Tissue

Specimens including cortical tubers (n = 16) and MRI normal perituberal tissue (n = 10) were obtained from TSC patients with medically refractory epilepsy who were undergoing epilepsy surgical resection. In one case, hippocampus with MRI normal architecture was resected as the epileptogenic area together with adjoining cortical tubers. The mean age of the 16 patients was 9.8 ± 1.5 years (range: 2.5 – 23 years), with 9 females and 7 males. The epileptogenicity of the resected tubers and perituberal tissue was established during invasive subdural intracranial monitoring of patients, as previously described [9]. Perituberal tissue did not show any abnormalities on detailed preoperative MRI examination using imaging protocols designed to maximize preoperative detection of malformations of cortical development. The location of the tuberal tissue and MRI normal perituberal tissue was confirmed intraoperatively by gross visual inspection, as well as by frameless stereotactic intraoperative MRI based surgical navigation. All patient protocols were approved by the Institutional Review Boards of Columbia University Medical Center and New York University Medical Center.

Histology and immunohistochemistry

Specimens were fixed in 4% paraformaldehyde in PBS for 14–18 h (4°C). 40 μm sections were prepared with a vibratome (Leica VT1000S) and stored in cryoprotectant solution at -20°C. Standard procedure for Nissl staining with cresyl violet was used for routine analysis of tissue.

Antibodies

Secondary antibodies include: conjugated to fluorophores: anti-mouse Alexa Fluor 488, 594, and 633; anti-chicken Alexa Fluor 594, 633; anti-rabbit Alexa Fluor 594; anti-goat Alexa Fluor 594, 633; anti-guinea pig Alexa Fluor 594, 633, all from goat or donkey (1:300, Molecular Probes, Eugene, OR).

Antigen retrieval procedure was performed with Antigen Unmasking Solution (Vector Laboratories, Inc., Burlingame, CA) by boiling for 40 sec in a microwave.

For light microscopy sections were pretreated in 3% H₂O₂ in methanol (30 min, at room temperature [RT]) and after blocking in 10% donkey serum (30 min, RT) incubated overnight (4°C) with primary antibodies. After washing sections were incubated with biotinylated secondary antibodies (1:1000, 1 hr, RT) and then, after washing, with avidin–biotin-peroxidase complex (ABC, Vector Lab.) (1 hr, RT). Peroxidase activity was visualized with 0.03% DAB (Sigma) with 0.005% H₂O₂ in 0.05 M Tris buffer. Sections were transferred to glass slides, dehydrated in a graded ethanol series and mounted in DPX (Sigma).

For double- and triple-immunofluorescence, after blocking with 10% normal goat (or donkey) serum (30 min, RT), free-floating sections were incubated in a mixture of primary antibodies raised in different species for overnight (4°C). For visualization, Alexa Fluor-conjugated secondary antibodies were used for 1 h at RT. Fluorescent Nissl reagent (NeuroTrace 640/660 deep-red, 1:150, Molecular Probes) was applied simultaneously with secondary antibodies in cases of double-immunostaining for visualization of general histological structure.

Blocking serum, primary, secondary antibodies, and fluorescent Nissl reagent were applied in 0.2 % Triton X-100 in PBS. Sections for fluorescent microscopy were mounted on slides in Vectashield (Vector Lab). To control the specificity of immunostaining, primary antibodies were omitted and substituted with appropriate normal serum. Slides were viewed using a confocal microscope (Bio-Rad Radiance 2000, Nikon E800).

Quantitative immunohistochemical analysis

We determined the size of microtubers using the diameter of the area occupied by astrocytes with high levels of GFAP or vimentin in the images (merged from the stacks of adjacent 6 images [1024 × 1024 pixel resolution, observed area 606 × 606 μm] captured at a distance of 1 μm from each other) obtained with a confocal microscope. In cases of oval-shaped microtubers, the longest diameter was measured. The border of microtubers was considered as the point at which the immunostaining of GFAP/vimentin decreased significantly. The long processes of astrocytes were excluded in the determination of borders. Furthermore, since most tubers contained astrocytes that did not stain with anti-GS antibodies, a double immunostaining for GS and either GFAP or vimentin showed the border clearly.

