Zebrafish model of tuberous sclerosis complex reveals cell-autonomous and non-cell-autonomous functions of mutant tuberin

Tuberous sclerosis complex (TSC) is a genetic disease characterized by hamartomas in multiple organs, including the brain, skin, kidney, heart and lung (Crino et al., 2006). These focal lesions represent non-malignant collections of cells that have undergone abnormal differentiation. Neurological features are generally severe, with many patients suffering from intractable epilepsy, autism, behavioral problems and mental retardation (Ess, 2006). These important neurological features are generally accepted to be due to brain hamartomas (termed ‘tubers’) that represent severe cortical malformations. TSC results from loss of function of either the TSC1 (encoding hamartin) or TSC2 (encoding tuberin) genes. Although often due to a spontaneous mutation, TSC can be inherited as an autosomal dominant disorder. According to the prevailing model, patients with TSC have an initial mutation in one copy of either the TSC1 or TSC2 gene, and this mutation is either inherited from a parent or spontaneously acquired early in development. A subsequent ‘second hit’ mutation or deletion then occurs in focal areas of various organs, leading to the development of a hamartoma. This loss of heterozygosity (LOH) model has been repeatedly demonstrated in kidney and lung hamartomas from patients with TSC, but supporting data in the brain has been quite elusive (Henske et al., 1996). These findings have led to proposals of alternative pathways of disease progression, including haploinsufficiency, post-translational modification of the TSC gene products (Ma et al., 2005) and possible dominant-negative action of certain mutant alleles (Govindarajan et al., 2005).

The TSC1 and TSC2 genes were named after genetic linkage studies determined that there were two independent loci that could cause TSC. Their gene products are essentially unrelated, possessing sequence homology only in their coiled-coil domains that mediate protein-protein interactions. Indeed, compelling evidence gathered over the last several years shows that hamartin and tuberin bind to one another forming a complex that can then inhibit the G protein Rheb, an activator of the TOR (target of rapamycin) serine/threonine kinase (Inoki et al., 2003; Zhang et al., 2003). In mammals, mTOR (mammalian TOR) is found within multiprotein complexes termed mTORC1 (contains Raptor and is highly sensitive to rapamycin) or mTORC2 (contains Rictor and is relatively rapamycin insensitive) (reviewed in Huang and Manning, 2008). The hamartin-tuberin–Rheb–TOR pathway is highly conserved in Drosophila, yeast, mice and humans (Nobukini and Thomas, 2004; van Slegtenhorst et al., 2004), with the mTORC1 signaling pathway being constitutively upregulated in TSC patients (Crino et al., 2006). The identification of mTOR as a key signaling pathway regulated by hamartin-tuberin greatly aided research in the pathogenesis of TSC and its role during development. Active areas of current research are delineating the functions of mTOR within mTORC1 and mTORC2 (Huang and Manning, 2009) and how dysregulation of these complexes leads to human neurological disease. This burgeoning information coupled with established mouse and rat models of TSC has fuelled translational research, leading to the use of mTORC1 inhibitors as therapeutics for TSC (Bissler et al., 2008; Davies et al., 2008; Franz et al., 2006; Zhou et al., 2009).

Although mutations in either TSC1 or TSC2 are sufficient to cause dysregulation of mTOR, patients with TSC2 mutations often manifest more severe disease than those with TSC1 mutations, suggesting that there are additional functions of tuberin that are currently unknown (Au et al., 2007). Multiple rodent models of TSC have been developed to study Tsc1 and Tsc2 gene function. Although informative, conventional homozygous mouse knockouts of either Tsc1 or Tsc2 are lethal by embryonic day 12 (Kobayashi et al., 1999; Kobayashi et al., 2001; Onda et al., 1999). Such studies shed only limited light on the pathogenesis of brain hamartomas in TSC because these homozygous mutant mice die prior to any substantive stages of cortical development. Mice that are heterozygous for Tsc1 or Tsc2 mutations develop kidney pathology by 6–12 months of age but exhibit only minimal brain pathology (Onda et al., 1999; Uhlmann et al., 2002a). Similar results were seen for the Eker rat, a long-studied model of kidney disease that is due to an insertional mutation within the rat Tsc2 gene (Kobayashi et al., 1995). Comparable to the situation in mice, homozygous Tsc2-deficient Eker rats manifest an early embryonic-lethal phenotype, whereas heterozygous animals have minimal CNS manifestations (Tschuluun et al., 2007). To circumvent these limitations and study abnormalities in postnatal mice, multiple conditional-knockout models of TSC have been established by inactivating the Tsc1 or Tsc2 genes in specific cell types, including in neurons and astrocytes (Uhlmann et al., 2002b; Way et al., 2009; Wong et al., 2003). Although these approaches have better defined the role of specific CNS lineages in TSC, they have not adequately modeled tuber formation in the brain.

