Activity, Specificity, and Probe Design for the Smallpox Virus Protease K7L

Background: The K7L protease is required for smallpox virus assembly.

Results: K7L requires specific substrate recognition elements, dimerization, and nucleic acid cofactors for maximum activity.

Conclusion: We have identified contributors to optimal K7L protease activity and specificity.

Significance: This first characterization of a poxvirus processing protease demonstrates the importance of dimerization and extended substrate recognition in the control of activity and specificity.

Abstract

The K7L gene product of the smallpox virus is a protease implicated in the maturation of viral proteins. K7L belongs to protease Clan CE, which includes distantly related cysteine proteases from eukaryotes, pathogenic bacteria, and viruses. Here, we describe its recombinant high level expression, biochemical mechanism, substrate preference, and regulation. Earlier studies inferred that the orthologous I7L vaccinia protease cleaves at an AG-X motif in six viral proteins. Our data for K7L suggest that the AG-X motif is necessary but not sufficient for optimal cleavage activity. Thus, K7L requires peptides extended into the P7 and P8 positions for efficient substrate cleavage. Catalytic activity of K7L is substantially enhanced by homodimerization, by the substrate protein P25K as well as by glycerol. RNA and DNA also enhance cleavage of the P25K protein but not of synthetic peptides, suggesting that nucleic acids augment the interaction of K7L with its protein substrate. Library-based peptide preference analyses enabled us to design an activity-based probe that covalently and selectively labels K7L in lysates of transfected and infected cells. Our study thus provides proof-of-concept for the design of inhibitors and probes that may contribute both to a better understanding of the role of K7L in the virus life cycle and the design of novel anti-virals.

Introduction

The Orthopoxviridae family of viruses infects vertebrate and invertebrate animals. Humans can be infected by several poxviruses, but the most prominent of them are variola (smallpox), monkeypox, vaccinia, and molluscum contagiosum viruses (1). The poxviruses are encoded by a large double-stranded DNA genome that is translated into ∼250 individual proteins within the host cytosol (2). The life cycle of the virus includes several morphogenetic stages during one of which a spherical, immature particle transforms into a brick-shaped, infectious virion. This maturation has been associated with the proteolytic processing of several structural proteins by two viral proteases (3–8), but the enzymatic principles that govern this process remain poorly understood. Thus, the maturating peptidases are known to pack together with their targets inside the viral core (a protein compartment surrounding the DNA), but the processes that switch their activity on are still unknown. Furthermore, the importance of these peptidases for infectivity makes them attractive targets for the development of antiviral therapies against a broad spectrum of poxviruses (9–11). Consequently, a better understanding of their substrate specificities should be beneficial for drug design.

The two variola genes that encode peptidases are H1L and K7L,⁴ whereas the same genes in the vaccinia virus are called G1L and I7L, respectively (12). The H1L/G1L gene products are distantly homologous to the M16 family of bacterial and eukaryotic metallopeptidases (6), which either degrade oligopeptides or truncate the signal peptides from mitochondrion/chloroplast-secreted proteins (13). The function of poxviral metallopeptidase is poorly understood, but it appears to play a complementary role in virus maturation (8, 12, 14).

The gene products of K7L (variola) and I7L (vaccinia) are 423-residue proteases that belong to the MEROPS Clan CE (7, 15, 16). The two proteases differ at only four residues and are thus likely to be functionally very similar. Residues important in catalysis of other members of Clan CE are conserved in K7L/I7L and suggest that it is a cysteine protease (confirmed experimentally by Byrd and Hruby group (17)). Based on in vivo cleavage data, I7L processes one envelope protein (P21K (A17L)) and four core proteins (P4a (A10L), P4b (A3L), P17K (A12L), and P25K (L4R)) (7, 16, 18, 19) at sites with the AG-X motif (where X is generally a small residue). AG-X motifs occur in other vaccinia proteins as well, but they are not processed, suggesting that additional substrate motifs are required for recognition (12).

The closest non-viral homologs of K7L are the eukaryotic SUMO-specific proteases (15, 20) and the SUMO-specific and/or ubiquitin-specific proteases from pathogenic bacteria (supplemental Fig. 1A). The function of the SUMO-specific proteases is to process SUMO precursors and deconjugate SUMO from target proteins (21–24). In a similar way, the Clan CE proteases of pathogenic bacteria deconjugate ubiquitin-like proteins in the host cell and suppress host cell defenses. As a result, they function as virulence factors (25–28). In addition to the poxviruses, Clan CE peptidases have been identified in two other double-stranded DNA viruses: adenoviruses and African swine fever viruses (29, 30). The adenoviral peptidase (adenain) processes both viral and host proteins (31–34). There is evidence that some Clan CE proteases are regulated by substrate binding. Thus, the Saccharomyces SUMO-specific protease Ulp1 (ubiquitin-like protein) is activated by the binding of its substrate protein, SMT3, at the cleft formed by the N-and C-terminal subdomains (20). In turn, adenain is activated by binding of the 11-residue fragment of the core protein pIV (supplemental Fig. 1B) and viral DNA (29, 35, 36). But it is not known if this is a general feature of viral CE proteases, including K7L.

Biochemical characterization of the K7L (or I7L) protease has not been previously performed, due to the lack of pure, active enzyme. Here, we report the expression and biochemical characterization of the K7L variola protease. Our results lead us to propose a model for substrate recognition and regulation on the activation mechanism and additional mechanistic parameters of this essential processing protease.

