Engineering of Ancestors as a Tool to Elucidate Structure, Mechanism, and Specificity of Extant Terpene Cyclase

Structural information is crucial for understanding catalytic mechanisms and to guide enzyme engineering efforts of biocatalysts, such as terpene cyclases. However, low sequence similarity can impede homology modeling, and inherent protein instability presents challenges for structural studies. We hypothesized that X-ray crystallography of engineered thermostable ancestral enzymes can enable access to reliable homology models of extant biocatalysts. We have applied this concept in concert with molecular modeling and enzymatic assays to understand the structure activity relationship of spiroviolene synthase, a class I terpene cyclase, aiming to engineer its specificity. Engineering a surface patch in the reconstructed ancestor afforded a template structure for generation of a high-confidence homology model of the extant enzyme. On the basis of structural considerations, we designed and crystallized ancestral variants with single residue exchanges that exhibited tailored substrate specificity and preserved thermostability. We show how the two single amino acid alterations identified in the ancestral scaffold can be transferred to the extant enzyme, conferring a specificity switch that impacts the extant enzyme’s specificity for formation of the diterpene spiroviolene over formation of sesquiterpenes hedycaryol and farnesol by up to 25-fold. This study emphasizes the value of ancestral sequence reconstruction combined with enzyme engineering as a versatile tool in chemical biology.


S3
For purification of the variant library, a 96-well plate-based centrifugation protocol was used. In specific, cell pellets from 5 mL cultivation volume were resuspended in 600 µl B-PER Complete Bacterial Protein Extraction Reagent supplemented with 20 mM imidazole and incubated for 15 min at room temperature on a plate shaker (750 rpm). Lysates were cleared by 20 min centrifugation (4 °C, 2300 x g) and transferred to individual wells of a His MultiTrap Fast Flow plate (GE Healthcare Life Sciences, Sweden) that was previously equilibrated with wash buffer. The plates were incubated with the lysate for 20 min at room temperature on a plate shaker (250 rpm) and the flowthrough was removed by 4 min centrifugation (4 °C, 100 x g). Washing was performed twice by adding 500 µl of wash buffer and removing wash-fractions by centrifugation (500 x g, 2 min, 4 °C). Proteins were eluted by adding 150 µl of elution buffer and incubating the plates at room temperature for 3 min. Elution fractions were then collected in a fresh collection plate by centrifugation (500 x g, 2 min, 4 °C). The eluate was transferred to a PD MultiTrap G-25 plate (GE Healthcare Life Sciences, Sweden) that was pre-equilibrated with desalting buffer (50 mM Tris-HCl, pH 7.4 at 25 °C). Desalted fractions were obtained by adding 50 µl of desalting buffer to the samples and collecting the eluate in a fresh collection plate by centrifugation (1 min, 800 x g, 4 °C). Protein concentrations were determined spectrophotometrically at 280 nm as described above.
All protein preparations were kept at 4 °C upon purification until used for activity assays or flash-frozen in liquid nitrogen for crystallographic studies.

Thermostability Measurements
Thermostability measurements were performed by nano differential scanning fluorimetry (nanoDSF) on a Prometheus NT.48 nanoDSF instrument (NanoTemper Technologies, Germany). Desalted protein solutions (20 µM for SvS-A2 and SvS-WT, 4-6 µM for selected variants) were soaked into a glass capillary and thermal unfolding was recorded from 20 °C to 95 °C at a rate of 1 °C min -1 (excitation power of 25%) by monitoring the ratio of protein autofluorescence at 330 nm and 350 nm. Melting points were determined as the point equivalent to the maximum of the first derivative of the change of the 330 °C / 350 °C ratio.

Size Exclusion Chromatography with Multi Angle Light Scattering detection (SEC-MALS)
Purified and desalted enzymes (50 µl of 3-8 mg mL -1 ) were analyzed by size exclusion chromatography on a Superdex 200 Increase 10/300 GL column (GE Healthcare, Sweden) in an ÄKTAmicro chromatography system (GE Healthcare, Sweden) at a flow rate of 0.75 mL min -1 using filtered Phosphate Buffered Saline (pH 7.4) for elution. The chromatography system was coupled to a miniDAWN multi angle light scattering detector (Wyatt Technology, USA) for molecular weight detection.
The elution profile was determined based on the differential refractive index and molecular weights were determined using ASTRA 5.3.4.20 software (Wyatt Technology, USA).

