Developmental brain abnormalities in tuberous sclerosis complex: A comparative tissue analysis of cortical tubers and perituberal cortex

Authors

  • Véronique Ruppe,

    1. Department of Neurology, School of Medicine, New York University, New York, New York, U.S.A
    Search for more papers by this author
  • Pelin Dilsiz,

    1. Department of Neurology, School of Medicine, New York University, New York, New York, U.S.A
    Search for more papers by this author
  • Carol Shoshkes Reiss,

    1. Department of Biology and Neural Science, New York University, New York, New York, U.S.A
    Search for more papers by this author
  • Chad Carlson,

    1. Department of Neurology, School of Medicine, New York University, New York, New York, U.S.A
    Search for more papers by this author
  • Orrin Devinsky,

    1. Department of Neurology, School of Medicine, New York University, New York, New York, U.S.A
    2. Department of Neurosurgery, School of Medicine, New York University, New York, New York, U.S.A
    3. Department of Psychiatry, School of Medicine, New York University, New York, New York, U.S.A
    Search for more papers by this author
  • David Zagzag,

    1. Department of Neurosurgery, School of Medicine, New York University, New York, New York, U.S.A
    2. Department of Pathology, School of Medicine, New York University, New York, New York, U.S.A
    Search for more papers by this author
  • Howard L. Weiner,

    1. Department of Neurosurgery, School of Medicine, New York University, New York, New York, U.S.A
    Search for more papers by this author
  • Delia M. Talos

    Corresponding author
    1. Department of Neurology, School of Medicine, New York University, New York, New York, U.S.A
    Current affiliation:
    1. Department of Neurology, Perelman School of Medicine, University of Pennsylvania, Philadelphia, Pennsylvania, U.S.A
    • Address correspondence to Delia M. Talos, Department of Neurology, Perelman School of Medicine, University of Pennsylvania, 263 Clinical Research Building, 415 Curie Boulevard, Philadelphia, PA 19104, U.S.A. E-mail: talosd@mail.med.upenn.edu

    Search for more papers by this author

Summary

Objective

Genetic loss of Tsc1/Tsc2 function in tuberous sclerosis complex (TSC) results in altered mammalian target of rapamycin (mTOR) signaling and abnormal brain development. Although earlier studies have focused on characterization of cortical tubers, in this study we sought to examine the unique cellular and molecular features of the perituberal cortex in order to better understand its contribution to epileptogenesis, cognitive dysfunction, and autism.

Methods

Standard histologic and immunohistochemical labeling was used to assess structural abnormalities and cell-specific pattern of mTORC1 activation in surgically resected cortical tubers and perituberal cortex. Western blotting was performed to quantify the expression of the mTORC1 and mTORC2 biomarkers phospho-S6 (Ser235/236), phospho-S6 (Ser240/244), and phospho-Akt (Ser473), in addition to evaluating the differential expression levels of several neuronal and glial-specific proteins in tubers and peritubers, as compared to non-TSC epilepsy specimens.

Results

Tubers demonstrated mild to severe disruption of cortical lamination, the presence of pS6-positive dysplastic neurons and giant cells, an overall increase in mTORC1 and a decrease in mTORC2 activity, increased axonal connectivity and growth, and hypomyelination. Perituberal cortex presented similar histologic, immunohistochemical, and molecular features; however, they were overall milder. Axonal growth was specific for TSC and was negatively correlated with deficient myelination.

Significance

Our results show an extension of cellular dysplasia and dysregulated mTOR signaling in the perituberal tissue, and demonstrate for the first time aberrant connectivity in human TSC brain. This study provides new insights into the pathophysiology of neurologic dysfunction associated with TSC and supports the intrinsic epileptogenicity of normal-appearing perituberal cortex.

A PowerPoint slide summarizing this article is available for download in the Supporting Information section here.

image

Véronique Ruppe, PhD is an Associate Research Scientist at the Comprehensive Epilepsy Center, Department of Neurology, New York University Langone Medical Center, New York, NY.

Tuberous sclerosis complex (TSC) results from mutations in the TSC1 (9q34) or TSC2 (16p13.3) genes that code for Tsc1 and Tsc2 proteins, respectively.[1] TSC affects multiple organ systems, with prominent involvement of the brain, skin, kidneys, and lungs. Brain lesions include cortical tubers, subependymal nodules, and subependymal giant cell astrocytomas. Cortical tubers represent hamartomatous brain malformations characterized by disorganized tissue architecture, aberrant cellular growth and morphology, altered lineage differentiation, and abnormal cellular maturation.[2-4]

Neurologic impairment causes the greatest morbidity in TSC patients. Seizures affect up to 90% of TSC patients,[5] often with early onset and medical intractability. Neuropsychiatric disorders associated with TSC include mental retardation (50–60%), autism (40–65%), failure to acquire language, impaired social and emotional skills, aggressive behavior, attention deficits, anxiety and affective disorders, sleep disorders, and motor control disorders.[5-7]

