High metal substitution tolerance of anthrax lethal factor and characterization of its active copper-substituted analogue

Journal of Inorganic Biochemistry 140 (2014) 12–22

Contents lists available at ScienceDirect

Journal of Inorganic Biochemistry journal homepage: www.elsevier.com/locate/jinorgbio

High metal substitution tolerance of anthrax lethal factor and characterization of its active copper-substituted analogue Suet Y. Lo a, Crystal E. Säbel a, Michael I. Webb b,c, Charles J. Walsby b, Stefan Siemann a,⁎ a b c

Department of Chemistry and Biochemistry, Laurentian University, Sudbury, ON P3E 2C6, Canada Department of Chemistry, Simon Fraser University, Burnaby, BC V5A 1S6, Canada Department of Bioengineering, University of California, Berkeley, CA 94720, USA

a r t i c l e

i n f o

Article history: Received 19 March 2014 Received in revised form 27 May 2014 Accepted 16 June 2014 Available online 24 June 2014 Keywords: Anthrax toxin Apoprotein Copper Electron paramagnetic resonance (EPR) Metalloprotease Metal substitution

a b s t r a c t Anthrax lethal factor (LF) is a zinc-dependent metalloendopeptidase and a member of the gluzincin family. The current report demonstrates a high metal substitution tolerance of LF atypical of gluzincins and other zincdependent metalloproteases. Mn2+, Co2+, Ni2+, Cu2+ and Cd2+ were found to reactivate the apoprotein of LF to a level either comparable to or significantly higher than that noted for the native zinc enzyme. The most active form of LF was obtained with Cu2+, a surprising observation since most Cu2+-substituted zinc proteases display very low activity. Cu2+-substituted LF (CuLF), prepared by direct exchange and by apoprotein reconstitution methodologies, displayed a several-fold higher catalytic competence towards chromogenic and fluorogenic LF substrates than native LF. CuLF bound Cu2+ tightly with a dissociation constant in the femtomolar range. The electron paramagnetic resonance spectrum of CuLF revealed the protein-bound metal ion to be coordinated to two nitrogen donor atoms, suggesting that Cu2+ binds to both active site histidine residues. While ZnLF and CuLF (prepared by direct exchange) were capable of killing RAW 264.7 murine macrophage-like cells, apoLF and all metal-reconstituted apoprotein preparations failed to elicit a cytotoxic response. Competition experiments using apoLF/ZnLF mixtures demonstrated the propensity of apoLF to relieve ZnLF-induced cell death, suggesting that both protein forms can compete with each other for binding to protective antigen. The lack of cytotoxicity of apoLF and its metal-reconstituted variants likely originates from structural perturbations in these proteins which might prevent their translocation into the cytoplasm. © 2014 Elsevier Inc. All rights reserved.

1. Introduction Anthrax is an infectious bacterial disease caused by Bacillus anthracis, and has been the topic of considerable interest as an animal and human pathogen, as well as a potential component of biological weapons [1]. Three proteins, protective antigen (PA), edema factor (EF), and lethal factor (LF), collectively known as the anthrax toxin, play a pivotal role in the manifestation of the disease [2]. PA (in its heptameric or octameric form) is a pore-forming protein which mediates the entry of LF and EF from the endosome into the host cell cytosol [3–7]. EF is a calcium and calmodulin-dependent adenylyl cyclase responsible for the generation of supraphysiological amounts of cyclic AMP, thus resulting in a disruption of water homeostasis [8,9]. LF is a zincdependent metalloendopeptidase responsible for the cleavage of most members of the mitogen-activated protein kinase kinase (MAPKK) family of signalling proteins near their N-termini [10–13]. Furthermore, LF has recently been shown to facilitate inflammasome activation and macrophage death by removing an N-terminal segment from NODlike receptor protein 1 (Nlrp1) [14–17]. ⁎ Corresponding author. Tel.: +1 705 675 1151; fax: +1 705 675 4844. E-mail address: [email protected] (S. Siemann).

http://dx.doi.org/10.1016/j.jinorgbio.2014.06.009 0162-0134/© 2014 Elsevier Inc. All rights reserved.

The zinc ion in LF is coordinated to the side chains of His686, His690, and Glu735, and to a water molecule which serves as the nucleophile in the peptide bond hydrolysis reaction [13,18]. In addition, Glu687, which is part of the thermolysin-like HExxH consensus motif [19], has been proposed to serve as a general base in the catalytic mechanism of LF by activating and orienting the zinc-bound water molecule for proper nucleophilic attack on the carbonyl carbon atom of the scissile peptide bond [13]. In view of the aforementioned amino acid residues involved in zinc coordination and water activation, LF can be classified as a gluzincin [19]. Initial investigations on the metal requirement of LF using reconstitution assays involving apoLF have revealed the protein's stringent requirement for Zn2+ for catalytic function. Indeed, apoLF preparations supplemented with other transition metal ions such as Mn2 +, Co2 +, Ni2+ and Cu2+ were found not to regain a significant degree of catalytic competence (typically below 2% of the activity of native ZnLF) [20], although some of these ions were found to effectively compete with Zn2+ for protein binding in earlier radiolabeling studies [21]. Furthermore, Ca2+ and Mg2+ have previously been shown to be required to restore (some of) the activity of apoLF by Zn2+ and other divalent transition metal ions [20,22]. However, recent studies have revealed the requirement for these alkaline earth metals not to be an inherent feature

S.Y. Lo et al. / Journal of Inorganic Biochemistry 140 (2014) 12–22

of LF, but instead to be dependent on the protocol of apoLF preparation. For instance, while apoLF generated by dialysis following exposure of ZnLF to chelators exhibited a dependence on Ca2 + and Mg2 + for enzyme activation by Zn2 + or Co2 +, the apoprotein recovered by centrifugal filtration (subsequent to chelator treatment) was found not to rely on alkaline earth metals for full reconstitution of LF's catalytic activity [23]. The dichotomy of these results has been attributed to the presence of residual chelator in dialyzed (but not centrifugally-filtered) apoLF samples, which can lead to complications in the interpretation of reconstitution data [23]. Although Zn2+ and Co2+ have been found to reconstitute apoLF in the absence of any other supplementary metal ions, the ability of other transition metal ions to restore the catalytic competence of the apoprotein has not been (re)investigated. The current report describes a reinvestigation of the propensity of apoLF to regain its catalytic function upon supplementation with various transition metal ions. Our results demonstrate that, with the exception of Fe2+, all biologically relevant 3d metals are capable of restoring the apoprotein's function. Furthermore, reconstitution of apoLF by Cu2+ was found to lead to an active enzyme which significantly exceeds the catalytic competence of ZnLF, an unusual phenomenon given that the vast majority of Cu2+-substituted zinc proteases display markedly reduced activities compared to those recorded for their native zinc counterparts [24]. Consequently, Cu2 +-substituted LF was prepared by two independent methods (one involving the direct exchange of ZnLF's Zn2 + ion by Cu2 +, the other based on the reconstitution of apoLF with Cu2+), and characterized kinetically and spectroscopically. In addition, the effect of PA supplemented with apo- and metalsubstituted LF on macrophage cell viability was assessed revealing all apoLF preparations (be they metal-deficient or metal-exposed) to be non-cytotoxic. 2. Experimental 2.1. General Chromogenic anthrax lethal factor protease substrate II, S-pNA (Ac-GYβARRRRRRRRVLR-pNA, pNA = para-nitroanilide) was obtained from Biomatik Corporation (Cambridge, ON, Canada). MAPKKide was purchased from List Biological Laboratories (Campbell, CA). All other chemicals were obtained from Sigma-Aldrich (St. Louis, MO). All metal salts (chlorides of Mn2+, Cd2+ and Co2+; sulfates of Zn2+, Fe2+, Ni2+ and Cu2 +) were of analytical reagent grade, and contained ≤ 0.001% Zn2 +. All solutions were prepared using MilliQ ultrapure water (≥18.2 MΩ cm resistivity), and buffers were depleted of trace metals by treatment with Chelex-100 resin. 2.2. Preparation of ZnLF, apoLF and CuLF Wild-type LF (ZnLF) was isolated and purified as outlined previously [25], and contained 1.0 (±0.1) Zn2+/protein molecule. ApoLF was prepared as described earlier with some minor modifications [23]. In brief, ZnLF (10 μM) in Hepes buffer (50 mM, pH 7.4) was exposed to 10 mM ethylenediaminetetraacetic acid (EDTA) and 1 mM dipicolinic acid (DPA) for 48 h at 4 °C prior to removal of excess chelator and concentration of the protein using an Amicon Ultra-15 centrifugal filter (30 kDa MWCO; Millipore, Bedford, MA). The absence of protein-bound Zn2+ and residual chelator (which can interfere with the proper interpretation of metal reconstitution data [26]) following centrifugal filtration was ascertained with 4-(2-pyridylazo)resorcinol (PAR) according to published protocols [23]. ApoLF obtained in this manner contained ≤ 0.03 Zn2 +/LF molecule. CuLF was prepared by two independent methods, one involving the direct exchange of Zn2 + in ZnLF with Cu2+, the other relying on the reconstitution of apoLF with Cu2+. Regarding the first method, 2 mL of purified ZnLF (20 μM) was loaded onto a Q-Sepharose column (5 mL bed volume) equilibrated with Hepes buffer (50 mM, pH 7.0). Following immobilization of the protein,

