Main

GNAS hotspot mutations have been described in various epithelial neoplasms, including appendiceal, pancreatic, and colonic tumors, and are found in up to two-thirds of intraductal papillary mucinous neoplasms (IPMNs).1, 2, 3, 4, 5 GNAS is a proto-oncogene originally described in pituitary adenomas that encodes for the Gsα subunit of heterotrimeric G-proteins that transduce signals via the G-stimulatory pathway and increase production of cyclic AMP (cAMP).6 Hotspot mutations have been identified in codons 201 and 227, with both hotspots resulting in constitutive activation of the G-stimulatory pathway with increased cAMP production.6, 7

GNAS mutations in pancreatic cyst fluid have been shown to be a helpful biomarker in classifying pancreatic cystic lesions and are highly specific for IPMNs in that setting.1, 2 GNAS mutations have also been identified in invasive pancreatic adenocarcinomas arising from IPMNs but rarely in invasive carcinomas without associated IPMNs.1 GNAS mutations are present in ~20% of appendiceal mucinous tumors including low-grade appendiceal mucinous neoplasms and mucinous adenocarcinomas.3 Additionally, ~2% of colonic adenocarcinomas harbor hotspot GNAS mutations and have been associated with a villous morphology and right-sided location.4 Another study has found a correlation between GNAS mutations in colonic adenocarcinomas and a mucinous phenotype, with hotspot GNAS mutations seen in 20% of the mucinous colonic adenocarcinomas in the study.8 GNAS-mutated neoplasms in the colon and pancreas frequently harbor concurrent mutations in KRAS, and are seen concurrently in ~90% of GNAS-mutated colon adenocarcinomas and in 50% of GNAS-mutated IPMNs.1, 4

Five to ten percent of lung adenocarcinomas are invasive mucinous adenocarcinomas (IMA).9, 10, 11 Mucinous neoplasms of the lung can be diagnostically challenging, as these tumors can be morphologically and immunohistochemically identical to metastatic mucinous tumors of the gastrointestinal tract.12, 13, 14, 15, 16, 17, 18 Primary IMA of the lung have distinctive molecular alterations, including frequent pathway-activating mutations in KRAS, occasional loss of function mutations in NKX2-1, and rare rearrangements in ALK and NRG1.19, 20, 21, 22, 23, 24 Although GNAS mutations have been described in association with mucinous and non-mucinous tumors in multiple organs, pathway-activating GNAS mutations have yet to be characterized in lung carcinomas. Assessing whether GNAS mutations can also be associated with primary lung adenocarcinomas could be of potential use to determine primary site of origin of mucinous tumors in the lung. In this study, we interrogate 2352 lung carcinomas subjected to targeted somatic tumor sequencing panels from two institutions for the presence of activating GNAS mutations and evaluate their associated clinicopathologic features.

Materials and methods

Case Selection

This study was approved by the institutional review boards at Dana Farber Cancer Institute and Massachusetts General Hospital and included patients with lung carcinoma undergoing a targeted next generation sequencing assay performed on tumor tissue at the Center for Advanced Molecular Diagnostics (n=1282) (Department of Pathology, Brigham and Women's Hospital) and the Center for Integrated Diagnostics (n=1070) (Department of Pathology, Massachusetts General Hospital). Only cases that were determined to be a primary lung carcinoma based on clinical history, radiologic findings, and pathologic features were included.

Clinicopathologic Features

The following features were recorded in the cases examined: sex, age at diagnosis, clinical presentation, radiographic findings, smoking history, and clinical stage. Histological slides were reviewed, and classification was performed according to the International Association for the Study of Lung Cancer recommendations.25 IMA was defined by goblet or columnar cell morphology with abundant intracytoplasmic mucin in greater than 90% of the tumor cells.

Immunohistochemistry

Primary antibody for TTF-1 was applied to five micron glass slides of formalin fixed, paraffin-embedded tumor tissue (clone 8G6G3/1 Dako, Inc., 1:300 dilution in citrate buffer, pressure cooker). Dako EnVision+ (Agilent Technologies, Santa Clara, CA) was used for signal detection. Strong nuclear immunohistochemical staining for TTF-1 in tumor cells was considered positive.

