Urine Proteomics Identifies Biomarkers for Diagnosis and Fibrosis Severity in Pediatric Chronic Pancreatitis.

Chronic pancreatitis (CP) is a progressive inflammatory disorder of the pancreas that is characterized by painful end-stage sequelae, with tissue fibrosis leading to loss of endocrine and exocrine function and morphological changes on imaging. Long-term chronic inflammation from CP is a risk factor for pancreatic cancer. It has been estimated that up to a third of patients with acute pancreatitis (AP) are at risk of developing acute recurrent pancreatitis, which is a risk for CP in their lifetime ( 1 ). Most adult CP cases arise from excessive alcohol use or smoking, while CP in children is tightly linked to hereditary causes (e.g., PRSS1 and SPINK1 mutations), with genetics playing a major role in the pancreatitis course ( 2 ). Given the different risk profiles for the 2 groups, pediatric and adult CP should be studied separately ( 3 ). In all cases, however, the presence of fibrotic tissue that leads to a decrease in pancreas function is a diagnostic hallmark of CP, but tissue histology is rarely obtained from patients with suspected CP due to the need for invasive biopsy. Diagnosis of CP is challenging because current techniques include imaging that can overrepresent or underrepresent the extent of pancreatic fibrosis and disease progression, potentially leading to incorrect or delayed medical intervention ( 4 ). There remains a need to develop a diagnostic method for CP to effectively treat patients. The Ammann scoring system ( 5 ) attempts to quantify interlobular and perilobular fibrosis by grading both the extent and distribution of fibrosis within the pancreas on histology. Despite its utility, the application of histopathologic fibrosis scoring in clinical practice is limited because of the invasive nature of obtaining pancreatic tissue samples, which is rarely performed in patients with suspected CP, especially in the pediatric population. Moreover, most existing studies focus on adult CP cohorts, leaving a significant gap in our understanding of fibrosis development and severity in pediatric cases, where etiologies often differ, and genetic factors play a more prominent role. To date, no reliable circulating biomarkers have been established for assessing pancreatic fibrosis in patients with CP. Proteomics offers an advantageous methodology for identifying candidate marker proteins that could be disease biomarkers. Proteomics is a method by which proteins and peptides can be quantified using mass spectrometry (MS). Using recent developments in data-independent acquisition MS, biomarker protein identification has been shown to have improved reproducibility across large cohorts. Our prior research has identified CEL2A as a novel urine protein biomarker for the diagnosis of AP ( 6 ). Previous studies have attempted to identify novel diagnostic biomarker candidates for CP using biofluids but have shown confounding variables between sample groups or have lacked validation ( 7 ). In addition, there have been no attempts to identify markers that exclude age and lifestyle-related variables, such as alcohol use and smoking. We hypothesized that, in children, urine proteomics would differentiate patients with CP from patients with AP and controls and would reveal correlated results with tissue fibrosis.

The study had subjects enrolled in 2 study cohorts approved by the Institutional Review Board (IRB) at Cincinnati Children's Hospital Medical Center for the acute pancreatitis registry (IRB 2012–4050), and the Total Pancreatectomy Cohort Study (IRB 2016–9510). All patients provided written informed consent for participation in the study. The protocol and informed consent forms were approved by the institutional ethics committees before enrollment. The study was designed using principles in the STROBE guidelines for Strengthening the Reporting of Observational Studies in Epidemiology. This was a cross-sectional PRoBE phase 1 biomarker study ( 8 ) based on a cohort of pediatric and young adults less than 21 years of age presenting to a large tertiary freestanding Children's Hospital. Some measurements from a subsection of our cohort were previously published ( 6 ) in a study focused on identifying AP biomarkers. This work builds on that preliminary data. Our study included both male and female pediatric subjects, and findings are reported for both sexes. The cohorts identified included patients with CP, patients with AP, healthy controls (HC), or an acute extremity fracture (fracture pain controls [FRC]) injury. CP was defined as per the published