Zebrafish are a compelling model in which to study the role of the TSC1 and TSC2 genes as well as the pathogenesis of TSC. This is owing to the many experimental advantages offered by zebrafish, including the ability to observe directly embryonic development, ease of cell transplantation and the many genetic tools available to identify specific populations of CNS cell types (Baraban et al., 2007; Cooper et al., 2005). Use of zebrafish to study TSC has been limited to date to two characterized TSC1 homologs (DiBella et al., 2009). In this study, the authors used antisense morpholino oligonucleotides to knock down function of the zebrafish tsc1a gene, and found a role in the control of cilia length and kidney development. However, these experiments did not address whether loss of zebrafish tsc1 causes abnormalities of cell size or its impact on brain development.

To define the role of tuberin during development, we cloned and characterized the expression of the zebrafish tsc2 gene. We then used TILLING (targeting induced local lesions in genomes; supplementary material Fig. S1) (Till et al., 2003) to identify zebrafish that harbor a tsc2 nonsense mutation. This tsc2^vu242 mutant allele encodes a truncated form of Tuberin, lacking the highly conserved GTPase-activating protein (GAP) domain that mediates Rheb inactivation and TORC1 inhibition. Homozygous tsc2^vu242/vu242 mutants died during early larval stages, manifesting multi-organ pathology with increased cell size and ectopically positioned cells within the forebrain. Greatly increased levels of TORC1 were seen in mutant embryos, and treatment of mutant larvae with rapamycin, a potent TORC1 inhibitor, reversed the defects in cell size, indicating that the role of Tuberin in the TOR pathway is highly conserved in zebrafish. Although heterozygous tsc2^vu242/+ fish are viable and morphologically normal, they showed moderate increases in TORC1 signaling. This finding and additional evidence suggests that a threshold of increased TOR signaling is required to cause pathology. Finally, transplantation of mutant cells to wild-type hosts revealed both cell-autonomous and non-cell-autonomous forebrain abnormalities. Overall, our studies reveal that the Tsc2-TOR pathway is highly conserved in zebrafish and demonstrate the utility of zebrafish for defining pathological mechanisms that lead to complex multi-organ disorders such as TSC.

To better understand the role of TSC2 in normal development and disease, we first cloned full-length zebrafish tsc2 cDNA. The predicted protein overall shows 60% identity and 73% similarity to human tuberin (supplementary material Fig. S2). The hamartin-interacting, tuberin and GAP domains are all highly conserved, with 71%, 70% and 75% identical amino acids between the two species, respectively (Fig. 1M). Analysis of the spatiotemporal expression pattern by whole-mount in situ hybridization revealed that tsc2 transcripts are maternally deposited (Fig. 1A) and ubiquitously expressed during blastula, gastrula and segmentation stages (Fig. 1B-D). By 25 hours post-fertilization (hpf), tsc2 expression is highly enriched in the developing eye, forebrain, midbrain, lateral hindbrain region and otic vesicles (Fig. 1E-G), as well as in the blood vessels (Fig. 1H). At 36 hpf, strong expression persists in the brain (Fig. 1I) and transverse sections show that tsc2 is also expressed in the ventrolateral region of the hindbrain (Fig. 1J). At 3 days post-fertilization (dpf), we detected low levels of tsc2 expression in the eye (Fig. 1K), heart and intestine, in addition to the continued higher levels in the brain (Fig. 1L).

To study loss of tsc2 function in zebrafish, we employed TILLING to identify N-ethyl-N-nitrosourea (ENU)-induced mutations in the tsc2 gene (supplementary material Fig. S1). By screening 4608 ENU-mutagenized F1 zebrafish (Mullins et al., 1994; Solnica-Krezel et al., 1994), we found one fish heterozygous for C to A transversion, predicted to change the normal tyrosine (Y) residue at amino acid 1031 to a premature stop codon (Fig. 1M). Because this tsc2^vu242 nonsense mutation is predicted to remove the GAP domain but retain the Hamartin-binding domain, we hypothesized that the truncated protein will have either null or possibly dominant-negative activity (Govindarajan et al., 2005). Because the C-terminus of Tuberin contains the epitope for all currently available anti-Tuberin antibodies, one consequence of the tsc2^vu242 nonsense mutation is an inability to perform immunofluorescence and western blotting experiments to determine expression levels and localization of Tuberin protein. However, we confirmed the existence of mutant Tuberin by using an antibody directed against phospho-Tuberin (Ser939), because the mutant Tuberin retains this N-terminal epitope. Notably, the expression levels of phospho-Tuberin (Ser939) were not appreciably changed between wild-type and heterozygous or homozygous mutant fish at 32 hpf (supplementary material Fig. S3). To more precisely define Tuberin stability in wild-type and heterozygous or homozygous mutant fish, new antibodies that specifically recognize the N-terminus of zebrafish Tuberin will need to be generated.

tsc2^vu242/vu242 homozygous mutant embryos and subsequent larvae did not develop overt morphological defects until 5 dpf, although a few homozygous mutants did not fully inflate their swim bladder. The most prominent phenotype was then manifest by 7 dpf, when tsc2^vu242/vu242 mutant zebrafish exhibited completely deflated swim bladders and enlarged livers (Fig. 2C,D,L). All mutant embryos died by 11 dpf, possibly because of limited food intake that might be due in part to the swim bladder abnormalities or abnormal brain morphology and/or function (see below). tsc2^vu242/+ heterozygous fish are viable, develop into fertile adults and have no discernable phenotype. Therefore, tsc2^vu242 seems to act as a recessive embryonic-lethal mutation in zebrafish.

