Liquid Chromatography and Mass Spectrometry Behavior of Citrullinated Peptides
Like many acidic PTMs, citrullination is known to hinder trypsin cleavage near the site of modification . Unlike phosphorylation and other acidic modifications, which add negative charge near the basic (Arg or Lys) cleavage site, conversion to citrulline, which is neutral, reduces the basicity of the amino acid side chain, decreasing trypsin affinity to the site. It is possible to use other enzymes (Lys-C, Glu-C) without encountering missed cleavages . As the modified arginine is not cleaved, database search results that identify peptides with C-terminal citrullination require manual validation . Many of these sites are reclassified as deamidation of either asparagine or glutamine (exactly the same mass shift as citrullination) after manual validation. The stoichiometry of citrullination is such that, without enrichment, the noncitrullinated form of a peptide is dominant and the concentration of the citrullinated peptides may be outside the dynamic range of the mass spectrometer.
In cases where there are both citrullinated peptides and noncitrullinated counterparts, it has been shown that the citrullinated version of the peptide elutes later from a C18 RP column. This feature can be used to confirm the presence of citrullination. Bennike et al.  showed for 24 synthetic peptide sequences (48 peptides total, 24 citrullinated and 24 unmodified) that the retention times of the citrullinated peptides were longer than the noncitrullinated peptides in all but two cases. The maximum shift was observed for a triply citrullinated peptide from the PAD4 protein; a 3.6% increase in retention time was observed. Overall, for the 24 peptides, the average retention time shift was +1.8%. For a 60 min gradient that equates to a retention time shift of +65 s. With modern LC-MS/MS methods, these peptides would not overlap. Two of the peptides in the study by Bennike et al. were positional isomers (NMKEEMARHLcREYQDLLNVK and NMKEEMAcRHLREYQDLLNVK, where cR denotes citrullination site), in which the peptide sequence is identical but the site of modification varies. Interestingly, there was a difference of 0.6% in retention time between these two peptides. The peptide NMKEEMARHLcREYQDLLNVK elutes earlier of the two peptides, suggesting it is more hydrophilic. Like all PTMs, complete conversion of a target site from arginine to citrulline is unlikely. Given the mass shift of +0.984 Da, if the modified peptide does coelute with its unmodified form, with most mass spectrometers, the two would be coisolated for MS/MS fragmentation. (Typical isolation width for MS/MS analysis is m/z 1-3.) Coisolation would complicate the MS/MS spectrum and hinder if not completely negate the successful identification of the peptide and PTM . Because of the propensity for missed cleavages in tryptic digests of biological samples, very few of the citrulli- nated peptides will elute at the same time as the unmodified counterpeptide. This method is more likely to be of use for Lys-C or Glu-C digests.
As mentioned earlier, citrullination results in an identical mass shift to deamidation of asparagine and glutamine (0.984 Da). Consequently, identification of citrullinated peptides by protein database searches is more difficult. Protease digestion performed in ammonium bicarbonate, Tris-HCl, or triethylammo- nium bicarbonate results in significant amounts of artificial deamidation of asparagine , further complicating the database search. It must also be noted that deamidation of Asn and Gln also results in an increase in RP LC retention time in comparison to the unmodified peptides . The retention time shift observed was similar to those seen for citrullination (~3% increase).
CID  of citrullinated peptides has been shown to result in intense peaks corresponding to neutral losses from the precursor ion . The functional group of citrulline contains an N-CO bond. As mentioned earlier, CID fragmentation predominantly cleaves this bond on the peptide backbone. It is cleaved in citrulline side chain, leading to the loss of isocyanic acid (43 Da, HNCO) (see Figure 7.1). Hao et al.  fragmented synthetic citrullinated peptides with CID and observed the neutral loss of isocyanic acid from both the precursor and fragment ions. The neutral loss from the precursor confirms that the peptide is citrullinated and the loss from fragment ions can be used to localize sites of citrullination. Any fragment ion with a neutral loss means that the site of citrullination is located on the remaining portion of the peptide. To test the method, nucleophosmin (NPM) (a molecular chaperone) was analyzed. NPM is known to be citrullinated, but the site of modification was not known. Hao et al. treated recombinant human His6-NPM with PAD4 and digested the modified protein with trypsin. LC-CID-MS/MS resulted in the identification of the peptide SIcRDTPAK. The peptide was identified as both a singly and doubly charged ion, and in both cases a peak corresponding to the expected neutral loss was observed. As there is only one arginine in the peptide, the site can be localized. The unmodified peptide was also identified in an analysis of a non-PAD4-treated sample.
