Main

Huntington disease is a neurodegenerative disorder associated with the expansion of a polyglutamine stretch at the amino terminus of the huntingtin protein (Htt) to more than 36 residues1,2. The selective neuronal loss observed in the striatum seems to occur by apoptosis3,4,5. These striatal neurons have altered cytoskeletal neurofilament networks6 and Htt migrates from the cytoplasm to the nucleus as the disease progresses7. The molecular pathways that might explain the strong correlation between increased polyglutamine repeats within Htt and subsequent neurodegeneration are not known. Htt is cleaved by caspases 3 and 6 (ref. 8) and expression of a polyglutamine repeat fragment was shown to cause cell death in primary rat neurons in a caspase-8-dependent manner9. Moreover, relocation of activated caspase-8 to an insoluble neuronal cell fraction was found in patients with Huntington disease9.

Caspase-8 is a member of a family of cysteine proteases involved in the activation and execution of programmed cell death (see ref. 10 for a review). It contains two death-effector domains (DEDs) in its pro-domain11. The DED is a small protein–protein interaction domain that facilitates the assembly of protein components required for the execution of various cell-death pathways12,13. Caspase-8 has been well characterized for its participation in cell death mediated by Fas or the tumour necrosis factor (TNF) receptor11. It is recruited through its DEDs to these receptors at the plasma membrane, where it initiates an apoptotic cascade11.

Several Htt-interacting proteins have been identified in an attempt to understand the molecular basis of Htt-mediated toxicity. One, named Htt-interacting protein 1 (Hip-1), was recently shown to contain a pseudo death-effector domain-like motif (pDED)14. Overexpression of Hip-1 induced cell death in a pDED-dependent manner through activation of caspases14. Hip-1 is a protein of relative molecular mass (Mr) 120,000 (120K), with homology to Sla2p, a protein involved in cortical actin cytoskeleton formation15, vesicle transport16 and endocytosis17,18. In mammalian cells, Hip-1 and Htt were shown to colocalize to clathrin-coated vesicles19. Furthermore, Hip-1 contains consensus sites for direct binding to the clathrin heavy chain and AP2 (ref. 19). As the Htt polyglutamine tract becomes expanded, however, Htt has significantly less affinity for Hip-1 (ref. 20). Taken together, these findings suggest that one consequence of an expanded polyglutamine repeat in Htt is the release of the pDED-containing Hip-1 protein and subsequent induction of a yet-unknown apoptotic pathway.

To explore this possible new pathway, the pDED of Hip-1 was used as bait in a yeast two-hybrid library screen. This identified a novel partner for Hip-1, which we named Hippi (for Hip-1 protein interactor), which contains a pDED domain with striking similarity to that of Hip-1. We show here that a physical interaction between Hippi and Hip-1 occurs that enhances Hip-1-induced cell toxicity. We provide evidence that caspase-8 is recruited to the Hippi–Hip-1 complex and might account for the observed toxicity by launching the apoptotic pathway. The presence of Htt with an expanded polyglutamine repeat favours the formation of Hippi–Hip-1 oligomers, suggesting a potential molecular basis for the pathogenesis of Huntington disease.

Results

Cloning of Hippi.

To determine potential modulators of Hip-1 toxicity, the pDED domain of Hip-1 was used as bait to screen a human brain cDNA yeast two-hybrid library. Many clones recovered encoded the same DED-like motif and represented the C terminus of a new protein that we named Hippi (Fig. 1a). The N-terminal portion of this protein was cloned by carrying out consecutive 5′ rapid amplification of cDNA ends (RACE) PCR on a brain cDNA library. The complete nucleotide sequence of this new gene and the deduced amino-acid sequence have been deposited in GenBank (accession number AF245220). The sequence was confirmed from multiple cDNA library clones containing the full-length open reading frame (ORF) as well as expressed sequence tag (EST) clones (51232, 257718, 610175 and 682684). The Hippi ORF encodes a protein of 429 amino acids. The sequence from amino acid 126 to the C terminus shows 37% identity and 67% similarity with a Caenorhabditis elegans gene of unknown function located on cosmid F59C6 (gp:cef59c6 2) (data not shown). In addition to the pDED domain, the C terminus shows weak similarity to some cytoskeletal proteins such as a Rho-associated kinase Rock-2 (gsp:w56473), centromeric protein E (sw: q02224) and myosin heavy chain (sw:p14105).

Figure 1: Organization of Hippi and Hip-1.
figure 1

a, Schematic diagram of Hippi and Hip-1. Coiled-coil-forming domains are predicted but not shown. Predicted domains, including a leucine zipper, the talin-like domain and the pseudo death-effector domain (pDED) in Hip-1, and the myosin-like domain and the pDED in Hippi are shown. b, c, The pDED of Hippi and its amino-acid sequence homology to other death-effector domains. The entire amino-acid sequences of the Hippi pDED and other known DEDs are shown. The first and last amino acids of each DED in their respective proteins are indicated on the left and right of each sequence. The region corresponding to the fifth helix (α5) is indicated. The arrowhead in c indicates the position of the critical amino acid that distinguishes the first group of DEDs, which have a charged residue (K) at this position, from the second group, which have largely hydrophobic residues (L, V) at the equivalent position.

Full size image

Hippi tissue distribution.

Northern blot analysis revealed that Hippi mRNA is expressed in the brain (Fig. 2a), as previously shown for Hip-1 (ref. 20). Hippi mRNA is present in all regions of the brain but to a lesser extent in the spinal cord (Fig. 2b). It is also abundant in heart, kidney, lung, and to a lesser extent in all other tissues tested (Fig. 2a).

Figure 2: Distribution of Hippi mRNA and Hip-1 and Hippi protein in brain and other tissues.
figure 2

a, Tissue distribution of Hippi mRNA. A northern blot of total human RNA from adult tissues (upper panel) shows an approximately 1.8-kb transcript hybridized with a Hippi probe. The lower panel represents the same blot hybridized with a β-actin probe to confirm RNA integrity and relative loading. Molecular markers are indicated in kb on the left. PBL, peripheral blood leukocytes. b, Regional distribution of Hippi mRNA in human brain. A northern blot of total human RNA from regions of adult brain showing expression of Hippi mRNA (upper panel). β-actin mRNA was used as a loading control (lower panel). c, Tissue distribution of Hip-1 (upper panel; anti-Hip-1 polyconal antibody) and Hippi (lower panel; anti-Hippi polyclonal antibody) assessed by western blot from various normal human tissues (50 μg human protein Medley (Clontech)). Molecular mass markers are on the right.

Full size image

A polyclonal antibody against Hippi protein was produced by immunizing rabbits with a recombinant polypeptide corresponding to the C terminus of Hippi initially isolated from the yeast two-hybrid system. Western blot analysis revealed the presence of an approximately 55K protein in many tissues (Fig. 2c, lower panel). It is likely that the 55K band corresponds to Hippi as it migrates on SDS–PAGE with the polypeptide synthesized from our full-length Hippi cDNA clone (Fig. 2c, lower panel, lane 1). Appreciable amounts of Hippi are detected in brain, thymus, lymph node, lung, liver, skin and kidney. All tissues expressing Hippi protein also express appreciable amounts of Hip-1 with the exception of skin (Fig. 2c, upper panel). Hippi and Hip-1 therefore have a very similar pattern of tissue expression, consistent with a potential functional link between them.

Hippi is expressed in the Golgi apparatus of neurons where it partly colocalizes with Hip-1.

