Identifying cellular signalling molecules in developmental disorders of the brain: Evidence from focal cortical dysplasia and tuberous sclerosis

We understand little of the pathogenesis of developmental cortical lesions, because we understand little of the diversity of the cell types that contribute to the diseases or how those cells interact. We tested the hypothesis that cellular diversity and cell–cell interactions play an important role in these disorders by investigating the signalling molecules in the commonest cortical malformations that lead to childhood epilepsy, focal cortical dysplasia (FCD) and tuberous sclerosis (TS).


INTRODUC TI ON
Epilepsy caused by focal cortical lesions in early childhood is a major cause of disability. 1 Our understanding of the cellular pathogenesis underlying these cortical lesions is limited because we understand little of the diversity of abnormal cell types within them, or how these different cell types interact to cause the disease.
Two of the most common developmental cortical lesions are focal malformations, focal cortical dysplasia (FCD) and cortical tubers in tuberous sclerosis (TS). FCD is the most common cortical lesion leading to multidrug-resistant epilepsy in children 2 and is characterised by disorganisation of the cortex and variable cytological abnormalities depending on the subtype. It is classified into three main types, FCD I, II and III. In particular, FCD type II is characterised by abnormally large pathological cells, dysmorphic neurons (DNs) in type IIA and IIB, and balloon cells (BCs) in type IIB. 3,4 Patients with the genetic disorder tuberous sclerosis develop cortical lesions (tubers) that are histologically indistinguishable from the lesions seen in FCDIIB. 5 In both diseases, the lesions are thought to arise through activation of the mTOR pathway. In TS, this is due to germline mutations in either TSC1 or TSC2 6 and in some patients with FCD, by somatic mutations in the mTOR pathway. [7][8][9] Our understanding of the pathology of the lesions may be over-simplified as it lacks consideration of other cells that may contribute to disease pathogenesis. The full extent of the cell types implicated in FCD and their interactions within the lesional tissue is still currently unknown. For example recent studies have suggested an important role for inflammatory cells and inflammatory components as well as oxidative stress in the pathogenesis. 10 In particular, a recent study demonstrated strong staining for the inflammatory markers iNOS, xCT, TLR4 and COX-2 in BCs, DNs and smaller glial cells. 11 These markers have been shown to be involved in initiating inflammatory responses and to contribute to seizure development. 12 We sought to identify potential interactions between pathological cells in FCDIIB and cortical tubers in an attempt to understand the cellular microenvironment that might lead to the pathology seen in these tissues. We hypothesised that cellular diversity and cell-cell interactions play an important role in these disorders and characterised novel small cell populations in these lesions and determined whether the mTOR signalling pathway, known to be aberrant in FCD may affect them.

Sample collection
Both fresh and formalin-fixed paraffin-embedded (FFPE) tissue were obtained from the histopathology department at Great Ormond Street Hospital. All samples were taken from excess surgical material and were reviewed by an experienced paediatric neuropathologist and classified according to the International League Against Epilepsy (ILAE) classification of FCD. 4 Normally formed cortex, lacking specific pathology, from patients undergoing temporal lobectomy for Hippocampal Sclerosis (HS) was used as control tissue. Ethical approval was granted (REC Approval # 05/Q0508/129).

RNA preparation and Affymetrix™ Human Exon 1.0ST microarray
Fifteen cases of different FCD subtypes (5 FCDIIB, 5 TS and 5 FCDIIA) were analysed. Another five samples of normal neocortex removed during surgery for HS were used as controls as they contained normally formed tissue that has previously been exposed to seizures. Fresh frozen samples from the selected cases were cut in half. One half was used for RNA extraction, and the other was used to generate frozen haematoxylin-and eosin-stained sections for pathological review and diagnostic confirmation. RNA was extracted using the Qiagen miRNeasy mini-kit. Eluted RNA was subjected to quality control by analysis on a NanoDrop 1000 spectrophotometer and samples with a 260/280 ratio lower than 1.8 were excluded. Further quality control was carried out using the Agilent Bioanalyzer platform and samples with an RNA Integrity Number (RIN) lower than four were excluded. The microarray was performed by UCL Genomics, using total RNA extracted from all 20 cases. RNA samples were processed on Affymetrix exon arrays CCL2 expressing cells decreased following inhibition of mTOR, the main aberrant signalling pathway in TS and FCD.

