Gene Fusion Detection

Overview

Gene fusions often result from large genomic rearrangements such as structural variants. While WGS secondary analysis pipelines typically contain alignment and variant calling stages, very few of them contain dedicated gene fusion callers. When they are included, they are usually associated with RNA-Seq pipelines where gene fusions can be readily observed.

Since gene fusions are frequently observed in cancer and since many sequencing experiments do not include paired RNA-Seq data, we have added gene fusion detection and annotation to Illumina Connected Annotations.

The rich diversity in gene fusion architectures and their likely mechanisms can be seen below:

Publication

Kumar-Sinha, C., Kalyana-Sundaram, S. & Chinnaiyan, A.M. Landscape of gene fusions in epithelial cancers: seq and ye shall find. Genome Med 7, 129 (2015)

Approach

Illumina Connected Annotations uses structural variant calls to evaluate if they form either putative intra-chromosomal or inter-chromosomal gene fusions. Let's consider two transcripts, NM_014206.3 (TMEM258) and NM_013402.4 (FADS1). Both of these genes are on the reverse strand in the genome. The vertical bar indicates the breakpoint where these transcripts are fused:

The above explains where the transcripts are fused together, but it doesn't explain in which orientation. By using the directionality encoded in the translocation breakend, we can rearrange these two transcripts in four ways:

Only two of the combinations yields a fusion containing both the transcription start site (TSS) and the stop codon. In one case, we can even detect an in-frame gene fusion. If only unidirectional gene fusions are desired, only these two fusions can be detected. If enable-bidirectional-fusions is enabled, all four cases can be identified.

Interpreting translocation breakends

At first glance, translocation breakends are a bit daunting. However, once you understand how they work, they're actually quite simple. For more information, we recommend reading section 5.4 in the VCF 4.2 specification.

REF	ALT	Meaning
s	t[p[	piece extending to the right of p is joined after t
s	t]p]	reverse comp piece extending left of p is joined after t
s	]p]t	piece extending to the left of p is joined before t
s	[p[t	reverse comp piece extending right of p is joined before t

Variant Types

Specifically we can identify gene fusions from the following structural variant types:

• deletions (<DEL>)
• tandem_duplications (<DUP:TANDEM>)
• inversions (<INV>)
• translocation breakpoints (AAAAAAAAAAAAAAAAAATTAGTCAGGCAC[chr3:153444911[)

Criteria

The following criteria must be met for Illumina Connected Annotations to identify a gene fusion:

1. After accounting for gene orientation and genomic rearrangements, both transcripts must have the same orientation if enable-bidirectional-fusions is not enabled. They can have the same or different orientations if enable-bidirectional-fusions is set.
2. Both transcripts must be from the same transcript source (i.e. we won't mix and match between RefSeq and Ensembl transcripts)
3. Both transcripts must belong to different genes
4. Both transcripts cannot have a coding region that already overlaps without the variant (i.e. in cases where two genes naturally overlap, we don't want to call a gene fusion)

ETV6/RUNX1 Example

ETV6/RUNX1 is the most common gene fusion in childhood B-cell precursor acute lymphoblastic leukemia (ALL). Samples with this translocation are associated with a good prognosis and excellent response to treatment.

Publication

Sun C., Chang L., Zhu X. Pathogenesis of ETV6/RUNX1-positive childhood acute lymphoblastic leukemia and mechanisms underlying its relapse. Oncotarget. 2017; 8: 35445-35459

VCF

Here's a simplified representation of the translocation breakends called by the Manta structural variant caller:

##fileformat=VCFv4.1
#CHROM  POS     ID      REF     ALT     QUAL    FILTER  INFO
chr12   12026270    .   C   [chr21:36420865[C   .   PASS    SVTYPE=BND
chr12   12026305    .   A   A]chr21:36420571]   .   PASS    SVTYPE=BND
chr21   36420571    .   C   C]chr12:12026305]   .   PASS    SVTYPE=BND
chr21   36420865    .   C   [chr12:12026270[C   .   PASS    SVTYPE=BND

When you put these calls together, the resulting genomic rearrangement looks something like this:

JSON Output

The annotation for the first variant in the VCF looks like this:

