Drosophila Niemann-Pick Type C-2 genes control sterol homeostasis and steroid biosynthesis: a model of human neurodegenerative disease

Cholesterol, an essential component of eukaryotic cell membranes, also serves as the precursor of many steroid hormones and thus plays vital roles in many developmental processes (Farese and Herz, 1998). Cells in the body maintain proper cholesterol levels through elegant homeostatic regulatory systems. Defects in cholesterol homeostasis and metabolism have been linked directly or indirectly to many disease conditions.

Niemann-Pick type C (NPC) disease is one such cholesterol homeostasis-related disorder characterized by aberrant accumulation of free cholesterol in late endosome and lysosome-like compartments(Patterson, 2003). Normal cells take up exogenous cholesterol through the receptor-mediated low density lipoprotein (LDL) endocytic pathway. LDL-derived free cholesterol must then leave the endosomal compartment, a process that is blocked in NPC disease cells, to move to other membrane compartments, including the endoplasmic reticulum (ER), and to control homeostatic responses(Liscum and Faust, 1987). NPC disease is a progressive neurodegenerative disorder in which the degeneration of cerebellar Purkinje neurons is most prominent(Higashi et al., 1993). Although the link between cholesterol homeostasis defects and neurodegeneration remains enigmatic, the deficiency of oxysterol and/or neurosteroid has recently been implicated as partially responsible for this neurodegeneration (Griffin et al.,2004; Langmade et al.,2006).

Mutations in either of two different human genes, NPC1 and NPC2, result in Niemann-Pick type C disease, with NPC1mutations accounting for about 95% of known cases(Patterson, 2003). The large Npc1 protein has 13 transmembrane domains and a sterol-sensing domain (SSD)(Carstea et al., 1997; Loftus et al., 1997). Npc2, a small, secreted protein that binds cholesterol strongly, was first found as an abundant component of human epididymal fluid and later linked through human genetics to the inherited cause of NPC disease in about 5% of the families studied (Naureckiene et al.,2000). The crystal structure of Npc2 has been determined and found to contain a cavity that genetic analyses show to be the likely binding site for cholesterol (Friedland et al.,2003; Ko et al.,2003). Npc2 may serve as a lysosomal cholesterol transporter,rapidly transporting cholesterol to acceptor membranes(Cheruku et al., 2006). Although Npc1 and Npc2 are different types of cholesterol-binding proteins,they appear to be in a common pathway or process based on the virtually indistinguishable phenotypes of the human patients carrying one or the other homozygous mutation.

To uncover the disease mechanisms as well as the biological function(s) of NPC proteins, useful NPC disease models have been established in yeast, worms,flies and mice (Berger et al.,2005; Higaki et al.,2004; Li et al.,2004; Malathi et al.,2004). We and the L. Pallanck laboratory have previously created Drosophila NPC models using npc1a (also referred to as NPC1 - FlyBase) mutations (Fluegel et al., 2006; Huang et al.,2005). Drosophila and all other insects are unable to synthesize sterol from simple precursors. In order to synthesize the molting hormone 20-hydroxyecdysone (20E) and to sustain the growth and reproduction of the fly, sterol has to be obtained from food(Clark and Block, 1959). In Drosophila, npc1a is crucial for sterol homeostasis, as is mammalian NPC1. The fly mutants have a molting defect and homozygotes die as first-instar larvae due to a deficiency of the molting hormone 20E, the primary steroid hormone identified in insects to date. 20E plays crucial roles in insect oogenesis, embryogenesis and metamorphosis(Thummel, 1996). npc1a mutants can be rescued by feeding them excess 20E or either of two of its precursors: cholesterol or 7-dehydrocholesterol. Thus, the ecdysteroid deficiency is evidently due to an inability to access sufficient sterol precursor, a somewhat surprising result given the massive accumulations of sterol in punctated structures that are seen in the mutants by filipin staining. The simplest explanation is that the accumulated sterol, stored in multi-lamellar and multivesicular compartments, is not available for 20E synthesis.

Based on the findings in npc1 mutant worms, flies and mice, we proposed a cholesterol shortage model of NPC disease(Huang et al., 2005). The normal function of Npc1 protein may be to promote delivery of sufficient sterol to the ER and/or mitochondria, organelles in which specific steps of steroidogenesis occur. In the studies reported here, we examined the functions of Npc2 proteins in Drosophila. Our results further support the cholesterol shortage model proposed previously.

