Functional Characterization of Glucokinase Variants to Aid Clinical Interpretation of Monogenic Diabetes

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This is a very comprehensive article and well-designed study exploring the power of systematic integration of functional explorations of GCK variants to improve the classification of variants pathogenicity. This study integrates large amounts of published data and resources with newly generated functional data and a bioinformatics classification pipeline to improve the classification process. They show increased ability to classify GCK variants through this process. The proposed framework may be applicable to other genes and contexts, and certainly expandable with more systematic functional data that will be generated for gene variants.

 

The article adds valuable information and methods, and is very well written.

 

I have only one possible recommendation and some minor points that may help further improve the manuscript.

 

Suggestion:

In this study, the authors only considered variants with low RAI and/or RSI. Since some of these variants have clearly very high values (pathogenic?), could they also already extend their strategy to include these extremely high values, similarly to what they did for low values? This should in principle further improve their capacity to classify variants. This point is well mentioned in the discussion, but I wonder if this could already be taken in account based on existing data.

 

Minor points:

• In Table 1, the bold is not very clearly visible (at least to me). Another method for highlighting should be used: star (*) or something else
• Lines 248-250: “For 8 of the 9 variants previously classified by MDEP, the functional evidence does not alter their classification. For the remaining one, the addition of functional data allows reclassification from VUS to likely pathogenic”. In Table 2, I count 8 out of 10 variants…, and two variants with functional evidence revised. Please check.
• Some typo in the name of Brnich on line 97 (misspelled as Birnich).

Author Response

Comments and Suggestions for Authors

This is a very comprehensive article and well-designed study exploring the power of systematic integration of functional explorations of GCK variants to improve the classification of variants pathogenicity. This study integrates large amounts of published data and resources with newly generated functional data and a bioinformatics classification pipeline to improve the classification process. They show increased ability to classify GCK variants through this process. The proposed framework may be applicable to other genes and contexts, and certainly expandable with more systematic functional data that will be generated for gene variants.

The article adds valuable information and methods, and is very well written.

I have only one possible recommendation and some minor points that may help further improve the manuscript.

 We thank the reviewer for their thoughtful evaluation of our manuscript. 

Suggestion:

In this study, the authors only considered variants with low RAI and/or RSI. Since some of these variants have clearly very high values (pathogenic?), could they also already extend their strategy to include these extremely high values, similarly to what they did for low values? This should in principle further improve their capacity to classify variants. This point is well mentioned in the discussion, but I wonder if this could already be taken in account based on existing data.

We agree with the reviewer that the assessment of the variants which appear to have increased activity is likely to be very helpful in refining our ability to classify variants.  Unfortunately this will require considerable work that is beyond the scope of this current study due to the limited number of patients with more mild presentations of hypoglycaemia that have been evaluated.  We have been collecting more patient data and hope that we will be able to collaborate with our clinical colleagues to put this data together in the near future so we can have greater confidence in the relationship of relative activity levels (RAI) to a formal  diagnosis of hypoglycaemia. 

Minor points:

• In Table 1, the bold is not very clearly visible (at least to me). Another method for highlighting should be used: star (*) or something else

Thank you for the feedback we have made this change. 

• Lines 248-250: “For 8 of the 9 variants previously classified by MDEP, the functional evidence does not alter their classification. For the remaining one, the addition of functional data allows reclassification from VUS to likely pathogenic”. In Table 2, I count 8 out of 10 variants…, and two variants with functional evidence revised. Please check.

Thank you for spotting this error on our part.  We have corrected this. 

• Some typo in the name of Brnich on line 97 (misspelled as Birnich).

Thank you for flagging we have checked and Brnich is spelt correctly throughout the manuscript.  

