Detection of a Recurrent DNAJB1-PRKACA Chimeric Transcript in Fibrolamellar Hepatocellular Carcinoma

See allHide authors and affiliations

Science  28 Feb 2014:
Vol. 343, Issue 6174, pp. 1010-1014
DOI: 10.1126/science.1249484

Oncogenic Suspect Exposed

It can be difficult logistically to study the genomics of rare variants of common cancers. Nevertheless, Honeyman et al. (p. 1010) studied fibrolamellar hepatocellular carcinoma (FL-HCC), a rare and poorly understood liver tumor that affects adolescents and young adults and for which there is no effective treatment. FL-HCCs from 15 patients all expressed a chimeric RNA transcript and protein containing sequences from a molecular chaperone fused in frame with sequences from the catalytic domain of protein kinase A. The chimeric protein retained kinase activity in vitro. Such recurrent gene fusions in cancer may signal a role in pathogenesis and provide an opportunity for therapeutic intervention.


Fibrolamellar hepatocellular carcinoma (FL-HCC) is a rare liver tumor affecting adolescents and young adults with no history of primary liver disease or cirrhosis. We identified a chimeric transcript that is expressed in FL-HCC but not in adjacent normal liver and that arises as the result of a ~400-kilobase deletion on chromosome 19. The chimeric RNA is predicted to code for a protein containing the amino-terminal domain of DNAJB1, a homolog of the molecular chaperone DNAJ, fused in frame with PRKACA, the catalytic domain of protein kinase A. Immunoprecipitation and Western blot analyses confirmed that the chimeric protein is expressed in tumor tissue, and a cell culture assay indicated that it retains kinase activity. Evidence supporting the presence of the DNAJB1-PRKACA chimeric transcript in 100% of the FL-HCCs examined (15/15) suggests that this genetic alteration contributes to tumor pathogenesis.

Fibrolamellar hepatocellular carcinoma (FL-HCC) is a rare liver tumor that was first described in 1956 and that historically has been considered a variant of hepatocellular carcinoma (1, 2). It is histologically characterized by well-differentiated neoplastic hepatocytes and thick fibrous bands in a noncirrhotic background (3, 4). FL-HCC has a clinical phenotype distinct from conventional hepatocellular carcinoma and usually occurs in adolescents and young adults. Patients have normal levels of alpha fetoprotein without underlying liver disease or history of viral hepatitis (36). Little is known of its molecular pathogenesis. FL-HCC tumors do not respond well to chemotherapy (7, 8), and surgical resection remains the mainstay of therapy, with overall survival reported to be 30 to 45% at 5 years (1, 6, 8, 9).

To investigate the molecular basis of FL-HCC, we performed whole-transcriptome and whole-genome sequencing of paired tumor and adjacent normal liver samples. To determine whether there were tumor-specific fusion transcripts among the coding RNA, we ran the program FusionCatcher (10) on RNA sequencing (RNA-Seq) data from 29 samples, including primary tumors, metastases, recurrences, and matched normal tissue samples, derived from a total of 11 patients (table S1). There was only one recurrent candidate chimeric transcript detected in every tumor sample. This candidate transcript is predicted to result from the in-frame fusion of exon 1 from the DNAJB1 gene, which encodes a member of the heat shock 40 protein family, with exons 2 to 10 from PRKACA, the gene encoding the adenosine 3′,5′-monophosphate (cAMP)–dependent protein kinase A (PKA) catalytic subunit alpha. This fusion transcript was not detected in any of the available paired normal tissue samples (n = 9). This fusion is not found in the COSMIC database (11) and has not previously been reported in the literature.

