Probing the urinary proteome of severe acute pancreatitis

Abstract

Background. Proteinuria is a characteristic feature of severe acute pancreatitis (SAP) that may allow unique insights into AP pathophysiology. This study used a proteomic approach to differentiate the abundant urinary proteins in AP patients. Materials and methods. Urine samples were prospectively collected from 4 groups (5 SAP, 10 mild gallstone AP, 7 mild alcohol AP, 7 controls). Reverse-phase high-performance liquid chromatography (RP-HPLC) and matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (LC MALDI) were used to identify urinary proteins and determine any differences between the groups. Results. There were 17 RP-HPLC major peaks in SAP groups of significantly greater absorbance magnitude than the corresponding ones in mild and control groups. Various mass spectrometry methods were used to identify 21 different parent proteins from these SAP peaks. They included fibrinogen, serum amyloid A, insulin and calcitonin gene-related peptides. There were no identifiable protein peaks at the corresponding elution times in the mild pancreatitis and controls samples. Discussion. Proteomic techniques offer a unique unexplored window into AP pathophysiology. The utility of these proteins as markers of pancreatitis severity now need to be further investigated and the identification extended to the full urinary proteome as technology permits.

Introduction

Acute pancreatitis (AP) remains a common intra-abdominal disease with a complex pathophysiology. The overall outcome has improved, but specific treatment(s) remains elusive. The challenge is the early identification and treatment of patients who will develop severe acute pancreatitis (SAP) and infected pancreatic necrosis. This study documents the development of a new approach to investigating the pathophysiology of AP by applying laser desorption/ionization time-of-flight mass spectrometry (LC MALDI) proteomic techniques to urine from these patients.

The science of proteomics defines the protein complement expressed by the genome, the so-called proteome ¹. It is a powerful tool that not only identifies peptides and proteins, but also quantifies their up- and down-regulation. The isolation of proteins from complex mixtures has been possible using techniques including high performance liquid chromatography (HPLC) and two-dimensional polyacrylamide gel electrophoresis (2D PAGE). Subsequent identification of proteins and peptides has historically been more problematic and relied on N-terminal amino acid sequencing by Edman degradation. Although a robust and tested method, this technique is time-consuming, expensive and requires relatively large amounts of very pure protein sample. Proteomics has come of age because of two key developments: soft ionization mass spectrometers, which can now rapidly define the molecular weights and amino acid sequences of peptides from relatively small quantities of complex sample mixture ² and publicly available databases that can identify proteins directly from these peptide fragment masses ³.

The proteinuria associated with pancreatitis is well recognized but neither fully understood nor systematically characterized. Low molecular weight proteins have been demonstrated in the urine during the acute phase of the disease, but these were absent later in the recovery period ⁴. Others have shown that microalbuminuria during the first 36 h of hospital admission was likely to be associated with more serious complications ⁵. Subsequent studies have tended to identify specific isolated urinary proteins in patients with AP, including trypsinogen activated peptide (TAP) ⁶, carboxypeptidase A ⁶, thyrotropin releasing hormone ⁷, hydroxyproline and fibronectin ⁸. However, the characterization of the complete urinary proteome in AP has yet to be accomplished and is likely to be difficult due to the variability of proteins in urine, both qualitative and quantitative ⁹. The low abundance and post-translational modification of urinary proteins means that traditional proteomic techniques such as 2-DE PAGE and Edman sequencing are not as useful as when these techniques are applied to studies of serum and tissue samples.

The aim of this study was to investigate the utility of a recently acquired LC MALDI proteomic technology platform in the identification of high abundance urinary proteins in patients suffering from AP.

