ABSTRACT
Fusion transcripts are thought to be somatic and associated with cancer. However, they have been observed in healthy tissues at high recurrent frequencies. We have used SCIF (SplicingCodes Identify Fusion Transcripts) to analyze RNA-Seq data from 727 multiple myeloma (MM) patients of the MMRF CoMMpass Study. MTG1-SCART1, SCART1-CYP2E1, and TPM4-KLF2 have been detected in 96.1%, 95.7%, and 92.2% of 727 MM patients and formed fusion gene signatures. MTG1-SCART1 and SCART1-CYP2E1 are read-through from the same locus and the two most recurrent epigenetic fusion genes (EFGs) out of 187 EFGs detected in ≥10% of 727 MM patients. TPM4-KLF2 fusion gene, which was initially thought to be somatic, has been shown by a monozygotic twin genetic model to be a hereditary fusion gene (HFG) and the dominant genetic factor associated with MM. This work provides the first line of evidence that HFGs are the genetic factors and EFGs reflect the consequences between genetics and environments during development.
Introduction
Multiple myeloma (MM) is a genetically complex and incurable blood malignancy characterized by the infiltration and growth of malignant plasma cells in the bone marrow1. Multiple myeloma affected approximately 427,000 people in 2013 and resulted in 79,000 deaths around the world2. In 2011, the Multiple Myeloma Research Foundation (MMRF) initiated the MMRF CoMMpass Study (MMRFCS), which performed comprehensive genomic analyses and followed each newly eligible patient longitudinally from initial diagnosis for ≥8 years3. Next-generation sequencing (NGS) of MM genomes had shown that MM was heterogeneous and was associated with significantly mutated genes and significant focal gene copy number gains and losses4-6. They resulted in definite, nonrandom chromosomal fusions of IGHC with the loci of FGFR3, CCND1, CCND3, MAF, and MAFB gene7,8. Analysis of 718 MMRFCS patients revealed that MM patients of European ancestries had higher mutations of TP53 and IRF4 than the African MM patients9. Cheynen et al. analyzed RNA-Seq data of a cohort of 255 newly diagnosed MM patients and identified 5.5 fusion genes, 29% of which were discovered in at least two patients10. Nasser et al. used RNA-Seq data from 806 samples with 704 matching WGS specimens and created 1192 of the observed fusions, a few of which were recurrent11. These two studies reflected overall obstacles and crises of RNA-Seq analyses because each software system discovered only small portions of fusion transcripts compromised by system errors and could not distinguish low expression fusion transcripts from “spurious” fusion transcripts10,12-14.
To overcome these problems and remove all system errors generated by experiments and software15-17, Zhou et al. used a structured table equal to 0.1% of human genome sequences. They implemented SCIF (SplicingCodes Identify Fusion Transcripts) to identify small subsets of fusion transcripts with maximum accuracy, eliminating “spurious” fusion transcripts18. Using SCIF, Zhou et al. discovered large numbers of fusion transcripts, among which the KANSARL (KANSL1-ARL17A) gene had been validated as the first predisposition (or hereditary) fusion gene specific to 28.9% of populations of European ancestry18. In addition, many studies have shown that fusion genes had been observed in cancer and healthy samples8,19,20, suggesting that fusion genes had many faces and required further investigation. To study fusion genes more precisely, we defined a hereditary fusion gene (HFG) as the fusion gene that offspring inherited from parents and excluded read-through fusion genes21 generated via transcriptional termination failure. Since environmental and physiological factors regulated read-through, we defined a read-through fusion gene as an epigenetic fusion gene (EFG)22. This report shows that TPM4-KLF2, MTG1-SCART1, and SCART1-CYP2E1 genes have been detected in 92.2%, 96.1%, and 95.7% of 727 MM patients, respectively, and are the HFG and EFG signatures associated with MM. TPM4-KLF2 is an HFG and MTG1-SCART1 and SCART1-CYP2E1 are EFGs.
