Large-scale Identification of Endogenous Secretory Peptides Using Electron Transfer Dissociation Mass Spectrometry*

Mass spectrometry-based unbiased analysis of the full complement of secretory peptides is expected to facilitate the identification of unknown biologically active peptides. However, tandem MS sequencing of endogenous peptides in their native form has proven difficult because they show size heterogeneity and contain multiple internal basic residues, the characteristics not found in peptide fragments produced by in vitro digestion. Endogenous peptides remain largely unexplored by electron transfer dissociation (ETD), despite its widespread use in bottom-up proteomics. We used ETD, in comparison to collision induced dissociation (CID), to identify endogenous peptides derived from secretory granules of a human endocrine cell line. For mass accuracy, both MS and tandem MS were analyzed on an Orbitrap. CID and ETD, performed in different LC-MS runs, resulted in the identification of 795 and 569 unique peptides (ranging from 1000 to 15000 Da), respectively, with an overlap of 397. Peptides larger than 3000 Da accounted for 54% in CID and 46% in ETD identifications. Although numerically outperformed by CID, ETD provided more extensive fragmentation, leading to the identification of peptides that are not reached by CID. This advantage was demonstrated in identifying a new antimicrobial peptide from neurosecretory protein VGF (non-acronymic), VGF[554–577]-NH2, or in differentiating nearly isobaric peptides (mass difference less than 2 ppm) that arise from alternatively spliced exons of the gastrin-releasing peptide gene. CID and ETD complemented each other to add to our knowledge of the proteolytic processing sites of proteins implicated in the regulated secretory pathway. An advantage of the use of both fragmentation methods was also noted in localization of phosphorylation sites. These findings point to the utility of ETD mass spectrometry in the global study of endogenous peptides, or peptidomics.

of endogenous peptides that gain specialized functions not ascribed to original precursor proteins. For a precursor protein to generate such peptides, it must undergo specific cleavages and in some cases needs to be modified at specific sites (1). This limited cleavage, or proteolytic processing, represents an important cellular mechanism by which molecular diversity of proteins is increased at the post-translational level. In the postgenome era, it is being recognized that localization of processing sites in secretory proteins facilitates the identification of biologically active peptides. A standard approach to determining such sites is to use a panel of antibodies directed against different regions of a target protein (2). However, it is practically impossible to prepare antibodies that can thoroughly cover potential processing products arising from the precursor. Alternatively, mass spectrometry-assisted unbiased analysis of endogenous peptides may be a major step toward elucidating proteolytic processing (3).
In neurons and endocrine cells, a majority of biologically active peptides are released via the regulated secretory pathway. They are stored in secretory granules and await secretion until the cells receive an exocytotic stimulus. Owing to their compartmentalization, secretory peptides can be noninvasively recovered in culture supernatant. We have shown that a data set of endogenous peptide sequences that are collected by this procedure is applicable to infer processing sites, as well as to identify bona fide processing products (4). Rather than being digested, every endogenous peptide should be analyzed in its native form to understand how the peptide is generated and subsequently degraded. However, it remains a challenge to identify endogenous peptides because of size heterogeneity (ranging from 3 aa to 100 aa). For example, thyrotropin-releasing hormone is a small 3-aa peptide, human adrenomedullin occurs as a 52-aa peptide, and a 98-aa N-terminal propeptide from the atrial natriuretic peptide precursor is found in the circulation. Unlike digested protein fragments used in bottom-up proteomics, C termini of these endogenous peptides are not restricted to specific residues. Furthermore, proteolytic processing leads to the production of peptides containing multiple internal basic residues, for which collision induced dissociation (CID) 1 shows limited performance (5).
A solution to address this issue in endogenous peptide sequencing might be the use of electron transfer dissociation (ETD) tandem mass spectrometry, which has been shown to provide a more complete series of fragment ions and hence a more confident sequence identification, along with the ability to leave labile post-translational modifications intact (6 -10). The benefit of ETD in bottom-up proteomics has been increasingly documented, whereas endogenous peptides remain largely unexplored by ETD, despite the expectation that ETD would improve sequencing for larger peptides. In the few studies on endogenous peptides (11,12), ETD did not cover large peptides exceeding 5000 Da. Because we have used CID to facilitate the discovery of previously unknown biologically active peptides (3,13,14), we were interested to see if ETD would be helpful to identify endogenous peptides that have escaped identification by CID. Here we conducted a large-scale identification of endogenous secretory peptides, ranging from 1000 to 15000 Da, using CID and ETD. We describe the merits of using ETD, in connection with CID, in peptidomics studies. The most significant finding is the identification of a previously unknown peptide, VGF[554 -577]-NH 2 , which was sequenced solely by ETD. This peptide was found to have antimicrobial activity.

