elongated by two FAS II cycles, giving the product of methyl pimeloyl-ACP (Lin et al., 2010). The unusual primer, methyl malonyl-CoA (ACP) is produced in the first-committed reaction, i.e., BioC-catalyzed SAM-dependent methylation (Lin and Cronan, 2012). BioH, a prototypical member of pimeloyl-acyl carrier protein (ACP) methyl esterase, removes an extra methyl moiety from pimeloyl-ACP methyl ester (Me-pimeloyl-ACP) to release an intermediate product of pimeloyl-ACP (Agarwal et al., 2012, Sanishvili et al., 2003) (Figure 1A), which directly enters as a dedicated substrate of BioF, into the latter steps of biotin heterocyclic ring formation (Agarwal et al., 2012, Rodionov et al., 2002). The cleavage of Me-pimeloyl-ACP by BioH efficiently prevents its further elongation and functions as a gatekeeper of connecting FAS II with biotin biosynthesis (Lin and Cronan, 2010, Lin et al., 2010). Unlike the representative “BioC-BioH” pathway, many bioC-containing microorganisms lack bioH homologues, raising the possibility that non-homologous isoenzymes are present (Shapiro et al., 2012). As expected, no less than four additional enzymes have been discovered, which consistently belong to the superfamily of α/β hydrolase (Figure 1B) (Bi et al., 2016, Feng et al., 2014, Shapiro et al., 2012). These paradigmatic members separately refer to BioK in Synechococcus (Shapiro et al., 2012), BioG of Haemophilus influenzae (Shapiro et al., 2012, Shi et al., 2016), BioJ exclusively in Francisella (Feng et al., 2014), and BioV restricted to Helicobacter (Bi et al., 2016), respectively.

Figure 1.

A Role of BioJ in Biotin Synthesis and its Phylogeny

(A) Scheme for physiological role of BioJ in biotin biosynthesis pathway. ACP, Acyl carrier protein; Me-Pim-ACP, Methyl pimeloyl-ACP ester; Pim-ACP, pimeloyl-ACP; KAPA, 7-keto-8-aminopelargonic acid; DAPA, 7,8-diaminopelargonic acid; DTB, dethiobiotin; BioF, 7-keto-8-aminopelargonic acid (KAPA) synthase; BioA, 7,8-diaminopelargonic acid aminotransferase; BioD, Dethiobiotin synthase; BioB, Biotin synthase.

(B) Phylogeny of a family of BioJ-containing α/β-hydrolases. The protein sequences of pimeloyl-ACP methyl esterase family were collected from NCBI database. The maximum likelihood-based phylogenetic analysis was conducted using Clustal Omega (https://www.ebi.ac.uk/Tools/msa/clustalo), and the final output is given with MEGA (https://www.megasoftware.net/mega4).

Tularemia (i.e., rabbit fever) is a highly infectious zoonosis (Oyston et al., 2004), whose spread/transmission is mainly dependent on arthropod vectors, such as ticks (Santic et al., 2010). The causative agent of this disease refers to Francisella tularensis, a facultatively intracellular Gram-negative bacterium, which possesses an unusual ability to primarily infect macrophage cells within the host (Celli and Zahrt, 2013). As a category A bioterrorism agent in the United States, F. tularensis seems to be a most virulent bacterium because an inhalation of as few as 10 bacteria is sufficient to result in severe and even fatal disease (Jones et al., 2012). On bacterial uptake by and/or entry into macrophage, F. tularensis exploits multiple strategies to rapidly respond the limited nutrition (e.g., cysteine [Alkhuder et al., 2009] and biotin [Napier et al., 2012]) within the harsh micro-environment of phagocytic cells accompanied with the burst of reactive oxygen species (Celli and Zahrt, 2013). It is a prerequisite for F. tularensis as a cytosolic pathogen to escape into host cytosols for its successful proliferation and survival during the life cycle of infection within hosts (Celli and Zahrt, 2013, Ray et al., 2009). A genome-wide in vivo negative screen originally suggested that bioJ (formerly designated as FTN_0818) of F. tularensis is a genetic determinant essential for both intracellular replication and bacterial infection in mice (Weiss et al., 2007). Then, Napier et al. (2012) reported that the bioJ protein product links biotin biosynthesis to efficient escape from the Francisella-containing phagosome, implying biotin as a nutritional limitation factor during infection. Subsequently, we found that BioJ is a non-homologous isoenzyme of E. coli BioH, the best-studied Me-pimeloyl-ACP carboxyl-esterase (Feng et al., 2014). Consistent with that of BioH (Lin et al., 2010), our data in vitro and in vivo demonstrated that BioJ also functions as a gatekeeper in Francisella biotin synthesis and determines the chain length of the biotin valeryl side chain (Feng et al., 2014). A similar scenario was also seen in the other cytosolic pathogen, Mycobacterium tuberculosis, because removal of bioA, an essential gene of de novo biotin synthesis impairs the establishment and maintenance of its chronic tuberculosis infections (Woong Park et al., 2011). This finding represents metabolic evidence for the link of biotin synthesis to bacterial virulence. However, structural and mechanistic aspects of BioJ remains largely elusive.