Cortical tubers and adjacent perituberal tissue are postulated to impair brain function and cause epilepsy. Although epilepsy surgery has originally focused on resection of cortical tubers, there is now increasing evidence in support of the epileptogenicity of the perituberal tissue.[8-12] Perituberal areas are often part of the ictal-onset zone[9, 10, 12] and must be removed in order to achieve seizure freedom.[8]

Other neuropsychiatric symptoms, including autism and cognitive impairments, do not always correlate with the “tuber burden”[13] and may be associated with other, more subtle structural and functional abnormalities possibly involving distant brain regions. Indeed, loss of white matter integrity[14, 15] and microdysgenesis of the gray matter[16, 17] outside of cortical tubers has been recently documented. Furthermore, nontuber pathology seems to be highly correlated with neurocognitive deficits and autistic symptomatology.[18]

The Tsc1 and Tsc2 proteins form a complex that modulates the activity of the mammalian target of rapamycin (mTOR), a serine/threonine kinase that associates with a number of other regulatory proteins to form two distinct complexes: mTORC1 and mTORC2.[19] The major downstream effects of mTORC1 include activation of the translational apparatus through inactivation of the eukaryotic initiation factor 4E (eIF4E)-binding protein 1 (4E-BP1) and activation the ribosomal S6 protein kinase (p70S6K) and its downstream target ribosomal S6 protein.[20] The mTORC1 signaling is required for proper neuronal development and function, whereas dysregulation of this pathway impairs the development of neural circuits at multiple levels. In animal models, upregulated mTORC1 signaling due to decreased Tsc1/Tsc2 activity leads to increased neuronal size, altered dendritic arborizations and spine formation,[21, 22] enhanced glutamatergic neurotransmission,[21, 23] altered axonal pathfinding and growth,[24, 25] as well as impaired synaptic plasticity.[26, 27] Similarly, dysplastic neurons in human cortical tubers have larger soma size, hypertrophic dendrites, present electrophysiologic evidence for increased synaptic densities, and express altered levels of excitatory and inhibitory neurotransmitter receptors.[4, 23, 28-31]

The mTORC2 signaling controls the activity of several kinases including Akt, the protein kinase C (PKC), and the serum- and glucocorticoid-induced protein kinase 1 (SGK1), known for their roles in cell survival and growth, regulation of the actin cytoskeleton, and ion transport.[19] The role of mTORC2 signaling in the central nervous system (CNS) has been less intensively studied; however, several brain models of TSC have demonstrated decreased mTORC2 activity associated with loss of Tsc1/Tsc2 function.[22, 32, 33]

In this study, we took a systematic approach to further elucidate nontuber pathology, focusing on the epileptogenic perituberal cortex. We hypothesized that developmental cellular dysplasia, dysregulated mTOR signaling, and altered neuronal connectivity extend beyond the margin of the magnetic resonance imaging (MRI)–defined cortical tubers. We used standard histochemistry, immunohistochemistry, and quantitative Western blot analysis for different cell-specific markers to identify and characterize the cellular composition of tuberal and perituberal tissue, in conjunction with downstream effectors of mTORC1 and mTORC2 (i.e., phospho-S6 [Ser235/236], phospho-S6 [Ser240/244], and phospho-Akt [Ser473]). In order to distinguish the potential effects of seizure activity on nontuber pathology, we also included in our analyses samples from non-TSC epilepsy patients for additional comparison.

Materials and Methods

Human subjects

A total of 30 patients with chronic drug-resistant epilepsy who underwent resective surgery at New York University Langone Medical Center (NYULMC), New York, NY, from February 2007 to March 2012 were prospectively enrolled in our study (Table S1). The surgically resected neocortical tissue included specimens from subjects diagnosed with TSC (n = 23; 13 female and 10 male), as well as from patients with epilepsy not associated with TSC (n = 7; 4 female and 3 male). All patients underwent either a two-stage or a three-stage surgery with invasive intracranial monitoring using grids and strips.[8] MRI was performed in all cases prior to surgery. Tubers were defined as areas of hyperintense signal on T2-weighted images and hypointense signal on T1-weighted MRI. Intraoperatively, cortical tubers presented as firm, pale, glassy, and relatively avascular brain lesions and were often calcified. In 8 of 23 TSC patients, more than one cortical tuber was removed during the surgery, such that a total of 39 cortical tubers were available for evaluation. In 16 of 23 TSC patients, a significant amount of perituberal cortex (which did not correspond to a tuber on the MRI scan) was removed, allowing for an in-depth analysis of the epileptogenic areas outside of tubers. The perituberal cortex specimens were resected within approximately 1 cm from the tuber margin and were normal in gross appearance and consistency. Specimens collected from non-TSC epilepsy subjects presented no gross MRI or macroscopic abnormalities and exhibited neuropathologic features consistent with mild focal cortical dysplasia (FCD). Further classification of FCD specimens was carried out according to the system proposed by Palmini.[34] Four cases met the definition criteria for FCD type Ia and three cases were diagnosed as FCD type Ib.

