New clinicopathological associations and histoprognostic markers in ILAE types of hippocampal sclerosis

Abstract

Mesial temporal lobe epilepsy with hippocampal sclerosis (MTLE‐HS) is a heterogeneous syndrome. Surgery results in seizure freedom for most pharmacoresistant patients, but the epileptic and cognitive prognosis remains variable. The 2013 International League Against Epilepsy (ILAE) histopathological classification of hippocampal sclerosis (HS) has fostered research to understand MTLE‐HS heterogeneity. We investigated the associations between histopathological features (ILAE types, hypertrophic CA4 neurons, granule cell layer alterations, CD34 immunopositive cells) and clinical features (presurgical history, postsurgical outcome) in a monocentric series of 247 MTLE‐HS patients treated by surgery. NeuN, GFAP and CD34 immunostainings and a double independent pathological examination were performed. 186 samples were type 1, 47 type 2, 7 type 3 and 7 samples were gliosis only but no neuronal loss (noHS). In the type 1, hypertrophic CA4 neurons were associated with a worse postsurgical outcome and granule cell layer duplication was associated with generalized seizures and episodes of status epilepticus. In the type 2, granule cell layer duplication was associated with generalized seizures. CD34+ stellate cells were more frequent in the type 2, type 3 and in noHS. These cells had a Nestin and SOX2 positive, immature neural immunophenotype. Patients with nodules of CD34+ cells had more frequent dysmnesic auras. CD34+ stellate cells in scarce pattern were associated with higher ratio of normal MRI and of stereo‐electroencephalographic studies. CD34+ cells were associated with a trend for a better postsurgical outcome. Among CD34+ cases, we proposed a new entity of BRAF V600E positive HS and we described three hippocampal multinodular and vacuolating neuronal tumors. To conclude, our data identified new clinicopathological associations with ILAE types. They showed the prognostic value of CA4 hypertrophic neurons. They highlighted CD34+ stellate cells and BRAF V600E as biomarkers to further decipher MTLE‐HS heterogeneity.

Introduction

Mesial temporal lobe epilepsy with hippocampal sclerosis (MTLE‐HS) is often resistant to anti‐epileptic drugs 32. Epilepsy surgery results in a decrease in the frequency of the seizures or in their cessation in most cases of pharmaco‐resistant MTLE‐HS. In some patients, however, epilepsy may recur after surgery and cognitive impairments, especially memory deficits, may be observed 15, 46. The uncertainty of postoperative prognosis makes difficult the decision to operate on a case of MTLE‐HS. A better understanding of the histopathological heterogeneity and of its effects on seizure frequency and cognitive outcome after surgery is needed.

Clinical studies describe a variable etiology and history for MTLE‐HS patients 17, 45. The natural history often begins with an initial precipitating injury (IPI). IPIs in infancy—most frequently febrile seizures (FS)—are considered causative for MTLE‐HS 13. Spontaneous seizures occur after a latent period. Tumors or focal cortical dysplasias/FCD may also induce HS and MTLE 18.

The neuropathology of HS is heterogeneous. There are diverse patterns of neuronal loss, gliosis and changes in the granule cell layer (GCL) which have been linked to surgical outcomes 15. Several classifications have been proposed 47. An international consensus reached in 2013 4, 5, 9, 40 defined three patterns of HS (type 1 with severe neuronal loss in CA1 and CA4, type 2 with predominant loss in CA1, type 3 with predominant loss in CA4) and a pattern “noHS, gliosis only” without neuronal loss. Type 1 is the most frequent and typical aspect with most severe neuronal loss in the Ammon horn, dense gliosis, GCL dispersion and often CA4 neurons with increased size. Such CA4 neurons were previously described as “hypertrophic neurons” by Thom et al 38, 39. However, the ILAE Diagnostic Methods Commission established a consensual definition of “hypertrophic neuron” in the context of FCD 6 but not in the context of HS 5. The term “CA4 hypertrophic neurons” here refers to CA4 pyramidal neurons with increased size of the cell soma and nucleus.

CD34 immunopositive (CD34+) multipolar or stellate cells have been described in distinct forms of epilepsy‐related cerebral lesions including ganglioglioma, dysembryoplastic neuroepithelial tumors (DNET) 2, 29, multinodular and vacuolating neuronal tumors (MVNT) 16, 41 and FCD 10, 23. The nosology of lesions containing these CD34+ cells without another specific tumoral component is not consensual and they are interpreted by some authors as the periphery of a ganglioglioma or as a non‐specific/diffuse form of DNET 43. In this context, we preferred to use a descriptive definition of nodules of CD34+ stellate cells and of scarce CD34+ stellate cells. CD34+ cells express the intermediate filament Nestin but not Glial Fibrillary Acidic Protein, (GFAP)—a pattern reminiscent of the phenotype of immature neural cells 43. The tumoral, dysplastic or reactive nature of CD34+ cells in samples of epilepsy surgery remains debated and their contribution to the nosology of MTLE‐HS is not known.

We investigated clinicopathological correlations using the ILAE classification of HS types and of the pattern no HS in a monocentric series of 247 patients undergoing surgery for MTLE‐HS including hippocampus resection. We characterized the CD34+ cells in these lesions to enhance the nosology of MTLE‐HS. We previously reported that ILAE types did not have prognostic value for the post‐surgical outcome of epilepsy in this cohort 25. We tested if the histopathological features associated with ILAE types might have a prognostic value.