Materials and Methods
Materials
Human multiple myeloma RNA-seq dataset
Multiple Myeloma Research Foundation (MMRF) initiated the MMRF CoMMpass Study in 2011. MMRF CoMMpass RNA-seq data (phs000748.v5.p4) were downloaded from controlled-access dbGaP (https://www.ncbi.nlm.nih.gov/gap). Seven hundred twenty-seven newly-diagnosed multiple myeloma patients were identified and used in this study. 99% of 727 RNA-seq data were from the bone marrow of multiple myeloma patients.
RNA-Seq dataset of Genotype-Tissue Expression (GTEx)
The RNA-Seq dataset of GTEx healthy blood samples had 427 healthy individual blood samples. RNA-Seq datasets of GTEx’s blood samples (dbGap-accession: phs000424.v7.p2) were downloaded from NCBI (https://www.ncbi.nlm.nih.gov/projects/gap/cgi-bin/study.cgi?study_id=phs000424.v7.p2).
Bone marrow samples
Four MM bone marrow samples and five healthy bone marrow samples were a gift from Prof. Yinxiong Li, Guangzhou Institutes of Biomedicine and Health, Chinese Academy of Sciences, China.
Primers
All primers used in this project were designed by Primer3Plus (https://www.bioinformatics.nl/cgi-bin/primer3plus/primer3plus.cgi).
Computers
All computations were performed on Linux Desktop computers with 8 to 12 GB memories.
Methods
Software
SCIF (SplicingCodes Identify Fusion Transcripts) had been described previously18.
Recurrent Fusion Frequencies (RFF)
The recurrent fusion frequency of a fusion gene was equal to the number of samples with at least one isoform divided by the total number of samples.
Classifications of fusion transcripts
The detailed protocols had been described previously22.
Isolation of total RNA from bone marrows
Total RNAs of Bone marrow tissues were isolated by the Trizol reagent (Invitrogen, CA) described previously18.
cDNA synthesis
cDNA synthesis was carried out as described by Zhou et al. 18.
End-point PCR amplification
PCR amplification, DNA cloning, and sequencing were performed using the following primers: MS001F GAGACCATGGCTGACTACCT; MS001R GGACCAACACCACTCCGT; SC001F AATTCCACGAACATGAGCCC; SC001R CACATGAATGGGGCCAGAAG; TK001F GAGCGGCGCGAGAAAGTG; TK001R GTGCTTTCGGTAGTGGCG. The protocols and procedures had been described previously18. Prof. Yinxiong Li and his lab members performed RT-PCR and DNA sequencing.
Results
A brief review of enormous numbers of MMRFCS fusion transcripts
The MMRFCS RNA-Seq data (phs000748.v5.p4) consisted of 727 newly-diagnosed multiple myeloma (MM) patients. The MMRFCS RNA-Seq data had 144,069 million RNA-Seq reads, the average of which was 198.2 million RNA-Seq reads per patient. We used SCIF to analyze the MMRFCS RNA-Seq data. We identified 641,680 fusion transcripts supported by 1,405,768 unassembled raw RNA-Seq reads, each of which had 20-61 bp sequences identical to 5’ and 3’ genes plus splice site signals and correct orientations. Under these conditions, we performed extensive computational simulations and showed that our methods had removed potential experimental and computational artifacts. The false-positive rate was about 1% for cancer data and 1-2% for healthy samples. All false positives were from repetitive DNA sequences, pseudogenes, and alternative splicing.