The control group consisted of autopsy cases (n = 12; 3 female and 9 male) with no known seizure history (Table S2). All subjects died of nonneurologic causes and presented no evidence of brain malformations or any other diagnostic lesions of the CNS. The postmortem interval (PMI) was <24 h in most cases (mean 14.9 h). Control specimens were obtained from the Brain and Tissue Bank for Developmental Disorders at the University of Maryland, Baltimore, Maryland. In 5 of 12 cases we obtained multiple specimens/case, originating from different brain regions (total number of specimens n = 21). This study was approved by the NYULMC Institutional Review Board. Informed consent was obtained for use of resected brain tissue and for access of medical records for research purposes.

Standard histopathologic evaluation

Slides were stained with hematoxylin and eosin (H&E) and Luxol fast blue-H&E (LH&E). Additional slides were subjected to immunohistochemical staining with neuronal nuclear antigen (NeuN), neurofilament protein, and vimentin, using standard protocols[35] (see Supporting Information).

Double label immunohistochemistry

Tissue sections were stained with the following antibodies: growth associated protein (GAP-43), myelin basic protein (MBP), NeuN, nonphosphorylated neurofilament (SMI311), phosphorylated neurofilament (SMI312), phospho-S6 (Ser235/236) (pS6), synapsin I, synaptophysin, vimentin, and the vesicular glutamate transporter 1 (VGluT1). All staining was performed as described previously[29] (see Supporting Information). For pS6 quantitative analysis, at least 10 representative fields (each covering 0.325 mm2) were imaged per stained section, with a total of 2–5 tissue sections per specimen (i.e., 20–50 images per specimen, or 6.5–16.25 mm2 per specimen). Double positive cells (either pS6/SMI311 or pS6/vimentin) were counted for every image, and an average cell count per mm2 was calculated for each specimen. The SMI312 semi-quantitative analysis was carried out on at least six contiguous fields per section and at least two sections per specimen (i.e., at least 12 fields per specimen). The distribution of SMI312 positive fibers was evaluated by two independent investigators and the final decision was made by consensus. The cortex and white matter were graded separately using a 0 to 4 scoring system, taking into account both severity and distribution. The scores were attributed as follows: 0, normal; 1, mild increase; 2, moderate increase; 3, severe increase; 4, severe and widespread increase.[36] Age-matched nonepileptic control specimens were used as reference and assigned a score of 0.

Western blotting

Frozen brain samples were processed for whole cell protein extracts, as previously published[29, 37] (see Supporting Information). Blots were probed for actin, Akt, phospho-Akt (Ser 473), GAP-43, MBP, NeuN, oligodendrocyte transcription factor Olig2, SMI312, pS6 (Ser235/236), pS6 (Ser240/244), and S6. Raw values were normalized to actin to control for differences in protein loading. Ratios of pAkt (Ser 473)/Akt, pS6 (Ser235/236)/S6, and pS6 (Ser240/244)/S6 were calculated in a similar fashion. Normalized values were then expressed as a percentage of the age-matched and brain region–matched control samples run on the same blot (100%).

Statistical methods

Statistical analysis was performed using the GraphPad Prism v.6.0b software (GraphPad Software Inc., San Diego, CA, U.S.A.). Differences between two groups were compared using two tailed Student's t-tests. One-way analysis of variance (ANOVA) followed by Tukey's honest significant difference (HSD) post hoc analysis was performed to compare multiple groups. Linear regressions were performed to find significant correlations between two independent variables. Statistical significance was defined as p < 0.05.

Results

Cellular dysplasia extends beyond the margin of the MRI-identified cortical tubers

We hypothesized that because perituberal tissue is often resected to achieve seizure control, it may also display architectural abnormalities and contain giant cells and dysplastic neurons. Standard histologic examination of TSC (Fig. 1A,B,D,E) and control cases (Fig. 1C,F) was carried out on H&E stained sections and demonstrated mild to severe disruption of cortical lamination and the presence of dysplastic neurons and giant cells in all cortical tubers examined (n = 39; Fig. 1A,D). The perituberal cortex samples presented similar histologic features; however, they were overall milder (n = 16; Fig. 1B,E). In three cases the perituberal cortex presented abundant dysplastic neurons, but no evidence for giant cells (data not shown). Single labeling for NeuN further highlighted the laminar defects in both tubers and perituberal samples (Fig. 1G–I). The typical radial microcolumns, indicative of persistent fetal cortical architecture in TSC,[38] were frequently observed in cortical tubers (Fig. 1G), but less often in the perituberal regions (Fig. 1H).

