Identification of a stable complex between a [NiFe]‐hydrogenase catalytic subunit and its maturation protease

Salmonella enterica serovar Typhimurium has the ability to use molecular hydrogen as a respiratory electron donor. This is facilitated by three [NiFe]‐hydrogenases termed Hyd‐1, Hyd‐2, and Hyd‐5. Hyd‐1 and Hyd‐5 are homologous oxygen‐tolerant [NiFe]‐hydrogenases. A critical step in the biosynthesis of a [NiFe]‐hydrogenase is the proteolytic processing of the catalytic subunit. In this work, the role of the maturation protease encoded within the Hyd‐5 operon, HydD, was found to be partially complemented by the maturation protease encoded in the Hyd‐1 operon, HyaD. In addition, both maturation proteases were shown to form stable complexes, in vivo and in vitro, with the catalytic subunit of Hyd‐5. The protein–protein interactions were not detectable in a strain that could not make the enzyme metallocofactor.

Biosynthesis of the NiFe(CN) 2 CO active site cofactor, and its insertion into the precursor of the large subunit, requires the action of specialist accessory proteins [11]. Once the [NiFe] cofactor is in place within the large subunit it is proteolytically processed at its C terminus by another accessory protein, a specific endopeptidase, the action of which renders the cofactor loading pathway irreversible [12]. Proteolytic processing of [NiFe]-hydrogenases is critical and without it such enzymes remain completely inactive. Genetic and biochemical studies have suggested that each individual hydrogenase isoenzyme is processed by its own dedicated endopeptidase [13][14][15]. These accessory proteins typically remove an approximately 15 amino acid residue peptide ('assembly peptide') from the C terminus of the large subunit following a conserved histidine residue within the consensus motif DPCXXCXXH, the cysteines of which provide two of the ligands needed for co-ordination of the [NiFe] cofactor.
The crystal structures of processing proteases revealed metal-binding enzymes. E. coli HybD, for example, contains cadmium ions in the structure [16], and E. coli HycI has a calcium ion binding site [17,18]. Proteolytic cleavage of the large subunit by the endopeptidase has been considered 'nickel dependent' in so far as the [NiFe] cofactor must be loaded into the large subunit before proteolysis [19]. Moreover, because purified proteases do not contain metal, and metal binding has been considered a crystallographic artifact [11,16,17], it has been hypothesized that the metal-binding motif of the endopeptidases is used to recognize the presence of the complete [NiFe] cofactor within the large subunit precursor [11].
In this work, the S. enterica Hyd-5 system has been employed to further understand the relationship between a [NiFe]-hydrogenase large subunit and its cognate maturation endopeptidase. For Hyd-5, the maturation protease was predicted to be encoded by the hydD gene. Here, deletion of hydD is shown to hamper Hyd-5 biosynthesis, but HydD function can be partially rescued by the HyaD protein encoded within the operon for Hyd-1. Furthermore, new evidence is provided that HydD, and HyaD, can form unexpectedly stable complexes with the large subunit of Hyd-5. Surprisingly, genetic experiments suggest the interactions do not require the presence of the C-terminal extension on HydB. Indeed, stable HydB-HydD and HydB-HyaD complexes can be isolated where the Cterminal extension of HydB remains completely unprocessed. These data suggest more elaborate roles for these important accessory proteins beyond transient proteolysis.
Plasmids constructed and studied in this work are listed in Table S2. For construction of plasmids pBADHyaD and pBADHydD, the hyaD and hydD genes were amplified by PCR and cloned in vector pBAD24 [20] using EcoRI-SalI. Plasmids for bacterial two-hybrid analysis were constructed using the pUT18 and pT25 vectors [21]. The hyaD and hydD genes were PCR amplified, minus stop codons, and cloned into pUT18 using BamHI-EcoRI sites. The hydB gene was PCR-amplified and cloned into pT25 using PstI-BamHI sites. The first truncated form, HydB T1 , was amplified and cloned into pT25 vector using BamHI-SmaI. The second truncated form, HydB T2 , was cloned as a BamHI and KpnI fragment into the pT25 vector.
For copurification experiments, vectors were designed to allow HydB to be overproduced along with either HyaD or HydD His-tagged variants. The hydB gene, including the initiation codon but with the native UGA stop codon replaced by two consecutive UAA stop codons to prevent read-through, was amplified using primers described in Table S3. The forward primer included an artificial ribosome-binding site (RBS) and six-base spacer to the initiation codon. The PCR product was digested with EcoRI and SphI and cloned into pQE80 resulting in pQE80-HydB. Then, hyaD or hydD were PCR-amplified using forward primers that included an artificial RBS and six-base spacer to the initiation codon. The reverse primers included the sequence for the hexahistidine affinity tag. PCR products were separately cloned into the pQE80-HydB vector using HindIII-PstI sites, resulting in pQE80-HydB-HyaD HIS or pQE80-HydB-HydD HIS vectors.