Materials and methods

Study population

Adult patients with a primary episode of AP as defined by abdominal pain associated with either a serum amylase >1000 U/L or computed axial tomography scans (CT) demonstrating pancreatic inflammation/necrosis, were identified on admission to Auckland Hospital (an 800 bed tertiary hospital serving a 280 000 catchment population). Patients were excluded if the duration of illness was greater than 48 h; they were currently taking diuretic or anticoagulation medication; they had chronic renal failure (pre-pancreatitis serum creatinine >0.12 mmol/L); or they had diabetes, nephrotic syndrome, or evidence of acute renal failure at any stage of the urinary collection (serum creatinine >0.12 mmol/L or 24 h urinary volume <400 ml). Normal controls were healthy male staff volunteers. The aetiology of AP was classified as gallstones (demonstrated by ultrasound, endoscopic retrograde cholangiography (ERCP), or at laparotomy), alcohol (when average daily intake exceeded 50 g, or there was a recent binge event in the clinical history and absence of other causes), or ‘other’ (when another or no cause was identified). ‘Severe’ AP was defined as AP associated with organ failure and/or local complications such as pancreatic necrosis, abscess, or pseudocyst ¹⁰. APACHE II ¹¹ scores were calculated at admission to assess severity. The study was approved by the Auckland Ethics Committee and each patient gave informed consent.

Sample preparation

Urine was collected during each 24 h of hospitalization, from which 20 ml aliquots were extracted for analysis. The urine was initially centrifuged at 2000 rpm for 25 min to remove particulate material. The supernatant was then loaded onto SepPak Plus C18 cartridges (Waters Corporation, Milford, MA, USA) and eluted with 4 ml of 0.1% trifluoroacetic acid/80% acetonitrile. The acetonitrile was evaporated off before injecting the residual sample volume through a 0.20 µm filter (Sartorius MiniSart RC4 filter).

Reverse-phase high-performance liquid chromatography

Reverse-phase high-performance liquid chromatography (RP-HPLC) was used to determine urinary protein profiles of each patient. A 100 µl portion of prepared urine sample was injected onto a Jupiter C18 5 µ 300 Å 4.6 mm×250 mm column (Phenomonex NZ, Auckland) through a manual Rheodyne 200 µl loop and Applied Biosystems 140B Solvent Delivery System. Buffer A comprised 0.08% (v/v) trifluoroacetic acid (TFA) in MilliQ H₂O and buffer B was 80% acetonitrile/20% buffer A. The separation gradient was 10–70% buffer B over 30 min at flow rates of 1 ml/min. Absorbance at 214 nm was detected by 1000S Diode (AUFS 3) and traces printed onto Yokogawa 3057 recorder. Column eluent was collected every 30 s in 1.5 ml microcentrifuge tubes loaded into a BioRad Model 2110 fraction collector.

Repeat RP-HPLC purification runs were performed with diluted samples when absorbance peaks exceeded maximum AUFS scale levels. Those samples with complex peak patterns were submitted to further RP-HPLC with a shallower buffer gradient (20–50% buffer B over 30 min) to further purify peak components. Additional attempts were also made to further isolate those fractions that still comprised complex mixtures of proteins by injecting 50 µl of the collected fraction onto a narrow bore Jupiter reverse phase column (C18 5 µ 300 Å 2.0 mm×250 mm column; Phenomonex). The buffer gradient was started at 15% less than the concentration at which the fraction initially eluted off the RP-HPLC column, and increased to the eluting concentration over 30 min in a previously validated method ¹². Separated peaks were collected manually.

The urinary RP-HPLC elution profile traces were initially compared for major peak absorbance value differences. Group differences for this and other variables were determined by ANOVA with Tukey's post hoc testing, where p<0.05 was regarded as significant (Prism 4.0 for Mac, GraphPad Software Inc.).

Protein sequencing

Edman degradation is a technique where individual amino acids are sequentially cleaved from the N-terminus of the protein and identified to give an amino acid sequence. Acetonitrile from the RP-HPLC fractions was removed by vacuum centrifugation. Fractions were reconstituted with an equal volume of 0.1% TFA before 50 µl was loaded for analysis onto Applied Biosystems Procise Sequencer Model 610A 2.1.

