Oral cancer, especially oral squamous cell carcinoma (OSCC), is a high-impact disease in the oral cavity. OSCC accounts for ~90% of malignant oral lesions and is widely recognized as the most frequently occurring malignant tumour of oral structures. Each year, approximately 500000 new cases are diagnosed worldwide, with a 5-year survival rate of only 50% . In the early stages of OSCC, the tumour responds well to combination therapy, as evidenced by a 5-year survival of 80% in these patients. However, the response to treatment is much lower in advanced OSCC . Thus, there is a need to investigate the molecular mechanisms involved, to identify potential therapeutic targets as well as to discover biomarkers for the early detection of OSCC and subsequent monitoring of its progression.
Saliva has gained notable attention as a diagnostic fluid because it is easy to collect and process, minimally invasive and associated with low costs . It contains a large array of proteins, which may be useful for novel approaches to prognosis, clinical diagnosis and monitoring and management of disease. Comprehensive analysis of the human saliva proteome may contribute to the understanding of pathophysiologies and provide a foundation for the recognition of potential biomarkers of human disease [4–6].
MALDI-TOF-MS is a powerful technique that can be used to analyse proteins from saliva . It can detect low-molecular-mass peptides with adequate resolution and sensitivity, making it a useful tool for peptide pattern profiling. In addition, beads with peptide libraries, mesoporous silica particles  or a magnetic core , such as weak cation-exchanger magnetic beads, may be utilized for selective enrichment of low-molecular-mass peptides before MS analysis. Magnetic beads constructed on nanomaterial are a promising material among the various types of separation beads. This kit based on weak cation exchange (WCX) principle. Proteins and peptides in samples are captured by specific adsorption of magnetic beads in low salt and low pH solution and released in high salt solution, so as to capture proteins and peptides in the serum. The proteins and peptides can be analysed by MALDI-TOF-MS. Using a combination of magnetic beads and MALDI-TOF MS enables efficient and sensitive detection of peptides that are specific for certain conditions; indeed, we used this technique to successfully identify serum peptide profiles in pilot studies [10,11]. This procedure, which involves weak cation-exchanging magnetic beads for sample separation, MALDI-TOF MS for peptide profile detection and a database for construction of condition-specific peptidome models, is a powerful tool that enables early detection, diagnosis and determination of the prognosis of various diseases .
In the present study, we investigated differences in the salivary peptide (1–10 kDa) profiles of T1 stage OSCC patients and healthy subjects by MALDI-TOF MS using a magnetic bead-based peptidome analysis of saliva samples. We aimed to identify a panel of specific biomarkers for differential expression.
The general clinical characteristics of the subjects are presented in Table 1. The entire mass spectra of the extracted peptide samples from 63 subjects in the two groups were generated using the same instrument settings in the range of 1000–10000 Da (Supplementary Figure S1). Most peaks were detected in the range of 1000–3500 Da.
An average of 50 peptide mass peaks was found when the two groups were compared. Next, the peaks among the mass spectra were quantified and compared. Eight of these peptide mass peaks (1285.6, 1731, 1191.4, 1353.9, 1584.6, 1553.5, 1329.9 and 1432.2 Da) were significantly different between OSCC patients and healthy controls (Figures 1, 2 and Table 2). Among them, four peptides (at 1285.6, 1553.5, 1329.9 and 1432.2 Da) were up-regulated and four (at 1731, 1191.4, 1353.9 and 1584.6) were down-regulated in the OSCC patients (Figure 2).
3D m/z ratio-intensity maps showed the two significantly different peptides at1285.6, 1432.2 Da, which had a particular trend among the two groups. Green curve, heathy control group; red curve, 7-month group; blue curve, OSCC patients group.
The peak intensities of the two different groups showing an increasing trend in peak intensity at 1285.6, 1432.2, 1353.5 and 1329.9 Da and a decreasing trend at peak 1731, 1191.4, 1353.9, 1584.6 Da. (*P<0.05; **P<0.01).
