Partial catalytic Cys oxidation of human GAPDH [version 1; peer review: 2 approved]

Background: n-Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) catalyses the reversible NAD+-dependent oxidative phosphorylation of n-glyceraldehyde-3-phosphate to 1,3-diphospho-n-glycerate in both glycolysis and gluconeogenesis. Methods: Four distinct crystal structures of human n-Glyceraldehyde3-phosphate dehydrogenase (HsGAPDH) have been determined from protein purified from the supernatant of HEK293F human epithelial kidney cells. Results: X-ray crystallography and mass-spectrometry indicate that the catalytic cysteine of the protein (HsGAPDH Cys152) is partially oxidised to cysteine S-sulfonic acid. The average occupancy for the Cys152-S-sulfonic acid modification over the 20 crystallographically independent copies of HsGAPDH across three of the crystal forms obtained is 0.31±0.17. Conclusions: The modification induces no significant structural changes on the tetrameric enzyme, and only makes aspecific contacts to surface residues in the active site, in keeping with the hypothesis that the oxidising conditions of the secreted mammalian cell expression system result in HsGAPDH catalytic cysteine S-sulfonic acid modification and irreversible inactivation of the enzyme.


Introduction
Mammalian HEK293F cells are routinely used in conjunction with secreted protein expression vectors for recombinant protein production 1 . They owe their popularity to ease of handling, robust growth rate, excellent transfectability, high capacity for recombinant protein expression, low-cost media requirements and low levels of secreted contaminants 2 . Of course, in absence of secretion of the desired recombinant protein at hand, contaminants are the only proteins present in recombinant protein expression systems 3,4 . In the course of a research effort aimed at the purification of recombinant Chaetomium thermophilum ERAD mannosidase CtHTM1P from the supernatant of HEK293F cells, we purified, crystallised and determined crystal structures of human n-Glyceraldehyde-3-phosphate dehydrogenase (HsGAPDH , EC 1.2.1.12) 5 .
GAPDH catalyses the reversible NAD + -dependent oxidative phosphorylation of n-glyceraldehyde-3-phosphate to 1,3diphospho-n-glycerate in both glycolysis and gluconeogenesis 6 . This reaction occurs at an important transition point in glycolysis between the enzymatic steps that consume and generate ATP. Several studies indicate that the enzyme has pleiotropic functions independent of its canonical role in glycolysis 7-9 . In addition to the somatic cells HsGAPDH isoform, the human genome encodes a testis-specific isoform HsGAPDHS (G3PT_HUMAN (Uniprot O14556)), which is expressed only in the post-meiotic period of spermatogenesis. In multiple mammalian species the sperm isozyme shares about 70% amino acid identity with the somatic isozyme and plays a role in anchoring GAPDHS to the fibrous sheath in the principal piece of the sperm flagellum 10 .
GAPDH was one of the first enzymes to be crystallised 11,12 and one of the first enzymes whose structure was determined by X-ray crystallography 13,14 . A number of studies have reported the GAPDH catalytic Cys carrying oxidised post-translational modifications. In particular, cysteine sulfenic acid (CSX, see Figure 1A) has been observed in the structure of rabbit muscle GAPDH 15 (of course, by X-ray diffraction alone, this modification may be difficult to distinguish from S-hydroxycysteine (CSO, see Figure 1C)). Cysteine-S-sulfinic acid (CSD, see Figure 1B) has been observed in structures of rat GAPDHS and E. coli GAPDH 16 . Cysteine-S-sulfonic acid (CSU, see Figure 1D) modifies the catalyic Cys in a structure of Streptococcus pneumoniae GAPDH 17 .
In this paper, we present and discuss mass spectrometry and crystallographic evidence supporting partial oxidation of the catalytic Cys residue of HsGAPDH purified from the supernatant of a HEK293F cell culture. A fraction of the molecules shows Cys-S-sulfonic acid instead of Cys in the active site. The four HsGAPDH crystal structures we describe (which we label as P2 1 forms A-D, see Table 1) represent novel HsGAPDH monoclinic polymorphs and are the first HsGAPDH crystal structures  the mixture was incubated at room temperature for 20 minutes; the DNA/PEI mixture was added to the cell culture; the cell culture was made 5 µM kifunensine (an inhibitor of endoplasmic reticulum and Golgi mannosidases, Cayman Chemical 109944-15-2). The cell culture was split into three 500 mL Erlenmeyer flasks with 0.2 µm vent caps (Corning), with 150mL of culture in each flask, and incubated in an orbital shaker incubator at 37°C, shaking at 120 rpm, under a 5% CO 2 atmosphere. The cells' supernatant was harvested 4 days post-transfection by centrifuging at 4,000 x g for 5 minutes.

