Serological diagnosis and prognosis of severe acute pancreatitis by analysis of serum glycoprotein 2

Subjects

Characteristics of the 153 patients with acute pancreatitis and 651 controls collected at the intensive care unit of the department of surgery (Otto-von-Guericke-University Magdeburg, 05/1995–03/2013) and 101 healthy subjects (in.vent Diagnostica, Hennigsdorf, Germany) are given in Table 1. Apart from acute pancreatitis, the following recruited disease controls are relevant for the differential diagnosis of acute upper abdominal in particular epigastric pain: chronic pancreatitis, pancreatic neoplasms, gastric and esophageal cancer, as well as peptic ulcer.

Patients with acute pancreatitis were stratified retrospectively according to severity using clinical and imaging data [10], [19]. Furthermore, the etiology of acute pancreatitis was determined in accordance with international guidelines [10]. For 30/152 patients, 3 or more consecutive samples could be obtained which resulted in 128 follow-up samples [median observation period: 28 days, interquartile range (IQR): 28 days]. Disease duration from clinical onset until admission was determined for all patients. For 152 and 146 acute pancreatitis patients, CRP and PCT levels were obtained, respectively.

All clinical diagnoses were based upon standard clinical, imaging, endoscopic and histological criteria. The diagnosis of chronic pancreatitis was established using a scoring system whereby patients with acute exacerbations were excluded [20].

The study was approved by the Local Ethics Committee of Magdeburg Medical Faculty and complies with the World Medical Association Declaration of Helsinki. Aliquots of the sera were stored at −80°C.

Anti-GP2 antibody production

Human GP2a and GP2b were expressed recombinantly and purified by Ni-chelate chromatography [21]. Polyclonal antibodies to total GP2 (GP2t) were produced by immunizing rabbits with an equal mixture of GP2a and GP2b (Pineda Antikörperservice, Berlin, Germany).

Monoclonal antibodies recognizing GP2a (K9) and GP2b (P102) were generated by immunizing mice with immunoconjugates [22] and selected by ELISA [23]. The animal experiments were approved by the local government (V3-2347-A16-4-2012).

ELISAs for the detection of GP2a/GP2t

The monoclonal anti-GP2a antibody K9 or polyclonal rabbit anti-GP2 antibodies at 1 μg/mL were coated onto Maxisorb plates (Nunc, Roskilde, Denmark) at 4°C overnight. After blocking with 1% bovine serum albumin in 50 mM Tris-buffered saline (Sigma Co, Taufkirchen, Germany) at room temperature for 1h, serum samples diluted 1 in 100 in 50 mM Tris-buffered saline with 0.2% bovine serum albumin were incubated for 1h and washed with Tris-buffered saline containing 0.1% Tween20 (Sigma) to detect GP2a and GP2t. Horseradish-peroxidase labeled polyclonal anti-human GP2 antibodies were added for 1h and developed with tetramethylbenzidine for 15 min (Seramun, Heidesee, Germany). The optical density was read using a microplate reader (SLT, Crailsheim, Germany) at 450 nm/620 nm. Purified recombinant GP2a or an equal mixture of GP2a and GP2b were used as standard material ranging from 0.1 to 10.0 ng/mL for the GP2a and GP2t ELISAs, respectively. The only obtained monoclonal antibody to the shorter GP2b interacted readily with both immobilized isoforms but as solid-phase antibody with the soluble isoforms at a low binding strength only. This did not allow developing a sensitive GP2b ELISA.

For interference experiments, GP2 containing sera were spiked with human hemoglobin (0.0–2.5 g/L), triglycerides as 20% emulsion of soybean oil (5.7–25.0 g/L), unconjugated bilirubin (30.0–100.0 mg/L), and GP2’s urinary homolog Tamm-Horsfall protein (uromodulin, 0.0–1.0 µg/L) (Sigma Co, respectively).

Statistical analysis

Data were checked for normality by the Kolmogorov–Smirnov test. The two-tailed, Mann-Whitney and Kruskal–Wallis tests were used to test for statistically significant differences of independent samples in 2 and more groups, respectively. Spearman’s rank correlation test was applied for within group comparison. Prevalence comparison between groups was performed by two-tailed Fisher’s exact test. p-Values of less than 0.05 were considered significant. Assay performance was analyzed by receiver-operating characteristics (ROC) curve analysis (Medcalc, Mariakerke, Belgium).

Regression of GP2a/GP2t levels

Serum GP2a and GP2t levels were linearly related to each other [coefficient of determination (R²)=0.49] when all patients with acute pancreatitis and controls were analyzed (Figure S1). Regression analysis in acute pancreatitis patients alone revealed a higher R² of 0.72 compared to controls (R²=0.29) and a significantly different slope and intercept (p<0.0001, =0.0001, respectively).