To determine whether tsc2^vu242 mutation causes abnormalities of TOR kinase signaling, we evaluated the phosphorylation status of S6 ribosomal protein and 4E-BP1 because these proteins are well-established downstream effectors of the mTOR kinase within the mTORC1 complex (Inoki et al., 2002). Immunohistochemistry on 7-dpf embryos using antibodies specific to phospho-S6 and phospho-4E-BP1 showed that the tsc2^vu242/vu242 mutant embryos exhibited highly increased levels of both phosphoproteins in several organs, including in cells in the eye, brain and spinal cord (Fig. 2E-J), liver (Fig. 2L), intestine (Fig. 2L,P), and kidney (Fig. 2P), as compared with wild-type siblings (Fig. 2A,B,E-G,K,M,N). The elevated TORC1 activity was suppressed by injection of wild-type zebrafish tsc2 and at least partially suppressed with wild-type human TSC2 RNA (supplementary material Fig. S4). We next performed western blotting to evaluate levels of activation of the TORC1 pathway in protein extracts from tsc2^vu242/vu242 homozygous, tsc2^vu242/+ heterozygous and wild-type embryos at 7 dpf (Fig. 2Q). Consistent with immunohistochemistry results, phospho-4E-BP1 and phospho-S6 levels were highly elevated in tsc2^vu242/vu242 homozygotes. These results indicate that activation of the TORC1 pathway as a consequence of decreased tsc2 function is conserved in zebrafish. We also detected a moderate increase of TORC1 signaling (phospho-S6 and phospho-4E-BP1) in heterozygous larvae when compared with wild-type siblings (Fig. 2Q). Notably, tsc2^vu242/+ heterozygotes showed no overt morphological defects, including cell size, despite this enhancement of TORC1 activation, implying a threshold above which TOR signaling can affect organ development. This increased TORC1 activity in tsc2^vu242/+ heterozygotes is consistent with a mechanism of haploinsufficiency of the remaining wild-type allele. This has been previously demonstrated in mouse models of TSC (Ehninger et al., 2008).

To test whether the tsc2^vu242 nonsense mutation can interfere with TORC1 signaling, we overexpressed full-length or truncated Tuberin by injecting synthetic wild-type tsc2 or tsc2^vu242 RNA, or an additional truncated-Tuberin-encoding RNA, into wild-type embryos at the one-cell stage. We observed enlarged cells in the superficial enveloping layer in embryos injected with 400 pg of tsc2 RNA containing the tsc2^vu242 nonsense mutation, and these cells had increased levels of phospho-S6 during early development (10–11 hpf) (Fig. 3C,G) compared with wild type. A similar phenotype was also generated by overexpressing a similar truncated form of Tuberin, made from a NruI deletion construct of wild-type tsc2 (Fig. 3D,H). By contrast, embryos injected with 300 pg of wild-type tsc2 RNA did not show such abnormalities (Fig. 3B,F). These results suggested that truncated Tuberin generated by the tsc2^vu242 nonsense mutation or a deletion construct might interfere with the function of endogenous Tuberin. By 27 hpf, abnormal brain morphology with a dorsally expanded hindbrain was observed in embryos injected with RNA encoding the truncated forms of Tuberin (Fig. 3K,L) but not in those injected with wild-type tsc2 RNA (Fig. 3J). In addition, we observed slightly decreased levels of tsc2 transcripts in heterozygous fish and a marked decrease in homozygous mutants at 26 hpf (supplementary material Fig. S5E,F). Given the position of the premature stop codon, this is probably caused by nonsense-mediated mRNA decay (NMD) (Silva and Romao, 2009). Although levels of mutant Tuberin could then decrease over time compared with wild-type Tuberin, the amount of phospho-Tuberin (Ser939) protein (supplementary material Fig. S3) was relatively stable in mutant larvae at 26 hpf.

The enlarged liver seen in tsc2^vu242/vu242 mutant larvae (Fig. 2C,D) could result from either increased numbers or size of hepatocytes. Using BrdU-incorporation experiments, we did not detect increased cell proliferation in tsc2^vu242/vu242 mutants compared to wild-type embryos by 7 dpf (data not shown). To measure cell size, we crossed tsc2^vu242/+ carriers with fish harboring the Tg(β-actin:mGFP) transgene, which expresses green fluorescent protein (GFP) at the cell membrane in all tissues under the control of the β-actin promoter (Cooper et al., 2005). We crossed the resulting tsc2^vu242/+;Tg(β-actin:mGFP) fish and determined cell size in tsc2^vu242/vu242 homozygous mutants in various organs at 9 dpf. These analyses revealed increased size of hepatocytes by approximately twofold in tsc2^vu242/vu242 homozygous mutants as compared with wild-type siblings (Fig. 4A,B,G).

