Preoperative GNAS and KRAS Testing in the Diagnosis of Pancreatic Mucinous Cysts

With increased use and improvements in cross-sectional imaging technologies, cystic lesions of the pancreas are becoming frequently encountered in clinical practice (1). Although many cysts, such as pseudocysts and SCA, have a benign clinical course, others, such as intraductal papillary mucinous neoplasms (IPMN) and mucinous cystic neoplasms (MCN), represent precursor lesions to invasive pancreatic adenocarcinoma (2–4). Because of their risk of malignancy, international consensus guidelines for the management of IPMNs and MCNs were developed and recently updated (5, 6). These guidelines were based on the assumption that mucinous neoplasms can be diagnosed correctly on the basis of standard clinical, imaging, and laboratory criteria. Despite these measures, preoperatively distinguishing pancreatic cysts from one another can be challenging and, if incorrect, can pose a significant health risk to the patient. However, changes within the management guidelines recommend cyst fluid molecular analysis in those centers with expertise in endoscopic ultrasound–fine-needle aspiration (EUS–FNA) and cytologic interpretation (6).

Previously, we reported the results of our institution's clinical experience with DNA testing for KRAS mutations from pancreatic cyst fluid obtained by EUS–FNA (7). KRAS mutations had a specificity of 100% but a sensitivity of 54% for mucinous differentiation. In comparison, carcinoembryonic antigen (CEA), a cyst fluid tumor marker often used to distinguish mucinous from nonmucinous pancreatic cysts, had a specificity and sensitivity of 85% and 62%, respectively, with a cutoff of 192 ng/mL and above. The combination of both assays increased the sensitivity to 83% and maintained a high specificity of 85%. Although KRAS, in conjunction with CEA analysis, improved the diagnostic yield of pancreatic cyst fine-needle aspirates, additional markers are needed to improve the sensitivity of cyst fluid DNA analysis.

Next-generation and whole-exome sequencing technologies have uncovered recurrent mutations within the major neoplastic cysts of the pancreas (8–10). Consistent with our findings, Wu and colleagues (9) identified a high prevalence of KRAS mutations in IPMNs. Rather unexpectedly, the authors also found activating mutations within the oncogene, GNAS. GNAS mutations were highly specific for IPMNs and not seen in the other pancreatic cystic neoplasms (8). Furthermore, the presence of either a GNAS or KRAS mutation was identified in >96% of IPMNs. However, these studies were based on DNA obtained from postoperative pancreatic cyst fluid or microdissection of the cyst epithelial lining. Thus, the diagnostic utility of testing for GNAS and KRAS mutations in clinical practice remains unclear. We therefore have analyzed both GNAS and KRAS mutations within a large cohort of pancreatic cyst fluid obtained by EUS–FNA. In addition, these results were correlated with corresponding clinicopathologic features and other diagnostic modalities in the diagnosis of pancreatic cystic neoplasms.

Study approval was obtained from the University of Pittsburgh Institutional Review Board. Pancreatic cyst fluids tested obtained by EUS–FNA were consecutively accrued for routine KRAS mutational analysis between January 2006 and December 2013 at the University of Pittsburgh Medical Center (Pittsburgh, PA; ref. 7). In all cases, cyst fluid was submitted for molecular analysis by the endoscopist because of uncertainty as to whether a pancreatic cyst represented a cystic neoplasm. In other words, patients with pseudocysts who had documented history of abdominal trauma or high clinical suspicion for main duct IPMN by endoscopy and radiographic imaging were often refrained from ancillary molecular testing. Eighty-three pancreatic cyst fluid DNA specimens were selected for GNAS mutational analysis. These cases included 41 IPMNs, 9 IPMNs with an associated invasive adenocarcinoma, 16 MCNs, 10 cystic pancreatic neuroendocrine tumors (PanNET), 3 retention cysts, 2 pseudocysts, 1 lymphoepithelial cyst, and 1 SCA. The diagnoses for all pancreatic cysts were rendered on the basis of standard histomorphologic criteria of corresponding surgical resection specimens (11, 12). Of note, the results of KRAS testing from 29 IPMNs, 7 IPMNs with an associated invasive adenocarcinoma, 13 MCNs, and 7 cystic PanNETs were previously reported by Nikiforova and colleagues (7). An additional eight pancreatic cyst fluid specimens were selected that clinically corresponded to SCA based on standard radiographic imaging, endoscopic ultrasound, CEA level of <2.0 ng/mL, and cytopathologic findings (11–13). Furthermore, at the time of KRAS testing, the specimens were assessed for loss of heterozygosity (LOH) at the VHL locus. In six of eight (75%) cases, VHL LOH was identified.

