Biomarkers predict outcome in Charcot-Marie-Tooth disease 1A

Abstract

Background: Charcot-Marie-Tooth disease type 1A (CMT1A) is the most common inherited neuropathy, a debilitating disease without known cure. Among patients with CMT1A, disease manifestation, progression and severity are strikingly variable, which poses major challenges for the development of new therapies. Hence, there is a strong need for sensitive outcome measures such as disease and progression biomarkers, which would add powerful tools to monitor therapeutic effects in CMT1A.

Methods: We established a pan-European and American consortium comprising nine clinical centres including 311 patients with CMT1A in total. From all patients, the CMT neuropathy score and secondary outcome measures were obtained and a skin biopsy collected. In order to assess and validate disease severity and progression biomarkers, we performed qPCR on a set of 16 animal model-derived potential biomarkers in skin biopsy mRNA extracts.

Results: In 266 patients with CMT1A, a cluster of eight cutaneous transcripts differentiates disease severity with a sensitivity and specificity of 90% and 76.1%, respectively. In an additional cohort of 45 patients with CMT1A, from whom a second skin biopsy was taken after 2-3 years, the cutaneous mRNA expression of GSTT2, CTSA, PPARG, CDA, ENPP1 and NRG1-Iis changing over time and correlates with disease progression.

Conclusions: In summary, we provide evidence that cutaneous transcripts in patients with CMT1A serve as disease severity and progression biomarkers and, if implemented into clinical trials, they could markedly accelerate the development of a therapy for CMT1A.

Keywords: Charcot Marie Tooth disease 1A; biomarker; disease progression; disease severity; skin biopsy.

Conflict of interest statement

Competing interests: The authors declare no competing interests.

© Article author(s) (or their employer(s) unless otherwise stated in the text of the article) 2017. All rights reserved. No commercial use is permitted unless otherwise expressly granted.

Figures

Figure 1:. Previously identified potential cutaneous mRNA biomarkers tested in skin biopsies of 266 patients with CMT1A.

A Participating centers in Europe and the USA with number of assessed patients and contributed skin biopsies for biomarker analyses (see Table 1 for further details). Next to 266 patients from Europe that were assessed once (in blue) additional 45 patients from Europe and the USA were sampled twice within 2–3 years giving information on the progression (in green). 37 healthy humans were included as controls from two centers in Germany (grey). B Heatmap displaying normalized Ct values from the qPCR analyses of all patients (columns) for all genes (rows). High values (in blue) correspond to low expression while low values (in red) indicate high expression. Both dimensions, patients and genes, are re-ordered by means of correlation based hierarchical clustering to group by expression profile similarities. The dendrogram on the top shows the clustering of patients, the dendrogram on the left shows the clustering of the genes. The biggest cluster of eight similarly regulated genes is highlighted in magenta. The two rows at the top show the two housekeeper genes for reference. C Correlation of the gene expression profiles across the patients. Both, rows and columns show genes, the upper triangle shows the numerical values of the correlation coefficients the lower triangle visualizes the correlations coefficients where blue colored circles represent negative correlations and red colored circles represent positive correlations. The size of the circles indicates the strength of correlation. The genes are clustered by their correlation profile. Eight out of 16 genes cluster with correlation coefficients >0.7 (highlighted in magenta). Most notably, there is no negative correlation among the genes, except for the perfectly negatively correlated housekeeper genes.

Figure 2:. Cutaneous biomarker expression separates CMT1A disease severity and controls.

A Principal Component Analysis (PCA) was performed on the PCR data of the biomarker cluster identified in Fig.1B-C. A scree plot shows the percentage of the explained variance for each of the principal components (PCs). PC1 captures most of the information and explains 95% of the observed variance while PC2 explains already only 1% of the observed variance. B Mapping of the samples on the first two principal components from A with PC1 on the x axis and PC2 on the y axis. All samples are color coded by their CMT disease severity status (red: healthy, green: mildly affected (CMTNS <= 10), blue: moderately affected (10 > CMTNS < 20), purple: severely affected (CMTNS >= 20). The ellipses are 70% probability ellipses assuming two dimensional normally distributed data drawn separately for each of the 4 groups. C A support vector machine (SVM) classifier separating mild and severe CMT cases based solely on gene expression profiles of the eight biomarkers (Fig. 1B-C) was trained. Shown are the results from a 10 times repeated 10-fold cross validation. The presented ROC curve plots sensitivity (y axis) vs. specificity (x-axis). A perfect classifier would reach the top left corner (100% sensitivity and 100% specificity). The closest point (youden index) of the trained SVM reaches 90% sensitivity and 76.1% specificity.

Figure 3:. Disease progression is clinically detectable over a two to three year’s time period

A Progression of CMT measured by the scores CMTNSv1, CMTNS_full, CMTNS_mod, and CMTNS_signif with patients (colored by contributing center) on the y-axis and the score difference on the x axis where positive difference correspond to higher scores at the second examination. B The change of CMTNS is shown over time. The time between measurements ranges from 2 to 3 years (with one exception of only 1 year time difference). Each line represents one patient. C Regression coefficients with 95% confidence intervals from the fit of a linear mixed effect model for CMTNS with fixed effects BMI, gender, age, and time and a random effect accounting for the repeated measures in the patients. The factor ‘time’ has a significant influence on the CMTNS with an estimated increase in CMTNS of 0.75 per year.

Figure 4:. Cutaneous expression of selected biomarkers changes with disease progression.

A Mixed effects repeated measures regression models were fit per gene to the expression data. The effect of time was assessed controlling for BMI, gender, age, and center effects. Shown are the p values for the time effect (orange: adjusted p value significant) for all genes. B Shown is the progression data underlying the regression models from D displayed as Ct value (gene expression) as a function of time. Each colored line represents one patient and the black lines show the average patient at time 0 progressing by the modeled time effect. C A predictive random forest model that classifies patients into progressive patients (CMTNS increase) and non-progressive patients (CMTNS does not increase) was trained on the six genes with significant change in expression over time. Data was available for 19 progressive and 11 non-progressive patients. The consensus ROC curve from a 10 times repeated 10-fold cross validation reveals an AUC of 0.74 and the sensitivity/specificity at the youden index (marked with a dot) are estimated to be 63.2% / 100%.