Protein purification
The cell supernatant was made 1x PBS by addition of the appropriate volume of 10x PBS stock (obtained by dissolving five PBS tablets (Sigma Aldrich P4417) in 200 mL of deionised water). The pH was adjusted to 7.4 and the solution filter-sterilised through a 0.2 µm bottle-top filter. Nickel immobilised metal affinity chromatography (IMAC) was used as the first step of purification. A 5mL HisTrap HP column (GE Healthcare 17-5248-01) was equilibrated with five column volumes (CV) of binding buffer (50 mM sodium phosphate pH 7.5, 300 mM NaCl, 1 mM TCEP (Sigma Aldrich C4706)). The cell culture supernatant was loaded onto the column at room temperature at a flow rate of 3 mL/min. After a 5CV wash in binding buffer, the column was fitted to an ÄKTA purifier fast protein liquid chromatography (FPLC) machine in a 4°C cabinet and further washed with 6 mL of elution buffer (50 mM sodium phosphate pH 7.5, 500 mM imidazole (Honeywell Fluka 56750), 200 mM NaCl).
The wash fractions from the IMAC run (total volume 4.5 mL) were pooled and concentrated to 2 mL in a polyethersulfone spin concentrator of 10 kDa MW cut-off (Thermo Fisher Scientific 88513). The concentrated sample was filtered through a 0.22 µm spin filter and loaded on a 2 mL loop connected to an ÄKTA purifier FPLC machine in a 4°C cabinet. The sample was injected on a size exclusion chromatography (SEC) S200 16/60 column (GE Healthcare 28-9893-35) equilibrated in filter-sterilised and degassed 20 mM HEPES pH 7.5, 100 mM NaCl and 1mM TCEP (Sigma Aldrich C4706) buffer, and run down the column at 0.4mL/min flow rate, collecting 1 mL elution fractions. HsGAPDH-containing fractions were pooled and concentrated as described before to a volume of 70 µL, and protein concentration measured by loading 1.5 µL of sample on a NanoDrop 1000 spectrophotometer (Thermo Scientific). The absorbance was OD 280 =15.0, equivalent to a concentration of 18.8 mg/mL (calculated from HsGAPDH ϵ 280 =0.7963 (mg/mL) −1 cm −1 ).
A volume of 30 µL of the HsGAPDH obtained from the S200 SEC run was diluted to 500 µL in filter-sterilised and degassed buffer HEPES (Sigma Aldrich H3375) 20 mM pH7.5, NaCl 100 mM, TCEP 1mM and injected onto a Superdex S200 10/300 column equilibrated in the same buffer. The mass of the sample was detected on elution with an 18-angle multiangle light scattering detector (Dawn ® HELEOS ® II) coupled with a differential Refractive Index detector (Optilab ® T-rEX) (Wyatt Technology). Figure 2 illustrates the results of the run. The three fractions C2-C3-C4 eluting between 12.5 and 14 mL were pooled and concentrated to a volume of 45 µL -and a concentration of 3.0 mg/mL.

