An archaeal protein evolutionarily conserved in prokaryotes is a zinc-dependent metalloprotease

Most organisms typically employ 2–4% of their genetic resources to code for proteases, enzymes that break down polypeptides or proteins by cleaving peptide bonds. For example, the hyperthermophilic acidophile Sulfolobus solfataricus encodes more than 70 putative proteases [1,2]. During the past decade Sulfolobus has been developed into a model system in studying archaeal molecular mechanisms such that major research progresses have been made in studying chromosome replication [3], DNA damage repair [4,5], cell cycle control [6] and regulation of gene transcription and protein translation [7–9] with this model. Although proteases are important players in cellular life, only a few proteases encoded in this archaeon have been characterized, including one extracellular protease and four intracellular enzymes [10–14]. Currently most proteases predicted for S. solfataricus remain to be characterized both for their biochemical properties and for their physiological functions.

We are interested in a pair of putative proteases, Sso0660 and Sso0661 encoded in the genome of S. solfataricus [2]. In the current protease database (MEROPS; http://merops.sanger.ac.uk/), both ORFs (open reading frames) are classified into the unknown protease family U62 [1]. These ORFs are homologous with the putative proteases encoded by tldD/tldE genes which were identified from screening for tolerance for letD on the Escherichia coli F plasmid coding for the CcdB toxin [15]. An insight into the functions of these putative proteases was gained from investigation of maturation of a peptide antibiotic and a toxin–antitoxin system both of which are plasmid-borne features in E. coli. Mutants defective in function of TldD or TldE (also called PmbA) accumulated the precursor of the antibiotic [16,17], and the encoded proteins were also implicated in the post-segregational killing by the F plasmid, which comprises two cognate factors, a stable toxin CcdB and a labile antitoxin CcdA. The cell death is ‘addicted’ to the short-lived CcdA such that cells rely on the de novo synthesis of CcdA to survive. While investigation of tldD/E deletion mutants suggested that both proteins could be involved in degradation of the CcdA antitoxin in vivo [17], crystallographic analysis of Thermotoga maritima TldE (PmbA) failed to detect any co-ordinates for metal ions in the protein structure or any structural domain of a hydrolase [18]. Since there has not been any report on biochemical characterization of a TldD/E homologue in the current literature, whether or not any TldD or TldE encodes a protease remains to be tested.

Here we report for the first time that S. solfataricus Sso0660, a TldD homologue, encodes a metalloprotease and it contains an unusual zinc-binding motif and a C-terminal cysteine residue, both of which are of crucial importance to its protease activity.

Restriction and DNA modification enzymes were purchased from New England Biolabs, Fermentas or TransGen. Plasmid DNA was extracted from E. coli cells using an AxyPrep plasmid mini-prep kit. Oligonucleotides used in the present study were synthesized from Invitrogen (listed in Table 1) where DNA sequencing of recombinant plasmids was also performed.

E. coli DH5α and Rosetta strains were used as host for DNA cloning and for producing recombinant protein respectively. Bacterial strains were cultured at 37°C in LB (Luria–Bertani) broth containing 50 μg/ml kanamycin. Chloramphenicol was further supplemented to 17 μg/ml if applicable.

Sso0660 and Sso0661 genes were amplified from the S. safataricus P2 genome by PCR using Pyrobest DNA polymerase (Takara) and specific primers (Table 1). The resultant gene fragments were cloned to the E. coli expression vector pET30a, giving pET-660 and pET-661. Sequences of the cloned DNA fragments in the plasmids were confirmed by DNA sequencing.