We classified the fusion transcripts based on their recurrent frequencies observed in 727 MMRFCS MM patients to identify the potential fusion genes associated with MM. We discovered that 518 fusion transcripts ranged from 10% to 96.1%, twenty-one of which had ≥50% recurrent frequencies (Supplemental Table 1), suggesting that many fusion transcripts were highly recurrent and potential MM biomarkers. Based on their characteristics, fusion transcripts were initially classified into read-through fusion transcripts21 and fusion genes23 formed via genomic alterations, including deletions, inversions, intra-chromosomal, and inter-chromosomal translocations. These 518 fusion transcripts were classified into 284 fusion transcripts produced via genomic alternations and 234 read-through fusion transcripts. Their expression levels were higher than that of KANSARL isoform 1 and much lower than that of KANSARL isoform 2 (Supplemental Figure 1). The IGHV-WHSC1 expression level was six-fold higher than that of KANSARL isoform 2 and was more than 100-fold higher than 99% of 518 fusion transcripts, suggesting that these 518 fusion transcripts were not highly expressed (Supplemental Figure 1). The average expression levels of the fusion genes were slightly lower than that of the read-through fusion genes. Out of 518 fusion transcripts, 284 fusion transcripts were encoded by 275 fusion genes generated by genomic alterations, which included deletions (5.1%), inter-chromosomal translations (69.1%), intra-chromosomal translocations (13.1%), and inversions (12.7%). Only four fusion genes had additional ≥10% recurrent isoforms, one of which was NCOR2-UBC having six additional isoforms.
On the other hand, 234 read-through fusion transcripts were encoded by 188 EFGs, 23.8% of which had additional isoforms with recurrent frequencies of ≥10%. The recurrent frequencies of the fusion genes produced by genomic rearrangements ranged from 10% to 92.2% of 727 MM patients, the average of which was 41.8%. In comparison, the EFGs ranged from 10% to 96.1% of 727 MM patients, the average of which was 52.9%. The former average recurrent frequency was 11.1% higher than the latter one. The higher EFG recurrent frequencies reflected that EFGs were more easily induced by cellular stresses, had a much longer time to be accumulated, and reflected many early events.
Since the OAZ1-KLF2 fusion gene with a recurrent frequency of 58.1% was the second highest fusion gene generated by genomic alterations, the recurrent frequency of TPM4-KLF2 was 34% higher than that of OAZ1-KLF2. Hence, TPM4-KLF2 was the dominant fusion gene produced via genomic alterations. Based on their gene names, it was immediately apparent that MTG1-SCART1 and SCART1-CYP2E1 were two epigenetic (read-through) fusion genes (EFGs) from the same locus and were detected in 96.1% and 95.7% of 727 MM patients, respectively. They were at least 16.8% higher than the two closest ACSS1-C20orf3 and ERGIC1-RPL26L1 EFGs, with a recurrent frequency of 78.9% (Supplemental Table 1).
To evaluate if TPM4-KLF2, MTG1-SCART, and SCART1-CYP2E1 were associated with MM, we used the GTEx blood samples as healthy control. Table 1 showed that TPM4-KLF2, MTG1-SCART1, and SCART1-CYP2E1 were detected in 1%, 28.3%, and 10.8% of 427 GTEx blood samples, respectively. They were 92-fold, three-fold, and nine-fold higher than the GTEx’s ones. Statistical analysis showed these differences were significant (p<0.00002) and indicated that TPM4-KLF2, MTG1-SCART1, and SCART1-CYP2E1 were associated with MM. These data confirmed that TPM4-KLF2 was the dominant fusion gene formed via genomic deletion, while MTG1-SCART1 and SCART1-CYP2E1 were the dominant EFGs.
Recurrent fusion frequencies of TPM4-KLF2 HFG and MTG1-SCART1 and SCART1-CYP2E1 EFG in MM patients and GTEx blood samples.
Validation of TPM4-KLF2 fusion gene as a hereditary fusion gene (HFG) associated with MM
Fig.2a shows a schematic drawing of the standard genomic structure of the TPM4-KLF2 region located on 19p13.1. A duplicated region underwent deletion and generated a putative TPM4-KLF2 fusion gene genomic structure (Fig.1b), which was transcribed into putative TPM4-KLF2 pre-mRNAs (Fig.1c). As shown in Fig.1d, TPM4 exon 1b was spliced with KLF2 exon 3 to produce the TPM4-KLF2 main isoform. This TPM4-KLF2 isoform was predicted to produce a putative 97 aa tropomyosin 4 and kruppel-like transcription factor 2 hybrid protein (Supplemental Figure 2). This hybrid protein had a head of 47 aa coiled-coil region of tropomyosin 4 and a tail of 50 aa region of kruppel-like transcription factor 2. The subcellular protein locations predicted by PSORTII24 were in the nucleus and less likely in mitochondria and were more similar to those of kruppel-like transcription factor 2 than tropomyosin 4’s ones.