Flies were cultured on standard cornmeal medium at 25°C. npc2aand npc2b excision mutageneses were performed using standard methods,starting with the Bloomington Drosophila Stock Center lines KG05307 and KG00996 (Robertson et al.,1988). Three alleles of npc2a (npc2a²³⁹,npc2a²⁷¹ and npc2a³⁷⁶) and three alleles of npc2b (npc2b¹⁸, npc2b¹⁹ and npc2b²²) were isolated. All mutants were back-crossed to wild-type (Canton-S) flies three times before further phenotypic characterizations. The molecular lesions were determined using genomic DNA polymerase chain reactions and by sequencing. The coding regions of npc2a and npc2b are entirely deleted by the mutations in their corresponding sets of alleles.

Eight npc2-like genes (npc2a-h) were found by BLAST searches of the Drosophila melanogaster genome with the sequence of the human NPC2 protein. The cDNAs corresponding to the npc2 gene family were amplified by RT-PCR and sequenced. The protein sequences were then deduced from the cDNA sequences. The gene structures of all but one(npc2e) of the npc2 genes were predicted correctly in FlyBase. npc2e (CG31410) has an extra intron compared to the FlyBase prediction and the correct coding sequence aligns well with other Npc2 proteins. The UAST-npc2a and UAST-npc2b constructs were made by inserting full-length cDNAs into the EcoRI site of the pUAST vector.

First-instar wild-type and mutant larvae were collected into 1.5 ml tubes(100 larvae/tube) and kept at -80°C until assayed for whole body ecdysteroid content. Larvae were homogenized by sonication (Sonic Dismembrator, Fisher) and extracted exhaustively with both methanol and ethanol. Pooled solvents for each replicate of 100 larvae were evaporated under low pressure into 2 ml plastic tubes and the dried residue subjected to RIA employing the H22 antiserum (Warren et al., 1988). For each genotype, 600-800 larvae were used.

For npc2a; npc2b double mutants, each group of 200 first-instar larvae was placed on apple juice plates with baker's yeast paste containing supplementary sterols, as described previously(Huang et al., 2005), and the lethal phases were determined by larval spiracle and mouth hook development. The final concentrations for the supplements used were: cholesterol, 0.14 mg/g; 7-dehydrocholesterol, 0.14 mg/g; and 20E, 8 μg/g.

The sterol content in larvae was quantified by following a published protocol (Fluegel et al.,2006). The Amplex Red cholesterol assay kit (Molecular Probes) was used to assess sterol content in wandering third-instar animals. Larvae were collected and washed before being weighed and homogenized in 150 mM NaCl, 2 mM EGTA, 50 mM Tris pH 7.5, to make a 100 mg/ml larval homogenate. The homogenate was spun at 5000 rpm for 5 minutes to pellet cuticle debris, and the supernatant was used for sterol content assays. Fluorescence was measured with a fluorescence spectrophotometer with a 560/585 nm filter set.

For filipin staining of free sterols, tissues were fixed in 4%paraformaldehyde for 30 minutes, washed twice in PBS and stained with 50μg/ml filipin (Sigma) PBS solution for 30 minutes. Samples were then washed twice with PBS before mounting them in Vectashield mounting medium. For TUNEL analysis, aged brains were dissected and fixed (PBS, 4% paraformaldehyde) for 20 minutes at room temperature. Tissues were washed twice in PBS, once in H₂O plus 0.1% Triton X-100 and 0.1% sodium citrate for 10 minutes,and then twice in PBS. TUNEL analysis was performed by following the manufacturer's instructions (Boehringer Mannheim). TUNEL and neuron double labeling was performed using an antibody against the pan-neuronal marker Elav(Iowa Hybridoma Bank). Synaptotagmin staining (anti-Synaptotagmin, from Dr Hugo Bellen) was performed using standard techniques(Littleton et al., 1993). All images were collected using a compound microscope and a cold CCD camera.

For each genotype, 10 vials containing a total of 200 flies were passed into fresh vials every 4 days, at which time the number of dead flies was recorded.

Methods for in situ hybridization to detect mRNA in overnight embryo collections and for electron microscopy were described previously(Huang et al., 2005). Malpighian tubules from third-instar larvae were dissected first, then fixed in 4% paraformaldehyde and 2% glutaraldehyde in 0.1 M cacodylate buffer, pH 7.4, followed by further processing for electron microscopy.

