The tubby-like proteins, a family with roles in neuronal development and function

Currently, four tubby gene family members (Tub, Tulp1, Tulp2 and Tulp3) (Kleyn et al., 1996; Noben-Trauth et al., 1996; North et al., 1997; Nishina et al., 1998) have been identified in humans and mice. These four genes probably comprise the entire vertebrate family because only one gene has been found in Drosophila (GenBank accession number: AE003453) and C. elegans (AF143297), which is consistent with the proposed two genome duplications that occurred during vertebrate evolution (Gibson and Spring, 2000). Tubby-like genes are also found in other organisms (i.e. Gallus (AJ396643), Xenopus (AW638855), Zea Mays (AI737314) and Arabidopsis (AC005309)). Plants in particular appear to harbor a large number of tubby-like proteins (TULPs). The strong conservation of Tub across members (Fig. 1) and through evolution suggests that the TULPs carry out some basic function in cells in which they are expressed. This suggestion is strengthened by the occurrence of disease phenotypes that result from mutations within tub and Tulp1 (Kleyn et al., 1996; Noben-Trauth et al., 1996; Heckenlively et al., 1995; Ohlemiller et al., 1997; Hagstrom et al., 1998; Banerjee et al., 1998; Lewis et al., 1999; Paloma et al., 2000) and by the embryonic lethality of mice homozygous for a disrupted Tulp3 allele (Ikeda et al., 2001).

Determining the cellular function of this family is, therefore, important not only for understanding the mechanisms that cause the observed disease phenotypes but also for understanding how they maintain the normal function of TULP-expressing cells. Experimental evidence suggests that TULPs have cellular roles in vesicular trafficking (Hagstrom et al., 1999; Hagstrom et al., 2001), in mediation of insulin signaling (Kapeller et al., 1999) and in gene transcription (Boggon et al., 1999). Molecular studies also provide us with clues about how expression of Tub and translocation of the TUB protein is affected by pathways regulating feeding behavior and energy balance (Koritschoner et al., 2001; Santagata et al., 2001). Here, we discuss the recent advances in genetic and molecular studies and the future directions that might be explored in order to further our understanding of the role of tubby gene family, particularly in neuronal cells.

The original phenotype associated with a mutation in the Tub gene, maturity onset obesity associated with insulin resistance (Table 1) (Coleman and Eicher, 1990), was later expanded to include sensory neural defects (retinal and cochlear degeneration) (Heckenlively et al., 1995; Ohlemiller et al., 1997). As the tubby gene is predominantly expressed in neuronal cells, including those of the hypothalamus, it was suggested that the late-onset obesity in tubby mice might reflect defects in the neurons of the hypothalamic ‘satiety center’ (Kleyn et al., 1996). Pro-opiomelanocortin (POMC) and neuropeptide Y (NPY) mRNA levels are significantly reduced in the arcuate nucleus (ARC), whereas NPY levels are increased 30-fold in the dorsomedial (DMH) and ventromedial (VMH) hypothalamic nuclei of tubby mice compared with wild-type controls (Guan et al., 1998). Although a reduction in αMSH (melanin-stimulating hormone), an anorexigenic peptide derived from POMC, might be expected to result in increased body weight, the consequences of increased NPY levels in the VMH and DMH are not known. In addition, the αMSH expression changes were observed after the disease onset and, therefore, may not be causative. The exact role of TUB in the etiology of the obese phenotype thus remains unclear. As TUB is expressed at comparable levels in all neurons in the brain, it is possible that additional, more subtle, phenotypes remain to be discovered.

Retinal degeneration characterized by apoptotic loss of photoreceptor cells is a common phenotype observed in both C57BL6-tub/tub (B6-tub/tub) and C57BL6.129-Tulp1^tm1Jax (Tulp1^–/–) mice (Ikeda et al., 1999b; Hagstrom et al., 1999; Ikeda et al., 2000). Although there are differences in progression and severity of the degeneration, morphological features of photoreceptor cell abnormalities in both the spontaneous tub mutant and the Tulp1 knockout are similar (Heckenlively et al., 1995; Hagstrom et al., 1999; Ikeda et al., 2000).

