Skip to main content
Premium Trial:

Request an Annual Quote

QTL Analysis Finds Effects of Genetic Variation on RNA Attenuated at Level of Protein Expression


NEW YORK (GenomeWeb) – In a paper published this week in Science, researchers from Stanford University and the University of Chicago presented data demonstrating limited correlation between regulation of RNA and protein expression.

Using a combination of genomic, transcriptomic, ribosome-profiling, and proteomic data, the researchers looked at the impact of genetic variation among regulatory quantitative trait loci on RNA and protein levels, observing that while most such QTLs are strongly linked to transcript expression, they have significantly smaller effects on protein expression.

The findings suggest the presence of some sort of buffering mechanism designed to maintain stability in protein expression in the presence of genetic mutations, Yoav Gilad, a researcher at theUniversity of Chicago and author on the paper, told GenomeWeb.

The effort, Gilad said, stemmed from earlier research by him and his colleagues as well as by other groups into regulatory QTLs — QTLs that affect expression of RNA as well as other processes involved in transcription rates — across different species.

Part of this research involved comparing mRNA and protein expression in humans, chimpanzees, and rhesus monkeys. Analyzing the data from these comparisons, the researchers realized that they saw larger cross-species differences in RNA than protein expression. And this, Gilad said, led them to ask how these expression levels compared across different subjects from the same species.

Such an intra-species comparison was important in that it could provide insight into the mechanism responsible for this disconnect between RNA and protein levels, he said.

Based on the cross-species comparisons, "we couldn't distinguish between two [potential] mechanistic scenarios," he noted.

"One is compensatory evolution," Gilad said. "In that mechanism you have a mutation that changes the RNA levels, and then you have a second mutation that compensates for it by changing the protein levels in order to compensate for the [first mutation].

While this is "a complicated mechanism, it exists," Gilad said. "We've seen it."

The other mechanism is a sort of buffering in which "the architecture of the regulatory network is set up in such a way that it allows for the buffering against some input of noise," he said.

The [Science] study was able to distinguish between these two scenarios, because by looking only within humans the researchers could know that any observed variation would be too recent to have been caused by compensation, Gilad said. "So buffering becomes really the default mechanistic explanation."

For the study the researchers used HapMap Yoruba lymphoblastoid cell lines, collecting ribosome profiling data for 72 lines and SILAC mass spec-based protein expression data  generated by proteomics firm MS Bioworks for 62 lines. They also had genome-wide genotypes and RNA-seq data for all the lines.

Integrating this data, they identified 2,355 expression QTLs, which regulate expression of mRNA; 939 ribosome occupancy QTLS, which regulate ribosome occupancy; and 278 protein QTLs, which are associated with protein levels.

Their analysis of the eQTLs found that the average effect of variation in eQTLs across the human subjects was smaller at the protein level compared to the RNA-seq or ribosome levels, which were roughly comparable. From this, the authors concluded that "the majority of genetic variants affecting transcript levels also alter ribosomal occupancy, typically with a similar magnitude of effect," while "eQTLs have attenuated (or absent) effects on steady state protein levels."

The researchers also observed that variations in pQTLs had, on average, smaller effects on RNA expression than protein expression. Additionally, they identified a subset of pQTLs they termed protein-specific QTLs in which variations had an effect on protein expression but no detectable effect on mRNA levels.

Based on the results, "it seems that there is a fair amount of buffering in the system, for quite a few genes, [and that] for mutations that result in differences in RNA levels between individuals, the effect is somehow buffered at the protein level," Gilad said.

This is significant, he added, in that it suggests this buffering is a general, widespread phenomenon involved in transcript and protein regulation. He noted that while the researchers currently have no data explaining the function and evolution of this buffering approach, one reasonable speculation is that it might it allow for more stability in the face of genetic mutations.

"You can envision two scenarios," he said. "I want the protein level to remain X, and in one scenario I'm going to make sure that every step of the way is precise enough that it leads to that X, and every mutation that changes any step on the way is being selected against."

In the other scenario — the one suggested by the Science study —  the system is arranged "such that some level of noise or variation from mutations in all the steps preceding protein level X can be sustained, buffered, attenuated, taken into account, and not affect too much the end result," he said.

The study also highlighted the often observed disconnect between transcript and protein data. While RNA transcript levels have frequently been used as proxies for protein expression levels due to the relative ease of transcriptomic techniques like microarrays and RNA-seq compared to shotgun proteomics, many studies have suggested that the two levels of information are not particularly well correlated. The Science paper would seem to indicate this as well, though at the level of comparing individual genes across multiple subjects, as opposed to comparing many genes across a single subject.

Gilad said, however, that recent work has found that when technical noise is accounted for, correlation between RNA and protein levels improves significantly. On a related note, researchers from the University of British Columbia published a study last year in Molecular Systems Biology that found that RNA levels can be better correlated with their corresponding proteins by taking into account protein synthesis and degradation levels.

Still, Gilad said, while "you can learn a lot about regulatory principles or the genetic impact on regulation just by studying RNA," given the observed disconnect between RNA and protein regulation, "if you really want to follow up an [RNA-based] finding because it could have an impact on a disease study or another phenotype that you care about, you should probably look at the protein before you continue to spend time and money on that observation."