EXPERIMENTAL PROCEDURES
Sample Preparation-The human islet cell line QGP-1 (ϳ1 ϫ 10 7 cells) (15) was stimulated with 10 M carbachol plus 50 mM potassium chloride in serum-free Hank's balanced solution. Five minutes after stimulation, the medium was harvested to recover released peptides. The supernatant was quickly subjected to solid extraction as previously described (4). The eluate was separated by gel filtration highperformance liquid chromatography (HPLC) to five fractions containing peptides in the approximate M r range 1000 to 15000.
Liquid Chromatography-Tandem Mass Spectrometry-All data were acquired on an LTQ Orbitrap XL instrument equipped with ETD (Thermo Fisher Scientific, San Jose, CA). A nanoFrontier HPLC system (Hitachi, Japan) was connected to the mass spectrometer for liquid chromatography-tandem mass spectrometry (LC-MS/MS). About 300 to 500 ng of the peptide mixture was loaded onto a trap column (C18, 75 m ϫ 100 mm) and separated on a MonoSpray C18 tip (GL Sciences) using a 60-min gradient from 10 to 50% acetonitrile in 0.1% formic acid at a flow rate of 200 nl/min. A protonated ion of polycyclodimethylsiloxane with m/z 445.120025 was used for internal calibration throughout. The mass spectrometer alternated between a full FT-MS scan (m/z 400 -1500) and subsequent MS/MS scans. Scans were all recorded in the Orbitrap with a resolution of 100,000 at m/z 400. CID and ETD were performed on separate LC-MS runs. Cations were isolated with an isolation window of 5 m/z units and provided on a dynamic exclusion list for 2400 s after selected for at least two MS/MS scans. Singly charged precursors were excluded. Monoisotopic selection was disabled with an exclusion window setting to 1 Da. The four most intense ions were chosen for CID fragmentation. Automatic gain control was used to accumulate sufficient fragment ions (MS/MS target value, 2E5; maximum injection time, 1000 ms). For ETD, the two most intense precursor ions were subjected to MS/MS with an ETD activation time of 90 ms for doubly charged ions. Charge state-dependent activation time was applied. The ETD activation time was optimized for enhancing peptide identification by testing four different periods of time (70, 90, 150, and 200 ms) and also consistent with a previous report conducted on endog-enous peptides using an ETD-enabled LTQ-XL (10). Supplemental activation was used for all MS/MS scans (16). ETD spectra were acquired using one microscan per spectrum. Automatic gain control was also used (MS/MS target value, 5E5; fluoranthene, 1E6; maximum fill time, 1500 ms). Samples were not subjected to reductive alkylation. However, an aliquot of the sample was reductive alkylated using dithiothreitol and iodoacetamide to identify peptides derived from chromogranin B (CgB), whose N-terminal region contains two cysteine residues forming a disulfide bond.
Data Analysis and Peptide Identification-Peak picking, deisotoping, and deconvolution of MS/MS spectra were conducted using Mascot Distiller (version 2.2.3) with the default parameters for Orbitrap. Peak lists were searched against NCBI nr human (14,987,464 sequences; 5,132,678,026 residues, as of August 13, 2011) using Mascot (version 2.2.3) with no enzyme specification. Pyroglutamination, C-terminal amidation, N-terminal acetylation, and methionine oxidation were simultaneously allowed as variable modifications. Data were searched with a precursor mass tolerance of 5 ppm and product ion mass tolerance of 50 mmu. Peptides were considered identified with a Mascot expect value of less than 0.05. The false discovery rate for peptide matches above Mascot identity threshold using a decoy database search was 0% in all cases. The highest scored MS/MS spectrum was used to report the score of the unique peptide in supplemental Table S2. The percent fragmentation was calculated as reported (8), with N-terminal proline cleavage also included as a possible fragmentation channel. In Mascot notation, the even-electron c ion corresponds to the b ion plus NH 3 . Mascot does not try to match radical c-1 ions, which are usually much weaker. The charged radical zϩ1 ion is the y ion minus NH 2 . Hence, the most abundant z-type ions observed in ETD are radical zϩ1 and evenelectron zϩ2. For calculating percent fragmentation, z-type ions were counted only once if both zϩ1 and zϩ2 ions were assigned by Mascot for a given N-C␣ bond.