In this study, we aimed to close this knowledge gap. Here we report a high-resolution X-ray crystal structure of BioJ at 1.58 Å, illustrating a distinct architecture from the paradigm BioH gatekeeper of biotin synthesis. In addition to the biochemical role played by BioJ in Francisella biotin synthesis, structure-guided functional analyses define a catalytic triad (S151, D248, and H278) and a cavity for binding of its physiological substrate Me-pimeloyl-ACP. In summary, this finding extends our understanding of the biotin synthesis pathway and provides the structural basis for BioJ virulence factor, a potential drug target against the deadly infections with Francisella.

Results

BioJ Is a Gatekeeper in Biotin Synthesis

In total, five types of α/β-hydrolases have been assigned to demethylase of pimeloyl-ACP methyl ester. The phylogeny of these non-homologous isoenzymes suggested that they are evolutionarily distinct (Figure 1B). Unlike the well-studied BioH, which is distributed in ү-proteobacteria, BioJ is restricted to the zoonotic pathogen Francisella and constitutes a unique sub-lineage, Subclade V (Figure 1B). It seems very true that BioJ terminates an alternative route of pimeloyl moiety to be elongated, assuring its entry of this C7 chain into the latter steps of biotin synthesis pathway (Figures 1A and 2A). In vitro enzymatic assays elucidated that BioJ (but not BioZ) cleaves its physiological substrate Me-pimeloyl-ACP into Pimeloyl-ACP (Figure 2B). Similar to BioH (Lin et al., 2010), BioJ is also a promiscuous enzyme in that it removes the methyl moiety from the non-physiological substrates, acyl-ACP methyl esters (like C6 and C8, Figure 2C).

Figure 2.

The Francisella BioJ Catalyzes an Essential Reaction of Biotin Synthesis

(A) Schematic diagram for the biotin synthesis pathway in Francisella. The enzymatic reaction catalyzed by BioJ is underlined by a rectangle with pink background. SAM, S-adenosylmethionine; SAH, S-adenosylhomocysteine; BioC, Malonyl-ACP O-methyltransferase; Me-malonyl ACP, Malonyl-ACP methyl ester; Me-pimeloyl-ACP, Pimeloyl-ACP methyl ester; BioJ, the iso-enzyme of BioH, a pimeloyl-ACP methyl ester carboxylesterase; FASII, type II fatty acid synthesis pathway.

(B) In vitro enzymatic assays reveal that BioJ can hydrolyze pimeloyl-ACP methyl ester to pimeloyl-ACP, rather than BioZ. The BioZ of Agrobacterium is used as a negative control.

(C) Eighteen percent urea PAGE-aided analyses of substrate specificity of BioJ. The plus sign denotes addition of BioJ (or BioZ) enzyme, whereas the minus sign refers to no addition of the protein.

(D) Use of bioassay to visualize bacterial growth of the biotin auxotrophic strain ER90 on minimal media at varied level of biotin.

(E) Scheme of the system of in vitro biotin synthesis.

(F) Biotin bioassay suggests that BioJ enzyme works together with the ΔbioH extract to synthesize biotin (and/or DTB) in the presence of its physiological substrate pimeloyl-ACP methyl ester.