Mutant strain construction
Plasmids constructed in this work are listed in Table S2. Oligonucleotides used as primers are listed in Table S3. Inframe deletions of chromosomal hypD, hydD, and hyaD genes were constructed using pMAK705 [22]. DNA covering approximately 500 bp upstream of the gene to be deleted, including the translation start site, was PCR amplified using genomic DNA from the LT2a strain as a template and cloned into pBluescript-II KS + using BamHI-EcoRI sites. Next, DNA covering 500 bp downstream of the gene, including the terminator site, was amplified and cloned into the new vector using EcoRI-HindIII sites. The deletion alleles were moved into the pMAK705 vector as BamHI/ HindIII fragments and onto the chromosome of the S. enterica LB03 strain by homologous recombination [22].
The construction of LB03 derivative strains (DhypD, DhyaD, DhydD, and DhyaD DhydD) encoding a hemagglutinin (HA) tag at the N terminus of HydB was performed as follows. A first round of PCR reactions amplified DNA covering the up and downstream of the hydB gene using primers 1HA tag HydB _SacI_FW/2HA tag HydB _RV and 3HA tag HydB _FW/4HA tag HydB_XbaI_RV primers, respectively. The two PCR products were linked via their overlapping sequence coding for the HA tag (YPYDVPDYA) fused in frame in the corresponding site of hydB using primers 1HA tag HydB _SacI_FW/4HA tag HydB_XbaI_RV. PCR products were cloned in pBluescript-II KS + as SacI-XbaI fragments, moved to pMAK705 with SacI/XbaI and the HA-tag sequence incorporated into the chromosome by homologous recombination [22].
For E. coli MAE01 (MG1655 DcyaA::Apra R ) 500 bp upstream of cyaA was amplified and cloned into pKS vector as an EcoRI-XbaI fragment. Next, 500 bp downstream was amplified and cloned into the new vector using XbaI-Hin-dIII. Then, the deletion allele was subcloned into pMAK705 leading to yield pFGM1. An apramycin resistance cassette was amplified from pIJ773 [23], digested with SpeI, and cloned into XbaI-digested pFGM1 to give pFGM2 and the deletion was moved onto the MG1655 chromosome [22]. The E. coli MG1655 (DcyaA::Apra R DhypF::Kan R ) strain (MAE02) was constructed by P1 transduction the DhypF:: Kan allele from JW5433 [24] into the MAE01.

Hydrogenase activity assays
Hydrogenase activity was measured by H 2 -dependent reduction of benzyl viologen [25]. Starter cultures were grown aerobically in low salt LB at 37°C. A 1 : 1000 dilution was made in 20 mL of fresh media and arabinose, when necessary, was added to 0.02% (w/v) final concentration. Cultures were incubated at 37°C for 16 h anaerobically before washed cell pellets were suspended in 500 lL 50 mM Tris/HCl (pH 7.5). Protein concentration of cell suspensions was determined by a modified Lowry method [26] using the DC Protein Assay Kit (BioRad, Hercules, CA, USA).

Rocket immunoelectrophoresis
Rocket immunoelectrophoresis was performed as described [8]. Periplasmic fractions were obtained from 250 mL of bacterial cultures grown anaerobically in low-salt LB at 37°C for 16 h and prepared by a lysozyme/EDTA method [8].