Protein Structure Determination
The crystallization screens were carried out at 14.6 mg mL -1 protein concentration using a Mosquito crystallization robot and commercial screens (PACT, JSGC+ obtained from Qiagen and Wizard screen from Emerald) in sitting drop format by the vapor diffusion method. For crystal production the hanging drop setup was employed in 24-well cell culture plates (Sarstedt, Germany). Rod shaped crystals of SvS-A2 and derived variants were obtained using 0.1 M Bis-Tris-Propane, 0.2 M Na-phosphate 20% (v/v) PEG 3350, pH 6.5. The design variants of SvS-A2 were crystallized in similar conditions with variation of pH 6.3-7.2 and PEG3350 concentrations from 12% to 20%. Further details and cryo-protection of the protein crystals are provided in Table S1. Cocrystallization was attempted with the substrates FPP and GGPP by adding the substrates at 5 mM and MgCl2 at 5 mM concentration to the protein solution 30 min prior to setting up the crystallization drops and at the cryoprotection stage (Table S1). The X-ray diffraction datasets with the best resolution in the cases of Trp79Phe_Gly83Leu, Ala224Ile and Trp156Tyr variants are from these cocrystallization attempts with FPP, however no ligand complex was obtained.
The X-ray diffraction dataset to 2.3 Å resolution for SvS-A2 was collected at beamline ID23-1 of the European Synchrotron Radiation Facility (ESRF, Grenoble, France) at 100 K. The X-ray diffraction datasets for SvS-A2 variants Trp79F_Gly83Leu, Trp156Tyr, Ala224Ile and 205-209:DREMH/AQDLE to 2.27-2.6 Å were collected at the BIOMAX beamline of MAX-IV (Lund, Sweden) at 100 K. The X-ray data were indexed and processed with XDS 2 and subsequently scaled with AIMLESS from the CCP4i suite. 3 The crystals belong to space group P212121 with cell dimensions a = 75.3 Å, b = 105.5 Å, and c = 105.5 Å for SvS-A2 with minor variation for the design variants. The statistics of the data sets are given in Table 1.

S4
The structure of the reconstructed ancestral protein SvS-A2 was solved by molecular replacement using MOLREP 4 and the poly-Alanine model derived from the coordinates of selinadiene synthase (SdS, PDB-ID: 4OKZ 5 ), a bacterial class I sesquiterpene cyclase from Streptomyces pristinaespiralis.
The model was built and refined by employing Arp/wArp 6 and continued with manual model building using COOT 7 intervened by cycles of restrained refinement using REFMAC-5. 8 Water molecules were placed by COOT and subsequently checked by manual inspection of the molecular contacts and electron density. The final model contains a dimer of the terpene synthase all α-fold representing amino acid residues 7-357 in chain-A and 8-357 in chain-B, and 274 water molecules with final crystallographic R and Rfree values of 0.185 and 0.219, respectively. Disordered surface loops lacking electron density prevented the modeling of the residue ranges 232-238 and 310-326 chain-A; and the residue ranges 88-95, 232-241 and 311-326 in chain-B (Table S1). The structures of the designed variants ( Table 1, Table S1) were solved by using the SvS-A2 structure as search model and refined using a similar protocol as for SvS-A2, leading to the 2.27 Å structure of Trp79Phe_Gly83Leu, the 2.38 Å structure of 205-209:DREMH/AQDLE, the 2.40 Å structure of Trp156Tyr and to the 2.60 Å structure of Ala224Ile variants of SvS-A2 presenting approximately the same disordered residue ranges as the parental SvS-A2 species (Table S1). Macromolecular interfaces and solvent accessible surfaces were analyzed at the PISA server. 9 The protein models were validated with regard to stereochemistry and model quality using COOT and MOLPROBITY. 10 Statistics from the refinement and model quality are provided in Table 1, the final models and structure factors are deposited with the Protein Data Bank with accession codes 6TBD, 6TJA, 6TIV, 6THU and 6TJZ.

