Age-Related Changes in Human Crystallins Determined from Comparative Analysis of Post-Translational Modifications in Young and Aged Lens: Does Deamidation Contribute to Crystallin Insolubility?

Lens sample preparations

Lens proteins from a 93-year old human male with nuclear cataract (Grade III, Pirie scale) were separated into water-soluble and water-insoluble fractions, digested with trypsin, and 2-DLC MS/MS analysis performed as previously reported.³ Peptides were analyzed on two different instruments: an LCQ Classic ion trap and a Q-TOF hybrid mass spectrometer. This data set (43,518 spectra) is the same as that used by Searle, et al.³ and will be referred to as the “validation” data.

Three additional human lenses [3-day old, 70-year old normal, and 70-year old nuclear cataractous (Grade II, Pirie scale)] were obtained from the Lions Eye Bank of Oregon with IRB approval, dissected within 12 hours of death, graded for cataract severity, photographed, and frozen at -70°C until use. Individual lenses were homogenized on ice in a 20 mM phosphate, 1.0 mM EDTA buffer (pH 7.0) using 1.0 ml of buffer per lens, and the homogenate centrifuged at 20,000 g for 30 min to pellet the water-insoluble proteins. The water-insoluble pellet was resuspended once in the same volume of homogenizing buffer, and again pelleted. The resulting pellet was re-suspended by brief sonication (5 sec × 2) on ice, and the protein content of both soluble and insoluble fractions determined in triplicate using a BCA assay (Pierce Biotechnology, Inc., Rockford, IL, USA). 2.5 mg portions of the samples were dried by vacuum centrifugation and stored at -70°C until analysis.

Dried water-soluble or water-insoluble lens samples (from the three lenses mentioned above and additional samples from the 93-year old cataractous lens that was previously analyzed) were dissolved in buffer containing 8M de-ionized urea, 0.8 M Tris, 0.08 M methylamine, and 8 mM CaCl₂ (pH 8.5). Cysteine residues were reduced and alkylated by successive treatment with DTT and iodoacetamide, and sequencing grade modified trypsin (Trypsin Gold from Promega, Madison, WI, USA) added at a ratio of 1:25 protease to substrate, resulting in a dilution of the urea to a final 2M concentration. Digestion occurred during an 18 hour incubation at 37°C with shaking. Following digestion, formic acid was added to a final concentration of 5%, and peptides solid phase extracted using Sep-Pak Light cartridges (Millipore, Billerica, MA, USA).

2-DLC separations and MS/MS acquisition

The Sep-Pak cleaned digests of lens proteins were injected onto a 100 × 2.1 mm polysulfoethyl A cation exchange column (The Nest Group, Inc., Southborough, MA, USA) at a 200 μl/min flow rate. Mobile phase A contained 10 mM sodium phosphate (pH 3.0) and 25% acetonitrile, and mobile phase B was identical, except it contained 350 mM KCl. Following 5 minutes of loading and washing in mobile phase A, peptides were eluted using a linear gradient of 0-50% B over 45 min, followed by a linear gradient of 50-100% B over 20 min. One-minute fractions were collected, dried by vacuum centrifugation, and re-dissolved by shaking in 100 μl of 5% formic acid. Fractions at the beginning or end of the salt gradient were combined, based on UV intensities, to reduce the number of fractions to about 40 per lens sample. Portions of each fraction (10 μl) were analyzed by LC/MS using an Agilent 1100 series capillary LC system and an LCQ Classic ion trap mass spectrometer (Thermo Electron, San Jose, CA, USA) using a standard electrospray source fitted with a 34 gauge metal needle (ThermoFinnigan, Cat. No. 97144-20040). Samples were applied at 20 μl/min to a trap cartridge, and then switched onto a 0.5 X 250 mm Zorbax SB-C18 column (Agilent Technologies, Palo Alto, CA, USA) using a mobile phase A containing 0.2% acetic acid. The gradient was 7-35% acetonitrile over 90 minutes at a 10 μl/min flow rate. This shallow reverse phase gradient was capable of completely resolving unmodified and deamidated peptide forms, verified with several pairs of synthetic peptides where Asn or Gln residues were replaced by Asp or Glu, respectively. Data dependent collection of MS/MS spectra used the dynamic exclusion feature of the instrument's control software (repeat count equal to 1, exclusion duration of 3 min, list size of 25) to obtain MS/MS spectra of the three most abundant parent ions following each survey scan. DTA files were created using extract_msn (ThermoFinnigan) from 400 to 4000 Da, with a minimum peak count of 25, and charge state analysis (ZSA, ThermoFinnigan) was performed to assign higher charge states and greatly reduce the number of duplicate DTA files. Each lens fraction was analyzed identically, and approximately 60,000 MS/MS spectra were collected per lens sample for a total of 439,116 MS/MS spectra.