Protein crystallisation
X-ray data diffraction collection and processing All crystals were cryo-cooled by plunging them into liquid nitrogen. X-ray diffraction data were collected at beamline I03 of the Diamond Light Source in Harwell, England, UK, with an X-ray beam of wavelength λ=0.97622 Å and size 80x20 µM.
Other data collection parameters are listed in Table 2. X-ray diffraction data were processed with the autoPROC suite of programs version 1.0.5 21 with the following command line options: process -h5 Datapath/Data_master.h5 -M HighResCutOnCChalf -d DataOutputDirectory. An alternative Open Source suite that would enable X-ray diffraction data processing in equivalent ways is xia2 29 running DIALS 30 .
Structure determination and refinement P2 1 Form A: initial structure factor phases were computed by molecular replacement, searching for eight copies of the HsGAPDH monomer in PDB ID 1U8F in space group P2 1 using the program CCP4-Molrep version 11.7.02 31 with all default parameters. The eight copies of the HsGAPDH monomer are arranged in two tetramers in the asymmetric unit. Initial automated water addition and positional and individual B-factor refinement were carried out in autoBUSTER, version 2.10.3 32,33 . An alternative Open Source piece of software that would enable refinement against X-ray diffraction data in equivalent ways is Vagabond 34 . autoBUSTER was run with the following command line options: refine -m Data.
staraniso_alldata-unique.mtz -p Model.pdbautoncs -Seq GAPDH.seq -d OutputDir -l XPE. grade_PDB_ligand.cif ExcludeBadContacts='EXC LUDE *|203:* *|203:*' AutomaticFormfactorCorr ection=yes -l CSU.grade_PDB_ligand.cif. Automated non-crystallographic restraints were used throughout 35 , including water molecules (assigned to each chain using CCP4-Sortwater version 7.0.078, with all default parameters). At each catalytic Cys152 site, a 0.5:0.5 occupancy ratio mixture of Cys and Cys S-sulfonic acid was initially modelled in unbiased Fo-Fc residual density (see Figure 5). At each Cys152 site, occupancies for Cys and Cys S-sulfonic acid were then refined under the constraint that they sum up to 1.000±0.005. Crystal form A was deposited in the Protein Data Bank (PDB) with ID code 6YND. Refinement statistics are reported in Table 3. P2 1 Forms B,C and D: initial structure factor phases were computed by molecular replacement, searching with the P2 1 Form A tetramer (PDB ID 6YND) in space group P2 1 , placing one tetramer per asymmetric unit in P2 1 Forms B and D and two tetramers per asymmetric unit in P2 1 Form C, using the program CCP4-Molrep 31 version 11.7.02, with all default parameters. The same refinement protocol and restraints were used a described for P2 1 Table 3.

Mass spectrometry
Protein was purified by 1D-PAGE before Coomassie staining and excision for proteomic analysis. Protein was digested in-gel with the addition of 20 ng sequencing grade trypsin (Promega V5111). No reduction or alkylation was performed during digestion to help preserve the native state of oxidation. Resulting peptides were analysed over a 1 h liquid chromatography-mass spectrometry acquisition with elution from a 50 cm EasyNano C18 column as detailed in 36 onto a Thermo Orbitrap Fusion Tribrid mass spectrometer. Spectra were acquired in data dependent acquisition mode with sequential fragmentation of all selected peptide precursors using both higher-energy collisional dissociation and electrontransfer dissociation (ETD). All spectra were acquired in the Orbitrap mass analyser with internal calibration from the ETD reagent.
Product ion spectra were searched against the expected sequence of HsGAPDH (Uniprot P04406) appended to a custom in-house database. All pieces of software were run with default parameters, with the exception that propionamide (+71.037114) and dehydro (-1.007825 Da) were added as manually defined variable modifications. The analysis was carried out using Mascot version 2.6.1 (Matrix Science) 37 , PEAKS Studio X+,10.5 (Bioinformatics Solutions Inc.) 38 and Byonic™ 3.0 (Protein Metrics Inc) 39 search engines ( Figure 3A and Figure 4A). The same data analysis was also carried out with freeware, using a combination of MSConvert version 3.0 40 , SearchGui version 3.3.18 41 and PeptideShaker version 1.16.45 42 ( Figure 3B and Figure 4B). The resulting peptide assignments are equivalent to the ones obtained with the commercial pieces of software, see Figure 3 and Figure 4. Searching specified a precursor tolerance of 3 ppm and a fragment ion tolerance of 0.02 Da. Initial searches included variable modification of: S-Oxy Cysteine ( Figure 1A), Cysteine S-sulphinic acid ( Figure 1B), and Cysteine S-sulphonic acid ( Figure 1D). Searches were then expanded in PEAKS 38 to include 313 of the most frequently observed proteomic modifications and in Byonic™ 39 to include disulfide bonding and wildcard mass addition to Cys.
Peptide spectra matches were filtered to p-values (or equivalents) of <0.05. Individual spectra were manually inspected for site localisation or ambiguity before reporting. Global protein coverage of over 90% was achieved. Individual post-translational modification (PTM) modified spectra were manually inspected for site localisation specificity and potential PTM ambiguity before reporting.