Peptide Synthesis-Peptides were synthesized by the solid phase method (Sigma Genosys, Japan) using Fmoc (N-(9-fluorenyl) methoxycarbonyl) strategy, purified by reverse phase HPLC, and verified for correct synthesis by MS and amino acid analysis. Peptide purity was confirmed on separate HPLC systems.
Antibacterial and Antifungal Activity Assay-Each test peptide (up to 10 M) was assessed with the target microbes Enterococcus hirae (E. hirae), Micrococcus luteus (M. luteus), Staphylococcus aureus (S. aureus) 209P, S. saprophyticus KD, Escherichia coli (E. coli) B, E. coli K12, E. coli kp, and Pichia pastoris (P. pastoris) GS115 using Alamar-Blue TM (BioSource International, Camarillo, CA) (14,18). The optimal growth temperature was 30°C for M. luteus, S. saprophyticus KD and P. pastoris GS115 and 37°C for the other microbes. After growth in 3% tryptosoy broth (Eiken Chemical, Tokyo, Japan) for 16 h with shaking at each optimal temperature, cells were washed twice with 10 mM phosphate buffer, pH 7.0, and diluted to 8 ϫ 10 5 colony-forming units/ml in the same buffer. Twenty-five microliters of bacterial suspension were mixed with an equal volume of sample in the absence or presence of peptides, and incubated for 1 h. After incubation, 200 l of 3% tryptosoy broth containing 10% alamarBlue TM was added to the reaction mixture and further incubated for the period of time shown in parentheses: E. hirae, S. aureus 209P and E. coli B (4 h), E. coli K12 and E. coli kp (6 h), M. luteus (7 h), S. saprophyticus KD (7.5 h), and P. pastoris GS115 (20 h). Aliquots containing all assay reagents without microbes were used as blank. After incubation, the reactions were monitored by absorbance at 569 and 600 nm. Molar extinction coefficients of OD 569 and OD 600 in the oxidative condition are 80586 and 117216. Viability (%) was expressed using the following formula: viability (%) ϭ (117216 ϫ OD 569 Ϫ 80586 ϫ OD 600 with peptide/117216 ϫ OD 569 Ϫ 80586 ϫ OD 600 without peptide) ϫ 100. The classical colony formation assay was performed on M. luteus, E. coli K12 and P. pastoris GS115 as described (18). Cathelicidin and ␤-defensin-2 (Peptide Institute, Osaka, Japan) were used as control. RESULTS We studied a complex mixture of peptides secreted by cultured human endocrine cells that received an exocytotic stimulus. The culture supernatant was harvested 5 mins after stimulation and separated by gel filtration HPLC to five fractions containing peptides in the approximate M r range 1000 to 15000. LC-MS/MS was performed in duplicate for each fraction on an LTQ-Orbitrap equipped with ETD. For comparison, the same sample was independently analyzed by CID in duplicate runs. Both MS and MS/MS scans were recorded with Orbitrap to ensure sequencing of highly charged peptide ions (zϾ4). The high mass accuracy and resolution enabled by Orbitrap allowed searching the human NCBI nr database with no enzyme specification and acquire significant data, and despite this broad search space the false discovery rate for peptide matches above the identity threshold using a Mascot automatic decoy database search was 0%.