Identification of modified peptides

Spectra were searched against a database of crystallins and other abundant lens proteins (35 proteins; 12,000 amino acids). The MS-Alignment algorithm⁵⁰ was used to generate a list of putative modifications in the sample of the form (aa, Δ), where Δ is a characteristic offset of PTM on amino acid aa. We examined each modification that occurred in at least 100 out of the 439,116 spectra with match quality score greater than zero. We rejected modifications that were judged to be artifacts (addition or subtraction of amino acids at peptide termini, or abundant modifications incorrectly assigned to an adjacent amino acid, etc.), and compiled a list of putative modification sites in the lens proteins.

To refine the list of modifications, we performed restrictive searches using the list of putative modifications provided by MS-Alignment. Sometimes “new” modifications could be explained by combinations of previously reported modifications. For instance, the modification (M, +58) resulted from an oxidation (+16) of an N-terminal methionine residue that was acetylated (+42). To screen for sites that could be explained in this way, we used InsPecT⁵¹ to perform the searches allowing up to three modifications per peptide from the restricted list of modifications. Parent ion mass corrections and a fragment ion mass tolerance of 0.5 Da were used for ion trap data. Fragment ion mass tolerances were 0.1 Da and no parent ion mass corrections were performed for QTOF data.

Validation of identified modifications

While existing MS/MS database search tools rigorously assess statistical significance of peptide interpretations, evaluating reliability of MS/MS interpretations of modified peptides remains an open problem. We chose a conservative approach to evaluate the reliability of identified modified peptides and performed manual validation on all four of the “validation” data sets. The lists of modified peptides from InsPecT for each of the four fractions were converted to data formats that could be read by the MS/MS spectrum viewer from OpenSea ³. The OpenSea spectrum viewer was configured to highlight y-ions, b-ions, a-ions, and loss of water or ammonia from major ions. For visualization purposes, fragment ions were highlighted if they were within 0.25 or 0.40 AMU of the calculated masses for the QTOF or ion trap, respectively. Modification sites, modification mass shifts, and the mass differences between the measured parent ion and the calculated modified sequence were displayed along with the spectra. Interactive visualization tools allowed zooming regions of spectra, measurement of the mass differences between spectral peaks, and simultaneous viewing of paired spectra of modified and unmodified peptides.

Manual validation accepted modified spectra satisfying the following criteria:

1. Correct parent ion mass shift values,

2. Appropriately shifted y- and b-ions confirming the modification mass and site,

3. Consistent fragment ion pattern between unmodified peptides and modified peptides (keeping point #5 below in mind),

4. Modifications located at peptide termini (or beyond the low and high mass cutoffs for the ion trap) were rejected if there was any potential assignment ambiguity,

5. The presence (or absence) of known modification specific fragmentation ion spectral features:

   1. Neutral loss of 43 Da associated with carbamylation

   2. Neutral loss of 98 Da from phosphorylated serine

   3. Neutral loss of 64 Da from oxidized methionine

   4. Enhanced cleavage C-terminal to deamidated asparagines,

6. Higher mass fragment ions were required to have realistic isotopic distributions,

7. Missed trypsin cleavages were required for modified arginine or lysine residues,

8. Peptides containing modified basic sites (K, R, N-terminus) were required to elute in an earlier SCX fraction (if the corresponding unmodified peptides from missed trypsin cleavage could be identified)

Manual validation would have taken too much time to be practical for analyses of additional lenses so an alternative automated validation strategy was developed. OpenSea, X! Tandem, and InsPecT searches of the same data sets were performed and spectral annotations compared on a spectrum by spectrum basis.

X! Tandem searches were configured using the 24 most frequent mass shifts and affected amino acids reported from MS-Alignment determined during manual validation. Due to the large number of potential modifications necessary to account for the InsPecT results, only peptides with tryptic termini were allowed in the X! Tandem searches. In addition, “blind” OpenSea analyses³ of the data sets were also performed using previously reported parameters for ion traps³ without amino acid substitutions allowed. Both X! Tandem and OpenSea used a human species subset of the current Swiss-Prot protein database (12,077 entries). Without invoking any scoring thresholds, the interpretations where all three programs agreed on modification and site were considered correctly assigned (“SameSite” consensus). A diagram of the analysis workflow is shown in Figure 1.