Immunohistochemical studies in mice revealed that Hippi is present in neurons and neuropil throughout the brain. Regions positive for Hippi immunoreactivity included cortex (Fig. 3a), striatum (Fig. 3b), globus pallidus, hypothalamus and cerebellum (data not shown). Within each region, there was some heterogeneity of intracellular Hippi staining; some neurons had Hippi in the extranuclear space only (small arrows in Fig. 3a) and others had Hippi in the nucleus as well (thick arrowhead in Fig. 3a). The extranuclear Hippi staining (brown in Fig. 3c; nucleus counterstained in blue) was perinuclear and patchy, suggesting association with organelles.

Figure 3: Expression of Hippi in mouse brain.
figure 3

Light micrographs showing Hippi immunohistochemical staining in cortex (a) and striatum (b) from a normal adult C57/Bl6 mouse. c, As in a except that the section has been counterstained with cresyl violet to show cell outlines and the time of the immunoperoxidase reaction has been reduced to show the patchy perinuclear Hippi staining better. d, e, Electron micrographs of Hippi subcellular localization. Hippi immunogold particles are mostly found associated with the Golgi apparatus. f, Low-power view of Hippi staining (green) in layer six cortex and lack of staining in corpus callosum (bottom of image). g, Low-power view of staining for the astrocyte marker GFAP primarily in corpus callosum. h, Merge of f and g showing that Hippi staining is predominantly neuronal. i, High-power view of Hippi-stained cortical pyramidal neuron. j, The same neuron labelled with FluoroGold, identifying it as a cortico-striatal projection neuron. k, Merge of i and j showing that a functionally identified cortico-striatal neuron expresses high levels of Hippi protein. l, FluoroGold-labelled striato-pallidal projection neurons. White arrowheads outline the body of a simple striato-pallidal neuron. The same neurons immunolabelled for Hip-1 (m) and Hippi (n). o, Merged image of l, m and n showing that a functionally identified striato-pallidal neuron is expressing high levels of both Hippi and Hip-1 proteins. Scale bars represent 25 μm (a, b); 10 μm (c); 40 μm (h); 5 μm (k); 10 μm (o).

Full size image

Consistent with the preferential distribution of Hippi in neuron-rich structures in the brain, it was abundant in layer 6 of the cortex (Fig. 3h) but absent from the underlying corpus callosum, which is composed mainly of glia and axons with only a few neuronal cell bodies. Electron microscopy revealed that Hippi was mainly found associated with the Golgi apparatus (Fig. 3d) in most neurons. Hippi-specific immunogold particles were also seen free in the cytoplasm or associated with vesicles or plasma membrane (data not shown). Within the Golgi, Hippi was found on the outside of cisternae (Fig. 3e) and mostly in the cis-Golgi. Immunofluorescence studies on human neuronal NT2 cells confirmed the presence of Hippi in the Golgi (Fig. 4). Patchy Hippi immunostaining was seen outside the nucleus (Fig. 4a), and these patches were localized within the Golgi (Fig. 4b, c). Interestingly, Hip-1 is also found around the nucleus in human NT2 cells (Fig. 4d), in a region that stains with the Golgi-specific marker (Fig. 4e, f). Co-immunolabelling of human neuronal NT2 cells for Hippi (Fig. 4g) and Hip-1 (Fig. 4h) confirmed that the proteins are found together in the Golgi apparatus (Fig. 4i), although there are significant regions of non-overlap as well. The overlap calculated by quantitative analysis of the reconstructed (three-dimensional) NT2 cell images was approximately 6%. There was similar partial colocalization of Hippi with Hip-1 in primary rat striatal neurons (Fig. 4j–l).

Figure 4: Subcellular colocalization of endogenous Hip-1 and Hippi in human neuronal NT2 cells.
figure 4

a–c, Colocalization of endogenous Hippi and the Golgi in human neuronal NT2 cells using rabbit polyclonal anti-Hippi (red) (a) and Golgi marker (C5 ceramide) in green (b). Images from a and b are merged in c to show the region of colocalization (yellow). d–f, Colocalization of endogenous Hip-1 and the Golgi in human neuronal NT2 cells using mouse monoclonal anti-Hip-1 (red) (d) and Golgi marker (green) (e). Merged images in f show colocalization (yellow). g–i, Co-immunolocalization of endogenous Hip-1 and Hippi in human neuronal NT2 cells using rabbit polyclonal anti-Hippi (red) (g) and mouse monoclonal anti-Hip-1 (green) (h). The boxed regions in images g and h are merged in i to show colocalization (yellow). The arrow in g indicates the perinuclear structure where Hippi is localized. j–l, Co-immunolocalization of endogenous Hip-1 and Hippi in primary rat striatal neurons using rabbit polyclonal anti-Hippi (red) (j) and mouse monoclonal anti-Hip-1 (green) (k). Merged image in l shows colocalization (yellow).

Full size image

Hippi is expressed in cortico-striatal and striato-pallidal projection neurons.

Hippi is highly expressed in striatal medium spiny neurons (Fig. 3b). In addition, high levels of Htt were identified in cortico-striatal neurons, which are susceptible to death in Huntington disease21,22. We thus wished to know if those neurons also express Hippi. Cell bodies of cortical neurons projecting to the striatum were labelled by retrograde uptake of FluoroGold injected into the striatum (Fig. 3j). The same neurons were found to express Hippi (Fig. 3i, k). Hip-1 has also been previously shown to be present in cortico-striatal neurons (F.G.G., R.S., S.X., C.-A.G., B.R.L., M.M., A.S.H., J.T., J.P.V., V.H., D.M.R., S.R., M.R.H., D.W.N., unpublished observations). Neurons projecting from the striatum to the globus pallidus are the most vulnerable to death in Huntington disease. The cell bodies from these neurons were identified by retrograde uptake of FluoroGold injected into the globus pallidus (Fig. 3l). These same neurons also expressed both Hip-1 (Fig. 3m) and Hippi (Fig. 3n). Tripartite coincidence labelling of these striato-pallidal neurons is shown in Fig. 3o. Hippi and Hip-1 are therefore both present in neurons that are vulnerable to premature death in Huntington disease.

Hip-1 and Hippi may define a new class of DED.

As described above, the Hippi C terminus contains a sequence with similarity to other known DEDs (see Fig. 1b). Hippi pDED shows 39.2% similarity and 26.6% identity to the Hip-1 pDED, and 34.9% similarity and 21.0% identity to other known DEDs. These values are relatively close to the averaged similarity (41.7%) and the averaged identity (29.5%) between the two DEDs in FLIP/Usurpin (a natural inhibitor of caspase-8 activation), caspase-8 and caspase-10. The similarity between Hippi and Hip-1 pDEDs is particularly striking in the putative fifth helix, where they both share the same sequence, SKLKEK (Fig. 1b). The charged residue (lysine; K) at the equivalent position in helix 5 of the pDED of Hip-1, the Hip-1 homologue Hip-12 (refs 23,24) and Hippi contrasts with the equivalent position in other DEDs, where a hydrophobic residue (leucine (L) or valine (V)) is usual (Fig. 1c). Analysis of the nuclear magnetic resonance structure of the FADD DED revealed that this position is exposed on the outside of the protein and might represent a critical point of contact between two DEDs25.

Specificity determinants within Hippi and Hip-1 pDEDs.

