Cell Survival Assays
Approximately 20,000 HeLa cells per well were seeded in a 96-well, white, flat-bottomed tissue-culture plate (Sigma, UK) in a volume of 100 μl culture medium (DMEM lacking phenol red, (Thermo Fisher Scientific, UK) supplemented with 2 mm glutamine and 10% fetal calf serum, plus penicillin/streptomycin). Cells were incubated for 18 h at 37 °C at 5% CO2. Dilutions of either aspirin or salicylic acid were made in culture medium to final concentrations of 20 mm, 10 mm, 5 mm, 2 mm, 1 mm, and 0.5 mm. A zero drug dilution was made containing only DMSO at the same concentration as in the dilutions (aspirin and SA were dissolved and stored in DMSO). To begin exposure to salicylate cell culture medium was replaced with the salicylate dilutions in quadruplicate. Cells were cultured at 37 °C and 5% CO2 for 6, 24, or 48 h. To assess cell viability 100 μl ATP assay buffer (50 mm Tris/phosphate pH 7.8, 16 mm MgCl2, 2 mm DTT, 2% v/v Triton-X-100, 30% v/v (37.8% w/v) glycerol, 1% w/v BSA, 0.25 mmd-luciferin, 8 μm sodium pyrophosphate tetra-basic decahydrate, 500 ng/ml Luciferase) was added to each well, before sealing with clear film and agitating at 900 rpm and 20 °C for 10 min. Luminescence was measured using an EnVision Multilabel plate reader (Perkin Elmer, UK). Readings were normalized to the zero-drug control for each set of replicates. In experiments using co-treatment with KDAC inhibitors, bufexamac was used at 0.25 mm or nicotinamide at 20 mm, cells were exposed for 24 h, and salicylate dilutions of 20 mm, 10 mm, 5 mm, 2.5 mm, 1.25 mm, and 0.625 mm were used.
Synthesis of Aspirin-d3
2.18 g salicylic acid (Sigma Aldrich) was mixed with 5g acetic anhydride-d6 (Sigma Aldrich) in an Erlenmeyer flask before addition of 8 drops (∼400 μl) of 85% orthophosphoric acid. The solution was mixed by agitation and warmed to 70 °C in a water bath for 15 min. Unreacted acetic anhydride-d6 was destroyed by addition of 14 drops (∼700 μl) cold ultra-pure water. Crystallization was initiated by addition of 14.5 ml ultra-pure water and transfer of the solution to an ice-bath for 30 min. Crystals of aspirin-d3 were removed from solution by filtration onto filter paper under suction on a Buckner funnel. Crystals were washed with ∼20 ml ice-cold ultra-pure water, then dried by 15 min suction. Recrystallization of Aspirin-d3 was then performed: The crystals were dissolved in 10 ml ethanol while gently warming. Recrystallization was triggered by addition of 27 ml warm ultra-pure water, followed by slow cooling at ambient temperature, then rapid cooling on ice. Crystals were filtered out of solution onto filter paper using a Buchner funnel and suction for 15 min, then dried on a fresh sheet of filter paper in a glass beaker loosely covered with tissue-paper for 2 days. Unlabeled Aspirin was synthesized by an identical method.
NMR Analysis of Aspirin
1H and 13C spectra were obtained with a Bruker Avance II 400 MHz, Bruker Avance 500 MHz or a Bruker Avance III 500 MHz spectrometer with the solvent peak used as the internal standard. NMR spectra were processed using TopSpin 3.1 (PC version) or MestReNova.
MS Analysis of Aspirin
Saturated solutions of commercially sourced aspirin, in house synthesized aspirin, and in house synthesized aspirin-d3 were made in acetonitrile. 1:1000 dilutions in acetonitrile of each was analyzed on a QExactive mass spectrometer in negative mode using a direct infusion flow rate of 20 μl·min−1, linked to a HESI-II probe in Ion Max API source (Thermo Scientific) running at 4 kV, 300 °C and S-lens RF level of 50. Settings for full MS scans were; scan range = 172–185 m/z, resolution = 70000, AGC target = 1e6, maximum injection time = 20ms. Spectra were acquired over 25 scans and viewed in Xcalibur Qual Browser.