Mass spectrometry

Mass spectrometry separates proteins on the basis of their mass and can fragment them into ions that can be used to identify the amino acid sequence. Proteins are initially ionized by either laser ionization of a dried crystalline mixture (matrix assisted laser desorption ionization or MALDI), or from a solution of highly charged droplets sprayed in the presence of a strong electric field (electrospray ionization or ESI). The ions may then be accelerated along a flight tube that separates them by their varying velocities based on individual mass-to-charge ratios (the so-called time-of-flight or TOF). A certain degree of ion fragmentation occurs along the flight tube due to inherent ion instability and post-source decay (PSD). PSD can be used to determine amino acid sequence and can be enhanced by chemical-assisted fragmentation (CAF) (Ettan™ CAF™ MALDI Sequencing Kit, Amersham Biosciences), which adds a sulphydryl group to the N-terminus of tryptic digested peptides, augmenting the PSD signal of y-series ions. A further advancement is the quadrupole time-of flight (Q-TOF) technique that gives a more reproducible and characteristic fragment pattern that enables identification of amino acid sequences when used in conjunction with computerized databases such as Mascot and ProID.

In this current study all fractions were trypsin digested before mass spectrometry. The fractions were reconstituted in 50 µl of 100 mM ammonium bicarbonate before the addition of 0.5 µg sequencing grade porcine trypsin (Promega; 0.5 µl of 1 µg/µl in 1% acetic acid). The solution was then incubated at 37°C for 2 h before the reaction was stopped by the addition of 1 µl of 100% TFA. The solution was vacuum centrifuged to near dryness and the protein was reconstituted in 0.1% TFA. Tryptic peptides were sulphonated with Ettan CAF-MALDI sequencing kit (Amersham Biosystems, Uppsala, Sweden) to enhance PSD identification. In brief, peptides were bound to a prepacked reverse-phase C18 matrix in pipette tips (ZipTip™, Millipore Australia Pty Ltd, Sydney, Australia) and O-methylisoureahydrogen sulphate solution (pH > 11.7) was loaded onto the ZipTip™. Tips were incubated in a sealed tube at 37°C for 2 h to block the ɛ-amino group of lysine residues and leave the N-terminal amino group intact. Freshly made CAF-labelling reagent was then loaded onto the ZipTip™ and allowed to react for 3 min at room temperature to sulphonate the N-terminus. The ZipTip™ was washed of excess CAF-labelling reagent and the derivatized peptide was eluted for further analysis. The sulphonation adds 136 Da to peptides containing a C-terminal arginine.

MALDI-TOF PSD was performed on an Applied Biosystems Voyager DE-PRO. Fractions were reconstituted with 0.1% TFA, mixed 1:1 in 1% α-cyano-4-hydroxycinnamic acid (αCHCA), spotted onto a Teflon plate (0.5 µl) and allowed to dry at room temperature. Protein masses were determined in positive reflector mode, accelerating voltage 20 000 V, grid voltage 62–75%, mirror voltage ratio 1.12, guide wire 0.002%, extraction delay time 50 ns, laser intensity 2000–3000, laser repetition rate 20 Hz, acquisition mass range 750–10 000 Da. Near-point external mass calibrations were performed daily with angiotensin and insulin standards.

For PSD analysis the instrument was set for PSD and the ion selector was set to the m/z values of the precursor ions sequentially. The number of voltage segments required was determined by the precursor ion mass and the lowest mass to be analysed. PSD spectra were thus acquired in a set of 10–18 spectra using decreasing reflector voltages. Laser power was adjusted so that unit mass resolution was maintained, and 100–300 laser shots were collected for each voltage segment. The resultant spectra were manually interpreted by measuring the masses between each y-ion peak, and the resultant amino acid sequence was entered into NCBI protein-protein BLAST search (www.ncbi.nlm.nih.gov/BLAST).