Integrated ion current measurements

Theoretical monoisotopic masses of unmodified and deamidated peptides were used to extract single-ion chromatograms using QualBrowser (ThermoFinnigan) for a 1.0 Da range centered on the specified mass values. Chromatogram peaks were integrated and summed across all SCX fractions (typically 3 or 4 fractions) where the peptides of interest were observed. The total ion current of each peptide form was used to estimate the extent of deamidation (deamidated ion current divided by sum of both unmodified and deamidated ion currents), expressed as a percentage.

Analysis of previously acquired 93-year old human cataractous lens data sets

MS-Alignment was configured to find blind modifications of masses from -200 to +200 Da, and used to analyze the “validation” data set. A list of putative modifications (mass shifts and affected amino acids) was used by InsPecT in a restrictive search allowing up to three modifications per peptide. A total of 378 distinct modification sites were identified from analysis of the 43,518 spectra⁵⁰ (four separate 2-DLC analyses). Manual validations of all modified peptides were performed as described above. A site was considered valid in a particular 2-DLC run if at least one spectrum contained sufficient evidence to confirm peptide sequence and localize the modification. Examinations of all spectra associated with a specific modification were used to classify the site as “yes”, “no”, or “maybe”. Combined evidence across the four data sets was used to access whether or not the site was considered confidently identified. Manual validation results are listed in Supplemental Table 2.

Modifications were grouped into likely in vivo modifications and known in vitro artifacts (N-terminal carbamylation, N-terminal pyroglutamic acid, potassium adducts, and parent ion loss of water and ammonia). Of the 378 modification sites present in all lens proteins, there were 219 putative in vivo PTM sites. The number of those sites passing manual validation was 149, and 101 sites were in crystallins. We emphasize that these sites are the most reliable PTMs and it is likely that some of the rejected sites are actually false negatives.

The automated validation strategy was tested on the “validation” data. The total number of modification sites reported by MS-Alignment/InsPecT was 378, there were 228 sites (either in vivo or in vitro) that passed manual validation, 196 sites had “SameSite” consensus status (X! Tandem, OpenSea, and InsPecT predictions agreed), and 178 of the consensus sites also passed manual validation. The consensus validation procedure had slightly lower sensitivity compared to manual validation (196/228=0.86), had a reasonable specificity (178/196=0.91), and, most importantly, could be fully automated for high-throughput PTM studies. The consensus results are also listed in Supplemental Table 2.

Analyses of additional human lenses

To verify that previously unreported modifications were age-related and not in vitro artifacts, 2-DLC analyses of three additional lenses (3-day old, 70-year old normal, and 70-year old cataractous) and repeat analyses of the 93-year old cataractous lens were performed. Due to the small number of lenses analyzed, no attempt was made to find modifications unique to cataracts. For each of the three aged lenses, proteins were separated into water-soluble and water-insoluble fractions. The young lens contained such a small amount of water-insoluble protein that only the water-soluble fraction could be analyzed. The seven ion trap 2-DLC runs produced 439,116 MS/MS spectra which were searched for PTMs using MS-Alignment/InsPecT. A total of 63,936 spectra had MQScores exceeding 1.0 and 18,719 spectra contained modifications (29%).

Although replicate analyses of each lens were not performed to access reproducibility of the reported results, we adopted several methods to ensure that PTM sites were reliably identified. The large number of SCX fraction for a less complex mixture, such as the lens, resulted in frequent redundant observation of modified peptides, and we required a minimum of two modified spectra for each candidate PTM site. On average, approximately 3 distinct peptide forms (tryptic, semi-tryptic, and missed cleavage peptides) were observed per reported PTM site, increasing the confidence in the detected sites. Modifications were individually tabulated for each lens, and only modifications that were observed in at least two of the three aged lenses were reported. Agreement of results from different analysis software can reduce random identifications, and all annotations having an MQScore greater than 1.0 were tested for “SameSite” consensus by three different analysis programs.

We used the young 3-day old lens as a control to classify the PTMs into likely in vitro modifications and putative in vivo PTMs. Many types of modifications were seen with high frequency in the 3-day old lens and in the aged lenses, and were classified as in vitro artifacts (carbamylation, salt adducts, loss of water, loss of ammonia, and pyroglutamic acid). Since some proportion of the detected deamidations could be introduced during sample processing or be non-deamidated forms of the peptides misidentified by the ion trap, deamidation abundance estimates (described below) for the aged lenses were compared to the corresponding abundance estimates for the 3-day old control lens, and results filtered to remove potentially misidentified deamidation sites. Carbamylated lysine residues were treated in a similar fashion.