(G) BioJ (rather than BioZ) can restore the ability of ΔbioH in biotin (DTB) synthesis. The bioassay was performed using the reporter strain ER90 (ΔbioF bioC bioD). The red formazan deposit is due to the reduction of the indicator tetrazolium, and this suggests the presence of biotin supply (or synthesized in the in vitro system) conferring bacterial growth of the biotin auxotrophic strain ER90 on the non-permissive condition.

(H) Functional expression of Francisella bioJ rescues its growth of the E. coli bioH mutant on the non-permissive condition of biotin-free medium. E. coli strains used here included the single ΔbioH mutant and a double mutant (ΔbioC/ΔbioH). To test the function of bioJ in vivo, the strains expressing bioJ (or empty vector) were stripped on minimal media with/without 4 nM biotin and maintained at 30°C for 36h. Vec, pBAD322 vector; Ara, arabinose; DTB, Dethiobiotin.

To address the role of BioJ in vivo, an indicator strain-based biotin bioassay was used, in which a visible deposition of a red formazan suggests bacterial viability (Figures 2D and 2E) (Feng et al., 2014). This strain ER90 (ΔbioF ΔbioC ΔbioD) is biotin auxotrophic, whose growth depends on the availability of biotin or its precursor dethiobiotin (DTB) (Figure 2D). The addition of BioJ into the in vitro system of biotin synthesis (Figure 2E) conferred bacterial growth of the indicator strain on the biotin-free, non-permissive condition (Figure 2F), which is consistent with our earlier observation (Feng et al., 2014). Evidently, the growth potency by BioJ is quite similar to scenarios with 5 pmol of biotin (Figure 2D) or its precursor DTB (Figure 2G). However, BioZ cannot work as BioJ does in our assays (Figures 2B and 2G), suggesting a different mechanism. Finally, the plasmid-borne bioJ expression restored the growth of the E. coli ΔbioH mutant but not the double mutant of ΔbioC/ΔbioH (Figure 2H), indicating that BioJ is functionally exchangeable with BioH. Given the fact that BioJ displays BioH-like activity, we therefore believe it is a unique gatekeeper of biotin synthesis (Feng et al., 2014).

Overall Structure of BioJ

The crystals of BioJ diffracted the X-ray to 1.58 Å with the space group of P12₁1 (Table 1) and contained just one copy of BioJ per asymmetric unit (ASU). We solved the structure of BioJ by molecule replacement using the structure of esterase 2 (EST2) from Alicyclobacillus acidocaldarius (PDB:1QZ3) as the search model (De Simone et al., 2004). In the finally refined model, we located 299 amino acids of BioJ except the first methionine and the loop of E213 to S217 (relevant data collection and refinement statistics are presented in Table 1). As a member of the α/β-hydrolase superfamily enzymes, BioJ, BioH, and BioG come from different bacteria and lack significant sequence identity (19% identity shared by BioJ and BioH; 15.7% identity between BioJ and BioG), but all of them contain two domains, a core domain and an α-helical lid domain. The structural comparison gave the root-mean-square deviation (RMSD) of 2.2 Å (165 Ca atoms between BioJ and BioH) and 2.7 Å (125 Ca atoms between BioJ and BioG), respectively. BioH and BioG have a core domain with just a seven-member central β-sheet flanked on either side by α-helices (or loop) and covered by an auxiliary domain (Figures 3A–3D). In contrast, the core domain of BioJ contains eight central β-sheets (Figures 3E and 3F), which is sandwiched in the middle by three helices on each side (h4, h12, and h13 on one side; h5, h6, and h11 on the other side, Figure 3E).

Table 1.