Construction of Missing Loops and Metal Ion Cluster for SvS-A2 and SvS-A2 Surface Variant
The crystal structure of SvS-A2 has unresolved residues, thus preventing proper description of the 3D structure of the entire system. Mg 2+ ions were included in several crystallization and soaking experiments, yet no metal ions were found in the crystal structures. To obtain a complete protein structure, the full amino acid sequence of SvS-A2 was employed to build a model of missing loops and the protein termini using YASARA. [11][12] Default parameters were employed for modeling of the loops and termini. Hydrogen atoms were added to the structure according to the pKa of residues and a pH of 7. Side chains for the unresolved parts were built and the loop structures were optimized by exploring a large number of conformations using YASARA rotamer libraries to generate an initial structure. Subsequently, side chains and loops were subjected to a combined steepest descent and simulated annealing minimization, keeping backbone atoms fixed. Finally, a fully unrestrained simulated annealing minimization was performed on the entire system. Convergence in minimization stages was reached when the energy improved less than 0.05 kJ mol -1 during consecutive steps. All refinement steps were conducted using a knowledge-based force field included in YASARA (YASARA2 force field) which has been optimized for structure prediction, refinement, energy minimization and validation. 11 The tri-Mg 2+ ion cluster was built into the minimized structure by superposition onto SdS (PDB-ID: 4OKZ 5 , Cα root mean square deviation (rmsd) of 1.26 Å over 312 residues). Lastly, the structure was minimized using the same force field as above and docked to the substrate GGPP (described in detail below). Missing loops in the SvS-A2 surface variant were generated in YASARA in the same way; the metal ion cluster and substrate were included from superposition with the final docked SvS-A2 model. Figures were generated using The PyMOL Molecular Graphics System, version 2.3.4 Schrödinger, LLC.

Molecular Docking
In order to obtain a refined model of SvS-A2 (full model including loops and metal ions) complexed to the substrate, GGPP was docked into the active site of monomeric SvS-A2 using the Molecular Operating Environment (MOE) 2015.10. 13 All described energy minimizations were performed using the AMBER10 force field to within an RMS gradient of 0.01 kcal mol -1 Å -1 and using a dielectric constant of three for the protein and an external dielectric of 80 to simulate an implicit water environment. 14 GGPP was constructed and energy minimized. Five hundred poses were retained for each ligand using the alphatriangle placement methodology with binding affinity (G) as scoring function as embedded in MOE. In the docking experiments, flexible ligand structures were generated using a Monte Carlo algorithm, while the protein was held fixed at its energy-minimized geometry. The top scored orientations of the products were then energy minimized, followed by energy minimization of the entire system. Since no productive prefolded conformation of GGPP could be obtained in this automated way, the product spiroviolene was built and successfully docked using the same settings as described. Subsequently, the experimentally suggested reaction mechanism for spiroviolene formation was followed in the backward direction to obtain a productive starting conformation of GGPP. 15 To this end, bonds were manually broken and formed to generate the suggested intermediates and finally GGPP. For each generated intermediate (and GGPP) energy minimization was performed, keeping the ligand and any protein atoms within a radius of 8 Å around the ligand fully flexible. Then the entire system was energy minimized. The intermediates were also generated by breaking and making bonds going in the forward direction, starting from productively prefolded substrate. This process was repeated iteratively until the conformation of intermediates converged when following the reaction in forward and backward direction. The process was repeated following the suggested intermediates in the formation of hedycaryol (based on the mechanism of another bacterial hedycaryol synthase) 16 , starting by superpositioning SvS onto PDB-ID: 4MC3 to generate initial coordinates of the substrate. For analysis of water as base/nucleophile, the entire system was placed in an explicit water box during minimization for spiroviolene 8, hedycaryol 14 and their preceding cations 7 and 13. Water molecules with a radius of up to 9 Å around the ligand were kept for further analysis.

Substrate Binding Affinity (∆G)
The monomeric SvS-substrate complexes were refined by a final energy minimization (RMS gradient of 0.01 kcal mol -1 Å -1 ) using the LigX interface in MOE. The protein atoms far from the substrate were kept fixed, while protein atoms in the binding site (defined at a distance of 8 Å from the substrate) were treated as fully flexible. This allows to account for protein flexibility, that is induced fit. After refinement, the binding affinity was calculated employing the GBVI/WSA ∆G scoring function.