To further characterize the candidate fusion transcript, we directly examined those RNA-Seq reads that mapped to PRKACA and DNAJB1. We examined PRKACA transcript levels with DESeq2 (12) and found that they were increased relative to normal in tumors from all nine patients tested [P value adjusted for multiple testing (pAdj) < 10−12, range three- to eightfold]. To determine whether the increased expression was attributable to a specific isoform of PRKACA, we quantified reads mapping to different exons and evaluated differential expression using DEXSeq (13). In all nine patients, there was an increase in the expression of exons 2 to 10 of PRKACA in the tumor relative to exon 1 and relative to the expression in normal tissue (Fig. 1A, left). This exon expression pattern does not correspond to a known isoform of PRKACA. Rather, it reflects an increase in PRKACA transcripts lacking the first exon, which encodes the domain that engages the regulatory subunits of PKA. All reads mapping to PRKACA in normal tissue were either contained within exons or bridged the junctions between adjacent exons at annotated splicing sites (Fig. 1B, left, blue). All tumor samples additionally had reads mapping from the start of the second exon of PRKACA to a point ~400 kilobases (kb) upstream of the coding region, which corresponded to the end of the first exon of DNAJB1 (marked with an asterisk in Fig. 1B). Examination of the exon expression of DNAJB1 in tumor samples revealed a decrease in the number of reads in exons 2 and 3 relative to exon 1 (Fig. 1, A and B, right). The data on the differential exon expression and the data on the RNA-Seq reads spanning the 400-kb distance that bridges these two genes further support a structural variation resulting in a chimeric transcript incorporating DNAJB1 and PRKACA sequences.

Fig. 1 RNA-Seq read coverage from fibrolamellar hepatocellular carcinoma and adjacent healthy liver tissue.

(A to C) Plot of reads mapped to chromosome 19 in the region encoding, on the negative strand, the genes PRKACA (chr19:14,202,499 to 14,228,558) and DNAJB1 (chr19:14,625,580 to 14,640,086) from the normal tissue (blue) and FL-HCC tissue (red). (A) Normalized RNA-Seq read counts from nine pairs of tumor and adjacent tissue demonstrate a consistent increase in tumor relative to normal in the reads mapping to exons 2 through 10 of PRKACA and a decrease in the reads of exon 1. Exons are shown as orange and green blocks [see (D)]. Normalized read counts are plotted per exon part (nonoverlapping portions of exons in all isoforms in ENSEMBL annotation; indicated by empty boxes). Transcript structure (solid color boxes) indicates most likely dominant isoform as inferred by RNA-Seq read coverage. Lines indicate the average normalized read count per exon part (in dominant isoform) for normal and tumor samples. (B) Sashimi plot (39) of RNA-Seq read coverage at PRKACA and DNAJB1 loci for patient 9. Solid peaks depict reads per kilobase per million reads mapped (RPKM) within individual exons. Reads that bridge different exons are shown as arcs. In every tumor sample (nine out of nine) and in none of the normal tissue sample (zero out of nine), there are reads mapped from the end of exon 1 of DNAJB1 to the start of exon 2 of PRKACA (read counts indicated for patient 9). (C) There is an additional set of reads from patient 4 that map from the second exon of DNAJB1 to the start of the second exon of PRKACA (read counts indicated). Indistinguishable results are observed in metastasis tissue from this same patient. (D) RNA-Seq read mapping predicts the production of four transcripts: a native DNAJB1 (green); a native isoform 1 PRKACA (orange); a predominant chimera with the first exon of DNAJB1 and exons 2 to 10 of PRKACA; and, in a subset of patients, a minority transcript with the first exon and part of the second of DNAJB1 and exons 2 to 10 of PRKACA. (E). Sanger sequencing of RT-PCR products from FL-HCC samples confirmed in seven out of seven patients a chimera transcript joining the end of exon 1 of DNAJB1 and the start of exon 2 of PRKACA.

In tumor samples from patients 4 and 14, there were indications of a second splice variant spanning PRKACA and DNAJB1 (Fig. 1C). In addition to the reads from the end of exon 1 of DNAJB1 to the start of exon 2 of PRKACA (Fig. 1C, middle red plot marked with an asterisk and Fig. 1D, the predominant chimera), there were reads that started in the middle of exon 2 of DNAJB1 and mapped to the start of exon 2 of PRKACA (Fig. 1C, middle red plot marked with a double asterisk and Fig. 1D, the minority chimera). For patient 4, we also sequenced four different metastases and observed reads that spanned the same regions (Fig. 1C, bottom red plot). These findings predict the presence of one predominant chimera incorporating only the first exon of DNAJB1 and a second, minority chimera, incorporating both the first and a portion of the second exon of DNAJB1. Both chimeras continue with exons 2 to 10 of PRKACA.