All the eight mass peaks were used to establish the diagnostic model using the radial basis function method. Two peaks (1285.6 and 1432.2 Da) exhibited the most significant difference (P<0.05, by t test) between the two groups compared with the other combinations of peptides. Thus, we used these two peptides to establish a fitted curve. 2D-cluster plot analysis demonstrated represents the best separating peaks in 2D spaces (Figure 3), whereas 3D view of principal component analysis (PCA) scores plot analysis indicated a well differential distribution of mass peaks between controls and OSCC patients (Figure 4). Columns represent samples; rows are m/z peaks as indicated by the average molecular mass. The shape of the two figures showed the well-separated locations of the samples from the two groups, indicating that the fitting results were satisfactory.
The scatter plots showed a well-fitting curve of two peaks with a significant difference (P<0.01, by t test) in their distribution between healthy controls and OSCC patients.
3D view displays of the principal component analysis of peptide profiles using BE software. Blue spots represent control individuals; red spots represent OSCC patients.
Detection of oral cancer at an early stage is important for successful clinical therapy . Patients with OSCC often present with advanced-stage disease, which is associated with poorer prognosis. Late-stage OSCC also requires more aggressive therapy, which results in increased functional disability. Conventional diagnostic techniques, including direct inspection and imaging technology such as positron emission tomography-computed tomography, are limited in their ability to detect early stage OSCC and are ineffective for screening high-risk populations . Screening tools are needed that combine high sensitivity and specificity and are sufficiently non-invasive and inexpensive to enable widespread use.
In recent years, interest in saliva for clinical purposes as an alternative to other body fluids, such as blood and urine, has increased. WS is a complex biological fluid due to the many processes involved in its production. In addition to the exocrine components, there are several non-exocrine contributors such as desquamated epithelial cells, intact and partial blood cells, gingival fluid and possibly fluid entering the oral cavity through mucosal seepage. This renders diagnosis of disease by the analysis of saliva both challenging and attractive. Saliva was found to be similar in microbial profile to the soft tissues . This was a significant finding from the study of the OSCC-free population . So the screening test of salivary peptides for OSCC is appealing. MS-based proteomics is a high-throughput method used to analyse salivary proteomics and has been employed in the study of protein/peptide spectra, biological marker spectra, as well as single biological markers for complicated diseases such as cardiovascular and cerebrovascular diseases, OSCC and neuro-degenerative diseases.
To date, more than 2000 peptides have been discovered in the salivary peptidome [16–18]. By mapping the corresponding protein entries, it has been possible to assign those peptides to 695 non-redundant protein species . Since the 1970s, salivary peptides have been grouped into six structurally-related major classes , namely, histatins, basic proline-rich proteins (bPRPs), acidic proline-rich proteins (aPRPs), glycosylated proline-rich proteins (gPRPs), statherin and cystatins [20–22]. Salivary PRPs, as well as bPRPs, aPRPs and gPRPs, are usually identified from the small peptide fraction (< 3 kDa). Some PRPs, along with statherin and histatin-1, appear to actively participate in tooth mineralization. Histatins, especially histatin-3 and histatin-5, which are found in high amounts in saliva, are strongly anti-fungal . The cystatin class comprises five major isoforms (S, C, D, SA and SN), which have strong bactericidal and virucidal properties . Statherin, a multifunctional molecule that possesses a high affinity for calcium phosphate minerals, such as hydroxyapatite, contributes to the maintenance of the appropriate mineral solution. Defensins are a family of low-molecular-mass (3–4 kDa) cationic proteins with antibiotic, anti-fungal and anti-viral properties. They are involved not only in innate immunity against infections but also in adaptive immunity, inflammation and wound repair .
In the present study, WCX magnetic beads and the MALDI-MS technique were employed to investigate WS samples from OSCC patients and healthy controls. The components extracted by the WCX magnetic method could be either low-molecular-mass peptides or fragments resulting from proteolytic activity occurring in the WS after secretion into the oral cavity. WCX magnetic beads separate the proteins and/or peptides of different isoelectric points from complex biological fluids with specific anionic ligands. The techniques of MALDI-TOF-MS combined with WCX magnetic beads incorporate both of their advantages ; the low cost, the simple purification, could capture more proteomes than other methods especially in the low-molecular-mass range ; sensitive, fast and essential for clinical use  allowed the identification of comprehensive ‘fingerprints’ of protein profiles within biological fluids and were used to identify biomarkers of various diseases. The effectiveness of this combination of techniques has been confirmed in many saliva-based peptide profile identification studies [29,30]. We examined 40 T1-stage OSCC saliva samples and 23 healthy control samples. Eight m/z peaks were found to be significantly different between the groups. Four of these were up-regulated and four were down-regulated. The mass peaks of 1285.6 and 1432.2 were detectable in all OSCC samples at a high intensity, but seldom in the healthy subjects, suggesting that these represent markers of OSCC and may play a role in the occurrence and development of this disorder. The 1731 and 1353.9 mass peaks were detected in the majority of healthy subjects, but seldom in OSCC patients so it may be others biomarkers to detect OSCC.