Targeted Tumor Genomic Sequencing (‘OncoPanel’) at the Center for Advanced Molecular Diagnostics

OncoPanel tumor profiling was performed as previously described.26 Indexed sequencing libraries were prepared from a 50-ng sonically sheared DNA sample using Illumina TruSeq LT reagents (Illumina, Inc., San Diego, CA). A custom RNA bait set by Agilent SureSelect (Agilent Technologies, Santa Clara, CA) was used to enrich libraries for exons and select introns of 282 genes implicated in cancer biology by using solution-based hybrid capture (list of genes and transcripts, Supplementary Table 1). Massively parallel sequencing was performed using Illumina HiSeq2500 (Illumina, Inc.) with 100 × 100 paired-end reads to achieve a mean target coverage greater than 50 ×. Sample reads were analyzed using a custom pipeline including Picard, BWA, and GATK (Broad Institute, Cambridge, MA) for sorting and alignment. Mutation analysis for single-nucleotide variants was performed using MuTect version 1 0.27200 (Broad Institute, Cambridge, MA). Insertions and deletions were called using Indelocator. Variants were filtered to exclude known germline variants in the Single-Nucleotide Polymorphism database by MuTect and variants that occur at a population frequency >0.1% in the Exome Sequencing Project database.

Targeted Tumor Genomic Sequencing at the Center for Integrated Diagnostics

Mutation hotspots and exons from 39 genes were targeted using Anchored Multiplex PCR.27 Total nucleic acid containing total RNA and genomic DNA were extracted from formalin-fixed, paraffin-embedded tissue by using the Agencourt FormaPure Kit (Beckman Coulter, Indianapolis, IN, https://www.beckmancoulter.com) before undergoing end-repair, adenylation, and adapter ligation. Two hemi-nested PCR reactions were then performed using one primer specific to a sequence in the gene of interest and one specific to a universal sequence in the adapter, for each PCR. This library was then sequenced on the Illumina MiSeq (Illumina, San Diego, CA) and 2 × 150 base paired-end reads were aligned to the human genome hg19 reference sequence with the BWA-MEM software package (version bwa-0.7.5a).28 Single-nucleotide variants were detected with MuTect and indels with a custom indel-detection algorithm.29 In regions with sufficient read coverage, this approach is validated to detect indels and single-nucleotide variants with allelic frequencies as low as 5%. The complete list of the genes and exons interrogated is listed in Supplementary Table 2.

Statistical Analysis

Categorical data was analyzed using Fisher's exact test. Statistical analysis was two-sided, and a P-value of <0.05 was considered statistically significant.

Results

Clinicopathologic Features of GNAS-Mutated Lung Carcinomas

The combined cohort of lung carcinomas (n=2352) included the following histotypes: 79% adenocarcinoma, 12% squamous cell carcinoma, 7% non-small cell lung carcinomas not otherwise specified, 1% large cell neuroendocrine carcinoma, and 1% adenosquamous carcinoma. Hotspot GNAS mutations involving codons 201 and 227 were identified in 11 of 1282 lung carcinomas (0.9%) at the Center for Advanced Molecular Diagnostics and in 8 of 1070 lung carcinomas (0.7%) at the Center for Integrated Diagnostics (total 19/2352, 0.8%), including 5 at codon 227 and 14 at codon 201 (Table 1). Abdominal and pelvic imaging study results were available for review in 18/19 patients with 15/18 (83%) having completely negative findings, 1 with multiple liver and bony metastases (case 2), 1 with multiple liver, pancreas, and adrenal gland lesions thought to represent metastases (case 9), and 1 with a 1 cm cystic lesion in the pancreas (case 17). Follow-up MRI imaging was performed on the cystic pancreatic lesion, which revealed a side-branch IPMN. This side-branch IPMN was considered to be indolent in nature and unrelated to the patient's lung tumor. In all 19 GNAS-positive cases, the patients were clinically determined to have primary lung tumors.

Table 1 Clinicopathologic and molecular features of GNAS-mutated lung adenocarcinoma
Full size table

Patients with GNAS-mutated lung carcinomas had a median age of 66 (range 24–78). GNAS-mutated lung carcinomas occurred predominantly in female patients (16/19, 84%). Ten of 19 GNAS-mutated lung carcinomas (53%) were classified as IMAs, whereas the remaining 9 tumors were adenocarcinomas without mucinous histological features (Figure 1). No cases met the criteria for colloid adenocarcinoma in the material that was available for review. TTF-1 immunohistochemistry was performed when tissue was available (n=11) and was positive in 6/7 non-mucinous adenocarcinomas (86%) and 1/4 IMA (25%) (P=0.09, Figure 2). Patients with GNAS-mutated non-mucinous adenocarcinomas were more likely to have a history of smoking (9/9, 100%) than those with IMAs (2/10, 20%) (P<0.001) (Figure 3).

Figure 1
figure 1

GNAS-mutated lung carcinoma histology. (a) Case 3 with GNAS p.Gln227Leu, non-mucinous adenocarcinoma histology with papillary and micropapillary architecture. (b) Case 4 with GNAS p.Gln227Leu, non-mucinous adenocarcinoma histology with acinar architecture. (c) Case 12 with GNAS p.Arg201Cys and KRAS p.Gly12Asp, invasive mucinous adenocarcinoma. (d) Case 17 with GNAS p.Arg201Cys and KRAS p.Gly13Asp, invasive mucinous adenocarcinoma.