EXPERIMENTAL PROCEDURES

Recombinant Plasmids

The full-length K7L cDNA sequence was generated by QuikChange mutagenesis (Stratagene) of a vaccinia virus I7L template. The K7L construct was cloned into the pGEX-4T vector (GE Healthcare), which encodes an N-terminal glutathione S-transferase (GST) purification tag followed by a thrombin cleavage site (Fig. 1A). The proteases I7L/K7L of variola and vaccinia viruses differ by only two non-conserved mutations, G75R and T308I; all other substitutions are strain-specific, and we presume that they are unlikely to affect protease activity. The substitutions are far from a putative substrate binding site (based on threading the K7L sequence onto crystal structures of proteases from Clan C57); see the supplemental Fig. 1 for a more complete description. The gene encoding the wild-type p25K construct (C-terminal FLAG-tagged) and the wild-type and C328A K7L mutant (each N-terminally FLAG-tagged) were cloned under the control of the T7 polymerase promoter into the mammalian expression vector pTF7-3. The cloning of p25K led to a mutation of Ser² to Ala. The P4a and P4b genes were cloned into pcDNA3.1 vectors. The P25K gene was also cloned into the bacterial expression vector pET151-TOPO (Invitrogen), providing it with an N-terminal His tag and tobacco etch virus protease cleavage site. The identities of the constructs were validated by DNA sequencing.

FIGURE 1.

Expression of K7L and analysis of its oligomeric state. A, shown is a GST fusion construct used for expression and purification of K7L. B, shown is SDS-PAGE of GST-K7L (lane 1), K7L after removal of GST by thrombin (lane 2), and K7L cross-linked with BS3 (bis[sulfosuccinimidyl]suberate) (lane 3). C, gel filtration of thrombin-cleaved K7L on Superdex S200-HR1030 is shown. Numbers on the side (B) or top (C) indicate molecular mass standards.

Expression of K7L Protease and P25K Protein in Escherichia coli

The pET151-p25K and pGEX-K7L plasmid constructs were expressed in E. coli BL21 Codon plus-RIL cells (Stratagene). Expression was induced using 0.4–0.6 mm isopropyl-d-galactopyranoside for 17 h at 17 °C. Cells were then collected by centrifugation, resuspended in 50 mm Hepes buffer, pH 7.3, containing 15% glycerol, 500 mm NaCl, 1 μm leupeptin, 1 mm PMSF, and 1 mm 2-mercaptoethanol), and lysed using a French press. The pellet was removed by centrifugation (45,000 × g for 1 h). The GST-K7L construct was purified from the supernatant fraction using affinity chromatography on a glutathione-Sepharose 4B column (GE Healthcare). The protein were eluted with 50 mm Tris-HCl buffer, pH 8.0, containing 10 mm reduced glutathione, 10% glycerol, and 500 mm NaCl. The purified samples were cleaved for 16 h at 4 °C using bovine thrombin (Sigma). Because of the thrombin cleavage site, the resulting proteins contained an additional Gly-Ser at the N terminus. After inactivation of thrombin by 0.5 mm 4-(2-aminoethyl)-benzenesulfonylfluoride, the cleaved proteins were further purified using a HiTrap heparin HP column (GE Healthcare). Proteins were eluted using a 0.2–1.0 m gradient of NaCl concentration in 30 mm HEPES buffer, pH 7.6, containing 10% glycerol. Samples were stored in 2 mm DTT and 50% glycerol at −80 °C.

His-tagged P25K was purified using a HiTrap Ni²⁺-chelating HP column (GE Healthcare). The protein was eluted using a 20–250 mm gradient of imidazole in 30 mm Tris-HCl buffer, pH 8.0, containing 500 mm NaCl and then cleaved for 16 h at 4 °C using tobacco etch virus protease. Due to the cleavage preferences of the tobacco etch virus protease, the N terminus of the resulting P25K protein included a six-residue extension (Gly-Ile-Asp-Pro-Phe-Thr). The digested samples were dialyzed against 30 mm Tris-HCl buffer, pH 8.0, containing 500 mm NaCl. Residual His-tagged P25K protein and tobacco etch virus protease were removed from the samples using a HiTrap Ni²⁺-chelating HP column. The digested P25K was further purified on a Superdex 75 HR 16/50 column (GE Healthcare) equilibrated with 50 mm HEPES buffer, pH 7.6, containing 0.3 m NaCl. Samples were stored at −80 °C in the same buffer augmented with 50% glycerol. Protein concentration was estimated from its A₂₈₀ using an ND-1000 spectrophotometer (NanoDrop); extinction coefficients were estimated by ProtParam.

Cell-free Expression of Vaccinia P4a, P4b, and P25K Proteins

A TNT coupled transcription/translation cell-free system (Promega) was used to synthesize vaccinia wild-type P25K using the pTF7-3 plasmid as template and P4a and P4b using the pcDNA3.1 plasmid as template. Biotinylated Lys-tRNA was used to label the synthesized proteins and to facilitate their detection by a streptavidin-horseradish peroxidase (HRP) chemiluminescent kit (Promega).

Purification of DNA and RNA

Vaccinia virus was purified from the infected BHK-21cells using sucrose gradient centrifugation (37). Total RNA was extracted from insect SF9 cells using an RNeasy Mini kit (Qiagen). The DNA and RNA samples were additionally purified by ethanol precipitation.