13

its Zn2+ ion was allowed to exchange with Cu2+ by passing 25 mL of CuSO4 (0.1 mM) in Hepes buffer through the column using a flow rate of 1 mL/min. Excess Cu2+ and released Zn2+ were removed by washing the column with 50 mL of Hepes buffer containing 100 mM NaCl. The metal-exchanged protein was eluted from the column using 20 mL of Hepes buffer containing 350 mM NaCl prior to being concentrated (to 100–180 μM) by Amicon Ultra-15 filtration. In addition to the preparation of CuLF by direct exchange (CuLFde), the protein was obtained by reconstitution of apoLF with Cu2 + (CuLFrec). In a typical procedure, 125 μL of apoLF (150 μM) in Hepes buffer (50 mM, pH 7.4) was reconstituted with 0.96 mol-equivalents of Cu2+ in a step-wise fashion by slowly supplementing the protein with 5 × 5 μL of an aqueous Cu2+ stock solution (720 μM) at room temperature. The step-wise addition of small aliquots of the metal ion solution to apoLF was necessary to avoid the precipitation of Cu(OH)2 in Hepes-buffered solutions at neutral pH [27]. The homogeneity of all LF preparations was assessed by sodium dodecyl sulfate polyacrylamide gel electrophoresis according to the method of Laemmli [28]. The concentration of LF was determined spectrophotometrically at 280 nm using an extinction coefficient of 74,200 M−1 cm−1 [23]. 2.3. Enzyme assays The enzymatic activity of LF was determined spectrophotometrically (at 405 nm) in Hepes buffer (50 mM, pH 7.4) at room temperature as described in the literature, using S-pNA (10 μM) as the chromophoric substrate [23,29]. In a few instances, the catalytic competence of the enzyme was assessed also with the aid of the fluorogenic MAPKKide substrate [30]. MAPKKide is a peptide derived from the MAPKK-2 substrate of LF harbouring an N-terminally linked o-aminobenzoic acid moiety (donor fluorophore) 2,4-dinitrophenyl group, C-terminal to the cleavage site, serving as the acceptor chromophore. In a typical assay, ZnLF (50 nM) or CuLFde (10 nM) in Hepes buffer (50 mM, pH 7.4) was incubated for 1 min at room temperature prior to the addition of MAPKKide (5 μM final concentration), and the increase in fluorescence intensity at 420 nm (following excitation at 320 nm), a consequence of the decrease in intramolecular quenching, was recorded using an OLIS RSM-1000 spectrofluorometer (Bogart, GA). Steady-state kinetic parameters (KM, kcat) for the hydrolysis of S-pNA by CuLFde and CuLFrec (both 10 nM) were estimated by (non-linear least squares) fitting of the initial velocity data to the Michaelis–Menten equation using Grafit 4.0 (Erithacus Software Ltd., Staines, UK), and a Δε405 nm value of 9920 M−1 cm−1 [29]. 2.4. Determination of metal content The Zn2+ content of ZnLF and apoLF was determined with the aid of PAR according to published protocols [23]. The metal content (both Cu2+ and Zn2+) of CuLFde was assessed using the Zincon/EDTA method as described in the literature with minor modifications [31]. In brief, CuLFde at a final concentration of 10 μM was introduced into borate buffer (50 mM, pH 9.0) containing urea (6 M) and Zincon (40 μM). Following incubation of the mixture for 5 min at room temperature, the absorbance at 615 nm (i.e., at the isosbestic wavelength, at which the extinction coefficients of the Zn2+:Zincon and Cu2+:Zincon complexes are identical) was determined. The total concentration of metal (Zn2+ and Cu2+) in the sample was calculated on the basis of the linear relationship between the absorbance at 615 nm and the concentration of metal in a set of Zn2 + (or Cu2 +) standards (0–20 μM) prepared and measured under analogous conditions. The concentration of Cu2+ was subsequently assessed by adding 1 mM EDTA (which is capable of rapidly demetallating Zn2+:Zincon, but not the Cu2+:Zincon complex [31]) to the protein sample, followed by the measurement of the absorbance at 615 nm. The concentration of Zn2+ in the sample was calculated by subtracting the obtained concentration of Cu2+ from the concentration of total metal (i.e., Cu2+ and Zn2+).

14

S.Y. Lo et al. / Journal of Inorganic Biochemistry 140 (2014) 12–22

2.5. Determination of dissociation constant

3. Results

The apparent dissociation constant of CuLF (Kapp d ) was estimated in a fashion analogous to that utilized in the assessment of the Kd values for ZnLF [23] and Co2+-substituted LF [25] using a metal buffer consisting value of CuLF was of Cu2+ (50 μM) and excess DPA. In brief, the Kapp d determined by recording the enzyme's catalytic activity at various concentrations of free Cu2+ in the assay using the following equation:

3.1. Reconstitution of apoLF by divalent transition metal ions

h

i 2þ Cu free : Relative activity ¼  2þ  Cu free þ K app d

ð1Þ

The concentration of free Cu2+ was adjusted to the desired values by changing the (total) concentration of DPA ([DPA]tot). [DPA]tot was determined by solving Eq. (2) and substituting the obtained concentration of free DPA ([DPA]) into Eq. (3), h i 1 Cu2þ β 2 ½DPA þ β1 ½DPA þ @1−  2þ  tot A ¼ 0 Cu free 0

2

h i 2þ ½DPAtot ¼ ½DPA þ Cu

free

  2 β1 ½DPA þ 2β2 ½DPA

ð2Þ

ð3Þ

where [Cu2+]tot denotes the total concentration of Cu2+ in the assay (i.e., 50 μM), and β1 and β2 represent the conditional stability constants of the Cu2 +(DPA) and Cu2+(DPA)2 complexes (at pH 7.4: β1 (Cu) = 1010 M−1; β2 (Cu) = 2 × 1016 M−2 [32,33]), respectively. The catalytic activity of the enzyme was assessed in a typical assay (using 10 μM S-pNA) after exposing the apoprotein (35 nM) to Cu2+ and DPA for 1 value of CuLF was obtained by a non-linear least h at 20 °C. The Kapp d squares fit of the enzyme activities to Eq. (1) using Grafit 4.0. In view value of the non-negligible affinity of Hepes buffer for Cu2+, the Kapp d by a factor was converted to a conditional Kd value (Kcd) by dividing Kapp d of 42 [27]. 2.6. Cytotoxicity assays RAW 264.7 murine macrophage cells were grown at 37 °C in RPMI supplemented with 5% fetal bovine serum and 1% penicillin/ streptomycin. A confluent 10 cm dish of cells was resuspended in 10 mL of medium and 100 μL of this cell suspension was added to wells of a 96-well plate. After overnight incubation, the medium was replaced with 100 μL of fresh medium containing 10−8 M protective antigen (purified as described previously [34]) and various concentrations of LF, and incubated for 4 h at 37 °C in 5% CO2 atmosphere. Cell viability was assessed using the CellTiter 96 Aqueous One Solution Cell Proliferation assay (Promega, Madison, WI). 2.7. EPR spectroscopy EPR measurements were performed on a Bruker EMXplus spectrometer operating at X-band (~ 9.4 GHz) with a PremiumX microwave bridge and HS resonator. Measurements were conducted at 80 K using a Bruker ER 4112HV helium temperature-control system and continuous-flow cryostat. Spectral simulations were performed using the MATLAB-based package EasySpin [35]. Prior to measurement, all protein samples were supplemented with glycerol (glassing agent) and subsequently frozen in liquid nitrogen. Each sample (final volume of 300 μL) contained 60–65 μM CuLF in Hepes buffer (50 mM, pH 7.4) and 30% (v/v) glycerol.

Early investigations into the metal requirement of LF have demonstrated its catalytic function to rely on Zn2 +, and that other divalent transition metal ions such as Mn2+, Co2+, Ni2+ and Cu2+ are incapable of restoring the protein's catalytic competence when added to apoLF preparations [20]. In contrast, two recent reports have shown that substitution of Zn2 + in LF by Co2 + yields an enzyme about twice as active as native ZnLF [23,25], a phenomenon also documented for other Co2 +-substituted zinc proteases of the thermolysin family [24,36,37]. However, the propensity of apoLF to be activated by other divalent transition metal ions was not investigated in these reports. Hence, the current study sought to re-examine the effect of various transition metal ions on the activation of apoLF by exposing the apoprotein to various concentrations of Mn2 +, Fe 2 +, Co2 + , Ni 2 +, Cu2 +, Zn2 + and Cd2 + for 1 h prior to recording enzymatic activities. As shown in Fig. 1, all metal ions except Fe2 + were found to activate apoLF. Indeed, the maximal activities observed with these metals were found to be either comparable to that of ZnLF (for Mn2 + and Cd 2 + ) or significantly higher than that of the Zn 2 +containing protein (for Co2 +, Ni2 + and Cu2 +). With the exception of apoLF supplemented with Mn 2 +, maximal activities were observed at a metal concentration of 10 μM. In the case of Mn2 +, the highest activity recorded required addition of 100 μM of the metal ion. This observation suggests that Mn 2 + binds more weakly to apoLF than the other metals employed in this study, a finding not surprising in view of the metal ion's (far left) position within the Irving–Williams series [38,39]. In the case of Fe2+, the activity of apoLF exposed to this metal ion was virtually indistinguishable from that of the apoprotein (13% residual activity) regardless of whether assays were performed under aerobic or anaerobic conditions (to prevent oxidation of Fe2+ to Fe3+). In this context it is important to note that the apoprotein's residual activity is not a consequence of incomplete metal removal (since apoLF contained less than 0.03 Zn2+ per LF molecule), but rather a result of the protein's picomolar affinity for Zn2+ [23], thus rendering apoLF capable of efficiently scavenging the metal ion from the assay medium (which, even under metal-free conditions, is known to contain nanomolar levels of Zn2+ [24]). To address whether Fe2 + binds to apoLF but is incapable of facilitating substrate hydrolysis, or whether the metal ion is incapable of being accommodated in LF's active site, the effect of Mn2+ supplementation on the catalytic function of Fe2+-exposed apoLF was investigated. The choice of Mn2+ as the competing metal ion in this study is warranted in view of it being capable of binding and reactivating LF, and it being a relatively weak competitor to Fe2 + (a consequence of the metals' position relative to each other in the Irving–Williams series [38]). As Fig. 1 shows, incubation of apoLF with equimolar amounts of Mn2+ and Fe2+ was found to reactivate LF to a degree virtually identical to that noted for its reconstitution by Mn2+ as the sole source of metal. This result strongly suggests that the increase in LF's catalytic competence is solely due to the presence of Mn2+, and that Fe2+ (although expected to possess a higher affinity) is incapable of competing with Mn2 + for LF's metal-binding site. Thus, it would appear that Fe2 + is the only (biologically relevant) 3d transition metal ion which is incapable of binding to apoLF. In the case of Cu2 +, addition of this metal ion to apoLF led to a dramatic increase in the protein's activity, exceeding that of ZnLF by a factor of 4.5. This observation is very unusual for Cu2 +-substituted zinc proteases since Zn2+ replacement by Cu2+ is typically accompanied by a several-fold diminution of enzymatic activity [24,40]. In light of this unusual feature (of Cu2 + being capable of hyperactivating apoLF), a more thorough characterization of CuLF by kinetic and spectroscopic means was initiated.

