Comparisons between epilepsy and control groups
White matter myelination and glia
On qualitative analysis of TLE cases, MAG showed more intense labelling of myelinated axons in the gyral core and cortical radial fibres than the deep white matter (Fig. 1A-C) whereas more uniform labelling was noted in control PM white matter (Fig. 1D). PLP findings were similar to MAG but with an overall reduced intensity (Fig. 1F-I). Relatively uniform patterns of axonal labelling were noted in both core and deep white matter in TLE with neurofilament markers (NF-L, NF-M and NF-L) (Supplemental Fig. 5C-E). Olig2 showed diffuse increase in white matter oligodendroglia in TLE (Fig. 1K-N) and Iba1, increased white matter microglia relative to cortex and controls (Fig. 1P-S). PDGFRβ highlighted pericytes in relation to vessels in addition to scattered single multipolar-glial cells through the white matter and cortex (Fig. 1U-W). In TLE cases there was a clear gradient, with increased numbers in deep compared to gyral core white matter and qualitatively fewer PDGFRβ-positive glial cells in controls (Fig. 1X). NeuN labelled single interstitial neurons in both core and deep white matter in TLE and control groups (Supplemental Fig. 5A). Tenascin C showed predominant expression in the white matter matrix, extending into the deep cortex in some areas around blood vessels and glial cells (Supplemental Fig. 5B).
Quantitation of MAG was significantly lower in TLE than NEC and EPC in STG and MTG core and deep white matter ROIs (p < 0.001 and p < 0.05); PLP was significantly lower in TLE than NEC in these regions (p < 0.05) but not EPC (Fig. 1E,J, Suppl Table 2). In contrast, Olig2 density, PDGFRβ and Iba1 LI, were all significantly higher in white matter ROI in TLE than controls (p < 0.001) but with greater differences noted between TLE and NEC than EPC groups (Fig. 1O,T,Y, Suppl Table 2).
White matter microangiopathy
In some TLE cases scattered medium sized white matter vessels showed variable degrees of hyaline thickening of the vessel wall (vascular sclerosis/SVD) on H&E (Fig. 2A) although this was not evident in all vessels or cases. PVS expansion, with pigment laden macrophages and corpora amylacea was noted around some vessels, including those without sclerosis (Fig. 2B-D). COL4 + highlighted the basal lamina of white matter vessels of all calibre (Supplemental Fig. 1C). Frequent findings were ‘splitting’ of the COL4 + layer to produce a double-layer, particularly in vessels with expanded perivascular spaces and sclerosis (Fig. 2E-G). This microvascular dissection was not appreciated with PDGFRβ or SMA (Fig. 2H). PDGFRβ (Fig. 2I,K) and SMA (Fig. 2H,J) labelling was noted in arterioles and also the smallest capillaries. Vessels showed continuous circumferential labelling with SMA or PDGFRβ (termed Type 1) or discontinuous labelling (Type 2) in both TLE and controls (Fig. 2I,K and Supplemental Fig. 4D,E). Double labelling with PDGFRβ/SMA in selected cases and controls confirmed co-localisation of labelling in some arterioles and capillaries (Fig. 2R-U), however many smaller capillaries were PDGFRβ+/SMA- (Fig. 2U), whereas PDGFRβ-/SMA + vessels were not observed in either TLE or controls.
Vascular sclerotic index
In cerebrovascular disease an SI value of greater than 0.3 represents mild, and greater than 0.5 severe SVD [10]. The mean SI was 0.35 in TLE and significantly higher than controls (NEC 0.31, EPC 0.28; p < 0.001, Fig. 2L). The SI was not higher in three TLE patients with identified cerebrovascular disease risk factors (hypertension and/or Type 1 diabetes) than those without. Mean vessel diameters and PVS were lower in TLE than control groups but not the ratio between these values (Fig. 2M-O).
Vascular size, types, and densities
The median diameter of COL4 vessels was lower in TLE across all white matter ROI compared to control groups (p < 0.05 to 0.0001, Supp Table 2, Fig. 2P). COL4 and SMA vascular densities were both significantly higher in TLE than NEC in the MTG core but not other ROI (p < 0.005, p < 0.05, Mann Whitney test) (Supplemental Fig. 6A,B, Suppl Table 2). Classification of different vessel sizes quantified on PDGFRβ, showed greater differences in the deep than core regions in TLE than NEC, including higher capillary and arteriole densities in the MTGD (Mann Whitney, p < 0.05, Supplemental Fig. 6C). These findings support reduced mean size of small vessels in TLE, but without a uniform increase in small vessel densities across white matter ROI.