Medical records were reviewed for each case to document patient demographics, radiographic imaging, endoscopic ultrasound findings, fluid viscosity (qualitative assessment noted by the endoscopist at the time of fine-needle aspiration), CEA analysis, and cytopathologic diagnosis. For cytology specimens, specimen adequacy was assessed in all cases using a three-tiered system: satisfactory, less than optimal, and unsatisfactory. Satisfactory was defined as the presence of sufficient epithelial cells and or mucin representative of the target cyst. Less than optimal consisted of scant epithelium in the absence of mucin, but with at least a few histiocytes present. Unsatisfactory specimens were virtually acellular and lacked mucin. A cytopathologic diagnosis of at least suspicious for a neoplastic mucinous cyst was made by individual cytopathologists at the University of Pittsburgh Medical Center based on the presence of extracellular mucin and/or mucinous epithelium with varying degrees of cytologic and architectural nuclear atypia (14–16). Clinical impression, radiographic imaging, endoscopic findings, and CEA levels were available to the cytopathologist before a diagnosis was rendered.

KRAS testing was performed at the time of pancreatic cyst fluid submission as described by Nikiforova and colleagues (7). Briefly, total genomic DNA was isolated from 200 μL of cyst fluid by column separation and according to the manufacturer's directions and instructions (Qiagen). The quantity of isolated DNA was assessed using a NanoDrop 1000 spectrophotometer (Thermo Scientific). For the detection of mutations, 10 to 50 ng of DNA was amplified with primers flanking exon 2 of the KRAS gene (forward primer 5′-GGTGAGTTTGTATTAAAAGGTACTGG-3′ and reverse primer 5′-TCCTGCACCAGTAATATGCA-3′) and exon 3 of the KRAS gene (forward primer 5′-TGAAGTAAAAGGTGCACTG-3′ and reverse primer 5′-GCATGGCATTAGCAAAGACTC-3′). The detection of GNAS mutations were performed using primers flanking codon 201 at exon 8 (forward primer 5′-TGACTATGTGCCGAGCGA-3′ and reverse primer 5′-AACCATGATCTCTGTTATATAA-3′). PCR products were sequenced in both sense and antisense directions using BigDye Terminator v3.1 cycle sequencing kit on the ABI 3730 (Applied Biosystems) according to the manufacturer's instructions. The sequence electropherograms were analyzed using Mutation Surveyor software (SoftGenetics, LLC).

At the time of KRAS mutational analysis, determination of LOH at the von Hippel-Lindau tumor-suppressor gene, VHL, was performed as previously described (17–19). Ten to 50 ng of DNA was amplified with PCR primers based on the use of 2 microsatellite markers: 3p25:D3S2303[L17972] and 3p26:D3S1539[L16393]. Postamplification products were electrophoresed and relative fluorescence was determined for individual alleles (ABI 3100 Genetic Analyzer; Applied Biosystems). The ratio of peaks was calculated by dividing the value for the shorter sized allele by that of the longer sized allele. Thresholds for significant allelic imbalance were determined using normal (non-neoplastic) buccal swabs specimens from each patient. Peak height ratios falling outside of 2 standard deviations beyond the mean for each polymorphic allele pairing were assessed as showing significant allelic imbalance.