We also observed that the spinal cord in tsc2^vu242/vu242 mutants at 7.5 dpf was clearly larger than that in wild-type siblings (Fig. 4E,F,I). To clarify again whether the spinal cord phenotype resulted from enlarged neuronal cells bodies or an increased number of neurons, we measured cell area (Fig. 4H) and counted the number of cell nuclei in the spinal cord (Fig. 4J) of Tg(β-actin:mGFP) transgenic wild-type and mutant embryos. These analyses revealed that neuronal cell bodies in the brain and spinal cord of tsc2^vu242/vu242 mutants were approximately 1.5-times larger than wild-type neurons (Fig. 4C-F,H). However, there was no change in the number of spinal cord neurons (Fig. 4J). These results indicate that tsc2 mediates cell size rather than cell proliferation in the liver and spinal cord during embryogenesis.

To determine what aspects of the tsc2^vu242/vu242 mutant phenotype are due to excess TORC1 signaling, we employed rapamycin, a specific inhibitor of mTORC1 (Huang and Manning, 2008). In zebrafish, rapamycin treatment before 3 dpf causes a general developmental delay and seems to specifically interfere with intestinal development (Makky et al., 2007). To avoid such adverse effects, we started rapamycin treatment at 3.5 dpf using rapamycin concentrations ranging from 5–1000 nM. After 24 hours of rapamycin treatment, embryos were genotyped and transverse trunk sections were examined for TORC1 signaling by determining levels of phospho-S6 using immunofluorescence. We observed a dose-dependent reduction of phospho-S6 levels in the mutant embryos (Fig. 4K). After 4 days of rapamycin treatment, the phenotype of enlarged hepatocytes seen in 7-dpf tsc2^vu242/vu242 mutants was also suppressed in a dose-dependent manner, with normalization of hepatocyte size seen with 50 nM of rapamycin (Fig. 4L,M). Statistical analysis confirmed significant changes in each treatment condition when compared with wild-type and untreated tsc2^vu242/vu242-mutant control fish (Fig. 4M). Together these observations indicate the increased size of hepatocytes and liver in tsc2^vu242/vu242 zebrafish mutants are largely due to the elevated level of TORC1 activity during early larval stages.

Almost all patients with TSC have structural brain lesions that include cortical tubers, white matter abnormalities, subependymal nodules and subependymal giant cell astrocytomas (Bourneville, 1880; Mizuguchi, 2001). To investigate whether the forebrain architecture in zebrafish is affected by the inactivation of tsc2, we collected coronal sections of wild-type and tsc2^vu242/vu242 homozygous mutant larvae at 7.5 dpf (Fig. 5). In the wild-type forebrain, the structure of pallium and the dorsal division of subpallium were well developed with a clear boundary between the gray and white matter (Fig. 5A,B; n=14/15). The telencephalic ventricle was also clearly defined (Fig. 5A). By contrast, the corresponding brain regions in the tsc2^vu242/vu242 mutants showed poorly defined midline structures of the gray matter and telencephalic ventricles (Fig. 5E). The pallium and dorsal subpallium were also disorganized and less compact in tsc2^vu242/vu242 mutants compared with wild type (Fig. 5E; n=18/18), reminiscent of the abnormal cortical lamination seen in patients with TSC (Crino, 2004). Moreover, the white matter between pallium and dorsal subpallium was disrupted by numerous scattered neuronal cell bodies, suggestive of abnormal neuronal migration (Fig. 5E,F). In zebrafish, this region is homologous to the dorsal mammalian brain at the interface between cortex and subcortical white matter (Mueller and Wullimann, 2009). The dorsal subpallium was also wider in the mutants but the average number of cells in this region was not significantly changed (data not shown), suggesting that this tissue was poorly organized and less compact than in the corresponding wild-type dorsal subpallium. The most anterior part of the diencephalon was also disorganized, including the pallium, eminentia thalami and the lateral forebrain bundle area (Fig. 5C,D,G,H). Finally, we analyzed more-posterior structures of the mutant brain. We did not detect any significant defects in the morphology of the posterior diencephalon extending to the hindbrain region of tsc2^vu242/vu242 mutants relative to wild-type siblings (supplementary material Fig. S6). These observations indicate that the anterior part of the forebrain was preferentially affected in tsc2^vu242/vu242 mutants compared with posterior brain regions.

