Many observers are predicting that whole-genome sequencing will become part of standard medical care within the next decade (59
). That potential has heightened interest in interpretation of genetic variants that might provide insight into assessing future risks of illness, refining present diagnoses, or predicting drug response. Hundreds of common variants shown to be associated with disease risk or quantitative traits have emerged in the last 5 years as a result of the GWAS approach. However, most of these have been shown to have a relatively modest effect on risk, and there appears to be “missing heritability” for many diseases and traits that may at least in part be attributable to rare variants, some of which may have larger effects (2
While successful discovery of such rare variants has been achieved in some studies by focusing on individuals at the extremes of a quantitative trait (6
), a major challenge for the future is how to interpret such variants when they occur in a less selected population. For that purpose, we studied a clinical population modestly enriched for cardiovascular disease risk and looked for rare variants in GCKR
Variants in GCKR
have been previously shown to affect a range of metabolic processes. For common variant p.Pro446Leu, it has been relatively easy to detect such associations because of the availability of phenotype information on large numbers of individuals. As an alternative approach to finding rare alleles, sequencing GCKR
in individuals with extremely high triglycerides has shown utility in relating such variants collectively to phenotypes (11
). However, for GCKR
or indeed any gene, determining whether a specific mutation is functionally important in an individualized clinical setting remains a major challenge. Bioinformatic methods such as mutation prediction algorithms and assessment of evolutionary conservation are useful preliminary tools in analysis of whether a particular variant might have functional consequences, but these approaches showed limitations in accuracy and consistency between prediction programs both in this study and in the study by Johansen et al. (11
Accordingly, we utilized existing information about GKRP function and undertook individual molecular characterization of all 18 GCKR variants from the ClinSeq project. We observed defects in cellular localization for the majority of these variants (12/18) as YFP-tagged GKRP constructs in HeLa cells and refined our analysis by GCK coexpression and kinetic characterization. Functional rare variants could be broadly subdivided into putative LOF and GOF subtypes (Table ).
Potential GOF mutations in conserved C-terminal residues of GKRP abolished GKRP nuclear localization and GCK sequestration, while p.Arg612Cys-GKRP showed no significant differences in kinetic properties compared with WT GKRP (Figure D and Figure ). This suggests that the region surrounding residues 607–612 could be part of the unknown mechanism by which GKRP is localized to the nucleus. As only 2 individuals carried potential GOF mutations (Table and Supplemental Figure 9), further studies will be needed to determine the phenotypic effects of such variants. However, physiologically, these mutations increase cytoplasmic GKRP and thus may serve to decrease GCK activity by decreasing both the pool of sequestered nuclear GCK and of active, cytoplasmic GCK. This would be predicted to decrease hepatic glycogen, triglyceride, and cholesterol synthesis.
Analysis of the subset of ClinSeq individuals heterozygous for rare GCKR
LOF variants collectively showed a significant increase in total cholesterol, LDL cholesterol, and triglyceride levels (Table ). These variants were both functionally and phenotypically similar to p.Pro446Leu, showing reduced expression, reduced nuclear localization, potential reduction in GCK sequestration, and reduced interaction with F6P and/or F1P. As has been proposed for p.Pro446Leu (13
), reduced nuclear localization and GCK sequestration would likely increase fasting hepatic glucose uptake and disposal through synthetic pathways including de novo lipogenesis. For p.Pro446Leu, the phenotypic effect was much stronger on triglycerides than on fasting glucose; presumably the same phenomenon is present in our collection of rare LOF variants, as even with small numbers, we were able to detect the effect on lipids, but not on fasting glucose. However, while variants within this group are qualitatively similar, they displayed a range in the magnitude of cellular and kinetic effects.
The most severe loss-of-function variants, such as p.Val103Met, appear to form very little, if any, functional protein, characterized by low cellular fluorescence and dramatically reduced ability to inhibit GCK. Physiologically, these variants may be indistinguishable from null mutations and might therefore be compared with heterozygous Gckr
-knockout mice. Gckr+/–
mice show reduced liver GCK levels and activity (24
) as well as trends toward lower hepatic glycogen content and higher blood glucose levels 30 and 60 minutes after an oral glucose tolerance test (24
). The reduction in glycemic control is likely attributable to a decrease in the Gkrp-bound Gck nuclear pool that is normally mobilized in response to a glucose challenge. This loss of nuclear stabilization and/or glucose-dependent translocation has been shown to be associated with impaired glucose tolerance in rodent models (24
). Phenotype comparisons and results of genotyping for the severe LOF subgroup were consistent with these findings.
As the common p.Pro446Leu variant has been shown to have both cellular and kinetic effects (13
), it is useful to consider the effect of this variant in cis
or in trans
with rare variants. Experiments assessing Leu446 in cis
suggested this variant may amplify defects in protein expression, localization, and activity, but will have fairly mild effects on F1P and F6P interaction. Kinetic results modeling the effect of Leu446 in trans
suggested intermediate effects on activity and phosphate ester response (Supplemental Table 7). Using our cellular model system, it is difficult to assess the effect of Leu446 in trans
on GCK sequestration. However, as loss-of-function variants are similar to p.Pro446Leu (Table ), cellular effects of Leu446 in trans
with a LOF rare variant are likely to include reduced GCK sequestration (43
), GKRP activity, and inhibition from both chromosomes. Phenotypic effects may range from those of Leu446 homozygotes (as suggested by GWAS) to those approaching Gckr
-knockout mice for more severe rare variants.
ClinSeq sequencing, combined with previous studies and emerging 1000 Genomes data, suggests it is unlikely there are additional common nonsynonymous GCKR
variants in the general western European population (11
). Follow-up genotyping confirmed the rarity of individual variants, but highlighted the importance of considering ethnicity for replication of rare variants. Accordingly, our findings supported a collective analysis of rare variants to explore relationships with phenotypes. However, functional characterization revealed that not every variant is likely to have the same biochemical characteristics and therefore the same phenotypic consequences (Table ). This is an important limitation of in silico predictions, as bioinformatic methods often cannot distinguish between different types of functional variants. The separation of phenotypes based on functional classification in the ClinSeq cohort suggests the mutational load of low-frequency GCKR
variants may still play a significant role in heritability of human glucose and lipid traits.
Some of the challenges highlighted by this study are likely to be amplified as more sequencing data becomes available. Resequencing studies will identify a large number of rare variants per individual, many of which may be novel. Phenotyping a large number of individuals for each rare variant will often not be practical. Unless there are reliable computational, cell biological, or biochemical methods for determining the functional consequences of a variant (as we have been able to do here with GCKR), it will generally be difficult to interpret the significance of rare sequence changes. While the era of complete genome sequencing holds much promise for identifying heritable risk factors and ushering in the era of personalized medicine, the leap from sequence discovery to functional inference and medical consequence will often not be trivial.