Figure 1.

Histopathologic features of cortical tubers and perituberal cortex. (AC) Hematoxylin and eosin (H&E) staining showing a lower magnification view of cortical dysplasia in tubers (A) and peritubers (B), as compared to control brain tissue (C). (DF) H&E staining showing a higher magnification view of clustered dysplastic neurons and giant cells in tubers (D) and peritubers (E), but not in the control cortex (F). (GI) NeuN immunostaining highlighting the laminar defects in tubers (G) and perituberal cortex (H), in contrast to preserved cortical architecture in control cortex (I). Note: typical cortical tuber radial microcolumns in G. Scale bars represent 50 μm in AC, 20 μm in DF, and 100 μm in GI.

Immunohistochemistry for pS6 (Ser 235/236), a marker of mTORC1 activation hereafter referred to as pS6, demonstrated that pS6-positive dysplastic cells were present and similarly distributed in both tubers (n = 21) and perituberal tissue (n = 7) (Fig. 2A,B). The pS6-expressing cells were immunopositive for nonphosphorylated neurofilament SMI311 (Fig. 2C,E), the intermediate filament protein vimentin (Fig. 2D–F), a marker of immature cells throughout the brain, or both. The average number of pS6/SMI311 double positive cells (mean ± SEM; Fig. 2G) was higher in tubers (6.78 ± 1.17/mm2 of stained tissue) relative to perituberal tissue (2.97 ± 1.52); however, the difference was not statistically significant (p > 0.05). Similarly, the density of cells coexpressing pS6 and vimentin (Fig. 2H) was not statistically different between the two groups (8.74 ± 2.80/mm2 of tissue in tubers and 5.87 ± 2.44/mm2 in peri-tubers; p > 0.05). Control specimens showed no pS6-positive dysplastic cells, but rather low pS6 immunoreactivity in cortical pyramidal neurons and glial cells, and moderate pS6 levels in endothelial cells of blood vessels (Fig. S1).

Figure 2.

Cellular dysplasia and altered mTOR signaling in tubers and perituberal cortex. (AF) Representative tuber (A,C,E) and perituber (B,D,F) coronal sections double labeled with pS6 (green) and cellular markers SMI311 and vimentin (red) showing widespread cellular abnormalities in both tubers and peritubers. The nuclear 4′,6-diamidino 2-phenylindole (DAPI) staining appears in blue. Scale bars represent 50 μm. (GH) Cell counts of pS6/SMI311 (G) and pS6/vimentin (H) double positive cells in cortical tubers (red dots) and peritubers (blue dots) demonstrating no significant differences between the two groups. (IK) Western blot quantification of pS6 (Ser235/236)/S6 (I), pS6 (Ser240/244)/S6 (J) and pAkt (Ser473)/Akt (K) in tubers and peritubers compared to age-matched, region-matched controls. Values are expressed as mean ± SEM. (L) Representative Western blot bands. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Hyperactivation of mTORC1 is associated with decreased mTORC2 signaling in both tubers and perituberal cortex

To further evaluate the degree of altered mTOR signaling in the perituberal cortex, we performed quantitative western blot analysis for individual mTORC1 and mTORC2 downstream targets in tubers, peritubers, and control tissue. Levels of mTORC1 activation were measured by assessing the phosphorylation of S6 at two sites, Ser 235/236 and Ser 240/244, relative to total amounts of S6 protein (Fig. 2I,J). The pS6 (Ser 235/236)/S6 ratios were significantly increased in both tubers (275% of control, n = 4; p < 0.001) and perituberal cortex (191% of control, n = 4; p < 0.01), with significantly higher levels in tubers, compared to peritubers (p < 0.05). The pS6 (Ser 240/244)/S6 ratios demonstrated a more modest increase (168% of control for tubers, n = 4; p < 0.05 and 144% of control for peri-tubers, n = 3; p > 0.05), and no significant difference between tubers and peritubers.

To determine whether mTORC2 activity was decreased in human TSC brain tissue, as it is in mouse models of TSC,[22, 32] we quantified mTORC2 activation by measuring the levels of pAkt (Ser 473) relative to total levels of Akt (Fig. 2K). The pAkt (Ser 473)/Akt ratios were similarly decreased in both tubers (47% of control, n = 4; p < 0.0001) and peritubers (57% of control, n = 4; p < 0.0001). These results, together with the histologic and immunohistochemical abnormalities described earlier emphasize the significant neuropathology of the epileptogenic perituberal areas.