Bacterial two-hybrid interaction assay
Bacterial two-hybrid analysis was performed as described [27] except bacterial cultures were grown anaerobically at 30°C for 16 h in static Hungate tubes filled with 10 mL of LB media. Protein-protein interactions were quantified by b-galactosidase activity assays [28].

Protein purification
Protein purification was carried using 5 L cultures of E. coli FTD147 (pQE80HydB-HydD HIS ) or FTD147 (pQE80HydB-HyaD HIS ) grown anaerobically at 37°C in Duran bottles filled to the top with low salt LB media. Protein production was induced with 2 mM (final concentration) of isopropyl b-D-thiogalactopyranoside at the outset. After 16 h, cells were harvested and suspended in Buffer A (50 mM Tris/HCl, pH 7.5, 150 mM NaCl, and 75 mM imidazole) containing protease inhibitors (Complete-mini, EDTA-free; Roche, Basel, Switzerland), 50 lgÁmL À1 of lysozyme and 10 lgÁmL À1 of DNase I. Cells were broken using Emulsiflex-C3 homogenizer and soluble fractions prepared by ultracentrifugation at 134 000 g at 4°C. The soluble extract was loaded onto a 5 mL His Trap HP column (GE Healthcare, Little Chalfont, Buckinghamshire, UK) equilibrated with buffer A. Bound proteins were eluted with a 50-mL linear gradient of the same buffer containing 500 mM imidazole. Fractions containing the His-tagged proteins were pooled and concentrated in a Vivaspin 20 (Millipore Inc., Billerica, MA, USA) filtration device (10 kDa molecular weight cutoff).

Protein analytical methods
Proteins were separated by standard SDS/PAGE [29] or Bis-Tris gels [30]. Proteins were visualized in-gel using Instant Blue stain (Expedeon, San Diego, CA, USA) or, if necessary, were transferred to nitrocellulose [31]. Western immunoblots were performed using an HRP-conjugated anti-hexaHis monoclonal (1 : 10 000 dilution; Abcam, Cambridge, UK) and an anti-HA Epitope Tag Monoclonal (1 : 2000 dilution; Sigma-Aldrich, St. Louis, MO, USA). Blots were developed using a Clarity Western ECL Substrate Kit (BioRad Laboratories) and data collected with the GeneGNOME camera system (Syngene, Cambridge, UK). For protein identification by tryptic peptide mass fingerprinting, protein samples of interest were resolved in 12% (w/v) acrylamide, stained with Instant Blue, and analyzed as a service by FingerPrints Proteomics Facility, University of Dundee.