The disruption of gray and white matter organization that we found in the pallium, subpallium and thalamic regions could be caused by intrinsic loss of Tsc2, but the heterogeneity of human tubers suggests that additional mechanisms might be possible. To address this issue, we made mosaic zebrafish by transplanting tsc2^vu242/vu242 cells into wild-type host embryos (schematic in Fig. 6A). Donor cells were obtained from a tsc2^vu242/+;Tg(h2afv:GFP) transgenic line, in which all cell nuclei express GFP (Pauls et al., 2001). Tsc2^vu242/+;Tg(h2afv:GFP) fish were intercrossed and their progenies were used as donor embryos in the transplantation experiment (Fig. 6B-N). During transplantations at the late blastula stage (4 hpf), donor cells were taken from the animal pole region and transplanted to the same region because most of the cells in this region normally give rise to forebrain and retina (Woo and Fraser, 1995). Analyses of chimeric embryos at 7.5 dpf revealed that tsc2^vu242/vu242;Tg(h2afv:GFP) mutant cells, marked by GFP expression in the nuclei, showed a strongly increased level of TORC1 activity, as indicated by increased levels of phospho-S6 (Fig. 6F-L; n=20/20). We also observed a marked disruption of gray-white matter borders in a subset of the chimeric embryos (Fig. 6I, rectangle, enlarged in J–N; n=3/18), similar to the phenotype observed in tsc2^vu242/vu242 mutants (Fig. 5E-H). Strikingly, we also noted several wild-type host cell bodies (GFP negative), which, as expected, showed no increase in phospho-S6 staining, that were surrounded by mutant cells within the lateral forebrain bundle (Fig. 6J-N). By contrast, wild-type and tsc2^vu242/+ cells transplanted into wild-type hosts did not have increased TORC1 signaling and did not cause any brain abnormalities (Fig. 6B-E; n=6/6). These results suggest that tsc2^vu242/vu242 homozygous mutant cells activate TORC1 in a cell-autonomous manner but also in a non-cell-autonomous manner, which can cause wild-type cells to be abnormally positioned within the white matter. Because such abnormalities are seen in the brains from patients with TSC (Park et al., 1997; Shepherd et al., 1995), this observation further supports the intriguing possibility that LOH might not be absolutely required in all cells for the formation of brain hamartomas (tubers) in TSC (Talos et al., 2008).

Given the predominance of rodent models of human disease and the marked difficulty in performing transplantation experiments in these models, it remains unknown whether Tsc1 or Tsc2 mutant cells can produce the focal brain hamartomas that are almost universally seen in patients with TSC. To address whether tsc2^vu242/vu242 mutant cells can produce ‘tuber-like’ abnormalities, we transplanted membrane GFP-tagged tsc2^+/+;Tg(β-actin:mGFP) and tsc2^vu242/vu242;Tg(β-actin:mGFP) donor cells into wild-type host embryos at the blastula stage. The resulting chimeras were raised to adulthood and sacrificed 1 year after transplantation. We analyzed transverse sections of the chimeric brains with respect to TORC1 activation and tissue architecture. These analyses revealed that the transplanted wild-type Tg(β-actin:mGFP) cells exhibited normal levels of TORC1 pathway activity and did not cause any obvious defects in the host brain (Fig. 7A,B). By contrast, transplanted tsc2^vu242/vu242;Tg(β-actin:mGFP) cells showed elevated levels of TORC1 activity compared with the surrounding wild-type cells (Fig. 7C,F). Moreover, we noted clusters of transplanted mutant cells in the gray-white matter boundary (Fig. 7G,H) and within the gray matter in a series of sections (Fig. 7C,D). These GFP-positive cell clusters were detected through 100 μm of contiguous sections (data not shown), suggesting that they formed an extensive cluster. There seemed to be at least two types of transplanted cells in each cell cluster: relatively small and rounded cells were present (Fig. 7E) but there were also elongated and enlarged cells with dendrites, suggestive of astrocytes and/or glial cells (Fig. 7G). We were unable to confirm the lineage of these intriguing cells because there are few antibodies currently available in zebrafish that are specific to glial cell populations. These brain lesions with mixed cell types and increased TORC1 signaling suggest the formation of brain hamartomas. Further examples of these lesions in the thalamic region are notable for evidence of further disruption of brain architecture (Fig. 7I-M). These mosaic experiments indicate that tsc2^vu242/vu242 mutant cells can produce local lesions with increased TORC1 signaling as well as more-extensive malformations in the adult wild-type brain.

TSC is a multifaceted human disease that is an important model system for the study of brain malformations as well as the pathogenesis of epilepsy and autism. Advances in these fields require relevant animal models to test hypotheses related to TSC1 and TSC2 gene function during development and homeostasis. Mouse models have been employed to great effect (Meikle et al., 2005; Meikle et al., 2008; Way et al., 2009; Wong, 2008), but fundamental issues concerning TSC1 and TSC2 function remain unanswered. In particular, the striking neurological aspects of TSC necessitate tractable model systems in which neuronal and glia differentiation can be closely examined and manipulated. To this end, we targeted the zebrafish tsc2 gene, given the unique experimental advantages of this model organism and the marked conservation of tuberin at the amino acid level (supplementary material Fig. S2). The tsc2^vu242 mutant allele described here encodes a truncated Tuberin, lacking the GAP domain, which mediates TORC1 inhibition. The resultant tsc2^vu242/vu242 mutants died during early larval stages, manifesting multi-organ pathology, including increased cell size in various tissues. Greatly elevated levels of TOR kinase activity were seen in tsc2^vu242/vu242 mutants, whereas only a moderate increase was observed in tsc2^vu242/+ heterozygous embryos. These heterozygous fish seem to have a normal lifespan and have a morphologically normal appearance. This suggests that there is a threshold above which increased TOR signaling causes organ pathology.

