Variants on protein structures part 4: MET exon 14 skipping

Some variants change no amino acid. They break a splice site, the cell splices the exon out of the RNA, and the protein comes out shorter. In lung cancer, many different DNA changes around exon 14 of MET all make the RNA join exon 13 straight to exon 15. Exon 14 encodes tyrosine 1003, which the enzyme CBL binds to mark the MET receptor for degradation (Kong-Beltran et al. 2006). We look at reads from two lung cancer cell lines in the Cancer Cell Line Encyclopedia: NCI-H596, which skips the exon, and EBC-1, which keeps it. We:

• open MET in the protein browser beside structure 3BUX, CBL bound to the stretch of MET around tyrosine 1003
• add DNA and RNA-seq reads as alignment tracks, with arcs for the splice junctions
• highlight exon 14 and follow it onto the structure
• copy the view to the splice site and sort the DNA reads on its first intron base

Prerequisites#

• for your own JBrowse, the protein3d and msaview plugins from the plugin store; genomes.jbrowse.org already loads them
• for your own reads, hisat2 and samtools

# Debian/Ubuntu
sudo apt install hisat2 samtools
# or conda, which has every tool above
conda install -c bioconda -c conda-forge hisat2 samtools

The data#

Nothing to download: the tracks read the BAM files by URL. They hold the reads around MET exons 13 to 15 from CCLE (Ghandi et al. 2019), aligned to hg38 with HISAT2. A BAM of your own, with its .bai index beside it, loads the same way.

Aligning your own RNA-seq reads#

A splice-aware aligner writes each read that crosses an intron as one alignment with a gap, and JBrowse draws those gaps as junction arcs. HISAT2 writes a sorted BAM that JBrowse loads as an alignments track:

# -x: a HISAT2 index of hg38; -1/-2: the read pairs
hisat2 -p 8 -x hg38 -1 reads_1.fq.gz -2 reads_2.fq.gz | samtools sort -o rna.bam -

# the .bai index lets JBrowse fetch one region at a time
samtools index rna.bam

Opening MET beside CBL#

The protein browser opens a gene's exons, an alignment of the protein across species and a structure, linked:

• open genomes.jbrowse.org/protein-browser/?gene=MET&ref=9606&structure=3bux and click Open in JBrowse ↗
• pick Assembly in the protein view's menu, then the entry for assembly 5, marked dimeric. 3BUX opens on its first assembly, the MET stretch alone; assembly 5 adds CBL
• close the MET orthologs view with the × on its title bar, and the ClinVar and AlphaMissense tracks with the × on each track's label
• type chr7:116771458-116771694 chr7:116771809-116772029 chr7:116774841-116775151 into the location box, which keeps exons 13, 14 and 15 with a little intron on each side

Adding the reads#

Each BAM goes in through File → Open track..., given its URL. Its .bai index is found beside it:

• open https://jbrowse.org/demos/metex14/NCI-H596_DNA.bam, then NCI-H596_RNA.bam and EBC-1_RNA_5pct.bam from the same folder
• on each RNA-seq track, turn off Show → Pileup and turn on Show → Sashimi read counts in the track menu, which leaves the coverage and an arc for each junction, labelled with its number of reads

Following exon 14 onto the structure#

A highlight on the genome reaches every linked view:

• drag across exon 14 on the scale bar and pick Highlight region
• click the link chip at the top of the band, pick Edit label... and type exon 14

The residues of 3BUX that exon 14 encodes take the highlight. Rotate the structure until they face you; they lie in a groove on CBL.

The cause at the splice site#

The first two bases of an intron are almost always GT, and the splicing machinery reads them to find the end of the exon. To look for a change there in the DNA reads, copy the view to the end of exon 14:

• pick View options → Copy view in the view's menu; the copy opens under the first view, so scroll down to it and close its RNA-seq tracks
• type chr7:116771975-116772005 into the copy's location box
• right-click a read at chr7:116,771,990 and pick Sort by → Base

In panel A, the NCI-H596 arcs leave exon 13 and land on exon 15, over a dip in coverage across exon 14, while the EBC-1 arcs enter exon 14 and leave it. In panel B, every NCI-H596 read has T where the reference has the G of the intron's GT. Without exon 14 the receptor lacks tyrosine 1003, the residue CBL binds in panel C, so CBL marks less of the receptor for degradation (Kong-Beltran et al. 2006).

Your own reads#

• start from the arcs: an arc that jumps an exon in one sample and not in another is the skip, whatever its cause
• then open the DNA reads at the exon's two ends; a skip can come from a base change at the splice site, as here, or a deletion that removes it
• any gene the protein browser opens takes the same steps, and Launch protein view in a gene's right-click menu picks another structure

Reproduce it end to end#

scripts/build_metex14.sh builds the three BAM files from CCLE's public data in SRA. It:

1. Downloads hg38 chromosome 7 and builds a HISAT2 index of it.
2. Streams the reads CCLE placed around MET exons 13 to 15 from SRA, without downloading whole runs.
3. Aligns them to hg38 with HISAT2, the RNA-seq reads spliced and the DNA reads not, then sorts and indexes each BAM.
4. Keeps 5% of the EBC-1 reads, which brings its depth near NCI-H596's.

curl -fO https://raw.githubusercontent.com/GMOD/jbrowse-components/main/scripts/build_metex14.sh
bash build_metex14.sh

See also#

• Tutorial: Variants on protein structures part 1: sickle cell from GWAS hit to haemoglobin S
• Tutorial: Variants on protein structures part 3: cancer hotspots in KRAS
• Tutorial: Proteins on genomes.jbrowse.org
• User guide: Highlights

External links#

• Protein browser
• Cancer Cell Line Encyclopedia
• jbrowse-plugin-protein3d

Citations#

• Ghandi et al. (2019). Next-generation characterization of the Cancer Cell Line Encyclopedia
• Kim et al. (2019). Graph-based genome alignment and genotyping with HISAT2 and HISAT-genotype
• Kong-Beltran et al. (2006). Somatic mutations lead to an oncogenic deletion of met in lung cancer
• Ng et al. (2008). Structural basis for a novel intrapeptidyl H-bond and reverse binding of c-Cbl-TKB domain substrates

Feedback on this tutorial is welcome: contact us.