Material and Methods

Patients and samples

The histopathological features of hippocampectomy samples of all adult patients (i) with MTLE‐HS (ii) treated with surgery at the Pitié‐Salpêtrière Hospital between 1995 and 2015 and (iii) with available histological material were analyzed (Supporting Information Figure 1). Surgical samples of hippocampal tumors, glial scars, vascular malformations were excluded. Three surgical approaches were used: anterior temporal lobectomy (ATL, n = 127), transcortical selective amygdalohippocampectomy (n = 114) and transylvian selective amygdalohippocampectomy (n = 6) 25. The procedures taken to protect personal data in this work were approved by the French National Technologies and Civil Liberties Commission. Collected data included age, gender, medical history, seizure types, surface and when performed, intracranial EEG records, presurgical neurological and cognitive status, presurgical anti‐epileptic treatment and postsurgical outcome. No history of limbic encephalitis was noted. MRI findings were reviewed by a neuroradiologist (AB) for MVNT and “BRAF V600E+ HS” as defined hereafter.

Pathology

Hippocampal samples were examined microscopically after hematoxylin and eosin staining and immunostaining for GFAP and NeuN. Temporal lobe was examined in patients treated by ATL. Samples were included only when CA1 and CA4 hippocampal subfields could be evaluated. Samples lacking CA1 or CA4 and fragmented samples were excluded (Supporting Information Figure 1). The anatomic definition of CA1, CA2, CA3 and CA4 was as defined previously 5. The diagnosis of tumors was defined by WHO 2016 22. FCD was classified as previously defined 6. Two neuropathologists (FB and ALCG) separately attributed each sample to a ILAE type of HS 5, and reviewed slides together, using a multi‐head microscope, so that a consensus conclusion could be agreed if initial classifications diverged. Unresolved cases were reviewed by a third neuropathologist (CD).

The neuronal loss in CA1, CA2, CA3, CA4 and in the GCL was evaluated semi‐quantitatively from 0% to 100% by 10% increments in comparison with controls. As controls, post‐mortem samples of hippocampus from patients without neurological symptoms and without brain lesion were used. Consent was obtained from the legally authorized family members. Changes in the distribution of granule cells were grouped under two subheadings: “dispersion” when the layer was more than 10 neurons thick 45 and “duplication” when two layers could be readily distinguished. “Broadening” of the GCL means both “dispersion” or “duplication.” When both dispersion and duplication were present, the sample was classified in the duplication subgroup. We also took note of CA4 hypertrophic neurons.

Immunohistochemistry

Formalin‐fixed paraffin‐embedded tissue sections (3 μm of thickness) were deparaffinized and immunolabelled with a Ventana Benchmark XT stainer (Roche, Basel, Switzerland). The secondary antibodies were coupled to‐peroxidase with diaminobenzidine as brown chromogen or to alkaline phosphatase with Fast Red (ultraView Universal Alkaline Phosphatase Red Detection Kit, VentanaVR) as red chromogen. Primary antibodies were as follows: BRAF V600E (mouse monoclonal VE1, 1:100, Abcys Eurobio), CD31 (monoclonal mouse JC70A, 1:30, Dako) CD34 (monoclonal mouse QBEnd10, 1:50, Dako), ELAVL2 (rabbit polyclonal, 1:500, Proteintech), GFAP (monoclonal mouse 6F2, 1:500, Dako), Internexin alpha/INA (monoclonal mouse 2E3, 1:100, Quartett Biochemicals), Nestin (monoclonal mouse 10C2, 1:2000, Sigma‐Aldrich), NeuN (Monoclonal mouse MAB377, 1:100; Merck Millipore), OLIG2 (rabbit monoclonal EP112, 1:200, Epitomics), P53 (monoclonal mouse, DO‐7, 1:100 Dako), SOX2 (rabbit monoclonal SP76, 1:300, Sigma‐Aldrich), Vimentin (monoclonal mouse V9, 1:50 Dako). The quality of immunolabelings was assessed by an external positive control for BRAF V600E (a BRAF V600E mutant colic adenocarcinoma was added on each slide) and P53 (an IDH mutant and P53 mutant anaplastic astrocytoma was processed on another slide). The quality was assessed by internal positive controls for the other antibodies. The CD34 immunolabeling was unsuccessful in 8 cases (Supporting Information Figure 1). It was tested only on the hippocampus, the extra hippocampal region was not tested in cases of ATL. CD34 immunolabeling was classed as “CD34–” if only endothelial cells were positive, as “CD34+ scarce” if only a few isolated extravascular stellate cells were detected, or as CD34+ nodular if the extravascular stellate cells detected were grouped in nodules in the hippocampus. We further characterized all “CD34+ nodular” samples (n = 8) and a subset of “CD34+ scarce” samples (n = 11) by BRAF V600E, P53, INA and ELAVL2 immunolabelings.

Statistical analysis

Quantitative variables are shown as mean ± standard deviation. The nominal variables were tested with chi‐squared test or exact test of Fisher when appropriate. Continuous variables were tested by Kruskall–Wallis test for three groups, the Mann–Whitney test for two groups and the t test when variable was normally distributed. Spearman's correlation coefficient was used to test the correlation of continuous variables. The postsurgical outcome was evaluated by Kaplan‐Meier plots for the Engel I and Ia classes across time. Differences of outcome were assessed by the log‐rank test. The significance level was on‐adjusted P < 0.05, two‐sided test. Statistical analysis was performed with the software StatView 5.0.