Results
Since the first crystal structures of the lobster GAPDH enzyme 43-45 several crystal structures of prokaryotic and eukaryotic GAPDHs have been determined. At the time of this writing, the PDB contains more than 142 GAPDH entries, 11 of which are of human GAPDH isoforms (see Table 1). All human GAPDH crystal structures described so far contain tetramers, sitting in crystal sites of three different symmetries: point group 1 (four copies of the monomer all slightly different from each other); point group 2 (dimer of dimers); and point group 222 (four identical copies in a tetramer of exact 222 symmetry). HsGAPDH crystals of form A and form B grew from the HsGAPDH protein sample purified from the supernatant of HEK293F cells by two chromatography steps (IMAC+SEC); crystals of forms B and form C grew from the same HsGAPDH protein sample after an extra SEC-MALS chromatography step, see Figure 2. Our HsGAPDH crystal structures all contain tetramers of point group 1 (two tetramers per asymmetric unit in forms A and C, and one tetramer per asymmetric units in forms B and D). Overall, the four novel crystal forms contain 24 crystallographically independent observations of the HsGAPDH molecule, adding to the 29 ones already present in the PDB at the time of this writing (see Table 1). The overall rmsd Cα is 0.170 Å across these 24 crystallographically independent copies of HsGAPDH.
The protein sample from which crystal forms C and D were grown was analysed by mass spectrometry. In addition to peptides corresponding to unmodified Cys152 (see Figure 3), we detected ions derived from fragmentation of the Cys152 containing tryptic peptide 146 IISNASCTTNCLAPLAK 162 , which suggests Cys152 S-sulfonic acid modification (Figure 4). Annotation is complicated by the presence of a further modification at Cys156 with a mass addition of +71.036 Da, as directly observed in the fragmentation spectrum by the mass spacing between the y-6 and y-7 ions (Figure 4). The +71.036 Da mass addition at Cys156 is best annotated as acrylamide adduct (propionamide) imparted during PAGE separation 46 rather than being a native modification at this residue. The measured mass of the peptide was 1869.8623 Da and the theoretical mass for the unmodified peptide is 1718.8695 Da. Including the annotated acrylamide adduct, this leaves 79.9568 Da unaccounted for, which is an extremely close match to S-sulfonic acid modification of Cys (+79.9568 Da). A potential ambiguity for the assignment is the similarity in mass of phosphorylation (+79.966331); however, phosphorylation would equate to mass error of 9.5 mDa as opposed to just 0.018 mDa for S-sulfonic acid. Unlike in the case of the acrylamide adduct, the +79.9568 Da mass addition cannot be observed in any of the individual b-or y-ions in the fragmentation spectrum (Figure 4), implying complete neutral loss of the moiety between precursor selection and fragmentation. Although neutral loss is frequently observed with Ser and Thr phosphorylation, it is rarely seen with 100% efficiency, further suggesting sulfation of Cys152 as the most credible assignment. Relative quantification of modified versus unmodified peptide forms is not possible based on the mass spectrometry data alone, due to unknown but likely significant differences in relative response resulting from the changes in the physico-chemical properties of the Cys152containing ions upon Cys-S-sulfation.  Peaks suggesting Cys152 S-sulfonic acid modification are also visible in the initial crystal electron density Fo-Fc difference maps of crystal forms A, C and D 1 , contoured at 2.5σ level or higher, in close proximity to the Cys152 γ sulfur atom of most HsGAPDH chains in the asymmetric unit (see Figure 5). The orientation of the Cys-S-sulfonic acid moiety in the active site is not unique, but it varies from chain to chain across the three crystal forms. Figure 6A illustrates the modification in the eight copies of HsGAPDH in crystal form C. The Cys152 Ssulfonic acid chain makes only loose non-bonding contacts to residues in the active site (see also Figure 5). Modelling of the Cys152 residue as a superposition of Cys and Cys-S-sulfonic acid with variable occupancies in crystal forms A, C and D enables estimation of the extent to which the crystals contain the Cys152 modification. The average occupancy for the Cys152-S-sulfonic acid modification over these 20 crystallographically independent copies of HsGAPDH is 0.31±0.17 (see Table 3). The HsGAPDH monomer with the largest occupancy for the modification (Occ C ys−S−SO3 =0.72) is monomer H in crystal form C. This predominantly Cys-S-sulfonated HsGAPDH monomer superposes with an rmsd C α =0.423 Å (over 334 C α atoms) with the Streptococcus pneumoniae GAPDH monomer carrying the same modification 17 , see Figure 6B. When comparing the modified structure to the one of HsGAPDH with unmodified Cys152 (PDB ID 1U8F) no significant local changes to active site side chains are observed either (rmsd C α =0.127 over 333 C α atoms), see Figure 6C. None of the HsGAPDH molecules in the crystals contain NAD + nor any bound ligand in the active site.