a). Schematic diagram of typical TPM4→KLF2 genomic structures. Horizontal red and black arrows represented TPM4 and KLF2 genes, respectively. Horizontal gray arrows indicated genes between TPM4 and KLF2 genes; b). A potential deletion resulted in the TPM4-KLF2 fusion gene; c) A putative TPM4-KLF2 pre-mRNA structure. Black triangles represented introns. Numbers above the rectangles indicated exon positions of TPM4 and KLF2 genes, respectively; d). TPM4-KLF2 mRNA was from their fusion junction. The sequences under the solid boxes were from RNA-Seq. e). RT-RNA PCR amplification of TPM4-KLF2 fusion gene. f). DNA sequencing of cloned RT-PCR products validated the TPM4-KLF2 fusion gene. Red and black boxes represented TPM4 and KLF2 exons, respectively. The red and black letters were TPM4 and KLF2 sequences, respectively.
a) Schematic diagram of the genomic structure of PAOX→MTG1→SCART1→CYP2E1. Horizontal gray, red, black, and green arrows represented PAOX, MTG1, SCART1, and CYP2E1, respectively; b) Failed transcriptional terminations may produce MTG1→ SCART1→CYP2E1 pre-mRNA transcripts. Red, black and green rectangles represented MTG1, SCART1, and CYP2E1 exons. Black triangles represented introns; c) MTG1-SCART1 fusion transcripts. Red and black rectangles indicated MTG1 and SCART1 exons, respectively. The red and black letters were MTG1 and SCART1 exon sequences, respectively. The sequences under the rectangles were from RNA-Seq; d) SCART1-CYP2E1 fusion mRNA. The sequences under the rectangles were from RNA-Seq. Black and green rectangles indicated SCART1 and CYP2E1 exons, respectively. The black and green letters were SCART1 and CYP2E1 exon sequences, respectively; e) RT-PCR amplification of MTG1-SCART1 and SCART1-CYP2E1 fusion transcripts of MM patients; f). DNA sequencing of cloned RT-PCR products validated the MTG1-SCART1 fusion gene. g). DNA sequencing of cloned RT-PCR products validated the SCART1-CYP2E1 fusion gene. Vertical arrows indicated the fusion junctions.
As Table 1 showed, the TPM4-KLF2 fusion gene had been observed only in four individuals out of 427 GTEx blood samples, which was about 1% of the GTEx’s blood samples, suggesting that TPM4-KLF2 fusion transcripts were produced from a potentially rare somatic deletion. To validate the TPM4-KLF2, we first selected four MM patients as a starting point. We independently performed RT-PCR amplification on the bone marrow RNAs isolated from the four MM patients, all of whom were TPM4-KLF2 positive. Fig.1e showed one of the four RT-PCR amplifications. All four RT-PCR products were cloned and sequenced independently. Fig.1f showed one of four TPM4-KLF2 sequences and demonstrated that the sequence had the identical TPM4-KLF2 fusion junction observed from RNA-Seq data. To study TPM4-KLF2 fusion gene frequencies in healthy control, we performed RT-PCR analyses on the five healthy bone marrow samples, and consequently, five out of the five healthy controls were shown to be TPM4-KLF2 positive. The healthy samples being TPM4-KLF2 positive were significantly higher than the GTEx one. A literature review showed that TPM4-KLF2 was first reported in acute lymphoblastic leukemia by Roberts et al.25. Locher et al. reported that TPM4-KLF2 was detected in 30% of acute myeloid leukemia samples, and all three normal bone marrow samples26.