We wished to confirm that the portion of Hippi responsible for its interaction with Hip-1 in the two-hybrid system is the C-terminal pDED. Removal of pDED severely compromised Hippi's ability to bind Hip-1 pDED (Fig. 5a; Hippi ΔDED). The Hippi pDED alone also interacts efficiently with the Hip-1 pDED, but not with caspase-8 DED-A or caspase-8 DED-B (Fig. 5a and data not shown). It is thus possible that the K residue found in helix 5 of Hip-1 and Hippi pDEDs but not in other DEDs could be critical for their interaction and for specificity in pDED binding interactions. We mutagenized the K at position 409 in the Hippi pDED to L, the residue found in most other DEDs, and observed a fivefold reduction in the ability of the K409L Hippi mutant to interact with wild-type Hip-1 (Fig. 5b). Conversion of the corresponding Hip-1 K to L fully restored the ability of the K409L Hippi mutant to interact with Hip-1 (Fig. 5b). The single K-to-L mutation in Hip-1 or Hippi was not sufficient to allow interaction with either of the caspase-8 DEDs (data not shown), suggesting that other structural differences might prevent the interaction. The K at position 409 in Hippi pDED was also mutagenized to aspartic acid (D), the residue found in the Fas/Apo-1/CD95 death domain at the equivalent position26. The K409D Hippi mutant had reduced ability to interact with wild-type Hip-1 (Fig. 5b). The interaction was fully restored when both pDEDs contained a D at that position (Fig. 5b). The presence of a K at that critical position in helix 5 of Hip-1 and Hippi pDEDs, but not in other known DEDs, might thus define them as members of a new subclass of DED. Members of that subclass would specifically interact with each other but not with members of the other subclass owing, in part, to the selectivity conferred by this key helix-5 residue.

Figure 5: Structural requirements for interaction of Hip-1 and Hippi.
figure 5

a, Hip-1 pDED interacts with Hippi pDED in yeast. Liquid-culture yeast two-hybrid assays with o-nitrophenyl-β-D-galactopyranoside (ONPG) as β-galactosidase substrate were carried out in Y190 yeast expressing the Gal4 DB–Hip-1 pDED fusion protein and the indicated Gal4 AD fusion proteins. The level of interaction between the two fusion proteins is expressed in units of β-galactosidase activity. A representative experiment (n = 3) is shown. Standard deviation (error bars) is calculated for each strain from triplicate samples. b, The helix-5 compatibility residue in the pDED of Hip-1 and Hippi determines their level of interaction. The same procedure as in a was used to measure the level of interaction between Hip-1 pDED and Hippi pDED fusion proteins carrying various residues at the critical position indicated in Fig. 1b. The top line under the histogram indicates the amino acid present in Hip-1 pDED and the lower line that present in Hippi pDED. A representative experiment (n = 4) is shown. Standard deviation (error bars) is calculated for each strain from triplicate samples.

Full size image

Hippi interacts with Hip-1 in cells.

We next carried out co-immunoprecipitation experiments to determine whether Hippi could physically interact with full-length Hip-1 in cells (Fig. 6a). In transfected HeLa cells expressing Hippi, immunoprecipitation of endogenous Hip-1 brought down Hippi (Fig. 6a, blot II, lane 2). Similarly, in cells transfected with Hip-1, low but detectable amounts of Hip-1 were observed in endogenous Hippi immunoprecipitates (Fig. 6a, blot I, lane 4). Significantly greater amounts of co-immunoprecipitated Hip-1 or Hippi were found when both Hip-1 and Hippi were co-transfected in the same cells (Fig. 6a, blots I and II, lane 5). Removal of the Hippi pDED slightly reduced (by approximately twofold) Hippi's ability to co-immunoprecipitate with Hip-1 (Fig. 6a, blots I and II, lane 6). The effect of the removal appeared more severe (> 4-fold reduction) when transfected Hippi lacking the pDED was co-immunoprecipitated with endogenous Hip-1 (Fig. 6a, blot II, lane 3). The observed effects were not due to a difference in the amount of Hip-1 or Hippi expressed in the various permutations (Fig. 6a, bottom two panels).

Figure 6: Interaction of Hippi and Hip-1 in cells.
figure 6

a, Association of Hip-1 and Hippi in HeLa cells. The indicated Hip-1 and Hippi constructs were transiently expressed in HeLa cells that were then lysed in NP-40-containing buffer. Hip-1 in Hippi immunoprecipitates was assessed by immunoblotting with anti-Hip-1 antibodies (blot I) and Hippi in Hip-1 immunoprecipitates (IP) was detected with anti-Hippi antibody (blot II). The two lower panels correspond to cell lysates directly loaded on the gel and blotted with Hip-1 or Hippi antibodies. Molecular mass markers are on the right. b, Structural requirements for association of Hip-1 with Hippi in 293 cells. Various Hip-1 and Hippi constructs were transiently expressed in 293 cells. Hip-1 in Hippi immunoprecipitates was assessed by immunoblotting with anti-Hip-1 antibodies (top panel). Expression of equivalent amounts of Hip-1 and Hippi mutant proteins was confirmed by western blot on 50 μg cell lysate using anti-Hip-1 (middle panel) or anti-Hippi (bottom panel) antibodies.

Full size image

Structural requirements for the interaction between Hip-1 and Hippi.

Because the removal of the Hippi pDED did not completely abolish its interaction with Hip-1, we looked for other regions in the proteins that contributed to binding. A series of Flag-tagged Hippi constructs were coexpressed with various Hip-1 deletion mutants in human embryonic kidney 293T cells (293 cells) (Fig. 6b). The Hippi protein was Flag-tagged at its N terminus to allow immunoprecipitation experiments to be carried out without interference by our anti-Hippi antibodies, which recognize a significant portion of the Hippi C terminus. In this cell system, appreciable amounts of Hip-1 were recovered in Hippi immunoprecipitates (Fig. 6b, lane 3). (Note that Hip-1 often appears as a complex of three distinct bands in SDS–PAGE. The two lower bands are likely to result from post-lysis degradation or are Hip-1 isoforms24.) As observed in HeLa cells, deletion of the Hippi pDED reduced Hippi's ability to interact with Hip-1 in 293 cells (Fig. 6b, lane 4). The removal of the C ter minus of Hippi, corresponding approximately to the last third of the polypeptide, completely abolished its ability to interact with Hip-1 (Fig. 6b, lane 5). This C-terminal region of Hippi contains two subregions: the pDED at the extreme C terminus, and another region proximal to the pDED, which we have indicated as the myosin-like domain (MLD), which represents most of the similarity with the cytoskeletal proteins mentioned previously (see Fig. 1a). Although we cannot rule out the possibility that the MLD interacts directly with Hip-1, it is also possible that it is necessary for the proper subcellular localization of Hippi, bringing it close to Hip-1 and thus facilitating pDED interactions. Similarly, the complete removal of the Hip-1 pDED did not abolish its ability to interact with Hippi (Fig. 6b, lane 6). Deletion of the C-terminal talin-like domain of Hip-1, however, dramatically reduced its ability to interact with Hippi (Fig. 6b, lane 7). This C-terminal domain has homology to cytoskeletal proteins such as talin27 and ERM (ezrin, radixin, moesin) proteins28,29,30, which interact with the actin cytoskleleton28,31,32. The talin-homology domain of Hip-1 might also be important for its proper subcellular localization and proximity to Hippi. Note that the residual interaction between Hippi and Hip-1 lacking the talin-like domain completely disappeared upon removal of the Hippi pDED (Fig. 6b, lane 9). This suggests that the residual interaction was probably mediated through Hip-1 and Hippi pDEDs. Together, these results suggests that the C terminus of both proteins is essential for their capacity to associate with one another in cells, and that, in addition to mutual pDED binding, this interaction may be facilitated by structures outside the pDEDs.

Mutant forms of Htt modify the extent of interaction between Hip-1 and Hippi.