International Study Group of Pediatric Pancreatitis: In Search for a CuRE (INSPPIRE) criteria and required imaging features of CP including the presence of imaging findings indicative of CP including but not limited to ductal calcifications, pancreatic ductal dilation plus one of the following: (i) pain consistent with pancreatic origin, (ii) endocrine dysfunction, and (iii) exocrine dysfunction ( 9 , 10 ). Based on this, AP was defined by meeting 2 of 3 criteria (pain, laboratory evidence by lipase or amylase of more than 3× normal levels, and imaging findings). Physicians completed surveys regarding clinical data for patients with the characteristics of AP presentations. The nonpancreatitis groups included individuals without chronic pain syndromes or significant chronic diseases. The fracture group comprised patients presenting to the emergency department within 48 hours of a painful extremity fracture confirmed by imaging. Samples were obtained from AP patients within 48 hours of presentation, and samples from patients with CP were obtained on the day of surgery for total pancreatectomy with islet cell transplantation (TPIAT). Data were stored in the Research Electronic Data Capture database (REDcap, Nashville, TN). Patient demographic data were collected at the time of admission and detailed in the Results section. Cohort biostatistical data were analyzed using SAS, version 9.4 (SAS Institute, Cary, NC). Owing to skewed distributions and/or small sample sizes, continuous variables were summarized as medians and interquartile ranges. Categorical variables were presented as counts and percentages. χ 2 or Fisher exact tests were used for group comparisons of categorical variables. For continuous data, nonparametric Kruskal-Wallis tests were used for group comparisons. Urine samples were processed by acetone precipitation and digested on suspension traps using trypsin. Plasma samples were processed using the Mag-net protocol ( 11 ) followed by on-bead digestion. Peptides were analyzed with a block randomized design using data-independent acquisition MS. The discovery cohort used a Thermo Fusion, while the validation cohort used an Exploris 480 mass spectrometer configured with a chromatogram library workflow ( 12 ) and interpreted using Prosit predicted spectra ( 13 , 14 ). Quantification used median normalization, and statistical significance was assessed using Bonferroni-corrected t -tests to control the familywise error rate, with the goal of preventing any false positives (familywise error rate < 0.05). Full protocols, including reagent compositions, instrument settings, and software parameters, are detailed in the Supplementary Methods ( http://links.lww.com/CTG/B461 ). At the time of TPIAT, pancreas tissue was collected for clinical examination per standard protocol and was sent to the pathology laboratory for review by a single pathologist for internal data consistency. Hematoxylin-eosin-stained slides were created for each case and labeled from the head, body, or tail of the pancreas. In 80% of cases, samples from 2 of these regions were available. The Ammann scoring system was followed to systematically grade fibrosis ( 5 , 15 ). Briefly, perilobular and intralobular fibrosis were graded separately. Fibrosis involving less than 50% (focal) of the pancreatic parenchyma was staged as 1 (mild), 2 (moderate), and 3 (marked). Fibrosis in more than 50% (diffuse) of the pancreatic parenchyma was staged as 4 (mild), 5 (moderate), and 6 (marked). The final fibrosis score was the sum of 2 components, perilobular (scored from 1 to 6) and intralobular (scored from 1 to 6), and was further categorized into 3 groups: 0–4 (low), 5–8 (medium), and 9–12 (high). Examples of these scores are illustrated in Figure 5 a. Quantitative values from 46 CP samples were used to develop a linear model to predict Ammann scores. Twelve CP-specific proteins were selected based on minimum redundancy, and 10,000 multiple linear regression models were built using bootstrapping. Coefficients from these models were averaged to generate a single predictive model. This process was then repeated 10,000 times using 12 randomly selected proteins to determine a background correlation distribution. Data, analytic methods, and study materials will be made available to other researchers on request. Raw proteomics data are publicly available on MassIVE using data set identifier MSV000096607 (reviewer password: “pancreas”) and at ProteomeXchange using the data set identifier PXD058660. Raw proteomics data file annotations are listed in Supplementary Data 1 ( http://links.lww.com/CTG/B461 ).