Results

General features of the whole series

Our series included 247 patients with mean age at surgery of 38.2 ±10.9 years and gender ratio of 1:1. Its features are summarized in Table 1 and Supporting Information Table 1. The mean age at epilepsy onset and the mean duration of epilepsy were 12.3 ± 9.3 years and 26.0 ±11.7 years, respectively. Both parameters were distributed unimodally (Supporting Information Figure 1). The mean duration of epilepsy was stable across time (rho = 0.08 P = 0.22).

Table 1.

Clinicopathological features in 247 patients with MTLE. n was precised if it was inferior to the total number of patients (247).

	n (%)
Clinical features
Male	121 (49)
IPI	191/246 (78)
IPI age (years)	2.0 ± 3.1
Febrile seizure	137/246 (56)
Family history of epilepsy	50/175 (29)
Age of onset of epilepsy (years)	12.3 ± 9.3
Epilepsy duration before surgery (years)	26.0 ± 11.7
Right side of hippocampectomy	119 (48)
Number of seizures per month	19 ± 30
Number of antiepileptic drugs	1.7 ± 0.8
Status epilepticus	19/211 (9)
Abnormal hippocampus on MRI	219/227 (96)
Cognitive impairment:
None	17/233 (7)
Verbal	88/233 (38)
Non‐verbal	48/233 (21)
Global	80/233 (34)
Histological features
CA1 neuronal loss	81% ± 19
CA2 neuronal loss	39% ± 19
CA3 neuronal loss	58% ± 24
CA4 neuronal loss	62% ± 22
GCL neuronal loss	25% ± 23
GCL pattern normal	56/239 (23)
GCL pattern dispersion	158/239 (66)
GCL pattern duplication	25/239 (10)
Hypertrophic CA4 neurons	106 (43)
Dysplasia	1 (0,4)

Histopathological correlates of ILAE types

ILAE type 1 was found in 75% of cases. ILAE types 2, 3 and the pattern noHS were present in 19%, 3% and 3% of the samples, respectively (Table 2 and Supporting Information Table 2). The kappa coefficient testing the inter‐observer reproducibility of ILAE type assessment was 0.612 (generally considered as “substantial” agreement). The distribution of neuronal loss across ILAE types was similar to the previously reported one (Supporting Information Figure 2). The size of CA4 hypertrophic neurons was significantly increased compared to CA4 pyramidal neurons of a control hippocampus (Supporting Information Figure 3). CA4 hypertrophic neurons were detected more often in type 1 than in either type 2, 3 or noHS (type 1: 57%, type 2: 15%, type 3: 14%, noHS: 0%, P = 2 × 10⁻⁷, Table 2). In type 1, CA4 hypertrophic neurons were associated with higher neuronal loss in CA1 (P = 0.047). One dysplasia was observed in a type 3 HS patient and corresponded to an increased density of neurons in CA3 associated with abnormal ganglionic aspect of neurons and abnormal astrocytic cells.

Table 2.

Clinical and pathological features associated with ILAE types. n was precised if it was inferior to the total number of each group. Abbreviation: sEEG = stereo‐electroencephalography.

	ILAE type 1 n = 186	ILAE type 2 n = 47	ILAE type 3 n = 7	ILAE type noHS n = 7	P value
Clinical features
IPI	144 (77%)	42 (89%)	3/7 (43%)	2/6 (33%)	0.001
Febrile seizure	104 (56%)	32 (68%)	0 (0%)	1/5 (17%)	6 × 10−4
sEEG	14/172 (8%)	1 (2%)	1 (14%)	5 (71%)	6 × 10−5
Abnormal hippocampus on MRI	168/170 (99%)	43/43 (100%)	6 (86%)	2 (29%)	7 × 10−8
Histological features
CA1 neuronal loss (%)	85 ± 8	84 ± 9	10 ± 17	0 ± 0	<10−4
CA2 neuronal loss (%)	44 ± 18	30 ± 16	3 ± 8	0 ± 0	<10−4
CA3 neuronal loss (%)	69 ± 16	34 ± 18	35 ± 34	0 ± 0	<10−4
CA4 neuronal loss (%)	72 ± 12	32 ± 9	47 ± 23	0 ± 0	<10−4
GCL neuronal loss (%)	29 ± 23	18 ± 21	17 ± 29	0 ± 0	<10−4
Hypertrophic CA4 neurons	98 (57%)	7 (15%)	1 (14%)	0	1 × 10−7

Histopathological correlates of granule cell layer alterations

The GCL was classified as of normal width (23%), dispersed (66%) or duplicated (10%, Table 3, Supporting Information Table 3, Supporting Information Figure 3). The GCL was significantly more likely to be dispersed or duplicated in type 1 than in noHS (type 1 81%, no HS 2/7 29%, P = 0.005). GCL dispersion was associated with higher neuronal loss in the CA2 and CA3 fields (P = 0.045 and P = 0.003, respectively). GCL broadening was associated with higher neuronal loss in the CA4 field (P = 0.003).

Table 3.