ESI-Q-TOF was performed on a quadrupole orthogonal acceleration time-of-flight tandem mass spectrometer (QSTAR XL, Applied Biosystems). Tryptic peptides were analysed by electrospray ionization in the Q-TOF mass spectrometer using a nanoelectrospray ion source. The capillary voltage was set to 1800 V and the curtain gas was set at 15 (Units). Low-energy collision-induced dissociation was performed using nitrogen as a collision gas at a recorded pressure of 6 psi, and the collision energy was optimized for different by ramping up the Q0 energy. Data acquisition was controlled by Analyst QS software (Applied Biosystems) using an automated acquisition mode for MS and MS/MS experiments. Obtained spectra were searched against a database using Pro ID with confidence > 73%. Any matches were then checked against Mascot database where p<0.05 was taken as proof of identity.

Results

Urine samples were collected from 5 patients with severe acute pancreatitis (SAP) (median age 69 years, range 60–78, 5 males), 10 patients with gallstone mild acute pancreatitis (MAP) (median age 65 years, range 44–78, 7 males) and 5 patients with alcohol MAP (median age 39 years, range 19–58, 4 males), and 7 control subjects (median age 42 years, range 35–50, 7 males). Patient samples were collected on the first day of admission. The 24 h urine volume of SAP patients was 904±272 ml (mean±SEM); gallstone MAP was 555±175 ml; alcohol MAP was 606±118 ml (p=0.45, one-way ANOVA). One patient died (SAP). The hospital stay for SAP was median 31 days, range 2–100, median 4 days, range 3–6 (gallstone MAP), and median 5 days, range 2–7 (alcohol MAP), respectively. The APACHE II score for the SAP group on admission was 10.4±1.5 (mean±SEM), 2.2±0.6 (gallstone MAP), and 0.6±0.3 (alcohol MAP) (p<0.0001). SAP was caused by gallstones (two patients), alcohol (one), and other (two patients). All SAP patients were admitted to intensive care for nutritional support (one total parenteral and four enteral nutrition) and prophylactic antibiotics (meropenem; Merrem, AstraZeneca, Australia) were given to all of them. Three SAP patients eventually had surgical intervention later in their admission for infected pancreatic necrosis, one had percutaneous drainage of acute fluid collection, one had an endoscopic retrograde choledochopancreatogram, and one required venovenous dialysis after 5 days of admission.

RP-HPLC

There was no single common pattern to the RP-HPLC elution peak profile in any group, nor could a single pattern be used to discriminate between the aetiological subgroups. However, there were 17 major peaks in the SAP group traces that had significantly greater absorbance unit (AU) values at 214 nm than those at identical elution times in both the mild groups and controls (Figure 1). These peaks eluted at differing buffer concentrations spanning unique positions at 22–43% buffer B and had an AU mean of 2.32±0.22 (±SEM) compared with those peaks from an equivalent extracted urine volume and elution time in the other groups (0.49±0.23 for mild gallstone AP, 0.19±0.04 for mild ethanol AP, and 0.20±0.04 for controls, respectively, p<0.0001). These large peaks were invariably found to contain a complex mixture of proteins and further RP-HPLC was able to partially resolve the individual protein components.

Figure 1. .

Representative RP-HPLC traces of urine from control, mild acute pancreatitis and severe acute pancreatitis patients. The gradient increase of acetonitrile (ACN) concentration from 10% to 50% over 30 min is shown. Absorbance 214 nm.

Edman sequencing

Three peaks were initially submitted for Edman degradation sequencing. Only one peptide was successfully sequenced and it was identified as collagen α1 (I) precursor.

Mass spectrometry

Of the 160 spectra obtained, 77 (48%) could be identified manually or by database searches. All proteins required trypsin digestion for CAF modification and database searching from ESI Q-TOF. There were 77 peptide sequences identified from the 17 major peaks isolated by HPLC above; 23 were identified by MALDI-TOF PSD with CAF, and 43 by ESI Q-TOF alone. These peptides corresponded to 21 different proteins (Table I). Of note, collagen I (α1 and α2) chains and fibrinogen sequences were each found in four of the five SAP patients, calcitonin gene related peptide (CGRP) in two SAP, haemoglobin in two SAP, and serum amyloid A in two SAP patients. α1 Anti-trypsin was found in three SAP patients. No proteins were identified from the urine in the mild cases or control urine at the same RP-HPLC elution positions.