Unlike tubby and Tulp1^–/– mice, B6.129-Tulp3^tm1Jax (Tulp3^–/–) mice exhibit embryonic lethality, with failure of neural tube closure characterized by neuroepithelial apoptosis, specifically in the hindbrain and the caudal neural tube (Ikeda et al., 2001). Tulp3 is ubiquitously expressed throughout embryonic development. The earliest observable phenotype of the knockout mice is a significant reduction in the number of βIII-tubulin-positive neurons in the hindbrain at embryonic days (E) 9.5-10.5. Tulp3 appears, therefore, to be essential for the maintenance or function of normally differentiating neuronal cell populations. The neuroepithelial cell population is known to be spatially subdivided into distinct classes of neuron (Briscoe and Ericson, 2001). As apoptosis is restricted to the ventral region of the neuroepithelium in the hindbrain of Tulp3^–/– embryos, there is selective cell death in specific cell types in this targeted mutant model.

Taken together, the genetic analyses suggest that the primary site of action of the tubby family genes is in neuronal cells, including hypothalamic, sensory and differentiating embryonic neurons. ‘Apoptosis’ of neuronal cells is common to all mutations in tubby family genes. Do mutations within Tulp genes directly drive the cell death pathway or does a functional defect lead to the eventual cell death? In the case of the sensory neural defects, the latter alternative is likely. Auditory brainstem response (ABR) and electroretinographic (ERG) recordings in tubby mice show a functional loss of hearing and vision, respectively, before cell death (Ikeda et al., 1999a) (A.I. et al., unpublished).

The severity of the phenotypes observed in tubby mice is affected by the genetic background (Ikeda et al., 1999a; Nishina et al., 2000). For example, the profound hearing loss in B6-tub/tub mice is completely rescued in tub/tub homozygous progeny derived from intercrossing B6-tub/tub mice with the mouse strains AKR/J, CAST/Ei and 129/Ola. Linkage analysis identified a major modifier locus – modifier of tubby hearing 1 (moth1) (Ikeda et al., 1999). A single moth1 allele from several different inbred strains protects tubby mice from the hearing loss associated with a B6-derived recessive moth1 allele. Homozygosity of the B6 moth1 allele itself cannot induce hearing impairment in wild-type B6 mice. This suggests that the allelic difference in the moth1 gene of B6 compared to the other inbred strains, therefore, is not critical for maintaining their sensory neurons but becomes significant only when the tubby gene is mutated and unable to function properly. In addition, the moth1 locus maps to a region on mouse chromosome 2 that also contains a gene that protects tubby mice from retinal degeneration (Nishina et al., 2000). This finding suggests that the same gene on chromosome 2 is involved in a common pathway through which retinal and cochlear degeneration are induced in tubby mice.

The C-terminal region is highly conserved among TULPs (Fig. 1), suggesting that this region might contain a functionally important domain. This idea is supported by genetic studies. The mutation within the tub gene is at a splice donor site in the last intron, which results in the replacement of the last 44 amino-acid residues with 24 that are not observed in the normal protein (Kleyn et al., 1996; Noben-Trauth et al., 1996). The fact that the targeted tub null allele (Stubdal et al., 2000) has the same phenotype as the spontaneous mutant confirms that this splice-site mutation is also the molecular basis for the spontaneous loss-of-function mutation. In addition, the same type of splice-site mutation was observed in the TULP1 gene among patients affected with retinitis pigmentosa (RP) 14 in a large Dominican Republic kindred (Banerjee et al., 1998). A clustering of mutations in the C-terminal region was also found in spontaneous cases of RP, again emphasizing the functional importance of this domain in Tubby family members.

To probe the function of this conserved region, Boggon et al. (Boggon et al., 1999) crystallized the C-terminal domain of TUB and performed X-ray crystallographic analyses. The electron-density map obtained was interpreted as a unique protein structure, a 12-stranded β-barrel conformation filled with a central hydrophobic core that traverses the entire barrel. Two prominent features of the folded protein are a large groove of positively charged residues and a smaller region of negatively charged residues on the opposite side of the groove. This positively charged groove is expected to form protein-protein or protein-DNA binding sites. Additional molecular biological experiments showed that the TUB C-terminal domain has the potential to bind double-stranded DNA oligomers (Boggon et al., 1999). Given that a chimeric protein composed of the N-terminus of TUB or TULP1 and a GAL4 DNA-binding domain can activate transcription, Boggon et al. (Boggon et al., 1999) suggested that TUB might be a novel class of transcription factor (Fig. 2b). So far, there is no experimental evidence that TUB or TULP1 can recognize specific DNA sequences, and targets for TUB transcriptional activation have not yet been identified. Nevertheless, it is tempting to speculate that the reduced Pomc and Npy mRNA levels in the arcuate nucleus are the result of reduced TUB transcriptional activity in the tubby mutant.