Discussion and conclusions
GAPDH is mainly a cytoplasmic enzyme, but it has been reported to moonlight as a cell surface and/or extracellular enzyme in healthy human cells 49,50 , in human cells subjected to stresses such as micronutrient starvation, hypoxia, infection, and cancer 50-54 , and to act as a extracellular effector in fungi and bacteria [55][56][57][58][59][60][61][62] . In our study, HsGAPDH was purified from the supernatant of HEK293F cells treated with the mannosidase inhibitors kifunensine and kept for 4 days at 37°C after transfection with a DNA plasmid for recombinant expression of an endoplasmic reticulum glycoprotein. Although kifunensine is toxic to cells only in concentrations higher than the one used (5µM), during the 4 days between transfection and harvesting of the supernatant the experimental conditions are likely harsher than the basal levels of oxidative stress in the HEK293F cell culture media 63 .
Stable oxidation of the GAPDH catalytic cysteine has been observed in a number of studies, for example to Cys sulfenic acid mediated by thiolate oxidation 64 : the modification is reversible but it confers the modified cysteine electrophilic properties similar to the ones of sulfenyl halides 65 . For example, GAPDH catalytic cysteine modification to sulfenic acid converts the enzyme from dehydrogenase to an acyl phosphatase [66][67][68] . In the presence of excess oxidant, further oxidation of sulfenic acid to sulfinic acid or sulfonic acid is possible. However, in contrast to sulfenic acid, higher oxidation states of catalytic Cys residues lead to enzyme inactivation and Cys S-sulfonation is generally considered an irreversible form associated with protein misfolding, degradation, and pathology 65 . In particular, S-sulfonated GAPDH has been reported to translocate to subcellular domains where it does not normally occur, where it may stimulate a "gain of function" that could provoke apoptosis 49 .
It is unclear whether the Cys S-sulfonic acid modification we observe in HsGAPDH purified from the supernatant of HEK293F cells represents a physiological response of the cells to the recombinant secreted protein expression system conditions, or whether it is simply the outcome of a protracted oxidative environment on the pool of extracellular enzyme instead. GAPDH allostery is mediated by intra-dimer contacts (the active site of one dimer subunit is in close proximity to residues in the neighbouring subunit). The observation that HsGAPDH tetramers pack in the crystals in ordered fashion even in presence of partial Cys-S-sulfonic acid modification, together with the overall structural similarity of the Cys S-sulfonated and native monomeric enzymes, suggests that the secreted HsGAPDH is constantly inactivated by catalytic Cys S-sulfonation in the expression system oxidising extracellular environment.

Data availability
Underlying data X-ray data of forms This project contains the following underlying data:
The Introduction and Discussion are overloaded with references to old articles, but there is not enough references to later works. For example, there are references to the first works of 1969-1972 years on GAPDH isolation and on acylphosphatase activity of GAPDH.

5.
Methods. Cloning and isolation of recombinant mannosidase is excessively detailed in Methods. It should be emphasized that the aim of this work was to isolate GAPDH from the conditioned medium and indicate what is critical for the isolation of this enzyme (for example, the importance of using reducing agents to prevent the oxidation of GAPDH). The authors did not decipher the name of the used reducing agent (TCEP -Tris (2-Carboxyethyl) phosphine), which complicates the understanding of the article. 6.
Discussion and conclusion. "However, in contrast to sulfenic acid, higher oxidation states of catalytic Cys residues lead to enzyme inactivation " -an incorrect phrase, since the oxidation of the catalytic cysteine to sulfenic acid also leads to disappearance of the dehydrogenase activity, although reversible 7.

If applicable, is the statistical analysis and its interpretation appropriate? Yes
Are all the source data underlying the results available to ensure full reproducibility? Yes Are the conclusions drawn adequately supported by the results? Yes