Article pubs.acs.org/jpr
Quantification of Citrullination by Means of Skewed Isotope Distribution Pattern Marlies De Ceuleneer,† Katleen Van Steendam,† Maarten, Dhaenens,† Dirk Elewaut,‡ and Dieter Deforce*,† †
Laboratory for Pharmaceutical Biotechnology, Ghent University, Ghent, Belgium Department of Rheumatology, Ghent University Hospital, Ghent, Belgium
‡
S Supporting Information *
ABSTRACT: Citrullination is a post-translational modification of arginine, resulting in a loss of positive charge and a 1 Da mass increase. Research on citrullinated proteins is crucial in rheumatoid arthritis, an autoimmune disease characterized by the presence of antibodies against citrullinated proteins. However, little is known about the location or quantity of deiminated arginine residues in these proteins. Since citrullination gives rise to a mass gain of only 1 Da, the isotope pattern of the citrullinated and the noncitrullinated version of a peptide will overlap. However, the difference between the theoretical, or noncitrullinated, and the measured isotope pattern can be used to quantify the amount of citrullination. We developed a method to quantify citrullinated peptides by means of their skewed isotopic distribution pattern. The method was first optimized with synthetic peptides, after both direct infusion and RP-HPLC separation on an ESi-QqTOF mass spectrometer. Additionally, we analyzed synovial fluid samples from rheumatoid arthritis patients and were able to quantify citrullinated peptides originating from citrullinated fibrinogen, a well-known antigen. KEYWORDS: citrulline, rheumatoid arthritis, label-free quantitation, posttranslational modification
■
INTRODUCTION Citrullination (or deimination) plays a role in many physiological processes, most notably apoptosis and epigenetic regulation, but also in pathologies such as cancer, multiple sclerosis, and rheumatoid arthritis.1 It is carried out by peptidyl arginine deiminases (PADs), but this reaction might be selfregulating: PADs reach maximum citrullination efficiency 3 h after their activation by calcium-influx, after which they autocitrullinate.2 Due to this phenomenon, it is likely that citrullination of arginine residues is never complete. Therefore, it is of interest to investigate to which extent this posttranslational modification takes place. We chose to investigate citrullination in the context of rheumatoid arthritis (RA). This autoimmune disease is characterized by the presence of antibodies against citrullinated proteins (ACPA) in 70−75% of patients, with a specificity of 95−99%.3 However, citrullination in itself is not specific for RA, and citrullinated proteins have been found in the synovial fluid of both RA and spondylarthropathy patients.4 The question remains whether the nature of the antigen or the antigen load might play a part in the fact that some of these patients develop antibodies against citrullinated antigens. In order to investigate whether the extent of citrullination of a certain antigen plays a role in autoimmunity, the need arises for quantitative © 2012 American Chemical Society
techniques that can help determine a possible threshold for triggering of autoimmune reactivity. Different problems arise in the quantification of citrullinecontaining peptides. Citrullination causes a 1 Da mass gain and the loss of a positive charge, which changes the properties of a citrullinated peptide during LC-MS. The different interaction with the RP-HPLC column can cause a shift in retention time, causing the citrullinated counterpart to elute at a later, but sometimes overlapping, time span. While it is possible to adjust the LC-gradient to ensure complete separation of both peaks to facilitate identification by MSMS, this would be timeconsuming to optimize for all citrullinated peptides present in a sample and would be impractical for complex samples where the citrullination status might not always be known for all peptides. Moreover, since citrullination in vivo is likely present in very limited quantities,5 the complete separation will most probably result in a low ion count for the citrullinated peptide, hampering correct identification by tandem MS. Additionally, the loss of a positive charge will change the charge state of the peptide, which means that the citrullinated and noncitrullinated version of the same peptide will not Received: May 16, 2012 Published: September 17, 2012 5245
dx.doi.org/10.1021/pr3004453 | J. Proteome Res. 2012, 11, 5245−5251
Journal of Proteome Research
Article
maximum of 1 mL of sample buffer (50 mM TrisHCl pH 6.8, 2% SDS, 10% glycerol, 5% β-mercaptoethanol), incubated for 5 min at 95 °C, loaded onto a 10% SDS-PAGE gel (Criterion Tris-HCl gel, IPG+1lane, BioRad), and run for 30 min at 150 V followed by 1 h at 200 V. After electrophoresis, a small part of the gel was used for Western blotting and the rest was cut into 0.5 cm wide horizontal strips.
