In this study, we provided first evidence for pathological cortical thinning in DM2 patients, which showed mostly overlapping regions but varying extents compared to DM1. Regarding clusters of cortical thinning compared to healthy controls both disease types showed largest cluster extent in occipital lobes accompanied by a stronger focus in the temporal lobes in DM2 and larger alterations in the frontal lobes of DM1 patients.

To our knowledge, this is the first study analyzing cortical thinning in DM2. Regarding cortical thinning in DM1, our results are to a large extent in line with those of comparable studies, although differences in detail may be due to differences in study design or patient selection: a previous study by Serra et al. showed reduced CT in the precuneus, the angular gyrus, the superior temporal gyrus and the medial frontal gyrus bilaterally, the right precentral gyrus, the right posterior and anterior cingulate cortex, and the left superior parietal lobule [38]. Although the anterior part of the cingulate cortex showed no significant differences in our sample, the other named brain regions revealed also decreased thickness. Pronounced thinning was also present in the occipital cortex in our DM1 sample which was not shown to be significant by Serra et al. [38]. However, they found a relationship between CT in the lateral occipital cortex and the performance at the Social Cognition Battery. In addition, another study by Zanigni et al. showed a “prevalent involvement of parietal-occipital region” especially the lateral occipital parts [12], which were highly significant in our study, too. Another study by Yoo et al. showed reduced CT in superior temporal cortices (including the supramarginal cortex), precuneus, lateral occipital cortex, superior frontal, precentral, and postcentral cortices in the patient group compared to controls [23], covering all brain regions that were also significant in our sample.

In our study sample, reduced visuospatial functioning and deficits of the non-verbal visual working memory accompanied occipital cortical thinning. Although visual deficits were not reported, this might be evidence for disturbances of visual processing within the occipital lobes. Further, the temporal lobe was affected by cortical thinning, which was more pronounced in DM2, were patients showed a high number of pathological results for the non-verbal visual working memory test. Especially, the medial temporal lobe plays a pivotal role in visual working memory, which was shown in our patients. Taken together, this provides evidence for disturbances of the occipito-temporal region in DM which has a large impact on processing of visuospatial information through the ventral visual pathway. It has been recently described using tractography analyses that the medial occipital longitudinal tract has a major function for visuospatial learning as it seemed to carry visual information between the early visual cortex and parahippocampal area [39].

Our HC were relatively younger than DM1 patients, but we think that this does not have a major impact on our results. Cortical thinning was more pronounced in DM and even more sever in DM1 despite DM2 patients were even a slightly older than DM1 patients. Based on our additional analysis between young and older HC, we proposed that our shown disease effects were not mainly driven by age-related effects. Thus, we showed that CT in HC who were younger than 33 years was not significantly decreased in occipital brain regions, instead a strong focus on frontal brain regions and post- and precentral gyri was obtained. This might not completely replace a perfectly age-matched control group, but with regard to our amount of available data, this additional analysis should provide more reliability of our results and weaken the influence of age for interpreting our results.

Direct comparison of CT between DM1 and DM2 yielded differences in the frontal lobes, the precuneus and the para- and precentral gyri where CT was lower for DM1 than for DM2. Like other parietal areas, the precuneus is relevant for visuospatial functions [40]. Clinically, our DM1 sample showed more pronounced daytime sleepiness and less spatial visualization abilities compared to DM2, though other parameters were either comparable or more decreased in DM2 such as the non-verbal visual working memory. In a recent study, acute sleep deprivation was shown to be associated with decreased functional connectivity between the precuneus and the middle frontal gyrus in healthy subjects [41], but also other studies analyzing healthy subjects found relationships between frontal lobe metabolic decreases and sleep deprivation [42]. Since these metabolic changes were only partially reversed by recovery sleep, we hypothesized that long-term consequences of cortical thinning might be involved. Thus, our DM1 patients who are suffering from daytime sleepiness, which is often accompanied by a disturbed sleep rhythm, might have subsequently developed CT alterations. Sleep disturbances are also known to be a factor for promoting dementia, which could be avoided by appropriate treatment. Therefore, treating sleep disturbances in DM 1 and DM2 may be a potential therapeutic approach in order to prevent progression in cortical thinning.

