Background
Myotonic dystrophy type 1 (DM1) is a neuromuscular disease with a worldwide average prevalence of around 1 in 8000 people and a high unmet clinical need [
1]. DM1 is considered the most frequently occurring adult-onset form of muscular dystrophy. This degenerative multisystem disease is characterized by a wide range of symptoms including myotonia, muscle weakness and dystrophy, fatigue, apathy, cataracts, obesity, and insulin resistance. Next to a severe decrease of life quality, DM1 patients suffer from a reduced life expectancy mostly due to problems with cardiac and respiratory function. Currently, no curative therapy exists.
DM1 is caused by the expansion of a CTG trinucleotide microsatellite repeat in the 3′ UTR of the DM1 protein kinase (
DMPK) gene [
2‐
4]. Unaffected individuals carry up to approximately 37 CTG triplets in
DMPK, while in DM1 patients this ranges from 50 to even a few thousand repetitions. Depending on the inherited repeat length, DM1 can become manifest at birth or early in life but more frequently becomes apparent in adulthood [
1]. In general, the disease manifestation is earlier and more severe with longer repeat expansions. Interruption of the CTG repeat by variants such as CCG or CGG is associated with milder symptoms [
5]. The expanded CTG repeat is thought to cause disease mainly via an mRNA gain-of-function mechanism, in which aberrant hairpin structures formed by long CUG repeats are central [
6‐
9]. Directly or indirectly, these hairpin structures dysregulate the function of RNA binding proteins from the muscleblind-like (MBNL) and CUGBP Elav-like (CELF) families, leading to widespread disturbed RNA processing and consequently altered functions of various proteins [
10‐
12]. Although proven to be the disease-causing mutation, clinical symptoms of DM1 are only moderately associated with the CTG repeat or the dysregulation of specific proteins which suggests an involvement of other mechanisms in symptom expression [
13‐
16].
While there are many promising therapeutic oligonucleotides, small molecule drugs, and gene therapies in the (pre) clinical pipeline for some of the signs and symptoms of DM1, none is expected to reach widespread clinical application soon. Physical training and increasing activity are currently being applied to relieve DM1 symptoms with marked improvements in relatively mildly affected DM1 patients [
17,
18], which has furthermore shown to induce biochemical responses in DM1 mouse models [
19,
20].
The to-date largest clinical trial in DM1 was OPTIMISTIC: Observational Prolonged Trial In Myotonic dystrophy type 1 to Improve Quality of Life-Standards, a Target Identification Collaboration [
18]. The OPTIMISTIC clinical trial included over 250 well-characterized DM1 patients from four centers in Europe, where the effects of cognitive behavioral therapy (CBT) and optional graded exercise therapy were closely monitored over 16 months via more than twenty outcome measures. Notably, the CBT intervention was tailored towards the specific needs of the patient in a shared decision-making process between the patient and the psychotherapist, allowing for a personalized intervention. The trial has shown significant, yet heterogenous improvements for various signs and symptoms, as well as the capacity for social activity and participation in DM1 [
21].
Here, we set out to find molecular profiles associated with the disease-causing CTG repeat and therapy response based on full blood mRNA sequencing before and after the CBT intervention of 27 patients from the OPTIMISTIC cohort. Given the accessibility of peripheral blood, it has increasingly been used for the successful identification of disease biomarkers for a variety of neurological and psychiatric disorders such as Duchenne muscular dystrophy (DMD), Huntington’s disease, major depressive disorder, and DM1 [
22‐
25]. Furthermore, the multisystem nature of DM1 is known to be reflected by various laboratory abnormalities of blood samples, supporting the relevance of peripheral blood for the identification of disease-relevant information [
26]. We analyzed gene expression levels as a function of CTG repeat size (as a proxy for disease load/severity) and of the therapy response. Next, we combined these findings and compared the results to various previously published datasets. We were able to identify 608 genes significantly associated with the CTG repeat and further illustrate that 97 of these genes returned towards more normal expression levels in clinical CBT responders.