Kinetic Analysis of Aspirin and Salicylic Acid in Cultured Cells and Cell Medium
HeLa cells were cultured in six well plates (9.5 cm2 area per well) in DMEM supplemented with FCS to 10% such that final cell counts at the end of the time-course were 1 to 1.5 million. Cells were exposed to 5 mm aspirin (unlabeled) for 0.25, 0.5, 1, 2, 6, 16, or 24 h by replacement of 2 ml medium. To assess concentrations of aspirin (ASA) and salicylic acid (SA) in both culture medium and cells the following method was used for each well of cultured cells: 1 ml culture medium was removed and snap-frozen before storage at −80 °C. The remaining medium was removed and the cells washed three times with 1 ml ice-cold PBS. Cells were lysed by addition of 100 μl lysis buffer (20 mm Tris pH 7.5, 150 mm NaCl, 1% triton-X100, plus Roche protease inhibitor mixture), and incubated on ice with orbital shaking for ∼10 min. Lysates were removed from wells and snap frozen before storage at −80 °C. This process was repeated in triplicate (n = 3) for each time point. For each time point a fourth well of cells was grown in parallel to allow estimation of cell numbers and cell volume using an automated cell counter (Countess, Invitrogen, UK). These values were used to calculate total cell volume, which was between 2 and 3 μl depending on the time point.
ASA and SA were measured in medium and lysate samples using UV absorption with in-line UPLC (Acquity) running a BEH-C18 50 × 2.1 mm 1.7 μm particle column. A 3 min gradient protocol was used running from 2 to 95% acetonitrile (in 0.1% formic acid) then back to 2%. For our system SA presented a λmax = 304 nm and retention time (RT) 1.49 min and ASA a λmax = 276 nm at RT 1.42 min. Calibration curves were defined over a range of ASA and SA concentrations from 0.1 to 1000 μg/ml. Sample concentrations of ASA and SA were calculated by reference to these calibration curves.
Measured data were modeled using Copasi (v4.16 - build 104) (20). Data were fit to a simple model for the diffusion of aspirin and salicylic acid between medium and cells and for the hydrolysis of aspirin in both compartments as shown in supplemental Fig. S1A. The five species shown in supplemental Fig. S1A, and two compartments were defined; Medium (2 ml) and cells (2 μl). Experimental data were weighted manually toward the more confident measurements from the culture medium (weight 1.0), and away from the cellular measurements (weight 0.2) due to both the inherent errors of the lower concentration measurements (actual concentration measurements in the cell extracts were in the order of 10 times lower than for the medium measurements) and the fact that these values included the extra variable of cell counting and size measurements. The seven rate constants shown in supplemental Fig. S1A were modeled along with the starting concentrations of SA and ASA in the medium. All other starting concentrations were fixed at zero. The in and out rate constants for ASA were constrained to be equal, as were those for SA. Repeated iterations of parameter estimation resulted in the fits shown in supplemental Fig. S1B, and the final rate constant values shown in supplemental Fig. S1C. Parameter scanning was used to predict the effect of the starting concentration of aspirin in the medium on the maximum cellular concentration of aspirin (supplemental Fig. S1E).
Anti-acetylated lysine antibody (250 μg.ml−1) (ImmuneChem, Burnaby, Canada) was used at 1:2000 dilutions 16 h 4 °C in immunoblotting experiments. Anti-Rabbit HRP (Sigma) was used as secondary antibody at 1:2000 dilutions for 1 h at room temperature.