Table I. Proteins identified in the urine of patients with acute pancreatitis.

Protein identity	Peptide sequence used to identify protein	HPLC elution time (% buffer B)*	Method of sequencing
α1 antitrypsin	SPLFMGK	28%, 38%	ESI Q-TOF
Apolipoprotein A1	DEPPQSPWDR	28%, 38%	ESI Q-TOF
CGRP	SNFVPTNVGSK	23%, 41%	ESI Q-TOF
Chain P human platelet profilin	PPPPPPPPPP	22%	ESI Q-TOF
Collagen α1 (I) precursor	PPGPPGPPGPPG	11%	CAF PSD
	AHGDRGEGP	26%	Edman
	GFSGLDGAK	29%	ESI Q-TOF
	GETGPAGPAGPIGPVGAR	33%	ESI Q-TOF
	GFSGLDAK	38%	ESI Q-TOF
	GQPGVOGFPGPKGANGEPGK	38%	ESI Q-TOF
Collagen α2 (I) precursor	GPAGPSGPAGKDGRR	21%	CAF PSD
	GLHGEFGL	30%	CAF PSD
	GHNGLQGLPGIAGHHGD	32%	CAF PSD
	IGQPGAVGPAGIR	33%	ESI Q-TOF
	GLPGVAGAVGEPGPLGIAG	43%	CAF PSD
Collagen α2 (IV) precursor	PGPPPVILPGMKDIKGEK	33%	CAF PSD
Collagen α1 (XII)	PGPGGRRPGFPG	38%	CAF PSD
Cytotoxic T lymphocyte protein 4	GIASFVBEYASPGKATEVR	32%	ESI Q-TOF
Fibrinogen	VNDRWK	39%	ESI Q-TOF
Fibrinogen α precursor	OADEAGSEADHEGTHSTK	22%	ESI Q-TOF
	ESSSHHPGIAEPSR	32%, 35%	ESI Q-TOF
	GSESGIFTNTK	33%, 36%	ESI Q-TOF
GDP-mannose 4,6 dehydratase	REFPEKSFW	27%	CAF PSD
Haemoglobin A chain	VLSPADKTNVK	27%, 38%	ESI Q-TOF
	VGAHAGEYGAELER	34%, 41%	ESI Q-TOF
	VGAHAGEYGAEALER	36%, 38%	ESI Q-TOF
	TYFPHFDLSHGSAQVK	38%	ESI Q-TOF
	VDPVNFK	38%	ESI Q-TOF
Haemoglobin B chain	VVAGVADALAHKYH	32%, 36%	CAF PSD
	VVAGVANALAHKYH	39%	ESI Q-TOF
	VVAGVANALAHKY	27%, 32%, 36%	ESI Q-TOF
	VVAGVNAAIAH	33%	CAF PSD
	VNVDEVGGEALGR	39%	ESI Q-TOF
	SAVTALWGK	41%	ESI Q-TOF
Haptoglobin precursor	GSFPWQAK	39%	ESI Q-TOF
Immunoglobin heavy chain	AEDTAVYYCAKRPPGYSSSWER	36%	ESI Q-TOF
Insulin	GEFYTPK	33%	ESI Q-TOF
Isocitrate dehydrogenase (S. enterica)	LVRRAAIEYAITNAR	38%	ESI Q-TOF
Lysine decarboxylase (S. enterica)	NALGILGGPKR	27%, 33%	ESI Q-TOF
Serum amyloid A	SGKDPNHFRPAGLPEK	24%, 28%, 33%, 36%	ESI Q-TOF
	LTGHGAEDSLADQAANK	28%, 33%	ESI Q-TOF
	GPGGAWAAEVISNAR	29%	ESI Q-TOF
	DPNHFRPAGLPEK	33%	ESI Q-TOF
	GAEDSLADQAANK	33%	ESI Q-TOF
	SGRRDPNHFRPAGLPEK	33%	ESI Q-TOF
	FGHGAENSDANEAA	35%	CAF PSD
	FFGHGAEDSLADQA	39%	CAF PSD
Uromodulin	VLNLGPITR	36%	ESI Q-TOF