Increased axonal connectivity and reduced myelination in cortical tubers and perituberal cortex

Recent studies in TSC patients using diffusion tensor imaging (DTI) have revealed widespread abnormalities within the otherwise normal-appearing white matter,[14, 15] whereas several mouse models provide additional evidence of abnormal circuitry in the TSC brain.[24, 39] However, such cellular changes had not been validated in human specimens.

In order to characterize axons in human TSC brain, we first examined the expression of the phosphorylated neurofilament SMI312, a pan-axonal marker[40] (Fig. 3A–F). In the cortex (Fig. 3A–C), SMI312 demonstrated a strikingly different distribution pattern in tubers (n = 17) and peritubers (n = 8) when compared to nonepileptic controls (n = 4). The TSC neurons possessed longer and more elaborated axonal branches, particularly near the soma. Numerous axon segments also appeared thicker and their trajectories were completely disorganized. Larger number of collateral axonal branches and hypertrophic axon segments were also observed within the superficial white matter portion of tubers and peritubers (Fig. 3D–F). Semi-quantitative analysis of SMI312-positive profiles performed on a subset of TSC cases (n = 6) demonstrated overall higher average scores relative to age-matched nonepilepsy controls in both cortex (p < 0.001) and white matter (p < 0.01); however, there were no significant differences between cortical tubers and perituberal regions: 2.66 ± 0.33 in tubers versus 2.33 ± 0.33 in peritubers for cortex (p > 0.05; Fig. 3G) and 1.66 ± 0.21 in tubers versus 1.5 ± 0.34 in peritubers for the superficial white matter (p > 0.05; Fig. 3H). Most of the aberrant axonal branches were immunopositive for the excitatory neurotransmitter marker VGluT1 (Fig. 3I), which is consistent with the excitatory phenotype of TSC dysplastic neurons.[4, 28] Most axons had button-like terminals that were immunoreactive for the presynaptic markers synapsin I (Fig. 3J), as well as synaptophysin (Fig. 3K). These data suggest that excitatory synapse formation is enhanced in both tuberal and perituberal areas.

Figure 3.

Altered axonal connectivity in tubers and peritubers. (AF) Immunolabeling with axonal marker SMI312 (red) in tubers (A,D), peritubers (B,E), and control samples (C,F) showing longer, disorganized axonal branches and hypertrophic axon segments in both TSC cortex (AC) and white matter (DF). (GH) Semi-quantitative analysis of SMI312-positive fibers in TSC cortex (G) and white matter (H) demonstrating higher then control (score 0) mean values, but no significant different average scores between tubers (red dots) and peritubers (blue dots). (I) Double immunostaining for SMI312 (red) and the vesicular glutamate transporter VGLuT1 (green) demonstrating substantial co-localization. (JK) Double labeling for SMI312 (red) and either synapsin I or synaptophysin (green) showing button-like axonal terminals coexpressing both markers. DAPI staining (AF and IK) is shown in blue. Scale bars in AF and IK represent 50 μm. **p < 0.01, ***p < 0.001.

In addition, double label immunohistochemistry for myelin basic protein (MBP) and SMI312 demonstrated strongly reduced MBP expression in both tubers and peritubers compared to control tissue (Fig. 4A–C). In TSC, the MBP-positive fibers appeared scattered throughout the cortex, and a significant proportion of axons were unmyelinated (Fig. 4D–I). Although these defects in myelination were much more pronounced in the cortex, we found that MBP expression was also markedly reduced in tuberal and perituberal white matter (Fig. 4J–L). This pattern was consistent with the relative lack of mature myelin in tubers and peritubers observed by LH&E histochemical staining (data not shown).

Figure 4.

Reduced myelination in cortical tubers and perituberal areas. (AI) Double immunolabeling for myelin basic protein (MBP; green) and SMI312 (red) showing less myelinated axons in cortical tubers (A,D,G) and perituberal cortex (B,E,H), when compared to control cortex (C,F,I). (JL) MBP (green) and SMI312 (red) double labeling demonstrating severe hypomyelination of the superficial white matter adjacent to cortical tubers (J) and perituberal cortex (K), relative to age-matched controls (L). Scale bar represents 50 μm.

A comparative analysis of TSC and non-TSC epilepsy pathology

Because seizure activity, especially when associated with neuronal loss, can induce axonal growth and neuronal network reorganization,[41, 42] it is unclear whether the increased connectivity in tubers and peritubers is due solely to mTOR dysregulation. To assess the potential contribution of seizure activity to alter intracortical connectivity, we included samples from patients with chronic epilepsy not associated with TSC (Table S1) in all our subsequent analyses. In these patients, the mean age at surgery was significantly higher (10.7 years), compared to TSC (4.7 years; p < 0.05), and there was also a significantly older age of seizure onset (4 years), relative to the TSC group (0.5 years; p < 0.05). In addition, the FCD patients experienced a relatively longer seizure history, defined as the time interval between the onset of seizures and surgery (6.8 years for FCD patients and 4.2 years for TSC), but on average presented with less frequent seizures (4.3 seizures/day in FCD group and 7.2 seizures/day in TSC); however, none of these trends were statistically significant.