Clinicopathological features associated with granule cell layer changes. n is precised if it was inferior to the total number of each group

	Normal GCL thickness n=56	GCL broadening n=183	Statistical tests Norm. vs. Abnorm. P value	Dispersion of GCL n=158	Duplication of GCL n=25	Statistical tests Norm. vs. Disp. and Dup. P value
Clinical features
Sex (male)	34 (61%)	83 (45%)	0.044	73 (46%)	10 (40%)	0.11
Intracranial infection	12/55 (22%)	16 (9%)	0.008	11 (7%)	5 (20%)	0.005
Gen. seizure	40 (71%)	131/182 (72%)	0.94	107/157 (68%)	24 (96%)	0.02
Stat. epilepticus	1/48 (2%)	17/156 (11%)	0.08	13/136 (10%)	4/20 (20%)	0.04
Abnormal hippocampus on MRI	45/50 (90%)	166/169 * (98%)	0.02 *	144/147 (98%)	22/22 (100%)	0.04
Histological features
ILAE type 1	34 (61%) *	144 (79%) *	0.01 *	125 (79%) *	19 (76%)	0.04 *
ILAE type 2	14 (25%)	33 (18%)		27 (17%)	6 (24%)
ILAE type 3	3 (5%)	4 (2%)		4 (3%)	0 (0%)
ILAE noHS	5 (9%) *	2 (1%) *		2 (1%) *	0 (0%)
CA1 neur. loss (%)	76 ± 27	82 ± 16	0.74	81 ± 17	86 ± 7.1	0.48
CA2 neur. loss (%)	35 ± 20	40 ± 19	0.13	41 ± 20	33 ± 13	0.045
CA3 neur. loss (%)	51 ± 26	61 ± 23	0.02	63 ± 24	51 ± 19	0.003
CA4 neur. loss (%)	53 ± 25	64 ± 20	0.003	64 ± 21	65 ± 19	0.011
GCL neur. loss (%)	27 ± 30	25 ± 21	0.63	24 ± 21 (n=157)	27 ± 21	0.74

*GCL broadening is significantly more frequent in type 1 compared to type noHS (P = 0.005). GCL dispersion is significantly more frequent in type 1 compared to noHS (P = 0.008).

Abbreviations: GCL = Granule Cell Layer; Gen. seizure, generalized seizure; neuron. loss, neuronal loss; Stat. epilepticus, status epilepticus.

Histopathological correlates of CD34 immunoreactive stellate cells

Three patterns were recognized: “CD34–,” “CD34+ scarce” and “CD34+ nodular” (Table 4; Supporting Information Table 4 and Figure 1). The CD34+ cells presented ramified cytoplasmic processes with a stellate aspect. The nucleus had slight anisokaryosis and nuclear membrane irregularities. Nestin immunolabeling was positive in endothelial cells but also revealed a stellate pattern in areas with CD34+ stellate cells (n_CD34+nodular=7, n_CD34+scarce=10). Co‐immunolabeling confirmed a co‐expression of Nestin and CD34 by stellate cells (n_{CD34+ nodular}=4, n_CD34+scarce=3). No stellate pattern was observed with Vimentin (n_{CD34+ nodular}=7, n_CD34+scarce=11 Figure 1) nor CD31 (n_{CD34+ nodular}=5, n_CD34+scarce=7 data not shown) whereas both markers are expressed by endothelial cells. Co‐immunolabeling showed a co‐expression of CD34 and SOX2 in some CD34+ stellate cells (n = 3 Figure 2) whereas SOX2 was negative in endothelial cells (n = 19). Co‐immunolabeling showed a co‐expression of CD34 and OLIG2 in some CD34+ stellate cells (n = 3 Supporting Information Figure 4) whereas OLIG2 was negative in endothelial cells (n = 10). Together CD34+ stellate cells are Nestin+, SOX2+/–, OLIG2+/–, CD31–, Vimentin‐ and can be distinguished from endothelial cells that are CD34+, Nestin+, SOX2–, OLIG2–, CD31+ and Vimentin+.

Table 4.

Clinical and pathological features of hippocampal sclerosis according to CD34 status. n is precised if it was inferior to the total number of each group.

	CD34 negative n = 196	CD34+ stellate cells n = 43	P value	“CD34+ scarce” stellate cells n = 35	“CD34+ nodular” stellate cells n = 8	P value
Presurgical clinical features
Fetal distress	6/189 (7%)	2/41 (5%)	0.63	0/33 (0%)	2 (25%)	0.02
Dysmnesic aura	59/176 (34%)	19/36 (53%)	0.037	12/28 (43%)	7 (88%)	0.006
sEEG	13 † (7%)	8 † (19%)	0.031	7 † (20%)	1 (13%)	0.034
Abnormal hippocampus on MRI	178/182 (98%)	34/38 (89%)	0.032	26/30 (87%)	8 (100%)	0.021
Histological features
ILAE type 1	156 ‡ (80%)	23 ‡ (53%)	0.0002	22/35 § (63%)	1/8 § (12%)	3 × 10−6
ILAE type 2	34 ‡ (17%)	12 ‡ (28%)	<10−4	9/35 (26%)	3/8 § (38%)	0.002
ILAE type 3	4 (2%)	3 (7%)	0.06	0/35 (0%)	3/8 § (38%)	0.044
ILAE noHS	2 ‡ (1%)	5 ‡ (12%)	0.03	4/35 § (11%)	1/8 § (12%)	0.21
CA1 neuronal loss(%)	83 ± 16	70 ± 30	<10−4	74 ± 28	45 ± 42	0.003
CA2 neuronal loss (%)	40 ± 19 (153)	29 ± 22 (35)	0.01	31 ± 22 (29)	20 ± 18 (6)	0.008
CA3 neuronal loss(%)	60 ± 23 (134)	49 ± 31 (33)		51 ± 30 (28)	40 ± 37 (5)
CA4 neuronal loss(%)	64 ± 21	48 ± 27		50 ± 27	43 ± 29
GCL neuronal loss(%)	27 ± 24 (188)	17 ± 22 (42)		19 ± 22	9 ± 17

^†The “CD34+ scarce” group and the “CD34+” group were significantly associated to sEEG compared to “CD34 negative” group.