CAF-PSD, chemical assisted fragmentation-post source decay; Edman, Edman degradation sequencing; ESI Q-TOF, electrospray ionization quadrupole time-of-flight; HPLC, reverse-phase high-performance liquid chromatography; CGRP, calcitonin gene-related peptide.

*Multiple elution (% buffer B) entries are given when the same peptide sequence was eluted at different times by HPLC.

Interestingly, many of the peptides had significant amino acid sequence homology yet considerable differences in apparent hydrophobicity as based on elution time. For example, the haemoglobin B sequence VVAGVANALAHKYH was found at 39% buffer B, but the similar sequence VVAGVADALAHKYH (different by only one residue at position 7 (N for D)) eluted earlier at 32% and 36%. Furthermore, many of the peptides with identical sequences displayed different hydrophobicities. For example, haemoglobin B chain VVAGVANALAHKY eluted at three different concentrations of acetonitrile (27%, 32% and 36%) yet had the same mass (1148 Da) when subjected to mass spectrometry.

Discussion

This study successfully demonstrated the utility of a modern proteomic approach to SAP by identifying a series of proteins in the urine of AP patients using RP-HPLC and mass spectrometry. The RP-HPLC profiles of the urine from patients with SAP collectively showed 17 major peaks that were clearly identifiable compared with those found at similar elution times in MAP and control samples. These marked urinary proteomic differences were present despite there being no overt evidence of acute renal failure at the time of collection in the SAP patients. These major peaks contained a mixture of proteins and peptides that were resolved to varying degrees by mass spectrometry. The protein fragment masses from 21 different proteins were ultimately identified, with the most frequent being collagen I (α1 and α2) and various fibrinogen fragments. These results are encouraging but also demonstrate that the full proteome is likely to remain a challenge due to an inability to obtain recognizable mass signature identifications on many targets. This is probably a consequence of the fragmented nature of the urinary proteins combined with the known propensity for multiple amino acid side chain modifications ¹³, which the current databases were unable to resolve by mass signature alone at this resolution. This limitation is likely to be resolved in the future as computer databases become more sophisticated hence allowing a further unique opportunity to investigate AP.

Proteomics is a rapidly advancing field that is a direct offshoot of the recent characterization of the genome. Having access to a gene blueprint via openly available databases has enabled the characterization of proteins in a variety of human diseases and the identification of new biomarkers. Underlying the success of proteomics is the development of high-throughput technologies that accurately and rapidly identify small amounts of protein from complex mixtures. The discovery and development of soft ionization methods ¹⁴ for proteins has led to mass spectrometry superseding Edman degradation for the identification and sequencing of proteins. At the time of commencing these studies, these techniques had not yet been applied to AP and this study set out to provide proof of principle that it might have future application for research into this disease.