The presented widespread reduction in FA provided evidence for microstructural alterations in both DM1 and DM2. As previously shown, this gives an indication of decreased white matter integrity [43]. As it has been already shown by Minnerop et al. [19], motoric tracts were strongly affected by decreased FA in DM1. In our study, both patient groups showed abnormal neuronal integrity in temporal lobes, which was also more pronounced in DM1. Similarly, cerebellum and brainstem were affected in both groups, but this tended to be less in DM2 again. In general, the more pronounced FA decrease in DM1 could be caused by its genetic effect which might have a larger impact on several distinct cell types in DM1, addressed recently by Lopez-Martinez et al. [44]. As described in their review, alternative splicing misregulations occur in cells of DM1 patients in different tissues such as brain tissue. Missplicing of the NMDA receptor 1 or the microtubule-associated protein tau was described to contribute to memory impairment or tauopathy-like degeneration of brain tissue, respectively. Since tau aggregations have been described to be associated with WM microstructural alterations in Alzheimer’s disease [45, 46], the tau-pathology could be also related to a pronounced FA decrease in DM1 leading to a widespread impact on several brain structures.

Since white matter was proved to be affected beyond white matter T2 lesions, FA findings indicate pathological disturbances of microstructural barriers to be present also in normal appearing white matter. However, the meaning and underlying causes for such microstructural damage remains unclear. In general, we observed similar lesion distributions for both disease types, with temporal white matter lesions seemed to be only occurring in DM1, which has been already shown by Minnerop et al. [19].

Similar to our study, Minnerop et al. (2011) found widespread differences in DTI measures between DM1 and HC, and less between DM2 and HC [19]. Comparing DM1 and DM2, they found less widespread differences between the two groups compared to our study. Although they did not describe the direct comparison between DM1 and DM2 regarding MD and RD changes, the comparison of both groups to HC yielded similar results to those we found. Thus, MD and RD were much more increased in DM1 than in DM2 when compared to HC.

In another investigation of cerebral FA alteration in a small sample of DM1 and DM2, the tendency of lower FA in DM1 compared to DM2 was shown, though only a rough division into larger brain lobes was performed and only five DM2 patients were included [47]. To our knowledge, no other study performed a comparison of white matter alterations between both disease types.

Other studies focused only on DM1 patients. For example, Baldanzi et al. (2016) also found extensive FA decreases compared to HC [48]. In another study investigating tractography, disruptions of white matter integrity were described for all tracts [49].

Both groups showed a relationship between disease duration and global MRI measures, but with different focuses. Thus, lower cortical thickness was associated with disease duration in DM2, while DM1 showed a significant relationship between disease duration and global FA decrease. This might be evidence for a different focus of brain alterations in the two disease types. Changes over time should be further analyzed in a longitudinal manner. Only two studies investigated longitudinal brain changes in DM1 [11, 50], and only one study included both types [51]. For example, Labayru et al. [50] found a progressive microstructural WM impairment in DM1, as global FA values were decreased over time.

In addition, a higher grade of muscular impairment was associated with decreased CT in both DM1 and DM2, but not with FA, which gave rise to the assumption that CT might be an imaging marker more related to muscle weakness and disability measured by MRS than global WM microstructural degradation. Still, there might be specific regions connected to muscular impairment rather than global WM integrity, which was not analyzed here. Based on our results, it would be worth to further investigate regional CT in relationship to physical constraints.

Surprisingly, CTG repeat length was positively related to CT in DM1. This contradicts other studies [23] and might be influenced by the inclusion of five childhood-onset patient. Childhood-onset patients showed significantly higher number of CTG repeats than adult-onset patients (Table 2) and simultaneously higher CT values (Supplement S5). Thus, there seemed to be an effect of high CTG repeat lengths on global CT, which had a high impact on the overall trend of the relationship between CTG repeats and CT and overshadowed the assumed negative association. Higher CT in childhood-onset DM1 might be explainable by the assumption that in younger ages, patients are able to compensate for atrophy and therefore they might not develop a rapid cortical thinning over a similar duration of the disease course as adult onset patients. No differences with regard to FA could be obtained for high or low numbers of CTG repeats, so that there might exist a distinguishable impact of repeat lengths on either FA or CT. We included both types of disease onset of DM1 in the other analyses as our childhood-onset patients showed only mild cognitive deficits and had no noticeable findings in their test results.

As far as we know, no differences between both patient groups regarding education have been described elsewhere. In contrast to the assumption that cognitive impairment in DM1 is usually more pronounced in DM1 than in DM2 the educational level of our DM1 patients was higher than in our DM2 patients. On the one hand, this could be related to a selection bias introduced by recruiting patients being interested and willing to participate in a clinical study, but this would account for both groups. On the other hand, our group of DM1 patients was obviously not a negative selection of cognitively impaired patients and therefore the neuropsychological and MRI findings are even more interesting and valuable. The findings indicate that in our DM1 patients, CTG repeat expansion sizes within the range of 120–1000 repeats were not generally associated with lower intelligence, frontal and memory dysfunction or reduced cognitive performance related to daytime sleepiness. This contradicts previous studies in which some neuropsychological test results were correlated to CTG repeat size [52, 53]. This could be related to genetic mosaicism [54] or selection bias towards inclusion of more educated patients into the study. Nevertheless, the educational levels in our DM1 patients and their neuropsychological test results underline that there is a strong need for individual neuropsychological testing to reduce inadequate presumptions and also for the evaluation of educational and professional levels in larger cohorts of DM1 and DM2 patients.