The hamartin-tuberin complex normally inhibits mammalian TORC1 activity, which controls many cellular processes, including cell growth, proliferation and differentiation (for reviews, see Huang and Manning, 2008; Huang and Manning, 2009). We found that the GAP domain of zebrafish Tuberin is also required for inhibition of zebrafish TORC1. This substantiates the concept that TOR function is highly conserved and that upstream and downstream aspects of this pathway can be effectively studied using zebrafish. One potential problem is that many genes in zebrafish have been duplicated during evolution. This can confound analyses, particularly for loss-of-function experiments. This is seen, for example, with zebrafish tsc1, for which two paralogs, tsc1a and tsc1b have been described (DiBella et al., 2009). However, the zebrafish tsc2 gene does not seem to have a paralog, facilitating loss-of-function analyses as described here.

We found that the expression of tsc2 in zebrafish is quite similar to that seen in mammals, with an early and widespread expression pattern that becomes more restricted and maximal in the central nervous system. One striking difference in zebrafish, however, is the apparent deposition of maternal tsc2 mRNA in fertilized eggs, which is usually minimal in mammals (reviewed in Tadros and Lipshitz, 2009). This might explain why zebrafish that are homozygous for the tsc2^vu242 allele can survive to approximately 11 dpf and are able to attain relatively advanced developmental stages compared with that seen for Tsc2-deficient mice and rats (Kobayashi et al., 1999; Kobayashi et al., 2001; Onda et al., 1999). A maternal contribution of tsc1 or tsc2 mRNA might further explain why ‘knock-down’ experiments directed against tsc1 using translation-blocking antisense morpholino oligonucleotides were associated with an earlier onset of and more-severe phenotypes (DiBella et al., 2009) (and our unpublished data). Alternatively, unlike in mammals, the tsc2 gene might not be absolutely required during early zebrafish development. Definitively addressing this issue will require the generation of zebrafish embryos lacking expression of both maternal and zygotic tsc2 genes.

On the basis of pathological data from patients with TSC, crucial aspects of any relevant model of TSC include elevated TOR (mTOR) activity and increased cell size within mutant organisms. This second feature should then be reversible by treatment with TORC1 (mTORC1) inhibitors such as rapamycin (Meikle et al., 2008; Zeng et al., 2008). Our zebrafish tsc2^vu242/vu242 embryos fulfill both these criteria, with dysregulated TORC1 signaling, as evidenced by increased levels of phospho-S6 and phospho-4E-BP1. Although this is seen in various organs throughout the mutant embryos, the greatest upregulation was detected in the brain and liver. Additional evidence for increased TORC1 signaling is provided by the enlarged size of neurons and hepatocytes in tsc2^vu242/vu242 zebrafish. This increased cell size was normalized by treatment with rapamycin at low concentrations, indicating TORC1 dependence (Fig. 4). These findings underscore the striking conservation of tsc2 sequence, as well as function, throughout evolution. In addition, these findings highlight the utility of zebrafish for the study of TSC and other human disorders resulting from dysregulation of the TOR kinase. It should be noted that not all cells of the tsc2^vu242/vu242 mutant zebrafish are increased in size. Whereas hepatocytes and spinal cord neurons, for example, are clearly enlarged, other cell types were unchanged. This might be explained by a differential cell-autonomous effect from increased TORC1 signaling and expression of other components of TORC1. Moreover, despite some enlarged organs, the overall body size of the mutant larvae does not seem to be increased, again suggesting variable requirements for Tuberin in distinct cells and tissues during zebrafish development.

Despite recent progress and the development of several rodent models of TSC, many important questions still remain about the role of TSC1 and TSC2 genes during development. In particular, the pathogenesis of tubers remains poorly understood. These brain malformations are almost universal manifestations in patients with TSC and probably underlie the severe neurological features of epilepsy and autism that are seen in these patients (Crino, 2004; Crino et al., 2006). In this study, we demonstrate that zebrafish that are homozygous for the tsc2^vu242 allele have strikingly abnormal brain development with disorganization of the gray and white matter. These results are quite reminiscent of the cortical defects seen within human cortical tubers. It will be important in future studies to delineate the cellular basis of the forebrain abnormalities in tsc2^vu242/vu242 mutants and the relevance of these mechanisms to tubers seen in human TSC patients. In particular, it will be imperative to investigate whether ectopically positioned neurons within mutant forebrains arise via abnormal migration and/or abnormal differentiation. The transparent nature of zebrafish embryos coupled with the availability of multiple fluorescent-tagged proteins will greatly facilitate such analyses (Garcia-Lecea et al., 2008; Tsai et al., 2006).

