Peptidomic analysis of endogenous plasma peptides from patients with pancreatic neuroendocrine tumours

Rationale Diagnosis of pancreatic neuroendocrine tumours requires the study of patient plasma with multiple immunoassays, using multiple aliquots of plasma. The application of mass spectrometry based techniques could reduce the cost and amount of plasma required for diagnosis. Methods Plasma samples from two patients with pancreatic neuroendocrine tumours were extracted using an established acetonitrile‐based plasma peptide enrichment strategy. The circulating peptidome was characterised using nano and high flow rate liquid chromatography/mass spectrometry (LC/MS) analyses. To assess the diagnostic potential of the analytical approach, a large sample batch (68 plasmas) from control subjects, and aliquots from subjects harbouring two different types of pancreatic neuroendocrine tumour (insulinoma and glucagonoma), were analysed using a 10‐min LC/MS peptide screen. Results The untargeted plasma peptidomics approach identified peptides derived from the glucagon prohormone, chromogranin A, chromogranin B and other peptide hormones and proteins related to control of peptide secretion. The glucagon prohormone derived peptides that were detected were compared against putative peptides that were identified using multiple antibody pairs against glucagon peptides. Comparison of the plasma samples for relative levels of selected peptides showed clear separation between the glucagonoma and the insulinoma and control samples. Conclusions The combination of the organic solvent extraction methodology with high flow rate analysis could potentially be used to aid diagnosis and monitor treatment of patients with functioning pancreatic neuroendocrine tumours. However, significant validation will be required before this approach can be clinically applied.

the high-abundance proteins from samples prior to analysis. Without removing these highly abundant components, the analyses will be hampered by the liquid chromatography/mass spectrometry (LC/MS) instrument detecting signature peptides from these proteins, at the expense of missing the low-abundance components. Historically, plasma protein extraction methods have involved selective protein removal via immunoaffinity of albumin, immunoglobulins and the top 12 to 14 most abundant proteins. 3,4 Although effective, the nature of these extraction devices means that throughput is low, and sample volume is limited to tens of microliters per extraction, 5 which is not conducive to the analysis of peptides in the low pg/mL range.
Therefore, these methods are inappropriate for peptidomics. Other sample handling procedures (for both human and rodent plasma) have used size-exclusion-based extraction techniques, such as ultrafiltration spin devices. 6,7 Some plasma peptidomic studies have employed solidphase extraction (SPE), whereby plasma is extracted by SPE followed by high-performance liquid chromatography (HPLC) fractionation and subsequent enzymatic digestion prior to LC/MS analysis. 8 Although these previous methodologies have shown effectiveness at extracting peptides from plasma, they are not considered high throughput, and in the case of immunoaffinity extraction, have to be performed consecutively. A high-throughput approach for enriching the low molecular weight (LMW) plasma peptidome/proteome is the use of protein precipitationeither through solvent-or acid-based precipitation. 9,10 The organic solvent approach has demonstrated the capability of detecting peptides in plasma in the low pg/mL range on its own 11 or with a subsequent SPE phase using selected reaction monitoring (SRM)-based analyses. 12,13 These approaches have shown that plasma pre-treatment combined with highly targeted triple quadrupole analysis is capable of detecting specific peptides at low pg/mL concentrations, although very few untargeted peptidomics analyses have been published. Studies by Albrechtsen and colleagues have demonstrated the ability to detect gut-derived peptide hormones in the circulation using untargeted approaches, but these protocols involved large sample volumes and extensive method workflows. 8,14 Whilst this method proved to be effective, it did not detect the peptides intact (protease digestion was performed), therefore complicating data interpretation. A study by Parker et al. used 200 μL of plasma from multiple time-points that were combined using a tandem mass tag protocol. 10 Intact peptides from the proglucagon gene were detected in this study but only the glicentin-related polypeptide (GRPP) was identified after significant sample workup. In some clinical settings it is not feasible or practical to obtain large plasma sample volumes, and analysis methodologies would be required to do more with less starting material.
Pancreatic neuroendocrine tumours (PNETs) are rare neoplasms that account for 1-2% of all pancreatic tumours. 15 PNETs may be sporadic or part of an inherited tumour syndrome such as multiple endocrine neoplasia type 1 or von Hippel Lindau syndrome. Furthermore, PNETs can be subdivided into non-functioning or functioning tumours, the latter referring to tumours capable of secreting peptide hormones or biogenic amines into the circulation usually in a dysregulated manner to produce defined tumour syndromes with varying clinical manifestations. For example, pancreatic islet α-cell tumours that overexpress proglucagon are typically associated with the glucagonoma syndrome, a rare disease characterised by necrolytic migratory erythema, impaired glucose tolerance, thromboembolic complications and psychiatric disturbance. 15 Currently, diagnosis of a PNET is made through a combination of clinical examination and history as well as radiological, histological and biochemical investigations. The latter are based, primarily, on immunoassays directed towards a limited number of peptide hormone biomarkers. However, these are of limited diagnostic and prognostic value and there is an unmet need for new and innovative methodologies that can measure and identify novel circulating biomarkers that may be of greater clinical utility. 16 This study describes the application of a well-characterised and validated organic solvent based peptide extraction method to the analysis of plasma samples derived from patients with functioning PNETs. Plasma from one of these patients was previously studied extensively using immunoassay-based techniques to characterise circulating peptidic components from the pancreas, revealing high concentrations of proglucagon-derived peptides (Case study 1). 17 Plasma from a second patient diagnosed with a metastatic pancreatic glucagonoma, that had not been previously characterised by immunoassay, was also analysed (Case study 2), along with plasma from a subject harbouring a benign insulinoma and plasma samples taken from a cohort of healthy control subjects. Samples were analysed by low-and high-throughput LC/MS methodologies to assess the feasibility of developing these techniques for use in a clinical setting.

