Giant Cells: Contradiction to Two-Hit Model of Tuber Formation?

Hamartomatous brain lesions, such as cortical tubers, subependymal nodules (SENs) or subependymal giant cell astrocytomas (SEGAs) are a hallmark of tuberous sclerosis complex (TSC), an autosomal dominant genetic disorder with a high sporadic case rate. So far, it has been elucidated that TSC is due to mutations in either of the two genes, TSC1 on chromosome 9q34 (van Slegtenhorst et al., 1997) or TSC2 on 16p13 (European Tuberous Sclerosis Consortium, 1993), encoding hamartin and tuberin, respectively. Formation of tubers and nodules occurs mainly during brain development, and seems to be nearly complete at the time of birth. It has been reported that cortical tubers and subependymal nodules are present in 19 and 31 weeks gestation fetus (Chou and Chou, 1989; Park et al., 1997).

Cortical tubers expand the gyri and cover the margin between the gray and white matter. Tubers are only occasionally seen in the cerebellum. Their calcification is commonly found, changing the tuber into a hard structure (thus, the name of the disease: “tuberous sclerosis”). Less frequently cortical tubers can degenerate into cystic lesions, which is not, however, connected with malignant transformation. Tubers can be limited to the cortex or the subcortical white matter. Originally, cortical tubers were believed to be pathognomonic, but according to current findings distinguishing these tubers from isolated cortical dysplasia on the basis of radiographic brain imaging and histological studies is difficult, although possible, at least in typical cases (Yagishita and Arai, 1999).

Subependymal nodules (SENs) are the most common brain lesions, which appear in about 90% of TSC cases. They are covered with a thin ependymal layer and contain elongated or swollen glial cells and their processes, giant or multinucleated cells, and sometimes, calcium depositions, although they are rarely calcified in the first year of life. SENs do not grow, but calcify progressively. By the age of 20, most of them are calcified.

SEGAs, whose incidence in TSC is about 5–15%, differ from subependymal nodules in their size and tendency to enlarge, which results in the clinical presentation of hydrocephalus. It is assumed that lesions greater than 12 mm are classified as SEGAs. The most important criterion of differentiation, however, is progressive enlargement of a lesion. Such a change in tumor size may lead to an increase of intracranial pressure, and result in significant morbidity and mortality. Differentiation between SENs and SEGAs is best performed on the basis of MRI investigation.

Cortical tubers are benign, while one malignant case of SEGA has been recorded (Telfeian et al., 2004). From our previous observations, we hypothesize that subependymal nodules may grow and differentiate into SEGAs (Roszkowski et al., 1995). Nevertheless, even benign brain lesions in TSC lead to seizures, mental retardation and autism (Jozwiak et al., 1998). Pathogenesis of brain lesions in TSC is uncertain, and much effort is currently being put into identifying molecular and developmental mechanisms leading to the appearance of characteristic giant cells, found in tubers, SENs and SEGAs. Because of numerous immunohistochemical and ultrastructural similarities (Bender and Yunis, 1980; Chou and Chou, 1989; Jay et al., 1993; Hirose et al., 1995), it has been hypothesized that giant cells in tubers and SEGAs share the same profile of differentiation and may be of the same cellular lineage (Lee et al., 2003).

HISTOLOGICAL EVALUATION OF GIANT CELLS

In histological evaluation of TSC brain lesions at least three cell populations can be identified: astrocytes, dysmorphic neurons, and giant cells, the last type being characteristic for tuberous sclerosis. Many of abnormally shaped neurons are stellate or multipolar cells not characteristic of normal cortex. Eosinophilic giant cells, which extend short thickened processes, are 5–10 times as large as normal neurons. The first ultrastructural research on giant cells from a subependymal tumor and a cortical tuber argued for their astrocytic origin (Trombley and Mirra, 1981), as evidenced by multipolar processes containing glial filaments and glycogen, as well as formation of hemidesmosomes with pia and vascular basement membranes. Later studies employing immunohistochemistry confirmed that subpopulations of giant cells express such glial markers as glial fibrillary acidic protein (GFAP) or S-100 protein (Roske et al., 2003; Sharma et al., 2004; Takahashi et al., 2004).