A PCR approach described previously [19] was employed to generate site-directed mutations in Sso0660. Two back-to-back primers were designed for generating each mutant gene (H228F, E229D, H233Y and C416G; Table 1), one of which carried the desired mutation (forward primer, fwd) whereas the other did not (reverse primer, rev). Full-length plasmids containing the designed mutations were amplified using the TransGen FastPfu DNA polymerase (TransGen Biotech) with pET-660 carrying the wild-type Sso0660 gene as the template. The resultant linear PCR products were phosphorylated with T4 DNA polynucleotide kinase and ligated with T4 DNA ligase to give circular plasmids, which were used to transform E. coli DH5α. Four resultant transformants were analysed for each cloning experiment. DNA sequencing of the mutant genes confirmed that all analysed plasmids carried the designed substitution mutations.

Expression plasmids derived from pET30a were transformed into the E. coli Rosetta, yielding strains for protein overexpression experiments. These strains were grown in LB broth at 37°C. When the attenuance of the culture at 600 nm (D₆₀₀) reached 0.6–0.8, the synthesis of recombinant protein was induced by adding 0.8 mM IPTG (isopropyl β-D-thiogalactoside). The induction was at 16°C for overnight. The cell mass was collected by centrifugation at 8000 g for 10 min and resuspended in the binding buffer of 50 mM sodium phosphate, 500 mM NaCl and 20 mM imidazole, pH 7.4. Cells were disrupted using a high-pressure homogenizer. After two passages, the resultant cell lysate was subjected to centrifugation at 12000 g for 20 min, yielding soluble and insoluble fractions of the cell lysate. The recombinant protein in the soluble fraction was purified directly by affinity chromatography with a pre-packed HisTrap column as described previously [20]. Eluted fractions containing purified recombinant proteins were pooled together and dialysed against the buffer of 50 mM Tris/HCl, 10% glycerol, pH 7.0, at 4°C overnight. The protein preparations were analysed for homogeneity by SDS/PAGE. After adding glycerol to 25% (v/v), the purified recombinant proteins were stored at −20°C until use. Only a small amount of the recombinant protein is present in the soluble fraction.

To purify the re-natured protein from the inclusion bodies, the insoluble fraction was washed with the buffer containing 50 mM sodium phosphate, 2 M urea, 1 mM EDTA, pH 8.0, and pelleted again (12000 g at 4°C for 10 min). Then, the pellet was dissolved with a denaturing buffer [50 mM sodium phosphate, 8 M urea, 5 mM DTT (dithiothreitol), 20 mM imidazole, pH 8.0], and the resultant suspension was subjected to protein purification by HisTrap affinity chromatography following the same procedure described above, except that all buffers used in this purification procedure contained 8 M urea. The purified protein was refolded in a stepwise dialysis against 50 mM sodium phosphate containing a decreasing concentration of urea (6, 4 and 2 M respectively), with each dialysis conducted for at least 4 h at 4°C. The final dialysis was in 50 mM Tris/HCl, 10% glycerol, pH 7.0. Remaining insoluble proteins were removed by centrifugation (12000 g for 10 min at 4°C), yielding re-natured proteins to be used for biochemical characterization.

Proteolytic activity of purified proteins was determined using either azocasein or FITC–BSA as substrate. In the azocasein assay, a reaction volume of 100 μl was prepared in a Microfuge tube, containing 2–10 μg of protease samples and 2% azocasein dissolved in 50 mM Tris/HCl, pH 7.0. After incubation at 55°C for 1 h, the reaction was stopped by adding 100 μl of 30% trichloroacetic acid to precipitate unreacted substrates. After removing precipitates by centrifugation (12000 g, 5 min), the enzymatic products in the supernatant were converted into coloured products by adding 1 M NaOH and quantified at 440 nm using a NanoDrop ND-1000 spectrophotometer. The specific activity of the enzyme (units·mg⁻¹·h⁻¹) was expressed as the amount of enzyme required to produce an absorbance change of 1.0 under the assay conditions. Contents of purified recombinant proteins were estimated using a MicroBCA kit from Pierce following the manufacturer's instructions.