Size Exclusion Chromatography of K7L

K7L aliquots (50 μl of the 7 and 20 μm samples) were injected into a Superdex-200 HR10/30 column (Amersham Biosciences) equilibrated with 30 mm Tris-HCl, pH 8.0, containing 300 mm NaCl. The column was eluted at 0.5 ml/min using an ACTA FPLC system. Molecular weight standards (Bio-Rad) were used to calibrate the column.

Cross-linking of K7L

K7L samples (2 μm) were cross-linked for 10–30 min at 20 °C in 20 mm HEPES buffer, pH 7.6, containing 300 mm NaCl, using 0.2 μm bis[sulfosuccinimidyl]suberate (BS3; Pierce). Reactions were stopped using 0.1 m Tris. Samples were separated on 4–12% gradient Bis-Tris SDS-PAGE and Coomassie- stained.

Sedimentation Equilibrium Experiments with K7L

Sedimentation equilibrium experiments were performed using a ProteomeLab XL-I analytical ultracentrifuge (Beckman-Coulter). K7L (0.075, 0.15, and 0.3 mg/ml) in 30 mm Tris-HCl buffer, pH 8.0, containing 200 mm NaCl was analyzed with and without 25% glycerol. The samples were loaded in the 6-channel equilibrium cells and spun at 20 °C for 24 h in an An-50 Ti 8 rotor at 18,000 rpm.

K7L Proteolysis of P25K

Purified P25K (8.3 μm) was coincubated with K7L (0.3 μm) in 30 mm Tris-HCl buffer, pH 8.0, containing 2–45% glycerol, 100 mm NaCl, and 2 mm DTT. Where indicated, 1 mm MgCl₂ and 0.1 mg/ml RNA and vaccinia DNA were added to the reactions. Reactions were stopped by adding SDS-loading buffer and then separated using 4–12% denaturing Bis-Tris SDS-PAGE gels (Invitrogen). Gels were Coomassie-stained and then digitized using a gel imager (Alpha Innotech). The specificity constants (k_cat/K_m) were estimated using the Wω approximation for time-dependent product accumulation (described in Ref. 38) and the assumption of [S] < K_m. The latter was validated by measuring inhibition of K7L-mediated cleavage of Ac-ISAG-7-amino-4-carbamoylmethyl coumarin (ACC) by P25K (see “Results”).

Synthesis of the Fluorogenic Peptides

All substrates containing the ACC fluorophore were synthesized using solid phase synthesis as described previously (39, 40). Purity and structure were confirmed using analytical HPLC and mass spectrometry. The tetrapeptide substrates were also analyzed by ¹H NMR (see supplemental Methods).

Synthesis of FAM Peptides

FAM-labeled peptides were synthesized using an Apex 396 multiple peptide synthesizer on a Rink Amide MBHA resin (0.72 mmol/g) (NovaBiochem) using a standard Fmoc (N-(9-fluorenyl)methoxycarbonyl) protocol and diisopropylcarbodiimide/1-hydroxybenzotriazole coupling (see supplemental Methods).

Synthesis of Activity-based Probes

Amino acid side-chain-protected biotin-AEEAc-FYA-COOH, biotin-AEEAc-EDTIFFA-COOH, and Biotin-hex-EDCIFYA-COOH peptides sequences were synthesized on a solid support using 2-chlorotrityl chloride resin by coupling using HATU (2-(1H-7-azabenzotriazol-1-yl)-1,1,3,3-tetramethyl uronium hexafluorophosphate) in N,N-dimethylformamide of the appropriate peptide sequence to vinyl methyl ester (obtained as previously described in Ref. 41. Deprotection of the peptides was carried out in 92.5:2.5:2.5:2.5 (v/v) trifluoroacetic acid/triisopropylsilane/phenol/water. Purity and structure of the synthesized material were confirmed using analytical HPLC and mass spectrometry. Results are: biotin-AEEAc-FYA-vinyl methyl ester (VME), electrospray ionization (ESI)-MS = 868.2, HPLC t_R = 13.00 min, yield = 47%, biotin-AEEAc-EDTIFFA-VME, ESI-MS = 1310.55; HPLC, t_R = 16.23 min, yield = 43%; biotin-hex-EDCIFYA-VME, ESI-MS = 1296.55; HPLC t_R = 16.39 min, yield = 51%.

Activity Assays Using the Fluorogenic Substrates

K7L activity was measured in duplicate in 50–100-μl volumes in 96-well plates. The reaction mixtures contained K7L (0.1–0.4 μm), a fluorescent substrate (50–100 μm), glycerol (10–45%), and Tris-HCl buffer, pH 8.0, containing 50–100 mm NaCl and 2 mm DTT. The samples were incubated at 25 °C for 5–60 min. The substrate cleavage rate was continuously monitored at λ_ex = 360 nm and λ_em = 460 nm using a Spectramax Gemini EM fluorescence spectrophotometer (Molecular Devices). V, k_cat, and K_m were calculated from the cleavage data using the equation V = k_cat[E][S]/(K_m + [S]).