^‡CD34+ stellate cells were significantly more frequent in type 2 and in noHS than in type 1.

^§“CD34+ scarce” stellate cells were significantly associated with noHS compared to type 1 (P = 0.004). The “CD34+ nodular” group was significantly associated with type 2 (P = 0.023), type 3 (P = 0.0002) and noHS (P = 0.0004) compared to type 1.

Abbreviations: sEEG = stereo‐electroencephalography.

Figure 1.

CD34+ cells are associated with rarer patterns of hippocampal sclerosis. A–D, I–J. A CD34+ nodular case. CD34 immunopositive/CD34+ stellate cells formed a nodule (solid arrowheads in (A‐B). Nestin immunolabeling showed a frequent extravascular stellate pattern in the area of the CD34+ nodule (solid arrowhead in C) whereas Vimentin immunolabeling showed no extravascular immunolabeling (D). E–H, K–L. A CD34+ scarce case. E. An isolated CD34+ extravascular stellate cell is observed (open arrowheads in E, F). Nestin immunolabeling showed a scant stellate extravascular pattern in the area of the CD34+ stellate cell (open arrowhead in G) whereas Vimentin immunolabeling showed no extravascular immunolabeling (H). Double immunolabelings showed the co‐expression in stellate cells of Nestin (brown) and CD34 (red) (I), and SOX2 (pink) and CD34 (brown) (J) in a CD34+ nodular case. Double immunolabelings showed the co‐expression in stellate cells of Nestin (brown) and CD34 (red) (K, higher magnification in insert), and of SOX2 (pink) and CD34 (brown) (L) in a CD34+ scarce case. M–P. Distribution of CD34 immunolabeling (CD34 negative in black, CD34+ scarce in orange and CD34+ nodular in green) for type 1 (M), type 2 (N), type 3 (O), the pattern noHS (P). Q–S. A CD34+ nodular case with BRAF V600E immunopositive neurons. Q. A moderate immunolabeling of scarce cells by P53 (solid arrowheads). The BRAF V600E immunolabeling is positive in neurons (solid arrowheads in r,s). T. A CD34+ scarce case with BRAF V600E+ neurons (open arrowhead). Scale bars: (A,E) 400 μm; (B–D,F–H) 100 μm; (I–L,T) 50μm; (Q,R) 200μm; (S) 20μm.

Figure 2.

Classification of CD34+ hippocampal sclerosis according to BRAF V600E and INA status. A,C,E,G–I. A case of “BRAF V600E+ HS” with BRAF V600E positive neurons. B,D,F,J–L. A case of hippocampal MVNT. A,B. CD34 immunolabeling showed a nodular pattern of CD34 positive stellate cells (solid arrowheads). C,D. NeuN immunolabeling showed a neuronal loss that is predominant in CA4 (solid arrowheads) corresponding to HS ILAE type 3. E. Absence of INA positive abnormal neurons. F. Presence of numerous abnormal INA+ neurons (solid arrowheads). G–I. Absence of abnormal neurons according to NeuN, ELAVL2 and INA immunolabelings in the white matter of CA2 (open arrowheads) in a BRAF V600E+ HS. J–L. Presence of Neun Negative, ELAVL2 positive, INA positive abnormal neurons (solid arrowheads) in the white matter of CA2 in a MVNT. Scale bars: (A–F) 2.5 mm; (G–L) 200 μm. Abbreviations: INA = internexin alpha; MVNT = multinodular and vacuolating neuronal tumor.

One hundred and ninety six samples were “CD34–,” 35 were “CD34+ scarce” and eight cases were “CD34+ nodular.” Samples with CD34+ stellate cells were significantly associated with type 2 (26%) and with noHS (71%) rather than with type 1 (13%, P = 0.0002). “CD34+ scarce” was associated with noHS (57%) compared to the type 1 (12%, P = 0.004). “CD34+ nodular” labeling was more frequently observed in the type 2 (7%, P = 0.023), in the type 3 (43% P = 0.0002) and in noHS samples (43%, P = 0.0004) rather than in the type 1 (0.6%).

BRAF V600E positive mononucleated neurons were observed in the CD34+ nodular group (n = 3/8), and in the CD34+ scarce group (n = 2/11): they were located in dentate gyrus, cornu ammonis and subiculum and were associated with P53 positive cells (n = 4, Figure 1) but no ganglioglioma or DNET was found. These HS (n = 1 type 1, n = 3 type 2, n = 1 type 3) are further called “BRAF V600E+ HS.” By contrast, in three other CD34+ nodular samples, the neuronal markers INA and ELAVL2 were intensely positive in abnormal, ectopic and often vacuolated neurons in the white matter of the subiculum, alveus and fimbria whereas NeuN was negative in these cells (Figure 2; Supporting Information Figure 4). These samples (n = 2 type 3, n = 1 noHS) were further considered as the hippocampal localization of MVNT. The BRAF V600E+ HS and hippocampal MVNT were mutually exclusive. They were not observed in CD34– samples (n_{BRAF V600E} = 11, n_INA = 13).