The numbers of (1) giant cells (detected with GS and p-S6 immunostaining), (2) p-S6+, p44/42+, and SPARC+ astrocytes in microtubers (detected in double immunostained sections for p-S6/p44/42/SPARC and GFAP), (3) CD68- and IBA1-immunolabeled microglial cells (in microtubers in double immunostaining with GFAP, and in tuber border area in double immunostaining with GS) were quantified in the images merged from a stacks of adjacent 6 images (1024 × 1024 pixel resolution, observed area 606 × 606 μm) captured at a distance of 0.5 μm from each other. Only cells with clearly outlined nuclei were counted.

The size of a giant cells was quantified based on OD of the area occupied by the cell body immunoreactive for p-S6 and GS. In perituberal areas, the data on giant cells in microtubers and giant cells located in normal parenchyma were pooled together because we did not find significant differences between these types of cells. Images were obtained from stacks of 6 images (1024 × 1024 pixel resolution, observed area 148 × 148 μm, step 0.5 μm). Only those cells that were unequivocally indentified as giant cells (based on a large size, irregular shape, and immunostaining) were taken into consideration.

To determine the level of GS, CD68 and Iba1 immunoreactivity, optical density (OD) was measured in the area of interest (in microtubers, occupied with reactive astrocytes with high levels of GFAP) in the images prepared as above (area of observation was 295 × 295 μm) that were transferred to Adobe Photoshop, grayscaled and used for analysis of OD with Scion Image Beta 4.02 (public domain). Level of EAAT2 immunoreactivity was measured based on OD obtained from merged stacks of adjacent 6 images (1024 × 1024 pixel resolution, observed area 148 x 148 μm) captured at a distance of 0.25 μm from each other). EAAT2 and GFAP immunoreactivities correspond to different cellular compartments (cytoplasm and plasma membrane, respectively) and the surface occupied by GFAP and EAAT2 inversely correlate (more cytoplasm in the area used for analysis of OD, less plasma membranes). To determine the OD of EAAT2 we used an equation: OD_EAAT2 = OD_EAAT2/(S_AOI − OD_GFAP − OD_Nissl), where S_AOI stands for the area of the image where OD was determined, OD_GFAP and OD_Nissl stand for area occupied with GFAP OD and nuclei, respectively. Thus we have calculated arbitrary units, that indicate the EAAT2 OD for arbitrary units of plasma membrane.

More than 6 images were used from each case from every area of interest (perituberal area, microtubers, tubers, tuber borders) for quantification of cell number and OD.

Statistical analysis

Data are expressed as mean ± SEM. Differences in the incidence were analyzed by Chi-square test. Continuous parameters were analyzed with Student’s t-test or with one- way ANOVA where appropriate. P < 0.05 was considered significant.

Defining the borders of cortical tubers

To discriminate between perituberal cortex and tubers, it is first critical to demarcate the cortical tubers themselves in histological sections. Based on immunohistochemical analysis, we have found that astrogliosis is the best criterion for morphological discrimination between tubers and the surrounding neocortical parenchyma (Figure 1). In addition to providing a clear histological delineation of tubers, this proposal is in line with MRI data where both T1- and T2-weighted images correlate with the level of tissue sclerosis/gliosis [13]. Several markers of astrocytes were used to determine tuber astrogliosis and accordingly tuber borders. These included markers typical for gray matter protoplasmic astrocytes such as glutamine synthetase (GS) and the astrocyte specific glutamate transporters (EAAT1 and EAAT2), all of which showed profoundly lower levels in tubers (Figure 1). In addition, markers that are seen in reactive/gliotic astrocytes such as glial fibrillary acidic protein (GFAP), vimentin, CD44, αB-Crystallin, and S100 revealed striking increases in tuber tissue (Figure 1, only some shown). Tuber borders were thus clearly outlined, segregating perituberal protoplasmic astrocytes from tuber gliosis (Figure 1). We observed that the borders of tubers were not always smooth, displaying uneven profiles with many peripheral protrusions (Figure 1a,b,c,e). Long processes of tuber astrocytes penetrated into the perituberal parenchyma and intermingled with neighboring protoplasmic astrocytes (Figure 1f). Furthermore, small foci of astrogliosis (microtubers, see below) were regularly found near the tuber borders (Figure 1a,b,c).