In a total of 20 runs, 2311 and 1733 peptides (including redundant identifications) were identified by CID and ETD, respectively. In Table I, these peptides were sorted according to precursor charge states and fragmentation methods. No peptide was identified using ETD of doubly charged ions although supplemental activation was applied. Aside from this, there was no apparent difference in charge distribution between CID and ETD. More than two-thirds of the peptides were quadruply or more highly charged (69% for CID and 76% for ETD). This charge distribution is skewed, relative to that commonly observed with digested peptides used in bottom-up proteomics, most of which are doubly or triply charged ions. To assess the degree of peptide dissociation, we calculated the percent fragmentation values, defined by Coon et al. (8) as the number of observed fragment ions (band y-type ions for CID and c-and z-type ions for ETD), for a peptide sequence, divided by the total theoretically possible number of fragment ions. The percent fragmentation was plotted as a function of m/z values of the identified precursors ( Fig. 1, supplemental Table S1). In CID, charge states and precursor m/z values showed little influence on percent fragmentation. Consistent with the previous study (8,10), ETD percent fragmentation was decreased as precursor ion m/z increases regardless of charge state. Although surpassed by CID in the overall number of identification, ETD provided more extensive fragmentation; nearly 12% of the ETD-identified  peptides have percent fragmentation values greater than 75, as compared with 2% of the CID-identified peptides.
After removing redundant identifications caused by either peptide overlaps between two adjacent gel filtration fractions or ionization of the same peptide into different charge states, we concluded that CID and ETD elucidated 795 and 569 unique peptides, respectively, with the overlap being 397 (Fig.  2, supplemental Table S2). Peptides larger than 3000 Da accounted for 54 and 46% in CID and ETD identifications, respectively. Notably, nearly 96% of the peptides arose from secretory precursor proteins. Judged by the number of peptides, our analyte appeared to contain relatively large amounts of VGF and chromogranin B (CgB) (supplemental Table S2). They contain multiple pairs of basic residues, some of which have been described as cleavage sites for processing to biologically active peptides (2,19).
The identified peptides derived from VGF and CgB were aligned to each precursor sequence and illustrated in Fig. 3. They are shown by gray (CID) and black (ETD) lines and stacked to highlight potential processing sites. Multiple lines of the same length indicate redundant identifications. By inspecting the cleavage sites revealed by these lines, we inferred primary processing sites that involve at least one basic residue (dotted lines, Fig. 3). The VGF sequence was almost completely covered by clusters of peptides sharing N termini or C termini, whereas CgB showed less coverage. No peptide was ascribed to the signal sequence. Previously, a limited number of VGF-derived peptides were reported in peptidomics work on human cerebrospinal fluid using CID (20 -22). We conducted a more detailed study by analyzing the culture supernatant from a human endocrine cell line and tissue extracts in normal rat brain and gut (3,4,23). In earlier work, several CgB-derived peptides have been isolated from extracts of human pheochromocytoma and neuroendocrine tumors (24 -26). The specific processing sites that have been proposed by those studies are indicated in Fig. 3. Although differences in species and tissue sources should be taken into account, these sites extensively overlap the potential processing sites revealed by the present study.
On close inspection some peptides were identified solely by ETD, an example of which is VGF[554 -577]-NH 2 (Fig. 3, supplemental Table S2). This C-terminally amidated peptide underwent extensive fragmentation at every possible N-C␣ bond except the Pro-Ser (positions 572-573), whereas CID showed poor fragmentation (Fig. 4). The lack of ETD cleavage N-terminal to proline is consistent with the previous work on proline-containing synthetic peptides (7). In triplicate analyses, this peptide, detected as quadruply and pentuply charged ions, had a percent fragmentation of 54 to 61% and 56 to 63%, respectively. It might be noteworthy that N-terminally truncated 13-and 14-residue peptides sharing the Cterminal amide structure were solely identified by CID. Antibody specifically recognizing the amide structure was generated to characterize major peptide forms. Major peptides immunoprecipitated with the antibody from the culture supernatant had the molecular masses corresponding to the theoretical masses of VGF[554 -577]-NH 2 and [565-577]-NH 2 (supplemental Fig. S1). Similar results were obtained with the supernatant from the human medullary thyroid carcinoma cell line TT (data not shown). We thus concluded that these two peptides represent major processing products.