The X-Ray Crystallographic Data Collection and Refinement Statistics

Dataset	BioJ
PDB ID	6K1T
Data Collection
Beamline	BL-17U1, SSRF
Wavelength	0.979
Resolution rangea	51.32–1.58 (1.67–1.58)
Space group	P1211
Cell dimensions
a, b, c (Å)	43.94, 67.31, 55.57
α, β, γ (°)	90, 112.55, 90
Total reflections	1,12,765 (17,680)
Unique reflections	37,190 (5,609)
Multiplicity	3.0
Completeness (%)	91.5 (95.6)
Mean I/sigma(I)	6.7 (3.6)
Wilson B-factor	17.46
R-merge	0.129 (0.745)
R-meas	0.179 (1.022)
R-pim	0.123 (0.695)
CC1/2	0.972 (0.620)
Refinement
Reflections used in refinement	36,992 (3,847)
Reflections used for R-free	1,808 (172)
R-work	0.2067 (0.2591)
R-free	0.2350 (0.2999)
Number of non-hydrogen atoms	2,659
Macromolecules	2,443
Solvent	216
Protein residues	299
RMS (bonds)	0.006
RMS (angles)	0.81
Ramachandran favored (%)	96.61
Ramachandran allowed (%)	3.05
Ramachandran outliers (%)	0.34
Rotamer outliers (%)	0.00
Clash score	4.32
Average B-factor	20.62
Macromolecules	19.97
Solvent	27.99

^aHighest resolution shell is shown in parentheses.

Figure 3.

Comparison of Overall Structures of BioH and BioJ, Two pimeloyl-ACP Methyl Ester Carboxylesterases

Ribbon structure (A) and topological diagram (B) of the E. coli BioH. Overall structure of BioH (A) consists of two domains, the capping domain and the α/β core domain. The helices 4, 5, and 6 of the capping domain are colored green, whereas the α/β core domains are labeled orange in α-helices and cyan in β-sheets. In the topological diagram of BioH (B), an arrow, rectangle and line refer to β-sheet, α-helix, and loop, respectively. The four residues that make salt bridge with ACP are underscored with red triangle in capping domain helix 4 and 5. The catalytic triad (S82, D207, and H235) located in the core domain is featuring with red dots. Overall architecture (C) and topological scheme (D) for the Haemophilus influenzae BioG. The lip domain of BioG is composed of four α-helices (α3–α6) and highlighted by green. In the α/β core domain, there are seven β-sheets (cyan) circled by three α-helices (grayish peach). In the topological diagram of BioG (D), typical catalytic triads (S65, D175, and H200) are labeled with red dots. Overall architecture (E) and topological scheme (F) for the Francisella BioJ. In the lip domain of BioJ, the α-helices 1, 2, and 3 are highlighted by yellow, whereas the α-helices 7, 8, 9, and 10 interacting with ACP are colored green. In the α/β core domain, the eight β-sheets (cyan) are circled with six α-helices (pink). Three basic amino acids are marked by red triangle in the helix 6 plus loop between sheet 6 and helix 7. Typical catalytic triads (S151, D248, and H278) are labeled with red dots. PDB entry is 1M33 for BioH, 6K1T for BioJ, and 5GNG for BioG.

BioH is architecturally similar to that of BioG (Figures 3A–3D), whose RMSD value is 1.078 Å. Meanwhile, the RMSD value of BioJ is 1.110 Å for BioG, and 1.175 Å for BioH, respectively. The major difference is described as follows: four (BioH, Figures 3A and 3B) or three (BioJ, Figures 3E and 3F) of the six α-helices flanking both sides of the core β-sheet are replaced with long loops in BioG (Figures 3C and 3D), thus forming an unusual α/β-hydrolase fold in BioG (Shi et al., 2016). The lid domain of BioJ caps on its core domain by three helices (α1-α3) at the N terminus and four helices (α7-α10) from the C terminus (Figures 3E and 3F). Topological comparison further illustrates obvious difference among the lid domains of BioH (Figure 3B), BioG (Figure 3D), and BioJ (Figure 3F). In relation to the counterpart of BioH that consists of four continuous helices (spanning from K121 to T185, Figure 3B) (Agarwal et al., 2012), BioJ exhibits a bigger lid domain comprising two separate parts (1–42 and185–231, Figure 3F). In fact, the lid domain of BioG is the smallest in that it is of only 57 residues (Y101–Q157, Figure 3D) (Shi et al., 2016).