S.Y. Lo et al. / Journal of Inorganic Biochemistry 140 (2014) 12–22

15

Fig. 1. Reactivation of apoLF by transition metal ions. ApoLF (1 μM; metal content b 0.03 Zn2+/LF) was incubated in Hepes buffer (50 mM, pH 7.4) in the absence and presence of the transition metal ions indicated in the figure for 60 min at room temperature prior to dilution of the mixture to achieve a final concentration of 50 nM with respect to the enzyme. Following the addition of S-pNA (10 μM), the initial rate of the reaction was monitored at 405 nm. Activities are expressed relative to that observed for apoLF supplemented with Zn2+ (1 μM), and represent the mean (±1 standard deviation) of three independent experiments.

3.2. Preparation of CuLF by direct exchange The suitability of two independent methods, direct exchange and apoprotein reconstitution, for the preparation of CuLF in quantities large enough to perform more thorough kinetic, thermodynamic and spectroscopic analyses of the Cu2+-substituted enzyme was explored. The former method of allowing a zinc protein to directly exchange its metal ion with another one is (conceptually) the simplest approach to metal substitution, and has been successfully exploited in the preparation of Co2 +-substituted LF [25]. Hence, preliminary studies were aimed at elucidating the dependence of ZnLF's catalytic activity as a function of the concentration of Cu2+ in the assay medium. As shown in Fig. 2A, exposure of ZnLF to 1 μM Cu2+ led to a significant increase (by ~ 60%) in the rate of substrate hydrolysis compared to that noted in the absence of the metal ion. LF's activity was found to increase with increasing Cu2+ concentration up to 12 μM, an observation indicative of the facile displacement of LF's Zn2+ ion with Cu2+. A further increase in the concentration of Cu2 + led to a significant and abrupt decrease in LF's catalytic function. Indeed, a concentration of 18 μM Cu2 + was found to be sufficient to diminish the activity of LF from 300% (at 12 μM Cu2+) to levels below those documented for ZnLF unexposed to Cu2+ (~50% residual activity). To investigate whether such an inhibitory effect is restricted to Cu2+ or is observed also with Zn2+, the influence of the latter metal ion on ZnLF's activity was recorded. As shown in Fig. 2A, supplementation of ZnLF with Zn2 + was found to inhibit the enzyme with an IC50 (half-inhibitory concentration) value of 12 μM, a value close to that observed for the inhibition of the protein by Cu2+. Since the observed hyperactivation of LF by direct exposure of the zinc enzyme to Cu2+ is indicative of facile metal ion exchange, the preparation of CuLF in larger quantities by such methodology was pursued using the protocols outlined previously for the preparation of Co2 +substituted LF [25]. Hence, ZnLF was immobilized on Q-Sepharose and subjected to an excess of Cu2+. Following removal of any extraneous metal ions, LF was recovered in the 350 mM NaCl fraction. Metal analysis revealed the recovered protein preparations to be virtually devoid of Zn2+ (≤0.03 Zn2+/LF) and to contain 0.74 (±0.05) Cu2+ ions per protein molecule. This result is indicative of the essentially complete Cu2+mediated displacement of Zn2 + from LF. The somewhat surprising

Fig. 2. Influence of Zn2+ and Cu2+ on the activity of ZnLF (A) and CuLF (B). Assays were performed by pre-incubating LF (50 nM for ZnLF, 20 nM for CuLFde) in Hepes buffer (50 mM, pH 7.4) for 30 min at room temperature in the presence of either Zn2+ (diamonds) or Cu2+ (circles) at the indicated concentrations, prior to monitoring the protein's activity with S-pNA (10 μM). Activity values shown are expressed relative to that of the enzyme (ZnLF in (A); CuLFde in (B)) in the absence of supplementary Zn2+ or Cu2+, taken as 100%. Error bars represent ±1 s.d. of three independent experiments.

16

S.Y. Lo et al. / Journal of Inorganic Biochemistry 140 (2014) 12–22

observation of a partially unoccupied metal-binding site in LF can most likely be attributed to the displacement of some LF-bound Cu2 + by Hepes buffer [27] and/or chloride ions present at significant concentrations in both the wash and elution media. Nonetheless, supplementation of these CuLFde preparations with 0.26 (±0.05) mole-equivalents of Cu2 + led to the complete incorporation of the metal ion into the apoprotein, as evidenced by the metal content (1.0 Cu2+/LF) and the accompanying increase (by ~ 35%) in the enzyme's activity (data not shown). Thus, these observations suggest that the fraction of apoprotein in CuLFde can be easily reconstituted by the addition of Cu2+. 3.3. Metal-mediated inhibition of CuLF As mentioned earlier, ZnLF was found to be inhibited by Zn2+ and Cu2+ (see Fig. 2A). To investigate whether CuLFde is prone to inhibition by these metal ions, the dependence of the enzyme's activity on the concentration of Zn2 + and Cu2 + in the assay medium was assessed (Fig. 2B). In the case of Cu2+, the activity of CuLFde was found to remain constant up to a concentration of 10 μM. An increase in the concentration of the metal ion beyond 10 μM led to a significant decrease in the enzyme's catalytic competence as evidenced by the protein being virtually inactive at 20 μM Cu2+. It is interesting to note that the concentration range in which this diminution of LF's activity occurred coincides with that previously documented for the interaction of Cu2 + with ZnLF. In the case of Zn2 +, exposure of CuLFde to this metal ion was found to lead to a marked reduction in the enzyme's activity at sub-micromolar concentrations, suggesting that Zn2 + is capable of displacing Cu2+ from the protein. Furthermore, the observation of an activity plateau between 1 and 10 μM of Zn2+ suggests the exchange of Cu2+ with Zn2+ to be complete above 1 μM of the metal. In addition, the degree of residual activity (~20–25%) recorded within this concentration range is indicative of CuLFde being 4–5 times as active as its Zn2+-containing counterpart, a finding consistent with that documented above for the reconstitution of apoLF with Cu2 + (see Fig. 1). At Zn2 + concentrations greater than 10 μM, the enzyme was found to be inhibited in a manner similar to that encountered with ZnLF (see Fig. 2A). 3.4. Preparation of CuLF by reconstitution of apoLF In addition to the preparation of CuLF from ZnLF by direct exchange, the Cu2+-substituted enzyme was generated by exposure of apoLF to Cu2+ (CuLFrec). Although a straightforward approach, a few comments regarding the supplementation of the apoprotein (in quantities required for spectroscopic investigations; see below) with Cu2+ are warranted. For instance, Cu2 + at concentrations of ~ 100 μM in buffered medium of neutral pH was found to be prone to precipitation in the form of copper hydroxide, a phenomenon well documented in the literature [27]. Hence, direct supplementation of apoLF (at concentrations ≥ 100 μM) with one equivalent of Cu2 + led to precipitation of Cu(OH)2, and thus to incomplete reconstitution of apoLF. To avoid these complications, apoLF was reconstituted with 0.96 eq of Cu2+ in a step-wise fashion by successively supplementing the apoprotein slowly with five aliquots of an aqueous Cu2+ solution.