The MRI findings were available for two cases of BRAF V600E+ HS and showed a typical hippocampal sclerosis (Figure 3). They were available for two cases of hippocampal MVNT. They showed T2 and FLAIR hyperintensities involving hippocampus and extrahippocampal structures and a pseudocyst and were compatible with a DNET or MVNT (Figure 3).

Figure 3.

Radiological features of BRAF V600E+ hippocampal sclerosis and hippocampal MVNT. A‐B. BRAF V600E+ hippocampal sclerosis. Coronal T2 sequence showing discrete atrophy, loss of digitation and T2 hyperintensity of the right hippocampus (vertical arrows). C–E. Hippocampal MVNT. Coronal T2 (C), FLAIR (D) and T1 (E) sequences show discrete atrophy of left internal temporal structures (vertical arrows), associated with extensive T2 and FLAIR hyperintensities involving the hippocampus, parahippocampal gyrus, fusiform gyrus and part of the inferior temporal gyrus (oblique arrows). A pseudocyst (arrowheads) is visible within the fusiform gyrus, hyperintense on T2 and hypointense on T1 and FLAIR images.

Presurgical clinical correlates of ILAE types

We analyzed relations between the different patterns of HS and the presurgical clinical data (Table 2 and Supporting Information Table 2).

In the type 1, most patients presented an IPI (77%) with a majority of FS (56%). In the type 1 with FS as IPI, CA2 neuronal loss was significantly positively correlated with the age at FS (n = 69, P = 0.0035, rho = 0.36, Supporting Information Figure 1). In the type 2, patients presented the highest prevalence of IPI (89%) with a majority of FS (n = 32/47 68%). In the type 3, we noted no FS, a low prevalence of IPI (43%) and a trend for an older age of onset and at surgery (18.6 ±11.2 and 40.9 ±16.1, respectively). In the pattern noHS, there were fewer IPI (33% P = 0.001), fewer FS (17% P = 6 × 10⁻⁴) and less abnormal MRI aspect (29% P = 7 × 10⁻⁸) compared to types 1 or 2. Stereo‐electroencephalographic studies (sEEG) had been performed more frequently in noHS (71%) than type 1 (8%) and type 2 (2% P = 6 × 10⁻⁵). There was a trend for an older age at onset (17.3 ± 10.0) than in type 1 and 2. The duration of epilepsy was shorter in noHS than in types 1 and 2 but not at a level of significance (duration_{type 1} = 26.0 y; duration_{type 2} = 26.6 y; duration_{type noHS} = 24.5 y; P = 0.64).

We analyzed the relationships between different patterns of GCL alterations and the presurgical clinical data (Table 3 and Supporting Information Table 3). GCL broadening (ie, dispersion or duplication) was more frequent in female patients (male 71%, female 82%, P = 0.044). GCL duplication was significantly associated with a history of status epilepticus (normal GCL 2%, dispersion 10%, duplication 20%, P = 0.04) and with the occurrence of generalized seizures (normal GCL 71%, dispersion 68%, duplication 96%, P = 0.02). Normal GCL pattern was associated with a normal MRI (normal GCL 10%, broadening 2%, P = 0.04).

We analyzed the relations between different CD34+ cells patterns and the presurgical clinical data (Table 4 and Supporting Information Table 4). The “CD34+ nodular” group presented more frequent dysmnesic auras (CD34– 34%; CD34+ scarce 43%; CD34+ nodular 88%; P = 0.006). In the “CD34+ scarce” group, sEEG had been more often performed (CD34– 7%; CD34+ scarce 20%; CD34+ nodular 13%; P = 0.037) and the MRI was more often normal (CD34– 2%; CD34+ scarce 13%; CD34+ nodular 0%; P = 0.022).

Prognostic value of histopathological lesions

In the type 1, CA4 hypertrophic neurons were predictive of a worse prognosis for Engel I (P = 0.045) and Engel Ia (P = 0.0046) (Figure 4). In the types 1 and 2, GCL dispersion and duplication had no prognostic value. The presence of CD34 stellate cells was associated with a trend for a better prognosis for the Engel Ia class (P = 0.06).

Figure 4.

Prognostic value of histopathological lesions of hippocampal sclerosis in the postsurgical outcome. Kaplan‐Meier plots are shown for Engel I and Engel Ia classes of postsurgical outcome: the ratio of patients fulfilling the criteria of Engel I (A,C,E) and Engel Ia (B,D,F) classes are depicted during the postsurgical follow‐up (time in years). A,B. In the type 1, presence of CA4 hypertrophic neurons (blue curve, n = 89) had a significant worse prognosis than absence of CA4 hypertrophic neurons (purple curve, n = 83). C‐D. in the types 1 and 2, absence of broadening of the dentate gyrus granule cell layer (purple curve, n = 44), their dispersion (blue curve, n = 137) and their duplication (black curve, n = 24) had no prognostic value. E‐F. Presence of CD34+ stellate cells (blue curve, n = 43) has no prognostic value for the Engel I class but showed a trend for a better prognosis than the absence of CD34+ stellate cells (purple curve, n = 170) for the Engel Ia class.

Discussion

Here, we report correlations between clinical and pathological data in samples from a large monocentric series of MTLE patients with HS patterns classified according to ILAE guidelines. These data confirmed some previously reported correlations and identified novel links between distinct forms of HS and the occurrence of histological lesions, CD34+ stellate cells, seizure symptoms, epilepsy severity and postsurgical outcome.