Quantitative Western blot analyses for SMI312 (Fig. 5A) were consistent with the immunohistochemistry, demonstrating increased SMI312 expression in tubers (205% of control, n = 6; p < 0.01), and to lesser extent in peritubers (133% relative to controls, n = 6; p > 0.05). In contrast, non-TSC epilepsy specimens demonstrated no change in SMI312 expression relative to control tissue (114% of controls, n = 7; p > 0.05), suggesting that seizure alone does not cause an increase in axonal connectivity.

Figure 5.

Altered cellular composition in tubers and peritubers, relative to non-TSC epilepsy and control samples. Western blot quantification of neuronal and glial markers SMI312 (A), growth cone protein GAP-43 (B), NeuN (C) MBP (D), and Olig2 (F), and the negative correlation of MBP to GAP-43 expression (E). Values are expressed as mean ± SEM. Representative western blot bands are show in G.*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Next, we quantified the expression of axonal growth cone protein GAP-43, which is highly expressed in growing axons.[43] The GAP-43 levels (Fig. 5B) were similarly significantly higher in tubers (177% of control, n = 5; p < 0.01), but less in peritubers relative to control (140% of control, n = 6; p > 0.05). The mechanism underlying axonal GAP-43 upregulation appears again to be mTOR-dependent, as non-TSC epilepsy specimens did not exhibit increased GAP-43 expression (90% of control, n = 6; p > 0.05).

To test whether the differences in the expression of these two axonal proteins could be explained by differences in overall neuronal densities, we performed quantitative analysis for NeuN (Fig. 5C), a nuclear protein expressed exclusively in differentiated neurons.[44] NeuN levels were slightly lower, although not significantly, in both tubers and peritubers relative to controls (66% and 72% of controls, respectively, n = 4 per group; p > 0.05). In non-TSC epilepsy cases, NeuN expression was comparable to controls (107%, n = 7; p > 0.05). The relative decrease in NeuN expression in TSC samples is consistent with the lack of NeuN positivity in undifferentiated giant cells,[4] suggesting a defect in neuronal differentiation and migration (Fig. 1G–I) rather than neuronal loss. It also suggests increased axonal branching per individual neuron in both tubers and peritubers.

In addition, we quantified MBP levels in TSC and non-TSC specimens (Fig. 5D) and found significantly lower expression levels in both tubers (29% of control, n = 6; p < 0.0001) and peritubers compared to controls (40% of control, n = 7; p < 0.001). Of interest, MBP levels were also significantly lower in non-TSC epilepsy specimens relative to controls (45% of control, n = 7; p < 0.001).

Decreased MBP expression was negatively correlated with GAP-43 in tubers and peritubers (p < 0.01; Fig. 5E), but not in epilepsy cases without TSC (p > 0.05; data not shown), consistent with the known inhibitory effects of GAP-43 signaling on myelination during normal brain development.[43, 45] To determine whether the decrease in MBP expression reflects a primary loss of oligodendrocytes, we quantified the expression of Olig2 (Fig. 5F), a transcription factor expressed in the entire oligodendrocyte lineage, independent of maturational status.[46] In contrast to the MBP deficiency, we did not observe any significant differences in Olig2 expression in tubers (89% of controls, n = 4; p > 0.05) and peritubers (92% of controls, n = 4; p > 0.05). Similarly, decreased MBP expression in non-TSC cases was not associated with decreased Olig2 expression (105% of controls, n = 7; p > 0.05), raising the possibility that seizure activity may directly affect oligodendrocyte maturation or inhibit myelin production.

Discussion

In this study we demonstrate in a large series of cases the presence of significant cellular and molecular abnormalities extending beyond the MRI-defined cortical tubers. These include abundant dysplastic neurons and giant cells, dysregulated mTOR signaling, increased axonal connectivity, and hypomyelination. Although there is increasing evidence of prominent perituberal epileptiform activity,[8-12] controversy persists about the anatomic and physiologic substrate of these changes. Although resection of residual perituberal epileptogenic tissue may be required to achieve seizure freedom,[8, 11] others argue that abnormal electrical activity originating in or involving the perituberal region may be due to “secondary epileptogenesis,” and may be reversible following tuber resection alone.[47] We demonstrate here that cortical tubers present the most severe cellular and molecular abnormalities; however, we also provide evidence of significant perituber pathology, suggestive of intrinsic epileptogenicity.