Table 1 Steady-state kinetic parameters of LF-mediated hydrolysis of S-pNA. Data shown represent the mean of three independent experiments (±1 s.d.). Enzyme

KM (μM)

kcat (s−1)

kcat/KM (M−1 s−1)

Reference

ZnLF CoLF CuLFde CuLFrec

3.5 (±1.4) 2.8 (±1.2) 7.3 (±1.9) 8.0 (±1.6)

3.7 (±0.5) 7.1 (±0.9) 21.2 (±2.0) 23.6 (±2.2)

1.07 × 2.54 × 2.90 × 2.95 ×

106 106 106 106

[23] [25] This work This work

for ZnLF and CoLF, respectively. In addition, while the catalytic efficiency of CuLF was 2.7 times higher than that of ZnLF, those of both the Cu2+and Co2+-substituted enzymes are similar, a phenomenon explicable on the basis of similar increases in both KM and kcat values when substituting Co2+ for Cu2+. To investigate whether the significantly increased catalytic competence of CuLF compared to ZnLF can be observed also with a different, more MAPKK-like substrate, the specific activities for the CuLFde- and ZnLF-mediated hydrolysis of MAPKKide, a fluorogenic peptide designed from the MAPKK-2 template [30], and of S-pNA (control) were recorded. As Table 2 shows, CuLFde was found to possess a 5.7-fold higher specific activity than ZnLF with S-pNA serving as a substrate, a feature which parallels the sixfold increase in the kcat value when substituting Zn2 + for Cu2 + in LF (see Table 1). The specific activity of the ZnLF-mediated hydrolysis of MAPKKide was found to be 0.09 μmol min− 1 mg−1, a value close to that determined previously for a MAPKKide variant bearing a fluoresceinthiocarbamoyl and a DABCYL group serving as the fluorophore and chromophore, respectively [30]. More importantly, substitution of Zn2+ by Cu2+ in LF led to a much larger increase in the specific activity in the case of MAPKKide (12.4-fold) than with S-pNA (5.7-fold). Hence, the highly elevated catalytic competence of CuLFde compared to that of ZnLF is a feature observable with two structurally distinct LF substrates. The apparent dissociation constant of CuLF was determined using a series of metal-buffered solutions with a 1000-fold excess of Cu2 + (over protein) and excess dipicolinic acid serving as the chelator to adjust the concentration of free Cu2+ in the assay medium. As shown value of CuLF was found to be 340 (± 20) fM, and in Fig. 3, the Kapp d was hence 3.5 times lower than that recorded for ZnLF in a previous study [25]. Since Hepes buffer exhibits a weak, yet not negligible, affinvalue for CuLF represents an ity for Cu2+ ions [27,42], the recorded Kapp d overestimation. To correct for the Hepes–Cu2+ interaction, a conditionby a correction factor, al Kd value (Kcd) was calculated by dividing Kapp d which depends on the concentration of Hepes and the pH. In the case of a 50 mM Hepes buffer of pH 7.4 employed in the current study, the correction factor is 42 [27]. Hence, the Kcd value for CuLF was determined to be 8.1 (±0.5) fM.

3.6. EPR spectroscopy As shown in Fig. 4, the EPR spectrum of CuLFrec shows a distinctive Cu signal. Closer examination of the spectrum reveals the presence of two species. The first of these dominates the spectrum (88% of total signal intensity) and was simulated with a rhombic g tensor, 2+

3.5. Steady-state kinetic parameters and dissociation constant of CuLF Following the preparation of CuLF by direct exchange and apoprotein reconstitution, its catalytic parameters with respect to the hydrolysis of S-pNA were determined. As shown in Table 1, the catalytic parameters of both CuLFde and CuLFrec are virtually identical. While the KM values for ZnLF and CoLF are similar, that of CuLF (irrespective of the method of preparation) was found to be approximately 2-fold higher, a phenomenon observed also with Cu2 +-substituted serralysin [41]. Furthermore, the kcat value for CuLF-mediated hydrolysis of S-pNA is approximately six and three times larger than that documented previously

Table 2 Comparison of specific activities of ZnLF and CuLFde. Specific activities (expressed in μmoles of substrate hydrolyzed per minute per mg of LF) were assessed by performing typical enzyme assays as described in the Experimental Section. Data shown represent the mean of three independent experiments (±1 s.d.). Enzyme

ZnLF CuLFde

MAPKKide

S-pNA

Specific activity (μmol min−1 mg−1)

Relative activity

Specific activity (μmol min−1 mg−1)

Relative activity

0.09 (±0.01) 1.12 (±0.21)

1.0 12.4

1.24 (±0.07) 7.12 (±0.66)

1.0 5.7

S.Y. Lo et al. / Journal of Inorganic Biochemistry 140 (2014) 12–22

17

Fig. 3. Determination of the apparent dissociation constant (Kapp d ) of CuLF. ApoLF (35 nM) in Hepes buffer (50 mM, pH 7.4) was supplemented with Cu2+ (50 μM) and excess DPA (concentration calculated using Eqs. (2) and (3)) for 1 h at 20 °C prior to introduction of S-pNA (10 μM) into the assay. The relative activities for CuLF (solid circles) were fit to Eq. (1) by non-linear regression (solid line). Data shown represent the mean of three independent experiments (±1 s.d.). For the sake of comparison, the data documented previously for the Kd value of ZnLF [23] is included in the figure (open circles).

g = [2.33, 2.08, 2.04], consistent with a Cu2+ species. While the hyperfine structure from the Cu2+ ion is evident at g1 and g3 with A1(63Cu) = 14 mT and A3(63Cu) = 7.5 mT, the hyperfine couplings at g2 are too small to be resolved. The EasySpin program used in these simulations [35] includes contributions from both copper isotopes (63Cu, I = 3/2, 69.17%, gN = 1.4849; 65Cu, I = 3/2, 30.83%, gN = 1.5877) [43]. However, the copper hyperfine couplings from each isotope are generally overlapping within the line widths of the spectra shown here due to the isotopes having identical nuclear spins and very similar nuclear g values. A closer examination of the g2 region of the EPR spectrum reveals a superhyperfine structure that is barely resolved within the intrinsic line-width. As shown in the inset in Fig. 4, this hyperfine structure can be resolved in the second-derivative spectrum, revealing a five-line pattern of splittings. The superhyperfine structure cannot be simulated as a 63Cu, 65Cu hyperfine structure. However, it is well simulated by hyperfine splittings from two 14N nuclei (I = 1, 99.632%) [43] with A2(14N) = 1.6 mT (see inset of Fig. 4). Hence, the observation of the rhombic g tensor of this species and the spectral evidence for two Cu2+-coordinating nitrogen ligands, strongly suggest the Cu2+ ion to be coordinated to LF in a non-square-planar geometry via the imidazole moieties of two histidine residues. In addition to the (dominant) spectroscopic feature described above, a small contribution (12% of total signal intensity) from a second signal with uniaxial symmetry (g⊥ = 2.05, g|| = 2.34, A⊥(63Cu) = 0.1 mT, A||(63Cu) = 15 mT) was also observed (see Fig. 4). This signal is consistent with literature reports of Cu2+ in Hepes buffer [27], and our own measurements (data not shown), indicating that a small amount of the metal remains uncoordinated to the protein in CuLFrec. It is important to note that the EPR spectrum of CuLFde (Supplementary Figure S1) was almost identical to that recorded with CuLFrec (shown in Fig. 4). Indeed, the spectrum of CuLFde can be simulated with essentially the same spectroscopic parameters but with a higher proportion of non-protein bound (aqueous) Cu2+ (23% instead of 12% for Cu2+-reconstituted apoLF). 3.7. Cytotoxicity studies To assess whether apoLF and metal-substituted LF preparations are cytotoxic, their effect on the viability of the murine macrophage-like cell line RAW 264.7 was investigated in the presence of PA. As shown

Fig. 4. EPR spectrum and spectral simulation of Cu2+-reconstituted apoLF. The experimental EPR spectrum was recorded at 80 K at a microwave frequency of 9.383 GHz, a microwave power of 2.0 mW, a time constant of 10.24 ms, and a modulation amplitude of 0.8 mT. The spectrum shown represents the average of 60 scans of 30 s each. A) The spectrum of protein-bound Cu2+ was simulated using g = [2.33, 2.08, 2.04], A(63Cu) = [14.0, 0.1, 7.5] mT, A(14 N) = [0, 1.6, 0] mT, line widths = [31.0, 0.2, 20.0] mT, A strain = [5, 0, 16] mT, and a weighting factor of 0.88. B) The spectral simulation of free Cu2+ in solution was performed using g = [2.335, 2.05, 2.05], A(63Cu) = [15.0, 0.1, 0.1] mT, line widths = [7,25,25] mT, A strain = [1, 0, 0] mT, and a weighting factor of 0.12. Inset (bottom): Second derivative of the experimental spectrum in the g2 region. Inset (top): Simulation of the 14N hyperfine structure with A(14 N) = [0, 1.6, 0].

in Fig. 5, both ZnLF and CuLFde were found to reduce cell viability with comparable efficiency in cytotoxicity assays. In contrast, both apoLF and CoLF (prepared by direct exchange) were observed to be noncytotoxic. It is important to note that storage of native ZnLF for 48 h at 4 °C and Amicon filtration (i.e., conditions used in the preparation of apoLF, except for the omission of chelator) did not reduce the protein's cytotoxicity. Hence, the lack of cytotoxicity of apoLF is not an artefact originating from the protocol employed in the preparation of the apoprotein. To investigate whether apoLF is capable of regaining its cytotoxicity upon reinsertion of a metal ion, apoprotein preparations were

18

S.Y. Lo et al. / Journal of Inorganic Biochemistry 140 (2014) 12–22

Fig. 5. Cytotoxicities of metal-depleted and metallated LF preparations. Cell viabilities were determined using the murine macrophage-like cell line RAW 264.7 in the presence of PA (10−8 M) and apoLF, native LF (ZnLF), CuLFde or CoLF (obtained by direct exchange as reported previously [25]) at the concentrations indicated in the figure. Error bars denote ±1 s.d. of three independent experiments.