The presence of the predicted dominant chimeric RNA transcript was validated by Sanger sequencing of reverse transcription polymerase chain reaction (RT-PCR) products in all seven tumor samples tested from six patients (Fig. 1E), including one tumor and recurrence pair. The same chimeric transcript was confirmed by Sanger sequencing in an eighth newly acquired primary tumor from a patient whose samples were not previously analyzed by RNA-Seq or whole-genome sequencing.

In every tumor sample, in addition to the presence of one or both predicted RNA chimeras, we also observed RNA-Seq reads consistent with transcripts covering the entire length of PRKACA, which suggested that the cells still contain a wild-type (WT) copy of the gene. These results are most consistent with a heterozygous deletion in chromosome 19 resulting in a single copy of the chimeric transcript and single copies of both the WT DNAJB1 and the WT PRKACA.

We next searched for potential structural variants in the FL-HCC genome by performing whole-genome sequencing on paired tumor and adjacent normal liver. We used the program Delly (14) to search for structural variations on chromosome 19. In 8 of 10 tumor samples, there were between 2 and 23 paired-end reads that spanned ~400 kb and mapped to both DNAJB1 and PRKACA. In addition, in all eight of these tumor samples, split reads mapped to the deletion fusion point (table S2). The program, SplazerS (15) identified 3 to 17 split reads mapping between these genes in all 10 tumor samples. The predicted deletions ranged in size from 401,552 to 409,262 base pairs (bp), with a telomeric breakpoint between chr19: 14,218,306 and 14,226,300 (hg19) and a centromeric breakpoint between chr19:14,627,567 and 14,628,632. These deletions were not detected in adjacent normal liver samples.

The presence of the predicted deletion in the tumor DNA was validated by PCR and Sanger sequencing in eight out of eight samples tested (Fig. 2B). In all patients, the precise breakpoints of the deletion mapped to different genomic coordinates (table S2). For all patients, the data were consistent with a deletion originating in the first intron (n = 6) or the second exon (n = 4) of DNAJB1 and terminating in the first intron of PRKACA (Fig. 2A).

Fig. 2 DNA sequence analysis of fibrolamellar hepatocellular carcinoma DNA.

(A) Mapping of the size and location of the breaks in the DNA between the DNAJB1 and the PRKACA genes. (B). PCR followed by Sanger sequencing confirmed a deletion of ~400 kD in each patient. Each deletion results in a fusion that starts either in intron 1 or exon 2 of DNAJB1 and ends in intron 1 of PRKACA. The break is in a different location in all patients. Note that the sequencing reads are shown for the positive strand. However, both DNAJB1 and PRKACA are coded off the negative strand.

To determine whether the chimeric RNA transcript was translated into a chimeric protein, we performed Western blot analysis. Proteins extracted from tumor and adjacent normal liver samples were separated by SDS–polyacrylamide gel electrophoresis (SDS-PAGE) and probed with an antibody to the carboxyl terminus of PRKACA. Normal and tumor samples showed a band corresponding to a size of 41 kD (Fig. 3A), the predicted size of isoform 1 of WT PRKACA protein. In nine of nine previously sequenced tumor samples, there was a slower moving band at ~46 kD, consistent with the predicted larger molecular size of the predominant chimeric protein (patients 1, 3, 4, 5, 6, 7, 10, 11, and 12). This higher molecular mass band was not observed in any of the normal samples. Additional pairs of tumor and normal tissue not originally sequenced (patients 13, 14, and 17) were analyzed and showed the same larger band in the tumor but not in the normal tissue. The larger band was also seen in two metastases from patient 2.

Fig. 3 Tumor-specific expression of a protein consistent with the predicted DNAJB1-PRKACA chimera.