Conclusions:
Our findings highlight previously uncharacterised small cell populations in FCD and TS which express specific signalling molecules. These findings indicate a new level of diversity and cellular interactions in cortical malformations and provide a generalisable approach to understanding cell-cell interactions and cellular heterogeneity in developmental neuropathology.

Analysis of expression data
Hierarchical clustering analysis and heatmaps Expression data from Affymetrix Human Exon 1.0 ST v2 arrays were read into R using the oligo package. Data were normalised using the RMA method as implemented in oligo. Annotation for the arrays was retrieved from the huex10sttranscriptcluster.db R package and Affymetrix transcript cluster IDs within the dataset were matched to their respective gene symbols for downstream analysis. A minority of gene symbols were duplicated, due to multiple exons from the same gene being represented on the array.
In these instances, the duplicate entries were collapsed into one gene-level entry using the sum of the exon-level expression data. Subsequently, the median absolute deviation was used to determine the top 5000 most variable genes across the cohort.
Unsupervised consensus clustering was then performed using the ConsensusClusterPlus R package using the WardD2 implementation of Ward's clustering method, a maximum K of 6 and 10,000 repetitions. After consensus clustering, the resulting groups were used for differential expression testing, performed using the limma R package and following the protocol described in the package vignette. Cases were divided into three groups based on consensus clustering (groups 1, 2 and 3). Differential expression comparisons were generated with a cut off of 0.05 for false discovery adjusted p-value was used to determine significance.

DAB Immunohistochemistry
DAB immunohistochemistry (IHC) was used to validate markers identified by IPA in tissue sections and organotypic brain slice cultures.
Antibodies were selected based on online reviews in CiteAb (https:// www.citeab.com/), and those used are listed in Supplementary material 1. Immunohistochemical staining was performed using a Leica Bond-Max TM autostainer (Leica Microsystems) with appropriate positive and negative controls. Antigen retrieval methods were optimised for each antibody.

Scoring of IHC staining
For evaluating the IHC staining, the whole tissue area (containing grey and white matter) was used for analysis. The staining of extracellular markers was semi-quantitatively assessed and the intensity of staining was scored as follows for each cell type (0 = negative, 1 = weakly positive, 2 = moderately positive, 3 = strongly positive).
The mean score for each stain was calculated across different cell types and brain structures (BCs, DNs, normal neurons, small glial cells, endothelial cells, neuropil and white matter). For organotypic slice cultures, IHC staining was automated and images were then evaluated. Slides were digitalised and exported as 10× tiff figures.
The entire area was then automated and quantified using Fiji software and the particle or area quantification macro was used depending on the cellular localisation of the marker. Radar plots were used to compare the IHC scores between BC and control groups.

Immunofluorescence (IF)
FFPE slides were dewaxed in xylene and rehydrated through a graded alcohol series. Antigen retrieval was performed in a pressure-cooker at 125°C in EDTA-Citrate buffer pH 6.2. 10% hydrogen peroxidase was used to block endogenous peroxidase activity. Sections were blocked in 40% heat-inactivated sheep serum diluted in PBS with 0.2% bovine serum albumin, 0.15% glycine, 0.1% Triton X-100 at room temperature for 1 h. Sections were then incubated with primary antibodies at 4°C overnight in a sealed and humid chamber. The following morning, sections were rinsed with

Slice culture treatment with the mTOR inhibitor (Everolimus)
Six surgical samples from three patient cases (1 sample from 1 FCDIIB case and 5 independent samples from 2 TS cases) were used to prepare slice cultures using the technique by Stoppini et al. 15 Briefly, fresh samples were cut into 400μm-thick sections using a tissue chopper (Nickle Laboratory Engineering Co. Surrey). The standard thickness of organotypic brain slice cultures in the literature ranges from 100 to 400 µm. 16,17 We found that in our hands that 400 μm was the optimal thickness to maintain tissue integrity while allowing a detailed threedimensional analysis of the cell types and interactions between them in the tissue. This is in keeping with many other studies that have also found this tissue thickness to be optimal for slice cultures. [18][19][20][21][22][23][24] For optimisation experiments (supplementary_6), six individual samples derived from six patients were collected. The culture media was changed daily. Five days of post-treatment, slices were fixed in 10% non-buffered formalin before being processed for paraffin embedding.