These high healthy TPM4-KLF2 frequencies contradicted our original idea that TPM4-KLF2 was a somatic genomic alteration and led us to speculate that TPM4-KLF2 was similar to the hereditary KANSARL fusion gene. Such high TPM4-KLF2 frequency compelled us to use monozygotic (MZ) twins with identical genetic materials as a genetic model to study human hereditary fusion genes (HFGs) systematically 22. We had identified a total of 1180 HFG from 37 pairs of MZ twins22. Interestingly, TPM4-KLF2 was detected in 15 (40.5%) of 37 pairs of MZ twins and 54.1% of 74 twin siblings22. These data had mathematically ruled out that TPM4-KLF2 was from a somatic genomic deletion. Hence, we could determine that TPM4-KLF2 was a hereditary fusion gene.
To investigate how many fusion genes generated via genomic alterations were potential HFGs, we first overlapped the top five fusion genes with recurrent frequencies of ≥50% with 1180 MZ twin HFGs. Four of them were present in the MZ twin HFGs and were TPM4-KLF2, OAZ1-KLF2, TRNAN35-FAM91A3P, and FOSB-KLF2. Only VPS37B-UBC was not present among 1180 MZ HFGs. Next, we overlapped 275 fusion genes with recurrent frequencies of ≥10% with 1180 MZ HFGs, respectively. 87 (31.6%) of 275 fusion genes overlapped with the MZ HFGs and hence, were HFGs (Supplemental Table 2). Since the MZ HFGs were only tiny portions of potential 5×109 HFGs22, real HFGs associated with MM would be significantly higher. Since somatic fusion gene IGHV-WHSC1 had only 8.5% recurrent frequency and four of the five fusion genes with recurrent frequencies of ≥50% were HFGs, we estimated that ≥80% of 275 fusion genes generated via genomic alterations were inherited from their parents and not from somatic genomic abnormalities, suggesting that there were ≥220 potential HFGs associated with MM. These large numbers of HFGs and potential HFGs associated with MM suggested that multiple myeloma was a complex genetic disease.
Characterization of MTG1-SCART1 and SCART1-CYP2E1 EFGs
The most frequently observed fusion genes were MTG1-SCART1 and SCART1-CYP2E1, detected in 96.1% and 95.7% of 727 MMRFCS MM patients. Based on their gene names, it was immediately apparent that MTG1-SCART1 and SCART1-CYP2E1 fusion transcripts were read-through products from the same locus. Fig.2a showed that MTG1, SCART1, and CYP2E1 genes (red, black, and green rectangles represented MTG1, SCART1, and CYP2E1) clustered together located on the 10q24.3 plus-strand. MTG1, SCART1, and CYP2E1 genes encoded mitochondrial GTPase 1, scavenger receptor 1, and cytochrome P450 2E1. Failed transcriptional terminations produced a single read-through pre-mRNA containing three genes or two separate MTG1-SCART1 and SCART1-CYP2E1 read-through pre-mRNAs (Fig.2b).
Fig.2c showed that cis-splicing of MTG1-SCART1 read-through pre-mRNAs resulted in MTG1-SCART1 isoform (designed as MTG1-SCART1 isoform 1) as discovered from the above RNA-Seq analysis. MTG1-SCART1 isoform 1 would contain the first ten MTG1 exons and the last 14 SCART1 exons and produce an open frame of 461 aa hybrid polypeptide, which consisted of 256 aa mitochondrial GTPase 1 and 205 aa scavenger receptor 1 with an isoelectric point of 10.17 and molecular weight of 50 kd. PSORTII24 predicted 34.8%, 30.4%, and 17.4% in the cytoplasm, nucleus, and mitochondria, respectively, significantly different from mitochondrial GTPase 1 of 73.9% and 13% in the cytoplasm and nucleus21. On the other hand, Fig.2d showed that the exon 1-15 of SCART1 was putatively spliced together with the exon 2-10 of CYP2E1 to produce the SCART1-CYP2E1 isoform (named SCART1-CYP2E1 isoform 1). Since the SCART1-CYP2E1 fusion junction sequences were located at the SCART1 3’ UTR region (Fig.2d), the SCART1-CYP2E1 isoform 1 would produce a normal scavenger receptor 1. Based on their fusion junctions, MTG1-SCART1 and SCART1-CYP2E1 were predicted to produce a mitochondrial GTPase 1-scavenger receptor 1 hybrid protein and a normal scavenger receptor 1, respectively.