Tuber-specific cellular abnormalities in perituberal MRI-normal tissue

Most of the observations presented in this section were obtained from MRI-normal, epileptogenic perituberal tissue (n = 10) that was dissected during surgery separately from tubers. Macroscopic tubers were never found in these samples of tissue.

Microtubers are common pathological feature of perituberal tissue

In contrast to the highly gliotic tubers, the perituberal gray matter was populated mainly with protoplasmic astrocytes displaying high levels of GS and low levels of GFAP (Figure 2a,b). To detect abnormal cells typical for tubers within perituberal tissue, we used several markers specific for neurons and glial cells and known to be increased in giant cells and cytomegalic neurons [14,15], as well as p-S6, to look for mTOR activation [16]. In every section from every specimen of perituberal tissue, we found aberrant cells characteristic of tubers: giant cells, gliotic astrocytes, and cytomegalic neurons.

In the perituberal gray matter, giant cells were usually surrounded by fibrous-like astrocytes with long processes displaying high levels of GFAP and CD44. These small astrogliotic islands were clearly outlined from the neighboring normal parenchyma containing protoplasmic astrocytes (Figure 1a,b,c; 2). We previously designated these microscopic islands as “microtubers” [11], in contrast to the canonical cortical “macrotubers” detected by MRI. They were relatively homogeneous in size, with an average diameter of 284.7 ± 17.3 μm (n = 70; range: max 461.6 μm, min 158.6 μm). If we consider the shape of microtubers as roughly spherical, we estimate that one microtuber is composed, on average, of ~ 27 astrocytes and includes ~ 20 neurons.

Two types of microtubers were distinguished based on the shapes and immune profile of astrocytes. The majority (∼80%) of microtubers (which we designated as type I) contained many astrocytes with long processes that radiated for several hundred micrometers into the adjacent gray matter, which was occupied by protoplasmic astrocytes (Figure 2a,b,d). The processes of these cells were devoid of the miniature lamellipodial-like processes that are a characteristic feature of protoplasmic astrocytes. Such structural feature was especially obvious when cells were immunostained for the plasma membrane glycoprotein CD44 (Figure 2b1). A minority (~20%) of microtubers (designated type II) were largely composed of astrocytes with processes of normal length endowed with many miniature lamellipodial leaf-like extensions that produced the typical bushy-like appearance of protoplasmic astrocytes (Figure 2a,c). However, in contrast to typical protoplasmic astrocytes, these cells were CD44+ (Figure 2c). It is worth noting that some type II microtubers contained only a few (2–4 in a plane of inspection) reactive-like astrocytes neighboring a giant cell (Figure 2e).

We suggest that the astrocytes with long, non-branched processes in type I microtubers are similar in many ways to the CD44+ long-process/interlaminar astrocytes in gray matter and/or to fibrous astrocytes in white matter, whereas astrocytes with processes of protoplasmic astrocytes size and shape, but CD44+, in type II microtubers are reactive protoplasmic astrocytes. To test this hypothesis we used immunostaining for SPARC/osteonectin, a glycoprotein we have found to be a characteristic marker of CD44+ interlaminar and fibrous astrocytes in human brain [17]. Indeed, type I microtubers contained many SPARC+ astrocytes whereas only a few SPARC+ cells were observed in type II microtubers (11.8 ± 0.824 per microtuber in type I vs 0.824 ± 0.3 in type II, p < 0.001) (Figure 3a,b). In addition, all SPARC+ astrocytes were CD44+ and had clearly outlined, long main branches without lamellipodial-like processes (Figure 3c). It should be noted that many giant cells also showed immunolabelling for SPARC (Figures 3b,d; 4c,d).

Astrocytes in both types of microtubers were characterized by diminished immunoreactivity for glutamate transporters (EAAT2 and EAAT1) (only EAAT2 was quantified, OD, arbitrary units, 0.1 ± 0.03 in microtubers vs 0.6 ± 0.3 in surrounding normal parenchyma, p < 0.001) and for GS (OD, arbitrary units, 1199.7 ± 296.6 in microtubersvs 9874.4 ± 1321.8 in surrounding normal parenchyma, p < 0.001) (Figure 3e-g). This decrease in GS and glutamate transporters was particularly prominent in the central part of microtubers, where many neurons were surrounded by astrocytes lacking these proteins (Figure 3e-g).