Statistical analyses to assess differences between GNAS- and/or KRAS-mutant and wild-type cysts were compared using the Fisher exact test for dichotomous variables using SPSS Statistical software, version 21 (IBM). All tests were two-sided, and statistical significance was defined as a P < 0.05.

The study cohort consisted of the following: 50 IPMNs, 16 MCNs, 10 cystic PanNETs, 9 SCA, 3 retention cysts, 2 pseudocysts, and 1 lymphoepithelial cyst. In 9 of 50 IPMNs, an associated invasive pancreatic ductal adenocarcinoma was present. All pancreatic cysts except eight of nine SCA were surgically resected and diagnosed on the basis of standard pathologic criteria (12). The remaining eight cysts were diagnosed clinically using accepted radiographic and endoscopic ultrasound findings, a CEA level of <2.0 ng/mL, and cytologic smears consistent with a SCA (13). Furthermore, at the time of KRAS testing, these eight SCA specimens were assessed for LOH at the VHL locus. In six of eight (75%) cases, VHL LOH was identified.

At the time of EUS–FNA, patients ranged in age from 20 to 87 years (mean, 60.2 years) and were predominantly female (55 of 91, 60%). The pancreatic cysts ranged in size from 0.7 to 9.9 cm (mean, 3.0 cm) and were distributed throughout the pancreas. This included 26 (29%) located in the head of the pancreas, 9 (10%) in the uncinate, 4 (4%) in the neck, 26 (29%) in the body, and 26 (29%) in the tail. Although sufficient for molecular studies, the amount of cyst fluid was insufficient for CEA analysis in 21 (23%) cases. In addition, 34 (37%) specimens were either less than optimal (33%, n = 30) or unsatisfactory (4%, n = 4) for cytopathologic diagnosis. The primary reason for specimen inadequacy was predominantly due to scant-to-absent cellularity.

Point mutations in GNAS, KRAS, or both were detected in 18 (20%), 36 (40%), and 11 (12%) cases, respectively. GNAS mutations consisted of either p.R201H (n = 9, 50%) or p.R201C (n = 9, 50%; Fig. 1). Mutations in KRAS were predominantly at codon 12 (34 of 36, 94%). These included p.G12D (n = 17, 47%), p.G12V (n = 12, 33%), p.G12R (n = 4, 11%), and p.G12F (n = 1, 3%). A mutation in codon 13 of KRAS was found in two (6%) cases and corresponded to p.G13D. No KRAS mutations in codon 61 were detected.

In total, 16 of 41 (39%) IPMNs and two of nine (22%) IPMNs with an associated invasive adenocarcinoma harbored GNAS mutations. KRAS mutations were identified in 28 (68%) IPMNs, 7 (78%) IPMNs with invasive adenocarcinoma, and 1 of 16 (6%) MCNs. Mutations in both genes were found in 10 (24%) IPMNs and 1 (13%) IPMN with invasive adenocarcinoma. And, as summarized in Table 1, mutations in either GNAS or KRAS were present in 34 (83%) IPMNs, 8 (89%) IPMNs with invasive adenocarcinoma, and 1 (6%) MCN. GNAS and KRAS mutations were absent in cystic PanNETs, SCA, retention cysts, pseudocysts, and a lymphoepithelial cyst. Univariate analysis showed that mutations in either GNAS or KRAS were associated with higher occurrence in males (58% vs. 23%; P = 0.001), increased mean patient age at diagnosis (66.5 vs. 54.5 years; P < 0.001), smaller mean cyst size (2.6 vs. 3.4 cm; P = 0.02), increased fluid viscosity (79% vs. 25%; P < 0.001), and elevated CEA (71% vs. 38%; P = 0.008). There was no statistically significant correlation between GNAS or KRAS status and pancreatic cyst location (P = 0.14), focality (P = 0.13), or specimen adequacy for cytopathologic diagnosis (P = 0.52).