Our results also differ from those of Jou et al. . They found three significantly different peaks in OSCC patients and healthy controls. Two of these (m/z=2919, 4373) were up-regulated and one (m/z=5592) was down-regulated. The differences between the two studies may be attributable to the different age and sex of the participants involved and/or to the different methods used to extract low-molecular-mass peptides. The 95.7% of participants enrolled in Jou et al.'s study were men (45 male, two female) with a mean± S.D. age of 50.79±10.20 years for the OSCC patients and 76.7% were men (23 male, seven female) with 44.9±10.1 years for the healthy control subjects, whereas our study used 42.5% men (17 male, 23 female) with a mean ± S.D.
Age of 56.25±14.23 years in OSCC patients and 56.5% men (13 male, 10 female) with 54.74±11.83 years in healthy controls. It is possible that the WS proteome changes with age . The small peptides (<10 kDa) used for MS analysis in our study were extracted from WS samples by WCX magnetic beads, whereas the saliva samples used in Jou et al.'s study were precipitated by using C8-magnetic beads. Nevertheless, the different processing methods could lead to artificial losses and modification of the samples, which could influence the results significantly .
The peptide sequence identifications made in the present study have led to interesting speculations. The mass peaks of 1285.6 and 1432.2 were both identified as histatin-3 by matching these peaks to the mass spectrum database of Bioyong Technologies Inc. histatin-3 belong to the histatin family which are a class of peptides named according to their high histidine content  that were identified in human saliva approximately 30 years ago [32,33]. Histatin family consist of 12 members found in the saliva secreted by the salivary glands of humans and higher primates, are localized in human oral tissues . Histatin-3, which is 32 residues in length, is encoded for by the histatin-3 precursor (HIS2) gene . Histatin-3 could kill Candida albicans, the most common and the most pathogenic oral Candida species [36,37]. However, histatin-3 are also active against other yeasts and fungi, including Candida glabrata, Candida krusei, Saccharomyces cerevisiae and Cryptococcus neoformans [38,39] and some bacterial species, including Streptococcus mutans, Porphyromonas gingivalis and Actinobacillus actinomycetemcomitans [40,41].
It has been reported that microorganisms, especially Candida species, are closely associated with OSCC [42–46]. Patients with OSCC tend to possess significantly raised concentrations of certain bacteria in their saliva [15,47,48]. Previous studies by various investigators have demonstrated a significant correlation between oral candidiasis and oral squamous carcinoma in a number of studies [49,50]. Rehani et al.  identified Candida as a possible factor in the development of OSCC. Marttila et al.  found that Candida colonization frequency and density were higher at oral mucosa of OSCC patients than in healthy controls'. Oral microorganisms inevitably up-regulate cytokines and other inflammatory mediators that affect the complex metabolic pathways and may thus be involved in carcinogenesis . It has been suggested that Candida species play a role in oral carcinogenesis by triggering nitrosamine compounds to activate specified oncogenes, thereby initiating oral neoplasia [51,53,54]. Previous studies were also demonstrated that the secreted anti-microbial proteins responsible for combating oral candidiasis include the salivary histatins [55,56]. So it suggests that the high level of histatin-3 in OSCC patients' saliva our findings were modulated by the raised concentrations of oral candidiasis.
In addition, histatin-3 are also involved in cell proliferation through the regulation of heat shock cognate protein 70 (HSC70) and cyclin-dependent kinase inhibitor 1B (p27Kip1) in oral cells  and could also bound to HSC70 inhibits HSC70-mediated activation of toll-like receptor (TLR) 4 signalling activation . TLRs are a family of transmembrane proteins that recognize a variety of endogenous and microbial agents. The TLR 4 could lead to more aggressive, invasive behaviour of OSCCs . It indicated that the histatin-3 may be involved in the progression of OSCC by interacting with TLR 4.