This series of pharmacoresistant MTLE patients had a similar mean age at epilepsy onset, a longer duration of epilepsy and an older age at surgery than for some previous reports. Supporting Information Table 5 lists associations between clinical and pathological data confirmed in this study 3, 4, 11, 27, 37, 40. The good interobserver reproducibility of ILAE classification in our series and the confirmation of main previous findings from other teams strengthen the interest of this methodology to compare data between teams of epilepsy surgery. We included cases with analyzable CA1 and CA4 because we can classify them into neuronal loss predominant in CA1 (type 2), in CA4 (type 3) or in both (type 1). CA2 and CA3 were not analyzable in some cases, which could limit our conclusions because we cannot exclude that some of these cases have neuronal loss predominant in CA2 or CA3. However, ILAE classification does not define HS by predominant neuronal loss in CA2 or CA3 and we did not observe it in our large series. Although we evaluated neuronal loss only by a semi‐quantitative visual evaluation, neuronal loss in CA2, CA3 and GCL, was more extreme for type 1 than type 2 and more severe for type 2 than noHS as reported 4. GCL dispersion or duplication was more frequent for type 1 and associated with a higher neuronal loss in CA4 as previously reported 3, 8, 24, 40, 42. We observed a trend for more IPI (including FS) in type 2, as did the study of Thom et al 40. We observed a trend for GCL dispersion in older patients, as previously reported 3. Samples with noHS were associated with shorter epilepsy duration, and less IPI (including FS). In a well‐defined group of type 1 with FS, we found a positive correlation between age at IPI and CA2 neuronal loss: our data suggests that the survival of CA2 neurons is enhanced when an IPI occurs earlier.

Hypertrophic CA4 neurons were associated with type 1 and a higher neuronal loss in CA4 in our series as previously reported 7, 30, 38, 39. They are thought as an alteration induced by epilepsy rather than dysplastic neurons 7, 30. mTOR pathway activation in the dysmorphic neurons of tubers or FCD causes hypertrophy and epileptogenesis 19, 26. This activation can also be induced by epilepsy 33, 34 and was described in CA4 hypertrophic neurons of HS 21. We identified CA4 hypertrophic neurons as a new pejorative prognostic factor for postsurgical outcome in type 1. These abnormal neurons could be a biomarker of a more severe injury epilepsy and of mTOR activation. Further studies are required to decipher the pathophysiological role of CA4 hypertrophic neurons.

GCL broadening has been linked to seizures and suggested to reflect either newly generated neurons or abnormal somatic migration of mature granule cells 3, 38. We detected GCL broadening more frequently in female than in male patients. Gender‐linked effects of estrogens on hippocampal neurogenesis have been described 1, 12, 35, 36. Estrogen modulates hippocampal neuronal networks in part via Brain Derived Neurotrophic Factor, a neurotrophin that increases granule cell neurogenesis, ectopic granule cells and transmission mediated by mossy fibers 14, 31, 44. Our data show that GCL duplication is associated with generalized seizures and status epilepticus, raising a possible link between this histological lesion and more severe epilepsies. Alternatively, generalized seizures may favor a dual layer of granule cell bodies.

We examined CD34+ stellate cells. Their Nestin+, SOX2+/–, OLIG2+/–, CD31–, Vimentin‐ immunophenotype favors an immature neural cell type. We observed BRAF V600E positive neurons in three “CD34+ nodular” and two “CD34+ scarce” HS. As there was no radiological or histopathological evidence of tumors in these cases and as the predominant lesions were neuronal loss and gliosis, we named these cases “BRAF V600E+ HS.” In vivo electroporation studies in the murine embryonic cortex showed that mild activation of Extracellular‐signal‐regulated kinase/ERK signaling driven by BRAF V600E or Ras led to the production of abnormally specified neurons whereas strong activation of ERK signaling drives gliomagenesis 20. These observations open the intriguing hypothesis that the presence of a BRAF V600E mutation in a subset of hippocampal neurons could drive epilepsy without tumorigenesis. Among CD34+ cases, we described three cases of hippocampal MVNT. The tumoral or dysplastic etiology of this lesion is debated 16, 41. We confirmed the absence of BRAF V600E mutation and the CD34 and INA immunopositivity of MVNT 16, 28, 41. We identified an association between the “CD34+ nodular” group and dysmnesic auras suggesting that the CD34+ cells are associated with a pre‐ictal activity in the hippocampus. We observed that samples with “CD34+ nodular” cells were linked to type 2, type 3 and to noHS in accordance with the previously reported association of diffuse DNET (which is one possible interpretation of CD34+ cells) with atypical HS 43. We also extended this association of atypical HS with “CD34+ scarce” cells. CD34+ stellate cells are thus associated with the pathogenesis of atypical HS. We found a trend for a better postsurgical outcome associated with CD34+ stellate cells suggesting that they correspond to a focal epileptogenic zone curable by epilepsy surgery. Together, these results reinforce the interest in the CD34 marker for the nosology of HS. They also urge to investigate the role of CD34+ cells in the pathogenesis of MTLE associated to atypical HS.