Homology Modeling of SvS-WT
Different templates were used to build homology models of the extant enzyme: (i) the obtained crystal structure of SvS-A2 (PDB-ID: 6TBD) including modeled loops and the Mg 2+ cluster docked to GGPP (SvS-WT-Hom1) (ii) an engineered surface variant of SvS-A2 (PDB-ID: 6TIV) including modeled loops and GGPP as well as the Mg 2+ cluster (SvS-WT-Hom2) (iii) SdS, the enzyme (PDB-ID: 4OKZ) in the PDB with highest sequence identity to SvS, including fully resolved loops as well as Mg 2+ cluster bound to a substrate analog (SvS-WT-Hom3). Templates provided during the homology modeling process were monomeric since accurate substrate-bound models obtained from the docking experiments were monomeric. Parameters for construction of the homology models in YASARA were set to default values, while setting the maximum number of templates to be used to one and providing the respective input template structure as PDB file. Confidence in the model was primarily assessed using the obtained Z-score for dihedral angles. The quality of the models was further assessed by Ramachandran plot analysis (RAMPAGE server), Verify 3D analysis and ERRAT 3D-1D profile. [17][18][19][20][21][22] Generation and manual docking of intermediates and substrates to the final homology model (SvS-WT-Hom2) was performed as described above for SvS-A2.

Product Characterization
The major products of enzymatic conversion were identified by mass spectrometry (MS) and flame ionization detection (FID) analysis on a dual detector GCMS-QP2010 Ultra (Shimadzu, Japan), using an Rxi-5ms capillary column (30 m length, 0.25 µm thickness, 0.25 mm i.d., Restek, USA). Purified and desalted SvS-A2 protein (2 µM) was incubated for 180 min in 2 mL capped glass vials (30 °C, 1200 rpm in Thermomixer (Eppendorf, Germany)) with 60 µM of either FPP or GGPP in reaction buffer (50 mM Tris-HCl, 1 mM MgCl2, pH 7.4 at 25 °C) in a total reaction volume of 200 µl. Reaction products were immediately extracted by adding 2x 300 µl hexane with a Hamilton syringe, vortexing and centrifuging (room temperature, 9000 x g, 10 min) and transferring the total amount of solvent phase to a GC-vial with a glass Pasteur pipette. For MS-characterization, 1 µl of the solvent phase was injected at 150 °C or 250 °C with a split ratio of 1.5. The column oven temperature was increased from 50 °C to 250 °C at a rate of 8 °C min -1 and then further raised to 330 °C at a rate of 15 °C min -1 and kept at 330 °C for 5 min under a pressure controlled flow of helium (100.0 kPa). Electron ionization was performed at an ion source temperature and transfer line temperature of 200 °C and a full scan of m/z-values ranging from 20 to 400 was recorded. Products were identified by comparing the resulting spectra to spectra in published literature and the NIST library version 11/11s. For identification of farnesol a (2E,6E)-farnesol standard was used.
Linear retention indices (Van den Dool-and Kratz-indices) were determined to identify farnesol, hedycaryol and spiroviolene by GC-FID analysis of extracts spiked with an analytical alkane standard (C8 -C20). 23 To this end, 1 µl of spiked extract was injected at 250 °C with a split ratio of 10.0. The column oven temperature was programmed based on a protocol published by Baer et al. 16 : the temperature was kept at 50 °C for 5 min and then raised to 320 °C at a rate of 10 °C min -1 and kept at 320 °C for 3 min under a pressure controlled flow of helium (160.9 kPa). The FID-detector was set to 350 °C using a makeup flow of 30 mL min -1 helium. Linear retention indices were then calculated based on relative retention of target compounds in reference to the two adjacent alkane peaks. The identity of hedycaryol was finally confirmed by analysis of thermal rearrangement to elemol 24 . The injection temperature was 150 °C, spectra were recorded using both full scans and selected ion monitoring (m/z values of 59, 93 and 161).

Product Quantification by GC-FID
Products were quantified by comparing GC-FID peak areas to the peak area of an internal decane standard and accounting for estimated relative response factors (GC instrumentation described above). 25 To this end 1 µl of extracted solution was injected at 250 °C with a split ratio of 10.0. The column oven temperature was increased from 50 °C to 250 °C at a rate of 8 °C min -1 and then further raised to 330 °C at a rate of 15 °C min -1 and kept at 330 °C for 5 min under a pressure controlled flow of helium (160.9 kPa). The FID detector was set to 350 °C using a makeup flow of 30 mL min -1 of helium.

