Recurrent somatic mutations of PRKAR1A in isolated cardiac myxoma

Background Cardiac myxomas are benign tumors that commonly arise within the left atria. Familial cardiac myxomas are a part of Carney Complex (CNC), an autosomal dominant multiple neoplasia syndrome caused by germline mutations in PRKAR1A. Seven percent of cardiac myxomas are associated with CNC. To date, the genetic basis of isolated cardiac myxomas (ICM), however, has not been fully elucidated. Methods We investigated the genetic profile of ICM using whole exome sequencing (WES). Suspected mutations were confirmed using targeted sanger sequencing. To further examine the presence of PRKAR1A mutations in ICM, we performed targeted sequencing in an additional 61 ICM specimens. Results 87.5% (7/8) of ICM harbored mutations in PRKAR1A. Three of the 8 ICM harbored biallelic somatic mutations of PRKAR1A, including c.607_610del:p.Leu203fs (pathogenic) + c.C896G:p.Ser299X (pathogenic), c.952delT:p.Leu318fs (pathogenic) + c.769-2 A>G (pathogenic) and c.178-1 G>C (pathogenic) + c. 550+1 G>C (pathogenic). Four of 8 tumors harbored monoallelic PRKAR1A mutations, including c.523_524insG:p.Tyr175_Val176delinsX (pathogenic), c.C920A:p.Ser307X (pathogenic), c.30delG:p.Glu10fs (pathogenic) and c.C289T:p.Arg97X (pathogenic). No identical variants were observed across the 8 ICM samples. Interestingly, none of these variants have been previously described in familial cardiac myxomas. In order to confirm our findings, directed sequencing of 61 ICM specimens was subsequently performed. Sixty-four percent (39/61) of ICMs tumors contained inactivating PRKAR1A mutations. Conclusion Our findings suggest that loss-of-function mutations of PRKAR1A may play a vital role in the formation of isolated cardiac myxomas.


INTRODUCTION
Cardiac myxomas (CMs) are benign cardiac tumors that are characterized by stellate to plump, cytologically bland, mesenchymal cells situated in a myxoid stroma [1,2]. These tumors often arise as a solitary mass situated in the left atria and have been described to occur more commonly in women than men [3,4]. Isolated cardiac myxomas (ICM) tend to occur in individuals in the sixth to seventh decade of life and can present with syncope, palpitations, dyspnea or heart failure [5]. Approximately 7% of CMs occur in association with carney complex (CNC), a tumor predisposition syndrome characterized by the development of pigmented cutaneous lesions, myxomas and multiple endocrine neoplasms [6]. Investigation of the pathogenesis of CNC revealed that 70% of affected individuals harbored germline mutations in PRKAR1A, a gene that encodes the regulatory subunit (type I-alpha) of protein kinase A (PKA) [7][8][9][10]. Indeed, PRKAR1A mutations have been described in two-thirds of CNC-associated CMs. Loss of function mutations in PRKAR1A lead to increased PKA activation thereby promoting CREB phosphorylation and upregulation of the MAPK and Rb/E25 signaling pathways. This dysregulation is believed to play an important role in tumorigenesis [11]. Genetic aberrations of PRKAR1A have been investigated in ICM [12][13][14]. Several authors have reported that ICMs do not harbor PRKAR1A mutations and have suggested that the genetic basis of ICM may differ from CNC associated CMs [13,14]. In contradistinction, a recent study suggested that 31% of ICM may contain inactivating mutations in PRKAR1A [5]. Despite this report, whether PRKAR1A mutations are involved in the pathogenesis of ICM remains controversial and a subject of ongoing debate.