The three consecutive arginine residues in VGF[554 -577]-NH 2 prompted us to test antimicrobial activity, along with the related peptides (Table II). VGF[554 -583], identified in the present study, was tested because it was C-terminally flanked by the dibasic residues that correspond to the N-terminal processing site of VGF[586 -615]/AQEE-30 (Fig. 3a). VGF[554 -577]-NH 2 showed antimicrobial activity against M. luteus and P. pastoris, comparable to the established antimicrobial peptides ␤-defensin-2 and cathelicidin, whereas Cterminally extended VGF[554 -578] and [554 -583] were less potent. Up to the concentrations tested, VGF[565-577]-NH 2 was not active against any of the test microbes (Table II and Fig.  5). To examine whether VGF[554 -577]-NH 2 is bactericidal or just bacteriostatic, we performed a classical colony formation assay. It has been established that the bactericidal peptide concentration revealed by classical colony formation assays and the alamarBlue TM assay shows a good agreement (17). The peptide showed strong antimicrobial activity in the colony assay with IC 50 of 1.9 and 1.4 M against M. luteus and P. pastoris, respectively (data not shown), indicating that it is bactericidal.
It remains a challenge to pinpoint a post-translationally modified site among multiple candidate sites in a sequence, because informative fragment ions for unequivocal localization are not always observed in MS/MS spectra. A total of 13 phosphopeptides were identified using CID and ETD without any phosphopeptide enrichment procedure (Table III). As a consequence, five unique modified sites were identified from six peptides, of which Ser130 of CgB identified in 4684.98-Da

FIG. 2. Mass distribution of CID-and ETD-identified peptides.
Only unique peptides were counted. (Inset) Venn diagram illustrating the number of unique peptides identified by CID and ETD. and 5067.25-Da peptides and Ser64 of the calcitonin precursor identified in 6297.01-Da peptide have not previously been described. Two phosphorylation sites in CgB, Ser335 of 6109.65-Da and Ser617 of 7049.27-Da peptides, were previously described (27), but represent fortuitous cases where CID, rather than ETD, allowed successful identification. Ser405 of the CgB 3730.65-Da peptide, previously determined as a phosphorylation site through conventional techniques (26), was confidently reconfirmed by ETD (supplemental Fig. S2). We got evidence that the VGF-and secretogranin II-derived peptides are phosphorylated, but failed to localize modified sites.
Except for phosphopeptides, the unique peptide sequences listed in supplemental Table S2 excluded any peptide for which the second-ranked sequence also has a Mascot expectation value of 0.05 or lower. However, we encountered one exceptional case; CID of a positively charged ion (ϩ6) with m/z 808.24 resulted in the identification of the amidated peptide from gastrin-releasing peptide (GRP) isoform 1 preproprotein[95-139]-NH 2 (calculated monoisotopic mass 4843.377), with a Mascot score of 120 (expt. 3.5E-08). Ranked second was the peptide from GRP isoform 2 preproprotein[95-138] (calculated monoisotopic mass 4843.369), with a score of 96 close to the top (expt. 9.1E-06). The two peptides share N-terminal 27 residues but have alternatively spliced C-terminal halves (Fig. 6). The fragment ions produced by CID were preferentially observed the N-terminal region so that distinguishing these two possibilities was difficult. In ETD, the first rank was the peptide from GRP isoform 1 (score 289, expt. 4.5E-25) and the second was from GRP isoform 2 (score 82, expt. 0.00022). Although both hits still exceed a stringent threshold (expectation Ͻ 0.001, rather than a default value of 0.05), we concluded that this peptide arises from the isoform 1 because of the extensive matching to theoretical ions (Fig. 6). It might be worth mentioning that GRP isoform 1 preproprotein[54 -139]-NH 2 (calculated monoisotopic mass 9514.789), which is N-terminally adjacent to the amidation motif of neuromedin C, was also identified by CID (supplemental Fig. S2). DISCUSSION We have shown that mass spectrometric analysis of endogenous secretory peptides provides a basis for identifying bi-

FIG. 3. Identified peptides derived from VGF (GI: 17136078) (A) and CgB/secretogranin-1 (GI: 36439) (B).