Under the assumption that both apoLF and ZnLF bind with similar affinities to PA, the addition of the apoprotein to ZnLF in cell killing assays should in principle lead to a diminution of the effective concentration of ZnLF available for binding to PA. For instance, the addition of 5 × 10− 9 M apoLF to 1.25 × 10− 9 M ZnLF ([ZnLF]/([ZnLF] + [apoLF]) = 0.2) would be expected to decrease the effective concentration of ZnLF by a factor of 5, and thus to 2.5 × 10−10 M. Indeed, the degree of cell viability observed under these conditions (~ 50%) is in general agreement with that recorded with 2.5 × 10−10 M ZnLF in the absence of apoLF (~45%; see Fig. 5). Furthermore, supplementation of 3.13 × 10− 10 M ZnLF with 5 × 10− 9 M apoLF ([ZnLF]/([ZnLF] + [apoLF]) = 1/17) would yield an effective ZnLF concentration (1.84 × 10−11 M) insufficient to induce cell death (see Fig. 5), a feature in accordance with such ZnLF/apoLF mixtures being essentially noncytotoxic (see Fig. 6). In summary, the cytotoxicity studies using various metal-containing LF preparations and apoLF indicate that only ZnLF and CuLFde are cytotoxic, and that apoLF, although being capable of binding to PA and hence competing with ZnLF for interaction with the translocase, is incapable of inducing cell death. 4. Discussion

reconstituted with Zn2+, Co2+ and Cu2+ prior to performing cell viability assays in a manner analogous to that described above. Although all reconstituted proteins were active with respect to their ability to hydrolyze S-pNA (with activities being virtually identical to those documented in Fig. 1), none of them were found to be cytotoxic (data not shown). Since binding of LF (via its N-terminal domain) to the oligomeric PA prepore represents a critical step in the translocation of the enzyme into the host cell cytosol [4–6], it was of interest to assess whether the lack of cytotoxicity displayed by apoLF (or its reconstituted forms) is a consequence of the protein not being capable of binding to PA. Hence, the propensity of apoLF to interact with PA was assessed by allowing the apoprotein (at a fixed concentration of 5 × 10− 9 M) to compete with cytotoxic ZnLF for binding to the translocase (Fig. 6). While the addition of an equimolar amount of apoLF to ZnLF led to a small increase in cell viability (from 14% to 25%), 4-fold dilution of ZnLF (to 1.25 × 10− 9 M) resulted in a significant reduction of cell death as evidenced by a doubling of the cell viability to ~ 50%. Further 4-fold dilution of ZnLF (to yield a 16-fold excess of apoLF over the holoprotein) was found to essentially provide complete protection of the cells from the cytotoxic effect of the holoprotein as judged by the increase in cell viability from 38% to 96%. These results clearly suggest that apoLF is capable of not only competing with ZnLF for binding sites on the PA prepore, but also relieving ZnLF-mediated cytotoxicity.

Fig. 6. Effect of apoLF on ZnLF-mediated cytotoxicity. Cell viability was assessed using ZnLF at the indicated concentrations (light grey bars) and ZnLF supplemented with 5 nM apoLF (dark grey bars). Error bars represent ±1 s.d. of three independent experiments.

The substitution of Zn2+ in zinc-dependent enzymes by other transition metal ions has provided an elegant means of gaining insight into the structural and/or mechanistic aspects of zinc protein function [24,44,45]. Among the 3d metal ions, Co2+ and to a lesser extent Cu2+ have been utilized to replace the spectroscopically silent Zn2+ (d10) to generate proteins amenable to, in particular, electronic and EPR spectroscopies [46]. The choice of the metal ion to be used to substitute for Zn2+ often depends on the focus of the study. In the case of spectroscopic investigations pertaining to protein function, the replacement of Zn2+ by Co2+ is preferred (over that by Cu2+) since most Co2+-substituted zinc proteases maintain a high degree of catalytic competence, whereas incorporation of Cu2+ most often leads to severely reduced enzyme activities [24,40]. Although initial investigations on metal substitution of LF have demonstrated a very stringent requirement for Zn2+ [20], more recent studies suggest that apoLF is capable of being reactivated by exposure to Mn2+ [47] and Co2+ [23]. Indeed, the isolated and partially characterized Co2+-containing protein (CoLF) has been found to be twice as active as its native counterpart [25], a finding not unexpected in view of the similarity of its active site architecture to that of the prototypical gluzincin thermolysin [13,48], which has been shown to possess a twofold higher activity with Co2+ [36]. The results of the current study are in accordance with those previously documented for Mn2 + [47] and Co2+ [23,25] in that exposure of apoLF to the former metal ion yielded an enzyme with an activity comparable to that of the zinc protein, whereas supplementation with the latter ion led to a twofold higher activity than that recorded with native LF. More importantly, the observations recorded in this study show that LF possesses an exceptionally high metal substitution tolerance since all metal ions (except for Fe2+) are capable of restoring enzyme function following their exposure to the apoprotein. Indeed, as Table 3 demonstrates, zinc peptidases such as thermolysin and angiotensin-converting enzyme (ACE) have been found to be marginally active with Ni2+, Cu2+ and Cd2+ [36,49]. On the other hand, metal substitution studies on tetanus neurotoxin (TeNT) and botulinum neurotoxin serotype B (BoNT/B) have revealed a somewhat higher substitution tolerance [50,51]. It is interesting to note that the active site of the clostridial neurotoxins TeNT and BoNT closely resembles that of LF by virtue of the presence of a second-shell active site tyrosine residue [52]. This amino acid residue in LF (Tyr728) has been implicated mainly in the protonation of the amine leaving group following peptide bond cleavage [13,52,53], although recent theoretical calculations suggest that it might be involved in stabilizing the tetrahedral intermediate (as has been postulated for thermolysin) [54]. Nonetheless, the observation of LF and TeNT being

S.Y. Lo et al. / Journal of Inorganic Biochemistry 140 (2014) 12–22

19

Table 3 Activities of some metal-substituted zinc proteases of the gluzincin and astacin families. ACE, angiotensin-converting enzyme; LC, light chain; N.R., not reported; N.D., not determined. Enzyme

LF Thermolysin ACE TeNT BoNT A (LC) BoNT B DPP IIIe Astacin Serralysinf

Activities of metal-reconstituted apoproteins (%)a

Reference

Apo

Zn2+

Mn2+

Fe2+

Co2+

Ni2+

Cu2+

Cd2+

13 b1 b5 N.R.c 15 b1 b2 3 b2

100 100 100 100 70 100 100 100 100

100 10 25 120 20 16 N.D. N.D. 4 (35)

16 60b b5 7 0 N.D. N.D. N.D. N.D.

220 200 55 100 N.D. 33 120 140 81 (3300)

220 b2 b5 61 N.D. N.D. 85 3 7 (197)

460 b2 b5 26 N.D. 51 100 37 13 (2600)

105 b2 b5 N.D.d N.D. N.D. N.D. N.D. 6 (17)

This work [36] [49] [50] [56] [51] [33] [57] [41]

a Except for BoNT A (LC), values are expressed as the percentage of the activity of the Zn2+-activated apoprotein. In the case of BoNT A (LC), values are expressed as the percentage of the activity observed with the native zinc protein. b Activity achieved with a 30-fold molar excess of Fe2+. c Although the actual value has not been reported for apoTeNT, the residual activity was found to be low. d Although data on the activity of TeNT following exposure of the apoprotein to Cd2+ is not available, it has been shown that the metal ion (even at 500-fold molar excess) is incapable of preventing the binding of Zn2+ to the apoprotein, an observation indicative of the very low affinity of the 4d transition metal for the protein's metal binding site [55]. e Unlike typical gluzincins harbouring the HExxH motif, DPP III contains a HELLGH motif. Values were estimated from the activity plots shown in Fig. 2 of Ref. [33]. f Activities were recorded with casein serving as the substrate. Values in parentheses denote (percent) activities observed with N-α-benzoyl-L-Arg-p-nitroanilide, and are based on kcat/KM values.

highly activated by Mn2 + has served as the basis for postulating a closer structural relationship of LF to clostridial neurotoxins than to thermolysin [52]. Furthermore, a notable metal substitution tolerance has also been observed with astacin and serralysin (see Table 3), both of which harbour a Tyr residue (Tyr149 and Tyr216 in astacin and serralysin, respectively) serving as a coordinating ligand in the active sites of these enzymes. As outlined earlier, the only metal ion examined which is incapable of binding to and reactivating apoLF appears to be Fe2+, a finding similar to that documented for both TeNT [50,55] and BoNT/A [56]. The metal substitution tolerance for Ni2 + has been found to be low for thermolysin and ACE, but appears to be significant for LF, dipeptidyl peptidase III (DPP III) and TeNT (see Table 3). The results of the current study also indicate that apoLF is activated by Cd2+. Although such observation might be surprising at first glance, a few carboxypeptidases have been shown to be enzymatically functional with Cd2+ [24]. The most striking finding with respect to the functional replacement of Zn2 + in LF is the exceptionally high activity of the Cu2 +-exposed apoprotein. Although the substitution of Zn2+ in many zinc proteases by Cu2+ is accompanied by a significant diminution (often by two orders of magnitude) in the enzymes' catalytic function [24,40], there are notable exceptions such as the gluzincins TeNT, BoNT/B and DPP III (see Table 3). Indeed, Cu2 +-substituted DPP III has been shown to be as active as its native zinc counterpart [33]. Furthermore, astacin and the astacin-like metalloendopeptidase serralysin have been shown to display significant catalytic function with Cu2+ (see Table 3) [57]. The unusually high degree of catalytic competence displayed by Cu2+-supplemented apoLF provided the impetus for a more thorough investigation of the physico-chemical properties of the Cu2 +substituted enzyme. For this purpose, CuLF was prepared via two independent approaches, one using direct exchange and the other relying on the reconstitution of apoLF (at high concentrations) with Cu2+. Regarding the former methodology, the facile exchange of an active site-bound Zn2+ ion with another divalent transition metal has been demonstrated for a variety of metalloproteases including thermolysin [36], BoNT [58,59] and LF [25]. The results of the current study suggest that Cu2+ can easily exchange with Zn2 + (and vice versa) as evidenced by the marked increase in catalytic activity upon exposure of ZnLF to Cu2+. It is interesting to note, however, that both Zn2+ and Cu2+ were found to inhibit LF in the mid-micromolar concentration range. The inhibition of many proteins, including zinc-dependent proteases, by Zn2+ is not an uncommon phenomenon [60]. Indeed, an excess of the metal ion has been found to inhibit many gluzincins including

