Pancreas. 2013 May;42(4):622-32.

Usefulness of MALDI-TOF/MS identification of low-MW fragments in sera for the differential diagnosis of pancreatic cancer.

Padoan A, Seraglia R, Basso D, Fogar P, Sperti C, Moz S, Greco E, Marchet A, de Manzoni G, Zambon CF, Navaglia F, Cristadoro L, Di Chiara A, Nitti D, Pedrazzoli S, Pavanello G, Plebani M.

Department of Laboratory Medicine, University of Padova, Italy.

 

Abstract

OBJECTIVES: To identify new biomarkers of pancreatic cancer (PaCa), we performed MALDI-TOF/MS analysis of sera from 22 controls, 51 PaCa, 37 chronic pancreatitis, 24 type II diabetes mellitus (DM), 29 gastric cancer (GC), and 24 chronic gastritis (CG).

METHODS: Sera were purified by Sep-Pak C18 before MALDI-TOF/MS Anchorchip analysis.

RESULTS: Features present in at least 5% of all spectra were selected (n = 160, m/z range, 1200-5000). At univariate analysis, 2 features (m/z 2049 and 2305) correlated with PaCa, 3 (m/z 1449, 1605, and 2006) with DM. No feature characterized gastric cancer or chronic gastritis. Ten-fold cross-validation binary recursive partitioning trees were obtained for patients’ classification. The tree (CA 19-9, age, m/z 2006, 2599, 2753, and 4997), built considering only patients with diabetes, allowed a distinction between DM [area under the receiver operating characteristic curve (AUC), 0.997], chronic pancreatitis (AUC, 0.968), and PaCa (AUC, 0.980), with an overall correct classification rate of 89%. The tree including CA 19-9, 1550, and 2937 m/z features, achieved an AUC of 0.970 in distinguishing localized from advanced PaCa. MALDI-TOF-TOF analysis revealed the 1550 feature as a fragment of Apo-A1, which was determined as whole protein and demonstrated to be closely correlated with PaCa.

CONCLUSIONS: The findings made demonstrate a role for serum peptides identified using MALDI-TOF/MS for addressing PaCa diagnosis.

PMID: 23271396

 

SUMMARY

Pancreatic cancer (PDAC) is the fourth cause of cancer-related deaths in industrialized countries. PDAC late diagnosis, which limits curative options, may depend on several factors, those mainly involved being: 1) no accepted screening strategy for sporadic PDAC; 2) organ confined tumours (stage I and some stage II) are often asymptomatic; 3) the only biomarker helpful forPDAC diagnosis, CA 19-9, has limitations in both sensitivity and specificity,  which do not exceed  79% and 82% respectively.

Diabetes mellitus (DM) or a reduced glucose tolerance are frequently found in PDAC. When long-standing,  DM is reported to increase the risk for PDAC while, when of new-onset, DM may be considered a consequence of this tumour. This latter observation have prompted some investigators to suggest that subjects with new-onset DM should be considered “at risk” and so they should undergone PDAC screening (1). However, to do this new biomarkers are advisable to enhance early PDAC detection rate.

Recent evidences have shown that sera may contain small proteins/peptides or proteolytic fragments, which may be a promising source for the discovery of new biomarkers. The bulk of these small peptides, also known as peptidome, could be easily analysed by matrix assisted laser desorption ionization (MALDI) time of flight (TOF) mass spectrometry (MS), which is an high resolution, high throughput technique that allows the detection of hundreds of serum/plasma component in a single run.

In this study the MALDI-TOF/MS peptidome from 51 PDAC patients was compared with that of 46 controls (HC), but also with those of 29 patients with gastric cancer (GC), 37 with chronic pancreatitis (ChrPa) and 24 with type II DM patients (DM). These latter groups were included to increase detectability of PDAC-associated biomarkers, avoiding the discovery of peptides only suggestive of the presence of inflammation and/or DM.