Purification of Acetylated Peptides from Cultured Human Cells
Twenty-one 150 mm diameter plates were used to culture HeLa cells to ∼70% confluency in 20 ml DMEM, 10% FCS. In batches of seven plates, the cells were exposed for 6 h with DMSO only, aspirin in DMSO or aspirin-d3 in DMSO, to a final DMSO concentration of 0.13% and aspirin concentration of 5 mm. Cells were washed three times with PBS and for each set of seven plates, the resultant cell pellet was lysed in four pellet volumes of 6 m urea, 2 m thiourea in 100 mm Tris/HCl pH 8.5 (lysis buffer). The three lysates were sonicated on ice (Branson sonifier, narrow tip, 40%) for a total sonication time of 140 s with 20 s on, 20 s off cycles. Protein yields were determined by Bradford's assay to be 40–45 mg per lysate. Samples were reduced by addition of DTT to 1 mm for 30 min at room temperature, followed by alkylation with 5 mm iodoacetamide during centrifugation at 20,000 × g for 30 min at room temperature in the dark. Any remaining debris was cleared from supernatants by 0.2 μm filtration. Per condition, 40 mg protein was carried forward. Each was digested by incubation with 1:200 (w:w) ratio LysC/protein (200 μg, Wako, Japan) at room temperature for 4 h. Peptide samples were diluted four times with 50 mm ammonium bicarbonate before digestion each with 1:400 (w:w) ratio trypsin/protein (100 μg - SIGMA trypsin gold) for 16 h at room temperature. Digestions were halted by acidification with addition of 10% trifluoroacetic acid (TFA) solution to pH ∼2–3 (to ∼0.6% TFA v:v). Precipitate was removed by centrifugation at 3000 × g for 15 mins before 0.2 μm filtration. For each sample peptides were purified by C18 reverse phase chromatography using spin columns (two Waters Sep-Pak, 6cc, 1 g cartridges per condition) as described by the manufacturers. Peptides were eluted from columns by 70% ACN in 0.1% TFA. Peptide concentrations were estimated using OD 260 and OD 280 measurements and the Warburg-Christian method. A volume equivalent to 300 μg peptide was removed for each batch (for use as “Crude” analysis) and these along with the remaining peptide samples were lyophilized in a vacuum centrifuge attempting to avoid over-drying. The 300 μg Crude samples were each resuspended to a concentration of 0.85 mg.ml−1 in 0.5% acetic acid, 0.1% TFA and carried forward for MS analysis. The remaining peptides were resuspended in 2 ml IP buffer (50 mm Tris/HCl pH 8.0, 100 mm NaCl). Any undissolved peptides were removed by centrifugation at 20000g for 30 min. Peptide solutions were requantified by the Warburg-Christian method to be 3.9–4.2 mg. ml−1. To purify acetylated lysine peptides, immune affinity chromatography was used. Briefly: 60μl anti-acetylated lysine agarose beads (ImmuneChem) pre-equilibrated with IP buffer was mixed with each peptide solution for 16 h at 4 °C. The resin was washed three times with 1 ml IP buffer before elution of peptides with three washes with 100 μl 0.1% TFA. Peptide solutions were desalted using two 4 ply STAGE tips per prep, and lyophilized peptide elutions resuspended in 60 μl 0.5% acetic acid, 0.1% TFA. These were carried forward for MS analysis as 'IP' samples for each treatment.
For the SILAC experiment investigating dynamics of aspirin-mediated lysine acetylation a single 150 mm dish of cells was cultured for each time point, except the 8h aspirin, 0h recovery condition, which had 8 plates because it was being used as a reference (see Fig. 6 C for experimental design). All treatments and purifications were carried out essentially as described above but scaling down to account for lower starting protein amounts. Five SILAC mixes labeled A to E were prepared, each with initial total protein amounts of 2–5 mg, and with the three SILAC conditions being mixed 1:1:1 (protein w:w:w) according to Bradford's assay.
Half-lives of aspirin-mediated lysine acetylation signals.A, Anti-AcK immunoblot of 6.25 μg (protein) crude cell lystes from HeLa cells treated with 5 mm aspirin for the indicated times, before change to medium lacking aspirin (recovery). Bands...
MS Analysis of Peptide Samples
Peptide samples were analyzed by LC-MS/MS on a Q Exactive mass spectrometer (Thermo Fisher Scientific) coupled to an EASY-nLC 1000 liquid chromatography system (Thermo Scientific) via an EASY-Spray ion source (Thermo Fisher Scientific). Peptides were fractionated on a 75 μm × 500 mm EASY-Spray column (Thermo Scientific) over various gradient lengths from 90 min to 240 min. The following describes the typical analytical set-up, but further specific details of MS run conditions can be found within the raw data files. Precursor ion full scan spectra were acquired over (m/z 300 to 1,800) with a resolution of 70,000 at m/z 200 (target value of 1,000,000 ions, maximum injection time 20 ms). Up to ten data dependent MS2 spectra were acquired with a resolution of 17,500 at m/z 200 (target value of 500,000 ions, maximum injection time 60 ms). Ions with unassigned charge state, and singly or highly (>8) charged ions were rejected. Intensity threshold was set to 2.1 × 104 units. Peptide match was set to preferred, and dynamic exclusion option was enabled (exclusion duration 40 s). The mass spectrometry proteomics raw data files have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository (http://www.ebi.ac.uk/pride/archive/) with the dataset identifier PXD003530 for the label-free site ID analysis, and PXD004995 for the occupancy and SILAC half-life analysis.