thermolysin [36], BoNT [56,61], TeNT [61] and DPP III [33]. Based on crystallographic studies of thermolysin and carboxypeptidase A, the molecular origin of Zn2 + inhibition in these enzymes (at least) has been attributed to the metal ion binding in a location vicinal to the active site [48,62,63]. Thus, it is not inconceivable that both Zn2 + and Cu2+ are capable of binding to a second, low affinity inhibitory metal binding site in LF with similar affinities. Such proposal may gain further support from the previous observation of additional metal binding sites in LF [53,64], BoNT [65] and TeNT [50]. The dissociation constant of CuLF (taking into account the interaction of Cu2 + with the buffer) was found to be 8.1 fM. Thus, the Kd value for CuLF is about 150 times and 9300 times lower than that determined for ZnLF (1.2 pM; [23]) and CoLF (75 pM; [25]), respectively. Hence, the affinity of the three transition metal ions for LF's active site increases in the order Co2+ b Zn2 + b Cu2+, a trend also documented for DPP III [33], and a finding in accordance with the Irving–Williams series [38]. EPR-spectroscopic investigations of CuLF have revealed a rhombic spectrum with g = [2.33, 2.08, 2.04], A(63Cu) = [14.0, 0.1, 7.5 mT], and a superhyperfine structure in the g2 region indicative of the coordination of two nitrogen donor centres (His686 and His690) to the metal ion. These spectral parameters are typical of type-II copper sites [66], and are similar to those reported for a variety of related enzymes. A particularly noteworthy observation derives from a single-crystal EPR study of Cu2 +-substituted carboxypeptidase A, which has reported principal tensor values of g = [2.31, 2.10, 2.04], and A(63Cu) = [14.4, ~0, 5.0 mT] [67], each of which are very similar to those of CuLF. As in the current study, A2 of the copper hyperfine coupling was too small to be resolved. A number of frozen-solution EPR studies of related Cu2+-substituted proteins also show spectra with similar g values and copper hyperfine coupling constants. In these reports, the EPR spectra have generally been described as uniaxial with g⊥ values close to the average of g1 and g2 in our study. As shown in Table 4, relevant examples include Cu2+-substituted carboxypeptidase A [68,69], thermolysin [70], DPP III [33], and astacin [57,71], which was also noted to have a rhombic splitting of g⊥. Comparison of the g values and values of A|| of the Cu2+-substituted enzymes above indicates that CuLF is more similar to carboxypeptidase A and astacin than to thermolysin and DPP III. The latter two enzymes have been proposed to harbour a distorted tetrahedral environment around the Cu2 + centre [68,70]. In contrast, the crystal structures of Zn2 +, Co2 + and Cu2 +-astacin show a five-coordinate metal centre with three histidine residues, one tyrosine residue, and one water ligand [57]. Furthermore, a higher resolution crystal structure of zinc

20

S.Y. Lo et al. / Journal of Inorganic Biochemistry 140 (2014) 12–22

Table 4 EPR spectroscopic parameters of Cu2+-substituted zinc proteases. Enzyme

g||

g⊥

A|| (mT)

A⊥ (mT)

Reference

Carboxypeptidase A Carboxypeptidase A Thermolysin DPP III Astacina Astacin

2.33 2.34 2.26 2.27 2.29 2.29

2.06 2.05 2.06 2.06 2.05 2.07

14.4 15.2 17.4 17.9 14.9 12.0

1.5 – – – – –

[68] [69] [70] [33] [57] [71]

CuLF

g1

g2

g3

A1

A3

2.33

2.08

2.04

14.0

7.5

This work

a

Although rhombic g values were not reported, the EPR spectrum of astacin was observed to have a rhombic splitting of g⊥.

carboxypeptidase A has revealed the metal ion to be pentacoordinated to two histidine residues, one water molecule and a glutamate residue being bound in a bidentate fashion [72] (see Supplementary Figure S2 for a comparison of the active site structures of LF, carboxypeptidase A and Cu2 +-astacin). Although a structure for its Cu2 + derivative has not been reported, it is not inconceivable that the coordination geometry is maintained upon the exchange of Zn2+ with Cu2+ in carboxypeptidase A. Hence, given the strong similarities in the EPR spectra of astacin, carboxypeptidase A and CuLF, it is tempting to suggest that the Cu2+ ion in CuLF is five-coordinate, a coordination (number) identical to that previously proposed for its Co2+-substituted analogue [25]. As outlined previously, the replacement of Zn2 + by Cu2 + in zinc proteases typically results in a profound reduction of catalytic function [24,40], a feature which has been often attributed to the preference of Cu2+ for planar or distorted octahedral/tetrahedral coordination ([69] and references therein). On the other hand, the same notion of coordinative distortion has been employed to argue in favour of the significant activity observed with Cu2 +-substituted astacin and serralysin, in which the metal ion is pentacoordinated in the resting state of these enzymes [41,57]. Here, the binding of the substrate is thought to lead to a distorted hexacoordinate transition state, which is stabilized as a consequence of the Jahn–Teller effect, thus lowering the activation energy of the peptide bond hydrolysis reaction [41]. As such, it is not inconceivable that the high activity of CuLF originates from a change from a five-coordinate Cu2 + centre in the resting state to a six-coordinate Jahn Teller-stabilized environment in the transition state. It is pertinent to note, however, that the coordination number/structure in itself might not be a factor governing the enzymatic activity of zinc proteases such as carboxypeptidases and thermolysin [40,70]. Instead, it has been proposed that the lower activities observed with some metal-substituted enzymes might be a consequence of subtle changes in protein structure, improperly oriented functional groups and/or a general lack of flexibility of the metal-binding motif [40,69]. Indeed, analysis of the crystal structures of Co2+, Mn2+, Fe2+ and Cd2+-substituted thermolysin has revealed a pentacoordinate environment for the first three metals and hexacoordination for Cd2+ [48]. While CoTL and FeTL displayed significant activities, the Cd2+ and Mn2 + analogues were found to be only marginally active (see Table 3), thus underscoring that the coordination number alone does not determine TL activity. However, the structures of CdTL and MnTL were found to be similar in view of a significant displacement of Glu143 (general base; equivalent to Glu687 in LF) from its position encountered in ZnTL, CoTL and FeTL [48]. Hence, it appears that metal-induced variations in the position of active site residue(s) are “regulating” the activity of metal-substituted TLs. Furthermore, studies on a Cu2+-substituted deletion mutant of DPPIII (del-DPPIII), which is inactive, have shown a similar glutamate side chain displacement, whereas such movement did not occur in Cu2 +-substituted DPPIII (which is active) [73]. The lack of activity in the del-DPPIII variant has been attributed to a high degree of rigidity of the Cu2+ coordination geometry, while the high metal substitution tolerance of Cu2+-substituted DPPIII (see Table 3) has been interpreted to arise from the flexibility of