To validate MTG1-SCART1 isoform 1 and SCART1-CYP2E1 isoform 1, we independently isolated total RNAs from the bone marrow of four MM patients. Fig.2e showed that one of four MM patient samples was used for RT-PCR amplification on MTG1-SCART1 and SCART1-CYP2E1. Four MM patients’ RT-PCR products were isolated, cloned, and sequenced independently. Fig.2f showed that one of four cloned RT-PCR MTG1-SCART1 sequences had identical fusion junction sequences from RNA-Seq. Similarly, Fig.2g showed that one of four MM patients had fusion sequences identical to SCART1-CYP2E1 isoform 1.
Table 1 showed that MTG1-SCART1 and SCART1-CYP2E1 in MMRFCS MM patients in the GTEx blood samples were three-folds and nine-folds more frequent than those in the GTEx blood samples. MTG1-SCART1 and SCART1-CYP2E1 were detected in 28.3% and 10.8% of the GTEx blood samples. The former was much more frequently detected than the latter and 2.6 folds higher, suggesting that MTG1-SCART1 and SCART1-CYP2E1 were not co-expressed and differentially regulated in the healthy population. Since TPM4-KLF2, MTG1-SCART1, and SCART1-CYP2E1 in MMRFCS MM patients had almost identical recurrent frequencies, TPM4-KLF2 might have affected the expression of MTG1-SCART1 and SCART1-CYP2E1. To investigate if TPM4-KLF2 was associated with their expression, we compared their gene expression patterns with those of 74 MZ twin siblings. Supplemental Table 3 showed that MTG1-SCART1 and SCART1-CYP2E1 were detected in 48.6% and 18.9% of 74 MZ siblings, respectively. The MTG1-SCART1 recurrent frequency was 2.6 folds of that of SCART1-CYP2E1.
In comparison, MTG1-SCART1 and SCART1-CYP2E1, detected in 96.1% and 95.7% of the MM patients, had almost identical recurrent frequencies and were almost co-expressed. The differences between MM patients and MZ twins suggested that their expressions were not directly associated with TPM4-KLF2. Other genetic and environmental factors contributed to their expression. Therefore, MTG1-SCART1 and SCART1-CYP2E1 were the unique EFG signatures specific to MM and could be used as potential biomarkers to monitor MM development and progression.
Like a traditional gene, these EFGs could undergo alternatively splicing. Supplemental Table 4 showed that MTG1-SCART1 encoded an additional 17 MTG1-SCART1 isoforms, while SCART1-CYP2E1 had additional eight isoforms (Supplemental Table 5). In addition, we found that MTG1-CYP2E1 fusion transcripts had been detected in 204 (28.1%) of 727 MM patients (Supplementary Table 1). However, no single copy of the MTG1-CYP2E1 fusion transcript had been observed in the GTEx blood samples, suggesting that MTG1-CYP2E1 reflected the MM progression. In a recent report, MTG1-CYP2E1 was a potential pathogenetic link between angiomyofibroblastoma and superficial myofibroblastoma in the female lower genital tract27. Since PAOX-MTG1 had been annotated as a PAOX-MTG1 EFG21, the PAOX-CYP2E1 locus had produced four EFGs, which underwent extensive alternative splicing and dramatically increased biological and functional diversities.
Discussion
This report used SCIF to analyze 727 MMRFCS RNA-Seq data and identified 641,680 fusion transcripts. Based on their recurrent frequencies, we discovered 518 fusion transcripts ranging from 10% to 96.1%, suggesting that enormous fusion genes may be associated with MM. Based on the types of fusion transcripts, fusion genes were classified into read-through21 and fusion genes generated via genomic alterations14. The expression levels of both EFGs and fusion genes were between those of KANSARL (KANSL1-ARL17A) isoform 1 and 218, suggesting that these recurrent EFGs and fusion genes were not highly expressed. It seemed that they had undergone a natural selection and were tamed. Among the high recurrent fusion genes, TPM4-KLF2, MTG1-SCART, and SCART1-CYP2E1 fusion genes were detected in ≥92% of 727 MMRFCS MM patients, had much higher recurrent frequencies than the rest of them, and formed the fusion gene signatures associated with MM.