Histologically, tubers have been characterized as discrete brain lesions presenting loss of cortical architecture and enlarged dysmorphic cell types, including dysplastic neurons and giant cells.[1] Both dysplastic neurons and giant cells demonstrate altered mTORC1 signaling; however, dysplastic neurons express exclusively neuronal markers, are able to form functional synaptic contacts, and can generate action potentials. In contrast, giant cells have an undifferentiated neuroprogenitor phenotype, expressing both neuronal and glial markers, do not form synaptic contacts, and are electrophysiologic silent.[3, 4, 30]

Most studies focus on characterizing the cellular pathology of cortical tubers, and little is known about the adjacent perituberal regions. The epileptogenicity of perituberal tissue and the lack of consistent correlations between “tuber burden” (i.e., number and size), seizure severity, and cognitive or behavioral impairments[13] led to the hypothesis that diffuse microstructural abnormalities in gray and white matter contribute to these neurologic manifestations. This hypothesis is supported by the observation that only 80% of TSC patients develop tubers.[1] Indeed, previous histopathologic analysis of autopsy nontuber cortex demonstrated focal areas of dyslamination, heterotopic dysplastic cells, or even small collections of giant cells and dysplastic neurons (“microtubers”), often located in proximity to cortical tubers.[16, 17]

Our study confirms the presence of cellular dysplasia outside of the MRI-defined cortical tubers. Furthermore, we found that the density of cells with mTORC1 hyperactivation was similar in the peritubers and tubers. Our results differ from the more modest changes reported in postmortem studies, and it is possible that this is mainly a feature of epileptogenic tubers. Epileptogenic tubers may be distinguished from nonepileptogenic tubers by increased volumes of 2-fluorodeoxy-d-glucose (FDG)–positron emission tomography (PET) hypometabolism relative to MRI tuber size.[11] These hypometabolic areas, not detected by MRI, are associated with abnormal cortical architecture and altered cellular composition. Extended areas of cortical dysplasia near epileptogenic tubers—especially the presence of abundant dysplastic neurons within the perituberal cortex—may create a dysfunctional network more prone to generate seizures, as these cells are often disoriented, have larger soma size, display longer dendrites, and express dysregulated neurotransmitter receptors, which may all predispose to aberrant intracortical connections.[4, 29-31]

In addition to inducing striking structural neuropathology, upregulation of the mTORC1 pathway may contribute to neurologic dysfunction in TSC by directly affecting neuronal and glial function.[48] In mice, the degree of mTORC1 pathway hyperactivation seems to correlate with the severity of neurologic dysfunction (e.g., earlier seizure onset, more severe epilepsy, and more pronounced behavioral abnormalities).[49, 50] Seizures, cognitive deficits, and autistic-like behavior in mouse models of TSC can be prevented or suppressed by treatment with the mTORC1 inhibitor rapamycin,[22, 27, 33, 50-53] even though this drug has no direct effect on neuronal excitability.[54] In addition, clinical trials using rapamycin have shown that although the drug does not have an impact on tuber size, it has a significant effect on epilepsy.[55]

Despite significant advances in our understanding of the role of mTORC1 signaling in TSC, little is known about the contribution of dysregulated mTORC2 activity in the pathogenesis of the disease. Inactivation of mTORC2 signaling in neural progenitor cells through conditional deletion of Rictor (rapamycin insensitive component of TOR) results in significant hypomyelination and abnormal behavior reminiscent of TSC, even in the absence of gross brain abnormalities or altered cortical layering.[56]

In this study, we show decreased mTORC2 activity in both tubers and peritubers, which is consistent with differential regulatory roles of Tsc1/Tsc2 on the mTORC1 and mTORC2 signaling.[57] Although decreased mTORC2 activation was demonstrated in mouse models of TSC,[22, 32] this is the first evidence of altered mTORC2 signaling in the human postnatal TSC brain. Of interest, analysis of human fetal tubers and Tsc2−/− mouse neuronal progenitor cells shows concomitant hyperactivation of both mTORC1 and mTORC2,[52] but this apparent discrepancy might be due to age-specific differences in the upstream activators of these pathways. Our data suggest that mTORC1-independent mechanisms may also contribute to the development of the disease. Such findings have implications for clinical trials using mTORC1-specific inhibitors (e.g., everolimus) for TSC.

Recent imaging studies provide compelling evidence for significant nontuber pathology in TSC patients. An increase in the apparent diffusion coefficient (ADC), which is a measure of disorganized white matter tracts, has been observed both in tubers and outside of tubers.[11] Furthermore, a decrease in fractional anisotropy (FA) or increased diffusivity (which reflects the degree of alignment of cellular structures within the fiber tracts and their structural integrity) has been described in TSC patients.[14, 58] Possible causes of increased diffusivity include increased spacing of axons, more permeable myelin sheaths, disorganized axons, decreased radius of axons, and hypomyelination.