The discovery cohort included 130 individuals, 28 with AP, 50 with CP, and 52 controls (21 HC and 31 FRC), and included some previously published proteomics results ( 6 ), where previous measurements were expanded on with fibrosis data. The validation cohort included an additional 36 individuals, 10 with AP, 15 with CP, and 11 HCs, measured 2 years later with an alternative proteomics platform. The sample sizes for the discovery cohort were chosen based on power analysis from a 6-sample test data set, while the sizes for the validation cohort were based on a power analysis of the discovery cohort data set. Baseline characteristics for both cohorts of the groups are presented in Table 1 . Of the validation cohort, we also had time-matched plasma samples from 5 AP, 14 CP, and 4 HC. Of the 65 total individuals with CP, 52 had pancreas tissues available for fibrosis scoring measurements. Clinical characteristics were included for the 3 fibrosis groupings: low, medium, and high (Table 2 ). We processed samples from these subjects according to the experimental outline in Figure 1 . Baseline participant characteristics of the discovery and validation cohorts Data presented as median (25th–75th percentile) or n (%). AP, acute pancreatitis; BMI, body mass index; CP, chronic pancreatitis; HC, healthy control. a Controls (31 extremity fracture patients, no pancreatitis history, and 21 HCs in the discovery cohort and 11 HCs in the validation cohort). Complete data unless otherwise specified by n = or stated denominator. In addition, 21 of 36 patients with CP had plasma and urine samples available from the same subject (14 CP, 3 AP, 4 HC). Clinical features of patients with CP with available fibrosis histology scores Data presented as median (25th–75th percentile) or n (%). Complete data unless otherwise specified by stated denominator. Experimental workflow. ( a ) The discovery cohort urine samples underwent a bottom-up proteomics sample preparation before mass spectrometry analysis for global protein quantification. ( b ) The validation cohort plasma and urine samples were processed, and significance was determined based on previous CP biomarker findings for blood studies. An example receiver operator curve plot for the best urine (S100P) and plasma (IL2RA) markers is shown. ( c ) CP patient samples who underwent TPIAT provided tissue samples, and these samples were given an Ammann score based on histology results. ( d ) Finally, the proteomics data were combined with the Ammann scores, and machine learning was used to generate a linear regression model between the 2 variables. Created in BioRender. AP, acute pancreatitis; CP, chronic pancreatitis; HC, healthy controls; FRC, fracture pain controls; IL2RA, interleukin 2 receptor subunit alpha; LC-MS/MS, liquid chromatography-tandem mass spectrometry; S100P, Protein S100-P; TPIAT, total pancreatectomy with islet cell transplantation. We detected a total of 2,137 library proteins with global urine proteomics of the discovery cohort, representing >95% of the observable human urine proteome ( 16 ). Of these proteins, we identified 1,760 across all 130 samples, where 34 had 2× fold change in pairwise comparisons with every control group (AP, HC, and FRC). This approach was intentionally stringent to eliminate nonreplicating expression changes. When samples from individuals with CP were compared with samples from controls, we found 4 proteins that were significantly upregulated above others consistently: isocitrate dehydrogenase (IDH1), calcyphosin (CAPS), synuclein gamma (SNCG), and protein S100-P (S100P) (Figure 2 and Supplemental Figure 1, http://links.lww.com/CTG/B461 ). In the validation cohort with analysis performed in a different laboratory, we similarly detected 2,010 proteins across 36 samples but could only quantify 29 of the 34 previously identified. We found that all of these proteins remained significantly different (Figure 3 a) and that the cohorts tracked together in relative expression levels (Figure 3 b). Volcano plots showing significant (familywise error rate <0.05) proteins that were either downregulated or upregulated in the