Isoform-Level Interpretation of High-Throughput Proteomics Data Enabled by Deep Integration with RNA-seq

Cellular
control of gene expression is a complex process that is
subject to multiple levels of regulation, but ultimately it is the
protein produced that determines the biosynthetic state of the cell.
One way that a cell can regulate the protein output from each gene
is by expressing alternate isoforms with distinct amino acid sequences.
These isoforms may exhibit differences in localization and binding
interactions that can have profound functional implications. High-throughput
liquid chromatography tandem mass spectrometry proteomics (LC–MS/MS)
relies on enzymatic digestion and has lower coverage and sensitivity
than transcriptomic profiling methods such as RNA-seq. Digestion results in
predictable fragmentation of a protein, which can limit the generation
of peptides capable of distinguishing between isoforms. Here we exploit
transcript-level expression from RNA-seq to set prior likelihoods
and enable protein isoform abundances to be directly estimated from
LC–MS/MS, an approach derived from the principle that most
genes appear to be expressed as a single dominant isoform in a given
cell type or tissue. Through this deep integration of RNA-seq and
LC–MS/MS data from the same sample, we show that a principal
isoform can be identified in >80% of gene products in homogeneous
HEK293 cell culture and >70% of proteins detected in complex human
brain tissue. We demonstrate that the incorporation of translatome
data from ribosome profiling further refines this process. Defining
isoforms in experiments with matched RNA-seq/translatome and proteomic
data increases the functional relevance of such data sets and will
further broaden our understanding of multilevel control of gene expression.