Pericyte distribution
SMA + and PDGFRβ + vessels were classified as type 1 or 2 based on complete or incomplete vascular coverage respectively. In TLE, PDGFRβ type 1 vessel densities were significantly greater, whereas type 2 PDGFRβ vessel densities were lower in deep white matter ROI in TLE compared to controls (p = 0.015 to p = 0.009) (Supplemental Fig. 6D,F, Supplemental Table 2); significant differences were not observed with SMA however (Supplemental Fig. 6H,J). There was evidence that Type 2 vessel were smaller in TLE than control groups in some ROI (SMA STGC p < 0.05; SMA MTGD p < 0.001; PDGFRβ STGD p < 0.05) whereas type 1 vessel diameters were larger than controls (PDGFRβ MTGC p = 0.04, STGD p = 0.04, MTGD p = 0.001) (Supplemental Fig. 6E,G,I,K). Paired tests showed significantly greater difference between type 1 and 2 vessels size with SMA in TLE cases than controls (p < 0.001 all ROI, Fig. 2Q). Analysis of double labelling for SMA and PDGFRβ showed higher densities of PDGFRβ+/SMA + and PDGFRβ+/SMA- in TLE than controls, reaching significance for type 2 vessels in the deep white matter (p = 0.028, Figure Supplemental Fig. 6I). These findings support increased pericyte vascular coverage of small vessels as well as altered relative expression of PDGFRβ and SMA in capillary pericytes in TLE.
Comparisons of gyral core to deep white matter
We further explored differences between gyral core and deep white matter for pathology variables, comparing average values across all gyri in TLE (Fig. 3) and control groups (Supplemental Fig. 7A,B). PLP LI was higher in the core than deep WM in TLE (p < 0.001) and controls (p < 0.05). For glial cells both Iba1 and PDGFRβ LI were increased in the deep white matter in TLE (p < 0.01) but no differences noted in controls whereas OLIG2 showed a core > deep gradient in control groups only (p < 0.05). For small vessels, notable findings were that type 1 vessel densities were higher in the deep than core white matter (for SMA and PDGFRβ (p < 0.01)) with an opposite gradient noted in controls (p < 0.05). In contrast PDGFRβ type 2 vessel densities were higher in the core white matter in TLE (p < 0.001) with no differences in controls. These observations highlight a superficial to deep white matter gradients for glio-vascular pathology in TLE which differs to controls.
Within both core and deep white matter, correlations between the vascular morphometric measures and glial/myelination were observed in TLE (Supplemental Fig. 7C,D). For example, SI positively correlated with PDGFRB glia and negatively with axonal neurofilament in the deep white matter. PVS negatively correlated with core PLP. Further correlations between glial, microglia, matrix protein Tenascin-C and small vessel densities suggest interactions in TLE.
RNA expression data in TLE
Epilepsy surgical and non-lesional postmortem cases showed clustering of cases into three groups with one group enriched with surgical cases (Fig. 1Za). 135 genes showed at least one-fold significant difference between postmortem and surgical cases (unadjusted P < 0.05, 40 genes when using Benjamini-Hochberg adjusted P value, P < 0.05, Fig. 1Zb), and bioinformatic analysis using Reactome revealed that these genes were associated with Developmental Biology and Signaling Transduction pathways including EGR2 and SOX10-mediated initiation of myelination and signaling by VEGF. Genes of interest associated with myelination and oligodendroglia including MAG, PLLP (proteolipid plasmolipin) and Olig2 showed significantly reduced mRNA expression in postmortem cases compared with surgical cases (p < 0.001), whereas PDGFRβ revealed higher gene expression in postmortem cases compared with surgical cases (p < 0.001) (Fig. 1Zc).