Discussion
The purpose of this study was the identification of DM1-specific therapeutic biomarkers in peripheral blood. The multisystem nature of DM1 is known to be reflected by laboratory abnormalities of peripheral blood, making it together with its accessibility a promising tissue for biomarker studies in this disease [
26]. Hence, we used blood samples of 27 DM1 patients from the OPTIMISTIC cohort to study the associations of gene expression with disease severity as well as the response towards the CBT intervention. In an effort to fairly represent the whole OPTIMISTIC cohort and to facilitate the generalizability of the study findings, a stratified random sampling procedure was implemented which resulted in a balanced patient cohort with respect to age, CTG-repeat length, sex, therapy response, and clinical site distributions. Patients with an interrupted CTG repeat were excluded in order to limit molecular heterogeneity induced by slower disease progression rates [
5]. Nonetheless, we identified substantial heterogeneity in molecular expression profile changes after the 10-month CBT intervention. Furthermore, we identified gene sets that were significantly associated with the disease-causing CTG repeat as well as with the average response towards the CBT intervention across different clinical outcome measures. Most interestingly, an overlap of 97 genes among these latter two gene sets has been identified, showing a clear trend of more normal expression levels in clinical responders. Based on these different gene sets, several biological pathways associated with DM1 have been discovered, as well as specific genes/gene families with ties to neuro(-muscular) disorders.
The OPTIMISTIC study has shown that DM1 patients significantly improve in their capacity for activity and social participation after the CBT intervention [
18]. It was furthermore hypothesized that CBT may directly or indirectly improve other biological systems affected by the disease. This hypothesis has been confirmed for muscles of the lower extremity, showing an increase in cross-sectional area as a result of the intervention [
58]. Here, we set out to further explore this hypothesis by investigating changes on the molecular level. These changes may be the result of increased physical activity, which has been linked to differences in gene expression in previous studies [
59], but may also be a more direct effect of the psychotherapeutic CBT intervention [
60].
In line with the results of an earlier study, we have illustrated that the clinical response towards the CBT intervention was rather heterogenous [
21]. A novel addition to this finding was the illustration that this heterogeneity extends towards changes in molecular profiles within a 10-month timeframe. Importantly, this heterogeneity could not be explained by changes in the cellular composition of the blood samples between the two time points, as the similarity of cell type composition has been verified. Additionally, this heterogeneity could not be explained by changes of different outcome measures such as the DM1-Activ-c, 6MWT, or compound response. While the CBT intervention likely played a part in this heterogeneity, the magnitude of this contribution could not be assessed due to the lack of a control group. Other factors, such as aging or seasonal effects, may also have contributed to this finding.
Across the different gene sets identified in this study, several of the genes with the lowest
p-values (
SLC39A8,
IRS2,
FBXO48) and one WikiPathway (transcription factor regulation in adipogenesis) were associated with insulin signaling or more broadly related to metabolism/adipogenesis [
61‐
63]. Dysregulation of insulin signaling has been linked to clinical features of DM1 and is an actively ongoing field of investigation [
64]. Aberrant insulin signaling has also been found in other diseases of the nervous system such as depression, with indirect improvements being observed after CBT [
65]. Interestingly, the anti-diabetic drug metformin has been shown to improve the mobility of DM1 patients with effect sizes of the 6MWT comparable to those observed in the OPTIMISTIC study [
66]. With increasing therapeutic interest in this area, our findings suggest that disease-relevant insulin signaling can be studied on a molecular level in blood samples, highlighting the utility of peripheral blood in this setting.
Similarly, across most of the gene sets, we identified several WikiPathways associated with immunological functions (cell-specific immune response, chemokine signaling pathway, IL-3, 4, and 5 signaling). While this may be in part due to a bias introduced by the profiled tissue, the immune system likely plays a role in the DM1 pathophysiology like for many other chronic diseases [
67]. As such, blood sample-based immunology studies may be an interesting field of future investigation.
The intersection of the genes significantly associated with the disease-causing CTG repeat, as well as the average CBT response across different outcome measures, revealed a subset of 97 genes. These genes are of particular interest for the identification of therapeutic biomarkers, as their disease association has been confirmed in an external dataset and they showed normalization of expression levels in clinical responders. Among the genes with the lowest
p-values associated to both CTG-repeat length as well as CBT response were
HDAC5,
DNAJB12, and
TRIM8. In total, the subset of 97 genes consisted of two HDACs (histone deacetylases,
HDAC5,
HDAC7). HDACs play an important role in transcriptional regulation and compounds that inhibit HDAC enzymes are being studied for their potential effect on a range of human diseases, including neurological disorders [
68]. The DNAJB12 protein is a member of the heat shock protein family, with some evidence supporting positive effects of their induction for muscular dystrophy and other muscle wasting conditions [
69]. The TRIM family protein TRIM72 has been shown to be an essential component of the cellular membrane repair in muscles, with evidence supporting some positive effects in mouse models of muscular dystrophy [
70]. Authors of the same study suggest the potential of other TRIM family members as potential targets in similar disease states, which may support the further investigation of
TRIM8 in DM1. Although mostly associated with therapy response,
RARA and
RXRA were also among the overlapping 97 genes. Stimulating retinoic acid signaling has been linked to muscle regeneration in mouse models via increased proliferation of fibro/adipogenic progenitor cells, highlighting the relevance of this pathway as another potentially DM1-relevant drug target [
71]. Taken together, these findings confirm the value of whole blood-based expression profiling for the discovery of therapeutic biomarkers in DM1.