automatically share the same charge state distribution. So far, only one research group has been able to quantify in vivo citrullinated peptides from rheumatoid arthritis synovial tissue.5 They estimate the citrullination of two peptides from fibrinogen-α (ESSSHHPGIAEFPSRGK and KREEAPSLRPAPPPISGGGYRARPAK) between 1.2 and 2.5% by means of label-free MS quantitation. However, they do not take into account possible differences in charge state distribution due to citrullination. Another attempt at quantification of citrullination was made by Andrade et al.,2 who used iTRAQ-labeling to identify citrullinated peptides in a mixture of autocitrullinated and noncitrullinated PADs. While these authors mentioned the possibility of using iTRAQ-labeling as a way to quantify citrullination, they did not use it as such and therefore had less need of taking into account possible differences in charge state distribution or overlap of MS spectra due to coelution of citrullinated and noncitrullinated peptides. Moreover, this method was carried out on artificial mixtures of noncitrullinated and in vitro citrullinated PADs with the aim to identify citrullinated residues in the citrullinated protein, and it has not been executed on in vivo samples. It is clear that there is a need for a straightforward method to quantitate citrullination in in vivo samples. In this work, we propose a practical way to estimate citrullination content from qualitative data by means of the skewed isotope pattern of a partially citrullinated peptide. When a peptide is partially citrullinated, the isotopic distribution patterns of the citrullinated and the noncitrullinated PADs will overlap and cause a deviation from the expected isotopic distribution pattern. We proved that this skewing in the isotope distribution pattern of the lowest charge state of a given peptide corresponds linearly with the percentage of the citrullinated form of that peptide present in the sample. We also applied this method to synovial fluids of RA patients, in order to verify the usefulness of our method for in vivo research.
■
Western Blot and Immunodetection of Antigens
Proteins were transferred to a nitrocellulose membrane by means of transblotting with CAPS (50 V, 3 h). Afterward, vimentin and fibrinogen were detected by means of immunodetection with polyclonal antivimentin antibodies (Abcam Ab7114) and polyclonal antifibrinogen antibodies (Abcam Ab 6666). Briefly, membranes were blocked with 1% BSA/0.3% Tween in PBS and incubated overnight with primary antibody diluted in blocking buffer (1/5000 for antivimentin, 1/10000 for antifibrinogen). After incubation with secondary antibody for 1 h (goat antirabbit, Pierce and rabbit antigoat, Abcam, respectively), bands were visualized by means of Supersignal West Dura Extended Duration Substrate (Pierce). Antimodified Citrulline Staining
For detection of citrullinated proteins on blot, the AntiModified Citrulline kit (Millipore) was used according to the manufacturer’s instructions. In Gel Digest
Horizontal strips from the gel corresponding to the location of fibrinogen and vimentin or other citrullinated proteins as determined by Western blot were subjected to in gel digestion. Selected bands were first washed three times in 250 mM TEABC (Sigma-Aldrich). After reduction with 25 mM TCEP (Sigma-Aldrich) for 2 h at 60 °C and alkylation with 200 mM MMTS (Sigma-Aldrich) for 1 h at room temperature, proteins were digested overnight with 2 μg of chymotrypsin (Promega) at room temperature or GluC (Promega) at 37 °C. Afterward, extraction of the peptides was performed sequentially with 50%, 75%, and 100% acetonitrile.
EXPERIMENTAL SECTION
RP-LC
Peptides
All separations prior to mass spectrometry were performed on a U3000 nano-HPLC device (Dionex) with a Pepmap C18 column (15 cm, 5 μm particle size, 100 Å pore size, Dionex). Buffers were 0.1% formic acid (A) and 0.1% formic acid in 80% acetonitrile (B). Synthetic peptides were separated in a short run (4% B to 60% B in 15 min, followed by 5 min of 100% B). All other samples were separated over 70 min (4−100% B).
All synthetic peptides were obtained from Thermo Scientific and were reconstituted in Milli-Q water (Millipore) and stored at −80 °C at a concentration of 10 mM. Patient Samples
Synovial fluids were collected from patients after informed consent, approved by the ethics committee of the Ghent University Hospital.
Mass Spectrometry Analysis
In Vitro Citrullination of Human Fibrinogen and Cell Lysates
Data were acquired on a Waters Premier (ESI-QqTOF) mass spectrometer after RP-LC separation. For tandem MS, collision induced dissociation was performed with custom collision energy profiles ranging from 25 eV to 55 eV for doubly charged peptides between m/z 400 and 1200 and ranging from 11 eV to 26 eV for triply charged peptides between m/z 435 and 1000.
In vitro citrullination was performed on human fibrinogen purified from plasma (Calbiochem, Darmstadt, Germany) or Hep-G2 cell lysates in deimination buffer (0.1 M trisHCl pH 7.4, 100 mM CaCl2, 5 mM DTT), by adding 7 IU of PAD from rabbit skeletal muscle (Sigma-Aldrich, St. Louis, MO, USA) per milligram of protein. After 2 h at 37 °C, the reaction was stopped by adding EDTA to a final concentration of 50 mM.