Compared with the C-terminus, the N-terminal region of members of the tubby family is less well conserved; however, clear similarities between the family members exist. In particular, the exons containing the acceptor sites for the N-terminal splice forms and the putative nuclear-localization signals are well conserved (Nishina et al., 1998) (Fig. 1). Consistent with the presence of potential nuclear localization signals, the N-terminus defines the nuclear localization of TUB (He et al., 2000; Santagata et al., 2001).

Protein sequences are highly conserved among tubby gene family members, which suggests that they perform the same type of function within the cells in which they are expressed. As described below, the observed mislocalization of rhodopsin in both tubby and Tulp1^–/– mice implies a common function (Hagstrom et al., 1999; Hagstrom et al., 2001) (A.I., P.M.N. and J.K.N., unpublished data). However, although expression of each gene appears to overlap in several cell types, genetic analyses suggest that they are not functionally redundant in all tissues. Although both TUB and TULP1 are expressed in the same hypothalamic neurons, tubby mice become obese but Tulp1^–/– mice do not, which suggests unique tasks for each protein (Ikeda et al., 2000). This idea is also supported by the observation that the distribution of TUB and TULP1 within the nucleus of hypothalamic neurons is not overlapping (Ikeda et al., 2000). In Tulp3^–/– embryos, only a specific population of neuronal cells undergoes apoptosis. One explanation for the restricted cell death is that the ubiquitously expressed Tulp3 is interacting with specific genes in a subset of developing neuronal cells. Alternatively, although on an organismal level there does not appear to be functional redundancy, it remains possible that family members can substitute for each other in specific cell types. In this context, it is interesting that abnormalities are observed only in adult tub/tub mice, even though the Tub gene is highly expressed in the neuroepithelium at embryonic day E10.5 (Sahly et al., 1998), a time at which the Tulp3 gene is also expressed (Ikeda et al., 2001). It is possible, therefore, that Tulp3 compensates for TUB during neuronal cell development.

Koritschoner et al. (Koritschoner et al., 2001) identified Tub as a potential T3-regulated gene in a differential display experiment. The level of expression of Tub is lower in hypothyroid rats, most significantly in Purkinje cells of the cerebellum. Normal expression levels were restored by T3/T4 treatment. Tub expression was also upregulated by T3 treatment of cultured neuronal cells. These results suggest that the expression of Tub is regulated by thyroid hormone, a hormone known to be important in the regulation of body weight (Krotkiewski, 2000) and neurosensory development (Forrest et al., 1996; Ng et al., 2001). However, the significance of these findings, although interesting, is not yet readily apparent, because the major difference in expression levels was observed in Purkinje cells. Although a role for the cerebellum in food intake operant behavior has been suggested (Morimoto et al., 1984; Lucchi et al., 1998; Mahler et al., 1993), whether acute thyroid regulation in the cerebellum affects feeding behavior is currently not known. Further studies of the regulation of Tub by thyroid hormone in other phenotypically relevant sites, such as the hypothalamic nuclei, could lead to important contributions to our understanding of the functional role of TULPs.

Kapeller et al. (Kapeller et al., 1999) reported that when the Tub gene was transfected into chinese hamster ovary cells that stably express insulin receptor (CHO-IR), the resultant TUB protein was phosphorylated at tyrosine residues in response to insulin and insulin-like growth factor treatment (Fig. 2a). Furthermore, in vitro studies showed that TUB is directly phosphorylated by the purified insulin receptor kinase, as well as by ABL and JAK2 kinases, but not by the EGF receptor or SRC kinases. The phosphorylated form of TUB appears to bind to the SRC homology 2 (SH2) domains of ABL, LCK and phospholipase Cγ. Therefore, TUB might function as an adapter protein, linking the insulin receptor to SH2-containing proteins. Since the tubby gene is prominently expressed in the brain, this signaling pathway is likely to be active in the neuronal cells in vivo. Interestingly, TUB and the insulin receptors colocalize in neurons of the hypothalamus (Kleyn et al., 1996; He et al., 2000; Unger et al., 1991; Carvalheira et al., 2001), a structure known to be important for regulating energy homeostasis. Recent studies using gene-targeted mice lacking insulin receptor substrate 2 (IRS-2) and brain insulin receptor (INSR) revealed that insulin signaling plays an important role in the energy homeostasis in the brain (Burks et al., 2000; Brüning et al., 2000). It will be interesting to see whether TUB phosphorylation and/or translocation to the nucleus is impaired in the brains of knockout mice in which INSR has been specifically ablated.