urine discovery cohort (light blue) relative to nonsignificant proteins (gray). A total of 34 proteins were significantly upregulated within pairwise tests of CP to all individual groups (dark blue) relative to AP ( a ), HC ( b ), and FRC ( c ). The 4 consistently most significant proteins are labeled (IDH1, S100P, SNCG, and CAPS). AP, acute pancreatitis; CAPS, calcyphosin; CP, chronic pancreatitis; HC, healthy controls; FRC, fracture pain controls; FDR, false discovery rate; IDH1, isocitrate dehydrogenase; S100P, protein S100-P; SNCG, synuclein gamma. ( a ) Urine proteome analysis box plots showing the delta of log 2 normalized intensities of CP samples and control sample median for each of the 34 significantly upregulated proteins (FC > 1, familywise error rate <0.05) in the discovery cohort. *Existing literature indicates significantly high expression in pancreatic ductal adenocarcinoma compared with controls. ( b ) Regression showing the median delta of the log 2 normalized intensities of CP and control sample median for the validated proteins found in both the urine discovery and validation cohorts. Control groups include AP, HC, and FRC in the discovery cohort and AP and HC in the validation cohort. AP, acute pancreatitis; CP, chronic pancreatitis; FC, fold change; HC, healthy controls; FRC, fracture pain controls. Attempts were made to limit common CP comorbidities when possible. The incidence rate of diabetes in the CP cohort was low enough to not be considered as a confounding factor (discovery: 1/50, validation: 3/15). Similarly, efforts were made to limit the effect of recurrent AP (recurrent pancreatitis) on the analysis of AP as a control group (discovery: 0/28, validation: 3/10). Considering t-distributed stochastic neighbor embedding (t-SNE) clustering of urinary protein expression in the 34 significant proteins from the discovery cohort, CP clearly differentiated from the other groups (Figure 4 a). Age was a significant differentiator ( P = 0.0006) of CP, specifically in the female fraction of the cohort, where nearly all female patients in the discovery cohort were postpuberty ( P = 0.0032), whereas this was not true for male patients (Figure 4 b). For this analysis, puberty was assumed to be ≥10 yo for females ( 17 ) and ≥11 yo for males ( 18 ). Further work is needed to explore how menstrual cycles are associated with pancreatitis. Of the 50 CP individuals in the discovery cohort, fecal elastase was reported for 29. Of these, 5 were determined to have exocrine pancreatic insufficiency (EPI) based on elastase levels below 100 µg/g stool. Although these individuals seem to have more extreme molecular phenotypes as shown by increased within-cluster distances (Figure 4 c), the low number of EPI cases did not allow subgroup analysis in a meaningful fashion. We observed no significant bias based on obesity (Figure 4 d, overweight: Z -score>1, obese: Z -score>2), reported pain (Figure 4 e, 0–10), or hospitalization rate (Figure 4 f, frequent rate defined as hospitalized >3 times in 18 months). Like others ( 19 ), we found that the fecal elastase level did not correlate with either scored pain or number of hospitalizations. Protein expression clustering using t-SNE for the 34 significantly upregulated proteins based on condition ( a ), age and sex ( b ), elastase levels ( c ), obesity ( d ), pain ( e ), and hospitalization rate ( f ). Puberty was assumed to be ≥ 10 yo for female patients and ≥11 yo for male patients. Individuals with fecal elastase levels below 100 µg/g were considered to have exocrine pancreatic insufficiency. Individuals with body mass index Z -scores >1 were considered overweight (84th percentile), while individuals with Z -scores >2 were considered obese (98th percentile). Pain at the time of sample collection was self-reported on a scale of 0–10. Individuals with frequent hospitalization rates were admitted more than 3 times in an 18-month period. Pain and hospitalization rate data are only available for patients with CP. AP, acute pancreatitis; CP, chronic pancreatitis; HC, healthy controls; FRC, fracture pain controls; t-SNE, t-distributed