The Npc2 protein has been conserved throughout much of eukaryotic evolution. Only one npc2 gene is present in yeast, worm, mouse and human genomes. Drosophila melanogaster, clearly a highly advanced organism, has a family of eight npc2-like genes, which we have named npc2a-h. We identified the gene family by BLAST searching with the sequence of human NPC2. We further confirmed and corrected the npc2-like gene structures by RT-PCR(Fig. 1A; also see Materials and methods). Protein sequences of Npc2a-h (CG7291, CG3153, CG3934, CG12813,CG31410, CG6164, CG11314 and CG11315; Fig. 1B) range from 22 to 36% identical to human NPC2.

Of the eight Npc2-like proteins, Npc2a (also referred to as NPC2 - FlyBase)has the highest sequence identity (36%) and similarity (53%) to human NPC2(Fig. 1B). Further protein sequence analysis within this protein family reveals that CG3934 (Npc2c),CG12813 (Npc2d) and CG31410 (Npc2e) form a subgroup clustered at cytogenetic locus 85F8 on chromosome III, while CG11314 (Npc2g) and CG11315 (Npc2h) form another subgroup clustered at locus 100A3 on chromosome III(Fig. 1A). Both groups of genes presumably arose from gene duplication events.

Crystal structure determination and mutational analyses have shown that Npc2 has three disulfide bonds and forms a hydrophobic core implicated in cholesterol binding (Friedland et al.,2003; Ko et al.,2003). All six disulfide bond-forming cysteine residues are absolutely conserved in Drosophila Npc2a-h proteins. At other positions shown to be functional in mouse Npc2, Npc2a-h proteins have some variation. For example, F66, V96 and Y100 (amino acid numbers correspond to positions in the mature Npc2 protein without the signaling peptide) of mouse Npc2 are located near the hydrophobic core and are involved in cholesterol binding (Ko et al., 2003). V96 is the same or highly similar in seven Drosophila Npc2 proteins except Npc2h, F66 is conserved or replaced by the similar amino acid Tyr (Y)in five Npc2 proteins (not in Npc2f, g and h), and Y100 is conserved or replaced by the similar Phe (F) in six Npc2 proteins (not in Npc2c and f)(Fig. 1B). D72 and K75 of mouse Npc2 are not required for cholesterol binding but are necessary for normal Npc2 function. D72 is conserved or replaced by the related amino acid Glu (E)in four of the Drosophila Npc2 proteins (Npc2a, e, f and h), while K75 is conserved in only three Npc2 proteins (Npc2e, g and h)(Fig. 1B). The variations of these key residues in Npc2 proteins may allow retention of cholesterol-binding ability while adding some capability to bind to sterols other than cholesterol. Evidence for functional conservation despite the changes in key residues is provided by the rescue of mammalian Npc2-mutant cells with an introduced yeast NPC2 gene, which also has changes in encoded key residues such as K75 and Y100 (Fig. 1B) (Berger et al.,2005).

The npc2-like gene family is present in other sequenced Drosophila genomes, such as D. yakuba, D. pseudoobscura and D. virilis, as well as genomes from other insect species, including Anopheles gambiae (at least eight npc2-like genes), Bombyx mori (at least three npc2-like genes) and Tribolium casteneum (at least three npc2-like genes). Together these data suggest possible multiple rounds of gene duplication events within Class Insecta.

The gene structures of Drosophila melanogaster npc2a-h reveal a pattern of evolution in the generation of introns within the coding region. Three genes (npc2a, g and h) have no intron. Two genes, npc2b and d, each have one intron in the same position(position 1 in Fig. 1B). Two others, npc2c and e, have two introns in the same positions(positions 1 and 2 in Fig. 1B). The eighth gene, npc2f, has three introns (positions 1, 2 and 3 in Fig. 1B). Interestingly, the intron positions (positions 1, 2 and 3 in Fig. 1B) in the D. melanogaster npc2-like gene family are almost identical to the intron positions of the vertebrate npc2 genes, including those from human,mouse, rat, chimpanzee, cow and zebrafish. By contrast, the intron position in ncr-2, the Caenorhabditis elegans homolog of npc2,is different (position 4 in Fig. 1B). Together, the chromosomal clustering of npc2 genes and the similarity of intron positions support the concept that the generation of the npc2 gene family was a result of multiround gene duplication events.