Gray lines, CID-identified peptides; black lines, ETD-identified peptides. Red boxes, pyroglutamination; yellow boxes, C-terminal amidation. Note that the maps include redundant identifications, excluding oxidized methionine-containing peptides. Processing sites proposed by previous studies in human VGF (20 -22) and human CgB (2, 24 -26) are indicated by open arrows above the precursor sequence. In (A), arrows indicate processing sites identified in rat brain studies (13,23). Asterisks indicate processing sites proposed in our previous peptidomics study of a different human endocrine cell line (3). Processing sites inferred from the present study are shown in vertical dotted lines. Magenta bars, consecutive basic residues. Pale magenta bars, single basic residues. ologically active peptides (3,13,14). In the present study, our immediate interest is to uncover endogenous peptides that escape identification by CID. Endogenous peptides remain largely unexplored by ETD or electron capture dissociation, whose performance for such peptides has not been thoroughly characterized. ETD was used on a low-end ion trap mass spectrometer to characterize venom peptides extracted from a marine snail (28), resulting in a fraction of small peptides (less than 3000 Da) being sequenced. Shen et al. recently used the same instrumentation as ours to improve endogenous peptide identification, but identified peptides came from doubly to quadruply charged ions in most cases (12). Thus little is known at present about the ability of ETD to dissociate endogenous peptides with higher charge states, despite the occurrence of large endogenous peptides (Ͼ5000 Da) in larger amounts than previously thought.
It has been generally accepted that ETD is more effective than CID for sequencing highly charged precursors or larger peptides in bottom-up proteomics (6 -9). However, our pep-

Endogenous phosphopeptide identification
Mascot scores of first to third hits (columns 6 to 8) are accompanied by expectation values shown in parentheses. N-term, N-terminal flanking amino acid ("Signal" indicates that the peptide is flanked by signal sequence); C-term, C-terminal flanking amino acid. Identified phosphorylation sites are marked in bold and underlined. In the last column, the asterisk indicates that either of the two N-terminal serine residues is phosphorylated but not uniquely determined. CT, calcitonin; SgII, secretogranin II. tidomic data indicate that CID was as potent as ETD in sequencing precursors having charges greater than three, with their proportions being 69% for CID and 76% for ETD. Furthermore, CID appears to be superior to ETD in total identification; 795 peptides for CID versus 569 peptides for ETD. For peptides larger than 7000 Da, CID uncovered a larger number of peptides than ETD (39 for CID and 15 for ETD). Good et al. reported that a linear decrease in percent fragmentation as a function of increasing precursor m/z was noted for ETD, with few peptides having precursor m/z values exceeding 850 identified (8). We also observed a similar but less remarkable decline; in our study peptides identified with precursors greater than m/z 850 still accounted for 30% (507 of 1733) in the total ETD identification. This difference in ETD performance may be partly explained by the fact that most of our secretopeptidome components have more internal basic residues than in vitro-digested proteins. The current limitation in ETD was particularly noted with precursors having higher m/z values (m/z Ͼ1300). Supplemental Fig. S3 shows CID and ETD of two pentuply charged precursors with m/z 1324.6 and 1411.7. They were unequivocally identified by CID as 6618.0-Da VGF-derived peptide and 7053.5-Da Nterminal propeptide of the somatostatin precursor, whereas the corresponding ETD spectra were so dominated by charge reduced species that no significant match was obtained.
We provided the most comprehensive data about VGF and CgB processing. Both fragmentation methods complemented each other to highlight potential cleavage sites more clearly than that provided by either alone. Because they represent major components in neuronal secretory vesicles and several peptides identified here have also been described as neuropeptides (2,19,29), our findings will contribute to neuropeptide profiling in neuropeptidomics. The VGF processing sites predicted here are consistent with the antibody-assisted biochemical characterization of major processing products in rat brain (23). For instance, the 7400 Da peptide (also known as TLQP-62 (29) 6. Deconvoluted CID and ETD MS/MS spectra of sextuply charged m/z 808.24. Alternatively spliced exons of the GRP gene produce two nearly isobaric peptides. Residues different from the top peptide are underlined in the lower peptide. Note that the second best matches for both fragmentations are below a Mascot expect value of 0.05, but that ETD showed a wider difference between the best and second best matches because of extensive fragmentations across the sequence.