stochastic neighbor embedding. We measured 21 paired urine and plasma samples from the validation cohort with global proteomics (14 CP, 3 AP, and 4 HC) to compare newly discovered urine markers with existing blood-based markers. For this, we measured 19 proteins previously studied in blood for CP diagnostic ability: interleukin 2 receptor subunit alpha (IL2RA) ( 20 ), MMP7 ( 21 ), ICAM1 ( 22 ), GDF15 ( 23 ), LGALS3BP ( 24 ), HP ( 25 ), PON1 ( 26 ), APOA2 ( 22 ), MMP9 ( 27 ), CX3CL1 ( 28 ), APOC1 ( 29 ), LTF ( 22 ), RETN ( 30 ), THBS1 ( 22 ), RARRES2 ( 31 ), AZGP1 ( 22 ), CRP ( 27 ), ADIPOQ ( 32 ), and PPBP ( 22 ). Although some of these proteins were previously studied only in serum, many were first measured in plasma. In our work, we found that 10 of 19 proteins produced higher expression levels in CP than the median control expression, where no proteins had higher than a tenfold median increase (Figure 5 a). By contrast, 27 urine proteins had CP median expression values above the median control expression, where SNCG had >25× median fold change (Figure 5 b). We generated receiver operator curves (ROCs) for each protein in their respective biofluid and found that 25 urine proteins had a larger area under the curve (AUC) than all the previously studied blood proteins (Figure 5 c). Error bars use 95% confidence intervals consistent with the Hanley and McNeal method ( 33 ). Consistent with previous data, we found that S100P, IDH1, CAPS, and SNCG were top-performing diagnostic markers, and all had ROC-AUC values >0.95. The best performing plasma marker was IL2RA (ROC-AUC = 0.80). Surprisingly, although we were able to measure these proteins in plasma, we found that they performed poorly when measured in urine (Supplemental Figure 2, http://links.lww.com/CTG/B461 ). Box plots showing the delta of log 2 normalized intensities of CP samples and control sample median for ( a ) each of the 19 proteins previously studied as CP biomarkers in serum, plasma, or whole blood found in the paired validation cohort plasma samples and ( b ) urine protein biomarker candidates found in the paired validation cohort urine samples. ( c ) The ROC AUC values for each of these proteins, where yellow denotes the novel urine proteins and red denotes the previously studied blood proteins. Error bars indicate a 95% confidence interval. The top performing proteins in urine were S100P (AUC: 0.994), HPGD (AUC: 0.979), SNCG (AUC: 0.974), CAPS (0.954), and IDH1 (0.954). AUC, area under the curve; CAPS, calcyphosin; CP, chronic pancreatitis; IDH1, isocitrate dehydrogenase; ROC, receiver operator curve; S100P, protein S100-P; SNCG, synuclein gamma. We obtained 46 CP pancreas tissue samples from patients in the discovery cohort and 6 from patients in the validation cohort. Examples of the histology for varying degrees of fibrosis can be seen in Figure 6 a. Since the validation cohort contained only a small range of scored samples, we considered the combined cohorts as a single fibrosis data set of 52 individuals. We built a linear regression model for predicting fibrosis using bootstrapping from 12 proteins with minimal redundancy (S100A6, GSR, GPD1L, IDH1, CAPS, SNCG, HPGD, HNMT, CRYM, GMPR2, PCBD1, and ATIC). More than 10,000 samples, these 12 proteins performed significantly better than linear regression models generated from 12 randomly selected proteins (Figure 6 b). We found that most of the 10,000 bootstrap models were positively correlated with fibrosis, with an average Pearson correlation coefficient >0.6. The aggregate linear model predicted Ammann fibrosis scores with R 2 = 0.6131 (Figure 6 c). ( a ) Representative photomicrographs of trichrome-stained pancreatic tissue samples. Blue-colored areas indicate fibrosis in CP samples of low, medium, and high fibrosis score groups (4× magnification). ( b ) Correlation distributions of predicted fibrosis scores based on the 12 significant CP proteins determined to correlate with fibrosis (blue) and 12 randomly selected proteins (red). ( c ) Regression model showing the prediction scores of proteins as they predict the fibrosis score in CP. CP, chronic pancreatitis.