When we performed RT-PCR amplifications on the healthy control, all five individuals were also TPM4-KLF2 positive. This high TPM4-KLF2 frequency led us to speculate that the TPM4-KLF2 fusion gene was similar to the KANSARL predisposition fusion gene18 and compelled us to use MZ twins to study HFGs systematically. We analyzed RNA-Seq data of 37 pairs of MZ twins and discovered 1180 HFGs22. Coincidently, one of many HFGs associated with MZ twin inheritance was TPM4-KLF2, which was detected in 15 of 37 pairs of MZ twins and 54.1% of 74 twin siblings22. The MZ twins’ high TPM4-KLF2 recurrent frequency had mathematically ruled out that TPM4-KLF2 was a somatic fusion gene. Hence, TPM4-KLF2 was determined to be a hereditary fusion gene. TPM4-KLF2 was first reported in 201225 and was later found in 30% of acute myeloid leukemia (AML) samples and all three normal bone marrow samples26, suggesting that TPM4-KLF2 was associated with AML. Hence, TPM4-KLF2 was a broad-spectrum HFG associated with multiple types of cancer and MZ genetics. PIM3-SCO226,28, NCOR2-UBC26, and OAZ1-KLF225,26 were previously-studied cancer fusion genes, also associated with the MZ twin inheritance and HFGs22. Hence, there was no such thing as a good HFG or a bad HFG whether an HFG caused what type of disease depended on other HFGs, developments, and environments.
To further understand the HFG role in MM, we showed that TPM4-KLF2, OAZ1-KLF2, TRNAN35-FAM91A3P, and FOSB-KLF2 were HFGs. TPM4-KLF2, OAZ1-KLF225,26, and FOSB-KLF2 with recurrent frequencies of≥50% suggested that disrupting KLF2 expression may be associated with MM. To check if there were more HFGs associated with MM, we overlapped 275 fusion genes of ≥10% were overlapped with 1180 MZ HFGs. 87 (31.6%) of them were HFGs. Since four of the top five fusion genes were HFGs and IGHV-WHSC1 had only 8.5% recurrent frequency, ≥80% of 275 fusion genes were predicted to be putative HFGs. Many previously-characterized somatic fusion genes, such as IGHV-WHSC18 and driver oncogenes, had much lower recurrent frequencies and were much later events. Hence, enormous numbers of HFGs were involved in MM and suggested that MM was a complex genetic disease, and TPM4-KLF2 HFG was the dominant genetic factor identified so far and a potential therapeutic target.
MTG1-SCART1 and SCART1-CYP2E1 were two EFGs from the same locus detected in 96.1% and 95.7% of MM patients, suggesting that both were almost co-expressed in MM. The MTG1-SCART1 and SCART1-CYP2E1 main isoforms potentially encoded a novel mitochondrial GTPase 1-scavenger receptor 1 hybrid protein and a normal scavenger receptor 1 protein, respectively, suggesting that both main isoforms had two different potential consequences. In addition, MTG1-SCART1 and SCART1-CYP2E1 EFGs had been found to have 18 and 9 isoforms detected in MM, respectively, which dramatically increased potential biological diversities. We compared data from MM patients and MZ twins to study if TPM4-KLF2 was directly associated with their expression. Supplemental Table 3 showed that MTG1-SCART1 and SCART1-CYP2E1 were detected in 48.6% and 18.9% of 74 MZ twin siblings. The former recurrent frequency was 2.6 folds of the latter, indicating that their patterns in MZ twins significantly differed from those in the MM patients. MTG1-SCART1 and SCART1-CYP2E1 had almost identical recurrent frequencies in MMRFCS MM and were the unique EFG signature specific to MM.