 

 

Reviewer 2 Report

Comments and Suggestions for Authors

The study systematically evaluates 25 GCK variants identified through exome sequencing or clinical diagnostic testing in the context of monogenic diabetes. Using gold-standard enzymology approaches, the authors characterize each variant’s glucose and ATP affinity, catalytic rate, cooperativity, and stability, followed by assessment of interactions with physiological and pharmacologic regulators. These functional data are integrated into Relative Activity Index (RAI) and Relative Stability Index (RSI) metrics and applied to a ClinGen Monogenic Diabetes VCEP decision tree to assign PS3/BS3 evidence strength. A substantial subset of variants demonstrates kinetic impairment or marked instability and is assigned PS3_Moderate or PS3_Supporting, whereas variants with near-wild-type parameters receive BS3_Supporting. When incorporated into the ACMG/MDEP framework, the functional results largely reinforce existing classifications but enable reclassification of one previously VUS variant to likely pathogenic. Comparisons with in-silico predictors (REVEL, AlphaMissense) and a yeast deep mutational scanning dataset highlight frequent discordance relative to biochemical assays. The authors further situate these results within a broader functional spectrum of GCK variants (≈50 total with comparable data), illustrating how gradations in activity and stability correlate with disease phenotypes. Overall, the manuscript argues that well-standardized functional assays, applied within curated ClinGen guidelines, can materially strengthen variant interpretation, while noting that current thresholds lack resolution for some mechanisms and that scalable functional approaches will ultimately be required.

Issues that should be addressed prior to publication:

1. The study relies heavily on single-enzyme in vitro assays, which may not fully capture context-sensitive behavior in hepatocytes or pancreatic beta cells, particularly for variants that alter trafficking, post-translational regulation, or degradation. It would strengthen the manuscript to briefly discuss how these mechanistic layers could affect classification, and whether complementary cellular or expression data exist.
2. Lab-to-lab variability is acknowledged but not quantified. Providing reproducibility metrics (e.g., inter-site coefficients of variation or Bland-Altman comparisons for shared controls) would increase confidence that functional cutoffs (PS3/BS3) are robust across experimental environments.
3. The RSI and RAI thresholds appear operational rather than statistically derived, and borderline cases (e.g., R275C) underscore their sensitivity to assay limits. Clarifying the rationale for the selected cutoffs, or including ROC-style or distribution-based justification (even preliminarily), would mitigate concerns about overinterpretation.
4. GKRP and GKA interaction results are treated as binary outcomes even when deviations from wild-type are observable. Providing quantitative criteria or confidence intervals, rather than qualitative statements, would reduce risk of under- or over-weighting regulatory effects in variant classification.
5. Table 1 is difficult to interpret because of its density and the number of kinetic parameters presented. It reads more like a raw data table than a results summary. Since Table 2 already provides the functional classification context, it may be clearer to move Table 1 to the Supplementary Materials and present a simplified version in the main text that highlights only the key metrics relevant to variant interpretation.
6. The application of ClinGen decision rules to variants with modest kinetic perturbations remains conceptually challenging. A more explicit discussion of how physiologic compensation from the wild-type allele and context-dependent GSIS thresholds scale with small changes in RAI/RSI would help clarify interpretation of mild variants.
7. The comparison with in-silico and deep mutational scanning tools demonstrates discordance but stops short of mechanistic insights. Where specific misclassifications exist, highlighting likely causes would provide a more informative evaluation of when these tools can or cannot be trusted.
8. Variant selection appears enriched for clinically suspicious alleles, which may inflate PS3 assignment. Please explain how the lack of a benign reference set affects the reliability of your thresholds.
9. Gain-of-function variants are acknowledged as under-defined within the current framework. Expanding this section to outline prospective functional criteria or candidate metrics (e.g., elevated RAI, altered GKRP sensitivity, stability differences) would be helpful for clinical interpretation of hypoglycemia-associated GCK alleles.