Full size image
Figure 2
figure 2

TTF-1 immunohistochemistry in GNAS-mutated lung invasive mucinous adenocarcinomas. (a,b) Case 10 with GNAS p.Arg201His and BRAF p.Val600Glu with positive TTF-1 immunohistochemistry. (c,d) Case 11 with GNAS p.Arg201Cys and KRAS p.Gly12Cys with negative TTF-1 immunohistochemistry.

Full size image
Figure 3
figure 3

Clinicopathologic features of GNAS-mutated lung carcinomas. IMA, invasive mucinous adenocarcinoma.

Full size image

Clinical presentation of IMAs included upper respiratory infection with cough in 2 cases, incidentally identified lung nodule in 4 cases, abdominal pain in 1 case, follow-up imaging for a history of esophageal adenocarcinoma in 1 case, and chronic cough with shortness of breath in 1 case (Supplementary Table 3). Radiographic presentation included multiple bilateral lung nodules in 4 cases, cavitary or cystic nodules in 3 cases, and groundglass opacities in 6 cases (Supplementary Table 3). Sixty percent of IMAs (6/10) were stage I at the time of diagnosis compared to 22% of non-mucinous adenocarcinomas (2/9); however, this correlation did not reach statistical significance (P=0.17).

GNAS codon 227 Mutated Lung Carcinomas Have Non-Mucinous Adenocarcinoma Histology

Five neoplasms had GNAS p.Gln227 mutations (4 p.Gln227Leu, 1 p.Gln227His) (Table 1). All of the GNAS p.Gln227 variants were transversion mutations (A>T (n=4), G>T (n=1)) and had non-mucinous adenocarcinoma histology (Figures 1a and b). The morphologies among these 5 cases included a mix of acinar, papillary, micropapillary, solid, and lepidic patterns. One case had a concurrent KRAS mutation (p.Gly12Cys), but did not show mucinous histologic features.

GNAS codon 201 Mutated Lung Carcinomas with Concurrent Raf/Ras Pathway Mutations are Associated with Invasive Mucinous Adenocarcinoma Histology

Fourteen neoplasms had GNAS p.Arg201 mutations (6 p.Arg201Cys, 6 p.Arg201His, 1 p.Arg201Leu, 1 p.Arg201Ser). Additionally, a subset of these tumors (12/14, 86%) had concurrent mutations in the Ras/Raf pathway (9 KRAS, 1 HRAS, 2 BRAF) (Table 1). All IMA had GNAS p.Arg201 mutations and concurrent Ras/Raf pathway mutations (9 KRAS, 1 BRAF) (Figures 1c and d). A large percentage (78%) of KRAS mutations in IMAs with concurrent GNAS p.Arg201 mutations were G>A transitions. Similarly, all of the GNAS p.Arg201 variants seen in IMAs were also transitions (G>A (n=4) or C>T (n=6)). Case 7 had GNAS p.Arg201Leu and HRAS p.Gln61Arg mutations, and while there was only a cytologic specimen available for review, no mucinous features were present. Additionally, Case 8 had GNAS p.Arg201His and BRAF p.Gly469Ala mutations but did not have histological features of IMA.

Additional Genomic Alterations in GNAS-Mutated Lung Adenocarcinomas

Recurrent mutations in TP53 and CDKN2A were identified in the cohort of 19 GNAS-mutated lung adenocarcinomas. TP53 mutations were seen in 3 of 5 (60%) adenocarcinomas with GNAS p.Gln227 mutations and 4 of 14 (29%) adenocarcinomas with GNAS p.Arg201 mutations. CDKN2A mutations were seen in 1 of 5 (20%) adenocarcinomas with GNAS p.Gln227 mutations and 2 of 14 (14%) adenocarcinomas with GNAS p.Arg201 mutations. TP53 variants included nonsense, splice site, and missense mutations, including TP53 p.Tyr220Cys, p.Arg249Gly, and p.Arg273Leu hotspot mutations. TP53 mutations were more likely to be seen in non-mucinous GNAS-mutated adenocarcinomas than in GNAS-mutated IMAs, 67 vs 10%, respectively (P=0.02). CDKN2A variants were frameshift or splice site variants, implying loss of tumor suppressor function (Table 1).