Correlation of GP2a/GP2t levels with disease duration

Both GP2a and GP2t demonstrated a significantly negative correlation with disease duration in acute pancreatitis (Figure S2) [Spearman’s coefficient of rank correlation (ρ): –0.22, –0.41, p=0.0053, <0.0001; respectively]. For further evaluation, patients with acute pancreatitis were stratified into three groups covering patients with disease duration of ≤3 days, with disease duration from 4 to 10 days and >10 days (Table 2). Onset of abdominal pain was defined as day 1 of disease.

Comparison of GP2a/GP2t levels in acute pancreatitis and control groups

Acute pancreatitis patients were age and gender matched with patients suffering from chronic pancreatitis, pancreatic neoplasms, insulinoma, and colon carcinoma (p>0.05).

Serum GP2a and GP2t by ELISA demonstrated significantly different levels in 153 acute pancreatitis patients and 752 controls (p<0.0001) (Figure 1A,B). Thus, median levels of GP2a and GP2t in acute pancreatitis patients demonstrated a significant decline from time period to time period (≤3 days, 4–10 days, >10 days) (p>0.05, respectively). Of note, several sera demonstrated GP2a and GP2t levels above the highest standard of 10 ng/mL and dilutions thereof could not be retested due to lack of material.

Figure 1:

Serum glycoprotein 2 isoform alpha (GP2a) (A) and total GP2 (GP2t) (B) levels in patients with acute pancreatitis (AP) and controls.

GP2a and GP2t were analyzed in 153 patients with AP, 651 disease controls, and 101 healthy subjects (HS) by enzyme-linked immunosorbent assay. Patients with AP were stratified into patients with up to 3 days of disease duration (n=12), between 4 and 10 days (n=47), and more than 10 days (n=94). As disease controls, 26 patients with chronic pancreatitis (CP), 125 with pancreatic neoplasms (PNpl), 118 with liver or biliary cancer (LBCa), 126 with gastrointestinal cancer (GICa), 40 with neuroendocrine tumor (NET), 40 with sarcoma (SA), 109 with benign liver or biliary diseases (bL/BD), 27 with bleeding or perforated peptic ulcer (PU), and 40 with peritonitis (PT) were investigated. (Optimized cut-offs for GP2a and GP2t obtained by receiver-operating characteristic curve analysis were illustrated by dashed horizontal lines. Data are displayed in Box-and-Whisker plots with far out values, defined as values that are smaller than the lower quartile minus 3 times the interquartile range, or larger than the upper quartile plus 3 times the interquartile range, displayed as solid triangles). p, Kruskal Wallis test, Conover post-hoc analysis.

Whereas the median GP2a and GP2t levels in acute pancreatitis patients until the 3rd day were significantly elevated in comparison with all control groups, the median GP2a level of patients until the 10th day was not significantly different from those in patients with pancreatic neoplasms, peptic ulcer, and peritonitis (p<0.05, respectively). In contrast, the corresponding GP2t level was not significantly different in patients with gastrointestinal cancer, neuroendocrine tumor, sarcoma and chronic pancreatitis (p<0.05, respectively). Furthermore, the median GP2t level in patients with liver/biliary cancer was even significantly higher.

Comparison of GP2a/GP2t assay performance

To consider the decline in GP2 levels over time, respective optimized GP2 cut-offs were obtained by ROC curve analysis employing acute pancreatitis patients till the 3rd day as positive criterion and all 752 controls as negative criterion (Figure S3). The comparison of ROC curves revealed a significantly higher area under the curve (AUC) for GP2a in contrast to GP2t (0.93 vs. 0.90, p=0.0427).

Optimal cut-off levels for GP2a and GP2t were determined at 0.7 and 2.3 ng/mL, respectively. Thus, GP2a possessed a significantly better accuracy with a sensitivity of 91.7% and a specificity of 96.7%. GP2t demonstrated the same sensitivity but a poorer specificity (92.6%). Altogether, these data resulted in positive/negative likelihood ratios of 24.6 and 0.09 for GP2a and 12.3 and 0.09 for GP2t, respectively.

Comparison of GP2a/GP2t positivity in acute pancreatitis and controls

The obtained cut-offs of 0.7 ng/mL (GP2a) and 2.3 ng/mL (GP2t) were used to define the prevalence of positives in patients and controls (Table 2). There was a decrease in the prevalence of GP2a and GP2t positives after the 3rd disease day which dropped to 44.1% and 42.4% till the 10th day, respectively.

The GP2a ELISA revealed significantly less false positives compared with the GP2t assay (28/752 vs. 56/752, p=0.0022). This was basically due to the significantly higher prevalence of positive GP2t findings in patients with liver/biliary cancer (18/118 vs. 5/118, p=0.0073) and particularly with hepatocellular carcinoma (9/40 vs. 1/40, p=0.0143).

GP2a/GP2t levels and positivity in etiological acute pancreatitis variants

GP2a and GP2t levels were significantly different in patients with varying etiology (p=0.0144, 0.0199, respectively). Patients with idiopathic and alcoholic illness demonstrated significantly elevated GP2a and GP2t levels in contrast to patients with biliary and post-trauma acute pancreatitis (p<0.05, respectively).