Our study took a novel approach to studying CP in children, where genetic risk factors are more prevalent, by performing proteomics analysis of minimally invasive biofluids (urine and blood) in our cohort. These experiments derived potential biomarkers that may help differentiate patients with CP from patients with AP and patients without pancreatitis (healthy subjects and extremity fracture conditions). Widely accepted biomarkers for CP do not currently exist because it has been historically difficult to separate CP-specific proteins from proteins altered by other confounding factors, such as alcohol use and smoking, as well as from other disease states. The fibrosis scoring data as a measure of tissue destruction in the CP cohort add to the novelty of our findings and show that a panel of proteins performs well as a disease diagnostic biomarker and a measure of fibrosis in pancreatic tissues. We confirmed previously known proteins in CP and discovered a new data set in childhood CP, with top-performing proteins being S100P, IDH1, CAPS, and SNCG. Urine is an attractive, noninvasive biomarker source that reflects protein changes caused by disease state ( 34 ). Although urine can be challenging for protein-level measurements because of protein concentration variability, our study used precipitation followed by protein concentration normalization to ensure equal protein in each sample regardless of the original biological levels. Moreover, to eliminate effect of variations in kidney excretion, none of our CP subjects had a diagnosis for an underlying kidney disease, and the urine samples were provided after an overnight fasting state. Previous studies attempted to characterize the urine proteome of adults with CP but were limited by a lack of validation ( 35 ) or sample size ( 35 , 36 ). Previously, we reported robust urine marker proteins for AP in a pediatric sample population ( 6 ). In this work, we reinvestigated this data set for CP markers and identified 34 potential candidates. Of these, 7 proteins have previously been shown as upregulated in CP through adult cohort studies using fluids and tissue (PCBD1 ( 37 ), HNMT ( 37 ), MAT2A ( 37 ), SCRN1 ( 37 ), CD151 ( 38 ), HSP90AA1 ( 39 ), and PARK7 ( 40 )). Gene Ontology (GO) protein set enrichment revealed that 16 of 34 significantly changing proteins were involved in cation binding (FDR = 5.3E-03) and 12 were involved in oxidoreductase activity (FDR: 4.3E-06). Interestingly, 14 of these proteins are highly expressed in pancreatic ductal adenocarcinoma (PDAC) subjects compared with controls in fluids and tissue (GPD1L ( 41 ), CD151 ( 38 ), GSR ( 22 ), HSP90AA1 ( 39 ), IDH1 ( 42 ), LGALS3 ( 43 ), MAT2A ( 37 ), PARK7 ( 40 ), PTGR1 ( 44 ), S100P ( 45 ), S100A6 ( 45 ), SCRN1 ( 46 ), SRI ( 47 ), and SNCG ( 48 )), underlining a possible link to PDAC in individuals with hereditary pancreatitis ( 49 , 50 ) or CP to PDAC ( 51 ). Since most proposed diagnostic protein markers for CP have been reported in blood ( 7 ), we measured a set of paired urine and plasma samples for 21 individuals collected in the same visit. In addition to the 34 candidate urine markers, we measured 19 previously reported plasma or serum CP protein markers: adipokines/chemokines/cytokines, glycoproteins/lipoproteins, and other potential protein biomarkers. Unsurprisingly, markers discovered in blood had poor performance in urine; the reverse was also true. Of the plasma proteins we targeted, soluble IL-2 receptor ( 20 ) (IL2RA) was the top-performing protein with an AUC of 0.801 to predict CP. By contrast, the best urine candidate, S100P, produced a nearly perfect AUC of 0.995 to predict CP. In this work, we attempted to associate quantitative protein abundances with histology scoring as a proxy for disease progression. Although no individual protein successfully predicted the severity of fibrosis in CP, a panel of multiple proteins may be useful for classifying patients. In addition, proteomics could be used as a preassessment to screen candidates for fibrosis, which currently requires invasive biopsies to fully assess the progression of CP. By using histology scoring on pancreas tissues, our quantitative proteomics results can be directly compared with disease progression. Compared with other studies, our population had a higher fibrosis score than another published study ( 52 ), which could be attributed to the nature of the referral center as a TPIAT center and the frequency of genetic mutations in children leading to higher fibrosis scores. Methods for measuring pancreas fibrosis previously included endoscopic ultrasound with elastography by measuring tissue stiffness ( 53 ). However, this technique has challenges in differentiating fibrosis from inflammation and neoplasia, operator dependency, and accessibility considerations. Our studies show promise for circulating protein measures and pancreas fibrosis grading. Our study has the limitations of being a single-center study with a small cohort, albeit relatively large for a pediatric study. One weakness is that at the time of cohort design, sex and age were not considered to have significant associations. Only after finalizing the full cohort, we observed that female patients exhibiting pancreatitis in our study were significantly older than male patients or other control populations (Table 1 ). In addition, we do not have longitudinal data to assess progression from early to late CP, and the range of fibrosis histology is biased toward severe fibrosis. In particular, since fibrosis histology was only performed immediately before a scheduled TPIAT surgery, our cohort contained a very few individuals graded for low fibrosis. These weaknesses are offset by having distinct derivation and validation cohorts and same-day-matched urine, plasma, and histology scoring data sets. However, owing to limited independent validation of fibrosis data availability, future validation of fibrosis prediction using urine biomarkers is necessary to assess reproducibility. Some proteins are either difficult or impossible to measure with global proteomics, either due to having low abundance relative to the sample dynamic range or due to incompatibilities with sample preparation methods. Human trypsinogen (PRSS1), for example, is challenging to measure because porcine trypsin is used to digest proteins before MS analysis. In conclusion, we have found several protein markers that perform well for distinguishing patients with CP from patients with AP and HCs in a pediatric cohort. Furthermore, we have found a set of proteomic markers in the urine that correlate with fibrosis scoring on pancreas tissues, suggesting potential for building future diagnostic criteria. These results may help unravel the underlying mechanisms of CP, present potential therapeutic targets, and identify insights into the linkage of pancreatitis to PDAC. These results underline a need for future multicenter and larger cohort studies to deeply characterize CP-specific proteins in pediatric and adult cohorts.