Activity Assays Using FAM-labeled Peptide Substrates

Assays were performed in duplicate 96-well plates. Reaction mixtures contained K7L protease (0.2–0.6 μm), FAM-peptide substrate (15 μm), and 30 mm Tris-HCl buffer, pH 8.0, supplemented with 15–45% glycerol, 50 mm NaCl, and 2 mm DTT. Samples were co-incubated at 25 °C for 5–30 min. Aliquots were withdrawn, and the enzyme was inactivated by 6 m guanidine-HCl. Aliquots were chromatographed using a C4 reverse-phase column (Higgins, Proteo 300, 250 × 4.6 mm). Fluorescence of eluted fractions was continually monitored at λ_ex = 490 nm and λ_em = 530 nm using a Gilson fluorescence detector (model 121) equipped with a narrow-band filter. Peptide yield was estimated from the peak area. The initial rate of the cleavage reactions (V_o) was then calculated based on the cleaved peptide yield using the equation V_o = A_p[S]/(A_s + A_p), where A_s is the area of the intact FAM-peptide, and A_p is the area of the FAM-digest product. Two experiments with independently prepared K7L samples were performed to obtain statistically significant data. Substrate depletion was minimized by collecting data below 30% product formation. Because K_m for the peptide substrates significantly exceeded the substrate concentrations used in the cleavage reactions, the k_cat/K_m specificity constant was estimated using the equation k_cat/K_m = V_o/[E][S].

To determine the molecular mass of the cleavage products and thus the location of the scissile bonds, the digest reactions were also analyzed by MALDI-TOF MS using a Bruker AutoFlex II MALDI-TOF mass spectrometer in a cyano-4-hydroxycinnamic acid matrix in 50% acetonitrile, 0.1% trifluoroacetic acid.

RESULTS

Recombinant Variola K7L Is a Dimeric Protease

GST-tagged K7L was expressed in E. coli as a soluble protein (1 mg/liter bacterial culture), purified, and cleaved with thrombin as described under “Experimental Procedures” (see Fig. 1). K7L migrated on SDS-PAGE with the expected size of 45 kDa (Fig. 1B). Size exclusion chromatography suggested a moleculare mass of 75–80 kDa, more consistent with homodimerization in solution (Fig. 1C). This was corroborated by cross-linking experiments, which demonstrated a dimer but no higher order polymers (Fig. 1B). Analytical ultracentrifugation (which is not affected by molecular shape) confirmed that K7L exists as a fairly strong dimer either in the absence or the presence of 25% glycerol (supplemental Fig. 2), with K_d values of 0.4 ± 0.2 μm (no glycerol) and 0.2 ± 0.1 μm (25% glycerol). At the concentrations used in our activity assays, K7L is, therefore, predicted to be principally dimeric. K7L could be stored for several months in 50% glycerol at −20 °C or −80 °C without loss of proteolytic activity. However, the enzyme lost activity at higher temperatures, with a half-life of ∼5 h at 23 °C and ∼30 min at 37 °C. Activity measurements were, therefore, performed at 20–25 °C.

The reported targets of K7L are P25K, P4a, and P4b. To determine the activity of recombinant K7L, we incubated it with in vitro translated P25K, P4a, and P4b proteins expressed from constructs encoding the highly similar vaccinia proteins (because variola DNA was not available). Variola and vaccinia P25K, P4a, and P4b differ by between 1–3%, and all sequence differences are distant from the cleavage sites; see supplemental Alignments. Because the sequences display such a high degree of identity, it is very likely that the activity of variola K7L toward vaccinia proteins would be essentially identical to the cleavage of the corresponding variola proteins. The activity assays were performed in the presence of 40% glycerol (Fig. 2). There is one putative cleavage site in P4b (Gly⁶⁰↓Ser⁶¹) and two in P25K (Gly¹⁸↓Ser¹⁹ and Gly³²↓Ala³³). In accord, we observed two cleavage products of the expected molecular mass for P25K (32 and 30 kDa; Fig. 2C) and one for P4b (64 kDa; Fig. 2D). Cleavage of one of the P25K sites was especially fast, and we observed slower cleavage of the other site and P4b by K7L. By contrast, P4a was not cleaved by K7L under our experimental conditions (Fig. 2E).

FIGURE 2.

Presumed viral targets of K7L and their in vitro cleavage patterns. A, P4a, P4b, P25K, and P17K are viral core proteins, whereas P21K is the envelope membrane protein. Alternative gene-based names (in parentheses) are given as well as the putative cleavage sites and their residue numbers. B, alignment of the cleavage sites is shown. Acidic residues that enhanced cleavage are in bold. C, D, and E, plasmids encoding the indicated proteins were transcribed and translated in vitro. Recombinant GST-K7L was added for the indicated times, and the reaction was terminated by boiling in SDS sample buffer. SDS-PAGE gels were developed with streptavidin-HRP to reveal processed and unprocessed substrate proteins. Bands common to both P4b and P4a gels belong to nonspecifically stained proteins. Cleavage reactions employed 0.3 μm GST-K7L at 25 °C in 50 mm NaCl, 2 mm DTT, 30 mm Tris, pH 8.0, buffer containing 40% glycerol and 12% v/v of the translation product.

Glycerol Enhances K7L Activity

To evaluate the activity of the K7L samples, we synthesized several peptidyl substrates corresponding to the putative P25K, P4a, and P4b cleavage sites (Table 1). The peptides were N-terminally acetylated and contained ACC as the fluorophore. Activity was measured by the release of the free fluorophore using a fluorimeter operating in the kinetic mode. K7L exhibits low activity against Ac-ISAG-ACC, but its activity was dramatically enhanced in the presence of 45% glycerol (Fig. 3A). Similar to glycerol, other osmolytes such as betaine and sucrose also increased activity against Ac-ISAG-ACC (Fig. 3B). As a result, routine activity measurements were performed in the presence of 40–45% glycerol at an optimal pH of 8.0 (Fig. 3C) unless otherwise indicated. Glycerol increased K7L activity against other fluorescent peptides distinct from Ac-ISAG-ACC and full-length protein substrates (data not shown), and thus we conclude that the osmolyte effect is on the enzyme, not the substrate.