Author Response

The study systematically evaluates 25 GCK variants identified through exome sequencing or clinical diagnostic testing in the context of monogenic diabetes. Using gold-standard enzymology approaches, the authors characterize each variant’s glucose and ATP affinity, catalytic rate, cooperativity, and stability, followed by assessment of interactions with physiological and pharmacologic regulators. These functional data are integrated into Relative Activity Index (RAI) and Relative Stability Index (RSI) metrics and applied to a ClinGen Monogenic Diabetes VCEP decision tree to assign PS3/BS3 evidence strength. A substantial subset of variants demonstrates kinetic impairment or marked instability and is assigned PS3_Moderate or PS3_Supporting, whereas variants with near-wild-type parameters receive BS3_Supporting. When incorporated into the ACMG/MDEP framework, the functional results largely reinforce existing classifications but enable reclassification of one previously VUS variant to likely pathogenic. Comparisons with in-silico predictors (REVEL, AlphaMissense) and a yeast deep mutational scanning dataset highlight frequent discordance relative to biochemical assays. The authors further situate these results within a broader functional spectrum of GCK variants (≈50 total with comparable data), illustrating how gradations in activity and stability correlate with disease phenotypes. Overall, the manuscript argues that well-standardized functional assays, applied within curated ClinGen guidelines, can materially strengthen variant interpretation, while noting that current thresholds lack resolution for some mechanisms and that scalable functional approaches will ultimately be required.

Issues that should be addressed prior to publication:

1. The study relies heavily on single-enzyme in vitro assays, which may not fully capture context-sensitive behavior in hepatocytes or pancreatic beta cells, particularly for variants that alter trafficking, post-translational regulation, or degradation. It would strengthen the manuscript to briefly discuss how these mechanistic layers could affect classification, and whether complementary cellular or expression data exist.

We thank the reviewer for this suggestion. We have expanded our coverage of these important points in both the introduction and discussion.  

 

1. Lab-to-lab variability is acknowledged but not quantified. Providing reproducibility metrics (e.g., inter-site coefficients of variation or Bland-Altman comparisons for shared controls) would increase confidence that functional cutoffs (PS3/BS3) are robust across experimental environments.

 

We agree that this would be highly desirable but we do not currently have this data so we have listed this as a limitation in the discussion. 

 

1. The RSI and RAI thresholds appear operational rather than statistically derived, and borderline cases (e.g., R275C) underscore their sensitivity to assay limits. Clarifying the rationale for the selected cutoffs, or including ROC-style or distribution-based justification (even preliminarily), would mitigate concerns about overinterpretation.

 

Thank you for raising this important point.  We have added a paragraph to the introduction to provide the rationale for the cut offs.  

 

1. GKRP and GKA interaction results are treated as binary outcomes even when deviations from wild-type are observable. Providing quantitative criteria or confidence intervals, rather than qualitative statements, would reduce risk of under- or over-weighting regulatory effects in variant classification.

 

We agree with the reviewer that this is a limitation of our current study but feel that there is insufficient data to assess this with the data available.  We have expanded our discussion of this as a limitation of our study and a need for work in this area. 

 

1. Table 1 is difficult to interpret because of its density and the number of kinetic parameters presented. It reads more like a raw data table than a results summary. Since Table 2 already provides the functional classification context, it may be clearer to move Table 1 to the Supplementary Materials and present a simplified version in the main text that highlights only the key metrics relevant to variant interpretation.

We thank the reviewer for this helpful feedback but respectfully disagree that this table should move to supplementary information.  The  table summarizes critical characteristics of the GCK variant proteins which will be important to others performing these assays.  We have taken onboard feedback from reviewer 1 and have increased the clarity of some of the important results captured in it. 

 

1. The application of ClinGen decision rules to variants with modest kinetic perturbations remains conceptually challenging. A more explicit discussion of how physiologic compensation from the wild-type allele and context-dependent GSIS thresholds scale with small changes in RAI/RSI would help clarify interpretation of mild variants.

 

We thank the reviewer for this feedback and have added further details on this to the introduction.

 

1. The comparison with in-silico and deep mutational scanning tools demonstrates discordance but stops short of mechanistic insights. Where specific misclassifications exist, highlighting likely causes would provide a more informative evaluation of when these tools can or cannot be trusted.

We thank the reviewer for this suggestion but feel it is beyond the scope of this current paper. 

 

1. Variant selection appears enriched for clinically suspicious alleles, which may inflate PS3 assignment. Please explain how the lack of a benign reference set affects the reliability of your thresholds.

 

We thank the reviewer for raising this important point.  Our study is in fact one of the first to study variants which have not been identified during diagnostic testing.  The vast majority (20/25) were identified through sequencing of 13K exomes half of which were individuals without type 2 diabetes.   This study reports the largest number of variants which are likely to be benign. 