For the fluorescence protease assay, we essentially followed the method described previously [21,22]. Briefly, a reaction mixture of 60 μl was prepared including 10 μl of FITC–BSA, 20 μl of 50 mM Tris/HCl buffer, pH 7.0, and 30 μl of enzyme solution with 2–10 μg of enzyme. After incubation at 55°C for 1 h, 120 μl of 5% trichloroacetic acid was added to each reaction mixture to terminate the reaction. Unreacted substrates were removed as described above. The fluorescence intensity of each sample was measured using a Shimadzu RF-5301PC fluorospectrometer with λ_ex at 490 nm and λ_em at 525 nm. Proteolytic activity was expressed in RFU (relative fluorescence units) with 1 unit arbitrarily defined as the amount of enzyme required for producing 1 optical density change under the assay conditions (RFU·mg⁻¹·h⁻¹).

The pH range tested for was 5–9. Two different buffers were used to generate different pH values: 50 mM phosphate buffer (pH 5.0 and 6.0) and 50 mM Tris/HCl buffer (pH 7.0–9.0). The assay started with incubating 2–10 μg of purified enzyme of Sso0660 with 2% azocasein in different buffers at 55°C for 1 h. Proteolytic activity of each sample was measured as described above.

The optimal temperature for the recombinant Sso0660 enzyme (2–10 μg) was determined similarly. Reaction mixtures containing the purified enzyme, 2% azocasein in 50 mM Tris/HCl buffer (pH 7.0) were prepared and incubated at various temperatures (40–100°C) for 1 h. The amount of products was estimated by spectrophotometry.

Nine metal ions including Zn²⁺, Cu²⁺, Ca²⁺, Mg²⁺, Co²⁺, Fe³⁺, Sr²⁺, Mn²⁺ and Hg²⁺ were studied for their effects on Sso0660 enzymatic activity at the final concentration of 5 mM. A master reaction was prepared from which aliquots were made. Metal ions were added to the aliquots individually, each with the final concentration of 5 mM. The resultant solutions were incubated at 55°C for 1 h. The amount of products was estimated by spectrophotometry as described above.

Several known protease inhibitors (see Table 3 for details) were assayed to test for their inhibition on Sso0660. Aliquots of the enzyme were pre-incubated with each inhibitor at 55°C for 1 h in the concentration indicated in Table 3. The residual activity of the enzyme was determined using the azocasein assay.

Sso0660 wild-type protein and its C416G mutant were analysed for metal ion cofactor using a TAS-990 atomic absorption spectrometer (Beijing Puxi General Instrumental Company). All solutions used for this experiment were prepared in plasticware with ultrapure water. Approximately 800 μg protein was used in each determination and the absorption of zinc ion was recorded at a wavelength of 213.9 nm. The results were the averages of measurements obtained for three different batches of wild-type and mutant proteins.

The S. solfataricus TldD homologue Sso0660 was overexpressed in E. coli as described in the Materials and methods section. As for many other thermophilic proteins expressed in a mesophilic host (e.g. [20]), Sso0660 recombinant protein forms predominantly inclusion bodies as a vast majority of the protein is present in the insoluble fraction of the cell lysate (Figure 1). We first purified the recombinant Sso0660 from the soluble fraction of the cell lysate and tested for its proteolytic activity. In an assay with azocasein as a substrate (azocasein assay), the purified Sso0660 protein showed a specific activity of 7.2 units·mg⁻¹·h⁻¹. Subsequently Sso0660 recombinant protein was also purified as a re-natured form from inclusion bodies. Protease assay with the re-natured protein showed a specific activity of 6.5 units·mg⁻¹·h⁻¹, which is slightly lower than the activity of the soluble protein. Testing both forms of recombinant Sso0660 protein with FITC-BSA as a substrate (fluorescence assay) yielded similar results, showing 1230000 and 1125000 RFU·mg⁻¹·h⁻¹ for the soluble and re-natured forms of enzyme respectively. Thus we conclude that TldD has proteolytic activity.