Of the 50 IPMNs within the study cohort, independent statistical correlations between GNAS and KRAS mutations with respect to patient demographics, endoscopic ultrasound cyst/fluid characteristics, cytologic findings, and histologic features were performed (Table 2). GNAS-mutant IPMNs occurred predominantly in male patients (83% vs. 41%; P = 0.007); however, no correlation with gender was seen in KRAS-mutant cysts (P = 0.77). No statistical differences in mean patient age were observed with either gene (GNAS, P = 0.66; KRAS, P = 0.36). With regard to endoscopic ultrasound and radiologic findings, there was no association between GNAS or KRAS mutations and mean cyst size (P = 0.43 and 0.48, respectively), location within the pancreas (P = 0.77 and 0.38), focality (P = 0.77 and 0.55), main duct dilatation (P = 1.00 and 0.11), the presence of a mural nodule (P = 1.00 and 0.67), fluid viscosity (P = 0.07 and 1.00), or CEA levels (P = 0.43 and 1.00).

By cytology, no significant differences between genetic mutation and specimen adequacy were observed (GNAS and KRAS, P = 1.00 and 0.765, respectively). Although this remained an issue for cytopathologic interpretation, a similar number of cases were diagnosed as at least suspicious for a mucinous cyst irrespective of mutational status (GNAS, P = 0.13; KRAS, P = 1.00). Upon surgical resection, there was no correlation between GNAS mutations and location within the pancreatic duct (P = 0.92). However, the prevalence of KRAS mutations was higher in IPMNs involving the branch duct (66% vs. 33%; P = 0.04) and lower in IPMNs involving the main duct (17% vs. 46%; P = 0.04). On the basis of histologic subtype, GNAS mutations associated with the intestinal subtype (P = 0.002) and, in fact, were found in all (100%) intestinal IPMNs within the study cohort. Conversely, no relationship between histologic subtype and KRAS mutations were identified (P = 0.20). Of note, oncocytic IPMNs were absent within the study cohort. No statistically significant correlations were seen with grade of dysplasia or association with invasive adenocarcinoma for either gene (GNAS and KRAS; P > 0.05).

In the preoperative setting, GNAS mutations had 100% specificity for IPMNs, but only attained sensitivity of 36% (Table 3). Mutations in KRAS had 98% specificity and 70% sensitivity. The combination of GNAS and KRAS mutations achieved a specificity and sensitivity of 98% and 84%, respectively. In comparison, the presence of multiple pancreatic cysts, a finding on endoscopic ultrasound often associated with IPMNs, had a specificity and sensitivity of 76% and 52%, respectively. In addition, increased fluid viscosity and elevated CEA had both a lower specificity (83% and 69%, respectively) and lower sensitivity (78% and 74%, respectively). A cytopathologic diagnosis of at least suspicious for a neoplastic mucinous cyst also had a lower specificity and sensitivity of 71% and 60%, respectively.

Overall, point mutations in GNAS and KRAS have a specificity of 100%, but a sensitivity of 65% for mucinous differentiation. Increased fluid viscosity, elevated CEA, and cytology had lower specificities of 96%, 90%, and 76%, respectively. However, increased fluid viscosity and elevated CEA had slightly higher sensitivities of 68% and 69%, whereas cytology was lower at 55%. Although there was insufficient cyst fluid present in 21 cases (23%), the combination of GNAS and KRAS testing with elevated CEA improved the sensitivity of both assays to 86% and maintained a high specificity of 90% for mucinous differentiation. The addition of fluid viscosity to DNA testing and elevated CEA increased the sensitivity (88%), but at a loss in specificity (86%). Using a multimodal approach, including DNA testing, CEA analysis, fluid viscosity, and cytopathology, further increased the sensitivity to 94%, but reduced the specificity to 67%.