Results
Evidence for overlapping roles of HydD and HyaD in the biosynthesis of Hyd-5 Hyd-5 is normally produced aerobically at very low levels [7], therefore the S. enterica LT2a strain was previously modified by the addition of a T5 promoter upstream of the hyd operon [7], and the transmembrane domain of the small subunit replaced by an affinity tag, generating LB03 [8]. This engineering was found to boost cellular hydrogenase activity in LB03 by five times in comparison to the basal activity recorded for LT2a ( Table 1).
The protein predicted to catalyze processing of Hyd-5 is HydD (STM1536) and an in-frame deletion in hydD was generated in LB03 to generate MAS02. Somewhat surprisingly, when the DhydD MAS02 strain was assayed, total hydrogenase activity levels were found to be 66% of that normally recorded for the parent strain ( Table 1). Note that deletion of the hypD gene, which is essential for cofactor biosynthesis, completely abolished all hydrogenase activity ( Table 1). The partial phenotype of DhydD could be complemented by hydD in trans (Table 1). These data indicate that HydD alone is not completely essential for Hyd-5 activity. The large subunits of Hyd-5 (HydB) and Hyd-1 (HyaB, STM1787) share 67% overall sequence identity (75% overall similarity). The processing step of Hyd-1 would be expected to be carried out by HyaD (STM1789), which itself shares 53% overall sequence identity with HydD. Thus, a double mutant (DhyaD, DhydD) was constructed (MAS04). The DhyaD, DhydD double mutant (MAS04) was found to be severely compromised in terms of Hyd-5 activity and showed only basal activity levels, similar to that exhibited by the LT2a strain under these growth conditions ( Table 1). The level of hydrogenase activity detected in the parental strain LT2a was attributed to the other hydrogenases expressed under these conditions (20% in relation to up-regulated LB03 strain), indicating that the double mutant (MAS04) was not compromised in this regard, although the assembly of the other individual hydrogenases was not investigated here. The DhyaD, DhydD double mutant could be complemented to similar levels (67-73% of hydrogenase activity measured in parent strain) by introduction of either hyaD or hydD on a plasmid (Table 1).
Engineering an epitope tag on the Hyd-5 large subunit To allow facile identification of the large subunit, new strains were engineered here that produce a HA-tagged version of the large subunit. Whole-cell hydrogenase activity was not affected by the addition of the epitope tag ( Table 2) and rocket immunoelectrophoresis showed similar levels of active Hyd-5 were present in the periplasms of the parent strain (LB03) and the modified strain (MAS05; Fig. 1C). Western immunoblotting also clearly identified HydB HA (Fig. 1A) and the processed form of the small subunit HydA HIS (Fig. 1B).
A further set of deletion alleles were introduced into the MAS05 strain (Table S1). First, a DhypD allele was introduced, thus completely blocking [NiFe] cofactor biosynthesis [11]. This led to unprocessed large and small subunits (Fig. 1A,B) and the complete loss of periplasmic Hyd-5 activity (Fig. 1C).
Next, single DhydD and DhyaD alleles were introduced into the hydB HA background (Table S1). The removal of HyaD had a minor effect on the assembly and activity of Hyd-5 ( Fig. 1), while deletion of hydD resulted in a reduction in the cellular levels of Hyd-5, as observed by rocket immunoelectrophoresis (Fig. 1C) and western immunoblotting (Fig. 1A,B), but the large and small subunits were processed as normal (Fig. 1).
Next, a DhyaD DhydD double mutant was constructed (Table S1). This strain was found to be completely devoid of periplasmic Hyd-5 activity (Fig. 1C)   and both the large subunit (Fig. 1A) and small subunit (Fig. 1B) remained in unprocessed forms. Hyd-5 assembly and activity could be recovered by addition of plasmid-borne hyaD or hydD to the double mutant (Fig. 1C), but it is notable that HyaD was unable to restore Hyd-5 to original levels (Fig. 1C).

Genetic evidence for a cofactor-dependent interaction between HydB and its maturation protease
The relationship between the Hyd-5 large subunit, HydB, and its maturation proteases, HydD and HyaD, was further studied using a bacterial two-hybrid assay based on reconstitution of Bordetella pertussis adenylate cyclase activity in an E. coli cyaA mutant [21]. HydD and HyaD were genetically fused to the N terminus of the T18 domain, and the full-length HydB protein was genetically fused to the C terminus of the T25 domain, of the adenylate cyclase (Table S2). A DcyaA E. coli reporter strain (MAE01) was generated (Table S1) before being cotransformed with pT25-HydB and either pUT18-HydD or pUT18-HyaD. The transformants were then grown anaerobically in rich medium and subsequent quantification of cellular bgalactosidase levels showed that the negative control sample (empty vectors) displayed only a basal level of activity ( Fig. 2A), while the strain producing the T25-HydB fusion together with the HydD-T18 fusion showed high levels of activity ( Fig. 2A), indicative of an interaction between HydB and HydD. The observed level of b-galactosidase induction was similar to that observed for the NarG-NarJ positive control interaction ( Fig. 2A) [32,33]. Next, the T25-HydB fusion was coproduced with the HyaD-T18 fusion anaerobically in the E. coli reporter strain ( Fig. 2A). In this case, an interaction between HydB and HyaD was also clearly detected ( Fig. 2A).
The role of the cleavable C-terminal assembly peptide on the Hyd-5 catalytic subunit was explored by  preparing aT25-HydB fusion that lacked the final 15 amino acids (termed HydB T1 ). Interestingly, coexpression of truncated T25-HydB T1 with either HydD-T18 or HyaD-T18 generated high levels of b-galactosidase activity, indicating the protein-protein interactions remained intact ( Fig. 2A). Next, the HydB large subunit was truncated still further by removal of the final 65 amino acids of HydB, including the DPCXXCXXH motif (HydB T2 ). In this case, coexpression of T25-HydB T2 with either HydD-T18 or HyaD-T18 lead to b-galactosidase activity being greatly reduced closer to that observed for the negative control ( Fig. 2A).
To investigate the role of the [NiFe] cofactor in the protein-protein interactions, the E. coli DcyaA reporter strain was further modified by incorporation of a DhypF allele to give strain MAE02. Next, the DcyaA DhypF reporter strain (MAE02) was transformed with the plasmids encoding fusions to the Hyd-5 components and the results of these experiments revealed that the levels of interactions detected were reduced to basal levels in all cases (Fig. 2B). Note, however, that the values obtained for the NarG-NarJ positive control, which is a biological system that does not utilize the hydrogenase [NiFe] cofactor, remained strong upon removal of hypF (Fig. 2B).