Identification of Peptides
Identification of peptides was performed by means of a Mascot Distiller (Matrix Science, version 2.3.2.0). Peaks were picked between m/z 50 and 100.000, with a correlation threshold set to 0.7 and a minimal signal-to-noise of 2. Baseline correction was applied for all spectra, and peaks were fitted by means of isotope distribution. Mascot searches were all performed against the SwissProt database (version 57.15, human taxonomy, 20 266 sequences) with methylthio on cysteine
Fractionation by Means of SDS-PAGE
When samples were too complex for direct infusion, fractionation was performed through gel electrophoresis. For synovial fluids, the sample was first treated with 1U hyaluronidase (Sigma-Aldrich) per 10 μL of sample (30 min, 37 °C). Afterward, the sample was dried and reconstituted in a 5246
dx.doi.org/10.1021/pr3004453 | J. Proteome Res. 2012, 11, 5245−5251
Journal of Proteome Research
Article
Figure 1. Calculated citrullination percentages for known mixtures of the citrullinated and noncitrullinated forms of synthetic peptides. (A) The process of calculating the extent of citrullination by means of the Excel template. Dashed lines denote the theoretical isotopic distribution pattern for the peptide STSRSLYASSPG (m/z 1212.5, [1+]); full lines denote the distribution pattern as measured in a sample containing 40% of the citrullinated peptide. In the left panel, the theoretical distribution pattern is shown for the peptide without partial citrullination. On the right, the theoretical distribution pattern is fitted to the measured distribution pattern, and from its alteration, the extent of citrullination is calculated. For this sample, containing 40% of STSCitSLYASSPG, the citrullination content was calculated to be 37.9%. (B) Calculated percentages of citrullination after direct infusion of samples containing increasing amounts of citrullinated peptide. Data are shown for all peptide charge states present in the samples. Trend lines were fitted by linear regression, and goodness-of-fit of the data to the calculated citrullination content is displayed (r2). For both peptides, the lowest charge state gives the highest correlation with the true citrullination content.
residues as fixed modification and deamidation of arginine, glutamine, and asparagine and oxidation of methionine as variable modifications. The peptide mass window was 0.3 Da, and the window for fragment ions was 0.45 Da. The #13C parameter was set to 0, to exclude the possibility that a peak caused by partial citrullination was wrongly annotated as the second peak in an isotope cluster. All reported peptides were significant (p < 0.05) in Mascot’s protein family summary.
the percentage of citrullination is calculated by dividing the summed intensities of the citrullinated form with the summed intensities of all forms of the peptide. About 20% of deamidated and deiminated peptides showed good chromatographic separation between the modified and the nonmodified forms for the gradient described above. Most of the peptides showed partial or complete overlap of the chromatographic peaks. When the modified and the nonmodified versions of a peptide did not coelute completely, data were obtained by integrating the extracted ion chromatograms of both together.
Calculation of Citrullination Percentages
For calculation of the amount of citrullination, the isotopic envelope mixture model described by Dasari et al.6 was implemented in Excel using the Solver Add-In to determine the extent of deamidation (Supporting Information). First, the theoretical isotope distribution is obtained from the “Isotope model” function in MassLynx and converted by the Excel template into a graph with as many data points as the experimental data based on hypothetical and user-defined intensity, mass error, and background noise values. Summed intensities are calculated for the noncitrullinated and the citrullinated forms of the peptide separately, where the latter will equal zero. The Solver Add-In, which optimizes the difference between two data points, is then used to minimize the intensity difference between the experimental and the theoretical data by altering these intensity, mass error, and background parameters. From this new theoretical distribution,
Specific Modification of Citrullinated Peptides
Specific modification of citrullinated peptides in mixtures was carried out by means of 2,3-butanedione (Sigma-Aldrich) in an acid environment as described previously.7
■
RESULTS AND DISCUSSION
Sample Preparation for the Analysis of Citrullinated Peptides
A factor that needs to be taken into account when analyzing citrullinated peptides is the amount of sodium ions in the sample. It became apparent that a citrullinated and a noncitrullinated version of the same peptide behaved differently in the presence of sodium and that the citrullinated peptide was 5247
dx.doi.org/10.1021/pr3004453 | J. Proteome Res. 2012, 11, 5245−5251
Journal of Proteome Research
Article
Figure 2. Optimization of the method after RP-nano-LC separation. (A) Comparison of low and high percentages of citrullination. Below 10% citrullination, no correlation can be found between the amount of citrullination present in the sample and the percentage calculated from the raw data. When the sample is citrullinated for 10% or more, a good linear correlation is found. Goodness-of-fit of the experimental data to the calculated citrullination content is displayed (r2). (B) Calculated citrullination percentages for known mixtures of the citrullinated and noncitrullinated form of synthetic peptides, measured after separation by nano-HPLC (n = 2−3). Trend lines were fitted by linear regression.