| Case study 1
This patient has been described previously. 17 Briefly, a 57-year-old female presented with a 12-month history of necrolytic migratory erythema, constipation, early satiety, nausea and vomiting and weight loss. She was also found to have profound hyperinsulinaemic hypoglycaemia. Radiological, histopathological and initial biochemical investigations confirmed a diagnosis of a well-differentiated glucagonsecreting pancreatic neuroendocrine tumour (grade 1) with widespread hepatic metastases. Subsequent gel filtration chromatography and analysis with immunoassays specific for proglucagon-derived peptides revealed elevated plasma concentrations of several peptides including proglucagon, glucagon, GLP-1 and GLP-2. A trial of octreotide (100 μg thrice daily), a short acting somatostatin analogue, successfully reduced plasma levels of proglucagon-derived peptides resulting in marked clinical improvement including abolition of the hypoglycaemia.

| Case study 2
A 54-year-old male was diagnosed with a metastatic pancreatic neuroendocrine tumour following investigations for weight loss and abdominal pain. Biochemical assessment of commonly measured fasting gut hormones and neuroendocrine tumour markers by immunoassay revealed isolated elevation of glucagon (179 pmol/L (normal range (NR): 0-50)), chromogranin A (223 pmol/L (NR: 0-60)), and chromogranin B (909 pmol/L (NR: 0-150)). He was commenced on a somatostatin analogue (Lanreotide 120 mg every 28 days) and had radiologically stable disease for the subsequent ten years. In 2017, surveillance cross-sectional imaging revealed disease progression (Figures S1A and S1B, supporting information). In addition, the patient reported symptoms suggestive of hypoglycaemia and hyperinsulism that were confirmed following a 12-h fast (plasma glucose 2.8 mmol/L; plasma insulin 30 pmol/L). Everolimus therapy was commenced which normalised glucose levels. The patient remains under close clinical, radiological and biochemical surveillance.

| Control subjects (including ethics and informed consent)
Control plasma samples were collected from 28 healthy adult volunteers as part of a larger study examining endocrine physiology after an oral glucose tolerance or liquid meal test (REC 13/EE/0195).
A mixture of samples was available from these individuals, where some were in the fasting state and some from 240 min after a 75 g oral glucose tolerance test, and some samples were obtained from individuals from repeat visits. Blood was collected in EDTA plasma tubes, centrifuged for 10 min at 4°C/3500 g and plasma aliquoted and snap-frozen on dry ice within 30 min of phlebotomy. A total of 62 plasma samples were selected for analysis from the 28 individuals.

| Plasma peptide extraction for nano-LC/MS analysis (Case study 1)
Plasma from Case study 1 17  FA in water and then 5% methanol in water with 1% acetic acid. The peptides were eluted from the cartridge using 2 × 30 μL of 60% methanol in water with 10% acetic acid. The eluant was evaporated and then reconstituted into 50 μL of 0.1% FA in water and a volume of 20 μL was injected onto a nano-LC/MS system.