RESULTS AND DISCUSSION
The development of next generation sequencing techniques has facilitated the discovery of previously unidentified cancer driver mutations [15]. We speculated that the discrepancy in the literature regarding the presence of PRKAR1A mutations in ICM could be due to lack of depth and coverage of previously utilized DNA sequencing methods.
In this study, whole exome sequencing (WES) was performed on 8 cardiac myxoma specimens derived from patients without any other manifestation of CNC. The median age of the patients at diagnosis was 51 years (range 44-60 years). Six of the 8 (75%) patients were female. All of the tumors arose within the left atrium, consistent with previous reports [16]. The pathological phenotype of these primary ICM tissues was confirmed by hematoxylin and eosin (H&E) staining ( Figure 1).
To explore the genetic variation in ICM, we performed whole exome sequencing as previously described [17]. The details of capture statistics and coverage are presented in Supplementary Table 1. After barcode-based sample deconvolution, sequence reads were mapped with BWA tools to the human genome (hg19). Sequence variant calls were performed by GATK after removing potential PCR duplicates. We called, on average 80,425 base substitutions and 8,061 insertions/ deletions per tumor. Missense-, nonsense-, or splice-sitealtering variants absent from the 1000 Genomes dataset were selected for further consideration. The majority of the identified variants were observed in the intronic and exonic regions, accounting for 44.81% and 33.77% of identified variants, respectively (Supplementary Table 1). These variants included synonymous, nonsynonymous, stop-gain, stop-loss, splice-site and frameshift mutations. The majority of detected mutations were non-synonymous, accounting for 750-800 identified mutations. This is in contrast to the 14-22 splice-site mutations that were identified in the ICM from each patient ( Figure 2A). Interestingly, we found that the majority of the base pair changes in the transcribed region were C>T (G>A) substitutions ( Figure 2B). To identify pathogenic mutations, we excluded single nucleotide polymorphisms (SNP), synonymous mutations, and variants in the untranslated, intronic and intergenic region. We identified 4391 variants, covering 3198 genes across all eight samples. Mutations that were suspected to result in altered gene product activity accounted for 12% of these variants. These included frameshift, stop gain and splice-site mutations ( Figure 2C). Additionally, we found 15 inactivating mutations involving PRKAR1A, which have been described in CNC related CM ( Figure 2D). Oncogenes such as BRCA1/2, JAK2, NOTCH2, EPHA2 and NCOR2 harbored 2 nonsynonymous mutations, which are extremely rare variants reported in the dbsnp database. Other mutated genes included GXYLT1 (1 frameshift, 1 stop gain and 3 non-synonymous mutations). Furthermore, we found inactivating mutations of ZNF880 and MUC3A in all 8 ICM tissue samples. AIM1L was mutated in 5 of 8 analyzed ICM. However, we found that these variants were germline mutations due to the presence of these mutations in patients matched peripheral blood DNA (data not shown).
Next, we investigated the effect of these inactivating variants on PRKAR1A expression in primary tumor tissues by immunohistochemistry (IHC). In agreement with the mutation analysis, PRKAR1A expression was not identified in the 7 tumors that contained somatic PRKAR1A mutations, suggesting the abrogation of PRKAR1A function in ICM (Supplementary Figure  2). To further assess the PRKAR1A mutation in ICM, we investigated the mutation status of PRKAR1A in an additional 61 ICM tumors. We amplified the 10 exons of PRKAR1A by PCR, and performed targeted Sanger DNA sequencing. In this cohort, 39 of 61 (64%) ICM specimens harbored inactivating PRKAR1A mutations ( Table 2). Similar to our WES data, no hotspot mutation was found in these samples. In addition, most of the PRKAR1A mutations were distributed across Exons 1-5. While expression of PRKAR1A varied across these samples, we speculate that some of them may have resulted in the truncation of the PRKAR1A protein due to the fact that most of the identified mutations were frameshift and nonsense mutations. For instance, case No.9 showed PRKAR1A positive due to the antibody, which may still be able to recognize 200-260 aa of truncated PRKAR1A protein.
Overall our results suggest that ICM may contain PRKARA1A mutations in approximately two-thirds of cases. The remaining samples may contain mutations at levels below our detection limit due to sample procurement. On the other hand, we speculate that there may be other driver mutations in the non-PRKAR1A mutated ICM specimens that have yet to be identified. Therefore, we examined the whole exome sequencing results from the patient whose ICM did not contain a PRKAR1A mutation. In this specimen, we found that the patient presented with a nonsense mutation in MXI1, a tumor suppressor gene that has been reported to be frequently inactivated in prostate cancer [18,19]. This mutation, however, was determined to be a germline mutation, supported by the presence of this mutation in matched blood sample. Interestingly, this patient also carried NOTCH2, EPHA2, NCOR2, FGFR4, and ARAF germline missense mutations. These non-SNP variants were not found in the other patients with PRKAR1A mutations. Unfortunately, we could not identify any clinical or pathologic differences between PRKAR1A mutated and non-mutated ICM patients. Although this is only one sample, this raises the possibility that ICM that do not harbor PRKAR1A mutations may be genetically heterogeneous with multiple genetic variants contributing to tumorigenesis. Thus, further work is necessary to define the clinical and pathologic differences between PRKAR1A-mutated and PRKAR1A non-mutated isolated cardiac myxomas. Our work, reported herein, clearly defines that PRKAR1A mutations are present in a substantial proportion of isolated cardiac myxomas.

Whole exome sequencing
Genomic DNA was isolated and sequenced using standard protocols for next generation sequencing (Compass Biotechnology, Co. Ltd. Beijing, China). Briefly, the Agilent SureSelect Human All Exon 60 Mb kit (Agilent Technologies, Santa Clara, CA, USA) was utilized to capture all exons. Shotgun libraries were established by shearing DNA and ligating sequencing adaptors bearing a 6 base-pair index, followed by polymerase chain reaction (PCR). Libraries were hybridized to SureSelect Human All Exon V6 solution-based probes, amplified, pooled, and sequenced on Illumina Hiseq X10 (PE 150 bp). The final targeted region was 60,456,963 base pairs, including approximately 20,000 genes (230,418 exons and splice junctions) in the genome. The average mean fold coverage was 203X. 97.9% of target bases were covered at least at once and 97.5% were covered at least 20X.

Validation of variants by Sanger sequencing
The germline and somatic variants identified by WES were further analyzed by Sanger Sequencing. Briefly, primers were designed by generunner software to amplify the genomic region using the extracted DNA from ICM tissues and blood samples. The PCR product was sent for automatic DNA sequencing.

Hematoxylin and eosin (H&E) staining
ICM primary tissues were fixed in 4% paraformaldehyde for 24h, washed in PBS and were embedded in paraffin. Two-micrometer sections were stained with hematoxylin and eosin following standard procedures.

Author contributions
JH performed the DNA extraction as well as sequencing analysis, MJS extracted the genomic DNA, performed PCR and targeted sequencing analysis. YYH, EYL, CSW and LG performed the pathological assay for ICMs. DC performed the bioinformatic analysis. MS and KP wrote the manuscript. ZPZ and YL designed all the experiments and wrote the manuscript.

ACKNOWLEDGMENTS
The study has been supported by the Hundred Talents Program of CAS and Innovative Research Grant to Scientific Research Center for Translational Medicine at Dalian Institute of Chemical Physics.