To address the potential roles of different NPC2-like proteins, the temporal and spatial expression patterns of the npc2a-h genes during embryonic stages was determined using whole-mount in situ hybridization. The data revealed that npc2a has the broadest expression pattern, whereas other npc2 genes are either not detectably expressed or expressed in restricted locations (Fig. 2). The npc2a gene provides a substantial maternal contribution of RNA,and is also ubiquitously expressed at all stages examined. High levels of npc2a expression were found in midgut, salivary gland and ventral nerve cord (Fig. 2A-D). npc2b is expressed at the highest levels in the trachea and hypopharynx (Fig. 2I). npc2g is specifically expressed in head mesoderm and fat body(Fig. 2G,H). npc2d and npc2h transcripts could be detected only in salivary gland(Fig. 2F), while npc2eis expressed in hindgut (not shown). The expression of npc2c and npc2f was not detected by in situ hybridization at any time during embryogenesis.

As npc1a is highly expressed in the ring gland, and ring gland expression of npc1a is important for ecdysteroid biosynthesis, the expression of npc2a-h in ring glands was examined. Brains and imaginal discs from wandering third-instar larvae were also examined. In contrast to npc1a, none of the npc2a-h genes was highly expressed in ring glands. We could detect moderate levels of gene expression in larval ring glands, brains and imaginal discs for several npc2genes, including npc2a and npc2b(Fig. 2K,L and data not shown).

Because npc2a has the broadest expression pattern among the eight genes studied, and the highest protein sequence similarity to vertebrate Npc2,we focused initially on characterizing npc2a function using mutant phenotypic analysis. Through P element imprecise excision we generated three deletion alleles (npc2a²³⁹, npc2a²⁷¹and npc2a³⁷⁶; Fig. 3A). The whole coding region of npc2a was completely deleted in each of the three alleles, yet homozygous mutant animals were viable and adults were fertile. Each allele was tested in trans to several different genetic deficiencies that remove the gene, and these genetic combinations were also viable and fertile. Whole-mount in situ hybridization with an npc2a antisense probe did not detect any RNA signal in homozygous npc2a mutant embryos, indicating that they are bona fide npc2a mutants (Fig. 2E).

We next examined the sterol distribution in npc2a mutants using filipin staining. Filipin, which stains non-esterified sterols, is often used to study sterol accumulation in NPC1 and NPC2 mutant mammalian cells. We previously used filipin successfully to determine the sterol distribution in Drosophila npc1a mutants and found that homozygous mutants have an abnormal sterol distribution similar to that found in mammalian NPC mutants. This is most easily seen by light microscopy as a punctate pattern of filipin-stained particles, and with electron microscopy as multi-lamellar structures (Huang et al.,2005).

In npc2a/npc2a mutant tissues, including salivary gland, midgut,Malpighian tubules, imaginal discs, brains, trachea and ovaries, a punctate pattern of filipin fluorescence was found(Fig. 3B-G). Most tissues had many such spots of accumulated sterol, except trachea, where we found fewer puncta. The filipin staining phenotype was similar to that of Drosophila npc1a mutant tissues and mammalian NPC mutant cells, indicating a conserved role for Drosophila npc2a in regulating efficient intracellular sterol trafficking. The sterol distribution abnormality in npc2a/npc2a mutants could be fully rescued by ubiquitous expression of a UAST-npc2a transgene (see below), indicating that this phenotype is indeed due to npc2a mutation.

We further examined the structure of mutant npc2a/npc2a cells using electron microscopy. Large multi-lamellar body and multi-vesicular body structures were found in npc2a mutant Malpighian tubules(Fig. 4), just as in homozygous npc1a mutants. The multi-lamellar structures were often clustered together to form large inclusions with or without electron-dense materials within (Fig. 4B and C,respectively). The similarities in cellular phenotypes and ultrastructural defects of npc1a and npc2a mutants further suggest the conserved roles of NPC genes in regulating intracellular sterol trafficking from Drosophila to mammals. As the homozygous mutants survive to adulthood, while npc1a/npc1a flies do not, there must be important differences between npc1 and npc2a phenotypes, and accumulation of sterol is not, by itself, adequate to cause death.

The apparently similar defects in sterol distribution in Drosophila npc1a and npc2a mutants raise the question: why do npc1a mutants die as first-instar larvae, while npc2amutants are viable and ultimately fertile? We have suggested previously that the first-instar larval lethality of npc1a is due to ecdysteroid deficiency, although this was inferred rather than measured directly(Huang et al., 2005). The difference in phenotypes between npc1a and npc2a homozygotes could reflect different ecdysteroid levels.