We also produced recombinant protein for the S. solfataricus TldE homologue Sso0661 and tested the purified protein for protease activities as for Sso0660. The purified Sso0661 protein consistently gave a low level of protease activity that was approximately 7% of the Sso0660 activity (results not shown). Moreover, whereas the purified Sso0660 recombinant protein appeared both as dimer and as monomer in a SDS/PAGE (Figure 1), the Sso0661 recombinant protein existed only as monomer (results not shown). These differences prompted us to examine their protein sequences more closely.

BLAST searches [23] of TldD/TldE homologues in the genomes of different Sulfolobales species, including S. acidocaldarius [24], S. islandicus [25,26], S. tokodaii [27], Metallosphaera cuprina [28] and Acidianus hospitalis [29] indicated that these organisms each encode 5 TldD/E homologues as for S. solfataricus. Four sets of distantly related TldD/Es were then chosen for sequence comparisons, including TldD and TldE (also named PmbA) of E. coli [30] and T. maritima [31] and the two pairs of TldD/Es (Sso0660, 0661, 1151 and 1152) of S. solfataricus [2]. Pairwise alignments of sequences of different members in TldD or TldE group revealed sequence similarity/identity of 19–26%/35–46% within each group. Multiple alignments of eight representative sequences using the Clustal W program [32] revealed several stretches of conserved sequences most of which are located within the C-terminal halves of these proteins (Figure 2). Six TldD-specific motifs include HEXXXHXXE, RMXNTXXXPG and four short sequence motifs (Figure 2), with one of the short motifs being GxC which contains the only cysteine residue of Sso0660. These findings prompted us to focus on biochemical characterization of the Sso0660-encoded enzyme.

As the re-natured Sso0660 recombinant protein was almost as active as the soluble recombinant protein both in the azocasein assay and in the fluorescent assay, re-natured recombinant protein was used in all subsequent experiments as a large yield of Sso0660 recombinant protein could be obtained from the inclusion bodies relatively easily.

First, we determined the optimal pH and temperature of the protease using the azocasein assay. When assayed for the pH range of 5–9, we found that Sso0660 is most active at pH 7 (Figure 3A). However, the enzyme also showed high activities at pH 6 and 9 (>75% of enzyme activity at pH 7), indicating that this enzyme is relatively insensitive to pH change.

The optimal temperature of the Sso0660 enzyme was also studied using the azocasein assay. As shown in Figure 3(B), the enzyme exhibits similar activities between 55 and 65°C. Relatively high proteolytic activities were observed at all other tested temperatures (>60%), except 95°C, for which the enzyme activity decreased to approximately 20% (Figure 3B). This indicates that Sso0660 recombinant enzyme is also relatively insensitive to the change of reaction temperature.

It should be noted that the temperature and pH optimums determined for Sso0660 with azocasein as the substrate may not be directly applicable to another substrate since optimal temperature and pH values of an enzyme can exhibit substrate-specificity, meaning that Sso0660 could exhibit different optimal values of temperature and pH when another substrate is to be used in the assay.

Next, we investigated the effects of various protease inhibitors on the activity of the Sso0660 protease. This was done by pre-incubation of the enzyme with each individual protease inhibitor at the indicated concentration at 55°C for 1 h, and residual protease activities in the samples were determined using the azocasein assay. PMSF and leupeptin did not show any inhibition. In contrast they stimulated the activity of the enzyme (Table 2), indicating that Sso0660 did not encode a typical serine protease. Neither pepstatin A nor bestatin inhibited the enzyme activity, demonstrating that the enzyme does not have an aspartic protease activity, nor does it have an aminopeptidase activity of leucine aminopeptidase, aminopeptidase B or tri-amino peptidase type.