Enzyme Kinetics
For kinetics of the reconstructed ancestral enzyme with either FPP or GGPP, experiments were performed in triplicate. Substrate concentrations ranging from 5 to 120 µM were prepared using a substrate stock solution (in 70% methanol/30% aqueous ammonia) in 2 mL capped glass vials in a total volume of 410 µl reaction buffer (50 mM Tris-HCl, 1 mM MgCl2, pH 7.4 at 25 °C). In all diluted substrate solutions the methanol concentration was adjusted to the maximum volume resulting from addition of methanol containing substrate stock within the experiment (corresponding to 4.2% in the reaction vials). The substrate dilutions were pre-incubated for 10 min at 30 °C in a thermomixer (Eppendorf, Germany) and the temperature of the thermomixer was verified with an external thermometer. Then, 10 µl of purified and desalted enzyme were added to each vial to reach a final enzyme concentration of 500 nM. After mixing by pipetting, 100 µl of solution were immediately withdrawn, transferred to a glass vial and extracted with 2 x 150 µl of hexane spiked with 10 µM decane as described above.
After withdrawing the initial 100 µl aliquot, the reaction was incubated at 1200 rpm shaking speed and 100 µl aliquots were withdrawn after 30, 60 and 90 min of incubation and extracted following the same protocol. Product concentrations in all extracts were quantified by GC-FID as described above and initial rates (stated in nanomoles per milligram per minute) were calculated from the slope of product accumulation over time (linear over the entire duration of the incubation).

Relative Diterpene Specificity
Relative diterpene specificity was determined in triplicates based on the accumulation of the major products (spiroviolene and hedycaryol) in an in vitro competition assay. An equimolar substrate mix of FPP and GGPP (60 µM each) was prepared in a total volume of 95 µl of reaction buffer (50 mM Tris-HCl, pH 7.4, 1 mM MgCl2) in 2 mL capped glass vials. The substrate concentration of 60 µM was chosen to achieve maximum detection signal on the GC-FID detector while ensuring that the enzyme was not saturated. The substrate dilutions were pre-incubated for 10 min at 30 °C in a thermomixer (Eppendorf, Germany) and the reaction was started by adding 5 µl of purified and desalted enzyme solution, resulting in a final enzyme concentration of 0.5 or 2 µM. The reactions were incubated at 1200 rpm, 30 °C for 180 min and then extracted as described above. Reaction products were quantified by GC-FID as described above and normalized by the enzyme concentration. Diterpene specificity was expressed as ratio of apparent second order rate constants of the major diterpene product (spiroviolene) over the major sesquiterpene product (hedycaryol) based on the following equation:

Malachite Green Screening of Variants
Freshly purified variants of SvS-A2 were screened for activity using a Malachite Green assay, adapted from Vardakou et al. 26 , based on the inorganic pyrophosphate released as a side product of terpene cyclization. As the Malachite Green assay detects inorganic phosphate, inorganic pyrophosphatase was added to the reaction buffer (50 mM Tris-HCl, 5 mM MgCl2, pH 7.4 at 25 °C) to a concentration of 50 mU mL -1 . Either FPP or GGPP (50 µM) and a negative control (no substrate) were prepared in 130 µl reaction buffer in a 96 deep-well plate format. The methanol concentration resulting from the addition of 50 µM substrate stock (70% methanol) in all wells corresponds to 1.8% (v/v). The plates were pre-incubated at 30 °C for 10 min without agitation. Subsequently, 20 µl of the different enzyme variants (in a range of ca. 0.2 µM to 3.9 µM) or buffer (enzyme-free negative controls) were added to the reaction wells to reach a final volume of 150 µl. Each reaction was run in triplicate and a standard dilution series ranging from 0.2 µM to 50 µM of dibasic sodium pyrophosphate was added to the plate to ensure that obtained signals are in the linear range of absorbance detection. The plates were sealed with a breathable film and shaken vigorously for a duration of 2 hat 30 °C. Malachite Green dye stock solution was used to freshly prepare Malachite Green development solution as described 26 and filtered through a 0.22 µm syringe filter before use. A colorimetric development plate was prepared by adding 100 µl of reaction buffer without pyrophosphatase into a fresh 96-well clear microplate (Greiner, Austria) and adding 36 µl of Malachite Green development solution to each well. Fifty microliter solution from each well on the reaction plate were transferred to the prepared development plate using a multi-channel pipette. After ca. 1 min, 7.5 µl of 34% (w/v) sodium citrate solution was added to each well as a color stabilizer. The development reaction was incubated for 20 min in the dark at room temperature on a plate shaker (750 rpm) before the absorbance was measured at 623 nm on a Spark plate reader (Tecan, Switzerland). The obtained signals were blanked against controls containing no enzyme and no substrate and signals were normalized to the enzyme concentration employed in each well. The sensitivity threshold was determined as 0.05 absorbance units after blanking and signals obtained below this threshold were considered as noise (represented with a dash in Figure 5).  [a] Z-score classification used according to YASARA output files: "Good" denotes a Z-score between -1 and 0, "Optimal" denotes a Z-score greater than 0. [b] 15 terminal residues were modeled but excluded from the Zscore calculation by YASARA.  Figure 1. The metal-binding DDxx(x)D motif (helix C) and NSE motif (helix H), as well as the capping loop (helix K) are shown opaque and are labelled. The thickness of the cartoon represents the Z-score at each residue, with thick cartoon indicating a lower Z-score and thin cartoon indicating a higher Z-score. In the top row an arrow in SvS-WT-Hom3 indicates the structural alterations in the metal binding site compared to the other models. In the bottom row the capping loop is shown transparent for clarity. Three arrows in the SvS-WT-Hom1 and SvS-WT-Hom2 models highlight regions of increased local Z-score (i.e. thinner cartoon) in SvS-WT-Hom2 compared to SvS-WT-Hom1.