On the other hand, however, the presence of rough endoplasmic reticulum, prominent paranuclear Golgi zones and dense secretory vessicle-like core granules are suggestive of neuronal features. Giant cells in tubers also express proteins that are typically found in immature neurons and neuroepithelial precursor cells, such as nestin, MAP 2C and N-methyl-d-aspartate (NMDA) 2D receptor subunit. Tuber specimens probed with antibodies recognizing the neural marker NeuN (neuronal nuclei), a protein binding with the DNA in mature neurons, exhibit a heterogenous staining pattern (Lee et al., 2003). Thus, a more precise classification has been proposed by Yamanouchi et al. (1997), who divided giant cells into two subtypes: “neuron-like giant cells,” having a round centrally placed nucleus with a single nucleolus, and Nissl substance in cytosol. In those cells neurofilament and MAP1B immunohistochemical staining was strong, while immunopositivity for nestin and vimentin was occasional. GFAP was found only incidentally. Another subtype was “indeterminate giant cells,” characterized by abundant cytoplasm, lack of Nissl substance, one or more eccentric nuclei, and positive staining for nestin, vimentin, MAP1B and, in majority, for GFAP. Those cells were rarely positive for neurofilament staining.

One possible explanation of these findings is that cellular markers indicate giant cells’ failure to terminally differentiate prior to migration into cortex. This hypothesis is also supported by the fact that SEGA cells stain with human anti-neuronal nuclear autoantibodies of types 1 and 2 (ANNA-1 and ANNA-2), serum markers of paraneoplastic syndromes associated primarily with small cell lung cancer (Laeng et al., 1998). Also expression of doublecortin in giant cells is revealing. Doublecortin (DCX) is a microtubule-associated protein required for neuronal migration during cortical development. DCX mediates the fetal migration of neuroblasts from the proliferative ventricular zone toward the pial surface and is transiently expressed in proliferating progenitor cells and newly generated neuroblasts (Brown et al., 2003). As the newly generated cells begin expressing mature neuronal markers, DCX expression decreases sharply. It has been reported that giant cells from the cortical lesions of tuberous sclerosis stained well with anti-DCX antibodies, suggesting restricted differentiation of these cells (Mizuguchi et al., 2002). Mixed phenotype and expression of various glial- and neuron-associated epitopes as well as epitopes characteristic for immature cells within lesions of the same morphology lets us hypothesize that such lesions contain cell lineages with the capacity to undergo divergent glioneuronal or neuroendocrine differentiation to a greater extent than do other mixed glial-neuronal tumors. It is also interesting to mention that HMB-45 (human melanoma, black) antigen found commonly in uterine and hepatic angiomyolipoma, peritoneal and ovarian epithelioid angiomyolipoma, clear cell epithelioid tumor of the kidney as well as cardiac rhabdomyoma associated with TSC (Weeks et al., 1994; Ribalta et al., 2000; Ji et al., 2001; Anderson et al., 2002; Hino et al., 2002; Cil et al., 2004) suggesting similar mechanism of their formation, is not found in SEGAs (Gyure and Prayson, 1997; Sharma et al., 2004). This fact is one of the premises arguing for different pathomechanisms in brain and extracerebral lesions found in TSC patients.

Differentiation of TSC neural progenitor cells into giant cells has been demonstrated in murine models. Onda et al. (2002) showed that TSC2 null murine neuroepithelial progenitor (NEP) cells grow without growth factors, express high levels of GFAP but low levels of early neuronal lineage markers. TSC2 null NEP cells, morphologically similar to giant cells in human tubers, exhibited differentiation into giant cells expressing both IIIβ-tubulin and GFAP. On top of that, TSC2 null giant cells and tuber giant cells have similar transcriptional profiles. TSC2 null NEP cells expressed high levels of phosphorylated S6K, S6, Stat3, and 4E-BP-1, and this upregulation was reversed by the treatment with rapamycin. In comparison, TSC2 ^+/− heterozygotes did not show potent activation of S6 kinase, which was found in TSC2 null cells. All these data let us hypothesize that giant cells constitute a heterogeneous line of cells originating from progenitor ependymal cells, having the ability to differentiate, in some degree, into neuronal or astrocytic cell lines.