Massive parallel sequencing of IPMNs has identified recurrent mutations in GNAS at codon 201, resulting in either an R201H or an R201C substitution (9, 10). Among the major pancreatic cystic neoplasms, these activating mutations have been shown to be highly specific for IPMNs (8). Consistent with these studies, we found that GNAS mutations had a specificity of 100% for IPMNs, whereas mutations in GNAS were absent in MCNs and SCA. In addition, we report, for the first time, the lack of GNAS mutations in cystic PanNETs, retention cysts, pseudocyst, and lymphoepithelial cyst. Although infrequent, these entities can enter the differential diagnosis of an IPMN, either clinically, radiographically, or cytologically. Thus, GNAS testing of pancreatic cyst fluid, obtained via EUS–FNA, represents a highly specific diagnostic modality in the identification of IPMNs.

However, within this study cohort, only 18 of 50 (36%) IPMNs harbored GNAS mutations. Wu and colleagues (9) identified GNAS mutations in 66% of IPMNs. Furthermore, in a follow-up study, Dal Molin and colleagues (20) found that 64% of IPMNs harbored GNAS mutations. A lack of sensitivity within our DNA detection technique may account for these discordant results. Indeed, Furukawa and colleagues (10) also performed whole-exome analysis of surgically resected IPMNs, but used Sanger sequencing and found a similar frequency (41%) of GNAS mutations within their cohort. Yet, KRAS mutations in this study were detected at a comparable rate in IPMNs to those reported by Wu and colleagues (9). An alternative explanation is that the amount of DNA from lysed and shed epithelium obtained by fine-needle aspiration may be less than that found in postoperative aspiration because of surgical manipulation. Despite evaluating a smaller sample size and lack of surgical follow-up data, Siddiqui and colleagues (21) also identified smaller proportion of GNAS-mutant IPMNs (44%) using pancreatic cyst fluid obtained by EUS–FNA. In addition, selection bias may also be a factor in the prevalence of GNAS mutations. Pancreatic cyst fluid DNA used for this study was obtained as part of routine clinical evaluation for KRAS mutations due to concern for a cystic neoplasm (7). In other words, pancreatic cyst fluid was submitted for DNA testing at the discretion of the endoscopist because of uncertainty as to whether a cystic lesion represented a mucinous cyst. Therefore, straightforward main and/or branch duct IPMNs by imaging were refrained from molecular analysis. Nevertheless, an argument can be made that in this scenario, molecular ancillary testing is not clinically indicated (or, only 36% of IPMNs within our patient population harbor GNAS mutations).

Despite the lower prevalence of GNAS mutations, the sensitivity and specificity for IPMNs by combining both GNAS and KRAS analyses was 84% and 98%, respectively. Other standard imaging, laboratory and pathologic techniques had both lower sensitivities and specificities. However, a note of caution should be taken when interpreting these results. As with any test, previous studies have shown that cyst classification based on DNA analysis alone can lead to false-positive results (22). Therefore, the assessment of GNAS and KRAS mutations should be made as part of a multidisciplinary approach including clinical, radiographic, and cytopathologic findings.

For mucinous cysts (both IPMNs and MCNs), point mutations in GNAS and KRAS had a sensitivity of 65% and a specificity of 100%. The lower sensitivity of molecular analysis for mucinous cysts as compared with IPMNs alone is largely due to the absence of both GNAS and KRAS mutations in MCNs. Within this study cohort, only 1 of 16 (6%) MCNs harbored a KRAS mutation. The relative lack of KRAS mutations identified in MCNs is, once again, contrary to those reported by Wu and colleagues (9) and can be explained by the same aforementioned reasoning for the lower rate of GNAS mutations in IPMNs. But more notably, these findings highlight the importance of identifying additional markers to improve the sensitivity of cyst fluid DNA analysis.