The importance of macro- and microstructural brain changes for cognitive functioning was shown by several associations between global CT or FA and pathologically relevant neuropsychological test results. Thus, complementary relationships were obtained for CT and FA, which were also differing between DM1 and DM2. Impaired selective attention, assessed by the d2-Test, was associated with global cortical thinning in DM1, but not in DM2, though the latter group showed a relationship to decreasing FA. Similarly, divided attention was negatively related to FA in DM2, but not in DM1. Therefore, attention deficits might underlie different mechanisms in the two disease types or have different extents in varying brain regions that might be functionally compensated by other regions. Again, based on our global analyses, it would be worth to further investigate regional CT and FA in relationship to neuropsychological alterations.

Previous studies that investigated the relationship between brain imaging measures and neuropsychology revealed heterogeneous results as reviewed by Minnerop et al. [17]. In one study, atrophy in the visual cortex was associated with reduced flexibility of thinking in DM1 and in DM2, depression was associated with brainstem atrophy and daytime sleepiness was correlated with atrophy in the middle cerebellar peduncles, pons, and the right medio-frontal cortex [9].

Regarding white matter integrity and cognitive decline, another study described a significant correlation between visual memory and WM integrity in the anterior cingulum, the splenium of the corpus callosum, and the precuneus [55]. Those analyses are indeed more specific compared to our global associations, which might be an explanation for our lacking relationship between, for example, selective attention and FA in DM1, and which would further point to the assumption of a compensatory effect by different brain regions. For DM2, results of the few existing studies are contradictory: in one study, no correlations between global brain volume and clinical parameters were found [21], whereas another study revealed significant correlations between brain volume and visuo-constructive abilities or psychomotor speed [16]. The involvement of GM abnormalities for visual abilities were confirmed in our study as cortical thinning was found to be accompanied by reduced visuospatial functioning. However, on one hand, small sample sizes in the mentioned previous studies of nine DM2 patients have to be considered. On the other hand, the comparability of the used neuropsychological tests are questionable.

Brain regions affected by white matter integrity alterations were widespread across the whole brain, so that an overlap with the more distinct cortical thinning was shown accordingly. Thus, microstructural changes seemed to be present beyond CT alterations. This is in line with previous studies suggesting that DM is a predominant white matter disease [12, 19]. On one hand, white matter changes might constitute an additional or underlying pathological mechanism with higher vulnerability. On the other hand, CT might provide a more specific and distinct biomarker for disease activity than DTI measures. Whether white matter changes were progressively increased and whether cortical thinning occurred later or to a lesser extent over time remains unclear and could be analyzed longitudinally. In the few existing longitudinal MRI studies, analyses did not comprise both CT and DTI measures [11, 50, 51], though it would provide evidence for their temporal evolution and their relationships.

Some limitations must be considered for interpretation of our results. First, this study might be based in absolute terms on a small sample of patients, although the number of patients is higher compared to the other studies. DM is a rare disease, which limits the possibilities to recruit patients that correspond to our strict inclusion criteria.

Second, no distortion correction could be done with raw data, because older scans were acquired without additional b0 with opposite phase encoding direction. This method was developed in the past years so that it was not common 8–10 years ago when our first dataset was acquired. Therefore, all data were handled the same and methods were not mixed.

Cortical thinning in DM is the main finding in our MRI analysis and was shown to be more widespread in DM1 as compared to DM2. There are overlapping pathological mechanisms in the two disease forms as shown by similarities in cortical thinning patterns. However, differences in micro- and macrostructural characteristics were observed as DM1 showed more pronounced deviations in FA and CT from HC. Our results might indicate the involvement of a high number of different neuronal cell types and of a more complex pathology. Cortical thinning in DM2 seemed to have a focus in the temporal lobe in addition to the occipital lobe, which is common in both disease types. These characteristics might help to further differentiate both types.

To sum up, we confirmed previously shown affection of brain white and grey matter in both disease types by use of larger sample size in comparison to previous studies and additionally showed extensive differences in white matter integrity between DM1 and DM2.