We next determined whether Hip-1 and Hippi are found in a multiprotein complex with Htt, or whether Hip-1 preferentially interacts with Hippi once it dissociates from Htt. It has been shown that Hip-1 binds tightly to normal forms of Htt (for example, Htt-Q(15) with 15 glutamine (Q) repeats; the disease-probability threshold being ≥ 35 glutamine repeats), but very inefficiently to disease-associated forms of Htt with polyglutamine expansion (for example, the pathogenic Htt-Q(128) with 128 repeats)20. In addition, it was shown that Hip-1 expressed in cells containing mutant Htt (Htt-Q(128)) was substantially more cytotoxic than in cells harbouring wild-type Htt (Htt-Q(15)), consistent with the hypothesis that Htt polyglutamine expansion makes Hip-1 more available for other interactions which result in apoptosis14. We therefore compared the amount of Hip-1 interacting with Hippi in 293 cells that were stably expressing mutant (Htt-Q(128)) and wild-type (Htt-Q(15)) constructs (corresponding to the N-terminal 548 residues of the Htt protein, which contains the polyglutamine-expansion region, the caspase cleavage cluster and the Hip-1-interacting domain). The stable cell lines were then transiently transfected with Hip-1 and Hippi and co-immunoprecipitation was carried out (Fig. 7). A significant and reproducible decrease in the ability of Hippi to interact with Hip-1 was observed in Htt-Q(15) cells compared to Htt-Q(128) cells (Fig. 7a). The average of four independent experiments indicates that the interaction between Hip-1 and Hippi in Htt-Q(15) cells is reduced by 47%, whereas in Htt-Q(128) cells it is increased by 45% (Fig. 7b), presumably due to excess available Hip-1. The transfected Htt-Q(128) may recruit endogenous Htt into cellular aggregates and liberate the associated Hip-1. Similar levels of Htt, Hippi and Hip-1 were expressed in all conditions. The presence of mutant Htt with an expanded polyglutamine stretch thus favours the association between Hippi and Hip-1.

Figure 7: The length of the Htt polyglutamine stretch governs the availability of Hip-1 to bind to Hippi.
figure 7

a, 293 cells stably expressing Htt with polyglutamine repeats of various lengths (15 or 128 repeats) were transiently transfected with Hip-1, Hippi or both. Co-immunoprecipitation of Hip-1 with Hippi in various conditions was examined by western blot using an anti-Hip-1 antibody. Expression of equivalent amounts of Hip-1, Hippi, and Htt mutants was confirmed by western blot on 50 μg cell lysate. Molecular mass markers are on the right. b, The intensity of the signal corresponding to the amount of Hip-1 co-immunoprecipitated with Hippi from 293 cells overexpressing wild-type Htt-Q(15) or mutant Htt-Q(128) was quantified by laser densitometry and compared to the signal from cells with endogenous level of Htt. A reference value of 100% binding was given to the level of interaction between Hip-1 and Hippi in 293 cells expressing only endogenous Htt (first column). The values are the average of four independent experiments and are significant, with P < 0.003 in t-tests.

Full size image

Hippi potentiates the pro-apoptotic effects of Hip-1.

It has been shown that expression of Hip-1 in cells results in apoptotic death14. We thus investigated whether coexpression of Hippi and Hip-1 would modify the extent of apoptosis seen with Hip-1 alone. 293T cells were transiently transfected with Hip-1 and Hippi, collected 48 h later and the extent of apoptosis evaluated by quantifying group II (effector) caspase catalytic activity (Fig. 8a) and group III (initiator) caspase activity (data not shown) and by morphological cell-death criteria (data not shown). Hip-1, but not Hippi, caused some cell death in this system (Fig. 8a, columns 3 versus 2). Coexpression of Hippi with Hip-1 tripled the toxicity seen with Hip-1 alone (Fig. 8a, column 4). The potentiating effect of Hippi depends on both the pDED and the MLD (Fig. 8a, columns 5 and 7). The pDED of Hip-1 was also necessary for its cytotoxic contribution (Fig. 8a, column 8). The talin-like domain of Hip-1 was necessary for the potentiating effect of Hippi, but did not affect the basal level of cytotoxicity (Fig. 8a, column 9). Therefore, for each protein, both domains that are important for their physical interaction are also critical for their combined pro-apoptotic cytotoxicity. We also observed that the K409L mutation in Hippi pDED, which abolishes its interaction with Hip-1 pDED (see Fig. 5b), completely abrogated Hippi's potentiating effect on Hip-1 toxicity. The cooperative effect of Hip-1 and Hippi in inducing cell death was also seen in post-mitotic, primary rat striatal neurons (Fig. 8b). Hippi and Hip-1 alone induced some cell death (whereas only Hip-1 did in fibroblasts) and together they resulted in substantial cell loss (around 75%). The ability of Hippi alone to mediate cell death may be due to the higher levels of endogenous Hip-1 in neuronal cells.

Figure 8: Hippi increases Hip-1-mediated toxicity in cells.
figure 8

a, 293T cells were transfected with the indicated Hippi and Hip-1 constructs and apoptosis was estimated by the extent of DEVDase activation of caspase. WT, wild type; FU, fluorescence units. b, Striatal neurons derived from E18 rat embryos were transfected with the indicated constructs plus pEGFP (at one-fifth the concentration of the Hippi/Hip-1 vectors). Cell survival was assessed by counting EGFP-positive neurons remaining on the dish by either confocal microscopy (data illustrated) or by flow cytometry after washing and detachment (data not shown).

Full size image

Caspase-8 associates with Hip-1 and Hippi.

A central role for caspase-8 in Huntington disease was recently proposed9. Because caspase-8, Hip-1 and Hippi all contain DED-like domains, we were interested to test whether caspase-8 could be recruited to the Hippi–Hip-1 complex and potentially contribute to Hippi–Hip-1-mediated cell death (Fig. 9). As caspase-8 is distributed widely in cells, it might participate in non-receptor-mediated apoptotic events. In NT2 neuronal precursor cells, for example, we found that endogenous Hippi, Hip-1 and caspase-8 were all present in the microsomal fraction, and that some of these molecules were pre-associated in a sub-apoptotic tripartite complex (Fig. 9a). When procaspase-8 was transfected into HeLa cells, it co-immunoprecipitated with endogenous Hippi (Fig. 9b, lane 2). Overexpression of caspase-8 and Hippi together did not lead to further increases in the amount of caspase-8 found in the Hippi immunoprecipitate (data not shown), indicating that an intermediary endogenous limiting factor may be necessary for the association of caspase-8 with Hippi. Removal of the caspase-8 pro-domain considerably reduced its ability to co-immunoprecipitate with Hippi (Fig. 9b, lane 4). Taking into account the high level of expression of the caspase-8 pro-domain deletion mutant compared to wild-type caspase-8 (Fig. 9b, lower panel), we estimate that caspase-8 lacking its pro-domain is 40-fold less efficient at interacting with Hippi than is wild-type procaspase-8. Transfected Hippi was also found in endogenous caspase-8 immunoprecipitates (Fig. 9c, lane 3) and Hippi lacking its pDED was still able to interact with endogenous caspase-8 (Fig. 9c, lane 4). In contrast, a Hippi deletion mutant lacking its whole C terminus, and previously shown not to interact with Hip-1 (see Fig. 6), was unable to co-immunoprecipitate with endogenous caspase-8 (Fig. 9c, lane 5). This suggests that Hippi's interactions with caspase-8 require Hip-1. Overexpressed Hip-1 was found in endogenous caspase-8 immunoprecipitates (Fig. 9d, lane 3). But overexpression of both caspase-8 and Hip-1 did not lead to increased amounts of Hip-1 in the caspase-8 immunoprecipitate (data not shown). This once again suggests that the Hippi–Hip-1 intereaction with procaspase-8 depends on an endogenous as-yet-unidentified intermediate. A deletion mutant of Hip-1 lacking its pDED interacted less efficiently with procaspase-8 compared to wild-type Hip-1 (Fig. 9d, lane 4). Removal of the talin-like domain in Hip-1, which is critical for the Hippi–Hip-1 interaction, did not influence Hip-1–caspase-8 interaction (Fig. 9d, lane 5). Together, these results indicate that, first, procaspase-8 can indirectly interact with the Hippi–Hip-1 complex through its pro-domain; second, the Hip-1 pDED is necessary for the recruitment of procaspase-8 to the complex; and third, that the Hippi interaction with procaspase-8 requires Hip-1.

