After enrolment the study cohort consisted of 122 participants, of whom 30 were classified as control participants, 29 as participants with type 2 diabetes without DPN and 63 as participants with type 2 diabetes with DPN (Table 1, Fig. 1). Between the groups, no significant differences were found regarding age, sex distribution, eGFR, lipoprotein a [Lp(a)], blood pressure, C-reactive protein and plasma cholesterol levels. Participants with type 2 diabetes showed higher fasting blood glucose, HbA1c and triacylglycerols as compared with control participants (Table 1).
Using human iPSC Schwann cells, we observed that high glucose alone was able to reduce the intracellular mRNA levels of PMP22, MPZ, PLP1 and MBP as well as MPZ protein expression significantly. This effect was even more pronounced with the addition of 2.5% fatty acids and could be partly ameliorated by the addition of insulin, mimicking potential conditions of the metabolic syndrome and type 2 diabetes (Fig. 2a,c,d,e,f). The decrease of intracellular mRNA levels of various myelination factors correlated significantly with the extracellular content found in the culture medium (Fig. 2a,b). In human nerve sections of the sciatic nerve of participants with and without DPN, the significant decrease of MPZ protein could be confirmed (Fig. 2g,h). Given this background, a study in humans trying to detect (extracellular) cmRNA for detecting and predicting the progression of DPN was initiated using the most promising targets derived from in vitro MBP, PLP1, PMP22 and MPZ. MBP cmRNA was detectable in all participants, but showed a high variability. PLP1 cmRNA could be detected in 79% of participants (96 of 122) and showed no association with DPN, whereas PMP22 cmRNA was detectable in only 58% (71 of 122) of participants but showed a significant decrease in participants with DPN (ESM Fig. 2). MPZ cmRNA was detectable in all participants and relative expression levels were 7.720 ± 2.609 in healthy control participants, 6.952 ± 2.323 in participants without DPN and 4.189 ± 2.528 in participants with DPN (p < 0.001 vs control participants and participants w/o DPN) (Fig. 3b). Multivariate distribution analysis revealed that reduced MPZ cmRNA levels were significantly associated with increased NSS and NDS, and decreased NCV of N. tibialis, N. peroneus and N. suralis (Table 2). Additionally, various QST variables correlated significantly with decreased MPZ cmRNA with a focus on tests reflecting the function of the myelinated A-δ fibres and A-ß fibres (Table 2). Furthermore, MRN analysis revealed that decreased MPZ cmRNA was significantly associated with decreased FA, consistent with NfL (Table 2, Fig. 4).
In order to study whether serum NfL is a useful marker for detecting and monitoring the progression of DPN, and also in order to test its performance against MPZ cmRNA, it was quantified in 119 participants. The values ranged from 3.67 to 55.2 pg/ml, with a mean value of 9.88 ± 3.72 pg/ml for healthy control participants, 12.51 ± 7.87 pg/ml in participants with type 2 diabetes without DPN and 14.69 ± 8.50 pg/ml in participants with DPN (p < 0.05 vs healthy control participants) (Fig. 3a). Multivariate regression analysis revealed that increased serum NfL levels were associated with increased NDS and decreased FA, as measured by MRN, and decreased NCVs of N. tibialis, N. peroneus and N. suralis (Table 2). Furthermore, various QST variables correlated significantly with increased serum NfL with a focus on sensory tests mediated by small unmyelinated C fibres (Table 2). MRN analysis revealed that increased serum levels of NfL were associated with decreased FA, which was reflected by either hypo- or hyper-intensive lesions of tibial and peroneal compartments of the sciatic nerve (Table 2, Fig. 4).
In order to test the discrimination between participants with and without DPN, receiver-operating characteristic (ROC) curve analysis was used. Using a cut-off value of 5.08, MPZ cmRNA achieved a sensitivity of 77.6% and specificity of 83.8%, with an AUC of 0.785, which was the best single marker regarding diagnostic performance (Table 3, Fig. 3c). In order to evaluate the improvement of already existing markers, ROC curves of combined markers were used: such as a standard model (diabetes duration, age, BMI, HbA1c), standard model + NCV of N. suralis or standard model + NCV of N. suralis + MPZ cmRNA. Using MPZ cmRNA, in addition to already established markers for the diagnosis of DPN, improved the diagnostic performance significantly from an AUC of 0.681 to 0.836 (Table 3, Fig. 3d). Due to the lack of discrimination between participants with DPN and without DPN, serum NfL showed only poor performance in a ROC analysis with an AUC of 0.564 (cut-off at 12.6 pg/ml).
Baseline laboratory and clinical profiles of participants of the follow-up study were comparable to those of the baseline cohort (ESM Table 3). At enrolment participants with high serum NfL levels displayed a hyperalgesic phenotype (gain of function), whereas participants with a loss of sensorimotor function (comparable to hypoalgesia) had significantly lower MPZ cmRNA levels (Fig. 5a,b). After 24 months, changes in QST were assessed (n = 90). Forty-three participants showed no change in symptoms, 21 participants experienced ‘more pain’ and 26 participants suffered a ‘sensory loss’ determined by extensive QST assessment, which are defined as changes in thermal/mechanical pain and/or detection, respectively (Fig. 5c). Prospective evaluation revealed that high NfL protein concentrations were associated with the development of increased pain after 24 months (Fig. 5e). Participants who developed a sensory loss after 24 months had significantly lower MPZ cmRNA levels at enrolment as compared with the ‘no change’ and ‘more pain’ groups (Fig. 5f). In a Kaplan–Meier curve, low MPZ cmRNA (cut-off <5.08) was significantly associated with the outcome of ‘sensory loss’ after 24 months (HR 6.519; χ2 = 13.03; 95% CI 2.53, 16.77) (Fig. 5d). Multivariate regression revealed that MPZ expression was independently associated with the development of sensory loss when tested vs age, male sex, BMI, diabetes duration, HbA1c and GFR. Participants with MPZ levels above the cut-off (>5.08) showed significantly lower risk (OR 0.651 [95%CI 0.39, 0.966], p < 0.05) for the loss of sensorimotor function (Table 4).
In order to validate the results of this pilot study, MPZ cmRNA levels were quantified in 33 external blinded serum samples of participants with type 2 diabetes of the prospective German Diabetes Study (ClinicalTrials.gov registration no. NCT01055093). Demographic, laboratory and clinical profiles of all participants showed no significant differences regarding age, sex, diabetes duration, BMI or HbA1c, but participants with diagnosed DPN had higher NSS and NDS as compared with those without DPN (ESM Tables 4, 5). The mean serum concentration of MPZ cmRNA was significantly lower in recently diagnosed participants with DPN compared with those without DPN (4.9 ± 2.9 vs 23.8 ± 7.6) (ESM Fig. 1d).