 

1. Gain-of-function variants are acknowledged as under-defined within the current framework. Expanding this section to outline prospective functional criteria or candidate metrics (e.g., elevated RAI, altered GKRP sensitivity, stability differences) would be helpful for clinical interpretation of hypoglycemia-associated GCK alleles.

 

We agree that this would be helpful but as outlined in our response to reviewer 1 we feel this is beyond the scope of this current manuscript. 

Reviewer 3 Report

Comments and Suggestions for Authors

Comments to the manuscript entitled “Functional Characterization of Glucokinase Variants to Aid Clinical Interpretation for Monogenic Diabetes”

The manuscript is interesting and extensive; however, it has minor issues that must be addressed before publication.

Specific comments

1) Lines 124- 126. The authors mention, “Out of the 11, three were completely inactive and did not respond to any concentration of substrate. These 11 variants had a relative activity index (RAI) lower than 0.5, and according to the decision tree”. The main problem in this work, of course, is that these variants have zero activity and are incompatible with life. The authors are advised to provide an explanation for this.

2) 2.1. Functional Characterization of GCK variants using gold standard in vitro assays: The authors are requested to provide saturation curves for each variant, as well as SDS-PAGE gels showing the purity obtained for each variant.

Table 1

3) The authors must explain why they obtained better yield (mg) compared to the WT in almost all variants. Furthermore, they will have to explain why, for some variants they indicated N/A, not available - for G170_K172dup, V203A, 169, and G258S, the affinity for glucose or ATP was so low that S0.5. If it could not be determined, then why do they indicate PS3_Moderate in Functional Evidence?

4) For variants A188T, V226M, and G258S, the authors indicate as Functional Evidence that PS3_Moderate, when the loss of Glucose S0.5 (mM), ATP kM (mM), kcat[A] (s-1), kcat[B] (s-1) is very significant with respect to WT; which suggests that supporting level criterion based on functional evidence in favor of pathogenicity (PS3_Supporting). This information should be verified.

5) Figures 2a and 2d. The authors should explain the purpose of these panels. The way the figure legends are described is difficult to understand. I suggest providing only the 2A and 2B in the main document.

Minor comments

6) Why did you add beta-mercapto-ethanol to the cells during transformation?

7) Please ensure that the name E. coli is in italics throughout the main document

 

Author Response

The manuscript is interesting and extensive; however, it has minor issues that must be addressed before publication.

Specific comments

1) Lines 124- 126. The authors mention, “Out of the 11, three were completely inactive and did not respond to any concentration of substrate. These 11 variants had a relative activity index (RAI) lower than 0.5, and according to the decision tree”. The main problem in this work, of course, is that these variants have zero activity and are incompatible with life. The authors are advised to provide an explanation for this.

We apologise for any confusion the variants studied are the equivalent of homozygous variants whilst the variants reported have mostly been reported in the heterozygous state.  As outlined in the introduction and discussion the wildtype allele provides compensation for the variant LoF allele which results in a very similar phenotype for heterozygous LoF variants despite the range of functional defects.   It is also worth stating that homozygous loss of function alleles have also been reported in humans and these are a cause of permanent neonatal diabetes presenting at birth.  The clinical presentation of the diabetes (i.e. diabetes severity)  is related to the functional severity of the variant because there is no wildtype allele to compensate.    Neonates are treated with insulin from birth.  Njolstad et al New England Journal of Medicine 2001; Sagen et al Diabetes 2003; Porter et al Journal of Pediatrics 2006; Raimondo et al Human Molecular Genetics 2015.   We have expanded the introduction to make this clearer. 

2) 2.1. Functional Characterization of GCK variants using gold standard in vitro assays: The authors are requested to provide saturation curves for each variant, as well as SDS-PAGE gels showing the purity obtained for each variant. Table 1

We have added SDS-PAGE gels to the supplementary data section.  Unfortunately we were only able to provide data for 17 of the 25 variants.  The proteins had been stored at -80C for some time and we note that 2 of them display some mild degradation (L314P and C213del) but their purity is evident (Supplementary Figure 2).    