Image J and statistics
ImageJ (https://fiji.sc/) was used to semi-quantify the level of CHI3L1, CCL2, pS6 and C-Caspase 3 staining. Slides were scanned using an Aperio CS2 slide scanner and exported at 10× magnification for further analysis. Statistical analysis of the slice culture staining was performed with SPSS version 22 using the Kruskall-Wallis test.

Passive clarity
The CLARITY technique was adapted from Deisseroth 25 and Gentleman et al. 26 Three cortical tubers from epilepsy surgery were sliced at a thickness of 400 μm and put into a hydrogel monomer comprised of 4% paraformaldehyde (PFA), 0.05% bisacrylamide (Bio-Rad), 4% (vol/vol) acrylamide (Bio-Rad) and 0.25% VA-044 photo-initiator (Wako) in PBS for 7 days. Tissue slices were transferred into a 6 ml container followed by a layer of olive oil to prevent oxygen interference during hydrogel poly-   Therefore, we undertook unsupervised clustering of the transcriptional profiles to identify the underlying biological groups, in a way that was not dependent on the variability of the cellular components in any individual sample ( Figure 1A).

Confocal microscopy 3D image construction
To determine the optimal number of groups that explained the variability in the data, we used the CDF plot ( Figure 1B). This revealed that cases clustered best into three molecularly defined groups and that a fourth group would not explain significantly more variability in the data. We analysed the differentially expressed genes between each of the three groups and these data are given in supplementary material 2.
There was a correlation between the underlying histopathology and the molecularly defined groups, for example the first group was a pure balloon cell group and consisted of four FCDIIB and three TS (group 1, n = 7), the second consisted of one FCDIIB, two TS, three FCDIIA and one HS case (group 2, n = 7), the third group consisted of two FCDIIA and four HS cases (group 3, n = 6) ( Figure 1A). However, as we predicted, this was not an absolute relationship between molecularly defined groups and histologically defined groups, possibly due to variability in sampling different types of cells in each sample.
We focussed on the molecularly defined groups to define expression patterns and in particular, focussed on differential gene expression between the pure BC and non-BC groups, that is group 1 and group 3.

Immunohistochemical analysis of differentially expressed genes
In order to understand cell-cell signalling in FCD, we focussed on DEGs that encoded for proteins located in the extracellular space.
In total, there were 133 DEGs encoding for extracellular molecules, immunoreactive for S100β and most expressed GFAPδ (83.6%), two glial progenitor cell markers. 28,29 In contrast, CHI3L1-positive cells were negative for neuronal markers (DCX, TuJ-1 and NeuN) and microglial markers (CD68 and Iba-1). These data suggest that CHI3L1-positive cells are, by phenotypic criteria, most likely to be glial progenitors.

Examining tissue architecture and cell types with organotypic slice cultures
In order to determine the functional relationship between these cells in FCDIIB and TS lesions, we used organotypic cultures. Slice cultures were optimised using non-lesional tissue according to established protocols. 15 We found optimal cell preservation with medium