There is agreement that anti-microbial treatment is important pre-, during and post-therapy for oral cancer patients [46,60–62]. Histatin-3 possesses potent anti-fungal and anti-microbial properties and has the advantage over conventional synthetic azole or polyene anti-fungals and anti-microbial of being a naturally occurring compound in man, with no known cross-reactivity with human cells or tissues . These qualities make it an ideal compound for development as an anti-fungal agent in the treatment of fungal infections of the oral cavity . An important consideration in the development potential of histatin as a therapeutic agent would be the determination of the in vivo mechanism, occurrence- and significance of resistance to this peptide.
In conclusion, our results suggested mass peaks of 1285.6 and 1432.2 Da which were both identified as histatin-3 in saliva as correlated with OSCC progression. However, the discovered candidate biomarkers need to be extensively validated with wider cases. Clearly, it is challenging to translate candidate biomarkers from proteomic investigations into real-world diagnostic or prognostic applications. Approval of use of histatin-3 as a biomarker to detect early stage of relies on the results of large-scale multicentre clinical trials. We plan to undertake such a study in the future.
Feng Chen conceived the study and revised the manuscript. Wei-Peng Jiang drafted the manuscript. Li-Xin Xu performed statistical analysis. Xin Peng revised the manuscript. Zhen Wang performed validation experiments. All authors read and approved the final manuscript.
This work was supported by the National Natural Science Foundation of China [grant numbers 81200762 and 81472527]; the National Supporting Program for Science and Technology [grant number 2014BAI04B06]; and the Science Foundation of the Third Dental Center, Peking University School and Hospital of Stomatology [grant number 011401].
Abbreviations: aPRP, acidic proline-rich protein; bPRP, basic proline-rich protein; gPRP, glycosylated proline-rich protein; OSCC, oral squamous cell carcinoma; TLR, toll-like receptor; WCX, weak cation exchange; WS, whole saliva
This is an open access article published by Portland Press Limited and distributed under the Creative Commons Attribution License 3.0.
The CSF peptidome constitutes only a minor fraction of the total protein contents of CSF. Several known bioactive peptides are present at concentrations in the pg/ml range, requiring the analysis of a few hundred microlitres of CSF to detect them by mass spectrometry. Because of the limited loading capacity of nano-LC columns (<1 µg), it is necessary to enrich the peptide fraction. Ultrafiltration using molecular weight cut-off (MWCO) filters provides a simple means to achieve this. Filters of different cut-off sizes (10 kDa, 30 kDa, 50 kDa) were evaluated. 50 kDa filters resulted in permeability of significant amounts of albumin (data not shown). While 10 kDa filters efficiently removed albumin, they retained a large part of the peptides in the mass range of interest (data not shown). According to the product documentation for the filters the recovery of a peptide of 1,350 Da is only approximately 76%. 30 kDa filters were found optimal for our application, achieving efficient removal of albumin, the most abundant CSF protein, without compromising peptide recovery.
From studies in plasma and CSF it is well known that many peptides bind to larger proteins and may therefore be retained in the ultrafiltration step , –. To improve the recovery of such peptides we investigated the effect of pretreating CSF samples with different concentrations of acetonitrile (ACN) and formic acid (FA), in order to dissociate the peptides from the carrier proteins prior to the ultrafiltration step. A CSF pool was divided into several 500 µl aliquots to which ACN or FA were added at different concentrations as described in the method section. The filtrates were analyzed by LC-MALDI MS and evaluated on the basis of number of detected compounds. Without any sample pretreatment, 2,445 compounds were detected (Figure 1 a). In a CSF sample treated with 20% ACN, the number of detected compounds increased to 3,543 (Figure 1 b). When the concentration of ACN was increased to 40% the amount of peptides in the flow through dropped significantly. A possible explanation for the observed decrease is that at this high ACN concentration, a large part of the sample proteins precipitate, resulting in co-precipitation of peptides. The highest numbers of compounds were detected when the CSF was incubated with 20% ACN. Addition of FA up to 5% to the samples also increased the number of detected compounds, however less than ACN. Increasing the FA concentration to 10% did not yield any additional improvement. Combining ACN and FA yielded poor results.
Figure 1. LC-MS profiles of CSF peptides.