| Plasma peptide extraction for high-throughput LC/MS analysis
An almost identical extraction method was performed on the sample from Case 2 (same plasma as described in case study 2); however, a larger volume of plasma was available for testing. Plasma (EDTA) was separated, aliquoted and rapidly frozen on dry ice for storage before extraction. Plasma was thawed on ice and aliquots of 250 μL were precipitated with 1 mL of 80% ACN in water containing 1 ng/mL of bovine insulin as an internal standard (Sigma Aldrich, Poole, UK). A lower ratio of plasma to ACN was used compared to the 50 μL plasma volume extractions so that the mixing is possible in a 2-mL 96-well plate. An additional 62 plasma samples from control individuals were extracted in parallel for comparative analysis of identified peptides.
Control samples were extracted into a single aliquot whilst the glucagonoma sample was extracted in duplicate as there was sufficient sample available. The sample used for the nano-LC/MS analysis (Case 1), and an additional aliquot from Case 1 after 4 days of octreotide therapy, were extracted in a single aliquot due to sample volume limitations. Plasma from the subject with insulinoma (Case study 3) was extracted in duplicate as a further comparator. All samples were spun at 3900 g for 10 min and the supernatants transferred and evaporated to dryness before the same SPE process was performed as stated above. However, the SPE eluant was not evaporated, but rather diluted by the addition of 75 μL of 0.1% FA in water to reduce the organic content percentage prior to direct injection onto the LC/MS system. A total of 50 μL was injected onto the LC/MS system in a high flow configuration.

| Nano-LC/MS analysis of plasma extracts
Peptide extracts were analysed using a Thermo Fisher Ultimate 3000

| High flow rate full scan LC/MS analysis
Extracts (50 μL) from the control cohort were injected onto a HSS T3 column (2.1 × 50 mm; Waters) at 15% A (0.1% FA in water, v/v) and 85% B (0.1% FA in ACN) at a flow rate of 300 μL/min and separated using a 6.5-min gradient to 40% B. The column was washed for 1.5 min at 90% B before returning to initial conditions at 8 min for a total run time of 10 min. IDA-based analysis involved a full scan of m/z 600 to 1600 with a resolution of 75,000, AGC of 3e6 and max fill time of 200 ms.

| Endogenous peptide identification
The nano-LC/MS files obtained from the two different extracts of value of 1% was used to filter the results, with a minimum of 1 unique peptide also required. However, the use of a 1% FDR value when using a non-specific digest approach is potentially questionable, as it could be overly punitive due to much higher chances of matching peptides in the decoy database. Therefore, where peptides from prohormones were identified, additional peptides from that protein with −10LogP values below the 1% FDR rate were included in the search results. The two high flow rate analysis files were searched using the same parameters.

| Proglucagon-derived peptides
Interestingly, multiple peptides derived from proglucagon were detected in the extracts (Table 2, Figure 1). A total of 16 peptides from the proglucagon peptide sequence were detected using the 1% FDR value; however, the Peaks software matched two extra peptides that were below the −10LogP value of 27.2 set by the 1% FDR setting.
In total, twelve peptides were detected from the N-terminal region of proglucagon which contained the GRPP sequence (proglucagon 1-30). These twelve peptides included full-length GRPP, and GRPP with amino acids missing at both N-and C-terminals, thus matching peptides identified previously by LC/MS to be secreted from human pancreatic islets. 18  presence; however, no peptide could be found that matched its proposed molecular weight. The same process was performed for oxyntomodulin, which has previously been detected in plasma samples using LC/MS approaches by our laboratory (data not shown), but was not found in this case after manually searching the LC/MS datasets.