A total of 491 putative PTM sites passed “SameSite” consensus and were seen in at least two of the aged lenses (Supplemental Table 3). After excluding likely in vitro artifacts, the remaining number was 352 sites. The number of PTM sites in major crystallin proteins was 220, which included 99 potential deamidation sites. Based on the number of assigned spectra, a chi-square test with one degree of freedom was used to identify the 46 deamidation sites that were more abundant (P>0.05) in the aged lenses compared to the 3-day old control. In the chi-square test, observed spectral counts were compared to expected values derived from average PTM rates in the two sets of data being compared. A p>0.05 would correspond to a chi square value in excess of 3.84. Similarly, 4 of 16 carbamylated lysine sites were more abundant in the aged lenses. Following this analysis, the final tally of identified aged human lens PTM sites was 155. The in vivo PTM sites for the nine major crystallins are shown in Figure 2 (α crystallins), Figure 3 (β crystallins), and Figure 4 (γ crystallins). There were four deamidation sites (N49, Q66, Q67, and N160) in γD seen in all three aged lenses with relative modification rates in excess of 10% that did not pass the chi-square test due to the low abundance of γD in the young lens. There were also 4 known sites of protein N-terminal acetylation in βA1/A3, βA4, and βB3 that were detected in the young control lens but did not pass the filtering criteria. The relatively low abundances of βA2, βB3, and γB resulted in too few modification sites (four) to warrant inclusion in the figures.

We detected several novel PTMs in the aged lenses that were absent in the 3-day old control lens. We identified numerous lysine modifications (carbamylation, carboxymethylation, and carboxyethylation) at many sites, and one example is shown for βB1 K117 in Figure 5. Novel arginine +55 Da modifications were observed at 11 sites and MS/MS spectra from three different peptides are shown in Figure 6. The lysine and arginine modifications occurred at sites of missed trypsin cleavage and, in some cases, the unmodified peptide was also observed. If a basic site in a peptide was modified, such as N-terminal carbamylation, there was a corresponding shift to earlier SCX elution times. In all cases where the unmodified missed cleavage peptide was observed, the lysine or arginine modifications caused a large (approximately 10 min) change to earlier SCX elution time indicating a modified basic site (data not shown).

PTM abundance estimates from spectral counting

The large numbers of modified spectra observed in this study suggested that PTM abundances could be estimated by spectral counting.^53-55 In this initial study, we have assumed that the relative ion currents of modified peptides would be similar to unmodified peptides, and that the selection process of ions for fragmentation would, therefore, be similar for all peptides. Use of synthetic peptides during chromatography testing (see Methods section) indicated that this assumption was valid for deamidated peptides in simple mixtures. We have also assumed that modified peptides undergo similar fragmentation to their unmodified forms and that identification rates by the analysis software would also be the same. However, more labile modifications such as phosphorylated serines (relatively less abundant in lens) would likely have suppressed spectral counts due to poorer fragmentation and identification. It is possible that absolute PTM abundances based on spectral counts could differ from their true abundances if these assumptions do not hold true and we have, therefore, compared relative abundances between lens samples processed under identical conditions to determine changes associated with aging and loss of crystallin solubility.

For each amino acid residue, the numbers of spectra matched to sequences containing that amino acid were tabulated for the unmodified amino acid and all distinct modified forms of that amino acid (if any). The modification abundance estimate was calculated as the ratio of the spectral count of a particular modified amino acid form to the total spectral count for that amino acid (all modified and unmodified forms). The spectral counts of the eleven major types of modifications at the 155 in vivo crystallin sites were averaged across the soluble and insoluble fractions of the three aged lenses for each type of modification, and are listed along with data from the 3-day old control lens in Table 1. The data in Table 1 are average modification abundances over the amino acids that had detectible modifications, and are not the overall modification abundances in the samples. Protein N-terminal acetylation (a co-translational modification), oxidation, and cysteine methylation were abundant, but deamidation was by far the major PTM present in aged human lens crystallins.