Creese et al.  used automatic detection of the peak corresponding to the neutral loss of HNCO from the precursor ion in a CID spectrum to trigger reselection of the same ion and fragmentation by supplemental activation electron transfer dissociation (saETD) . The saETD mass spectra provide additional confidence in the localization of a citrulline modification rather than deamidation of asparagine or glutamine. Creese et al. analyzed a set of four synthetic citrullinated peptides. These peptides were spiked into a commercially available mix of tryptic peptides derived from six proteins and a more complex tryptic digest of human saliva. In the analyses of both the six protein mixture and the saliva digest, all four of the peptides were selected and fragmented by CID. In each case, a peak corresponding to the neutral loss of HNCO was observed, and the precursor ions were subsequently fragmented by saETD. The data were searched using the SEQUEST algorithm , and for all but one of the peptides, the ETD mass spectra resulted in higher scores than the corresponding CID mass spectra. It was also noted that analysis of the CID spectra resulted in up to a 1.1% false-positive rate (FPR) (incorrect assignment of citrullination—all peptides identified as citrullinated were manually validated), whereas for all of the saETD analyses the FPR was 0%.
Jin et al.  developed a similar method in which (HCD of the precursor ion was triggered by the neutral loss of the isocyanic acid in a CID spectrum. It was noted that HCD fragmentation of citrullinated peptides resulted in a significant reduction in the intensity of the neutral loss peak. In addition, more backbone ions were observed in the HCD spectra. Using the neutral loss- triggered HCD method, Jin et al. were able to characterize previously unlocalized citrullination sites on glial fibrillary acidic protein (GFAP). Two samples were analyzed: a purified GFAP sample analyzed without modification and the same protein treated with PAD2 prior to analysis. The samples were digested with trypsin, Lys-C, and Glu-C. From the PAD2-treated sample, 17 citrullination sites were identified, and 5 sites were identified from the untreated sample. If an identified peptide contained either asparagine or glutamine, the mass spectrum was manually validated. The benefit of both neutral loss methods is the improved confidence that the triggering of the second fragmentation technique gives. Both ETD and HCD give greater confidence in identification (peptide sequence and modification site) than CID; however, observation of the diagnostic peak corresponding to the neutral loss in the CID mass spectra confirms the presence of citrulline.
One method capable of distinguishing citrullination from deamidation utilizes treatment with 2,3-butanedione and antipyrine . The ureido group of citrulline reacts with 2,3-butanedione in the presence of trifluoroacetic acid. This in turn reacts with antipyrine (Figure 7.5), increasing the mass of amino acid residue by 238 Da. It is then possible to detect this modification using UV detection at 464 nm radiation, and modified citrullinated proteins can be identified by Western blot with antimodified citrulline antibodies.
The 2,3-butanedione reaction with the ureido group of citrulline alone can be used to distinguish citrullination from deamidation as shown by De Ceuleneer et al. . The reaction results in a mass increase of 50 Da, easily
Figure 7.5 The modification of citrulline using 2,3-butanedione and antipyrine catalyzed by trifluoroacetic acid.
distinguished from deamidated and unmodified peptides by LC-CID-MS/MS analysis. The reaction is performed by mixing citrullinated peptides with 2,3-butanedione in the presence of trifluoroacetic acid. De Ceuleneer et al. observed 95% reaction efficiency when converting a synthetic peptide (STScRSLYASSPG) with a 16 h reaction. The reaction was also performed on the unmodified counterpart peptides to determine the specificity of the reaction. It was found that only the citrullinated peptide was modified. The synthetic citrullinated peptide was spiked into a “complex sample” (cytochrome c, Lys-C, and Glu-C digests) and the reaction repeated. The conversion rate was limited to approximately 70% even with an overnight incubation. Human fibrinogen was citrullinated in vitro using PAD from rabbit skeletal muscle (PAD enzyme details are not provided) and digested with Glu-C or Lys-C. After analysis by LC-MS/MS and subsequent data analysis with in-house software (msMod searches for the difference in peptide ion signal between the unmodified and modified samples), 15 peptides containing 17 citrullination sites were identified. Of the 15 peptides, only 9 were identified in a Mascot search. It was suggested that the other six peptides were of too low abundance to be selected for fragmentation.