Figure 9: Hip-1 and Hippi associate with caspase-8 and are dependent on it to mediate cellular toxicity.
figure 9

a, Microsomal membranes were isolated from cultured NT2 neuronal precursor cells by differential centrifugation. Endogenous Hip-1 was immunoprecipitated using anti-Hip-1 monoclonal antibody and Hippi and caspase-8 (C8) in the Hip-1 immunoprecipitates were determined by immunoblotting (lanes 2, 4). Controls (lanes 1, 3) were processed identically, except that the protein G–sepharose beads did not contain Hip-1 antibody. b, Caspase-8 (C8), Hippi and caspase-8 lacking its pro-domain (Δpro-C8) were transiently overexpressed independently in HeLa cells. Caspase-8 in Hippi immunoprecipitates was assessed by immunoblotting with anti-caspase-8 antibody (top panel). Expression of the various mutants was evaluated by western blotting the total lysate with anti-caspase-8 antibody (lower panel). Molecular mass markers are on the right. c, Caspase-8 (C8), wild-type Hippi, Hippi lacking its pDED (ΔDED), or Hippi lacking the C terminus (myosin-like domain (MLD) and the death-effector domain (DED) were transiently overexpressed independently in HeLa cells. The presence of Hippi in caspase-8 immunoprecipitates was assessed by immunoblotting with anti-Hippi antibody (top panel). Equivalent expression of the various mutants was confirmed by western blotting the total lysate with anti-Hippi antibody (lower panel). Molecular mass markers are on the right. d, Caspase-8 (C8), wild-type Hip-1, Hip-1 lacking its DED (ΔDED), and Hip-1 lacking its talin-like domain (Δtalin) were transiently overexpressed independently in HeLa cells. The presence of Hip-1 in caspase-8 immunoprecipitates was assessed by immunoblotting with anti-Hip-1 antibody (top panel). Equivalent expression of the various mutants was confirmed by western blotting the total lysate with anti-Hip-1 antibody (lower panel). Molecular mass markers are on the right. e, Apoptosis in HeLa cells transfected with the indicated permutations was assessed as indicated in the legend of Fig. 8. Typical cellular morphology is shown as insets. DNC8 indicates dominant-negative caspase 8. f, Established apoptotic pathways and their mediators.

Full size image

Hippi–Hip-1-mediated toxicity occurs through a caspase-8-dependent pathway.

As caspase-8 was found to interact physically with the Hippi–Hip-1 complex, we were interested to find out whether caspase-8 has a role in Hippi–Hip-1-mediated cellular toxicity (Fig. 9e, f). Apoptosis can be initiated through either the 'extrinsic' caspase-8-dependent pathway, which predominantly facilitates 'death receptor' signalling, or through the 'intrinsic' caspase-9-dependent pathway, which mediates death signals that are processed through the mitochondrion (Fig. 9f)33. The pro-apoptotic effects of the Hippi–Hip-1 complex (Fig. 9e, column 3) were attenuated by inhibitors of the extrinsic pathway, including a dominant-negative procaspase-8 mutant (procaspase-8 with an inactivated catalytic cysteine) and FLIP/Usurpin (Fig. 9e, columns 4, 5). In contrast, Bcl2, an inhibitor of the intrinsic pathway, did not affect the pro-apoptotic effects of Hippi–Hip-1 (Fig. 9e, column 6). These results are consistent with the observation that Hippi and Hip-1 associate with procaspase-8 in cells, and show that the pro-apoptotic effects of the Hippi–Hip-1 complex are mediated through the caspase-8 cell-death pathway.

Discussion

The molecular link between polyglutamine expansion in Htt and the consequent apoptotic death of striatal neurons in Huntington disease has been elusive. The pDED-containing protein Hip-1 has been shown to interact less efficiently with Htt bearing a long polyglutamine repeat20. Hip-1 itself has also been shown to induce cellular apoptosis when overexpressed in cells14. A current model of the molecular basis of Huntington disease is that polyglutamine expansion in Htt causes release of the pro-apoptotic protein, Hip-1, which would be toxic to sensitive striatal neurons. In this study, we have shown that a second protein, Hippi, is a molecular accomplice of Hip-1 that assists in engagement of the caspase-8 cell-death pathway. This molecular framework could provide the mechanism by which Htt polyglutamine expansion enhances the probability of neuronal apoptosis.

Many features of Hippi indicate that it might be a cellular partner for Hip-1. Both proteins have a domain with strong homology to other cytoskeletal proteins (the talin-like domain in Hip-120 and the myosin-like domain in Hippi) and these domains are essential for the interaction between the two proteins (Fig. 6). Both proteins bear a pDED that has the potential to interact with the other, but is incompatible with the classical DEDs found in caspases 8 and 10, FADD and FLIP/Usurpin (Fig. 1b). Hip-1 and Hippi proteins have very similar patterns of tissue distribution (Fig. 2) and are abundant in CNS neurons (Fig. 3)20. Both proteins partially colocalize in the Golgi apparatus in human neuronal cells (Fig. 4). They have the potential to directly interact in cells (Fig. 6) and this interaction is modulated by the presence of Htt (Fig. 7). Finally, they can act synergistically to induce activation of caspases and eventual cell death through caspase-8 (Figs 8, 9).

It was interesting to observe that although deletion of the pDED in Hip-1 or Hippi had only a modest effect on their ability to interact with each other, it totally prevented the synergistic cytotoxic effect seen when the proteins are coexpressed. This supports the possibility that the Hippi–Hip-1 complex serves as a docking site for recruiting other DED-containing proteins and initiating cell death. According to our yeast two-hybrid results, the Hip-1 pDED is unable to recruit caspase-8 directly, but it could well recruit procaspase-8 through an adaptor molecule that would contain both types of DEDs. This hypothesis is supported by the finding that caspase-8 is associated with the Hip-1–Hippi complex in a Hip-1 pDED-dependent but indirect manner. We have not yet been able to identify this putative adaptor polypeptide.

We propose the following working model as one component of the molecular basis of Huntington disease (Fig. 10). In healthy neurons, the large protein Htt acts as an anti-apoptotic factor, as suggested by the recent observations that conditional inactivation of Htt in mice leads to neuronal degeneration34, that wild-type Htt reduces the toxicity of mutant Htt in vivo35 and that overexpression of wild-type but not mutant (polyglutamine-expanded) Htt can reduce cell toxicity induced by Hip-1 (ref. 14). The anti-apoptotic effect could thus arise from the ability of wild-type Htt to interact stably with Hip-1, precluding the interaction of Hip-1 with Hippi. Polyglutamine expansion in the N terminus of Htt leads to the release of Hip-1 and the manifestation of its pro-apoptotic potential. Unbound Hip-1 would be free to interact with Hippi, which could enhance the toxicity of Hip-1 in sensitive striatal neurons, thus contributing to the progression of Huntington disease. A central role for caspase-8 in Huntington disease has previously been suggested9, which is entirely consistent with our observation that the Hippi–Hip-1 complex recruits procaspase-8 and initiates apoptotic cell death through this pathway. Activation of this pathway would activate proteolytic enzymes, such as caspase-3, resulting in initial cleavage of Htt8,36 to yield a truncated product that would enter the nucleus, further aggravating cellular toxicity through its modulation of transcription37. The Hippi–Hip-1 interaction could therefore act as a cytoplasmic trigger upstream of caspase-3 activation, initiating the cascade of molecular events that eventually leads to cell death.