the metal-binding motif [40,73]. Whether such conformational flexibility is the underlying cause for the exceptional metal substitution tolerance of LF and the high activity of its Cu2 +-substituted variant remains to be established. The cytotoxicity of CuLF (obtained by direct exchange) was found to be similar to that of ZnLF, whereas CoLF was found to be non-cytotoxic despite its inherently higher catalytic competence (compared to ZnLF) in enzyme assays. The lack of cytotoxicity of the Co2 +-substituted form of LF might be explicable on the basis of the lower affinity of Co2 + (compared to Zn2 +) for LF. Indeed, it has recently been shown that under the low pH conditions encountered in endosomes, a significant proportion of Zn2+ dissociates from LF, presumably due to the protonation of the Zn2+-binding imidazole moieties of His686 and His690 in the protein's active site [74]. Given the 60-fold lower affinity of the enzyme for Co2+ at neutral pH [25], it is not inconceivable that most (if not all) of the Co2 + ions are dislodged from PA pore-bound LF molecules at low endosomal pH prior to protein translocation. In the case of CuLF, significant dissociation of Cu2+ from the protein is not to be expected in view of the high (femtomolar) affinity of LF for this metal ion. It is interesting to note that while CuLF was found to be much more active than its native zinc counterpart, both proteins displayed similar cytotoxicities. Although the molecular origin underlying this observation remains to be established, it is plausible that an increased catalytic competence in enzyme assays does not necessarily translate into elevated cytotoxicities, a phenomenon governed by multiple complex biochemical processes (e.g., translocation, multiple substrate targets, downstream effects following MAPKK and/or Nlrp1 cleavage). Furthermore, the possibility of the Cu2 + ion in LF being displaced by Zn2+ during the duration of the cytotoxicity assays appears unlikely since (i) Cu2+ binds more tightly than Zn2+, and (ii) this process should have been observable also with CoLF for which facile direct exchange has been demonstrated previously [25]. The observation of CoLF being non-cytotoxic, hence, suggests that such exchange did not occur. The most perplexing finding of the cytotoxicity investigations concerns the lack of any cytotoxic effect with apoLF and its metalreconstituted variants. Based on competition studies with ZnLF and apoLF (see Fig. 6), it is apparent that the latter protein is capable of binding to PA, and of effectively competing with ZnLF for binding sites on the pore. In light of these observations, (at least) two possible causes for the relief of ZnLF-mediated cytotoxicity by apoLF can be envisaged (see Fig. 7). For instance, following binding of apoLF and ZnLF to the PA pore, both protein components might be translocated, but only ZnLF can contribute to cytotoxicity because of the apoprotein's inability to bind cytoplasmic Zn2+ (Fig. 7B). However, such possibility appears to be remote since cytosolic free Zn2+ pools are believed to be between tens and hundreds of picomolar [75,76], concentrations expected to be sufficient to reconstitute (at least most of) LF given its high affinity for the metal ion (Kd = 1 pM; [23]). Furthermore, if the failure to reinsert Zn2 + in the cytoplasm was the sole reason for the lack of any cytotoxic effect of apoLF, one would have expected (catalytically functional) Zn2 + or Cu2 +-reconstituted apoLF preparations to be cytotoxic (an effect not observed). Alternatively, apoLF, although capable of binding to PA, might not be translocated as a consequence of demetallation-induced structural perturbations. As such, apoLF might remain bound to the pore while translocation of neighbouring ZnLF molecules proceeds in the typical manner (Fig. 7C). The data obtained in the current study indicates that at a 4:1 molar ratio of apoLF and ZnLF, the observed degree of cytotoxicity is very similar to that of 5-fold diluted ZnLF in the absence of apoLF. This result suggests the diminution of cytotoxicity to be a consequence of the decrease in the effective concentration of ZnLF by a factor of five, a proposal in agreement with model C depicted in Fig. 7. As outlined earlier, both apoLF and (catalytically competent) reconstituted apoLF preparations were found to be non-cytotoxic. It thus appears that, once having undergone demetallation, LF is incapable

S.Y. Lo et al. / Journal of Inorganic Biochemistry 140 (2014) 12–22

21

Fig. 7. Schematic representation of plausible origins of the relief of ZnLF-mediated cytotoxicity by apoLF. For the sake of clarity, the octameric PA pore is shown with all four LF binding sites occupied. A) The pore is fully occupied by ZnLF. All four protein molecules are translocated and contribute to cytotoxicity. B) ZnLF and apoLF (when present) are translocated through the pore. The diminution of cytotoxicity is based on the inability of cytosolic zinc pools to reconstitute apoLF to the functional holoform. C) ZnLF is translocated, whereas apoLF remains bound to the pore. The effect of apoLF on the cytotoxicity in this case is the same as that shown in panel B.

of killing cells even if re-metallated. It is not unreasonable to assume that demetallation induces subtle structural changes in LF which persist even after protein re-metallation. Although the nature of such potential structural perturbations was not investigated in the current study, apoLF has previously been shown to possess a slightly lower melting point than ZnLF [53]. In addition, recent studies on the C-terminal core protease domain of LF have revealed that the demetallated form contains a slightly lower α-helical content and that binding of Zn2+ leads to an increased robustness of the enzyme active site [77]. Nonetheless, NMR and molecular dynamics simulations on the same protein have demonstrated that this core domain has a fold essentially identical to that observed in crystal structures of LF [78]. Given the computational and spectroscopic evidence, it appears unlikely that potential structural perturbations in LF's active site following demetallation (and subsequent remetallation) could be considered the cause of the lack of cytotoxicity, a notion consistent with the observed complete regain of catalytic competence of these protein preparations in vitro. In light of these considerations, it is plausible that demetallation-induced alterations in LF's structure occur at a location remote from the active site (e.g., the PA-binding domain). It is also not inconceivable that bivalent metal ions (other than the active site Zn2+ ion) stabilize the structure of the enzyme in solution, and that the removal of these ions leads to structural perturbations which impair LF translocation. It is interesting to note that in the case of LF's close relative BoNT/A, initial studies had shown that Zn2+ removal followed by reinsertion of the metal led to a significant loss of biological (in vivo) activity due to spectroscopically observed irreversible tertiary structural changes [65,79]. However, later investigations have revealed that the catalytic domain of apoBoNT/A can be internalized and becomes reactivated by intracellular Zn2+, suggesting that the removal of the metal ion is not sufficient to block internalization or intracellular Zn2+ insertion [59]. Furthermore, crystallographic studies on apoBoNT/B have provided evidence that the tertiary structure of the protein is maintained upon removal of Zn2 + from the holoform of the enzyme [80]. Similar to the situation prevailing for BoNT, Zn2 + does not appear to play a role in the determining the structure of TeNT since both apo- and holoforms were shown to have virtually identical physico-chemical properties [81].

5. Conclusions The current investigations on LF have revealed a high metal substitution tolerance atypical of that encountered with gluzincins and other related zinc proteases. The most catalytically active metal-reconstituted protein was obtained with Cu2 +. CuLF was found to be cytotoxic, bound the metal ion with femtomolar affinity, and displayed an EPR spectrum indicative of Cu2 + being bound to the active site in a fivecoordinate environment. Finally, apoLF and all its metal-reconstituted forms were non-cytotoxic. ApoLF effectively relieved the cytotoxic effect imposed by ZnLF, and as such might be regarded as a translocation inhibitor. The molecular origin(s) underlying these phenomena are unknown, but might be related to subtle structural changes in apoLF and its metal-replaced derivatives preventing translocation through the PA pore. Studies are currently underway to probe, in more detail, the effects of de- and remetallation on LF's structural integrity. Abbreviations ACE apo BoNT CoLF CuLF CuLFde CuLFrec DPA DPP EF LF MAPKK Nlrp1 PA PAR S-pNA TeNT TL ZnLF

angiotensin-converting enzyme apoprotein botulinum neurotoxin cobalt-substituted LF copper-substituted LF CuLF obtained by direct exchange CuLF obtained by apoLF reconstitution dipicolinic acid dipeptidyl peptidase edema factor lethal factor mitogen-activated protein kinase kinase NOD-like receptor protein 1 protective antigen 4-(2-pyridylazo)resorcinol lethal factor protease substrate II tetanus neurotoxin thermolysin zinc-containing (wild-type) LF