The LC retention time is shown on the y-axis and the m/z on the x-axis. (a) Without sample pretreatment, 2,445 compounds were detected. (b) When adding 20% ACN prior to the ultrafiltration step, the number increased to 3,543.
Identification of endogenous peptides by sequence database searching is more challenging than identification of proteins in tryptic digests. Because the peptides are not produced by cleavage by one specific protease, restrictions regarding enzyme cleavage specificity cannot be applied in the database search, increasing the number of peptide sequences to consider in the search by a factor of 100–1,000, and thereby decreasing the significance of the matches in the database searches. Performing a database search with Mascot and filtering for a false discovery rate (FDR) of 1% (corresponding to a Mascot MOWSE score of 41 in the search) resulted in 596 peptide identifications. However, upon inspection of peptide hits with lower scores it was found that several of these could be confirmed based on the characteristic fragment ions arising from the presence of specific amino acids and the location of charged residues in the amino acid sequence. For singly charged peptide ions, strong charge-remote fragmentation C-terminally to aspartic acid and, to somewhat lesser degree, glutamic acid for peptides containing arginine, produces characteristic b- or y- fragment ions, depending on the location of the arginine , . Prominent cleavage N-terminally to proline residues also produces strong b- or y- ions. Peptides that display these fragmentation characteristics often score poorly in database searches due to their selective fragmentation. However, these diagnostic fragment ions provide means to manually evaluate peptide specific matches retrieved by database searches.
For peptide identification, manual validation was performed on all database search results retrieved using a relatively low score threshold: ion score >15 and at least one peptide per protein with ion score >27. With these settings, a total of 4,185 peptide matches were retrieved. Manual validation was based primarily on the fragmentation rules discussed above for peptides containing aspartic acid/glutamic acid and arginine, and peptides containing proline. Upon inspection of the raw spectra, errors in the automatic peak detection were corrected, sometimes increasing the number of assigned peaks. In some cases validation was supported by similarity of the fragment ion pattern with that of an already identified peptide, covering partly the same amino acid sequence.
Figure 2 shows two examples illustrating the application of the described validation criteria for a strong and a weak peptide match. The fragment ion spectrum of m/z 3511.7738 (Figure 2 a) matched the peptide SVNPYLQGQRLDNVVAKKSVPHFSDEDKDPE from Neuroendrocrine protein 7B2 with an ion score of 194.9. Charge-remote fragmentation C-terminally to Asp-12, Asp-25, Asp-27, and Asp-29 due to the presence of arginine at the N-terminal part of the peptide (Arg-10) results in abundant corresponding b-ions, confirming the assignement of the peptide. The lowest scoring peptide match which was validated was the peptide KANDESNEHSDVIDSQELSKVSREFH (m/z 3000.3891) derived from Osteopontin, which received an ion score of 15.8. Prominent charge-remote fragmentation is observed C-terminally to Asp-14, Asp-11, Glu-8 (y-18), Glu-17 (y-9), in this case giving rise to the corresponding strong y-ions, y12, y15, y18, and y9, respectively, due to the presence of Arg-23 near the C-terminus of the peptide. Despite that the spectrum in the latter example only contains a few fragment ion signals the predictability of the fragment ion peak pattern based on the matched peptide sequence provides evidence for the correctness of the identification.
Figure 2. Peptide identification.
(a) MS/MS spectrum of an ion of m/z 3511.7738 matching the peptide SVNPYLQGQRLDNVVAKKSVPHFSDEDKDPE from Neuroendrocrine protein 7B2 with an ion score of 192.4. Charge-remote fragmentation C-terminally to Asp-12, Asp-25, Asp-27, and Asp-29 due to the prescence of arginine at the N-terminal part of the peptide (Arg-10), results strong corresponding b-ions. (b) MS/MS spectrum of an ion of m/z 3000.3891 matching the peptide KANDESNEHSDVIDSQELSKVSREFH with an ion score of 15.8. Charge-remote fragmentation is observed C-terminally to Asp-14, Asp-11, Glu-8 (y-18), Glu-17 (y-9), in this case giving rise to the corresponding strong y-ions, y12, y15, y18, and y9, respectively, due to the presence of Arg-23 near the C-terminus of the peptide.