Figure 10: Proposed model for Hippi/Hip-1 mediated apoptosis.
figure 10

Hip-1 is normally bound and sequestered by wild-type Htt (polyglutamine (Q) < 35 residues). Expansion of the Htt polyQ stretch to > 35 residues results in lower affinity for Hip-1, freeing Hip-1 to bind to Hippi. Hippi and Hip-1 form a complex that sequesters procaspase-8 and activates it, presumably by proximity-induced autocatalytic maturation, leading to apoptosis.

Full size image

We have thus identified a new apoptotic pathway involving proteins that are modulated by the state of polyglutamine expansion in the Htt protein. This pathway might account for a substantial part of the propensity of vulnerable neurons to undergo apoptotic death in Huntington's neurodegeneration.

Methods

Gal4–Hip-1-pDED cDNA construct

A Hip-1 pDED cDNA construct encompassing amino acids 280–375 (according to the published Hip-1 sequence20) was generated by PCR using the following synthetic oligonucleotides: 5′-AGCTGGATCCTGCAGCTGAAGGGCCACG TC-3′ and 5′- GCATGGATCCCATGGACACCTGTTTGGTCAC-3′. These primers created a Hip-1 pDED-containing fragment that was fused in-frame into the BamHI site of the Gal4 DNA-binding domain (BD) of the yeast two-hybrid vector pAS2-1 (Clontech, Palo Alto, CA).

Screening for Hippi

This vector was then used as a bait to screen a pre-transformed human brain Matchmaker cDNA library (Clontech). The pAS2-1 vector expressing the Hip-1 pDED–BD fusion protein was transformed into yeast strain PJ69-2A using the Yeastmaker yeast transformation system (Clontech) (lithium acetate method). Expression of the fusion protein used as bait was confirmed by western blot using a monoclonal antibody against the Gal4 BD (Clontech). Expression of the Hip-1 pDED–BD fusion protein in strain PJ69-2A gave no background growth; thus there was no need for 3-amino-1,2,4-triazole when grown on SD–Trp–His medium. Strain PJ69-2A expressing the Hip-1 pDED-BD fusion protein could thus be mated with the pre-transformed brain cDNA library (cloned in the pACT2 vector as a fusion with the Gal4 activation domain (AD) and transformed into the yeast mating strain Y187). The diploids were then spread on SD –Trp –Leu –His medium. After 4 days, some positive colonies were picked, purified on the same solid medium and assessed for β-galactosidase (β-Gal) activity using the β-Gal colony lift assay. Plasmids were then isolated from His+ β-Gal+ colonies using the Yeastmaker yeast plasmid isolation kit (Clontech). The pACT2-derivative plasmids isolated were then retransformed into the original PJ692A-pAS2-1 Hip-1 pDED–BD strain for confirmation of true positives. About 25 of the positive clones turned out to contain the same DNA fragment (that is, of the same length and same restriction pattern).

DNA sequencing, 5′ RACE and isolation of cDNA

Positive cDNA clones were sequenced with the Gal4 AD sequencing primer (Clontech) using the Big Dye Cycle Sequencing Dye Terminator kit and ABI 373 Stretch Sequencer (Applied Biosystems, Foster City, CA). All the clones sequenced had the same sequence out of frame with Gal4 AD. Surprisingly, we detected in those yeast clones a fusion protein of the right size (ORF + AD) using an anti-Gal4 AD monoclonal antibody (Clontech). Cloning the ORF back in frame with the AD led to the synthesis of more fusion protein of the right size, better growth on –His medium and stronger β-Gal activity. This suggests that the initially selected positive clones were able to execute frameshifts to produce enough of the appropriate fusion protein to allow two-hybrid interaction and growth of the mated yeast. The sequence turned out to be the C-terminal end of a new gene product (Hippi). Subsequently, primer walking (5′ RACE) was used to determine the remaining upstream sequence. The following primers were then used in a reverse transcription PCR (RT–PCR) reaction to clone full-length Hippi from human brain RNA: p1463 5′-GCTCTAGACTCGAGCCATGGCCGCAGAGGTCT GTG GGCCTGAGCCCACGCTGG-3′; and p1475 5′-GCTCTAGACTCGAGCCAT GGTTAATAAAAGCCTGTTGCTGGTTCTGGAATAACTGTGGC-3′. The 1.6-kb cDNA fragment was then cloned into the XbaI site of the pBluescript SK II vector (UMP 6014) for sequencing as described above.

cDNA library screening for Hippi

In an attempt to clone full-length Hippi and identify splice variants, clones were isolated by screening a lambda TriplEx mouse kidney cDNA library (5′Stretch plus, Clontech) on nylon membranes (Roche Molecular Biochemicals, Indianapolis, IN). The Hippi probe, composed of the largest ORF known (from Met 1 to Tyr 429), was generated by PCR and labelled with alkaline phosphatase using the AlkPhos Direct kit (Amersham Pharmacia Biotech, Piscataway, NJ). Hybridization was carried out overnight at 55 °C with the buffer provided in the AlkPhos kit. Positive clones, identified by chemiluminescence and autoradiography with the CDP-Star reagent, were purified through multiple rounds of dilution plating. Escherichia coli expressing Cre recombinase were infected with the lambda clones and the plasmids containing positive clones were isolated by in vivo excision and sequenced as above.

Determination of additional 5′ HIP-1 sequence

5′ RACE was carried out using a human brain Marathon-Ready cDNA library (Clontech) to determine whether 5′ coding sequence additional to that given for accession number U79734 existed. Primary 5′ RACE used a gene-specific primer 1 (5′-GGTTGCTGGAGCGGTAGAACAG-3′) and the supplied Marathon-Ready cDNA specific primer, Adaptor Primer-1. The primers used for the secondary 5′ RACE were a nested gene-specific primer 2 (5′-CTTTCAACTTTGTAAACTGCTCC-3′) and the supplied Adaptor Primer 2. Advantage Klentaq polymerase was used for both the primary and secondary reactions (Clontech). The resulting PCR product was TA-ligated into vector pCR2.1 and transformed into 'One Shot' cells (TA Cloning Kit, Invitrogen). Sequencing was carried out and, following verification, the new sequence (accession number AF365404) was used to make full-length HIP-1.

Full-length HIP-1

Full-length HIP-1 was reverse transcribed from human whole brain poly(A)+ RNA (Clontech) using the Gene Amp RNA PCR system (Perkin Elmer, Markham, Ontario, Canada). The synthetic primers were designed to contain flanking EcoRV–XhoI–NdeI nested restriction sites suitable for ligation into various commercial vectors. The amplimers were 5′-GCGATATCCTCGAGCATATGAAG CAGGTGCCCAACCCACTGCCCAAGGTGCTGAGC-3′ (forward) and 5′-GCGATATCCTCGAGCTATTCTTTTTCGGTTACCAC TTCTTGCAGTGTAGG-3′ (reverse) and the polymerase was Expand HiFi (Roche Molecular Biochemicals). The resulting fragment was subsequently purified from a preparative agarose gel, trimmed with XhoI (Roche Molecular Biochemicals), ligated into the XhoI site of pBluescript II SK+ and transformed into E. coli XL2-Blue cells (Stratagene, La Jolla, CA). The resulting construct was designated [full-length HIP-1 Met 1–Glu 1030]:[pBII SK+]:[XhoI]:[T7]:[MF-UMP#6027] ([construct]:[vector]:[insert site]:[sense orientation]:[identifier]).