MS Data Analysis
Raw MS data files were processed using MaxQuant software (version 126.96.36.199) (21, 22) and searched against UniProtKB human proteome (86749 sequences - 13/06/2012) and the built-in “contaminants” list. The variable modification for lysine acetylated by aspirin was defined in Andromeda to allow automated database searching for Acetyl-d3 K. Specificity was considered only for lysines, composition was set to H-1C2OHx3 (monoisotopic mass 45.029394924), position at peptide C termini was excluded, and two diagnostic peaks were defined; H8C7ONHx3 (128.1028942181) and H11C7ON2Hx3 (145.1294433196). For RAW data analysis in MaxQuant enzyme specificity was set to trypsin/P, cleaving C-terminal to lysine or arginine. Lysine and arginine were selected as special amino acid and a maximum number of three missed cleavages were allowed. Carbamidomethylation of cysteines was set as a fixed modification and oxidation of methionines, acetylation of protein N termini, acetylation of lysines and d3-acetylation of lysines were set as variable modifications. A minimum peptide length was set to seven amino acids and a maximum peptide mass was 5000 Da. A false discovery rate of 1% was set as a threshold at protein, peptide and site levels, and a mass deviation of 6 ppm was set for main search and 20 ppm for MS2 peaks. MaxQuant output files have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository (see above), and MS/MS spectra can be viewed on MS-Viewer (http://msviewer.ucsf.edu/prospector/cgi-bin/msform.cgi?form=msviewer) using the search keys 4iznb5zvfu (label-free study) and kjfmg24iwf (SILAC study).
The final list of aspirin-mediated lysine acetylation sites was created by filtering the MaxQuant output removing sites by the following criteria; (a) Peptides derived from the decoy database, (b) Sites with localization probability lower than 0.75, (c) Peptides with mass error greater than 2 ppm after mass recalibration. (d) Peptides of only bovine origin. (e) Peptides that had the best evidence for modification come from a sample not treated with aspirin-d3. Protein copy number values were calculated using the label-free proteomic ruler method (23) using a ploidy value of 3.4 based on the 76–80 chromosome estimate for HeLa cells, compared with the usual 46 for human cells.
Sequence Logo Analysis
Sequence logos were generated by plogo v1.2.0 (24). For aspirin sites 12069 31 residue sequence windows were submitted, of which 11370 were processed. For the endogenous sites 1473 were submitted, of which 1273 were processed. “Endogenous” sites were defined as those sites where the file containing the spectrum with the best localization probability was found in nonaspirin-treated cells. Background was the human proteome, n = 660373.
Secondary structure and solvent exposure predictions
All peptides were expanded to their whole protein “parent” and submitted to Jpred v4 (25) for protein secondary structure prediction. Jpred has length and time limits for making predictions where any protein longer than 800 amino acids or shorter than 20 amino acids, or where a prediction takes longer than 1 h will not get a prediction. Thus, a small number of proteins do not get a secondary structure prediction. DisEMBL disorder prediction (26) was performed via JabaWS (27). The central lysine residue for each peptide was then interrogated in its parent protein for its secondary structure and disorder prediction and assigned as being either in helical, beta strand or coil, and disordered or non-disordered region. Any peptides missing secondary structure predictions were ignored. The counts for each of the five states were summed across all peptides in a particular set and then compared between sets. As predicted secondary structure from Jpred can only be one of three states the comparison was at the level of proportion of each state between sets. The Wald method was used to determine the 95% confidence interval on the proportions of each state in each data set:
Where S is the number of lysines in a given state, and n is the total number of lysines in a given dataset. Similar methods were applied to solvent exposure calculations, which is also provided by Jpred. In this case the majority of lysines are predicted to be solvent exposed so comparisons were made on those with a predicted solvent exposure of 25% or less.