Many astrocytes in microtubers showed p-S6 immunoreactivity (Figure 4a,c,d). Especially prominent p-S6 expression was observed in microtubers located in deep cortical layers (255 of 349 astrocytes in V-VI layers vs. 46 of 123 astrocytes in upper I-IV layers, p < 0.001).

To gain additional insight into the characteristics of astrocytes in microtubers, we tested activation (phosphorylation) of p44/42 MAPK (p44), which was previously shown to be activated in tubers [18]. We found that many astrocytes in microtubers displayed a high level of p44 (Figure 4b,c). Microtubers located in deep cortical layers and in white matter contained more p44+ astrocytes than those in upper gray matter layers (56 of 121 cells in upper layers vs. 156/184 in deep layers, p < 0.001). Only a few (6/61) giant cells revealed immunoreactivity for p44 (Figure 4c). Many p44+ astrocytes were also immunopositive for p-S6 (in deep cortical microtubers ~ 80%, 214 of 305 p44+ astrocytes) (Figure 4c). Many SPARC+ astrocytes were also p-S6+ (Figure 4d).

The appearance of reactive-like astrocytes in the immediate vicinity of giant cells in microtubers raised the possibility that giant cells might modulate their environment. We examined microtubers for microglial reaction as a marker of inflammatory brain parenchymal disturbances. In 45/47 microtubers tested with CD68 and LN-3 immunostaining, there was no prominent activation of microglia detected (shown only CD68, Figure 5a). Optical density of CD68 immunolabelling did not differ in microtubers from surrounding normal parenchyma (6150 ± 885 arbitrary units per 1 mm² in microtubers vs 6030 ± 799 in normal parenchyma, P = 0.922). In only 2/47 microtubers, activated microglia were observed in large areas that included the microtubers but extended far beyond their boundaries (not shown). We also used immunostaining for Iba1, as a general marker for both resting and activated microglia, and did not find significant microglial changes in microtubers (Figure 5b). Numbers of Iba1+ microglial cells did not differ significantly from surrounding normal parenchyma (61.3 ± 7.66 cell per 1 mm² in microtubers vs 55.5 ± 8.08 in normal parenchyma, P = 0.62).

We assayed also whether giant cells contain tumor necrosis factor alpha (TNFα), as shown previously [18]. TNFα is a cytokine that is responsible for induction of reactive changes in astrocytes in variable brain pathologies [19]. We found that many giant cells were TNFα+ (in microtubers 14 of 21 cells, in macrotubers 77/115, p = 0.722) (Figure 5c,d).

In perituberal tissue, in addition to giant cells in microtubers, we also found scattered giant cells surrounded by normal appearing protoplasmic astrocytes (Figure 5e,f,g). There were <1/10^th as many of these isolated giant cells in comparison to giant cells within microtubers (24 versus 235 cells), and they were found predominantly in upper (I – III) cortical layers (19 of 24).

Giant cells in the perituberal gray matter (both in microtubers and isolated) never reached as large a size as those within tubers (374.3 ± 22.2 μm² (n = 20) vs 888.9 ± 40.88 μm² (n = 20), respectively, p < 0.001) and never had as long and thick processes as giant cells in tubers (compare those in Figure 5e-g with those in Figure 6a-c). We also did not find giant cells immunopositive for p75 neurotrophin receptor (p75) in perituberal tissue although p75+ giant cells were often observed in tubers (106/178 in tubers vs. 0/34 in microtubers, p < 0.001) (Figure 6c). Such differences in giant cells might indicate different developmental routes or different growth and ‘maturation’ of giant cells.

An examination of a single TSC patient’s hippocampus that had a normal MRI appearance but was resected as part of the epileptogenic zone revealed microtubers and cytomegalic neurons in Ammon’s horn (Figure 6d,e), as well as in the hilus (not shown). This further supports the idea that microtubers represent important parts of epileptogenic zones apart from tubers.