Interestingly, the genes significantly associated with the CTG repeat showed a moderate correlation with the genes associated with DMD body measurements of an external study. We hypothesize that these body measures are likely also correlated with age, which in turn reflects disease progression. This suggests that some of our results may therefore not be DM1 specific, but rather reflect non-specific molecular dysregulations shared across different (neuromuscular) disorders. This hypothesis is supported by the significant association between the CTG repeat with
MMP9, which is known to be a non-specific biomarker that has for instance been linked to cardiac remodeling after myocardial infarction, inflammation, and DMD [
72,
73]. We therefore deem further exploration of shared dysregulations as highly valuable, as this may lead to the discovery of therapeutic targets relevant to a variety of diseases.
Although DM1 is known as an alternative splicing disease, only four splice events have been significantly linked to the disease-causing CTG repeat in this study. This may be the result of relatively low
DMPK expression in blood [
74] and is in line with the absence of strong splice aberrations in blood from DM1 patients compared to controls [
25].
DMPK’s low expression in blood cells may also explain the lack of concordance between our disease severity-associated gene expression differences observed in blood with gene expression differences observed in the muscle and brain. So, while whole blood-based transcription profiling can identify disease-relevant molecular dysregulations, these dysregulations do likely not fully reflect the dysregulations observed in other tissue types. Yet, we found a high correlation of the CTG-repeat effect size with the DM1 phenotype effect size of a different blood-based study, as well as with a principal component derived body measure association of a DMD-based study. While the former validates our findings, the latter suggests the possibility of shared, disease-relevant, dysregulations across different neuro(muscular) disorders detectable in peripheral blood. If true, associated pathways might reveal highly interesting targets for drug discovery, as they may have a positive influence on multiple diseases.
Limitations of this study
To find disease-relevant gene expression in blood, we searched for linear associations with the disease-causing CTG-repeat length. While the CTG-repeat length is thought to be the main driver of molecular dysregulation, associations between the progenitor allele length of the CTG repeat with several clinical outcome measures, including DM1-Activ-c and 6MWT, have been found to be only weak-moderate [
13]. In line with the previously published challenges to directly relate gene expression to clinical phenotypes, we were not able to find significant, direct associations between clinical outcome measures and gene expression. Still, among the genes with the lowest
p-values for the DM1-Activ-c questionnaire was
GSKIP, a gene encoding for an inhibitor protein of the known DM1 drug target GSK3-β [
75‐
77]. Given the biological relevance of this finding, we deem it likely that the current study design was underpowered to study the association of gene expression with individual clinical outcome measures, especially when taking clinical and molecular heterogeneity into account.
The clinical heterogeneity in therapy response may in part be explained by the personalized nature of the CBT intervention, with therapy foci being tailored towards the needs and wishes of the individual patient. As a consequence, one might expect different outcome measures to be more appropriate for CBT efficacy assessments for different patients. Yet, the identification of molecular signatures associated with therapy response necessitates the use of the same clinical outcome measure. For this reason, and to average out some of the uncertainty inherently associated with the recording of the different outcome measures, we settled on the use of the compound response score. While the scaling assured a more or less equal contribution of each outcome measure to this score, we acknowledge that this combined score is biased by the outcome measures that were used in OPTIMISTIC.
Even though we statistically corrected for non-specific molecular changes between the two time points, the lack of RNA-seq profiles from the OPTIMISTIC control arm makes it difficult to state with certainty that the observed molecular changes are due to the therapy itself. However, this does not discount their value as potential therapeutic targets, as they are, regardless of the mediation factor, significantly associated with improved clinical status. Moreover, for this reason, we deemed studying the RNA-seq profiles of the OPTIMISTIC control arm to be less valuable, as these patients did not significantly clinically declined over the 10-month timeframe [
18].
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.