A histogram of the number of manually validated and rejected peptide matches as a function of ion score (Figure 3) shows that quite a large number of peptide hits with low ion score could be validated. A total of 730 peptide matches were validated. The 22% increase of assigned peptides by manual evaluation compared to the 596 peptides identified with an FDR of 1% in the Mascot database search is significant and suggests that, particularly for the analysis of singly-charged ions, search algorithms may be improved by implementing amino acid- and charge-based fragmentation schemes in their scoring algorithms.
Figure 3. Stacked histogram of all peptide matches as a function of ion score.
Peptide matches were retrieved that fulfilled the criteria: ion score >15 and at least one peptide match with ion score >27 per protein. The coloured segments indicate the number of validated (green) and rejected (red) peptide matches.
The identified peptides are listed in Table S1. The identifications comprised 626 unique peptide sequences (not taking into account post-translational modifications) originating from 104 proteins (Table 1). The mass spectrometric data is available via the PRoteomics IDEntifications database  (PRIDE, http://www.ebi.ac.uk/pride/, Accession number 24353).
Comparisons with other studies
Comparing the peptides identified in our study to those identified in the study by Zougman et al., one of the most comprehensive CSF peptidomic studies to date, showed that only 23% of the peptides identified in our study were present in the other data set (Figure 4 a). The large difference between the data sets may be attributed to the different ionization techniques used in the two studies (MALDI vs. ESI), to differences in the method used for preparation of the peptide extracts, and that different CSF samples were analyzed.
Figure 4. Comparison of the peptides identified in the current study to other published peptidomic and proteomic data.
Only 23% of the peptides identified in the current study were part of the peptide set reported by Zougman et al.. Comparison of the proteins represented by the endogenous peptides identified in the current study and in the study by Zougman et al. with the proteins identified in the CSF proteomic analysis by Schutzer et al. (b) reveals that both peptide sets have a high degree overlap with the proteomic set.
To assess how the CSF peptidome compares to the CSF proteome, we compared the proteins represented by the endogenous peptides identified in the current study and in the study by Zougman et al. with the extensive CSF proteomic analysis by Schutzer et al. comprising 2,462 proteins  (Figure 4 b). Both peptidomic data sets have high overlap with the proteomic set. Thus, on a protein level, the peptidomic approaches used in the two studies mainly identifies peptides derived from the same proteins that are found by global proteomic analyses.
Biological functions of the identified proteins
While the significance of the endogenous peptides in CSF is still largely unknown, several of the precursor proteins from which they derive are associated with various brain disorders and known cellular processes in the brain. A few examples are given below. While the presence in CSF of all of these proteins has been reported previously , several of the endogenous peptides derived from these proteins are reported for the first time.
Amyloid beta peptides.
The amyloid beta A4 protein (amyloid precursor protein, APP) plays a central role in the pathophysiological processes in AD, which is characterized by progrediating neuronal degeneration with amyloid deposits . By enzymatic processing of APP, the peptide amyloid beta (Aβ) 1–42 is generated which is highly prone to aggregation and is the major constituent of the amyloid plaques that form in the brain. Aβ1–42 is also used as a CSF biomarker for AD . In this study, we detected Aβ1–14, 1–15, 1–16, 1–17, 1–19, and 5–15 (Figure 5 a). The existence of several of these fragments have been recently reported in CSF in a study using immunoprecipitation in combination with mass spectrometry . The C-termini of these fragments span the α-secretase cleavage site and may thus be markers of non amyloidogenic APP processing that may be protective from AD . The identification of Aβ5–15 is the first report of this truncated form in CSF. The identification of Aβ peptides starting at position 5 is especially interesting since such peptides may represent the activity of an APP processing pathway that is up-regulated after inhibition of the major Aβ producing enzyme BACE1 ,  and CSF Aβ5-X peptides may be useful as pharmacodynamic markers in trials of BACE1-inhibitors .
Figure 5. Identified endogenous peptides from (a) Amyloid beta A4 protein isoform h precursor (A4_HUMAN, commonly referred to as APP) and (b) from Amyloid-like protein 1 isoform 1 (APLP1).
The peptides identified from A4_HUMAN are all located within the β-amyloid peptide, starting at or in proximity of the N-terminus of the peptide, defined by the BACE cleavage site. Correspondingly, four of the peptides identified in APLP1 are generated by BACE cleavage C-terminally to Arg-567.