FUNCTIONS Of TSC1 AND TSC2 PROTEINS

Interaction of hamartin and tuberin is necessary for the formation of intracellular tuberous sclerosis complex that can inhibit activation of mammalian target of rapamycin (mTOR) kinase, regulating nutrient uptake, cell growth, and protein translation. In spite of the fact that tuberous sclerosis complex is formed by both proteins, in the largest and most comprehensive research up-to-date comprising both mutational and phenotypic studies, we showed that TSC1 mutations are less frequent and are associated with a lower frequency of seizures and moderate-to-severe mental retardation, fewer subependymal nodules and cortical tubers, less-severe kidney involvement, no retinal hamartomas, and less-severe facial angiofibroma than TSC2 mutations (Dabora et al., 2001). This may be due to individual properties of both proteins. Both TSC genes are found to be tumor suppressor genes. However, hamartin possibly inhibits tumor formation by regulating cellular adhesion through the actin-binding proteins of ezrin-radixin-moiesin (ERM) family as well as the small GTP-binding protein (G protein) Rho (Lamb et al., 2000). Also tuberin is a GTPase-activating protein (GAP) which inhibits Ras-related family of small G proteins, such as Rap1, Rab5 and Rheb (Wienecke et al., 1995; Xiao et al., 1997; Inoki et al., 2003; Zhang et al., 2003). It has been shown that human malignant astrocytomas display high activity of G proteins, e.g. p21-ras, which is secondary to overexpression of receptor tyrosine kinases activating G proteins (Guha, 1998). Signals transmitted by p21-ras activate a number of pathways leading to tumorigenesis, such as proliferation or expression of angiogenic factors. The phenomenon of G protein activity leading to tumorigenesis is exemplified by neurofibromatosis type 1 (NF1), where biallelic inactivation of the protein NF1, being a negative regulator of p21-ras, is associated with tumor formation. Tuberin regulates also cell cycle, as the absence of this protein can both induce cells to pass through the G1/S transition of the eukaryotic cell cycle and prevent them from entering a quiescent state (Soucek et al., 1997) (Table I).

GENETIC MECHANISM OF TSC

Mutation detection studies have led to identification of mutations in nearly 90% of TSC patients (Jones et al., 1999; Dabora et al., 2001). The analysis showed some 470 mutations in both TSC1 and TSC2 genes; however, much more mutations were detected in the latter gene (ratio about 1:4) (Dabora et al., 2001; Kwiatkowski, 2003), mainly missense, in-frame deletion and large deletion mutations. This ratio may reflect the relative size of the two genes. No hotspots were found in TSC1 or TSC2 sequence, and none of the single mutations accounts for more than 2% of all mutations (Kwiatkowski, 2003). On top of that, some of the identified mutations were found to be polymorphisms, not causing pathogenic effects (Roberts et al., 2003).

In a recent study, we examined the expression of tuberin and hamartin in SEGAs from nine patients with TSC, indicating a loss of both tuberin and hamartin expression in the subependymal giant cell astrocytomas of patients with both TSC1 and TSC2 mutations (Jozwiak et al., 2004). This result is consistent with a two-hit model for the development of subependymal giant cell astrocytomas, stating that both alleles of a given TSC gene have to be mutated for the disease to occur (Knudson, 1971). Indeed, loss of TSC1 or TSC2 heterozygosity is well documented in bladder transitional cell carcinoma, kidney angiomyolipoma, and cardiac rhabdomyoma occurring in the course of TSC (Henske et al., 1997; Jozwiak et al., 2001; Knowles et al., 2003; Meikle et al., 2005). In our research, we performed immunohistochemical and genetic analyses on SEGAs from seven TSC patients, four with mutations in TSC1, and three with mutations in TSC2 (Chan et al., 2004). We found that SEGA cells show high levels of phospho-S6K, phospho-S6, and phospho-Stat3, all those proteins being downstream of and indicative of mTOR activation. On top of that, five of the six SEGAs also showed evidence of biallelic mutation of TSC1 or TSC2 which suggested that SEGAs are likely to arise through a two-hit mechanism of biallelic inactivation of TSC1 or TSC2, leading to activation of the mTOR kinase. Wilson et al. (1996) studied 26 apparently sporadic TSC cases, two TSC families non-informative for linkage analysis and two confirmed chromosome 16-linked TSC families. The diversity of mutation types found argues that TSC2 does not act in a classic tumor suppressor fashion. In addition, there was no specific correlation between different mutations and clinical presentation of the disease, which lets us think that the lack of wild-type protein (resulting from biallelic inactivation) is sufficient for the activation of mTOR and subsequent formation of brain lesions. These results led to current studies with rapamycin (or its analogues) used for targeted inhibition of mTOR in the treatment of TSC. Such a treatment in mouse models gave very promising results (Lee et al., 2005).