Constructs

An amino-terminally Flag-tagged version of Hippi was created by PCR on UMP 6014 (see above) using the following primers: p1498 5′-TCAGAATTCTGC GATGGACTACAAGGACGACGATGACAAGAC TGCTGCTCTGGCCGTCGTCACGACGTCG-3′ and an internal reverse primer (p1392) 5′-CCAATATATTTCAATGCTTCTTC-3′. An EcoRI–BspEI fragment from the PCR product was then exchanged in the initial vector (UMP 6014) and the corresponding region was fully sequenced to ensure that no mutation had been introduced.

Clones were generated as above by PCR-mediated template modification and were fully sequenced before use. Designations are in the following format: [construct]:[vector]:[insert site]:[identifier] Clones for two-hybrid assays were as follows : [Hippi E142-Y429]: [pACT2]: [EcoRI–XhoI]: [MF UMP#5857], [ΔDED Hippi E142-E341]: [pACT2]: [EcoRI–BamHI]: [MF UMP#7441], [Hippi K409L]: [pACT2]: [EcoRI–XhoI]: [MF UMP#5858], [Hippi K409D]: [pACT2]: [EcoRI–XhoI]: [MF UMP#5859], [Hippi pDED R331-Y429]: [pACT2]: [EcoRI–XhoI]: [MF UMP#5922], [Caspase-8 DEDa F3-E85]: [pACT2]: [EcoRI–XhoI]: [MF UMP#5915], [Hip-1 pDED Q281-M375]: [pAS2-1]: [BamHI]: [MF UMP#5630], [Hip-1 pDED K351L]: [pAS2-1]: [BamHI]: [MF UMP#5948], [Hip-1 pDED K351D]: [pAS2-1]: [BamHI]: [MF UMP#5949]. Clones for expression in mammalian cells were as follows: [Hippi-Flag M1-Y429+flag]: [pBI]: [EcoRV]: [MF UMP#7273], [Hippi ΔDED-flag M1-E341+Flag]: [pBI]: [EcoRV]: [MF UMP#7272], [Hip-1 M1-E1049]: [pBI]: [SalI]: [MF UMP#7236], [Hippi-Flag/ Hip-1]: [pBI]: [EcoRV–SalI]: [MF UMP#7278], [Hippi ΔDED-Flag/ Hip-1]: [pBI]: [EcoRV–SalI]: [MF UMP#7274], [Flag-Hippi flagM1-Y429]: [pCEP]: [HindIII–NotI]: [MF UMP#7012], [Flag-Hippi ΔDED FlagM1-E341]: [pCEP]: [HindIII–NotI]: [MF UMP#7831], [Flag-Hippi ΔC FlagM1-L241]: [pCEP]: [PvuII–NotI]: [MF UMP#7704], [Flag-Hippi ΔMLD FlagM1-E244+E330-Y429]: [pCEP]: [HindIII–NotI]: [MF UMP#7762], [Flag-Hippi K409L]: [pCEP]: [HindIII–NotI]: [MF UMP#7111], [Hip-1]: [pCEP]: [XhoI]: [MF UMP#7693], [Hip-1 ΔDED]: [pCEP]: [XhoI]: [MF UMP#7694], [Hip-1 Δtalin]: [pCEP]: [XhoI]: [MF UMP#7695]

Generation of Hippi polyclonal antibodies

The original Hippi ORF cloned by the two-hybrid system (amino acids 141–429) was used as a template to generate a fragment by PCR that was clonable into the pET11d vector with NcoI ends. The following synthetic oligonucleotides were used: 5′-TGCTACCATGGGCGGCCGCGTCGAAGAAC-3′ and 5′-TCTGACCATGGTTAATAAAAGCCTGTTGC-3′. The fragment was then cloned into NcoI in the bacterial expression vector pET11d. Protein expression was induced with isopropylthiogalactoside (IPTG) and purified as described elsewhere38. The purified protein was subsequently used to immunize rabbits and generate polyclonal antibodies (Covance, Denver, PA).

Purification of antibodies

Twelve lanes of a 8% SDS–PAGE were each loaded with 20 μg purified Hippi protein obtained from bacteria as described above. The proteins were then transferred to PVDF membrane by electroblotting. The membrane was cut into small pieces where Hippi (amino acids 141–429) polypeptide migrates at about 45K. These small pieces of PVDF were then blocked for 30 min in a BSA-WB solution (3% BSA in wash buffer (WB): 10 mM Tris-HCl pH 7.6, 0,5% NP-40, 150 mM NaCl). PVDF pieces were then washed twice for 4 min with WB. The PVDF membrane was then incubated overnight on a rotating wheel at 4 °C in 30 ml of a 1:20 dilution of the Hippi serum in WB. The membranes were then washed twice for 5 min in WB and three times with WB without NP-40. Antibodies were eluted by putting the membranes in 1.5 ml of a 0.2 M glycine pH 2.8 solution and frequent vortexing for 2 min. A Tris-HCl pH 12 solution (105 μl) was added to bring the pH to 7–8. The elution step was carried out twice and the two fractions were pooled and dialysed overnight against 1× PD solution (10× stock: 2 g l−1 KCl, 2 g l−1 KH2PO4, 80 g l−1 NaCl, 11.5 g l−1 Na2HPO4). This protocol is based on a method described previously39.

Northern blot analysis

Human 12-lane multiple tissue northern (MTN) blot and human brain multiple tissue northern (MTN) blot II (Clontech) were probed with a [32P]dCTP-radiolabelled PvuII–XhoI fragment extracted from the Hippi cDNA sequence (amino acids 322–429). The probe was prepared with the T7 QuickPrime kit (Pharmacia) according to the manufacturer's specifications. The blots were first incubated with 10 ml ExpressHyb solution (Clontech) for 30 min at 68 °C. Hybridization was in ExpressHyb solution (20 × 106 c.p.m. in 10 ml) at 68 °C for 1 h. Following hybridization, northern blots were washed twice with 2× SSC–0.05% SDS at room temperature for 10 min and twice again with 0.2× SSC–0.1% SDS at 50 °C for 20 min. Autoradiography was carried out overnight using MR Kodak film. Hybridization of the northern blot with β-actin as an internal control probe confirmed that the RNA was intact and had transferred.