The two obvious fluids to investigate in pancreatitis are serum/plasma and urine. The former presents a particular challenge because in blood only 22 proteins comprise 99% of the total protein ¹⁵, meaning that low abundance proteins are difficult to isolate or to monitor for pathological changes. One mass spectrometry technique (surface enhanced laser desorption ionization or SELDI) has already been applied to the serum of pancreatitis patients, showing that protein profiles differ ¹⁶. Urine offers a unique advantage in this disease since it is a micro-filtrate of serum and is arguably a more accessible body fluid. It is also an extremely relevant sample source, since renal function changes are known to be associated with AP ¹⁷. Until recently, applying proteomics to urine has been difficult as 2D PAGE is hindered by the low abundance of proteins in normal urine, the ‘contamination’ effect of the more ubiquitous albumin, and the small size of most urinary protein fragments. Novel approaches to characterize the proteome of normal urine have included concentration methods such as dye and acetone precipitation ¹⁸,¹⁹, and separation techniques such as protein fractionation ⁹. However, these are time-consuming and require large amounts of urine. Indeed, one study required 90 mg of urinary protein (comparable to 4.5 litres of urine) to identify 124 peptides ²⁰, making it impractical in critical illnesses such as pancreatitis. Severe pancreatitis is unique in that it is associated with significant proteinuria that negates the need for complicated concentrating procedures. The ability of new proteomic technologies to circumvent these obstacles without urine pretreatment has already been recognized ¹³, but this is the first study, to our knowledge, that has applied them to an investigation of a critical illness like AP. We reasoned that for this first proof of principle study a simple analysis of direct samples of urine collected and compared from mild and severe cases might reveal what abundant proteins appear in the urine early in the development of a severe attack. There was no difference in the total 24 h urine output of any of the study groups, so we deliberately chose not to concentrate multiple samples of normal or mild urine to artificially increase the protein content ²¹, as we were interested in what proteins appeared at abundant levels in realistic bedside patient samples. This would allow any findings to be transferred to the normal clinical setting.

Proteinuria has long been recognized as a feature of SAP and has even been proposed as a prognostic indicator of outcome ⁵. It is a reliable feature of early disease that has led researchers to speculate that a protein released in pancreatitis may be responsible for the changes in renal protein handling ⁴; a tantalizing hypothesis that may explain the onset of organ dysfunction in other areas. Unlike other studies of critical illness that have discovered a high correlation of albuminuria alone with severe disease ²¹, this study revealed there to be a complex mixture of urine proteins, many of them reflecting aspects of the general inflammatory process that is occurring in early severe pancreatitis. We found many peptide masses that did not match the existing databases. While many are expected to be due to chemical modifications to the proteins, it remains possible that one or more of these is from a potential ‘pancreatitis-specific’ novel urinary protein. It will be interesting to make careful comparison to the urinary proteomes of other inflammatory illnesses as they become available in the future, to determine if any of these signatures are specific to the pancreatitis process itself. It was of note that many of the parent proteins successfully identified through our study have been implicated previously in AP but usually in serum. This current study now confirms by this independent technique their signature in urine and thereby this fluid's potential to be probed for their measurement.

In this study there was no clinical evidence of overt renal failure, suggesting that the finding of these diverse urinary proteins in severe pancreatitis is an early feature and therefore worthy of close investigation for its prognostic utility. The spectrum of proteins displayed in the first 24 h of admission by the urine was entirely consistent with the evolving inflammatory process and confirms that this will likely be a useful new viewpoint for monitoring the evolution of the disease. For instance, serum amyloid A (SAA) is an acute phase reactant that has been implicated as a modulator of neutrophil function and oxidative injury in acute inflammation ²². Its role in pancreatitis has not been defined but elevated serum levels have been found to be a better predictor of outcome than C-reactive protein (CRP) ²³. To the best of our knowledge, this study is the first to describe the presence of an SAA signature in the urine of critically ill pancreatitis patients, and invites further investigation into its role and measurement in the urine as a severity indicator. In a similar manner, fibrinogen and its fragments have also been postulated as a predictor of outcome in AP, yet efforts with plasma measurements have been disappointing ²⁴. The findings we report here of readily identifiable fibrinogen-derived peptides in the urine of severe pancreatitis patients may therefore offer an alternative matrix in which to re-investigate the predictive value.

The presence of collagens I and IV in this study is in keeping with experimental studies into pancreatitis pathogenesis. Products of lipid peroxidation stimulate fibroblasts to synthesize increased levels of collagen I ²⁵ that have been shown in the peri- and intra-lobular regions of mouse pancreas 4 days after the intraductal injection of trypsin ²⁶. These have yet to be confirmed in human studies and our findings would point to the potential utility of urinary collagens as an early measurable change in human SAP.