Biochemical evidence for a stable interaction between HydB and its maturation proteases
The E. coli FTD147 strain (Table S1), that lacks the genes for the large subunits of Hyd-1, Hyd-2, and Hyd-3, was transformed with pQE80-HydB-HydD HIS and grown under anaerobic induction conditions to allow co-overproduction of HydB and HydD HIS . A soluble protein fraction was prepared and analyzed by immobilized metal ion affinity chromatography (IMAC; Fig. 3A). The HydD HIS protein was found to  copurify with a protein of~60 kDa (Fig. 3A). Tryptic peptide mass fingerprinting identified this protein as the S. enterica HydB polypeptide (score 43 810) with 96% overall sequence coverage (Fig. S1A). The mass spectrometry demonstrated that the C-terminal assembly peptide was still present on HyaB (Fig. S1).
A similar approach was taken for HydB and HyaD HIS (Fig. 3B). In this case, only a weakly staining protein was present at~60 kDa (Fig. 3B), but nevertheless tryptic peptide mass analysis identified S. enterica HydB (score 17 697) with 96% overall sequence coverage (Fig. S1B). Interestingly, a strongly staining band at~90 kDa was present as copurifying with HyaD HIS and it was observed to cross-react with the anti-His antibody used (Fig. 3B). Tryptic peptide analysis of this species identified S. enterica HydB, (score 5146 with 65% overall sequence coverage; Fig. S2), together with HyaD (score 1022 with 66% overall sequence coverage; Fig. S2). One interpretation of these data is that the strongly staining band at~90 kDa represents a stable complex between the large subunit and the maturation protease. It could be that the~90 kDa band is a covalent complex between HydD and HyaD, especially as HyaD has three nonconserved cysteines and the samples were not boiled before SDS/PAGE due to the presence of imidazole [34]. However, note that some protein oligomers remain stable in SDS/ PAGE without covalent bonds [35].

Cross-talk between accessory proteins during [NiFe]-hydrogenase biosynthesis
In organisms that contain several [NiFe]-hydrogenases, genetic analysis usually points to each large subunit gene being coexpressed with a gene coding for a specific maturation protease [11]. In this work, the organization of the operon encoding Hyd-5 suggested that hydD was the most likely candidate as a maturation protease [7]. However, data presented here show that, when Hyd-5 is overproduced under anaerobic conditions, the role of HydD can be partially compensated for by HyaD, which is encoded within the operon for Hyd-1 ( Fig. 1 and Table 1). This was rather unexpected since studies of E. coli Hyd-3 processing demonstrated that protease gene deletions (DhycI or DhycHI) were sufficient to block Hyd-3 biosynthesis [15] but had no effect on the assembly of other hydrogenases in that organism [36]. Likewise, the hoxM gene product is the essential maturation protease for the Hyd-5-like membrane-bound-hydrogenase from Ralstonia eutropha [37], but the coexpressed soluble hydrogenase requires the action of an alternative gene, hoxW, for its maturation [38]. It is worth considering, however, that although it has been shown here that S. enterica HydD and HyaD are at least partly interchangeable, it is unlikely that the two proteins would be coexpressed under physiological conditions [5]. Thus, it remains of critical importance that each differentially expressed isoenzyme is coproduced with its own dedicated maturation protease.