When testing for the effects of the irreversible thiol inhibitor E-64 [trans-epoxysuccinyl-L-leucylamido(4-guanidino)butane] and NEM (N-ethylmaleimide), each inhibitor reduced the activity of Sso0660 to 36–56%. Since there is only one cysteine residue in Sso0660, this suggested that the cysteine residue could be important for Sso0660 activity. As both monomeric and dimeric forms of Sso0660 are present in the purified recombinant protein (Figure 1), it was of interest to study whether the monomer or the dimer could be the active form of the enzyme. The protein dimer was converted to monomer by treating the protein with DTT and the resultant monomeric protein was assayed for proteolytic activity. The protein showed a very low proteolytic activity (6%), suggesting that the protein dimer is the active form of the enzyme. When the only cysteine residue in Sso0660 was substituted with a glycine residue, the resultant mutant enzyme C416G is a monomeric protein, retaining 12% of the residual activity (Figure 4C). This indicates that the cysteine residue is very important to the enzyme activity of Sso0660.

We also treated the enzyme with EDTA or o-phenanthroline, two specific inhibitors of metalloprotease that inhibit the enzyme activity by depleting metal ions from the active centre of the enzyme. Measuring the residual proteolytic activity of the treated Sso0660 samples indicated that while o-phenanthroline strongly inhibited the activity, EDTA abolished the activity completely (Table 2). This demonstrates that Sso0660 encodes a metalloprotease.

To study the effects of metal ions on the enzyme activity, nine different metal ions, including Zn²⁺, Cu²⁺, Ca²⁺, Mg²⁺, Co²⁺, Fe³⁺, Sr²⁺, Mn²⁺ and Hg²⁺, were added to the enzyme reaction individually, all in an excess amount (5 mM). The activity of Sso0660 was assayed using azocasein as the substrate with the results summarized in Table 3. Only zinc ion further stimulated the enzyme activity (up to 160%). All other tested metal ions showed a negative effect on the enzyme activity and the strongest inhibitory effects were observed for Mg²⁺, Fe³⁺ and Co²⁺ retaining only 3–9% residual activity whereas Mn²⁺, Cu²⁺ and Hg²⁺ exhibited a moderate inhibitory effect (63–84%). To reveal if zinc ion could be a cofactor of this protease, flame spectrometric analysis of zinc ion was performed for the purified Sso0660 protein. We found that zinc ion is associated with the protein in approximately 1:1 molar ratio, indicating that Sso0660 is a metalloprotease with Zn²⁺ as the cofactor.

Although Sso0660 lacks HEXXH, a common zinc-binding motif widely present in metalloproteases [33,34], it does contain the HEAIGH motif that could represent a variant version of the common zinc-binding motif. We conducted site-directed mutagenesis for the three conserved AA (amino acid) residues (His²²⁸, Glu²²⁹ and His²³³, Table 1) and constructed mutant genes for expressing H228F, E229D and H233Y mutant proteins. As for the wild-type enzyme, these mutant proteins also formed inclusion bodies. We purified re-natured mutant proteins from insoluble fractions and analysed them by SDS/PAGE. As shown in Figure 4(B), they formed both monomer and dimer as for the wild-type enzyme.

All three Sso0660 mutants were assayed for their protease activities using the azocasein assay. As shown in Figure 4(C), the activity was greatly decreased for three mutant enzymes, retaining 8.3–14.8% of the activity of the wild-type enzyme. These results suggest that HEAIGH motif functions as the active centre of the metalloprotease.

The identified HEXXXH motif may comprise an unusual metal ion-binding motif in TldD proteins although it contains an AA insertion between the two histidine residues that co-ordinate the metal ion cofactor in the more common zinc-binding motif HExxH of metalloproteases. Interestingly, a similar motif with the same consensus functions in co-ordinating a zinc ion at the active centre of eukaryotic DPP (dipeptidyl peptidase) III enzymes [35,36]. In these eukaryotic enzymes, the interactions between the HEXXXH motif and the zinc ion is to be further stabilized by a glutamate residue located approximately 52 AAs downstream of the motif [37]. However, this glutamate residue does not appear to be conserved in TldDs, or at least not at the same location (Figure 2).