TABLE 1.

K7L proteolysis of fluorogenic peptides

Kinetic parameters were evaluated in 30 mm Tris-HCl buffer, pH 8.0, 100 nM NaCl, 2 mm DTT, and 45% glycerol with a range of peptide concentrations and K7L from 0.2 to 1.0 μm. NM, not measured. Errors indicate observed standard deviation of three measurements.

Sequence source	Peptide sequence	kcat/Km	Km
		m−1s−1	μm
P25K, Gly18↓Ser19	Ac-FFAG-ACC	300 ± 5	100 ± 15
P25K, Gly32↓Ala33	Ac-VIAG-ACC	120 ± 30	NM
P4b, Gly60↓Ser61	Ac-ISAG-ACC	80 ± 30	830 ± 110
P4a, Gly614↓Ser615	Ac-FYAG-ACC	560 ± 90	NM
P4a, Gly697↓Thr698	Ac-TNAG-ACC	No cleavage	No cleavage
P25K, Gly18↓Ser19	Ac-TIFFAG-ACC	570 ± 30	7.5 ± 3
P25K, Gly18↓Ser19	Ac-EDTIFFAG-ACC	14,400 ± 300	26 ± 7
P4a, Gly614↓Ser615	Ac-RYFYAG-ACC	880 ± 140	NM
P4a, Gly614↓Ser615	Ac-CPRYFYAG-ACC	1010 ± 20	150 ± 30

FIGURE 3.

Stabilizing effects of osmolytes, salt, pH, and enzyme concentration on K7L activity. A, shown is activity (velocity as a function of Ac-ISAG-ACC concentration) in the presence of 10 and 45% glycerol. B, shown is a comparison of K7L activation by betaine, glycerol, and sucrose (velocity as a function of osmolyte concentration) in 35 μm Ac-ISAG-ACC. C, shown is the effect of NaCl concentration on K7L activity in 35 μm Ac-ISAG-ACC. The initial velocities were measured at 25 °C in 30 mm Tris, pH 8.0, and 2 mm DTT with the addition of the various components. C, shown is the effect of pH on K7L activity in the presence of 45% glycerol. D, shown is the effect of K7L concentration on the specific activity of K7L toward 30 μm Ac-EDTIFFAG-ACC in the presence of 10% glycerol.

Dimerization Enhances K7L Activity

We analyzed the effect of K7L dimerization on specific activity by observing the concentration dependence of the cleavage of an ACC-labeled peptide substrate (Fig. 3D). The dependence had a sigmoidal shape with AC₅₀ = 0.2 ± 0.1 μm, which corresponds to the K_D value determined by analytical ultracentrifugation (see above). This result clearly links the activity of K7L with dimerization.

Finally, we compared the activities of recombinant K7L produced in E. coli versus vaccinia virus-infected mammalian cells (supplemental Fig. 3). No significant differences were observed, indicating that K7L is not differentially activated in mammalian cells (e.g. by post-translational modification).

Probing Substrate Specificity

The substrate specificity of vaccinia I7L has previously been defined by in vivo cleavage assays of mutant P25K (5, 42, 43). However, because the published data did not provide quantitative comparisons of cleavage efficiencies, we performed an in-depth analysis of K7L. The P1 site of all putative natural K7L substrates is a Gly residue. To test the tolerance of K7L quantitatively, we mutated the conserved Gly at positions 18 and 32 of P25K to Ala and compared their cleavage efficiencies with wild-type P25K. As shown in Fig. 4A, both sites of WT P25K were fully cleaved in 1 and 9 h, respectively, whereas no evidence for cleavage of the double mutant was obtained after 9 h of incubation. This indicates a very strong preference for Gly in the P1 position of protein substrates.

FIGURE 4.

Subsite profiling of K7L and comparison with related protease substrate preferences. A, P25K and its double mutant, G18A/G32A, were produced in a cell-free in vitro transcription/translation system (Promega), cleaved by recombinant K7L-GST for the indicated times, and visualized after SDS-PAGE as described under “Experimental Procedures.” B, amino acid preferences at the P3-P1′ positions of the cleavage site Gly↓Ser¹⁸ of P25K are shown; Gly³² was always mutated to Ala to facilitate quantitative analysis. The analyses were conducted in same way as shown in the panel A. Single mutations were introduced at the P3-P1′ positions of the site Gly↓Ser¹⁸, as indicated on the horizontal axis. The vertical axis shows the relative cleavage rates (%). C, a library of peptides exploring the P2-P4 positions was constructed as described under “Experimental Procedures,” and relative activities of 0.5 μm K7L are expressed in relation to the most efficiently cleaved substrate at each position. All reactions were conducted in 40% glycerol. The horizontal axis shows amino acids written in the standard one-letter code; the unnatural amino acids are: O, norleucine; B, β-alanine. D, relative preferences are compared for related deSUMOylating and deubiquitinating proteases. Red indicates the highest activity; black indicates the lowest. The preferences are clustered by their similarity to K7L.