We have now directly measured ecdysteroid levels during the first-instar stages (38 hours after egg laying) of wild-type, npc1a/npc1a and npc2a/npc2a larvae. Compared to wild type, the npc1a mutant had low ecdysteroid titers (16.7±0.9 pg/100 mutant larvae versus 87.7±4.4 pg/100 wild-type larvae). The npc2a/npc2a mutant larvae had somewhat lower than normal ecdysteroid levels (53.3±3.6 pg/100 mutant larvae versus 73.8±4.1 pg/100 wild-type larvae)(Garen et al., 1977; Kraminsky et al., 1980; Neubueser et al., 2005). These results could explain why npc1a mutants die as first-instar larvae,i.e. cannot molt, while npc2a mutants are viable and are fertile as adults. Furthermore, the data support our previous hypothesis that the first-instar lethality of npc1a/npc1a mutants is due to ecdysteroid deficiency.

The ecdysteroid titer results do not explain why apparently similar defects in sterol distribution are associated with a shortage of sterol substrate for ecdysteroid biosynthesis in npc1a but not npc2a mutants. There are at least two possibilities. First, npc1a and npc2amay function differently in ecdysteroidogenesis, so that only npc1abut not npc2a is involved in sterol transport to the mitochondria. This could be true despite the apparently similar overall accumulation of sterol in filipin-stained compartments. Alternatively, the difference could be due to redundant functions of the multiple npc2 genes. Perhaps in the npc2a/npc2a mutants a substantial amount of sterol reaches the mitochondria, transported by other Npc2 family protein-mediated processes.

In order to test the gene redundancy hypothesis and to examine possible functions of a second npc2 gene, the function of npc2b was analyzed. npc2b is expressed in the tracheal system and hypopharynx(Fig. 2G), so we paid particular attention to the possible redundancy of npc2a and npc2b in these tissues. We found that npc2a and npc1a mutants have quite different patterns of sterol accumulation in larval trachea. Punctate filipin staining was readily observed in npc1a mutants but few sterol particles accumulated in the trachea of npc2a mutants (Fig. 5A,C).

To determine whether sterol accumulation in npc2a mutants is prevented by npc2b, we generated three npc2b deletion alleles (npc2b¹⁸, npc2b¹⁹ and npc2b²²) of npc2b by imprecise P element excision. The whole coding region of npc2b was completely deleted in these three alleles. In homozygous npc2b mutant animals, no in situ hybridization signal could be detected with an npc2b antisense probe,so as expected the new alleles of npc2b were nulls(Fig. 2). Like npc2ahomozygotes, npc2b homozygotes and flies carrying an npc2ballele in trans to a genetic deficiency were viable and fertile.

No sterol accumulation was observed in any npc2b mutant tissues,including the trachea, where we know the gene is preferentially transcribed(Fig. 5B). However, npc2a/npc2a; npc2b/npc2b doubly homozygous mutants had a large number of filipin-stained puncta in the trachea. The level of sterol particles was similar to sterol accumulation in npc1a/npc1a mutants(Fig. 5D). We conclude that npc2a and npc2b function redundantly in sterol trafficking,at least in trachea.

Although both single mutants were viable, fertile and were not developmentally delayed, npc2a; npc2b double mutants died as larvae or pupae and the third-larval instar was prolonged. Aside from a small percentage of animals (about 17%) that died in the first or second larval stage, the majority of npc2a/npc2a; npc2b/npc2b double mutants molted to the third instar quite normally. They remained in the third instar for 3-6 days, compared with about 2 days for wild-type animals. Twenty-six percent died while still in the third-instar stage, while the remaining 57% formed pupae (Fig. 6A). About a tenth of the mutant pupae developed to the adult stage, but they were sick and usually died within 2 weeks (Fig. 6A). For this reason we were not able to establish homozygous double mutant stocks.

Most of the npc2a/npc2a; npc2b/npc2b double mutants could be rescued by feeding them a diet enriched with cholesterol, 7-dehydrocholesterol or 20E (Fig. 6A). The prolonged third instar of the double mutants, together with the results of the rescue experiment, suggests that the ecdysteroid level is relatively low in the double mutants. In the presence of sufficient substrate, npc2a/npc2a;npc2b/npc2b mutants were evidently able to synthesize enough ecdysteroid for fairly normal development.

As with npc1a mutants, the insufficiency of sterol substrate appeared to be the main problem for npc2a/npc2a; npc2b/npc2b double mutants. The similarity of the double npc2 homozygotes to npc1a homozygotes suggests that npc1a has irreplaceable functions, while the two npc2 genes tested to date have somewhat redundant functions. Both Npc1 and Npc2 are necessary to regulate sterol homeostasis and carry out adequate biosynthesis of 20E.