The fraction of water-insoluble protein in human lens increases with age, and eventually becomes more abundant than the water-soluble protein.⁵⁶ A reasonable hypothesis for this is the accumulated damaging effect of PTMs on crystallin structure and stability. Starting with the 491 PTM sites listed in Supplemental Table 3, we tested for sites that were more abundant in aged lenses and also more abundant in the water-insoluble fraction. A chi-square test with one degree of freedom was used to identify sites more abundant (p>0.05) in aged lenses than the 3-day old control, and sites more abundant (p>0.05) in the water-insoluble fractions than in the water-soluble fractions. One carbamylated lysine, five methylated cysteine sites and 36 deamidation sites from lens proteins passed the tests. No other sites passed the filtering. Figure 7 shows the abundances for the 31 crystallin deamidation sites passing the above criteria that were also greater than 2-fold more abundant in the water-insoluble fractions compared to the water-soluble fractions.

PTM abundance estimates from spectral counting

There are few quantitative data reported for lens modification abundances making it difficult to access the accuracy of the spectral counting abundance estimates. In Lapko, et al.,⁸ the overall extent of methylation was reported for γD (56%) and γC (12%) in a similar aged lens (72-year old). These overall lens average values were calculated using the reported weights of each dissected region and taking a weighted averaging of the two abundance values. In comparison, our observed rates were 66% for C110 in γD, and 16% (C22) and 14% (C79) for two sites in γC, in good agreement with the above published values. In an earlier report,²⁴ a value of 31% was reported for γS, which is predominantly methylated at C26, and we observed a methylation rate of 43% at this site.

We were able to integrate peptide ion currents for 14 unmodified peptides and their deamidated forms taken from the sites shown in Figure 7. These peptides had a sufficiently intense 1+ ion which improved the accuracy of the extracted ion chromatograms. A comparison between the average deamidation abundances determined from the integrated ion currents and the abundances based on spectral counting is shown in Figure 8 for the water-insoluble fractions. A positive correlation (correlation coefficient of 0.83) indicated good agreement between the two methods to estimate deamidation abundances. We also checked the method using the (protein) N-terminal acetylated peptides in βB2, αA, and αB where the co-translational modifications were expected to have relative abundance estimates of 100%. We measured values of 92%, 95%, and 96%, respectively. We consider the spectral counts to be tentative abundance estimates until additional comparisons can be made with relative peptide ion currents and stable isotope ratio measurements that are planned.

Age-related changes in human lens

The small number of lenses processed so far provides a preliminary picture of the aging changes occurring in the human lens. PTM abundance estimates were averaged over all three aged lenses to minimize the unknown uncertainty associated with the spectral counting estimates, and comparisons between the water-soluble and water-insoluble fractions were made. Deamidation was found to be the major age-related PTM and was significantly more abundant in the water-insoluble fractions. Deamidation introduces a negative charge into the protein and may alter the tertiary structure.⁵⁷ Accumulated deamidations may lead to partial protein denaturation and disrupt protein-protein interactions^{22, 58} necessary for the highly ordered structure of the lens. It is clear from Figure 7 that there is a striking increase in deamidation rates associated with water-insolubility. Asparagine sites are expected to deamidate faster than glutamine sites,^{59, 60} and it is not surprising that 74% of the sites are asparagine, and the 12 most extensively deamidated sites were at asparagine residues.

Since deamidation causes only a +1 Da mass shift, it was possible that ion trap mass inaccuracy contributed to incorrect assignments of some deamidation sites. Several precautions were taken to reduce this effect. Reverse phase chromatography conditions were employed so that unmodified and deamidated peptides did not co-elute. Mass corrections using y- and b-ion pairing were done to reduce parent ion mass errors that could result in unmodified peptides being incorrectly interpreted as deamidated forms. Lastly, deamidation abundances in the aged lenses were compared to the 3-day old control lens, and only those sites where deamidation was increased relative to the young lens were reported (see Results section). Our criteria for accepting identified deamidation sites was extremely conservative since 24 of the 53 rejected sites were, in fact, previously reported sites of deamidation.

Another significant finding was the far greater extent of S-methylation than was previously known in lens. The estimated abundances for S-methylation in αA, βB1, and βB2 were very low compared to previously identified sites in the γ crystallins and βA3. The mechanism of S-methylation in the lens remains unknown, but the wide variety of methylated sites suggests an enzyme able to recognize cysteine sites independent of primary sequence or three-dimensional structure. There was no methylation in the 3-day old lens and additional lenses at intermediate ages are needed to determine when methylation first occurs in the lens.

Carboxymethyl lysine (CML) and carboxyethyl lysine (CEL) are advanced glycation end products and have been suggested as general markers of advanced age and oxidation.^{61, 62} This study is the first to identify specific CML and CEL sites in lens crystallins, however, there did not appear to be any correlation between CML/CEL abundances and protein insolubility (data not shown), and these modifications were far less abundant than deamidation or methylation (see Table 1). Methionine and tryptophan oxidation (+16) can occur during sample processing and we did not observe a significant increase in +16 oxidation rate between the 3-day old control lens and aged lenses.