Pathology and RNA expression in HS cases
Most cases had HS (33/44), but we did not observe significant differences between mean core and deep white matter pathology variables in the fewer cases without HS apart from PDGFRβ type 1 vessels which had increased diameter and reduced PLP in core regions in HS (p≤ 0.05) (Supplemental Fig. 8A). Gene expression analysis however showed more significant findings with reduced MAG, PLLP and Olig2 in HS compared to non-lesional epilepsy surgical cases (p < 0.05, Supplemental Fig. 8B,C) in addition to genes known to interact with MAG such as MBP, SOX10 (P < 0.05) and NGFR (P < 0.05) according to STRING database [65]. MAG expression and genes in myelination pathway positively associated with genes in the angiogenesis and activated microglia and cytokines pathways but with greater correlation in non-HS cases (Supplemental Fig. 8D-H). These findings support evidence for a relatively greater reduction of myelination in HS/TLE and interaction of myelination, angiogenesis pathways and neuroinflammation in non-HS cases.
DWI metrics in relation to white matter pathology: DWI measures showed regional differences in TLE with RD values higher in the gyral cores than deep white matter whereas AD, MD, FA and FIXEL were higher in the deep regions (Fig. 3). Linear regression analysis of DWI values with pathology variables revealed greater significance in core than deep white matter regions (Supplemental Table 3). Glial labelling (Iba1 and PDGFRβ) increased with higher diffusion parameters (AD, RD and MD, p < 0.05 to < 0.0001) and higher myelin (PLP) labelling with lower AD (p < 0.05) and FA values in the core (Fig. 4A-D). Regarding vasculature, type 2 vessel density regressed with FA and FIXEL measures in the core white matter: higher PDGFRβ and SMA with lower FA and greater FIXEL respectively (p < 0.05) Fig. 4E,F). In the deep white matter, greater mean PVS associated with lower FA (Fig. 4G). In addition COL4 vessel diameters positively correlated with MD and AD in both core and deep white matter (Fig. 4H,I). These observations suggest that although DWI differences between core and deep white matter may partly be explained by anatomical differences in axon bundle organisation, alterations of glial density and vascular structures in TLE further influence diffusion measures (Figure. 5G). In the ten cases with paired DWI and gene-expression data from deep white matter, positive association between myelination genes (MOG, MAG, MBP, PLLP) and FA and negative relationship with RD was noted (Supplemental Table 3).
White matter pathology and neuropsychometry
Nine of 43 patients were classified as having verbal decline in cognitive function pre-operatively and 9/43 with impaired working memory at the time of surgery. Using logistic regression analysis for mean deep and core white matter values for groups with or without cognitive decline, we found higher PVS measure (Fig. 5A) and SMA type 2 vessel densities (Fig. 5B) in the decline group, as well as reduced vascular diameters for deep type 1 vessels (SMA and PDGFRβ) and type 2 vessels in core and deep (PDGFRβ) (Fig. 5C-F). There was also a trend for lower MAG LI (Supplemental Table 5). There were no significant relationships for pathology variables and impaired working memory impairment at the time of surgery. These findings support a relationship between microvascular alterations, including vessel calibre in those patients with TLE who have declined in their general verbal intellect over time prior to surgery.
White matter pathology and epilepsy history
A positive correlation between age and SI was noted in the post-mortem controls only (Spearman’s r = 0.594). There was no correlation between SI or PVS and duration of epilepsy in the TLE group but small vessel measurements were more abnormal the longer the epilepsy, including increased vascular diameter (SMA, p < 0.05) and reduced vascular density (COL4, p < 0.05) in the core white matter (Supplemental Fig. 9A,D). Regarding vessel types, type 1 vessel densities increased with age whereas type 2 vessels decreased with duration of epilepsy (PDGFRβ, Deep; p < 0.05) (Supplemental Fig. 9C,F). In addition, reduced axonal labelling was observed with age (NF-M, Deep ; p < 0.05) and more pronounced the longer the epilepsy (NF-H, Deep; p < 0.01) (Supplemental Fig. 9B,E). There were no differences for vascular, glia axonal or myelin markers in relation to hemisphere side. Gene expression of MAG, PLLP, Olig2 and PDGFRβ was compared between patients with an age of surgery lower or equal to 40 years and over 40 years of age; the older cohort showed lower expression of MAG, PLLP, OLIG2 mRNA whereas PDGFRβ RNA level appeared to be slightly higher (Supplemental Fig. 9D). These findings suggest that longer duration of epilepsy is associated with age-related alterations of myelin and microvasculature changes.