| Pancreatic polypeptide
In Case 1, plasma levels of pancreatic polypeptide (PP) measured by immunoassay were 542 pmol/L, significantly higher than the reference range (NR: 0-300 pmol/L). The LC/MS analysis identified two peptides from the PP prohormone sequence and two additional peptides below the 1% FDR level, one of which corresponded to an N-terminal extended pancreatic polypeptide that has been detected in previous studies. 18

| Chromogranin A
Chromogranin A (CHGA) is involved in the storage and secretion of regulatory peptides, and is also subject to post-translational cleavage to form shorter peptides with unclear physiological roles. Chromogranin A is a marker of enteroendocrine cells in tissues and is used clinically as a plasma biomarker of NETs. 20 The plasma level of CHGA measured by immunoassay in Case 1 was reported as 146 pmol/L (NR: 0-60 pmol/L). The LC/MS analysis identified a total of nine peptides from the pro-peptide, three below the 1% FDR value (Table 2), including two previously characterised peptides from CHGA known as LF-19 and GR-44. Of the CHGA peptides identified in plasma, all but one were also identified in the previous analysis of peptides secreted from pancreatic islets, 18 and all peptides were detected in a peptidomics analysis of an islet cell line (QGP-1). 21 One detected CHGA peptide was an extended form of LF-19 with a 14 amino acid C-terminal extension. Multiple spectra were also obtained from the C-terminal region of vasostatin-2, which incorporates residues 97-131 of CHGA. This peptide is not specified in the database as a definitive CHGA peptide, but has been named as vasoconstrictioninhibiting factor (VIF). 22 The ability to detect CHGA peptides suggests that LC/MS could potentially be used in the diagnosis and monitoring of NETs.

| Secretogranin 1/chromogranin B
The plasma concentration of chromogranin B (CHGB) measured by immunoassay was reported as being higher than the normal range (205 pmol/L, NR: 0-150 pmol/L). 17 One peptide from this protein was detected above the 1% FDR (residues 90-132), which could potentially be a dipeptidyl-peptidase cleaved version of peptide 88-132 and was detected in the human islet cell line study. 21 Looking at lower −10L0gP values, three more peptides from CHGB were detected, one of which has been identified previously in pancreatic islets (residues 600-613). 18

| Glucose-dependent insulinotropic polypeptide (GIP)
A single peptide was identified from the GIP prohormone (22- were only 74, 10 and 14 pmol/L, respectively. 17 The peptide YY concentration, according to the previously reported immunoassay data, was 237 pmol/L, which is likely to be in the range of the LC/MS system, since other peptides were identified in the low hundred pmol/L range (e.g. intact glucagon was present at 107 pmol/L). 17 The reason for PYY not being detected is unknown, but it is possible this peptide may have degraded, or was too low in signal to trigger an MS/MS spectrum.

| Hepcidin
Multiple peptide matching sequences from hepcidin were identified; however, as the methodology did not include a reduction and alkylation step, the search result did not identify the 25 amino acid hepcidin. This peptide hormone is produced in the liver and believed to be important in the regulation of iron uptake. 23 Instead, the peptides that were identified came from the N-terminal region of the pro-peptide, and likely represent peptides formed in parallel with hepcidin by furin-mediated processing. 24 Hepcidin is a peptide with at most 25 amino acids from the C-terminal region of the prohormone and contains eight cysteines which form four internal disulphide bonds.
These internal bonds confound the PEAKS spectral matching identification software; however, performing reduction and alkylation on plasma samples extracted using the described method results in the identification of hepcidin 25 and some of its catabolites (data not shown).

| Thymosin B4
Thymosin B4 is a 4.9-kDa peptide involved in the organisation of Gactin fibres within the cytoskeleton of many cells in the body. 25 The presence of this peptide in plasma could be due to its release from dying cells, or from white blood cells or platelets circulating within the blood plasma. Although its source is unclear, it is present at concentrations high enough to be measured using a high-throughput LC/MS/MS methodology, should it be of interest to monitor either the peptide itself or its oxidised form.

| Neurosecretory protein VGF
Peptides that were detected from VGF were mainly from the Nterminal region immediately after the signal peptide. A total of eight peptides were identified, four of which were over the 1% FDR setting.
This protein is believed to be involved in the secretory process of regulatory peptides 26 and secreted peptides from VGF were identified in islets. 18 These peptides have previously been reported in plasma extracts from an exercise study. 10 None of the peptides that were identified have been classed as bioactive peptides from the VGF pro-protein. 26