DISCUSS ION
In this study, we explored the cellular diversity and cell-cell interactions playing a role in the pathology of developmental cortical lesions, specifically FCD and TS lesions. Using unsupervised transcriptional profiling, we clustered cases into molecularly defined groups and compared the differential gene expression between the most histologically diverse of these groups. Using these differentially expressed genes we identified extracellular signalling molecules and showed that the top 12 all showed differential expression in dysplastic compared to normal cortex when assessed by immunohistochemistry. In particular, we focussed on two secreted molecules, CHI3L1 and CCL2, that were expressed uniquely in small glial cells (glial progenitors and microglial respectively) only in dysplastic cortex and showed that they are under the control of the mTOR pathway. This suggests that there is a heterogeneous population of cells, regulated by mTOR signalling, and that there is a complex network of signalling underlying the abnormalities in dysplastic cortex.
We took an unbiased clustering approach to segregate cases as this allowed us to control for variability in the cellular composition of the samples. Moreover, it means we did not assume that the histologically defined subgroups represent uniform biological entities.
Through this analysis, we focussed on secretory molecules that were differently expressed between different groups as this is likely to reveal both cellular heterogeneity and networks of cell to cell signalling. While many molecules showed disease-specific differential expression (both by RNA expression and immunohistochemistry), the two that we focussed on, CCL2 and CHI3L1, were found to be Based on their morphology and previous literature, 30 we hypothesised that the CHI3L1-positive cells in FCD were most likely to be microglia or astrocytes. To test this hypothesis, a lineage-based immunofluorescence approach was undertaken, which showed that CHI3L1 expression did not co-localise with neuronal (DCX, TuJ-1 F I G U R E 5 Double immuno-labelling shows that CCL2 expression colocalised with microglial markers (Iba-1, MHC Class II, CD68) and a few with the astrocytic marker, GFAP. There was no colocalisation identified with neuronal markers, Neurofilament and TUJ-1. White arrowhead indicates CCL2 (+) cells, yellow arrowhead shows colocalisation with GFAP. Scale bar is 20 μm.
and NeuN) or microglial markers (CD68 and Iba-1) but did with glial and progenitor cell markers, suggesting the cells may be of glial progenitors.
In the central nervous system, up-regulation of CHI3L1 has been seen in many diseases, including encephalitis, 30 prion disease, 31 Alzheimer's disease, 32 amyotrophic lateral sclerosis, 33 multiple sclerosis, 34 trauma, 35 developmental status 36 and in glioblastoma. 37 The role of CHI3L1 in the central nervous system is unclear, but it has been suggested to be involved in tissue remodelling during inflammation, 38 to regulate excess inflammatory cell recruitment 35,39 and to promote cell survival. 40 Taken together with our data, this suggests that the CHI3L1-positive cells may either be part of an inflammatory response specific to FCDIIB/TS or may represent a developmentally abnormal cell population (in keeping with their progenitor cell phenotype).
We also found increased CCL2 expression in group 1. Double labelling studies showed that the majority of CCL2-positive cells co-localised with microglial markers (including CD68, Iba-1 and MHC class II) and a minority with astrocytic markers (GFAP).
The cells were mostly found in FCDIIB/TS cases, and some were found in a few FCDIIA cases but were not present in the normally formed cortex of patients with epilepsy. CCL2 has been previously described to be expressed not only in microglia but also in astrocytic cells and dysmorphic neurons. 41 Recently, a growing body of evidence has shown increased neuroinflammation in FCDIIB and TS. 41,42,44,45,46 The release of proinflammatory mediators (such as complement, cytokines, chemokines) not only triggers the immune system but also act as neuromodulators, which direct excitability and induced epilepsy. [47][48][49][50][51] The increase of the classical complement system components was reported in the TS a few years ago. 41 Several pieces of research also suggest that C1q and C3 might play a role in mediating the elimination of CNS synapses. [52][53][54][55] Moreover, the activation of both innate and adaptive immune responses in FCDII was also noticed. 42,44,56 Our data also indicate an up-regulation of markers involved in neuroinflammation

CO N FLI C T O F I NTE R E S T S
The authors declare no conflict of interest. The Editors of Neuropathology and Applied Neurobiology are committed to peer-review integrity and upholding the highest standards of review.
As such, this article was peer-reviewed by independent, anonymous expert referees and the authors (TSJ) had no role in either the editorial decision or the handling of the paper.

E TH I C A L S TATEM ENT
Ethical approval was obtained for this study. (REC Approval # 05/ Q0508/129).

DATA S H A R I N G
Most of the data is in the supplementary material. Additional data are available from the corresponding author.

PEER R E V I E W
The peer review history for this article is available at https://publo ns.com/publo n/10.1111/nan.12715.

O RCI D
Jessica C. Pickles https://orcid.org/0000-0001-7888-1723 Thomas S. Jacques https://orcid.org/0000-0002-7833-2158 F I G U R E 8 (A) Schematic diagram of the differences between the BCs and non-BCs (FCDIIA) groups according to the current literature. The difference between both groups is the presence of the balloon cells. In this proposed model, we suggest that there is more cellular heterogeneity in FCDIIB/TS and that the secretory factors CHI3L1 and CCL2 are produced by a population of small glia. (B) Our experiments indicate that these small glial and microglial cells reside close to BC's in FCDIIB and in cortical tubers from TS. We found that the expression of CHI3L1 and CCL2 diminished following inhibition of mTOR.