Discussion

GNAS mutations are most commonly associated with gastrointestinal neoplasms, including pancreatic IPMNs and invasive adenocarcinomas arising from IPMNs, colonic adenocarcinomas and mucinous appendiceal tumors.1, 2, 3, 4, 5, 8 Our study demonstrates that GNAS mutations are present in a small subset (0.8%) of primary lung carcinomas. As the genomic profiling of lung adenocarcinomas becomes more clinically utilized, it is important to recognize that GNAS variants can be identified in primary lung adenocarcinomas and do not necessarily indicate metastasis from a gastrointestinal site. Fifteen patients in this series underwent abdominal imaging without evidence of extrapulmonary disease. Although this study cannot absolutely exclude that these neoplasms may represent metastases from undiagnosed gastrointestinal sites, we consider this possibility unlikely given the patients' clinical evaluations and the known pathogenesis of pancreatic and other gastrointestinal mucinous neoplasms.

Mutations in both GNAS hotspots, p.Arg201 and p.Gly227, are seen in lung carcinomas. Approximately half of GNAS-mutated lung carcinomas have IMA histology, all of which have concurrent GNAS p.Arg201 and Ras/Raf pathway mutations. In contrast, GNAS p.Gly227 mutations are not associated with IMAs. The pathobiology behind this association is uncertain, as both p.Arg201 and p.Gly227 GNAS hotspot mutations have similar mechanisms of action, inhibiting GTPase activity, and thereby functioning as a dominant oncogene by constitutively activating Gs.6

Patients with IMA histology with concurrent GNAS p.Arg201 and Ras/raf pathway alterations are significantly less likely to have a history of smoking than those patients with GNAS p.Gln227 mutations and non-mucinous histology. Although KRAS mutations in lung carcinoma,30, 31 particularly KRAS transversion mutations (G>T,G>C) have a strong association with smoking,32, 33 7 of the 9 IMAs with concurrent GNAS p.Arg201 and KRAS mutations in our cohort had KRAS G>A transition mutations, which is more commonly seen in non-smokers. The presence of these ‘gastrointestinal-type’ KRAS mutations (G>A transitions) has been previously described in primary lung IMAs and should not be taken as evidence that a tumor is a metastasis from a gastrointestinal primary site.34 Additionally, it should be emphasized that the presence of a GNAS mutation cannot be used as a tool to distinguish primary lung adenocarcinomas from gastrointestinal or pancreaticobiliary metastases, as GNAS-mutated tumors from each of these sites have overlapping morphologic, immunophenotypic, as well as molecular findings. Therefore, careful clinicopathologic correlation is required in these scenarios to properly identify primary site of disease.

IMAs comprise a subset of lung adenocarcinomas with distinctive morphological and clinical features. IMAs are becoming an increasingly heterogeneous molecular group. In addition to common oncogenic mutations in KRAS, IMAs may infrequently harbor ALK and NRG1 rearrangements.24 Recurrent NKX2-1 (also known as TTF1) mutations have been identified in conjunction with KRAS mutations, and have been shown to promote pathogenesis and loss of TTF-1 expression in mucinous lung cancer.23 This study shows an association between IMA histology and mutations involving KRAS and GNAS p.Arg201, suggesting a novel mechanism by which cooperative KRAS and GNAS mutations promote primary lung carcinogenesis toward a mucinous or gastrointestinal phenotype. Understanding the association between secondary mutations in KRAS-mutated lung adenocarcinoma and morphological phenotype requires additional experimentation and in vivo models.

The clinical and biological significance of GNAS mutations in lung carcinoma is currently unknown. Some data suggests that patients with pseudomyxoma peritonei and GNAS mutations have a significantly shorter progression free survival than patients without GNAS mutations.35 Additionally, the same study has found that GNAS-mutated colorectal carcinomas have an aggressive clinical course. In contrast, GNAS-mutated IPMN-associated carcinomas follow a more indolent course than those with GNAS mutations36

Currently, therapies directly targeting GNAS mutations are not available. However, based on preclinical evidence, tumors with GNAS mutations may respond to inhibitors of the MAPK pathway.37 More data is needed to determine whether GNAS mutations in lung carcinoma portend a particular clinical outcome or will be responsive to a specific targeted therapy. Since IMAs are heterogeneous on the molecular level, and the histological features of GNAS-mutated lung adenocarcinomas are not distinctive, broad testing of lung adenocarcinomas should be employed to identify such cases, and will become increasingly necessary should a therapeutic agent become available.

Although GNAS mutations are present in an infrequent subset of lung adenocarcinomas, the overall incidence of lung cancer is high, and the prevalence of GNAS-mutated lung adenocarcinomas is expected to be significant. This group of neoplasms is important to recognize and correctly classify the primary site based on clinicopathologic features, which can be challenging in the setting of IMA histology and negative TTF-1 immunohistochemistry. Further studies will be needed to determine their clinical course, biologic behavior, and whether they will be amenable to targeted therapies.