Our results agree well with the hypothesis that Npc1 and Npc2 promote efficient intracellular sterol trafficking for ecdysone biosynthesis. To further pinpoint the roles of Npc2, we examined the sterol level in npc2 mutants and possible tissue-specific requirements for npc2. Despite the altered filipin staining patterns in many tissues,the overall level of sterol was not much different in Drosophila npc1a mutants compared to controls(Fluegel et al., 2006). We measured sterol levels in npc2 mutants and found a similar result:the overall level of sterol was not significantly changed in npc2a/npc2a or npc2b/npc2b single mutants or in npc2a/npc2a;npc2b/npc2b double mutants(Fig. 6B).

As npc1a is required in the ring gland for ecdysone biosynthesis,we analyzed whether npc2 genes are also important in this same crucial tissue or in others. We focused our analyses on the ring gland,nervous system and trachea. We used the Gal4 system to address tissue-specific requirements for npc2a and npc2b. Ubiquitous expression of UAST-npc2a or UAST-npc2b could rescue the lethality: 83% of the double mutants survived to adulthood in the presence of tub-Gal4>UAST-npc2a and 80% survived to adulthood in the presence of tub-Gal4>UAST-npc2b. Only 5% survived in double mutants lacking any transgene. The pattern of punctate filipin-stained sterol accumulation in trachea of double mutants was similarly rescued by the transgenes (data not shown).

Expression of UAST-npc2a or UAST-npc2b only in the ring gland, using the 2-286-Gal4 driver, rescued the lethality of the double mutant: 78% of the double mutants survived to adulthood in the presence of 2-286-Gal4>UAST-npc2a and 86% survived to adulthood in the presence of 2-286-Gal4>UAST-npc2b. These findings are consistent with the conclusion that a defect in ecdysone biosynthesis is the main cause of the larval lethal phenotype. By contrast, pan-neuronal expression of UAST-npc2a or UAST-npc2b did not show any rescuing activity.

In addition to cellular defects in cholesterol homeostasis, mammalian NPC mutants have neuronal and behavioral defects, including region-specific neurodegeneration, ataxia, dementia and early death. We examined Drosophila npc2 mutants in detail to search for potential neuronal phenotypes.

Drosophila neurodegenerative mutants are often associated with a short life span and numerous large vacuoles in the brain(Min and Benzer, 1999; Palladino et al., 2002). We assessed the adult life span of npc2a mutants. npc2a mutants displayed a slightly shorter life span compared with wild type(Fig. 7A). For example, by day 52 more than 60% of the wild-type flies were still alive compared with fewer than 10% of the npc2a/npc2a mutants. Fifty percent of the npc2a mutants died by day 36, a time when more than 90% of the wild-type flies remained alive.

We sectioned adult brains from 30-day-old npc2a/npc2a and wild-type animals to look for the presence of large vacuoles indicative of neurodegeneration. We found no evidence of any neurodegenerative vacuoles(data not shown). Reasoning that subtle neurodegeneration may not cause the formation of large vacuoles, we next used TUNEL staining to look for apoptotic cells in adult brains. Compared with wild type, we found few TUNEL-positive cells in 30-day-old npc2a/npc2a mutant brains(Fig. 7B). By contrast, many TUNEL-positive cells were present in 7-day-old npc2a/npc2a;npc2b/npc2b double homozygous brains and in tracheal cells that extended along the top of the brains (Fig. 7B). We double-stained mutant flies with antibodies against the pan-neuronal marker Elav and for TUNEL-positive cells. Most of the TUNEL-positive cells were neurons (Fig. 7C). Similar TUNEL-positive cells, indicative of neurodegeneration, were found in Drosophila npc1a mutants (data not shown). Thus Drosophila npc1/2 mutants faithfully display cholesterol accumulation and neurodegenerative phenotypes analogous to those of mammalian NPC mutants.

Mammalian Npc1 has been found in axons as well as presynaptic nerve terminals, and Npc1/Npc1 mutant mice have mild morphological changes in presynaptic nerve terminal (Karten et al., 2006). For this reason, Synaptotagmin staining of third-instar larvae was performed to examine neuromuscular junction (NMJ)structure and axon morphology. We found no difference in NMJ morphology in npc2a/npc2a mutants, but axonal transport defects were detected at a low frequency (two to three sites per animal). These defects took the form of accumulated Synaptotagmin within axon tracts(Fig. 7C). The significance of this phenotype for neural function remains to be learned.