(A) Immunoblot analysis. Protein extracts of fibrolamellar carcinoma (T) and adjacent liver tissue (N) were separated on SDS-PAGE and subjected to immunoblot analysis using an antibody to the carboxyl terminus of PRKACA. This analysis revealed the presence of the native PRKACA in all tumor, metastasis, and normal samples and the presence of one additional, apparent higher molecular mass band in all tumor samples (the predominant chimera). There is a second even higher molecular mass band, the minority chimera, in the two tumor samples that had demonstrated a second set of RNA reads mapping between exon 2 of DNAJB1 and exon 2 of PRKACA (patients 4 and 14). (B) Confirmation of chimeric protein. Protein extracts of fibrolamellar carcinoma (T) and adjacent liver tissue (N) were immunoprecipitated with an antibody to the amino terminus of DNAJB1 and run adjacent to total-cell extract on SDS-PAGE. These samples were then subjected to immunoblot analysis with an antibody to the carboxyl terminus of PRKACA (C) PKA activity of WT PRKACA and chimera are indistinguishable. HEK-293T cells were transfected with an empty control plasmid, a plasmid encoding WT PRKACA, or a plasmid encoding the chimeric DNAJB1-PRKACA. Cell extracts were diluted and assayed for PKA activity. The activity of the WT PRKACA and the chimera PRKACA-DNAJB1 are significantly higher [P < 0.001, two-way analysis of variance (ANOVA)] than background kinase activity. Samples were processed in triplicate ± SEM. (D and E) Immunofluorescence assay. The presence and distribution of PRKACA protein was examined with an antibody against the carboxyl terminus in (D) adjacent normal tissue and (E) FL-HCC liver tissue from patient 11, both imaged by confocal microscopy. The green areas correspond to PRKACA and the blue areas correspond to nuclei, which were stained with Hoechst stain. Similar results were seen in samples from additional patients (fig. S1). Scale bar, 20 μm.

We observed a third immunoreactive band in samples from two patients that migrated at ~50 kD (Fig. 3A, patients 4 and14), consistent with the minority chimera predicted by the RNA-Seq analysis. This larger band was present in addition to the bands corresponding to the native PRKACA protein and the predominant DNAJB1-PRKACA chimeric protein. This slower-mobility, apparently higher molecular mass band is consistent with a chimeric protein that includes amino acids corresponding to the first exon and portion of exon 2 of DNAJB1, as well as exons 2 to 10 of PRKACA (Fig. 1C, patient 4 shown).

To confirm that the higher molecular mass bands were the chimeric proteins, we tested reactivity with an antibody to the amino terminus of DNAJB1 and an antibody to the carboxyl terminus of PRKACA. Protein was immunoprecipitated with an antibody to DNAJB1, separated by SDS-PAGE alongside a sample of the total solubilized tissue, and then probed on a Western blot with an antibody to the carboxyl terminus of PRKACA (Fig. 3B). The whole-cell lysate again showed bands corresponding to both WT PRKACA and the chimeric protein. However, the sample immunoprecipitated with DNAJB1 showed only the higher molecular mass band (lanes labeled IP-DNAJB1, Fig. 3B) corresponding to the chimera. The WT PRKACA band was not present in the immunoprecipitate. Thus, the higher molecular mass band that is present only in tumor samples is a true chimeric protein resulting from the fusion of the amino terminus of DNAJB1 and carboxyl terminus of PRKACA.

To determine whether the DNAJB1-PRKACA chimeric protein retains kinase activity, we transfected human embryonic kidney–293T (HEK-293T) cells with plasmids encoding either WT PRKACA or the DNAJB1-PRKACA chimera. Control cells were transfected with an “empty” plasmid lacking the PRKACA or chimera genes, which provided a measure for the background activity level of PKA in the cells. PKA activity was measured by the ability of cell lysate to phosphorylate the fluorescent PKA substrate peptide LRRASLG, whose mobility shifts on a gel upon phosphorylation. In cells expressing either WT PRKACA or the chimera (Fig. 3C), PKA activity was significantly greater than controls (P < 0.001) (Fig. 3C). Both PRKACA and the chimera were each expressed from the same promoter, and the PKA activity was indistinguishable in cells expressing either protein. This demonstrates that the chimeric protein retains full PKA activity. Finally, we examined the expression of PRKACA protein in tumor and adjacent normal tissue by confocal fluorescence microscopy. Analysis with an antibody against the carboxyl terminus of PRKACA revealed a fluorescent signal that was consistently brighter throughout the cells in FL-HCC (Fig. 3E and fig. S1) than in the adjacent tissue normal liver (Fig. 3D).