CSF also contains several amyloid-like proteins, which have sequence homology to APP, and that to some extent undergo similar processing. One such protein is the amyloid precursor-like protein 1 (APLP1). APLP1 undergoes processing by BACE and γ-secretase generating Aβ-like peptides of which one has been shown to have potential value as a biomarker for AD . Several of these Aβ-like peptides were also identified in our study, four of which have N-termini at the BACE cleavage site (Arg-167) in APLP1 (Figure 5 b). The peptides derived from APLP1 and APP may thus be used in parallel in clinical studies to monitor the activity of these secretases to investigate if similar patterns of peptides are affected in different disease or to monitor effects of pharmaceutical compounds that target a given enzyme.
Another interesting finding is the identification of fragment 28–36 of cocaine- and amphetamine-regulated transcript (CART), covering part of the amino acid sequence of CART(1–39). This peptide is involved in regulating many processes including body weight, reward and endocrine functions . CART is together with NPY believed to regulate the leptin-mediated feeding response .
Neuropeptide Y is a highly abundant neuropeptide and a potent neuromodulator involved in several different processes, e.g. hunger, stress response, cardiovascular function, and circadian rhythms . Three peptides were identified that span part of the NPY 1–36 sequence: NPY 1–20, 3–22, and 5–22, none of which have been previously reported in CSF. It is believed that the intact NPY 1–36 is the active peptide although it has not been shown in intact form in CSF.
From myelin oligodendrocyte glycoprotein (MOG) we identified six peptides spanning part of the extracellular region with a common cleavage site C-terminally to the amino acid at position 81. MOG is found on the surface of the myelin sheath and has an yet unclear role in multiple sclerosis where autoantibodies against the protein are found . The levels of autoantibodies against MOG seem to be an indicator of disease intensity.
Peptides from the granin family have been discussed as CSF biomarkers for various diseases such as AD, multiple sclerosis, schizophrenia, and depression . The granins (Chromogranin-A (CMGA), Secretogranin-1 (SCG1), Secretogranin-2 (SCG2), Secretogranin-3, 7B2, NESP22, proSAAS, and VGF) are involved in regulated delivery of several key factors in CNS such as neurotransmitters, hormones and growth factors. Peptides from Neurosecretory protein VGF have been associated with different brain disorders, such as VGF26–62 with decreased levels in frontotemporal dementia; VGF378–397, , with decreased levels in AD; VGF23–62, with increased levels in schizophrenia and in depression; and SCG2529–566, with decreased levels in depression. In our study three of the peptides (VGF26–62, VGF23–62, and SCG2529–566) were found both as intact peptides and in truncated forms, while the third VGF peptide was only found as peptides exceeding the sequence length with 1–6 amino acids. In the case of SCG2 and CMGA, an alteration in three peptides is described as potential marker for multiple sclerosis, decreased SCG1441–493 and SCG1306–365 and increased CMGA194–213, . SCG1441–493 has also been found decreased in frontotemporal dementia . Although we did not identify any of SCG1441–493, SCG1306–365, or CMGA194–213 as intact peptides, we identified several forms spanning parts of these sites (Figure 6).
Nine fragments of the prion protein were identified. The detected peptides are fragments created after the signal peptide has been removed from the protein with some starting at amino acid position 23 and others starting more C-terminally in the octapeptide repeat region which is associated to prion disease . In CJD, which is characterized by spongiform degeneration and neuronal death leading to dementia, motor dysfunction, and eventually death, the prion protein plays a critical role. When the prion protein assumes an abnormal conformation it becomes very prone to aggregation, which starts an autocatalytic cascade that eventually produces neurotoxic species of the protein , .
Our results show that LC-MALDI MS can be used for monitoring large numbers of endogenous CSF peptides in sample volumes relevant to clinical studies. Several of the identified peptides derive from proteins involved in physiological and pathological processes in the CNS. The CSF peptidome contains information about peptides spanning other parts of the proteins than are found using bottom-up proteomic workflows, and may thus be a complementary strategy for identifying biomarkers of disease. Supplementing the method described here with stable isotope labeling methods for quantification, such as the tandem mass tag (TMT) or the isobaric tags for relative and absolute quantitation (iTRAQ) approaches may be a viable strategy for identification of new biomarkers in CSN disorders.