Structural changes in Schwann cells and nerve fibres in type 1 diabetes: relationship with diabetic polyneuropathy

Aims/hypothesis Our aim was to investigate structural changes of cutaneous Schwann cells (SCs), including nociceptive Schwann cells (nSCs) and axons, in individuals with diabetic polyneuropathy. We also aimed to investigate the relationship between these changes and peripheral neuropathic symptoms in type 1 diabetes. Methods Skin biopsies (3 mm) taken from carefully phenotyped participants with type 1 diabetes without polyneuropathy (T1D, n=25), type 1 diabetes with painless diabetic polyneuropathy (T1DPN, n=30) and type 1 diabetes with painful diabetic polyneuropathy (P-T1DPN, n=27), and from healthy control individuals (n=25) were immunostained with relevant antibodies to visualise SCs and nerve fibres. Stereological methods were used to quantify the expression of cutaneous SCs and nerve fibres. Results There was a difference in the number density of nSCs not abutting to nerve fibres between the groups (p=0.004) but not in the number density of nSCs abutting to nerve fibres, nor in solitary or total subepidermal SC soma number density. The overall dermal SC expression (measured by dermal SC area fraction and subepidermal SC process density) and peripheral nerve fibre expression (measured by intraepidermal nerve fibre density, dermal nerve fibre area fraction and subepidermal nerve fibre density) differed between the groups (all p<0.05): significant differences were seen in participants with T1DPN and P-T1DPN compared with those without diabetic polyneuropathy (healthy control and T1D groups) (all p<0.05). No difference was found between participants in the T1DPN and P-T1DPN group, nor between participants in the T1D and healthy control group (all p>0.05). Correlational analysis showed that cutaneous SC processes and nerve fibres were highly associated, and they were weakly negatively correlated with different neuropathy measures. Conclusions/interpretation Cutaneous SC processes and nerves, but not SC soma, are degenerated and interdependent in individuals with diabetic polyneuropathy. However, an increase in structurally damaged nSCs was seen in individuals with diabetic polyneuropathy. Furthermore, dermal SC processes and nerve fibres correlate weakly with clinical measures of neuropathy and may play a partial role in the pathophysiology of diabetic polyneuropathy in type 1 diabetes. Graphical Abstract Supplementary Information The online version contains supplementary material available at 10.1007/s00125-023-06009-z.

ROIvolume in the orthogonal max-projection confocal images.The density of nSC not abutting to nerve fibres (nSC-nerve complex, PGP9.5 -/S100 + /SOX10 + /DAPI + ) was estimated by counting the number of nSC-nerve complexes that were within the selected ROI, ∑Q -(nSCnerve), divided by the sampling ROIvolume in the orthogonal max-projection confocal images.
The sampling ROIvolume was measured by multiplying the area of ROI in the orthogonally projected 2D confocal images (∑a(ROIarea) with the thickness (21 µm) of original Z-stack images (disector height).The analysis was performed using Zeiss ZEN 3.

Comparison between methods for dermal nerve fibre and SC quantifications
To compare the results with previously published methods, we analysed the expression of subepidermal SC (including the subepidermal SC process (SSCP) and soma (SSCS)) and subepidermal nerve fibre (SNF) using a method optimized from a published study [1].The previously described method calculated the dermal structure density per length (not per volume as done here) and within 40 µm from the epidermal-dermal border (not 200 µm as done in DNF area fraction).Therefore, in addition to the method described above, we also quantified the solitary and total SSCS number density, SSCP density and SNF density per volume within 40 µm from the border (the modified methods), and per length within 40 µm from the border (the original methods).The details of the modified quantification methods are described below.

Comparisons between three groups: healthy controls, T1D and DPN (painless + painful)
Post hoc comparisons showed no significant difference between participants with painless DPN and painful DPN in any of the skin biopsy results.Therefore, to investigate the relationship between the skin biopsy quantifications and diabetic polyneuropathy, we combined all participants with DPN (regardless of their pain-status) into one single DPN group.
Same as the results in the four group comparisons, we also found significant differences in cutaneous SC quantifications (number density of nSC-nerve complexes, DSC area fraction, SSCP density per volume) and nerve fibre quantifications (DNF area fraction, SNF density per volume) between the three groups (all p and padj<0.05),but not in the number density of nSC+nerve complexes, solitary and total SSCS per volume (all padj>0.05;ESM Table 1).Furthermore, post hoc comparisons in the above-mentioned five quantifications within the three study groups showed significant differences between DPN and healthy control group, as well as between DPN and T1D group (all p and padj<0.05);but no difference between T1D and healthy control group (all p and padj>0.05;ESM Table 2).
3 (blue edition) software.NV(nSC+nerve) = ∑Q -(nSC+nerve) / (∑a(ROIarea) * disector height) NV(nSC−nerve) = ∑Q -( nSC−nerve) / (∑a(ROIarea) * disector height) 2) Total dermal Schwann cell (DSC) expression level was estimated by measuring the DSC area fraction.Briefly, the DSC area fraction was estimated by measuring the area that was occupied by the total DSC (∑a(DSCarea)), divided by the total area of region of interest (∑a(ROIarea)) in the dermal area within 200 µm below epidermal-dermal border from the orthogonally projected 2D confocal images.DSC area fraction = ∑a(DSCarea) / ∑a(ROIarea) Dermal nerve fibre area fraction Dermal nerve fibre (DNF) expression level was estimated by measuring the DNF area fraction (the same method as measuring the DSC area fraction).Using FIJI, an individual threshold was applied to each orthogonal average-projection confocal image to exclude as much as possible of non-specific bindings from the images.DNF area fraction was estimated by measuring the DNF occupied area (∑a(DNFarea)), divided by the total area of region of interest ∑a(ROIarea), in the dermal area within 200 µm below epidermal-dermal border from the orthogonally projected 2D confocal images.DNF area fraction = ∑a(DNFarea) / ∑a(ROIarea)
All continuous variables were non-normally distributed and were presented as median (IQR).After square root transformation, variables were analyzed using both ANOVA (p value) and ANCOVA (padj).value,adjusting for age, sex and HbA1c).Variables that could not be transformed were analyzed by Kruskal-Wallis H test (nonadjusted p value).Square root transformed (parametric) data are presented as adjusted p value and mean difference with 95% confidence interval analyzed by pairwise comparisons t tests adjusting for age, sex and HbA1c, and nonparametric data are presented as non-confounder adjusted p value by using Dunn's test.*p values < 0.05; **p values < 0.01; and ***p values < 0.001.