We next studied cleavage of tetrapeptide substrates and employed a positional scanning substrate library approach to define optimal preferences at the P2, P3, and P4 positions. The P1 position was fixed as Gly, and the P2, P3, and P4 positions were each randomly altered to one of 18–20 amino acids, including norleucine and β-alanine (Fig. 4C). Because we employed similar peptide concentrations, the relative contributions of the amino acid substitutions to activity could be directly measured. The results demonstrate a high preference for Ala at P2. Selectivity at P3 was lower, but hydrophobic residues were especially well tolerated. The P4 position was more restricted than P3 and demonstrated a clear preference for hydrophobic residues (Fig. 4C). These results agree well with the sequence motifs that exist in the natural cleavage targets of K7L (Fig. 2C) and distinguish K7L from closely related proteases in Clan CE (Fig. 4D) (39).

We next utilized our peptide-based observations in a mutational analysis of the putative natural substrate P25K. We mutated P3, P2, and P1′ positions in the Gly¹⁸↓Ser¹⁹ cleavage site while introducing an Ala mutation at Gly³¹ to silence any contribution of the secondary cleavage (Fig. 4B). P1′ in the Gly¹⁸↓Ser¹⁹ site demonstrated a clear preference for Ser and Ala, the most frequently occurring residues at all putative natural K7L cleavage sites (Fig. 2B). We also found that substitution of Ala at P2 with Gly did not affect the cleavage rate, confirming the proposal that cleavage of natural substrates may be influenced by exosites (12).

To corroborate these findings we synthesized individual tetrapeptide ACC substrates corresponding to the optimal tetrapeptide sequence and the cleavage sites of P25K, P4b, and P4a and co-incubated them with K7L. The k_cat/K_m parameters were then calculated from the cleavage data (Table 1), revealing a preference for Tyr and Phe at the P3 position. These results indicate that the K7L specificity in the P2, P1, and P1′ positions surrounding the scissile bond correlates well with cleavage sites of natural proteins sequences (Fig. 2, C–E). However, beyond this region, correlations break down.

Mapping the Extended Substrate Binding Site of K7L

The discrepancies in cleavage site specificities between peptide and protein substrates beyond the catalytic cleft suggested that more distant residues than P4-P1 might affect activity. To determine the effect of distant residue positions on K7L activity, we utilized two strategies. In the first strategy we extended the ACC-based peptidyl substrates to encompass the P1-P8 residues matching the P25K and P4a cleavage sites. Kinetic measurements reveal that extending both substrates to the P5 and P6 residues gave only a minor enhancement in the rate of peptide hydrolysis. In contrast, extending the substrates to the P7 and P8 positions promoted hydrolysis by up to 25-fold while leaving the hydrolysis of the P4a-based peptide almost unchanged. Notably, negatively charged residues in P7 and P8, derived from the P25K cleavage sites, were preferred over the neutral residues found at these positions in the site P4a and P4b sequences.

In a second strategy, we synthesized peptides that covered the P9-P4′ residues of P25K, P4a, and P4b, derivatized at their N terminus with FAM. This allowed us to monitor cleavage by PR-HPLC of timed samples at low μm substrate concentrations, permitting an estimate of the catalytic efficiency (k_cat/K_m) (see “Experimental Procedures”). The cleavage sites were independently confirmed by MALTI-TOF mass spectrometry. In keeping with the results presented in Table 1, peptides encompassing the Gly¹⁸↓Ser¹⁹ site of P25K were cleaved by K7L far more efficiently than peptides derived from other sites of P25K, P4a, and P4b proteins.

Interestingly, the FAM-LDRIITNAGTCTV peptide spanning the Gly⁶⁹⁷↓Thr⁶⁹⁸ cleavage site of P4a appeared resistant to K7L proteolysis. This was in accord with data on the Ac-TNAG-ACC peptide demonstrating that this site is highly unfavorable for cleavage by K7L (Table 2). In agreement, MALDI-TOF mass spectrometry data confirmed that the FAM-LDRIITNAGTCTV peptide was not cleaved even by very high levels of K7L (not shown). Similar to the Ac-EDTIFFAG-ACC substrate, removing P9-P7 residues from the optimal P25K peptide sequence reduced its cleavage rate by ∼40-fold. Significantly, extending substrates beyond the scissile bond in the prime direction (downstream of the scissile bond) had no major impact on cleavage rates, and extending from P2′ to P4′ provided only a modest (less than 2-fold) enhancement in catalysis (Table 2). Overall, our cleavage data for longer peptides correlate very well with K7L activity in vitro against protein substrates (P25K and P4b). Based on our data, it is likely that the Gly⁶⁹⁷↓Thr⁶⁹⁸ site of P4a protein is a suboptimal substrate of K7L.

TABLE 2.

K7L proteolysis of the P9-P4′ FAM-labeled peptide

Kinetic parameters were evaluated in 30 mm Tris-HCl buffer, pH 8.0, 100 nM NaCl, 2 mm DTT, and 45% glycerol with 10 μm peptide and K7L from 0.1 to 1.0 μm. O (norleucine) was used in place of Met to obviate oxidation of the latter during synthesis of DDLQMVIAG↓AKSK sequence. Errors indicate observed S.D. of two measurements. ND, not determined.