[600 -613](BAM-1745), and [617-676](CCB). All the peptides but CgB [326 -385] were identified in the present study. Flanked by either pairs or groups of basic amino acids, they are considered specific processing products. Despite the presence of potential cleavage sites across the sequence, no peptide has been described in the literature for the N-terminal 216-amino acid region. Our data revealed that the precursor undergoes extensive processing across the entire sequence, including previously reported processing sites (Fig. 3B). The reason for the lack of peptide identification for positions 540 -572 remains to be elucidated.
Some peptides were solely identified by ETD, including CgB-derived GAWK of 8788.13 Da (24) and VGF[554 -577]-NH 2 . This C-terminally amidated peptide had escaped MS/MS identification because of poor CID fragmentation in our previous peptidomic survey (3). In the present study, ETD was able to provide an almost complete series of ions to ensure confident identification; it has five proline residues, where cleavage N-terminal to proline is missing in ETD (7). Considering this proline effect, the peptide has 18 theoretically cleavable bonds rather than 23. As shown in Fig. 4, 17 of the 18 bonds were cleaved. Of note, related N-terminally truncated peptides were not revealed by ETD, demonstrating the advantage of using different dissociation techniques. The ability of ETD to cleave peptide backbone across the entire sequence was also demonstrated in distinguishing between the peptides derived from two different isoforms of the neuropeptide GRP gene, again reflecting the merits of ETD in peptidomics. The high mass accuracy and resolution attained by Orbitrap enables discrimination between zϩ1 and zϩ2 ions, ensuring in some cases that zϩ2 matches in Mascot are not 13C isotope peaks of zϩ1 matches.
In rat brain, TLQP-62/rat VGF[556 -617] was originally described as a major peptide recognized by antibody against the protein C terminus (29). The N-terminal 556Thr is 554Thr in human VGF. An appreciable number of the peptides starting with 554Thr were revealed by ETD, including VGF[554 -577]-NH 2 , [554 -583] and [554 -615] (Fig. 3A). Given the sequence identity between human and rat, the cleavage site (PR2TLQP) may represent a major processing site also in human. These peptides have not been described in peptidomic reports of VGF-derived peptides (3, 20 -22). On the other hand, the C-terminal cleavage site has recently been described through MS/MS identification of VGF[565-577]-NH 2 (3). Because 574RHYPGR579 agrees with the consensus motif for prohormone convertase (PC) 1/3 or 2 ((K/R)-(X) n -(K/ R)2, where n ϭ 0, 2, 4, or 6 and X is any amino acid other than Cys) (30) and is nested by a canonical amidation signal of 578GR579, generation of the amidated peptides can be explained by the general rule of proteolytic processing in the regulated secretory pathway (30). This is also supported by the fact that the complete set of enzymes involved in C-terminal amidation, PC1/3 or 2, carboxypeptidase E, and pepti-dylglycine ␣-amidating monooxygenase were expressed by the cell line used (not shown).
VGF[554 -577]-NH 2 showed antimicrobial activity against M. luteus and P. pastoris, comparable to major antimicrobial peptides ␤-defensin-2 and cathelicidin. The N-terminal 11 residues are considered essential for activity, because N-terminally truncated VGF[565-577]-NH 2 was inactive. In addition, the decreased activity observed with the two C-terminally extended peptides suggests that the C-terminal amide structure is important as well. It is known that amidated C termini, rather than common carboxyl termini, help to enhance electrostatic interactions between the peptide and a target bacterial membrane (31). This may also be the case with this peptide. The amidation motif is not found in rodent VGF sequences, but shared by primates (supplemental Fig. S1C), suggesting that the peptide is assigned some biological role unique to primates.
We were able to identify phosphopeptides of intermediate size (3500 to 8000 Da). Two phosphorylation sites, not described in the literature, were unequivocally localized. Of note, all but one (6109.65-Da peptide) of the 13 peptides were both N-terminally and C-terminally flanked by previously reported signal sequence cleavage sites or PC1/3 or PC2 consensus sites, strongly suggesting that they are bona fide secretory phosphopeptides. Although not explored, phosphorylated peptides quickly released via exocytosis may have some functional roles in the nervous and endocrine systems. Because larger peptides lead to less confident identification of modification sites than smaller peptides, ETD parameters should be further optimized. We conclude that CID and ETD complement each other to advance our understanding of endogenous peptides, which has received limited attention in the proteomics community.