A stable complex requires cofactor biosynthesis
The two-hybrid system used here has proven useful in dissection of metalloenzyme biosynthesis pathways and can detect positive interactions that have been subsequently quantified by biophysical techniques to have dissociation constants in the region of 1-1000 nM [32,33,[39][40][41]. In this work, an interaction between HydB and HydD could be readily detected using this system (Fig. 2). This behavior in the two-hybrid assay was initially surprising and counterintuitive. This is because one interpretation of the available literature would be that a maturation protease would recognize its large subunit substrate only transiently when the [NiFe] cofactor was successfully loaded [19,42,43]. Cleavage might be expected to follow quickly and thus the interaction between protease and large subunit might be short-lived, transient and unstable [43]. Instead, these data suggest the interaction is prolonged and stable under the conditions tested (Fig. 2). The two-hybrid data also indicate that binding is critically dependent on the presence of the [NiFe] cofactor, since no interactions could be detected in a cofactor mutant (Fig. 2B), but surprisingly points to the interaction being prolonged and stable in the presence of cofactor ( Fig. 2A). It is possible that either the large subunit or protease fusions are unstable or degraded in the absence of [NiFe] cofactor, and that this leads to the negative results obtained (Fig. 2B).
Furthermore, the two-hybrid assay clearly shows that the C-terminal extension on the large subunit cannot be the key docking position for the maturation proteases (Fig. 2). This observation helps to partly explain recent work on Hyd-2 assembly in E. coli [44]. In that work, the C-terminal extension from the Hyd-1 large subunit was transposed on to the Hyd-2 large subunit and processing of the chimera was reported to remain HybD dependent [44]. However, conversely, a similar experiment where the Hyd-2 large subunit extension was transposed onto the Hyd-3 catalytic domain resulted in a chimera that could not be processed by any maturation protease tested [15]. Interestingly, while HydD and HyaD retained the ability to bind to the mature form of HydB (i.e. that lacking the C-terminal extension) this interaction was also found to be dependent on an intact [NiFe] cofactor biosynthesis pathway, in so far as the bacterial two-hybrid data can be interpreted (Fig. 2). This was surprising because so-truncated large subunit has been previously reported to be completely devoid of the [NiFe] cofactor [42]. However, note that in this early work the hydrogenase large subunit in question was subjected to a partial purification procedure that may have led to loss of any loosely bound nickel cofactor that may have been present [42].
Co-overproduction of the Hyd-5 large subunit HydB with HydD HIS in an E. coli host strain resulted in the isolation of a stable complex between the two proteins (Fig. 3). These data corroborate the two-hybrid experiments (Fig. 2). Moreover, in this case, the proteolytic event was shown not to have taken place since the C-terminal peptide extension could be clearly detected by mass spectrometry (Fig. S1A). The ability to isolate such a stable protease-substrate complex is a fascinating one. A transient hydrogenase-maturation protease complex was noticed in vitro during a metal competition assay [43]; however, in most other cases reported, an inactive variant of a protease would have to be prepared in order to obtain a complex such as this. The observation that maturation proteases stably bind to, and do not quickly and transiently proteolytically process their target under these test conditions, adds an unexpected new detail to the hydrogenase biosynthetic pathway.

Supporting information
Additional Supporting Information may be found online in the supporting information tab for this article: Fig. S1. Identification of HydB copurified with Hyd-D HIS or HyaD HIS . Fig. S2. Identification of a HydB-HyaD HIS complex. Fig. S3. Sequence identity shared between Salmonella enterica maturation proteases. Table S1. Bacterial strains constructed and studied in this work. Table S2. Plasmids used in this work. Table S3. Oligonucleotides used in this work.