NPC disease is characterized by aberrant lysosomal storage of cholesterol and other lipids and by massive degeneration of Purkinje neurons in the cerebellum and, to a lesser degree, other neurons. Major intracellular trafficking defects involving at least the late endosomes and lysosomes that contain Npc1 protein are also observed. The link between the trafficking defects, sterol homeostasis defects, and neurodegenerative pathology is still a mystery, and there is considerable debate about which defect is primary,i.e. initiating.

In cells treated with the drug U18666A, which causes a phenotype much like NPC disease, the trafficking defects are readily visible before sterol accumulation (Ko et al.,2001). The trafficking defect may occur earlier than sterol accumulation and compartmentalization in the diseased state as well. The neurodegeneration could then be a consequence of either sterol accumulation or of other outcomes of defective trafficking. Evidence in favor of the latter idea comes from monitoring the degeneration of cerebellar Purkinje neurons in Npc1/Npc1 mutant mice (Ko et al.,2005). The Purkinje cells that die are not those that have the highest cholesterol accumulation. Other outcomes of defective trafficking may therefore kill neurons, such as a failure to transport sterol substrate to ER/mitochondria for steroid hormone synthesis, as we have suggested in a model of proposed cholesterol shortage (Huang et al., 2005).

Here we show that Drosophila npc2a and npc2b play redundant roles in regulating sterol homeostasis and 20E biosynthesis. The mutant phenotypes of npc2a; npc2b double-homozygous mutants support the proposed cholesterol-shortage model. Moreover, the apoptotic neurodegeneration observed in the fly mutants suggests a further similarity to mammalian NPC disease, and opens up the possibility of applying model organism genetics to understanding the disease process more completely and perhaps devising treatments.

A single gene encoding the cholesterol-binding protein Npc2 is present in many eukaryotic species, with the notable exception that a family of Npc2-like proteins arose within insects or their ancestors. The gene structure analysis of the Drosophila npc2-like gene family clearly indicates that the npc2-like genes were formed by multiple rounds of gene duplication. Why do insects have so many Npc2-like proteins and what are their roles?

In general, gene duplication allows the evolution of new gene functions. In that case, one copy can retain the original function of its ancestor and the other can gain new biological functions through further mutation. The prominent sterol accumulation phenotype in many tissues of the npc2amutant, the broad expression of npc2a, and the high degree of sequence identity between Npc2a and human NPC2 compared with the other seven Npc2-like proteins, all suggest that npc2a functions similarly to vertebrate npc2. From that perspective, the mystery is about the roles of Npc2b-h. Our study of npc2b demonstrates that npc2bis especially highly transcribed in trachea, and in that tissue it is partially redundant to npc2a with respect to sterol homeostasis. This is an incomplete answer to the origin of the gene duplications, because it is not clear why two genes are required. Other npc2 genes(npc2c-h) may also function partially redundantly with npc2abecause npc2a; npc2b double mutants have a weaker phenotype than npc1a mutants (larval/pupal lethal versus first-instar lethal). As insects are cholesterol auxotrophs and need external sterol sources for growth(Clark and Block, 1959), it is possible that some of the Npc2-like proteins may be involved in sterol uptake. The pattern of introns in the Drosophila npc2 gene family provides additional insight into their evolution by suggesting a possible sequence of gene duplication events. The intron-less npc2 genes (npc2a, g,h) may have come first, as the vertebrate genes also lack introns. Next to arise would be npc2 genes like npc2b and d that have a single intron in position 1. An additional intron appears at position 2 in npc2c and e, and the most elaborate gene, npc2f,has a third intron in position 3. Alternatively, the ancient gene may have had three introns, and the other genes have been generated by successive loss of introns. As the intron positions in vertebrate NPC2 genes are almost identical to those in Drosophila npc2 genes, one can speculate that they were generated in the same order through evolution.

As a classical lysosomal storage disease, NPC disease is characterized by the accumulation of large amounts of free cholesterol and other lipids in lysosome-like compartments. The search for the causes of this pathology focused mainly on potential cytotoxic effects caused by the accumulation of cholesterol and other lipids (Patterson and Platt, 2004). However, cholesterol-lowering drug treatments did not alleviate NPC disease progression and sometimes made it worse, arguing strongly against the original sterol-excess theory of the disease(Akaboshi and Ohno, 1995; Somers et al., 2001). To elucidate the molecular and cell biology of NPC protein functions, and shed light on the causes of NPC pathology, NPC models have been established in yeast, worms, flies and mice.