In conclusion, our data confirms some previously reported clinicopathological correlations of MTLE‐HS and opens new perspectives on the heterogeneity of this syndrome—perspectives that may improve the management of those patients (Figure 5). Type 1 is the most frequent, corresponds to the most severe neuronal loss and is associated with frequent IPI and FS, hypertrophic CA4 neurons and GCL broadening. Type 2 is strongly linked to IPI including FS, and to CD34+ stellate cells. A subgroup of type 1 and a subgroup of type 2 are characterized by GCL duplication, more frequent history of status epilepticus and a higher ratio of patients presenting generalized seizure. Type 3 and the pattern no HS are rare, and associated with rarer IPI, an older age at epilepsy onset and at surgery, and CD34+ stellate cells. MTLE with noHS is associated with more normal MRI and sEEG.

Figure 5.

Clinicopathological correlates of ILAE types of HS and of their subgroups. Distinctive clinicopathological features of the four ILAE types of HS are summarized. CA4 hypertrophic neurons are more frequent in type 1 and are associated with worse postsurgical prognosis. GCL duplication identified a subgroup of type 1 and a subgroup of type 2. It is associated with a higher prevalence of presurgical status epilepticus and/or generalized seizures. CD34+ stellate cells were more frequently observed in the rarest ILAE types: 2, 3, noHS. The “CD34+ scarce” pattern was associated with higher prevalence of normal MRI and stereo‐encephalographic studies. The “CD34+ nodular” pattern was associated with dysmnesic auras.

Supporting information

Additional supporting information may be found online in the Supporting Information section at the end of the article.

Figure 1. Clinical features of mesial temporal lobe epilepsy – hippocampal sclerosis. a. A flowchart shows the steps for the inclusion of samples in the study according to the indicated criteria. b. Distribution of the age of epilepsy onset in the patient population. c. Distribution of the duration of epilepsy. d. Epilepsy duration plotted against the date of surgery. e. Neuronal loss in the CA2 region of type 1 was positively correlated with age at the first FS (n = 69, P = 0.0035, rho = 0.36).

Figure 2. Neuronal loss evaluation in the ILAE types of hippocampal sclerosis. a‐h. NeuN immunolabeling in CA1 (a‐d) and CA4 (e‐h) in hippocampal sclerosis of type 1 (a,e), type 2 (b,f), type 3 (c,g) and in the pattern noHS, gliosis only (d,h). i. Percentage of neuronal loss in CA1, CA2, CA3, CA4 is shown as box and whisker plots for type 1 (blue), type 2 (red), type 3 (green) and type noHS, gliosis only (pink). By definition there is no neuronal loss in the pattern noHS, gliosis only. Although the ILAE classification is based on the predominance of neuronal loss in either CA1 and/or CA4 fields, we noticed that some neuronal loss was also present in the GCL and in the CA2 and CA3 regions. Neuronal loss in CA2 was significantly higher in type 1 than in type 2 (P < 10⁻⁴). It was higher in type 2 than in type 3 (P = 7 × 10⁻⁴) and than in noHS (P = 10⁻⁴). Neuronal loss in CA3 field was significantly higher in type 1 than in type 2 (P < 10⁻⁴) and higher in type 2 than in noHS samples (P = 2 × 10⁻⁴). Loss of dentate granule cells was significantly higher in type 1 than in type 2 (P = 0.002) and higher in type 2 than in noHS (P = 0.007). Scale bars: a‐h 200 μm.

Figure 3. Histopathological aspects of hypertrophic CA4 neurons and of broadening of the granule cell layer of the dentate gyrus. a‐c. Hematoxylin eosin, 400×. Normal CA4 neurons (a), hypertrophic CA4 neurons (b,c). d. The longer axis (μm) of the nucleus of CA4 pyramidal neurons was significantly increased in hypertrophic CA4 neurons of HS (20.8 μm ± 3.2 red) compared to CA4 neurons of a control hippocampus (14.5 μm ± 2.0 blue, P < 0.0001). e. The longer axis (μm) of the cell soma of CA4 pyramidal neurons was significantly increased in hypertrophic CA4 neurons of HS (52.6 μm ± 14 red) compared to CA4 neurons of a control hippocampus (27.8 μm ± 7.9 blue, P < 0.0001). f. The ratio between the longer axis of the nucleus and the longer axis of the cell soma of CA4 pyramidal neurons was significantly decreased in hypertrophic CA4 neurons of HS (0.42 ± 0.1 red) compared to CA4 neurons of a control hippocampus (0.55 μm ± 0.1 blue, P < 0.0001). g‐i. NeuN immunolabeling, 200×. Normal granule cell layer/GCL of the dentate gyrus (g), dispersion of the GCL (h), duplication of the GCL (i). Scale bars: a‐c 100μm, d‐e 100μm.

Figure 4. Expression of OLIG2 by CD34+ stellate cells. a‐c. Co‐expression of OLIG2 (pink) and CD34 (brown) by CD34+ stellate cells in a CD34+ nodular case. Scale bars: a‐c 50μm.

Figure 5. Histopathological aspect of hippocampal MVNT. The parahippocampic white matter (a,c) and the alveus (b,d) harbored nodules (black arrowheads in a,b) containing neuropile, vacuolated neurons and vacuoles (white arrowheads in c,d). Scale bars: a‐b: 1mm, c‐d: 100μm.

Table 1. Clinicopathological features of the series.

Table 2. Clinicopathological features associated with ILAE types.

Table 3. Clinicopathological features associated with granule cell layer changes.

Table 4. Clinicopathological features of hippocampal sclerosis according to CD34 status.

Table 5. Comparison of the positive results of different studies with ILAE type of HS.