The novel R+55 modification is similar in mass to previously reported⁶³ advanced glycation end-product modifications of arginine in lens, but higher mass accuracy measurements are necessary to determine the chemical structure of the modification. We were able to confirm serine and histidine +28 modifications³ at very low levels in the lens, and determining the identity of these modifications will also require further experiments.

Blind PTM search results

Blind searches of large data sets containing high frequencies of modified peptides, such as lens, produce rather extensive lists of putative modifications. Modifications on closely located amino acids often produce similar theoretical spectra, thus making it difficult to identify the exact position of modification. These can be difficult to assign reliably if the mass accuracy of the instrument is limited or the spectrum is of poor quality. Multiple modifications per peptide are even more complicated, and, unfortunately, occur frequently in aged lens samples. It is challenging to distinguish between a sum of known modifications and a potentially novel single-site PTM, a problem that grows when larger numbers of potential modifications are considered. Also, if partially tryptic peptides are considered, addition or subtraction of amino acids can be difficult to distinguish from larger PTM mass shifts.

Requiring PTM site agreement from multiple analysis programs was one way to address these issues, and was used in place of time-consuming manual validation. Since it was possible that the high sensitivity of the MS-Alignment/InsPecT analysis could result in less accuracy because of lower quality spectra, we chose two additional programs having different approaches to PTM detection. OpenSea tries to match PTMs to de novo sequencing results and only the higher quality spectra will be considered. X! Tandem uses a more traditional database searching method, and a restricted number of modifications and consideration of only tryptic peptides (due to performance limitations) may increase scoring accuracy. However, consensus is more likely for shorter peptides or peptide having only one or two possible modification sites (fewer ambiguous site possibilities), for modifications of a reasonable delta mass value (larger than +1 Da), and for higher quality spectra. This automated validation scheme likely had a high false negative rate, but was necessary to generate reliable annotations for the large number of modified MS/MS spectra in this study. Although many spectra identified by MS-Alignment/InsPecT did not pass our strict validation protocol, this does not imply that all rejected MS-Alignment/InsPecT interpretations were incorrect. There are undoubtedly many lens PTM sites that we rejected in this initial study, and future experiments will likely identify additional novel lens modifications.

The previously reported methylated arginine, methylated lysine, oxidized tyrosine, and phosphorylated tyrosine modifications by MacCoss et al.⁶ could not be confirmed. That study analyzed a 4-year old congenital cataractous lens using 2-DLC MS, and used different proteases to create overlapping peptides. Since a normal young lens was not processed as a control, it is difficult to determine if the unique modifications were associated with the congenital cataract. It is also possible that the modifications they observed were below the sensitivity of our experiment. However, there have been several analyses of aged lens since MacCoss et al. published their results, and none of those modifications have been reported in any aged lenses (see Supplemental Table 1). Unfortunately, deamidation was not considered in their analysis.

There were many PTMs identified in our data that are not shown in Figures 2 through 4. Many peptides were observed where loss of water or ammonia had occurred. There did not seem to be any correlation between S-18 sites and any of the known phosphorylation sites, so -18 and -17 PTMs were classified as in vitro artifacts in our analysis. We observed a large number of N-terminally carbamylated peptides, probably as a result of using urea as a denaturant. We distinguished in vivo carbamylated lysines from artifactual modifications by comparing carbamylation rates between the aged lenses and the 3-day old control lens processed under identical conditions, and their appearance at missed cleavage sites. There were several tryptic peptides that had glutamines located at the N-termini, and the most abundant forms of those peptides were often the pyroglutamic acid forms (a -17 Da N-terminal modification). Supplemental Table 4 is a partially annotated PTM frequency table generated from the complete MS-Alignment/InsPecT results and gives a sense of the PTM complexity in aged lens proteins.

There were many modifications detected at sites in non-crystallin lens proteins such as filensin and phakinin that passed “SameSite” consensus validation, but were not summarized here (see Supplemental Table 3). Also, some lower frequency modifications were not included in the X! Tandem searches, but may have been proper in vivo modifications (W+4 is a likely conversion of tryptophan to kynurinine, for example). In Supplemental Table 5, we list the putative sites of six additional modifications that were present in multiple crystallins and in peptides containing single modifications. Verification of these modifications will be the subject of future studies.