In summary, we have provided evidence for a 400-kb heterozygous deletion on chromosome 19 in 10 out of 10 FL-HCC patients tested. We detected a chimeric DNAJB1-PRKACA RNA transcript in 12 of 12 patients tested and a putative chimeric DNAJB1-PRKACA protein in 14 of 14 patients tested. The genomic deletion, the chimeric transcript, and the chimeric protein were not present in any normal liver samples tested. This chimera is predicted to incorporate the J domain of DNAJB1 and the catalytic domain of PRKACA. The promoter is from the DNAJB1 gene, which could explain why the chimeric transcript is expressed at higher levels than the WT PRKACA transcript (Fig. 1, A, B, and C).

PKA is a heterotetramer composed of two regulatory subunits and two catalytic subunits. In this configuration, the catalytic subunit is inactive until cAMP binding causes its release from the regulatory units. The DNAJB1-PRKACA chimera retains the functional catalytic domain and maintains full kinase activity, but it is missing the domain that binds the regulatory subunits of PKA. PKA phosphorylates numerous cytoplasmic and nuclear substrates, including members of the Ras, mitogen-activated protein kinase (16), estrogen signaling (17), and apoptosis pathways (18). PKA is also involved in signaling via endothelial growth factor receptor (19) and regulation of aromatase expression (20), both of which can be overexpressed in FL-HCC (2124). PRKACA has been implicated in epithelial-mesenchymal transition, migration, and invasion of lung cancer cells (25). A review of publicly available data sets from the Cancer Genome Atlas (26, 27) suggests that PRKACA is amplified in 12% of ovarian serous cystadenocarcinoma (28); 5% of uterine corpus endometrial carcinoma (29); 3% of adenoid cystic carcinoma (30); 2% of lung squamous cell carcinoma (31); 1% of sarcoma (32), colon and rectum adenocarcinoma (33) and breast invasive carcinoma (34). In FL-HCC, the deletion we observe in chromosome 19 has not been reported in comparative genomic hybridization of FL-HCC (3538), perhaps because of the limited resolution of the approach at the time those studies were performed.

There are currently no molecular diagnostic tests for FL-HCC. Because previous studies have detected PKA in the peripheral blood of cancer patients (39), this chimera may represent a diagnostic marker for FL-HCC. Surgical resection remains the cornerstone of therapy and patients who present with advanced-stage or metastatic disease have few treatment options. While the role of the DNAJB1-PRKACA chimera in the pathogenesis of FL-HCC has yet to be addressed, our observations raise the possibility that it contributes to the pathogenesis of the tumor and may represent a therapeutic target.

Supplementary Materials

Materials and Methods

Fig. S1

Tables S1 to S3

References (4048)

References and Notes

  1. Acknowledgments: Two coauthors on this study are patients with FL-HCC. We thank S. Tavazoie and H. Goodarzi for trouble-shooting, recommending protocols and software, and for helpful discussions of the data and the manuscript. We thank R. Darnell for support, insightful discussions of the data, and critical comments on the manuscript. We also thank the Pathology core facility at Memorial Sloan-Kettering Cancer Center (MSKCC), the Molecular Cytology core facility (MSKCC), and members of the fibrolamellar community for support. This work was funded by a grant from The Fibrolamellar Cancer Foundation, The Rockefeller University Center for Clinical and Translational Science grant 2UL1RR024143, and by an anonymous donor. We thank J. Panda for fund-raising efforts. C.N.T was supported by a Howard Hughes Medical Institute International Student Predoctoral Fellowship. The New York Genome Center and the authors (N.R., S.G., A.-K.E., and V.V.) have filed a patent application related to the use of the chimeric transcript in the detection of human cancers. Sequence data have been deposited into the database of Genotypes and Phenotypes, dbGaP (accession no. phs000709.v1.p1).
View Abstract

Navigate This Article