Several unusual targets were identified and included peptides identified as belonging to Salmonella enterica. Surprisingly this has been associated in various case histories of AP ²⁷. It is widely believed that bacterial translocation through the intestine is a source of infection once pancreatic necrosis has developed, yet this study suggests that systemic presence may occur earlier. Indeed intestinal integrity is known to be reduced early in pancreatitis ²⁸ and the role of enteric bacteria such as S. enterica also warrants further study.

Th e presence of insulin (pancreatic islet hormone) and calcitonin gene-related peptides (neuropeptide) was unexpected and to our knowledge this finding has not been reported previously in the urine of AP patients. It is therefore interesting to speculate on the prognostic value of such small peptides, particularly since one is of direct pancreatic origin.

Other parent proteins for the fragments described in this study have not been associated with AP but have been described in normal urine elsewhere. Indeed α-1 anti-trypsin, apolipoprotein, fibrinogen, haemoglobin, haptoglobin and uromodulin have all been identified in proteomic studies of concentrated normal urine ²⁰. Their presence in our study serves to validate our detection methodology. To date these proteins have only been identified in normal urine that has been greatly concentrated and so it seems likely that these normal urinary proteins are excreted in supranormal amounts in SAP given that we required no concentration step. Whether this is due to increased permeability of the glomerular filter, decreased proximal tubule reabsorption, or both, cannot be determined from this study. This raises the possibility that a spot urinalysis demonstrating these proteins could be an early indicator of renal involvement in the course of SAP before creatinine deteriorates.

Although this study was successful in overcoming some historical problems in identifying urinary proteins, it has raised a series of further interesting issues. First, there were several instances where mass measurements of some protein fragments indicated that they had undergone substantial and variable post-translational modifications. This problem has been reported before ²⁰,²⁹. The vast number of chemical modifications that include peptide oxidation, deamidation and carbamylation make many of the obtained mass spectrometry spectra in our study currently indecipherable. To overcome the problem of modifications and fragmentation, we used tryptic digestion to enable database recognition of unaltered or predictably altered smaller fragments. The main limitation that remains for this approach is the mass accuracy of mass spectrometers coupled with the capacity of the computerized databases to interpret peptide fragments and their chemical modifications, but it is conceivable that this will be resolved as new mass spectrometry technologies such as Fourier transformed mass spectrometry (FTMS), with increased mass accuracy down to 0.002 Da (2 ppm), are applied to biological samples and subsequent database resolution advances.

In an attempt to improve the identification process this study used ‘chemically assisted fragmentation’ to try to solve the problem of modified protein identification. Eliminating all but the ‘y ions’ from the mass spectrometry spectrum allows the mass of the modification to be isolated by simple manual interpretation. Furthermore, it enhanced fragmentation of the peptide that was generally poor with MALDI-TOF PSD. This technique has been reported in de novo sequencing before ³⁰ but never applied to urinary proteins.

Another inherent limitation of the techniques described in this current study was the need to trypsin digest all the urinary protein before identification. Although successful, our techniques did not allow direct identification of all of the initial parent molecules that supplied the subsequent trypsin-digested fragments. We could not therefore quantify to what extent the initial proteins were present in more than one modified state; although the multiple similar fragments and variable retention times suggest that this was probably a common occurrence.

In conclusion, this feasibility study has successfully applied recent proteomic techniques to identify many of the abundant protein signatures present in the urine of patients developing severe pancreatitis. Despite the limitations of mass resolution on current machines, databases and the intrinsic challenges of working with proteins in a urine matrix, this study provides the first key steps toward identifying the entire urinary proteome in AP. It also provides evidence that through the tools of modern mass spectrometry the urine now offers an exciting, albeit challenging new window into the pathogenesis of SAP. Encouraged by these findings the next task should be evaluation of these and other less abundant urinary proteome targets for their relative efficacy as severity markers and possible new directions for therapeutic design.