Structural Variation
Structural variants (SVs) change larger pieces of the genome. A segment can be lost, copied, flipped, moved somewhere else or swapped between chromosomes. They start where indels stop, at about 50 bases (again, the cutoff varies from tool to tool), and the largest span millions of bases and can be seen under a microscope. The 1000 Genomes Project found 2,100 to 2,500 SVs in a typical genome, covering about 20 million bases. That's far fewer variants than SNVs or indels, but more bases changed. Long-read sequencing finds many more, about 27,600 per genome, mostly smaller ones in repetitive sequence that short reads miss.
Breakpoints and Junctions
However complicated an SV looks, it comes from the same few steps: the DNA breaks, and the loose ends are joined back together in a new order. The places where the sample stops matching the reference are the breakpoints. Each new join is a junction, connecting two stretches of the reference that aren't normally next to each other. A junction has a direction. It can carry on reading the same way, or it can fold back and continue on the opposite strand, as the reverse complement. A few extra bases that come from neither side are sometimes stitched in at a junction, too.
Any SV can be described as a list of its junctions, and the pattern of junctions tells you what kind of SV it is. One more distinction helps. A balanced SV moves sequence around without gaining or losing any, so every base is still there, just in a different order. An unbalanced SV adds or removes sequence, which changes how many copies of it the cell has (see Copy Number Variation).
Kinds of Structural Variants
- Deletion. A segment is lost and the sequence on either side is joined. It takes one junction, and it's unbalanced.
- Insertion. New sequence appears at a single point. Many large insertions are mobile elements, stretches of DNA that copy themselves into new places in the genome. Compared with the reference, a typical person carries about 915 Alu and 128 L1 insertions, the two most common kinds.
- Tandem duplication. A segment is copied, and the copy lands right after the original. It takes one junction, and it's unbalanced: the segment now has an extra copy.
- Inversion. A segment is cut out, flipped end to end and put back in the same place. Both of its junctions fold back. Nothing is gained or lost, so it's balanced.
- Translocation. A segment moves to a new place. Most often it moves to a different chromosome, but it can also move within the same one, which takes three junctions. The moved segment can keep its direction or arrive inverted. When two chromosomes trade pieces and nothing is lost, the translocation is balanced.
- Single breakend. Only one side of a junction can be placed. One end sits in the reference, and the other runs into sequence that can't be mapped, such as a long repeat or DNA missing from the reference altogether.
- Complex. Some events have several junctions that fit none of these patterns. The most extreme is chromothripsis, in which one or a few chromosomes shatter and are stitched back together in a single event, leaving tens to hundreds of rearrangements. It turns up in at least 2–3% of cancers and in about a quarter of bone cancers.
Finding Structural Variants
Most SVs are longer than a short read, so a read rarely shows a whole event. Callers look for three kinds of indirect evidence instead:
- Split reads. A read that crosses a junction can't be aligned in one piece. The aligner places part of it in one spot and the rest somewhere else, and records the second part as a separate, supplementary alignment. Split reads pin a junction down to the exact base.
- Discordant pairs. In paired-end sequencing, the two reads come from opposite ends of a DNA fragment of known size, so they normally map facing each other a predictable distance apart. When the fragment spans a junction, the pair maps too far apart (a deletion), too close together (an insertion), facing the wrong way (an inversion or tandem duplication) or on different chromosomes (a translocation).
- Read depth. Unbalanced SVs change how many reads cover a region: fewer over a deletion, more over a duplication. Copy Number Variation covers this in detail.
Callers combine these signals, and SVs are still harder to find than SNVs or indels. Junctions often fall in repeats, where short reads can't be placed with confidence, and balanced SVs leave read depth unchanged. Long reads can span a whole event, repeats included, which is why they find so many more SVs.
Gene Fusions
When an SV's two breakpoints land inside two different genes, the junction can join them into a single gene fusion. The breakpoints are usually in introns. When the fused gene is transcribed and spliced, exons from the first gene are followed by exons from the second.
A fusion can do harm in two main ways. It can make a hybrid protein that combines domains from both genes, as in BCR-ABL1 below. Or it can put one gene under the control of another gene's promoter. In about half of prostate cancers, the start of TMPRSS2, which is switched on by male hormones, is fused to ERG, so ERG is made far more than usual. Fusions don't need translocations, either. TMPRSS2 and ERG sit close together on chromosome 21, and most often the fusion forms when the stretch between them is deleted. The EML4-ALK fusion in lung cancer comes from a small inversion on chromosome 2.
A rearrangement can also switch a gene on without fusing anything. In Burkitt lymphoma, a translocation puts MYC next to an immunoglobulin gene. Immunoglobulin genes are very active in B cells, the cells the lymphoma grows from, and their control switches keep MYC on, too.
BCR-ABL1
In 1960, Peter Nowell and David Hungerford noticed an unusually small chromosome in the blood cells of people with chronic myeloid leukemia (CML), and it was named the Philadelphia chromosome, after their city. In 1973, Janet Rowley showed that it comes from a translocation between chromosomes 9 and 22. The exchange joins the start of BCR, on chromosome 22, to most of ABL1, on chromosome 9. The shortened chromosome 22 that carries the fused gene is the Philadelphia chromosome.
ABL1 is a tyrosine kinase, one of the kinases that switch other proteins on or off by adding phosphates. Normally it's kept off by its own first section, a cap that folds back onto the kinase domain. The fusion loses that cap, and BCR adds a section that makes the fusion proteins cluster together. Together, these keep the kinase permanently on, sending the signals that tell blood cells to grow and survive. BCR-ABL1 is enough to cause the disease: mice given bone marrow carrying the fused gene develop a CML-like illness.
BCR-ABL1 was also one of the first cancer drivers to be targeted by a drug. The drug blocks the fusion's kinase, switching off the signal that had been stuck on, and in its first trial in CML, 53 of 54 patients on the higher doses had their blood counts return to normal. Ten years later, 83% of patients who started on the drug were still alive.
Finding Fusions
Fusions are often looked for in RNA-seq rather than DNA. In DNA, the breakpoints can fall anywhere in introns that may be tens of thousands of bases long, often in repeats. In RNA, splicing removes the introns, so the fusion shows up as two known exons joined together, and finding it also confirms that the fused gene is expressed. Fusion callers look for reads that cross the exon–exon junction and for read pairs with one read in each gene, and many also check against lists of known fusions. In the clinic, common fusions such as BCR-ABL1 are also tested for directly, with targeted assays.