Astrocyte heterogeneity in cortical tubers

As emphasized above, one of the most prominent features of tubers was a high level of astrogliosis. In 7/16 tubers, severe astrogliosis completely occupied the tuber gray matter (to pia surface) and almost every astrocyte was devoid of GS and glutamate transporter (EAAT1 and EAAT2) immunoreactivity. In the other 9/16 tubers, we found that upper cortical layers (usually layer 1 and 2) were populated with protoplasmic astrocytes and many microtubers were located near the tubers (Figure 1a,b,c). Small “islands” of protoplasmic astrocytes were observed within tuber highly gliotic tissue that produced a patchy pattern due to intermingled gliotic and non-gliotic areas (Figure 1e, 7a). Because we did not make 3-D reconstructions of the whole tubers we cannot rule out the idea that these island-like areas of normal brain parenchyma were not in reality deep invaginations of perituberal, normal parenchyma.

Although most tubers were composed of a homogeneous population of gliotic astrocytes, in some areas within tubers it was possible to discriminate groups of astrocytes displaying higher levels of vimentin, p-S6, and p44 than neighboring gliotic astrocytes (Figure 7b,c,d). These cells were grouped around giant cells and might correspond to microtubers with astrocytes that did not reach the “final stage of gliotic maturation” observed in neighboring highly gliotic tuber astrocytes. The average diameter of such groups was 329 ± 15.1 μm (n = 15; range: max 459.3 μm, min 236.7 μm), overlapping extensively with the dimensions of the perituberal microtubers. Indeed, these groups within tubers exceeded the average dimension of microtubers in the perituberal parenchyma by less than the size of one astrocyte domain (~100 μm, not including the area of overlapping processes from neighboring cells [21]).

The phenotype of astrocytes located at the tuber border zones was different from that of astrocytes in the central part of the tubers, as well as from that of normal protoplasmic astrocytes in the neighboring gray matter. These ‘transitional’ areas, which we speculate represent developing gliosis in tubers, were populated with astrocytes that had characteristic features consistent with ongoing astrocyte changes. 1) Only in these areas were there unusual astrocytes with shapes of fibrous astrocytes (cells with long processes devoid of miniature leaf-like protrusions) and yet with high levels of GS, EAAT1, and EAAT2, which are characteristic features of protoplasmic astrocytes (Figure 7e-h). Taking into consideration that the majority of these unusual astrocytes did not show immunoreactivity for SPARC (Figure 7f,h), we considered them to be astrocytes with mixed protoplasmic and fibrous features. 2) Many astrocytes in these areas were immunopositive for phosphorylated p44 MAPK. These p44+ astrocytes showed minimal immunoreactivity for GS and glutamate transporters. (Figure 8a-c). 3) p-S6+ astrocytes predominated in these areas and often immunoreactivity for p-S6 colocalized with p44 and SPARC (Figure 8b,d). Both p44+ and p-S6+ astrocytes expressed SPARC and revealed low levels of GS and glutamate transporters, whereas astrocytes with high levels of GS and glutamate transporters were SPARC immunonegative (Figure 7f,h; 8c,d). Thus, many astrocytes in these “transition zones” display features of both protoplasmic and long-process astrocytes. These may well reflect a phenotype in transition.

The level of microglial activation (based on immunoreactivity for CD68 and Iba1) in these transitional areas was similar to that in tubers and significantly exceeded that of the non-gliotic areas surrounding tubers (not shown).

Is perituberal tissue an epileptogenic area?

What is the functional significance of aberrant cellular elements in perituberal tissue? Do they participate in seizure initiation and/or progression? Seizure origin in TSC is the subject of much debate with several observations favoring perituberal tissue as a source of spontaneous epileptiform activity [5-9]. We found that perituberal tissue contains both microtubers and cytomegalic neurons, both of which may contribute to the epileptogenicity of this tissue. Electrophysiological analysis of tubers and cortical dysplasia has shown that only cytomegalic/dysplastic neurons revealed features of hyperexcitabilty, while giant/balloon cells were electrophysiologically silent [33]. Cytomegalic neurons in perituberal tissue may have similar or identical properties of their tuber counterparts and thus could potentially trigger ictal events. Similarly, the aberrant astrocytes that surround giant cells in “microtubers” are characterized by markedly diminished expression of EAAT1, EAAT2, and GS, proteins important for the uptake and metabolic conversion of glutamate to glutamine within astrocytes. We have found that similar astrocytes in tubers have deficient glutamate uptake [34]. Thus, even normal neurons located in microtubers may be predisposed to overexcitation because of the profound diminution of astrocytic glutamate clearance [35,36]. Our finding of microtubers and cytomegalic neurons in an epileptogenic hippocampal focus (Figure 6) also supports the thesis that these aberrant perituberal components might be the triggers of perituberal seizures in TSC.