Sera from PDAC or GC were enriched of peptides with respect to HC, ChrPa and DM, underlying the complexity of the cancer peptidome. A total of 160 peaks/features  with a molecular mass from 1200 to 5000 m/z were identified. Since the purpose of this study was to identify new peptidome signatures able to distinguish the studied groups, we performed several statistical analyses based on the MALDI-TOF/MS identified features. In the first step a single-marker strategy was adopted and differences in features’ presence or absence among groups were ascertained without taking into account their relative abundances. Among the 43 features differently expressed among groups, 17 were closely correlated with ChrPa and PDAC, being 11 positively and 6 negatively associated with pancreatic diseases. No specific feature was correlated with GC. In the second step, aimed to identify a classification panel, a multi-marker strategy was adopted, classification and regression trees being performed. This method recursively divides patients on the basis of positive or negative findings of a series of features, which cut-off are chosen maximizing classification results. The algorithm outputs are graphical trees, which contain branches and leafs. However, as this method is prone to overestimate the classification performances, we choose to validate results by the resubstitution and the leave-one out cross-validation methods.

Classification trees were run using as predictors not only the MALD-TOF/MS features but also CA 19-9, age and sex.

The first tree, obtained by considering all patients, included CA 19-9, age and the m/z 1865 and 3182 MALDI-TOF/MS features. With this tree we obtained a limited classification of PDAC and DM only (80% and 79% of cases), while it did not satisfactorily allowed the classification of the other groups (HC, 55%; GC, 52%; ChrPa,  54%). We focused then only on patients with diabetes. The obtained tree included CA 19-9, age and the features at m/z 2006, 2599, 2753 and 4997. The areas under the ROC curves of this classification tree were 0.968 for ChrPa, 0.997 for DM and 0.980 for PDAC. The overall correct classification rate was 89% and 65% by the resubstitution and the leave-one out cross validation, respectively.  These results supported the high performance rates of ROC curves and allow to propose this tree for the distinction of type II DM patients from those diabetic patients with PDAC or ChrPa. The last tree was constructed considering PDAC patients only, using tumour stage as dichotomous outcome: localized cancer (stages IA-IIB) vs advanced cancer (stages III-IV). The overall currect classification were 85% and 68% by the resubstitution and the leave one out cross-validation.

In a further analysis, we characterized by MALDI-TOF-TOF/MS four of the previously identified features (m/z 1530, 1550, 1778 and 2753). They corresponded to Clusterin, Apo-A1, complement C3 and albumin. Apo-A1 and C3 were measured in 172 sera (43 HC, 50 PDAC, 26 ChrPa, 24 DM and 29 GC) and in 69 new serum samples (8 HC, 31 PDAC, 27 DM and 3 GC). Apo-A1 values were stratified for age and sex. Low Apo-A1 were most frequently found in PDAC (43%) and GC (31%) than in HC (0%), ChrPa (23%) and DM (10%), in line with finding from Felix et al. (2).

Multivariate logistic regression analyses, including age, CA 19-9, Apo-A1 and C3 as predictors and PDAC as outcome, demonstrated that combined Apo-A1 and C3 are better predictors than CA 19-9 alone. Finally, focusing on diabetic patients only (57 PDAC, 17 ChrPa and 51 DM), we performed two multivariate logistic regressions using PDAC as outcome. The first included age and CA 19-9, while the second included age, CA 19-9, Apo-A1 and C3. The second model classified patients better than the first, being the areas under ROC 0.91 and 0.85, respectively.

In conclusion, bioinformatics applied to MALDI-TOF/MS profiling of sera allowed us to identify two possible candidate biomarkers (Apo-A1 and C3) to be used in addition to CA 19-9 for the identification of PDAC among patients with diabetes mellitus, albeit validation studied in large cohort of patients, focused mainly on those with new onset diabetes mellitus, are required.

 

References

  1. Pannala R, Basu A, Petersen GM, Chari ST. New-onset diabetes: a potential clue to the early diagnosis of pancreatic cancer. Lancet Oncol. 2009;10:88-95.
  2. Felix K, Hauck O, Fritz S, Hinz U, Schnölzer M, Kempf T, Warnken U, Michel A, Pawlita M, Werner J. Serum protein signatures differentiating autoimmune pancreatitis versus pancreatic cancer. PLoS One. 2013;8(12):e82755.

Figure 1

Multiselect Ultimate Query Plugin by InoPlugs Web Design Vienna | Webdesign Wien and Juwelier SchönmannMultiselect Ultimate Query Plugin by InoPlugs Web Design Vienna | Webdesign Wien and Juwelier Schönmann