A distinct functional role of the TULPs may also be conferred by their subcellular localization. In hypothalamic neurons, TUB is localized in the cytoplasm and nucleus (He et al., 2000), whereas in photoreceptor cells it appears to be only in the cytoplasm. In addition, although TUB was only detected in the nucleolus in the hypothalamic neurons, TULP1 was present in structures that are likely to represent the perinucleolar cap and coiled bodies or gems in the nucleus but not the nucleolus (Ikeda et al., 2000). Although it is currently not known why TUB is differentially localized in different cell populations, it is possible that regulatory factors that determine the localization of TUB confer multiple functions upon the TULPs. For example, we found that TUB is localized in the nucleus of the precursor cells of the retinal photoreceptor cells, which undergo proliferation, and TUB localization shifts to the cytoplasm when these cells differentiate (unpublished data).

TUB is localized in the nucleus as well as in the plasma membrane of cultured cells (Kapeller et al., 1999; Koritschoner et al., 2001; Santagata et al., 2001). Santagata et al. (Santagata et al., 2001) report that TUB is transported from the plasma membrane to the nucleus through G protein α_q (Gα_q) activation (Fig. 2b). Association of TUB with the plasma membrane is thought to occur after its binding to PtdIns(4,5)P₂ (phosphatidylinositol-4,5-bisphosphate), a phospholipid that is highly enriched in the plasma membrane. Release of TUB is triggered by receptor-mediated activation of Gα_q through the action of phospholipase C-β. These findings suggest potential candidates for receptors that might regulate TUB in vivo through Gα_q, such as serotonin (Tecott et al., 1995), bombesin (Ohki-Hamazaki et al., 1997), dopamine D1 (Sidhu, 1998), melanocortin 4 (Huszar et al., 1997) and melanin-concentrating hormone receptors (Chambers et al., 1999; Saito et al., 1999). These receptors are expressed in the hypothalamus and influence feeding behavior and energy homeostasis. Although transfection of serotonin receptor 2c (5HT_2c) induced nuclear translocation of TUB in vitro (Santagata et al., 2001), and a mild obesity phenotype is observed in 5HT_2c null mutant mice (Tecott et al., 1995), other hypothalamic receptors may also play a role in the obesity observed in tubby mice. There is much that still needs to be learned about the regulation of TULPs. Generally, nuclear localization of proteins is regulated by their binding proteins or conformational changes owing to post-translational modification such as phosphorylation. Phosphorylation of TUB may be important for its nuclear localization or, alternatively, additional proteins may be necessary for TUB translocation.

Extracellular vesicles accumulate near the inner segments of the photoreceptor cells in both Tulp1^–/– and tubby mice (Hagstrom et al., 1999; Ikeda et al., 2000). Hagstrom et al. (Hagstrom et al., 1999) hypothesized that the tubby family normally functions in intracellular vesicular trafficking of opsin and other transported factors (Fig. 2c). Indeed, ectopic localization of rhodopsin is observed (Hagstrom et al., 2001) (A.I., P.M.N. and J.K.N, unpublished) in both Tulp1^–/– and tubby mice. In addition, yeast two-hybrid analysis using TUB as bait identified proteins that are associated with secretory vesicles as binding partners (Hagstrom et al., 2000). These results support a role for TUB and TULP1 in protein trafficking in the photoreceptor cells of the retina.

Genetic and biochemical studies, which have focused on identifying a common function for the members of the tubby family, have provided insights into potential functions of the TULPs in a wide variety of neuronal cell populations. Continued functional studies aimed at elucidating tissue-specific functions may clarify how TULPs function in cells and how the mutations within these genes cause particular disease pathologies. For example, if TULPs act as transcription factors, do they recognize specific DNA sequences within specific tissues? What kinds of genes are regulated by TULPs? These genes will certainly become key factors in understanding the pathways through which TULPs act. But how can a role as a transcription factor be reconciled with the observed rhodopsin mislocalization and the localization of TUB to the nucleolus? At this point, our current understanding of the function of the tubby-like gene family is reminiscent of the three blind men and the elephant, who after examining different parts of the elephant came to vastly different conclusions about the nature of the beast. Mutants for each tubby family gene studied thus far show significant abnormalities and phenotypes that do not always overlap despite overlapping expression patterns. It is therefore not clear that all of the observed phenotypic characteristics of Tulp mutations can be ascribed to a single pathway with one function. Indeed, it seems possible that TULPS are multifunctional proteins that may play multiple, independent roles in normal cellular function and integrity.

We are grateful to Barbara K. Knowles and Edward H. Leiter for careful review of the manuscript. Research in our laboratory was supported by grants from Foundation for Fighting Blindness, NIH DK59641, and AXYS Pharmaceuticals Inc. Institutional shared services are supported by National Cancer Institute Support grant CA-34196.