TPM4-KLF2 HFG and MTG1-SCART1 and SCART1-CYP2E1 EFGs had formed the HFG and EFG signatures of MMRFCS MM. In order to initiate MM, cells must have TPM4-KLF2 HFG and a set of other MM HFGs (Fig.3a). Expression of these sets of HFGs resulted in read-through pre-mRNAs to produce EFGs such as MTG1-SCART1 and SCART1-CYP2E1 EFGs and maybe activated some dormant HFGs (Fig.3b). Without treatments and interventions, these cells expressed excessive EFGs and HFGs, resulted in genomic alternations, and produced somatic fusion genes. Cells underwent natural selection to lead to MM developments and progression (Fig3c). In comparison, if healthy individuals had only TPM4-KLF2 HFG and no other MM HFGs (Fig.3d), cells expressed much fewer EFGs and could tolerate more stresses (Fig.3e). Hence, they maintained typical genomic structures and produced few somatic gene mutations (Fig.3f). Fig.3 showed that individuals must have both a set of MM HFGs and favorable environments to initiate MM. Without any one of them, the cells would be divided typically. Therefore, we could use a drop of blood or a few cells to screen HFGs and EFGs associated with MM as early as the middle of pregnancy to monitor MM development and progression. Meanwhile, we could have enough time to develop different approaches or therapeutic drugs to prevent cell progression to MM. Hence, applications of HFGs and EFGs will potentially revolutionize cancer research, diagnosis, and treatments and potentially mark the beginning of a new era of cancer and other disease research.
a). Cells with TPM4-KLF2 and other HFGs produced fusion or truncated proteins and altered gene expression; b). During their development, the cells with altered gene expression promoted EFG expressions such as MTG1-SCART1 and SCART1-CYP2E1 and maybe activated some dormant HFGs to be expressed; c). The cells with excessive EFG expressions resulted in genomic rearrangements producing somatic fusion genes. The cells began to accumulate somatic fusion genes and gene mutations and underwent natural selection. Finally, the cells were amplified uncontrollably and led to develop MM; d). The cells from healthy individuals had no key HFGs associated with MM, resulting in fewer gene expression errors; e). The cells with more minor expression errors resulted in fewer cellular restraints represented by fewer EFGs; f). The cells with fewer EFG expressions maintained normal DNA repair mechanisms and had fewer gene mutations and genomic rearrangements. The green rectangle indicates the normal genes. The solid red and green rectangle presented the key HFG, such as TPM4-KLF2. The open black and red box showed the minor sets of HFG. The light green and blue rectangles and blue and green rectangles represented EFGs such as MTG1-SCART1 and SCART1-CYP2E1. The black and red rectangles and red and black rectangles indicated somatic fusion genes and gene mutations.
Disclosures
No authors declare a relevant conflict of interest.
Figure Legends
Supplementary Figure 1 Expression Levels of fusion transcripts with recurrent frequencies of ≥50%. KANSARL isoforms 1 and 2 were used as references.
Supplementary Figure 2 Hybrid protein encoded by TPM4-KLF2 HFG Underlined sequences represented the tropomyosin 4 amino acids. Asterisk was a stop codon.
Supplemental Table 1 Lists of fusion transcripts with recurrent frequencies of ≥10% discovered in 727 MMRFCS patients
Supplemental Table 2 Hereditary fusion gene list identified by overlapping between 1180 monozygotic twin hereditary fusion genes and fusion genes with recurrent frequencies of ≥10
Supplemental Table 3 Comparison of MTG1-SCART1 and SCART1-CYP2E1 recurrent frequencies between multiple myeloma and monozygotic twins
Supplemental Table 4 MTG1-SCART1 Isoform lists discovered in 727 MMRFCS patients.
Supplemental Table 5 SCART1-CYP2E1 Isoform lists discovered in 727 MMRFCS patients.
Acknowledgments
We have expressed our most profound appreciation to Ms. Kathy Giust for her private information and support. We profoundly appreciate Prof. Yinxiong Li and his lab members’ contribution to this work.