LOSS OF HETEROZYGOSITY IN BRAIN TUBERS

In spite of the above reports implying that second-hit mutation is sufficient for TSC growth formation, in case of brain lesions two-hit model of pathogenesis seems not to be the only one existing, as very often loss of heterozygosity (LOH) is not found in such cerebral pathologies (Henske et al., 1996; Wolf et al., 1997; Niida et al., 2001, Ramesh, 2003). In a research performed in an animal model of TSC, the Eker rat, Mizuguchi et al. (2004) tested whether cytomegaly of the giant cells in the cortical tuber was caused by deletion of the normal TSC2 allele and resultant loss of heterozygosity. In giant cells isolated individually, semi-nested polymerase chain reaction demonstrated the presence of the wild-type TSC2 allele. Also, immunohistochemical evaluation detected positive tuberin immunoreactivity. It has also been shown that even in patients with total loss of immunoreactivity (loss of heterozygosity) for hamartin and tuberin, e.g. in renal angiomyolipomas, giant cells from cortical tubers and SEGAs show positive staining for both TSC-associated proteins (Mizuguchi et al., 2002). Jansen et al. (2004) performed mutation analysis in a tuberous sclerosis patient within a resected cortical tuber, finding no second-hit mutation. However, the authors observed that giant cells showed predominantly nuclear hamartin, while tuberin was located in cytosol. It was postulated that such an accumulation of hamartin and tuberin in separate cellular compartments of giant cells may prevent formation of the hamartin–tuberin complex, resulting in mTOR activation and increased S6 phosphorylation, which was found in this research. These data provide an alternative mechanism for tuberogenesis. Another explanation is that brain tubers consist of a mixture of different cell types, including giant cells, dysplastic neurons and glial cells. So far, it has not been elucidated which of these cells are of primary importance in tuber formation, and which are reactive. Uhlmann et al. (2002) addressed a question of the influence of TSC1 and/or TSC2 heterozygosity on in vivo astrocyte proliferation potential in mice. It has been determined that both TSC1 ^+/− and TSC2 ^+/− heterozygotes had 50% more astrocytes than wild-type animals. The authors studied also whether compound (TSC1 ^+/−, TSC2 ^+/−) heterozygosity could lead to synergistic growth advantage and found that in this case the number of astrocytes increased by ∼100%, showing that TSC1 and TSC2 genes act cooperatively to negatively regulate astrocyte growth. Interestingly, TSC2 ^+/−, p53 ^+/− compound heterozygosity did not increase astrocyte number, although heterozygosity of both genes studied separately increased the number of cells by 50%. These conclusions agree with the results of Lantuejoul et al. (1997), who did not observe abnormal p53 immunoreactivity in a TSC-associated tumor. When astrocytes derived from TSC2 ^+/− mice and wild-type animals were cultured in vitro, no difference in proliferation rate was discovered. Most importantly, TSC2 ^+/− astrocytes did not demonstrate anchorage-independent, or autonomous growth advantage.

Lack of second-hit mutation and no mutation at all found in about 10% of TSC patients may be explained by mosaicism or occurrence of two cell populations with different DNA in the same organism. Most organisms have mosaic cells that are limited to some specific tissues and do not exhibit pathogenic effects. On the other hand, some chromosome changes can only exist in a mosaic form, because in a non-mosaic form they are lethal. It has been found that high rate of mosaicism may be predicted for conditions with a high percentage of new mutations (Hall, 1988). In fact, mosaicism for TSC2 is well known and has been found in about 30% of index patients with combined TSC2-polycystic kidney disease syndrome (Sampson et al., 1997; Cheadle et al., 2000), as well as isolated TSC1 and TSC2 mutations (Kwiatkowska et al., 1999; Verhoef et al., 1999).

A paper by Roux et al. (2004) sheds some new light on the influence of cellular tuberin deficiency and its role in tuber formation. The authors found that a G protein Ras induces the phosphorylation of tuberin and, in cooperation with the nutrient-sensing pathway, regulates mTOR effectors, such as p70 ribosomal S6 kinase 1 (S6K1). The mitogen-activated protein kinase (MAPK)-activated kinase, p90 ribosomal S6 kinase (RSK) 1, was found to phosphorylate tuberin at a regulatory site, Ser-1798. This phosphorylation inhibits the tumor suppressor function of the tuberin–hamartin complex, resulting in increased mTOR signaling to S6K1. Thus, it is possible that in case of TSC2 heterozygosity, the intracellular pool of protein is inactivated, at least partially, by G proteins, which leads to mTOR activation. Moreover, it has been demonstrated that some mutations in TSC2 lead to loss of GAP activity, while the ability of tuberin to form a complex with hamartin remains intact. Such a mechanism has been shown in case of Rheb (Zhang et al., 2003) and could result in increased activity of MAP kinase-dependent pathway, while maintaining physiological role of TSC complex. We hypothesize that this mechanism could be very important for the pathogenesis of TSC. However, further studies are necessary to explain whether this finding could significantly influence the progress or appearance of the disease.