Unfortunately we have been unable to locate the saturation curves for the experiments that were performed in Oxford.  This work was done prior to Dr Gloyn’s relocation to stanford in 2020 and although we have all the data for the kinetics we do not have the saturation curves.  This data is available for variants studied in Stanford and has been added. (Supplementary FIgure 3). 

3) The authors must explain why they obtained better yield (mg) compared to the WT in almost all variants. Furthermore, they will have to explain why, for some variants they indicated N/A, not available - for G170_K172dup, V203A, 169, and G258S, the affinity for glucose or ATP was so low that S0.5. If it could not be determined, then why do they indicate PS3_Moderate in Functional Evidence?

The yield of protein obtained is highly variable and can be due to a number of factors including bacterial growth.  For some variants the affinity for glucose (glucose S0.5)  is so low it is not possible to perform the kinetic assay as the concentration of glucose required for enzyme activity can not be produced: simply put we can not make a stock of glucose solution higher than 1.2M in our hands at room temperature (it won’t dissolve) which makes it difficult to have a final concentration in our assay of more than 400mM.   If the glucose S0.5 is that low then the RAI will be less than 0.5.  This has been modelled previously (Gloyn et al 2004).    

4) For variants A188T, V226M, and G258S, the authors indicate as Functional Evidence that PS3_Moderate, when the loss of Glucose S0.5 (mM), ATP kM (mM), kcat[A] (s-1), kcat[B] (s-1) is very significant with respect to WT; which suggests that supporting level criterion based on functional evidence in favor of pathogenicity (PS3_Supporting). This information should be verified.

All three variants have RAI of < 0.5 which means that they are classified as PS3_moderate which is the highest level of confidence that can be assigned with a functional assay for pathogenicity. 

5) Figures 2a and 2d. The authors should explain the purpose of these panels. The way the figure legends are described is difficult to understand. I suggest providing only the 2A and 2B in the main document.

Thank you for this feedback. We have moved these two figures to supplementary figure 1 as requested. 

Minor comments

6) Why did you add beta-mercapto-ethanol to the cells during transformation?

This is often added to improve the efficiency of transformation. 

7) Please ensure that the name E. coli is in italics throughout the main document

Thank you, we have corrected where required. 

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

My only suggestion was to possibly consider the variants with increased activity in their study. I take the point that this would require much work, and cannot be included in the present manuscript. I look forward to a future article on this point! 

The authors revised the minor points that I highlighted. I have no more comment.

This is a very nice and comprehensive article. 

 

Author Response

We thank the reviewer for their evaluation of our manuscript and are delighted that they find it improved are happy with our response and comment that "it is a nice comprehensive article".     

Reviewer 2 Report

Comments and Suggestions for Authors

The revision addresses several of my previous points, but a number of issues remain unresolved. The most significant is the lack of mechanistic explanation for the discordance between the in vitro enzymology, in-silico predictors, and deep mutational scanning data. The authors explicitly decline to provide any mechanistic insight, stating that such analysis is beyond the scope of the paper. Even if additional experiments are not feasible, a qualitative, mechanism-oriented discussion of the observed discordance would substantially strengthen the manuscript. If the authors are unwilling to address this issue, I would not be able to recommend the manuscript for publication.

Author Response

We thank the reviewer for their evaluation of our manuscript.  As the reviewers all acknowledge this is a valuable contribution to the literature and provides a rich datasource of high value for variant interpretation for monogenic diabetes.  

We understand the enthusiasm from the reviewer for more details on the comparison of the tools but this is not a primary outcome of our study.  We have however  included an additional paragraph to "speculate" on the reasons for the discrepancy between the tools drawing on previous studies which have performed a systematic analysis and concrete observations which can be drawn.  A more thorough post-mortem of the differences between the approaches requires a larger body of data. 

We trust that with this additional paragraph that our manuscript will now be acceptable for publication. 

 

Reviewer 3 Report

Comments and Suggestions for Authors

I thank the authors for considering the comments. The manuscript is substantially improved and can be accepted

Author Response

We thank the reviewer for their evaluation of our manuscript and are pleased they find it improved and suitable for publication.