Since Cys⁴¹⁶ is important to enzyme activity, we investigated if Cys⁴¹⁶ could contribute to the zinc cofactor co-ordination at the active centre. The C416G mutant protein was subjected to flame spectrometric analysis for zinc ion component. We found that the mutant enzyme contains an equivalent or slightly higher amount of zinc ion, indicating that Cys⁴¹⁶ does not play a role in co-ordinating the zinc cofactor. Therefore, we reasoned that the importance of Cys⁴¹⁶ in Sso0660 is solely for the formation of protein dimer.

It has been well documented that zinc-binding motifs require an additional site in metal ion cofactor co-ordination. In fact, differences in the usage of an additional site for metal ion coordination have been used as a criterion to classify metalloproteases [34,38]. Notably, there are several other AAs that are only conserved in TldD sequences (Figure 2). Probably one of the conserved AAs plays a function in zinc ion co-ordination in Sso0660. Taken together, our results indicate that TldDs constitute another group of HEXXXH motif-possessing metalloproteases, re-assigning TldD enzymes to a family of protease belonging to the metalloprotease class.

The TldD superfamily of putative proteases is widespread in archaea and bacteria. While approximately 60% of known bacterial genomes encode TldD and TldE homologues, in the domain of Archaea only members in Haloarchaeota and Nanoarchaeota lack a TldD or PmbA/TldE. Many species have more than one pair of tldD/tldE-like genes as summarized in the peptidase database (MEROPS) (http://merops.sanger.ac.uk/cgi-bin/famsum?family=U62) [1]. Whereas the widespread occurrence of TldD and TldE argues for conserved functions for this superfamily of protease, in vivo functions of TldD and TldE have been investigated only for the maturation of antibiotic microcin B17 and for degradation of the antidote CcdA, both of which are plasmid-borne features. We reasoned that the observed effects of TldD/TldE deficiency could reflect the fact that these genetic elements hijacked a cellular process involving TldD/E. A prime candidate for the TldD/E-involved cellular processes is that these enzymes could be responsible for regulating the activities of chromosome-encoded toxins and antitoxins, which are highly abundant in archaea and bacteria [39]. To date, several toxin/antitoxin systems have been being characterized in bacteria (reviewed in [40]) and a pioneering experiment with an archaeal system suggests that archaeal toxin/antitoxin systems function in similar mechanisms [41]. Thus, further studies of these toxin–antitoxin systems will unravel possible functions of TldDs and TldEs in the regulation of toxin–antitoxin activities in archaeal and bacterial organisms.

Alternately, these proteases can directly regulate important enzymes by protein maturation or degradation. For example, a metal-dependent protease activity has recently been implicated in regulating the DNA reverse gyrase activity in S. solfataricus and thereby affecting genome integrity [42]. As many ORFs are predicted as metalloproteases in the MEROPS database [1], the involved protease has to be identified experimentally. Fortunately, very versatile genetic tools for in vivo study of gene functions are available for S. acidocaldarius [43] and for S. islandicus [44]. For the latter, the genetic tools developed include targeted gene inactivation, genetic complementation and protein overexpression [45–49]. Employing both biochemical and genetic methodologies to further investigate TldD/E encoded in Sulfolobus species should shed light on the functional roles of this important group of protease in Archaea.

AA

amino acid

DTT

dithiothreitol

E-64

trans-epoxysuccinyl-L-leucylamido-(4-guanidino)butane

ORF

open reading frame

LB

Luria–Bertani

NEM

N-ethylmaleimide

REU

relative fluorescence units

Yongmei Hu, Wenyuan Han and Xu Feng conducted the experiments. All authors participated in experimental design and troubleshooting and data interpretation. Nan Peng, Yun Xiang Liang and Qunxin She provided research materials. Yongmei Hu, Wenyuan Han and Qunxin She wrote the paper.

We thank members of our laboratories for helpful discussions.

This work was supported the National Natural Science Foundation of China [grant numbers 31100050, 31100096, 31128011] and the Danish Council for Independent Research: Technology and Production Sciences [grant numbers 09-062332, 11-106683].