Functional Validation of Catalytic Triad in BioJ

The prototypical catalytic triad (S82, H235, and D207, Figure 4A) of BioH lies at α10, a loop between β6 and α8 and loop between β7 and α9 (Figure 3B), respectively. Although it is an atypical α/β hydrolase, BioG also has evolved into a similar catalytic triad (S65, D175, and H200, Figure 4B) (Shi et al., 2016). Not surprisingly, the structure of BioJ defines a canonical catalytic triad (S151, D248, and H278) at the interface of the core domain and the auxiliary domain (Figures 3F and 4C), which is analogous to the counterparts in BioH (Figure 4A) (Agarwal et al., 2012) and BioG (Figure 4B) (Shi et al., 2016). Multiple sequence alignment of BioJ homologs showed that they exhibit nearly 80% amino acids identity and consistently possess an identical catalytic triad candidate across different species of Francisella (Figure S1).

Figure 4.

Structural and functional definition of catalytic triad motif in BioJ

(A–C) An enlarged view of the catalytic triad motif in BioH (A), BioG (B), and BioJ (C). Structural snapshots in (A) and (B) were separately generated via 180° rotation of the enlarged views illustrated in Figures 3A, 3C, and 3E.

(D) Genetic complementation reveals that the catalytic triad (S151, D248, and H278) is required for BioJ function. Site-directed mutagenesis was performed to give three bioJ derivatives, namely, S151A, D248A, and H278A. The ΔbioH strains of E. coli expressing bioJ (or its point mutants) were plated on minimal media with/without 4 nM biotin and incubated at 30°C for 36h. Ara, arabinose; Vec, an arabinose-inducible vector of pBAD322.

The conserved steric configuration of catalytic pockets across BioH (Figure 4A), BioG (Figure 4B), and BioJ (Figure 4C) verified our earlier prediction for BioJ active sites guided by structural modeling (Feng et al., 2014). Indeed, the alanine substitution revealed that the point-mutants of BioJ (e.g., S151A and H278A) inefficiently hydrolyze its substrate of Me-pimeloyl-ACP (Figure S2). In the assays of genetic complementation, the wild-type version of bioJ allowed the ΔbioH mutant to grow well on the minimal agar plates lacking biotin. However, none of three point mutants of bioJ (S151A, D248A, and H278A) supported bacterial growth of the ΔbioH mutant on the biotin-free condition, which is almost identical to scenarios with only empty vector introduced (Figure 4D). In addition, the supplementation of exogenous biotin (4 nM) reversed/bypassed functional defection of catalytic triad (Figure 4D). Together, it constitutes a first structure-function proof of catalytic triad in BioJ.

Structural Insight into BioJ-ACP Interplay

Our long-term exploration had no success in obtaining the structure of BioJ complexed with Me-pimeloyl-ACP. However, the availability of the complex structure of the substrate pimeloyl-ACP methyl ester with the point mutant (S82A) of BioH (Figures S3, 5A, and 5B) rendered it possible to gain a putative glimpse of the interplay between the substrate gatekeeper BioJ and the ACP group of its substrate (Figures 5E and 5F). First, four basic residues (namely, R138, R142, R155, and R159) at the capping domain of BioH close to its cavity (Figures 5A and 5B) are mapped to form ionic interactions with the ACP group (Figure S3A). This is due to the formation of salt bridges between the side chain of basic amino acids with positively electronic density area (R138 and R142 in Figure S3B; R155 and R159 in Figure S3C) and the negatively charged ACP-α2 (Q13, D34, and D37 in Figure S3B; E46, I53, and D55 in Figure S3C). Using these criteria, a similar pattern of BioG contacting ACP is also defined by Shi et al. (Shi et al., 2016), which includes the three basic residues, namely, K118, K127, and R132 (Figures 5C and 5D). As expected, we also illustrated that BioJ has a similar positively charged interface with the potential to interact with the ACP moiety (Figures 5E and 5F). This interface is mainly constituted by the three key basic amino acids (K184, K221, and R223) in Cavity-1 at the lid domain close to cavity (Figures 5E and 5F). More importantly, the synergistic roles of the three residues are demonstrated in the assays for growth curves (Figure 5G) and bacterial viability on the agar plates carrying 4 nM biotin (Figure S4A) or lacking biotin (Figure S4B), following the genetic manipulation of bioJ with a series of combined alanine substitution (single, double, and triple). The alignment of BioJ with esterase 2 (EST2) from the thermophilic bacterium Alicyclobacillus acidocaldarius exhibits 30.95% identity (Figure S5). Although their catalytic triads are almost identical (i.e., S155, D252, and H282 in EST2; S151, D248, and H278 in BioJ), the substrate-interacting pocket between BioJ and EST2 varies greatly (Figure S5). In brief, the long hydrophobic tunnel in EST2 for the entry of the hexadecane moiety comprises ten critical residues (N15, S26, S35, L36, G82, V87, S211, L215, F284, and F287, Figure S5) (De Simone et al., 2004). By contrast, the three distinct residues (K184, K221, and R223) are implied to participate in the interplay between BioJ and its substrate pimeloyl-ACP methyl ester (Figure 5). This scenario is also seen with those of BioH (Agarwal et al., 2012) and BioG (Shi et al., 2016) (Figure 5).