Whereas fish that are homozygous for the tsc2^vu242 allele have overt multi-organ pathology, heterozygous fish are phenotypically normal. This broadly supports the prevailing model of TSC pathogenesis that an initial germline mutation in TSC1 or TSC2 is followed by a subsequent ‘second-hit’ in the remaining allele of either gene. This mechanism is consistent with an apparent requirement for homozygous inactivation of Tsc1 or Tsc2 in mouse models of TSC (Kobayashi et al., 1999; Onda et al., 1999). Mice heterozygous for Tsc1 or Tsc2 seem to undergo normal brain development without any abnormalities of neuronal layering or other structural lesions (Onda et al., 1999; Uhlmann et al., 2002a). However, an alternative mechanism to consider is haploinsufficiency, where the presence of only one functional copy of either TSC1 or TSC2 results in the pathological and clinical features of TSC. Broadly consistent with this notion, a moderate increase in mTORC1 signaling is seen in the hippocampus of Tsc2 heterozygous mice, which have structurally normal brains. These animals were also shown to have deficits in hippocampus-dependent learning tasks, and these deficits were suppressed by rapamycin treatment (Ehninger et al., 2008). Whereas zebrafish heterozygous for the tsc2^vu242 allele seem to undergo normal development and do not have enlarged cell size, they manifest moderately increased TORC1 signaling (Fig. 2; supplementary material Fig. S5). We reasoned that tsc2^vu242/+ heterozygous zebrafish might have decreased resistance to induced endoplasmic reticulum (ER) stress compared with wild-type zebrafish. Accordingly, we found that treatment with low doses of tunicamycin (to pharmacologically induce ER stress) significantly activated TORC1 signaling in tsc2^vu242/+ but not wild-type zebrafish (supplementary material Fig. S7). These results further support the concept that haploinsufficiency of tsc2 when combined with physiological insults can lead to the elevation of TOR signaling past the threshold to trigger pathological changes. However, a dominant-negative activity of the truncated Tuberin cannot be excluded at this time.

It is unknown how directly relevant any results from tsc2^vu242 mutant zebrafish are for patients with TSC. A subset of TSC patients clearly harbor TSC2 mutations that inactivate or remove the GAP domain but retain the hamartin-binding domain (Sancak et al., 2005). Such alleles have not yet been examined for functional consequences in animal models. This reflects the technical difficulty in studying human genetic diseases in transgenic mouse models and underscores the need for additional model systems such as zebrafish.

tsc2^vu242/vu242 mutant cells exhibit a cell-autonomous increase of TOR1 activity when transplanted into wild-type zebrafish. However, we also observed that transplanted tsc2^vu242/vu242 mutant cells were able, in a non-cell-autonomous manner, to cause mislocalization of wild-type cells in the forebrain. This result is particularly interesting because it suggests that some cells that contribute to tuber formation in the brains of patients with TSC might retain a wild-type TSC2 allele. The possible mechanism is unknown but might reasonably relate to factors secreted from mutant cells that alter the differentiation, migration or function of wild-type cells in the brain.

This new zebrafish model is well situated to begin to address many important questions in TSC and TOR biology and disease pathogenesis, and should complement studies that use rodent models. The conserved sequence and expression of zebrafish tsc2 coupled with the multi-organ pathology seen in tsc2^vu242/vu242 homozygous zebrafish and their response to treatment with rapamycin further illustrate the utility of this model of TSC. Many questions remain about TSC pathogenesis and the role of the TSC2 gene during brain development. The availability of this zebrafish model should facilitate new discoveries and catalyze the identification and testing of new therapies for neurological and non-neurological aspects of TSC.

Zebrafish strains used in this study included AB*, tsc2^vu242, Tg(β-actin:mGFP) (Cooper et al., 2005) and Tg(h2afv:GFP) (Pauls et al., 2001). Embryos were obtained from natural matings and raised at 28.5°C in egg water (0.3 g of sea salts/l).

A nonsense mutation in the tsc2 gene was isolated by screening 4608 F1 progeny of ENU-treated males using TILLING (supplementary material Fig. S1) (Till et al., 2003). A part of the tail was used for the genotyping of individual embryos after fixation in 4% paraformaldehyde. tsc2^vu242 was maintained in the AB* genetic background. The mutation introduces a C3087 to A transversion, which abolishes an HpyCh4IV restriction-enzyme recognition site. For genotyping of tsc2^vu242, we amplified a 151-bp fragment by PCR using the forward primer 5′-CCAGCACCACCTGCAGTCTGG-3′ and reverse primer 5′-CTCTTGGGCAGAGCAGAGAAGTTGG-3′ flanking the mutation site. Mutation was confirmed by absence of the HpyCh4IV restriction site.

Synthetic RNA was prepared, using mMessage mMachine kit (Ambion), from wild-type tsc2, mutant tsc2^vu242 and a truncated form of tsc2 generated by NruI restriction-enzyme digestion. Capped mRNAs were diluted in 0.1 M KCl solution containing 0.5% Phenol Red. 200–400 pg aliquots of RNA were injected into one-cell-stage embryos using standard protocols.

3 ng of antisense morpholino oligonucleotide was injected into one-cell-stage embryos to verify antibody specificity of phospho-Tuberin (Ser939). MO-tsc2 (5′-ACTCTTTACTGGGCTGTTTATTCAT-3′; GeneTools) was designed to specifically target the ATG start codon and inhibit translation of tsc2.

Embryos were fixed in 4% paraformaldehyde overnight and dehydrated in 100% methanol at −20°C. Whole-mount hybridization was performed using standard protocols (Jowett and Lettice, 1994). BCIP/NBT (Vector Laboratories) mixture was used as a chromogenic substrate. In situ images were acquired using a Zeiss Axioscope and Nikon COOLPIX 995 digital camera.