Aside from the change in mass, citrullination also causes the loss of a positive charge. This has important consequences for the charge of the peptide in an ESI source: the citrullinated form of a peptide will have a lower charge state than the noncitrullinated counterpart. This was investigated by means of two synthetic peptide pairs: SAVRARSSVTGVR/SAVRACitSSVTGVR and STSRSLYASSPG/STSCitSLYASSPG. The citrullinated and noncitrullinated peptide were mixed together with the percentage of citrullination ranging from 0% to 100%, increasing in 10% increments. Data were obtained after direct infusion into the mass spectrometer and analyzed by means of an excel template based on the isotopic envelope mixture model described by Dasari et al.6 In short, raw data are pasted into the template, together with the theoretical isotopic distribution of the peptide of interest, which is calculated by means of the “isotope model” function incorporated in MassLynx. By means of the Solver Add-in, the difference between both graphs is calculated and the amount of skewing of the isotopic pattern can be determined (shown in Figure 1A for STSRSLYASSPG (1+) with a theoretical citrullination content of 40%. The percentage of citrullination for this sample was calculated to be 37.9%). When analyzing the different charge states of a peptide, we could conclude that the lowest charge state of a given peptide showed the highest agreement with the true percentage of citrullination (Figure 1B). For this lowest charge state, the slopes of linear regression were 1.072 (± 0.1035) for SAVRARSSVTGVR and 1.003 (± 0.03557) for STSRSLYASSPG, indicating that the measured percentage was a good gauge of the true amount of citrullination present in a sample. It is not surprising that higher charge states are more biased toward the presence of arginine, since arginine can hold a charge and citrulline cannot.
more likely to form sodium-adducts (data not shown). Therefore, all samples were desalted for 30 min on a C18 trapping column before nano-LC separation and subsequent MS analysis. Additionally, the possibility that citrullination altered ionization capacity was investigated, since the loss of a positive charge might diminish ionization capacity. For correct quantification based on peak areas, ionization of both peptides needed for the calculation should be comparable.8,9 Summed peak intensities of all charge states of pure synthetic peptides were comparable for the citrullinated and the noncitrullinated form of the same peptide, showing that citrullination does not change ionization capacity (data not shown). Isotopic Distribution Pattern Is a Measure of Citrullination Content of Peptides
The 1 Da mass gain introduced in a peptide by citrullination of an arginine residue will alter the isotopic distribution pattern of this peptide when compared to its noncitrullinated counterpart. In the case of a complete enzymatic reaction, the spectrum will shift 1 Da but will retain the same isotopic distribution pattern, because the change in isotopic distribution introduced by substitution of a nitrogen and a hydrogen atom by an oxygen atom would be too small to be detected even by the most accurate mass spectrometers.10 However, when the citrullination of the peptide is partial, the isotopic distribution patterns of the citrullinated and the noncitrullinated form of the peptide will overlap. Since the presence in the environment of the different isotopes of an element is known and fixed, the isotopic distribution pattern of a peptide can be predicted with high confidence. If the isotopic distribution of a peptide is altered due to citrullination, the difference between the theoretical and the measured isotope profile can be used to calculate the amount of citrullination. 5248
dx.doi.org/10.1021/pr3004453 | J. Proteome Res. 2012, 11, 5245−5251
Journal of Proteome Research
Article
Table 1. Three Citrullinated Peptides from Our Proteins of Interest Could Be Identified and Quantified protein
peptide sequence
expectancy (p)
m/z (charge)
% skewing
VIME_HUMAN
IKTVETRDGQVINETSQHHDDLE LNDRFANYIDKVRFLEa TNTKESSSSHHPGIAEFPSRGKSSSY
0.00082 1.8 × 10−005 0.00039
888.7 (3+) 672.05 (3+) 898.09 (3+)
63.1% 100% 100%
FIBA_HUMAN a
No distinction could be made between citrullination and deamidation of asparagine, since both options were identified by Mascot.
Optimization of the Method after LC Separation of Peptides
Citrullinated Loci in in Vitro Citrullinated Vimentin and Fibrinogen
The above data were collected by direct infusion of synthetic peptides into the mass spectrometer, to provide proof-ofconcept of the method. However, when analyzing real samples, the addition of an RP-LC step prior to mass spectrometry analysis is preferred, to ensure maximal concentration of individual peptides. This results in a much lower sample load needed. Therefore, we expanded our method validation to include an LC separation step. First, we tried to evaluate the accuracy of our method when looking at small and large amounts of citrullination. As can be seen in Figure 2A for the peptide SAVRARSSVTGVR, there is no linear correlation once the amount of citrullination drops below 10%. Therefore, all calculated amounts that fall below 10% will be displayed as “