Guarantor of the article: Brian C. Searle, PhD, and Maisam Abu-El-Haija, MD, MS. Specific author contributions: M.G.M., B.C.S., M.A.-E.-H.: writing, first draft. All authors: writing, editing, and approval of the final submitted version. L.B., J.G., M.A.-E.-H.: sample collection and supervision of the cohorts. M.G.M., V.G.: sample preparation. V.G., M.S., R.S., A.T.: clinical data collection. M.G.M., L.Z., B.C.S.: molecular data collection. M.G.M., K.B.B., B.C.S., L.H.: data analysis. B.C.S., M.A.-E.-H.: supervision and funding. Financial support: This study was funded through resources from the National Institutes of Health grants K23 DK118190 (MA), R03 DK131156 (MA), R35 GM150723 (BCS), and S10 OD018056 (OSU Campus Chemical Instrument Center). Potential competing interests: This author discloses the following: Brian C. Searle is a founder and shareholder in Proteome Software, which operates in the field of proteomics. The remaining authors disclose no conflicts. Study Highlights WHAT IS KNOWN ✓ Chronic pancreatitis is an inflammatory condition often diagnosed with imaging, which is unreliable for early detection. ✓ No validated noninvasive biomarkers exist for diagnosing chronic pancreatitis or monitoring fibrosis, especially in children. WHAT IS NEW HERE ✓ This study identifies specific urine proteins that outperform existing blood-based markers in diagnosing. ✓ A subset of these urine biomarkers tracks with fibrosis severity, indicating their predictive power. ✓ By developing a multiprotein urine panel, we provide a measurable, noninvasive tool for tracking disease progression. ✓ These findings could guide new diagnostic protocols that improve early intervention, reducing reliance on invasive procedures. ✓ Chronic pancreatitis is an inflammatory condition often diagnosed with imaging, which is unreliable for early detection. ✓ No validated noninvasive biomarkers exist for diagnosing chronic pancreatitis or monitoring fibrosis, especially in children. ✓ Chronic pancreatitis is an inflammatory condition often diagnosed with imaging, which is unreliable for early detection. ✓ No validated noninvasive biomarkers exist for diagnosing chronic pancreatitis or monitoring fibrosis, especially in children. ✓ This study identifies specific urine proteins that outperform existing blood-based markers in diagnosing. ✓ A subset of these urine biomarkers tracks with fibrosis severity, indicating their predictive power. ✓ By developing a multiprotein urine panel, we provide a measurable, noninvasive tool for tracking disease progression. ✓ These findings could guide new diagnostic protocols that improve early intervention, reducing reliance on invasive procedures. ✓ This study identifies specific urine proteins that outperform existing blood-based markers in diagnosing. ✓ A subset of these urine biomarkers tracks with fibrosis severity, indicating their predictive power. ✓ By developing a multiprotein urine panel, we provide a measurable, noninvasive tool for tracking disease progression. ✓ These findings could guide new diagnostic protocols that improve early intervention, reducing reliance on invasive procedures.