| Secretogranin 2
Secretogranin 2 (SCG2) is a protein involved in neuroendocrine secretory granule processing, and is also believed to be the source of bioactive peptides. 26 A total of nine peptides were identified above the 1% FDR setting, and four additional peptides were matched.
One had a − 10LogP value of 15.77, which corresponded to the peptide manserin (residues 527-566). A potential DPP4-cleaved manserin peptide was also detected (529-566). The other SCG2 peptides identified were from the C-terminal region of the protein (569-610), as well as potential DPP4-cleaved versions. The peptide from the C-terminal region of SCG2 was previously identified as an islet-secreted peptide, 18 further validating its identification in plasma in this study. The two aliquots of plasma from Case study 2 were extracted and analysed using a high-throughput LC/MS and IDA-based approach, returning significant numbers of peptides for CHGA, which was the top hit. Other proteins that were identified included CHGB, SCG2, neurosecretory protein VGF, PCSK1N and hepcidin (Table S2, supporting information). The majority of peptides that were matched in the high flow rate analysis of plasma from Case study 2 had also been identified in the nano-LC/MS analysis of Case study 1. A major omission from this analysis was that Peaks did not return a hit for any glucagon gene products, despite the high plasma glucagon levels

| High-throughput full-scan peptide screening analysis
The in human plasma. 27 The peak area ratio values of the selected peptides were plotted and peptides that are commonly measured to diagnose glucagonomas displayed clear separation between the values obtained from the glucagonoma sample extracts and the controls, whereas other plasma peptides showed similar levels ( Figures 3A and 3B).
The peptide peak area ratios obtained in the high flow rate analysis suggest that this approach could be developed as a diagnostic tool for confirming the presence of a glucagonoma, as the level of glucagon, proglucagon 72-108, CHGA, CHGB and SCG2 peptides were substantially higher than in the controls. Interestingly, analysis of the pre-and post-octreotide treatment samples from Case study 1 showed that treatment with octreotide caused the plasma level of all peptides displayed in Figure 3A to fall after 4 days. Furthermore, inclusion of (A) (B)  Further validation would be required before this approach can be used as a diagnostic tool; however, our preliminary methodology definitely shows promise. The extraction method could also be improved further, as the recovery of insulin, C-peptide and glucagon was calculated as 45, 55 and 34%, respectively. As the methodology has been developed as a generic peptide extraction method, further improvements could be made to improve recovery of specific peptides.

| CONCLUSIONS
This plasma peptidomics methodology has demonstrated the ability to The LC/MS system identified a large glucagon peptide (1-61) and its degraded form, which was hypothesised in the previous report to be the cause of a large concentration of "total" glucagon defined using a C-terminal specific antibody. 17 By LC/MS, there was no evidence of glicentin or oxyntomodulin in these samples, despite our having detected these peptides in other analyses using the same extraction technique. The production of species containing the C-terminus of glucagon rather than oxyntomodulin is consistent with the tumour being derived from pancreatic alpha cells rather than the gut, as prohormone convertase 2 is responsible for proglucagon cleavage at the C-terminus of the glucagon sequence, 30 and is expressed in normal alpha-cells but not intestinal L-cells.
This methodology could potentially be used to study and diagnose other NETs that produce a range of prohormones and their cleavage products at supra-physiological levels. Given the marked heterogeneity that exists amongst neuroendocrine tumours, identifying novel biomarkers or combinations of biomarkers that enable patient stratification may have prognostic or therapeutic utility.
Developing innovate methodologies, such as the one we describe, may provide such an opportunity. In the current study, however, it is notable that the cases of metastatic glucagonoma we describe both had significant disease burden and it remains to be determined how well our methodology will perform in less advanced disease.
Significant further validation is required before the methodology would be applicable in a clinical setting, such as developing a quantitative method for glucagon-gene-derived peptides and the other secretogranin peptides, which was outside the scope of this study. Multiple synthetic peptides, and their equivalent stable isotope labelled internal standards, would need to be synthesised to develop and validate a quantitative methodology. Validation would also involve the analysis of larger patient sample numbers from this rare disease to demonstrate the selectivity and specificity of a mass spectrometric assay compared to the existing immunoassay methodology. However, the described organic solvent precipitation method followed by SPE has demonstrated its capability of enriching for multiple peptide species, therefore making it an ideal pre-analytical technique.
Eventually, a multiplexed and high-throughput method that targets multiple prohormone-derived peptides and vesicle-associated proteins could be developed to diagnose NETs and, importantly, monitor response to treatment more effectively than current radiology-based surveillance protocols. This would enable a single LC/MS analysis to be used rather than multiple immunoassays for each target peptide/protein. It would not only make the diagnosis pathway simpler, cheaper and require less plasma, but would also indicate the exact peptide sequences produced rather than relying on their varying cross-reactivity against a panel of antibodies.