Figure 5.

Structure-based search for binding sites of BioJ to its physiological substrate, methyl pimeloyl-ACP

(A–D) Electrostatic surfaces of the BioH (A) and BioG (C). Black dashed frame is used to highlight the positively charged surface of BioH forming salt bridges with the ACP domain of methyl pimeloyl-ACP near the substrate cavity. The PDB entries of BioH and BioG are 4ETW and 5GNG, respectively. An enlarged view of the positively charged basic residues of BioH (B) and BioG (D) that interact with the acidic ACP moiety of methyl pimeloyl-ACP. The side chains of basic amino acids (A and C) are given in a clockwise rotation of 90°.

(E) Surface structure of BioJ. A similar cavity surrounded with positively charged residues is labeled by black frame.

(F) Magnified view of positively charged, side-chain residues in BioJ. It is derived from (E) in the clockwise rotation of 90°. The three putative basic amino acids refer to K184, K221, and R223, respectively. The gradual changing polarity of the protein surface from negative to positive is marked by color from red to blue.

(G) Use of growth curves to evaluate a role played by the three basic residues-containing substrate cavity in the BioJ action. The recipient strain for bioJ and its mutants is STL24 (ΔbioH, Table S1), and plasmid vector is pBAD322 (Table S1). The addition of Arabinose (0.20%) induces the expression of bioJ (and/or its mutants) in STL24 that grows into M9 minimal media with glycerol as sole carbon source. It is given in means ±SD. Three independent experiments were conducted. The single, double, and triple mutants are colored green, magenta, and red.

In particular, the analyses of cavities and molecular docking were applied to infer the residues participating in the BioJ-ACP Interaction, which allowed us to observe the presence of a unique Cavity-2 in BioJ (Figure S6A), but not in canonical BioH or atypical BioG. It indicated that two channels went through the catalytic triad of BioJ, the two ends of which are Cavity-1 and a unique Cavity-2 (Figure S6A). Interactions with residues in Cavity-1 was not observed from the docking result, because the space of channel-1 formed by Cavity-1 is insufficient to accept the substrate pimeloyl-ACP methyl ester. This indicates that the configuration of BioJ may alter during the catalytic process. Unlike those of canonical BioH and atypical BioG, Cavity-2 consists of three continuous α-helices (α1, α2, and α3) from the N terminus of BioJ (Figures 3E and 3F). The channel generated by Cavity-2 is opposite to the other one formed by Cavity-1 (Figure S6A). The analysis of cavities suggests that BioJ creates a bigger channel-2 connected with Cavity-2 than the channel-1(Figure S6A). Of note, both of them are interconnected and go through the catalytic triad. Following the docking process, the analysis of electrostatic distribution further unveiled the occurrence of three basic acids (K29, K40, and R41) in the vicinity of Cavity-2 (Figures S6A and S6B). However, bacterial growth assays along with the alanine substitutions confirm that they are not implicated in the BioJ function (Figure S6C). This suggested that Cavity-2 is not exploited for interacting with the substrate of pimeloyl-ACP methyl ester but might be for expelling product of methyl.