Our studies of Drosophila npc2 genes are consistent with the sterol-shortage model proposed previously(Huang et al., 2005). In this model, sterols are trapped in aberrant organelles in NPC mutant cells, and therefore insufficient amounts of sterol reach the ER or mitochondria. In mammals, the lack of sufficient sterol in the ER triggers a homeostatic activation of transcription of genes that encode machinery for the synthesis and import of sterol, thus setting in motion a sustaining cycle of excess sterol, leading to more excess sterol. In flies and mice, the failure to bring sufficient sterol substrate to the ER/mitochondria could deprive cells of the ability to synthesize adequate steroid hormone. The consequences are different between mammals and flies, because the actions of steroids are quite different. In flies the principal steroid hormone is 20E, the molting hormone,so the defect is a failure to molt. In mammals the cerebellar Purkinje neurons are known to produce multiple neurosteroids, although their functions are far from clear (Tsutsui et al.,1999). Npc1/Npc1 mutant mice are deficient in neurosteroids, and administration of supplementary allopregnanolone reduces the symptoms of NPC disease (Griffin et al., 2004). Thus, both fly and mouse NPC mutants are steroid hormone deficient and both mutants can be rescued by exogenous steroid hormone treatment, suggesting strongly that cholesterol and the consequent steroid shortages play a central role in NPC disease.

Our studies reveal a new layer of ecdysteroid biosynthesis regulation, i.e. sterol substrate availability. The regulation of ecdysteroid biosynthesis and the downstream events that mediate ecdysteroid hormone action have been studied continuously for several decades using genetic and biochemical approaches (Gilbert et al.,2002). To date, many genes that affect 20E biosynthesis have been identified and characterized, and these can be grouped into four functional classes. The first class of genes includes upstream factors such as prothoracicotropic hormone (PTTH) that control whether the prothoracic gland should synthesize ecdysone or not. A PTTH mutant has not been isolated in Drosophila, but studies in other insects have clearly demonstrated the essential function of PTTH in ecdysteroid biosynthesis(Gilbert et al., 2002). The larval arrest phenotypes resulting from ablating Drosophila neurons that produce PTTH are consistent with a role in governing ecdysteroid biosynthesis (X.H. and M.P.S., unpublished). The second class of genes consists of the yet-to-be-identified PTTH receptor and the Ras signaling cascade that transduces the PTTH signal. Ras appears to act through its downstream effector Raf to control ecdysteroid biosynthesis(Caldwell et al., 2005). The third class of genes includes nuclear transcription factors and regulators,such as ftz-f1, ecd, woc and mld(Gaziova et al., 2004; Neubueser et al., 2005; Parvy et al., 2005; Wismar et al., 2000). The targets of these proteins are not well defined but may include the fourth class of genes, the Halloween genes (e.g. dib, sad, phm, shd, spo and spo2) that encode p450 enzymes that mediate the conversion of cholesterol to 20E through multi-step reactions in the ER and mitochondria(Chavez et al., 2000; Gilbert and Warren, 2005; Ono et al., 2006; Petryk et al., 2003; Warren et al., 2002).

The present study, together with our previous study on Drosophila npc1a, defines a fifth class of genes functioning to ensure a sufficient supply of sterol substrates for 20E biosynthesis. This class of mutants has intact 20E biosynthetic enzymes, as shown indirectly by our feeding and rescue experiments, but has insufficient sterol substrate for 20E production. Therefore, the ecdysteroid-deficient mutant phenotype can be suppressed by excess cholesterol or 7-dehydrocholesterol, as in npc1a or npc2 (a and b) mutants. Other members of this gene class may include some START domain-containing proteins as well as PBR, which are implicated in transporting sterol into mitochondria for steroid biosynthesis in mammals (Stocco,2001).

We are very grateful to Kaye Suyama and Matt Fish for technical assistance. We thank Dr Hugo Bellen for Synaptotagmin antibody. X.H. was supported by a Walter and Idun Berry Postdoctoral Fellowship. M.P.S. is an Investigator of the Howard Hughes Medical Institute. Research reported here was supported by grants from the Ara Parseghian Medical Research Foundation (M.P.S.), National Basic Research Program of China (973 program) #2007CB947200 and grant#KSCX1-YW-R-69 from the Chinese Academy of Sciences (X.H.), and grant#IBN0130825 from the National Science Foundation (L.I.G. and J.T.W.).