G305
[a] Helix numbering according to SvS-A2. Positions that differ between SvS-WT and SvS-A2 are indicated with X, residues that are conserved across all four enzymes are marked with asterisk. [b] Intermediates stabilized by putative π interactions, suggested from modelled pathways (Figure 4, Figure S7) are indicated in brackets; summary from enzyme library experiments indicated in italics.
[b] Literature indices are from DB-5 or HP-5MS columns, which are equivalent to Rxi-5ms column. [c] Hedycaryol is detected as its thermal rearrangement product elemol, generated in situ in the GC-injector 24 .
[d] The given LRI is obtained from an experimental standard run on the same column. Differences may occur due to rounding.  [a] The reaction intermediates complexed to the active site are shown in Figure 4. Numbering of atoms in intermediates is based on the numbering in GGPP.   Table S6. The proposed electron flow characterized from another hedycaryol synthase is represented with conventional arrows in the 2D depiction. 16 Residues involved in complexing metals and pyrophosphate in 9 are shown as sticks and labelled, but omitted for clarity in other panels. The hydrophobic cage around the intermediate is shown as sticks in 10 and for clarity only residues in close proximity of the cation are shown as sticks in the following panels. Anti-Markovnikov addition of the double bond at C1 in 12 is assisted by stabilization of the forming charge by the backbone carbonyl of Ala186 (shown as green sticks and by dashed line). Individual residues involved in cationic π-interactions are highlighted as cyan sticks. For nucleophilic water attack on 13 water molecules within a radius of 9Å of the ligand are shown. One water molecule bridges C11 and Asp87 (distance Owat•••C11 = 3.3 Å, distance Hwat•••OAsp87 = 1.6 Å). In SvS-WT, the water molecule is located closer to the pyrophosphate moiety than in SvS-A2 (Table S6). [a] The reaction intermediates complexed to the active site are shown in Figure S7. Numbering of atoms in intermediates is based on the numbering in FPP. The key specificity-switch residues identified in the library are labeled and colored in dark blue (Trp156Tyr to achieve GGPP specificity) and yellow (Ala224Ile to achieve FPP specificity). (b) The SvS-A2 active site is shown with GGPP docked (shown in dark blue sticks and transparent spheres). The side chain of Ala224 is shown as yellow sticks and transparent spheres. (c) The crystal structure of SvS-A2 Ala224Ile is superposed onto (b) and the side chain of Ile224 is shown as yellow sticks and transparent spheres, highlighting the steric clash that likely prevents cyclization of GGPP in the active site of the Ala224Ile variant.

Secondary shell mutations
[a] All variants carry an additional A89H exchange to preserve the metal binding site as in SvS-WT.