SILAC Analysis of Aspirin-mediated Acetylation Decay
MS data derived from the five SILAC mixtures described above were processed by MaxQuant essentially as described above for the label-free experiment. The match between runs, and requantify options were selected in MaxQuant to allow identifications among “experiments” to be used in all, and to provide ratios for modified peptides even in absence of a heavy counterpart. The reported list of identified sites was filtered to remove: (a) Peptides derived from the decoy database, (b) sites with localization probability lower than 0.75, (c) peptides with mass error greater than 2 ppm after mass recalibration, (d) peptides with no reported SILAC ratios. This removed 2061 sites. SILAC ratios were manually normalized using the median normalization factor created by MaxQuant for the Crude data, for each reported ratio. These normalized SILAC ratios for each “Mix” were used to calculate relative acetylation at each time point. Acetylation was set to 100% for time t = 8 h exposure to aspirin-d3 (0 h recovery), and % acetylated for all other time points was calculated by average of two methods using all three SILAC ratios reported in a single mix. For example, for “Mix D” (L = 8 h Aspirin, M = 8 h aspirin + 0.25 h recovery, and H = 8 h aspirin + 0.5 h recovery);
% acetylated at 8.25 h according to M/L ratio;
or % acetylated at 8.25 h according to H/L and H/M ratios;
% acetylated at 8.25 h,
% acetylated at 8.5 h according to H/L ratio;
or % acetylated at 8.5 h according to M/L and H/M ratios;
% acetylated at 8.5 h;
Data points corresponding to mixes without SILAC ratios were not included in the analysis. Data were fit to a single phase exponential decay by non-linear regression using GraphPad Prism version 6.0f (GraphPad Software, Inc.). Plateau was constrained to 0% acetylation and only positive K values were considered. Any sites reported by Prism as “Too few points,” “Hit constraint,” or “Not converged” were rejected. For the remainder, only sites with R squared value of 0.85 and a minimum of 9 data points were shortlisted. Any of these with acetylation profiles characteristic of exogenous proteins (extremely short half-lives and potentially of bovine origin), were rejected. Of a total of 6942 sites remaining after initial filtering, this secondary filtering left a list of 1480 sites of aspirin-mediated acetylation of human proteins with good quality decay data.
Acetylation site Occupancy Calculation
A SILAC experiment was conducted with HeLa cells grown in the heavy (H) condition treated with DMSO, and the light (L) condition treated with 5 mm aspirin-d3 for 8 h (Fig. 5B). A third SILAC condition (medium or M) was included, which was treated with 5 mm aspirin-d3 for 15 min. This was included in case of issues associated with requantification of acetylated peptides in H/L comparisons, but eventually acetylated peptide ratios were not required for occupancy calculation, so the M condition was unnecessary. This SILAC mixture is referred to as “occupancy mixture” or “Mix O.” The cells from Mix O and the resultant MS data were processed in parallel with those used for acetylation half-life analysis (described above). Acetylation site identifications from all mixtures (Mixes A to E and Mix O) were used to define a set of unmodified counterpart peptides as described in Fig. 5A. All unmodified counterpart peptides identified and quantified from Mix O were used to calculate acetylation site occupancy using equation 7. See Fig. 5C and 5D for derivation.
Occupancy% = [1 − (Pr/Ur)] × 100
The lysine site occupancy of aspirin-mediated acetylations is very low.A, For a protein (p) acetylated at lysine (K), digestion by trypsin yields a modified peptide (m). The unmodified protein can yield two unmodified counterpart peptides (u′...
Where Pr and Ur are the H/L ratios for the total protein and unmodified counterpart peptide of each acetylated peptide. For reference, occupancy calculations were also applied to a control set of peptides not designated as unmodified counterparts. These were used as a presumptive negative control set with which the known unmodified counterparts could be compared.
Other Statistical Analyses
Survival assays were made with n = 4 replicates and differences between conditions measured by Student's nonpaired two tailed t test assuming equal variance. For acetylated lysine half-life analysis, the 1480 sites were divided into 6 subsections based on half-life using 6 h bins: 1, 0–6 h; 2, 6–12 h; 3, 12–18 h; 4, 18–24 h; 5, 24–30 h; and 6, >30 h. Annotation of known acetylation, ubiquitination and SUMOylation sites, GO terms, Pfam families and KEGG pathways were annotated to each site in Perseus using the most up-to-date annotations files and comparison with published data (28, 29). Relative enrichment of these in each sub-section was calculated by Fishers exact test in Perseus. Only values with a Benjamini-Hochberg FDR of <1% in at least one sub-section were used in the heatmap. Hierarchical clustering was performed using the Euclidian method calculating distances by averages using Perseus-reported enrichment factor scores for each category in each subsection.
Comparisons among subsections for significant differences between numerical means of protein half-life (from reference (30)) used the Student's t test (comparing each sub-section with an equally sized random sample of the entire set). The two-tailed test was employed using either a two-sample equal variance method, or two sample unequal variance method depending on whether the variance ratio was less or greater than 1.5 respectively.
For occupancy calculations, comparisons between the values calculated for true unmodified counterpart peptides, and those not thought to be unmodified counterparts, the two-tailed unpaired Student's t test was applied using Welch's correction.