Embryos were fixed in 4% paraformaldehyde from overnight to 2 days at 4°C. Fixed embryos were embedded in 1.2% agarose/5% sucrose and saturated in 30% sucrose. Tissue blocks were frozen in 2-methylbutane chilled using liquid nitrogen. 10-μm sections were collected on microscope slides using a Leica cryostat. Sections were kept in −80°C before use. Sections were rehydrated in 1× PBS for 20 minutes at room temperature and blocked in 5% sheep serum/1× PBS for 1 hour. Sections were incubated with antibodies to phospho-S6 ribosomal protein (Cell Signaling #2215; Ser235/236; dilution 1:300), phospho-4E-BP1 (Cell Signaling #2855; Thr37/46; dilution 1:300) and phospho-Tuberin (Cell Signaling #3615; Ser939; dilution 1:300) overnight at 4°C, rinsed for 1 hour with PBS and then incubated overnight at 4°C with Cy3-labeled goat anti-rabbit secondary antibody (Jackson ImmunoResearch). Sections were then washed with 1× PBS for 1 hour and mounted in VECTASHIELD with DAPI (Vector Laboratories). Images were acquired using a Zeiss Axio Imager Z1 and Zeiss AxioCam MRm digital camera. Digital images were then processed using Adobe Photoshop CS2. All images received only minor modifications, with control and mutant sections always processed in parallel.

To compare cell size, images of cross sections were obtained and the outline of each cell was drawn manually using ImageJ. The number of pixels where then measured in each cell. For neuronal cell counting, sections were stained with DAPI, images of cross-sections obtained from 12 different embryos and cells were manually counted in Adobe Photoshop. Statistical analyses were performed using Student’s t-test.

Tunicamycin (Sigma) was dissolved in dimethyl sulfoxide (DMSO) as 1 mg/ml concentration and diluted with egg water at a final concentration of 200 ng/ml and 400 ng/ml. Tunicamycin treatment started at 7 dpf and continued until 8.5 dpf.

A stock solution of rapamycin (30 mg/ml in 100% ethanol; LC Laboratories) was diluted with egg water. Rapamycin treatment was initiated between 3.5 dpf and continued until 7 dpf depending on the specific experiment. Rapamycin-containing water was replaced every 24 hours.

A total of 20 embryos at 7 dpf were lysed by adding RIPA cell lysis buffer (18 μl per embryo of 10 mM Tris, pH 7.4, 150 mM NaCl, 1 mM EDTA, 0.1% SDS, 1% Triton X-100, 1% sodium deoxycholate) and pre-heated (100°C) in 4× Laemmli SDS sample reducing buffer (6 μl per embryo) (1×: 250 mM Tris-HCl, pH 6.8, 40% glycerol, 8% SDS, 20% β-mercaptoethanol, 0.004% Bromophenol Blue), then the tissue was macerated until suspended. For phospho-S6, lysates equaling 1.6 embryos were loaded per lane in a Tris-HCl 4–15% polyacrylamide gel. For phospho-4E-BP1, lysates equaling four embryos were loaded per lane in a Tris-HCl 18% polyacrylamide gel. The gels were run using the Bio-Rad Criterion system according to the manufacturer’s instructions. The gels were transferred onto polyvinylidene fluoride (PVDF) membranes. Subsequent blocking, antibody incubation and film exposure were performed according to the manufacturer’s recommendations. The antibodies used included anti-phospho-S6 ribosomal protein (Ser235/236; #4856, 2F9, Cell Signaling Technology; 1:1000), anti-4E-BP1 (#9644, 53H11, Cell Signaling Technology; 1:1000) and anti-β-actin as a loading control (#A5441, Sigma-Aldrich; 1:2000).

We used tsc2^vu242;Tg(β-actin:mGFP) and tsc^vu242;Tg(h2afv:GFP) lines for chimeric analysis by transplantation. 30–50 blastomeres from donors obtained by crossing tsc2^vu242;Tg(β-actin:mGFP) or tsc2^vu242;Tg(h2afv:GFP) fish were transplanted at 4 hpf into same-stage wild-type hosts. Host embryos were raised at 28.5°C. 1-year-old adult chimeric fish were euthanized by 20% 3-aminobenzoic acid ethyl ester (MESAB; Sigma); brains were then fixed in 4% PFA for analysis. Donor embryos expressing GFP-positive cells were fixed at 1 dpf and genotyped for tsc2^vu242 mutation. Host embryos and adult brains were mounted in 1.2% agarose/5% sucrose as described above. After determination of donor genotypes, 10-μm transverse sections of fixed chimeric embryos were collected using a Leica CM1900 cryostat microtome. Sections were stained with the anti-phospho-S6 antibody and mounted in VECTASHIELD with DAPI as described above and imaged using Zeiss Axio Imager Z1 or Zeiss AxioCam MRm digital cameras.

We thank L.S.-K. group members for discussions and for critically reading the manuscript. We thank our fish facility staff for excellent care. This work in the L.S.-K. lab was supported by the Zebrafish Initiative – Vanderbilt University Academic Capital Venture Fund and Martha Rivers Ingram Endowed Chair. K.C.E. was supported by the NINDS, NIH and the Tuberous Sclerosis Alliance.