Discussion

In relation to the bio operon, the genomic context of these gatekeeper-encoding genes is divergent, although they are consistently grouped into the superfamily of α/β-hydrolase (Shapiro et al., 2012). Unlike the paradigm E. coli bioH that is not integrated into bio operon, the counterpart of Pseudomonas appears within the bioBFHCD operon (Shapiro et al., 2012). Quite different from bioGC organization in Campylobacter and Haemophilus (Shapiro et al., 2012), the Francisella bioJ acts as a neighbor, but not within its bioBFCD operon (Feng et al., 2014). Thus, it seems likely that these gatekeeper enzymes (synthesizing pimeloyl moiety of biotin) are “wild cards” during the ongoing domestication of the bio operon. Probably, bioJ is a partially domesticated gene in Francisella (Feng et al., 2014). Intriguingly, together with the observation by Napier and coworkers (Napier et al., 2012), we speculated that this gatekeeper BioJ is associated with Francisella virulence (Feng et al., 2014). In addition, we found that efficient utilization of biotin by BplA (FTN_0568), a biotin protein ligase lacking a DNA-binding motif, is essential for the intracellular pathogen F. novicida to survive within macrophages and mice (Feng et al., 2015). Because the disruption of de novo biotin synthesis impairs the initiation and maintenance of chronic infection with Mycobacterium tuberculosis (Ren et al., 2016), it is reasonable to propose biotin as a nutritional (or restricted) virulence factor.

Intriguingly, the physiological requirement of biotin in Francisella (Feng et al., 2014) is much less than that of M. smegmatis (Wei et al., 2018) and Agrobacterium tumefaciens (Feng et al., 2013). That is because a single biotinylated AccB protein is predicted, whereas multiple biotinylated enzymes of M. smegmatis (Wei et al., 2018) and A. tumefaciens (Feng et al., 2013) are experimentally verified. Clearly, non-homologous isoenzymes of BioJ gatekeeper of biotin synthesis are distributed within a family of diversified pathogens (Figure 1), which is partially featuring with BioH of Salmonella and Vibrio, BioG of Campylobacter and Neisseria, and BioV of Helicobacter. Subsequently, the question to ask is if these isoenzymes participate in bacterial virulence. Probably it might consolidate the aforementioned hypothesis in view of a common role played by biotin metabolism in bacterial colonization, competitive infection, and survival within hosts.

The biochemical and structural data reported here represent the molecular basis for this atypical BioJ gatekeeper in biotin synthesis. A working model for BioJ action is proposed here, which is described with three putative steps as follows: (1) the recognition of BioJ to Me-pimeloyl-ACP; (2) the removal of methyl group from Me-pimeloyl-ACP by catalytic triad; and (3) configuration change-dependent exclusion of pimeloyl-ACP product from Cavity-1 and release of methyl group via Cavity-2 (Figure S6A). As for the final step, the core size of Cavity-1 presumably becomes smaller to extrude pimeloyl-ACP, whereas the core size of Cavity-2 is enlarged for the release of the methyl product. Because its overall architecture is quite different from those of BioH (Agarwal et al., 2012) and BioG (Shi et al., 2016), the high-resolution structure of BioJ we solved (Figure 3) extends our mechanistic understanding of the functional unification across BioH-like esterase within the distinct evolutionary placement (Figure 1). The fact that the conserved catalytic triad is shared among BioJ (Figure 4C), BioH (Figure 4A) (Agarwal et al., 2012), and BioG (Figure 4B) (Shi et al., 2016) suggests a promising anti-virulence drug target. Similarly, the predictive substrate-binding mechanism of BioJ offered us an avenue for the discovery of leading anti-bacterial drugs. Taken together, large-scale screen and/or computational design of small molecule inhibitors targeting these two motifs of the bio gatekeeper represent a new approach for combating the deadly infections with multidrug-resistant superbugs.

Limitations of the Study

Although a complex structure of BioH with Me-Pim-ACP provides direct evidence for its binding to physiological substrate (Agarwal et al., 2012), we have no success in securing a complex structure of BioJ with Me-Pim-ACP. Therefore, it is probable that cavity analyses (and molecular docking)-based insights are incomplete, with respect to an interplay between BioJ and its substrate. Also, in vitro enzymatic actions of all the mutant protein with defection in substrate binding are needed to consolidate the scenario seen with the assays of genetic complementation. In the near future, the picture of gatekeeper in biotin synthesis is relatively complete upon the availability of the BioV structure.

Methods

All methods can be found in the accompanying Transparent Methods supplemental file.

Published: September 27, 2019

Data and Code Availability

The accession number for the atomic coordinates of BioJ protein reported in this paper is PDB: 6K1T.