Protein, cleavage site	Peptide sequence	kcat/Km
		45% Glycerol	15% Glycerol
		m−1s−1
P25K, Gly18↓Ser19	EEDTIFFAG↓SISE	28000 ± 3000	1300 ± 400
P25K, Gly18↓Ser19	EEDTIFFAG↓SI	17500 ± 1500
P25K, Gly18↓Ser19	TIFFAG↓SISE	670 ± 220
P25K, Gly32↓Ala33	DDLQOVIAG↓AKSK	3500 ± 800	220 ± 100
P4b, Gly60↓Ser61	VDDDFISAG↓ARNQ	2000 ± 500
P4a, Gly614↓Ser615	YCPRYFYAG↓SPEG	2100 ± 500
P4a, Gly697↓Ser698	LDRIITNAG↓TCTV	ND
P4a, Gly697↓Thr698 mutant	LDRIITNAG↓TSTV	ND

Evidence for Allosteric Exosite-mediated Activation of K7L

The substantial stimulation of activity of K7L catalysis on substrates occupying the P7 and P8 positions suggested two possibilities; 1) extended concerted substrate recognition or 2) an exosite interaction acting as an allosteric switch. To address these hypotheses we tested whether P25K could enhance K7L proteolysis of the Ac-ISAG-ACC substrate. We co-incubated K7L (0.5 μm) with P25K (0.1–100 μm) and then measured cleavage velocity of the fluorescent substrate. The control proteins bovine serum albumin (BSA) and lysozyme had no effect on cleavage velocity, but the P25K protein enhanced the K7L activity by up to 2.4-fold, with a maximal stimulatory effect at 8 μm P25K. Higher P25K concentrations led to the inhibition of K7L activity against Ac-ISAG-ACC (Fig. 5A). In contrast, the FAM-EEDTIFFAGSISE peptide showed no stimulation of Ac-ISAG-ACC and instead inhibited with an IC₅₀ of 37 μm, similar to the concentration at which P25K begins to inhibit cleavage of the tetrapeptide reporter substrate (Fig. 5B). These findings are consistent with a small allosteric effect but indicate that the major mechanism of enhanced cleavage of long peptides is the presence of extended and concerted substrate recognition.

FIGURE 5.

Influence of full-length P25K and polynucleotides on K7L activity. A, activation by P25K of Ac-ISAG-ACC cleavage is shown. The concentration-dependent activation fits a bell-shaped dose-response curve, with K_A = 8 μm and K_I = 45 μm. BSA and lysozyme were used as controls of nonspecific binding. B, shown is inhibition by FAM-EEDTIFFAGSISE of Ac-ISAG-ACC cleavage, with data fitted to a first-order dose-response curve, K_I = 37 μm. The reactions were carried out in 30 mm Tris, pH 8.0, containing 40% glycerol, 50 mm NaCl, 0.2–0.5 μm K7L, and 50 μm the Ac-ISAG-ACC substrate. C, cleavage of P25K by K7L in the presence of various co-factors is shown. 0.3 μm K7L was co-incubated for the indicated times with 8 μm P25K in the presence of 1 mm Mg²⁺ (left panel), human RNA (0.1 mg/ml) + Mg²⁺ (middle panel), or viral DNA (0.1 mg/ml) + 1 mm Mg²⁺ (right panel). Experiments were performed 3, 15, or 40% glycerol as indicated. The digests were separated by SDS-PAGE. D, the images in C were digitized to calculate the respective k_cat/K_m values as described under “Experimental Procedures.” E, effects of glycerol, human RNA, and Mg²⁺ ions on the cleavage of Ac-EDTIFFAG-ACC by K7L are shown. K7L (0.1 μm) was co-incubated at 25 °C with the peptide (10 μm). The reactions contained 30 mm Tris-HCl, pH 8.0, 50 mm NaCl, and 2 mm DTT, and the indicated concentrations of glycerol. Where indicated, 1 mm MgCl₂ or 0.1 mg/ml RNA were added to the reactions.

Polynucleotides Stimulate Proteolysis of P25K by K7L

P25K is a high affinity binder of polynucleotides (K_d = 1–2 nm), including both RNA and DNA (44). To test whether polynucleotides affect the cleavage of P25K, we co-incubated K7L with P25K in the presence and the absence of RNA, double-stranded vaccinia DNA, and glycerol. DNA, and especially RNA (both at 0.1 mg/ml), substantially stimulated the cleavage of both P25K sites, and their effect was particularly strong at lower concentrations of glycerol (3–15%) (Fig. 5, C and D). It is possible that polynucleotides are required to stimulate K7L proteolysis of the P25K protein and that this stimulation does not require the presence of high glycerol. Importantly, RNA (and viral DNA; not shown) both inhibited cleavage of the Ac-EDTIFFAG-ACC peptide by K7L by ∼4-fold (Fig. 5E), suggesting that K7L has a propensity to bind polyanions. Moreover, polynucleotide stimulation of catalysis on P25K was much higher than the inhibitory effect toward the octapeptide substrate. Because the activation is protein-specific, we propose that viral DNA and RNA influence the cleavage of P25K by promoting its interaction with K7L at an allosteric site.

Peptide-based Inhibitors and Activity Probes

The K7L preferred-substrate fingerprint was used to design activity-based probes that would be recognized by the K7L protease. For this purpose, the FYAG and EDTIFFAG peptide sequences were functionalized with an N-terminal biotin tag and a C-terminal VME reactive “warhead” (Fig. 6A). The selectivity and efficiency of the biotin-AEEAc-EDTIFFA-VME was directly confirmed in labeling experiments using K7L. To visualize the labeled material, samples were subjected to Western blotting with streptavidin-HRP. Although the short probe failed to visibly label K7L, the longer biotin-AEEAc-EDTIFFA-VME probe at 25 nm was sufficient to label the active site of K7L, consistent with measured inhibition rate constants (Fig. 6B). Blocking of labeling by N-ethylmaleimide indicated that the probe targets the active site (Fig. 6B).