{"positions":[
{
  "chromosome": "12",
  "position": 12026270,
  "refAllele": "C",
  "altAlleles": [
    "[chr21:36420865[C"
  ],
  "filters": [
    "PASS"
  ],
  "cytogeneticBand": "12p13.2",
  "variants": [
    {
      "vid": "12-12026270-C-[chr21:36420865[C",
      "chromosome": "12",
      "begin": 12026270,
      "end": 12026270,
      "isStructuralVariant": true,
      "refAllele": "C",
      "altAllele": "[chr21:36420865[C",
      "variantType": "translocation",
      "transcripts": [
        {
          "transcript": "ENST00000396373.4",
          "source": "Ensembl",
          "bioType": "mRNA",
          "introns": "5/7",
          "geneId": "ENSG00000139083",
          "hgnc": "ETV6",
          "consequence": [
            "transcript_variant",
            "unidirectional_gene_fusion"
          ],
          "impact": "modifier",
          "geneFusions": [
            {
              "transcript": "ENST00000437180.1",
              "bioType": "mRNA",
              "source": "Ensembl",
              "geneId": "ENSG00000159216",
              "proteinId": "ENSP00000409227.1",
              "intron": 2,
              "hgnc": "RUNX1",
              "hgvsr": "ENST00000437180.1(RUNX1):r.?_58+274::ENST00000396373.4(ETV6):r.1009+3367_?",
              "directionality": "unidirectional"
            },
            {
              "transcript": "ENST00000300305.3",
              "bioType": "mRNA",
              "source": "Ensembl",
              "isCanonical": true,
              "geneId": "ENSG00000159216",
              "proteinId": "ENSP00000300305.3",
              "intron": 1,
              "hgnc": "RUNX1",
              "hgvsr": "ENST00000300305.3(RUNX1):r.?_58+274::ENST00000396373.4(ETV6):r.1009+3367_?",
              "directionality": "unidirectional"
            }
          ],
          "isCanonical": true,
          "proteinId": "ENSP00000379658.3"
        },
        {
          "transcript": "NM_001987.5",
          "source": "RefSeq",
          "bioType": "mRNA",
          "introns": "5/7",
          "geneId": "2120",
          "hgnc": "ETV6",
          "consequence": [
            "transcript_variant",
            "unidirectional_gene_fusion"
          ],
          "impact": "modifier",
          "geneFusions": [
            {
              "transcript": "NM_001754.5",
              "bioType": "mRNA",
              "source": "RefSeq",
              "isCanonical": true,
              "geneId": "861",
              "proteinId": "NP_001745.2",
              "intron": 2,
              "hgnc": "RUNX1",
              "hgvsr": "NM_001754.5(RUNX1):r.?_58+274::NM_001987.5(ETV6):r.1009+3367_?",
              "directionality": "unidirectional"
            }
          ],
          "isCanonical": true,
          "proteinId": "NP_001978.1"
        }
      ]
    }
  ]
}
]}

Field	Type	Notes
transcript	string	transcript ID
bioType	string	descriptions of the biotypes from Ensembl
exon	int	exon that contained fusion breakpoint
intron	int	intron that contained fusion breakpoint
geneId	string	gene ID. e.g. ENSG00000116062
hgnc	string	gene symbol. e.g. MSH6
hgvsr	string	HGVS RNA nomenclature

Gene Fusion Data Sources

To provide more context to our gene fusions, we provide the following gene fusion data sources:

• COSMIC
• FusionCatcher

Consequences

When a gene fusion is identified, we add the following Sequence Ontology consequence:

              "consequence": [
                "transcript_variant",
                "gene_fusion"
              ],

• If both transcripts have the same orientation, we label it as unidirectional_gene_fusion, if they have different orientations, we label it as bidirectional_gene_fusion
• If both unidirectional and bidirectional ones are detected, we label it as gene_fusion.

Gene Fusions Section

The geneFusions section is contained within the object of the originating transcript. It will contain all the pairwise gene fusions that obey the criteria outline above. In the case of ENST00000396373.4, there 7 other Ensembl transcripts that would produce a gene fusion. For NM_001987.4, there was only one transcript (NM_001754.4) that produce a gene fusion.

For each originating transcript, we report the following for each partner transcript:

• transcript ID
• gene ID
• HGNC gene symbol
• transcript bio type (e.g. protein_coding)
• intron or exon number containing the breakpoint
• HGVS RNA notation
• gene fusion directionality

tip

Before Illumina Connected Annotations 3.15, we provided HGVS coding notation. However, HGVS r. notation is more appropriate for these types fusion splicing events (see HGVS SVD-WG007).

          "geneFusions": [
            {
              "transcript": "NM_001754.4",
              "bioType": "protein_coding",
              "intron": 2,
              "geneId": "861",
              "hgnc": "RUNX1",
              "hgvsr": "NM_001754.4(RUNX1):r.?_58+274::NM_001987.4(ETV6):r.1009+3367_?",
              "directionality":"uniDirectional"
            }
          ],

The HGVS RNA notation above indicates that the gene fusion starts with NM_001754.4 (RUNX1) until CDS position 58 and continues with NM_001987.4 (ETV6). 1009+3367 indicates that the fusion occurred 3367 bp within intron 2.