22

S.Y. Lo et al. / Journal of Inorganic Biochemistry 140 (2014) 12–22

Acknowledgements We are indebted to Dr. J. Mogridge (Department of Laboratory Medicine and Pathobiology, University of Toronto) for performing the cytotoxicity assays. This work was supported by the Laurentian University Research Fund (LURF), and by the Natural Sciences and Engineering Research Council of Canada (NSERC, grant no. RGPIN/312629-2013) in the form of Discovery Grants (to C.J.W. and S.S.). References [1] M. Mock, A. Fouet, Annu. Rev. Microbiol. 55 (2001) 647–671. [2] R.J. Collier, J.A. Young, Annu. Rev. Cell Dev. Biol. 19 (2003) 45–70. [3] C. Petosa, R.J. Collier, K.R. Klimpel, S.H. Leppla, R.C. Liddington, Nature 385 (1997) 833–838. [4] J.A. Young, R.J. Collier, Annu. Rev. Biochem. 76 (2007) 243–265. [5] R.J. Collier, Mol. Aspects Med. 30 (2009) 413–422. [6] K.L. Thoren, B.A. Krantz, Mol. Microbiol. 80 (2011) 588–595. [7] J.G. Bann, Protein Sci. 21 (2012) 1–12. [8] S.H. Leppla, Proc. Natl. Acad. Sci. U. S. A. 79 (1982) 3162–3166. [9] C.L. Drum, S.Z. Yan, J. Bard, Y.Q. Shen, D. Lu, S. Soelaiman, Z. Grabarek, A. Bohm, W.J. Tang, Nature 415 (2002) 396–402. [10] N.S. Duesbery, C.P. Webb, S.H. Leppla, V.M. Gordon, K.R. Klimpel, T.D. Copeland, N.G. Ahn, M.K. Oskarsson, K. Fukasawa, K.D. Paull, G.F. Vande Woude, Science 280 (1998) 734–737. [11] G. Vitale, R. Pellizzari, C. Recchi, G. Napolitani, M. Mock, C. Montecucco, Biochem. Biophys. Res. Commun. 248 (1998) 706–711. [12] G. Vitale, L. Bernardi, G. Napolitani, M. Mock, C. Montecucco, Biochem. J. 352 (2000) 739–745. [13] A.D. Pannifer, T.Y. Wong, R. Schwarzenbacher, M. Renatus, C. Petosa, J. Bienkowska, D.B. Lacy, R.J. Collier, S. Park, S.H. Leppla, P. Hanna, R.C. Liddington, Nature 414 (2001) 229–233. [14] J.L. Levinsohn, Z.L. Newman, K.A. Hellmich, R. Fattah, M.A. Getz, S. Liu, I. Sastalla, S.H. Leppla, M. Moayeri, PLoS Pathog. 8 (2012) e1002638. [15] K.A. Hellmich, J.L. Levinsohn, R. Fattah, Z.L. Newman, N. Maier, I. Sastalla, S. Liu, S.H. Leppla, M. Moayeri, PLoS One 7 (2012) e49741. [16] B.C. Frew, V.R. Joag, J. Mogridge, PLoS Pathog. 8 (2012) e1002659. [17] M. Moayeri, I. Sastalla, S.H. Leppla, Microbes Infect. 14 (2012) 392–400. [18] B.E. Turk, T.Y. Wong, R. Schwarzenbacher, E.T. Jarrell, S.H. Leppla, R.J. Collier, R.C. Liddington, L.C. Cantley, Nat. Struct. Mol. Biol. 11 (2004) 60–66. [19] W.N. Lipscomb, N. Sträter, Chem. Rev. 96 (1996) 2375–2434. [20] S.E. Hammond, P.C. Hanna, Infect. Immun. 66 (1998) 2374–2378. [21] K.R. Klimpel, N. Arora, S.H. Leppla, Mol. Microbiol. 13 (1994) 1093–1100. [22] J. Kim, Y.G. Chai, M.Y. Yoon, Biochem. Biophys. Res. Commun. 313 (2004) 217–222. [23] C.E. Säbel, S. St-Denis, J.M. Neureuther, R. Carbone, S. Siemann, Biochem. Biophys. Res. Commun. 403 (2010) 209–213. [24] D.S. Auld, Methods Enzymol. 248 (1995) 228–242. [25] C.E. Säbel, R. Carbone, J.R. Dabous, S.Y. Lo, S. Siemann, Biochem. Biophys. Res. Commun. 416 (2011) 106–110. [26] D.S. Auld, Methods Enzymol. 158 (1988) 110–114. [27] M. Sokołowska, W. Bal, J. Inorg. Biochem. 99 (2005) 1653–1660. [28] U.K. Laemmli, Nature 227 (1970) 680–685. [29] F. Tonello, M. Seveso, O. Marin, M. Mock, C. Montecucco, Nature 418 (2002) 386. [30] M. Forino, S. Johnson, T.Y. Wong, D.V. Rozanov, A.Y. Savinov, W. Li, R. Fattorusso, B. Becattini, A.J. Orry, D. Jung, R.A. Abagyan, J.W. Smith, K. Alibek, R.C. Liddington, A.Y. Strongin, M. Pellecchia, Proc. Natl. Acad. Sci. U. S. A. 102 (2005) 9499–9504. [31] C.E. Säbel, J.M. Neureuther, S. Siemann, Anal. Biochem. 397 (2010) 218–226. [32] R.M. Tichane, W.E. Bennett, J. Am. Chem. Soc. 79 (1957) 1293–1296. [33] J. Hirose, H. Iwamoto, I. Nagao, K. Enmyo, H. Sugao, N. Kanemitu, K. Ikeda, M. Takeda, M. Inoue, T. Ikeda, F. Matsuura, K.M. Fukasawa, K. Fukasawa, Biochemistry 40 (2001) 11860–11865. [34] A. Kassam, S.D. Der, J. Mogridge, Cell. Microbiol. 7 (2005) 281–292.

[35] [36] [37] [38] [39] [40] [41] [42] [43] [44] [45] [46] [47]

[48] [49] [50] [51] [52] [53] [54] [55] [56] [57] [58] [59] [60] [61] [62] [63] [64] [65] [66] [67] [68] [69] [70] [71] [72] [73] [74] [75] [76] [77] [78] [79] [80] [81]

S. Stoll, A. Schweiger, J. Magn. Reson. 178 (2006) 42–55. B. Holmquist, B.L. Vallee, J. Biol. Chem. 249 (1974) 4601–4607. Y. Hashida, K. Inouye, J. Biochem. 141 (2007) 843–853. H. Irving, R.J.P. Williams, J. Chem. Soc. (1953) 3192–3210. S.I. Gorelsky, L. Basumallick, J. Vura-Weis, R. Sarangi, K.O. Hodgson, B. Hedman, K. Fujisawa, E.I. Solomon, Inorg. Chem. 44 (2005) 4947–4960. K.M. Fukasawa, T. Hata, Y. Ono, J. Hirose, J. Amino Acids 2011 (2011) 574816. H.I. Park, L.J. Ming, J. Biol. Inorg. Chem. 7 (2002) 600–610. H.E. Mash, Y.-P. Chin, L. Sigg, R. Hari, H. Xue, Anal. Chem. 75 (2003) 671–677. J.A. Weil, J.R. Bolton, Journal (2007) Pages. G. Parkin, Chem. Commun. (Camb.) (2000) 1971–1985. G. Parkin, Chem. Rev. 104 (2004) 699–767. I. Bertini, C. Luchinat, Adv. Inorg. Biochem. 6 (1984) 71–111. M.Y. Zakharova, N.A. Kuznetsov, S.A. Dubiley, A.V. Kozyr, O.S. Fedorova, D.M. Chudakov, D.G. Knorre, I.G. Shemyakin, A.G. Gabibov, A.V. Kolesnikov, J. Biol. Chem. 284 (2009) 17902–17913. D.R. Holland, A.C. Hausrath, D. Juers, B.W. Matthews, Protein Sci. 4 (1995) 1955–1965. P. Bünning, J.F. Riordan, J. Inorg. Biochem. 24 (1985) 183–198. F. Tonello, G. Schiavo, C. Montecucco, Biochem. J. 322 (1997) 507–510. C.C. Shone, A.K. Roberts, Eur. J. Biochem. 225 (1994) 263–270. F. Tonello, C. Montecucco, Mol. Aspects Med. 30 (2009) 431–438. F. Tonello, L. Naletto, V. Romanello, F. Dal Molin, C. Montecucco, Biochem. Biophys. Res. Commun. 313 (2004) 496–502. G.A. Dalkas, A. Papakyriakou, A. Vlamis-Gardikas, G.A. Spyroulias, Protein Sci. 18 (2009) 1774–1785. J.F. Wright, M. Pernollet, A. Reboul, C. Aude, M.G. Colomb, J. Biol. Chem. 267 (1992) 9053–9058. S.A. Ahmed, L.A. Smith, J. Protein Chem. 19 (2000) 475–487. F.-X. Gomis-Rüth, F. Grams, I. Yiallouros, H. Nar, U. Küsthardt, R. Zwilling, W. Bode, W. Stöcker, J. Biol. Chem. 269 (1994) 17111–17117. G. Schiavo, O. Rossetto, A. Santucci, B.R. DasGupta, C. Montecucco, J. Biol. Chem. 267 (1992) 23479–23483. L.L. Simpson, A.B. Maksymowych, S. Hao, J. Biol. Chem. 276 (2001) 27034–27041. W. Maret, Biometals 26 (2013) 197–204. P. Foran, C.C. Shone, J.O. Dolly, Biochemistry 33 (1994) 15365–15374. K.S. Larsen, D.S. Auld, Biochemistry 28 (1989) 9620–9625. M. Gomez-Ortiz, F.X. Gomis-Rüth, R. Huber, F.X. Avilés, FEBS Lett. 400 (1997) 336–340. S.K. Kochi, G. Schiavo, M. Mock, C. Montecucco, FEMS Microbiol. Lett. 124 (1994) 343–348. F.N. Fu, R.B. Lomneth, S. Cai, B.R. Singh, Biochemistry 37 (1998) 5267–5278. I. Bertini, A. Scozzafava, Met. Ions Biol. Syst. 12 (1981) 31–74. L.C. Dickinson, J.C.W. Chien, J. Am. Chem. Soc. 105 (1983) 6481–6487. R.C. Rosenberg, C.A. Root, P.K. Bernstein, H.B. Gray, J. Am. Chem. Soc. 97 (1975) 2092–2096. A. Schäffer, D.S. Auld, Biochemistry 25 (1986) 2476–2479. I. Bertini, G. Canti, H. Kozlowski, A. Scozzafava, J. Chem. Soc. Dalton Trans. (1979) 1270–1273. H.I. Park, L.J. Ming, J. Inorg. Biochem. 72 (1998) 57–62. H.M. Greenblatt, H. Feinberg, P.A. Tucker, G. Shoham, Acta Crystallogr. D 54 (1998) 289–305. J. Hirose, H. Kamigakiuchi, H. Iwamoto, H. Fujii, M. Nakai, M. Takenaka, R. Kataoka, M. Sugahara, S. Yamamoto, K.M. Fukasawa, Arch. Biochem. Biophys. 431 (2004) 1–8. L.H. Montpellier, S. Siemann, FEBS Lett. 587 (2013) 317–321. R.A. Colvin, W.R. Holmes, C.P. Fontaine, W. Maret, Metallomics 2 (2010) 306–317. A. Krężel, W. Maret, J. Biol. Inorg. Chem. 11 (2006) 1049–1062. P.V. Gkazonis, G.A. Dalkas, C.T. Chasapis, A. Vlamis-Gardikas, D. Bentrop, G.A. Spyroulias, Biochem. Biophys. Res. Commun. 396 (2010) 643–647. G.A. Dalkas, C.T. Chasapis, P.V. Gkazonis, D. Bentrop, G.A. Spyroulias, Biochemistry 49 (2010) 10767–10769. L. Li, B.R. Singh, Biochemistry 39 (2000) 10581–10586. S. Eswaramoorthy, D. Kumaran, J. Keller, S. Swaminathan, Biochemistry 43 (2004) 2209–2216. V. De Filippis, L. Vangelista, G. Schiavo, F. Tonello, C. Montecucco, Eur. J. Biochem. 229 (1995) 61–69.