FIGURE 6.

An activity-based probe labels K7L in vitro and in cell extracts. A, shown are chemical formulae of biotinylated peptide-based probes. Inhibition constants were calculated by observing the pseudo-first-order decay of activity (k_obs) in the presence of inhibitor concentration (I). B, upper panel, K7L (0.2 μm) was labeled with the indicated concentrations of short (S) and long (L) probes for 15 min at 25 °C and visualized with streptavidin-HRP by SDS-PAGE. Lower panel, the total enzyme was visualized by Western blotting using an anti-K7L antibody on the same gel. C, shown is labeling K7L and I7L in vaccinia virus-infected (multiplicity of infection = 5, 24 h) BSC-1 cells. Where indicated, recombinant K7L was co-expressed from a transfected pTF7-3 vector. 10 μm biotinylated EDTIFFAG-VME was added to homogenized samples, incubated for 15 min, and loaded onto SDS-PAGE. The biotinylated proteins were visualized with streptavidin-HRP. The black arrows indicate the positions of recombinant K7L and viral I7L on the gel. The open arrow points to the absence of the I7L band in a sample lacking the virus.

To test if the biotin-EDTIFFAG-VME probe labels cellular K7L, we added 10 μm concentrations of probe to lysates of BSC-1 cells infected with vaccinia virus. After a short incubation, the lysates were analyzed by Western blotting with streptavidin-HRP (Fig. 6C) and also with a K7L antibody (supplemental Fig. 4). The results demonstrate that the probe predominantly labeled a 48-kDa band in the virus-infected cells. Future experiments and probe optimization will be required to determine whether the probes described here are able to inhibit viral infectivity and/or replication.

Conclusions

K7L is designated by the MEROPS classification system as a member of the C57 family of peptidase Clan CE (45), a clan that contains processing proteases from adenovirus and African swine fever virus, deSUMOylating enzymes from yeast to humans, and deubiquitinating enzymes from plant and animal pathogenic bacteria (24–29, 35). K7L orthologs are present and presumably required in all orthopoxviruses. By employing transient expression and conditional lethal virus analyses, it has previously been demonstrated that I7L, the vaccinia ortholog of K7L, is required for core protein processing (7, 16, 18). A seminal study used extracts of virally infected cells containing I7L to demonstrate that the enzyme is a cysteine protease able to cleave core protein precursors P4a, P4b, and P25K (17). The authors were unable, however, to produce catalytically active I7L synthesized in vitro, leading them to conclude that a cofactor is required for activity.

Our success in producing catalytically active K7L by expression in E. coli leads us to propose that dimerization of K7L, and by inference I7L, is required for activity and that the in vitro translation system previously employed (17) produced insufficient concentrations of protein to enable significant dimerization. Moreover, our results strongly suggest that K7L does indeed require a cofactor for maximal activity. Although we do not know its identity, we have shown that glycerol and other fold-stabilizing osmolytes have a stimulatory effect. Furthermore, we found that polyanions have differential effects on the activity of the enzyme on protein vis à vis peptide substrates. Together, these characteristics point to a regulation of K7L that is multivariate and more complex than for other Clan CE members, possibly as a consequence of the stringent requirement for its activity to be localized predominantly or exclusively to the condensing virion (7, 16, 18).

Information about other viral proteases from Clan CE is exceedingly limited, and to date only the adenoviral-processing protease adenain has been characterized biochemically and structurally in significant detail (29, 35, 46, 47). Interestingly, some properties of K7L appear to be distinct from other Clan CE members, whereas others are related. Thus, K7L functions most efficiently as a homodimer, which is not a reported requirement for other Clan CE enzymes. However, in common with other Clan CE members, we have provided in vitro evidence that co-factors enhance K7L activity, although the details vary. Thus, adenain is activated by simultaneous binding of two co-factors, the 11-residue fragment of the core protein pIV and viral DNA (29), whereas our data suggest that both RNA and DNA oligonucleotides can enhance K7L activity toward its cognate substrate, P25K.

Our analysis of the extended substrate binding characteristics of K7L helps to explain the specificity of the protease for its predicted substrates, revealing that the presence of negatively charged residues at the P7 and P8 positions greatly enhances cleavage of some substrates. This implies that an exosite interaction distant from the active site is required for optimal cleavage of natural substrates. Consistent with our findings, neither of the peptides that span the P4a cleavage site, Gly⁶¹⁴↓Ser⁶¹⁵ nor the P4a protein itself, is cleaved by recombinant K7L.

K7L has been proposed as a promising target for smallpox therapy (10, 11). Indeed, a small molecule that prevents vaccinia core protein maturation has been reported (10, 11) and may be an inhibitor of I7L. Our biotinylated activity-based probe analysis demonstrates that the K7L is active and targetable in lysates from infected cells. In conclusion, our results provide many novel insights into the mechanism, specificity, and regulation of the K7L viral protease. Our work also provides a solid foundation as well as specific molecular tools (protease production methods, assay design, inhibitor design and synthesis) that can now be used to probe the function of K7L in a cell-based setting.