The reaction of citrullinated peptides with 2,3-butanedione and antipyrine is performed with the modifying reagents in large excess. It is therefore necessary for the sample to be purified prior to LC-MS/MS analysis. To remove the excess reagents, Stensland et al.  proposed desalting the sample with SCX chromatography followed by C18 RP LC to remove salts. Traditionally SCX is used as an additional dimension in complex proteomics analysis ; here it is solely used to remove excess reagent. Using this protocol, they were able to remove over 99% of the excess antipyrine. Stensland et al. used the aforementioned technique to modify citrullinated peptides created by incubating either bovine serum albumin (BSA) or MBP (human) with PAD4. To ensure identification of unmodified counterpeptides, the proteins were digested with Lys-C without incubation with PAD4. An alternating CID and ETD experiment was performed with the peptides separated on a C18 RP column. They observed that CID of peptides with modified citrulline residues resulted in an intense peak at m/z 201.1. This peak was attributed to fragmentation of the exocyclic methyl group to the C4 of the imidazolone moiety and was proposed as a potential trigger for other fragmentation techniques such as product ion- triggered ETD, as used for glycopeptide analysis . The ETD mass spectra of the modified peptides did not yield any fragments of the modification and as such could be used to localize the site of citrullination. LC-CID-MS/MS analysis of citrullinated BSA resulted in the identification of two citrullinated peptides (VPQVSTPTLVEVScRSLGK and LGEYGFQNALIVRYTcRK) whose fragmentation spectra were of sufficient quality to both identify the peptide and localize the site of modification. In both cases, an intense peak at m/z 201.1 was observed. The MBP Lys-C digest was analyzed using an alternating CID- ETD LC-MS/MS analysis. In this analysis, the ETD mass spectra were used to identify and localize the sites of modification, and the CID mass spectra were used to confirm the presence of the m/z 201 signature ion. From this analysis, four peptides were identified containing five sites of citrullination. The sample was also analyzed with a CID-only method from which only two of the four citrullinated peptides were identified. It has been shown by Holm et al.  that the use of 2,3-butanedione alone results in multiple additional ions, mass increases including 66, 116, and 162 Da, whereas the reaction with antipyrine only results in a mass increase of 238 Da. Holm et al. observed that the product in the reaction between citrulline and 2,3-butanedione is an intermediate species and the addition of antipyrine reacts with this intermediate, but it is proposed that other nucleophiles could react to produce these unwanted products if antipyrine is not present.
In 2005, Kubota et al.  developed a novel labeling method for the identification of citrulline. The method is reliant on in vitro citrullination of proteins by one of the PAD enzymes. The reaction is performed in 50% heavy water (H218O). When the PAD enzyme converts arginine to citrulline 50% of the citrulline will contain heavy oxygen. The sample is then digested with trypsin, chymotrypsin, and Glu-C prior to LC-CID-MS/MS analysis. The incorporation of 18O gives the peptide ion a distinct isotope pattern, which can be detected in the mass spectrometer. To ensure all fragment ions shared this isotope distribution, a wider isolation of the ion was used (5 m/z) prior to fragmentation. Kubota et al. used the Mascot search algorithm to identify citrulli- nated peptides from the data set. Mass spectra assigned to peptides with multiple arginine residues and ambiguous sites of modification were manually analyzed. Fragment ions retaining the citrulline had distinct isotope patterns.
The obvious drawback of this method is the need to citrullinate the proteins in vitro, and therefore it cannot be used for the analysis of endogenous citrullina- tion. This method has its uses for purified proteins, which are known to be citrullinated by particular PADs.