Recent whole-exome sequencing of cyst epithelium from the four major neoplastic pancreatic cysts has identified a limited number of genetic mutations that may be used diagnostically to classify each cyst type (8–10). IPMNs were characterized by mutations in GNAS, KRAS, and the E3 ubiquitin ligase, RNF43. RNF43 mutations were also identified in 40% of MCNs. The combination of GNAS, KRAS, and RNF43 testing would improve the sensitivity, while maintaining a high specificity for mucinous differentiation than GNAS and KRAS alone. In contrast, SCA had mutations in VHL or LOH in or adjacent to VHL, and did not contain mutations in GNAS, KRAS, or RNF43. Although not discussed in detail, six of eight (75%) SCA used within this study also demonstrated VHL LOH. Finally, solid pseudopapillary neoplasms were characterized by mutations in CTNNB1 and lacked GNAS, KRAS, RNF43, and VHL mutations. A five-gene panel that includes GNAS, KRAS, RNF43, VHL, and CTNNB1 could lead to a highly accurate diagnosis.

Although GNAS and KRAS testing can be useful in distinguishing mucinous from nonmucinous cysts, a critical role for cyst fluid DNA analysis has been to identify the malignant potential of a pancreatic cyst. Neither GNAS nor KRAS mutations correlated with grade of dysplasia or the presence of invasive adenocarcinoma within a mucinous cyst. Similar studies have not been performed with RNF43. However, Kanda and colleagues (23) reported the presence of mutations within the tumor-suppressor gene, TP53, in secretin-stimulated pancreatic juice samples collected from duodena of patients with high-grade lesions (pancreatic intraepithelial neoplasia grade 3 and high-grade IPMNs) and pancreatic cancer. TP53 mutations were identified in 50% of patients with high-grade lesions and 67% of patients with invasive pancreatic adenocarcinoma. No mutations in TP53 were detected in patients with low-grade IPMNs. In addition, the authors did not identify TP53 mutations in subjects with chronic pancreatitis, but prior studies using endoscopic retrograde cholangiopancreatography-collected juice have found TP53 mutations in a small percentage of individuals (24–26). Although TP53 mutations may prove to be a useful marker in determining the presence of high-grade lesions and invasive adenocarcinoma, additional studies are required to assess its accuracy in patients undergoing pancreatic cyst evaluation.

In summary, we report the largest series of GNAS and KRAS analysis from pancreatic cyst fluid obtained by EUS–FNA. Although the combination of GNAS and KRAS testing was highly specific and sensitive for IPMNs, these results should not be interpreted in isolation. We recommend integrating cyst fluid DNA analysis in conjunction with clinical, radiographic, and cytopathologic data. Regardless, the lack of sensitivity for MCNs underscores the need for additional markers to improve the detection of pancreatic mucinous neoplasms and toward appropriate patient management.

R.E. Brand is a consultant/advisory board member for Asuragen, Inc. No potential conflicts of interest were disclosed by the other authors.

Conception and design: A.D. Singhi, K.E. Fasanella, R.K. Pai, A. Khalid

Development of methodology: A.D. Singhi, M.N. Nikiforova, G.S. Mantha, A. Khalid

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A.D. Singhi, M.N. Nikiforova, K.E. Fasanella, R.E. Brand, X. Lu, G.I. Papachristou, A. Slivka, H.J. Zeh, K.K. Lee, A. Tsung, G.S. Mantha, A. Khalid

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A.D. Singhi, A. Tsung, A. Khalid

Writing, review, and/or revision of the manuscript: A.D. Singhi, M.N. Nikiforova, K.E. Fasanella, K.M. McGrath, N.P. Ohori, T.L. Bartholow, R.E. Brand, J.S. Chennat, G.I. Papachristou, A. Slivka, H.J. Zeh, A.H. Zureikat, K.K. Lee, A. Tsung, G.S. Mantha, A. Khalid

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A.D. Singhi, J.S. Chennat

Study supervision: A.D. Singhi, G.S. Mantha, A. Khalid

The authors thank Mrs. Robyn L. Roche for outstanding administrative assistance. In addition, the authors thank Drs. Samuel A. Yousem and Alyssa M. Krasinskas for helpful comments and suggestions.

This study was supported in part by a grant from the National Pancreas Foundation, Western Pennsylvania Chapter (to A.D. Singhi).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.