Consistent with these imaging studies, we demonstrate for the first time a complete disorganization of axonal projections and an overall increase in axonal connectivity in the TSC brain, most pronounced within cortical tubers. Most axons presented an excitatory phenotype and demonstrated the capability to form abundant synaptic contacts with the surrounding neurons, suggesting increased local excitatory innervation, in line with a recent report showing increased spontaneous excitatory synaptic activity in human TSC neurons.[31] Such effects are likely due to an upregulated neuronal growth program, downstream from mTOR, as axonal sprouting in non-TSC epilepsy cases was not observed. This hypothesis is further supported by the observation of increased synapse densities in hemimegalencephaly,[59] another hamartomatous malformation of the brain characterized by mTOR dysregulation.[60]

During normal development, axons are produced in excess and this process is closely associated with increased synaptogenesis.[61] By the end of the first postnatal year, these excessive connections are eliminated and a mature, more precise innervation pattern emerges. Failure to prune excessive excitatory connections during brain development has been associated with both chronic epilepsy[62] and autism,[63] suggesting that therapies directed to inhibit axonal growth or induce axonal pruning might be beneficial for patients with TSC.

Increased axonal growth and formation of aberrant excitatory synapses have also been observed in adult models of hippocampal epileptogenesis,[64] as well as in models of neocortical posttraumatic epilepsy.[65] Similar changes have been documented in human hippocampus from patients with temporal lobe epilepsy.[41, 66] It is notable that we found no evidence for increased axonal connectivity in our non-TSC epilepsy cases, despite no significant differences in average seizure duration and frequency between the two groups, suggesting that axonal sprouting may play a less important role in the pathophysiology of human neocortical epilepsies not associated with TSC.

Finally, we demonstrate a significant reduction in MBP expression in TSC and non-TSC epilepsy cases, consistent with either deficient oligodendrocyte maturation, or a defect in MBP production. Oligodendrocyte precursor cell maturation is influenced by a multitude of positive and negative axonal signals. During normal development, the inhibitory signals, including polysialylated-neural cell adhesion molecule (PSA-NCAM), Notch and GAP-43 signaling, decrease in parallel with progression of myelination, whereas under pathologic conditions they need to be removed or blocked in order for myelination to proceed.[67] Here, we found that GAP-43 levels are upregulated and negatively correlated with MBP expression in TSC, suggesting that in these patients myelination is likely inhibited by the growing axons. However, we cannot exclude the possibility that seizure activity may have contributed to these changes by directly affecting oligodendrocyte maturation or inhibiting myelin production.

The mechanism of deficient myelination in patients with chronic epilepsy without TSC is not clear, but it is interesting to note that after seizures, neurons secrete proteases such as metalloprotease 9 (MMP-9)[68] that can degrade extracellular matrix (ECM) proteins, including laminins, which are well known for promoting oligodendrocyte maturation and for stimulating myelin production.[69]

Myelination plays an important role in promoting and maintaining structural and functional axonal integrity. Myelin prevents axonal sprouting, thereby limiting synaptic plasticity, it increases the axonal caliber expansion by increasing the number of phosphorylated neurofilaments, and it promotes the clustering of voltage-gated Na+ and K+ channels at the paranodal-juxtaparanodal axonal domains which increases conduction velocity.[70, 71] In addition, myelinating oligodendrocytes secrete trophic factors that provide metabolic support and protection against excitotoxic damage through overactivation of axonal glutamate receptors.[72] Hypomyelination may have multiple negative effects, permitting the unrestricted growth of axons and reducing the velocity of action potential propagation, but also predisposing to late-onset axonal dysfunction and damage. Treatments that would provide axonal protection and/or promote myelination might be beneficial for both TSC and non-TSC epilepsy patients.

In summary, this study validates previous DTI studies showing widespread brain abnormalities in TSC and provides a deeper insight into the basic mechanisms of epilepsy and behavioral disabilities associated with the disease. In addition, our data is in line with the recent classification system of FCDs introduced by the International League Against Epilepsy (ILEA) Task Force,[73] which recognizes cortical dysgenesis adjacent to an acquired primary lesion as a new clinicopathologic entity called FCD type III. Nevertheless, this classification does not address dysgenesis adjacent to type IIb malformations, including TSC cortical tubers, and further extension of this classification might be warranted to include the types of lesions we are reporting here.

Acknowledgments

This work was supported by FACES (Finding a Cure for Epilepsy and Seizures) (to D.M.T.), the James Shaw Foundation (to D.M.T.), and the Tuberous Sclerosis Alliance (to H.L.W.). Control human tissue was obtained from the National Institute of Child Health and Human Development (NICHD) Brain and Tissue Bank for Developmental Disorders at the University of Maryland, Baltimore, MD, contract HHSN275200900011C, Ref. No. N01-HD-9-0011.

Disclosure

None of the authors has any conflict of interest to disclose. We confirm that we have read the Journal's position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.

Ancillary