Microtubers and the derivation of tuber gliotic astrocytes

Despite great progress in the molecular biology of TSC, less is known about the origin and development of phenotypic features of abnormal cell types in the TSC brain. Analysis of microtubers may help to elucidate how tuber gliotic astrocytes develop.

Cortical tubers are composed of tightly packed gliotic astrocytes that make immunohistochemical analysis of individual cell shape and properties very complicated, if possible at all. In contrast, in microtubers where cells are not so densely packed, we identified two types of astrocytes: 1) astrocytes with long, straight, main processes lacking miniature leaf-like protrusions (the majority of astrocytes) and 2) protoplasmic astrocytes revealing features of reactive cells (a minority of cells). The first, based on morphology and immune profile, are similar to the CD44+ long-process astrocytes found at the pial surface and in white matter [17,21]. We propose that the majority of gliotic tuber astrocytes are aberrant CD44+ long process/fibrous-like astrocytes that due to abnormal development appear and replace the normal protoplasmic astrocytes in gray matter. Thus, due to such aberrant development, astrocytes that usually are located in white matter and the pial surface populate the gray matter.

Although long-process astrocytes in normal gray matter were described by Ramon y Cajal [37], their appearance and functions in gray matter remain mysterious. It should be remarked that while the divergent development of neurons and glial cells (including astrocytes) is well studied [38,39], there is little known about the divergent origins of protoplasmic, CD44+ long process, and fibrous astrocytes.

The appearance of reactive-like astrocytes in microtubers raises critical questions about the relationship between giant cells and neighboring astrocytes. Why are giant cells often but not always surrounded by reactive protoplasmic astrocytes? What may be the cause of these reactive-like changes in astrocytes? One possibility based on our data is that giant cells may release cytokines, e.g. TNFα that promote local reactive changes in astrocytes. As is well known, TNFα is a powerful trigger of reactive changes in astrocytes [19].

Many astrocytes in microtubers and in tubers showed a high level of immunoreactivity for activated (phosphorylated) p44 MAPK. It should be noted that overactivation of p44 was consistently observed and considered as an important, although unexplained, mechanism in TSC-related neoplasms [40]. p44 and mTOR pathways have close, although not completely understood, interrelations. p44 may inactivate the TSC1-TSC2 complex through phosphorylation of TSC2 [41]; while mTOR can provide negative (through p70 S6K1 and PI3K, [42]) or positive (through protein phosphatase 2A, [43]) feedback on p44. At this time, it remains unclear how p44 MAPK activation may be related to astrocyte changes in TSC.

The heterogeneity of the astrocyte population in microtubers and tubers shown in this study is in line with a recent detailed analysis of the glial and neuronal populations in tubers [15]. These authors and our previous work [20] defined several types of aberrant astrocytes in tubers, including reactive astrocytes. Reactive changes in astrocytes, e.g. expression of p-S6 and p44, can be used as a good indicator of acute ongoing cellular alterations [19,44].

Are macrotubers built of microtubers?

Should microtubers be considered as elementary units or domains, which by consecutive aggregation create the main bodies of macrotubers? This “elementary unit” hypothesis is grounded largely on the anatomical observations: 1) rather homogeneous size of microtubers; 2) consistent association of microtubers with tuber borders; 3) appearance of microtuber-like areas with reactive astrocytes surrounding giant cells in tubers especially near their borders. We hypothesize that the CD44+ long-process astrocytes and the giant cells in individual microtubers comprise the progeny of one or more neural precursors in the embryonic ventricular zone (VZ) that have developed LOH at a TSC locus or have developed another mutation that interfere with their normal developmental program. VZ cells that remain heterozygous at the TSC 1 or 2 loci will generate normal cortical populations. Early genetic changes that give rise to large numbers of altered VZ cells may give rise to tubers, whereas late genetic changes, which give rise to one or only a few altered VZ cells may give rise to microtubers.