In contrast to GP2a, there was a significantly higher prevalence of GP2t in patients with idiopathic disease compared with biliary acute pancreatitis patients (9/21 vs. 8/54, p=0.0143).

Correlation of GP2a/GP2t with CRP and PCT

GP2a and GP2t levels were significantly correlated with PCT (ρ=0.21, 0.26, p=0.0110, 0.0012, respectively) and CRP values (ρ=0.37, 0.40, p<0.0001; respectively) in 152 and 146 follow-up samples of acute pancreatitis patients, respectively (Figure S4).

In general, GP2a and GP2t levels declined rapidly after the 3rd disease day. However, patients with severe illness demonstrated elevated GP2a levels until the 21st day with a steady decline over time (Figure 2A). CRP levels demonstrated a similar modulation corresponding with therapy success (Figure 2A–D). In two lethal cases, there was a strikingly similar behavior of GP2a and PCT levels at time points beyond the 10th day (Figure 2E,F). One patient showed a simultaneous increase and subsequent decrease of GP2a and PCT levels in response to pancreatic necrosectomy on day 34 (Figure 2E). Of note, increases in GP2a levels could be determined even later than the 50th day (Figure 2G,H). GP2t levels showed similar alterations like GP2a (data not shown).

Figure 2:

Serum glycoprotein 2 isoform alpha (GP2a), procalcitonin (PCT), and C-reactive protein (CRP) levels in follow-up samples of patients with acute pancreatitis between the 2nd and 49th disease day.

(A) Male, 47 years, alcohol abuse, severe acute pancreatitis with local complications. (B) Male, 31 years, alcohol abuse, severe acute pancreatitis with systemic complications. (C) Male, 48 years, alcohol abuse, severe acute pancreatitis with local complications. (D) Male, 31 years, alcohol abuse, severe acute pancreatitis with systemic complications. Acute pancreatitis patients with late increase of GP2a (>10th day). (E) Female, 65 years, biliary, severe acute pancreatitis with lethal outcome. (F) Female, 83 years, biliary, severe acute pancreatitis with lethal outcome. (G) Male, 59 years, alcohol abuse, severe acute pancreatitis with systemic complications. (H) Male, 26 years, alcohol abuse, severe acute pancreatitis with systemic complications.

Prediction of acute pancreatitis severity by GP2a/GP2t levels

Given the association of GP2a and GP2t levels with PCT and CRP, acute pancreatitis patients were stratified in 4 groups according to disease severity retrospectively (Figure 3).

Figure 3:

Association of serum glycoprotein 2 isoform alpha (GP2a) and total GP2 (GP2t) with severity of disease.

GP2a (A,C) and GP2t (B,D) levels were detected by enzyme-linked immunosorbent assays in acute pancreatitis (AP) patients with mild disease (I), severe disease with local complications (II), severe disease with systemic complications (III), and severe disease with lethal outcome (IV). Of note, disease duration was significantly different in the AP severity stages (p=0.0330) whereby severe cases with local complications had a significantly shorter disease duration at first presentation than severe acute pancreatitis patients with systemic disease (p<0.05). In contrast, the differences in etiology in the four severity stages of AP patients did not reach significance (p=0.0586). Optimized cut-offs for GP2a and GP2t obtained by receiver-operating characteristic curve analysis were illustrated by dashed horizontal lines. Data are displayed in Box-and-Whisker plots with far out values, defined as values that are smaller than the lower quartile minus 3 times the interquartile range, or larger than the upper quartile plus 3 times the interquartile range, displayed as solid triangles.

Remarkably, patients with lethal outcome did demonstrate a significantly higher prevalence of GP2a and GP2t compared with mild cases at admission (7/16 vs. 2/22, 6/16 vs. 2/22, p=0.0211, 0.0497; respectively). This resulted in an odds ratio of 7.8 [95% confidence interval (CI): 1.3–45.1, p=0.0222] for GP2a and 6.0 (95% CI: 1.0–35.3, p=0.0474) for GP2t positivity regarding the prognosis for a lethal outcome. When acute pancreatitis patients admitted until the 10th disease day were considered only, there was also a significantly higher GP2a positivity in lethal cases compared to mild disease (5/6 vs. 1/7, p=0.0291).

There were no significantly different GP2a levels in the four disease severity stages of the 153 patients with acute pancreatitis (p=0.1167) (Figure 3A). In contrast, GP2t levels differed significantly (p=0.0220) (Figure 3B). Thus, mild acute pancreatitis cases had significantly lower GP2t levels compared to severe illness with local complications and lethal outcome (p<0.05, respectively).

When all severe cases (n=131) were compared to patients with mild disease, GP2a and GP2t levels were significantly increased in the severe acute pancreatitis cohort (p=0.0329, 0.0042, respectively) (Figure 3C, D).