Immunocytochemistry on brain sections

Mice were deeply anaesthetized and perfused through the left cardiac ventricle with 200 ml 4.0% paraformaldehyde in 0.1 M phosphate buffer (PB). Brains were removed and post-fixed for 1 h. Brains were then cryoprotected in ascending 10–30% sucrose in 0.1 M PB. Coronal sections (50 μm) were cut using a freezing microtome. Sections were rinsed in 0.05 M PBS (pH 7.2) and processed for immunocytochemistry as follows. To reduce endogenous peroxidase activity and nonspecific antibody binding, sections were incubated in 3% hydrogen peroxide and then in PBS containing 4% normal goat serum (NGS) followed by PBS rinses. Sections were then incubated in 2% NGS–PBS containing anti-Hippi antibodies at a concentration of 1:100 and 50 μg ml−1 biotin for 48–72 h. Sections were then rinsed in PBS and incubated overnight in biotinylated goat anti-rabbit secondary antibody (Vector, Burlingame, CA) in PBS containing 2% NGS. Following rinses in PBS, sections were then incubated in avidin–biotin complex (Vector ABC Elite) for 4 h followed by several PBS rinses. After incubation in 0.05% 3,3′-diaminobenzidine tetrahydrochloride (DAB, Sigma) and 0.01% hydrogen peroxide in 50 mM Tris buffer for 5–15 min, sections were mounted on glass slides, dehydrated and coverslipped for light microscopy using a Leica DMRE. Some Hippi-immunoreacted sections were embedded in Epon for examination of semi-thin sections. To assess the expression of Hippi in vivo, brains of adult FVB/N mice were processed for combined immunocytochemistry with either a neuron-specific marker (NeuN, Chemicon) or astrocyte-specific marker (CY-3-conjugated glial fibrillary protein (GFAP), Sigma), and an antibody against Hippi.

Immunogold labelling for electron microscopy

Mice were deeply anaesthetized and perfused through the left cardiac ventricle with 200 ml 3.0% paraformaldehyde and 0.1% glutaraldehyde in 0.1 M PB. Brains were removed, post-fixed for 1 h, cut into 50 μm coronal sections on a vibrating microtome (vibratome) and processed for labelling. For ultrastructural analysis, we used pre-embedding immunogold labelling of Hippi. Sections were rinsed in PBS, processed according to the manufacturer's instructions with primary antibody used at 1:100. Ultra-small colloidal gold conjugated secondary antibody (Aurion, Wageningen, The Netherlands) was used to bind the primary antibody. Following post-fixation with 2.5% glutaraldehyde, gold particles in sections were intensified using R-gent SE-EM silver enhancement kit (Aurion). Sections were then fixed with 0.5% osmium tetroxide in 0.1 M PB for 15 min and processed for electron microscopy as described elsewhere40. Selected sections were then placed in 0.5% osmium tetroxide in 0.1 M phosphate buffer for 30 min, rinsed in PB, dehydrated in 25–100% ethanol followed by propylene oxide, infiltrated and flat embedded in Epon between sheets of Aclar and cured at 60 °C for 2–3 days.

Retrograde labelling

For retrograde labelling of striato-pallidal and cortico-striatal projection neurons, mice received stereotaxic injections of 0.25–1 μl FluoroGold (3% in sterile PBS) into the globus pallidus or lateral striatum one week before death. Mice were injected intraperitoneally with 100 units of heparin in sterile water, deeply anaesthetized with pentobarbitol or chloroform, and transcardially perfused with cold 3% paraformaldehyde in 0.1 M PBS. Brains were post-fixed for 24–48 h in the same fixative, and 30-μm sections cut on a vibratome. Sections were collected in sterile PBS at 4 °C, rinsed in 0.1 M PBS with 0.3% Tween 20, and incubated in blocking solution (0.1% PBS with 0.3% Tween 20, 3% whole goat serum, and 5% BSA) for 1 h. Sections were sequentially placed into primary antisera against Hippi (diluted 1:100 in block solution) or NeuN (dilution 1:50 in block solution), or Hip-1 (diluted 1:100 in block solution) or GFAP (dilution 1:200 in block solution), as indicated, for 24 h at 4 °C. After incubation with primary antibody, sections were washed several times in blocking solution and incubated in secondary antibody for 2 h at room temperature. Secondary antibodies (Molecular Probes) were used as follows: goat anti-mouse CY-3 with NeuN and Hip-1 at a dilution of 1:200, and goat anti-rabbit Alexa 488 with Hippi primary at a dilution of 1:200. Following further washes with 0.1 M PBS, sections were dry-mounted on gelatin-coated slides, dehydrated by serial ethanol washes, and permanently mounted with Fluoromount. Sections containing regions of frontal cortex, striatum and hippocampus were analysed using an upright fluorescence microscope (Zeiss), digital images captured on a CCD camera (Princeton Instrument Inc, Mamouth Junction, NJ), and processed into double immunofluorescence figures using northern exposure image program. No staining was observed in control sections when primary antibody was omitted.

Immunoprecipitation and immunoblotting

HeLa or human embryonic kidney 293T cells (293 cells) were transiently transfected with various constructs using Lipofectamine (Life Technologies Inc., Carlsbad, CA) according to the manufacturer's instructions. In all transfections the pan-caspase inhibitor z-VADfmk at 50 μM was present during transfection and throughout the 48 h expression period to prevent possible cell death induced by the constructs. Cells were collected 48 h later with a cell scraper and washed in PBS. They were lysed in NP-40-containing buffer (50 mM Tris pH 8, 1% NP-40, 2 mM EDTA) supplemented with 10 μg ml−1 each of the protease inhibitors leupeptin, aprotinin, N-tosyl-L-phenylalanine chlorometyl ketone (TPCK), N-tosyl-L-lysine chlorometyl ketone (TLCK), and phenylmethylsulphonyl fluoride (PMSF). Polypeptides were recovered by immunoprecipitation from equivalent amounts of total cellular protein using the antibodies indicated and protein A–sepharose (Amersham Pharmacia Biotech). (Standard controls (not shown) include: protein A–sepharose without antibody; beads loaded with preimmune serum; and beads loaded with irrelevant sera (anti-FADD or anti-caspase-3; all lanes shown were eluted from protein A–sepharose beads). Immunoprecipitates were washed three times with lysis buffer. Proteins were then eluted in SDS-containing sample buffer, boiled, electrophoresed in SDS–PAGE (Novex-Invitrogen, San Diego, CA), and transferred onto nitrocellulose membranes for immunoblotting. The blots were first incubated in blocking buffer (BB) (Tris-buffered saline pH 7.4 (TBS), 5% (w/v) milk (blotting grade blocker non-fat dry milk; Bio-Rad), 0.1% (v/v) Tween 20) for 1 h at room temperature and then incubated for another hour in primary antibody diluted in BB. After washing three times in 1× TBS with 0.1% (v/v) Tween 20 for 5 min, blots were incubated for 1 h at room temperature in goat anti-rabbit IgG coupled to horseradish peroxidase diluted 1:3,000 in BB. Blots were washed three times in 1× TBS, 0.3% (v/v) Tween 20 for 5 min, and three times in 1× TBS, 0.1% (v/v) Tween 20 for 5 min. Detection was by enhanced chemiluminescence (Amersham).

Cell toxicity assay

293 cells were plated in DMEM, 10% FCS, 50 U l−1 penicillin/streptomycin, 2 mM L-glutamine (all from Canadian Life Technologies, Burlington, Ontario) in 6-well dishes (Falcon), and transfected at 60% confluency using the CaPO4 method. Transfection efficiency was 80–95% as measured by β-Gal staining. The medium was changed 6 h post-transfection, and again 24 h post-transfection. At 48 h post-transfection, cells were scraped on ice, washed with cold PBS, and group II (effector) caspase activity was measured using the fluorescent substrate Ac-DEVD-AFC (Biomol, Plymouth Meeting, PA) at a final concentration of 50 μM. The plate was incubated in a Labsystems Fluoroscan Ascent fluorescent plate reader at 37 °C for 75 min, with a reading obtained every 5 min, using excitation at 385 nm and emission at 510 nm and Ascent Research edition 2.2.4 software. The DEVDase activity was normalized to protein concentrations. For post-mitotic neurons, striatal neurons from E18 rat embryos were isolated and grown in neurobasal medium for 7 days in culture, then transfected using the CaPO4 method. Half of the medium was removed before transfection, to be added back afterwards. pEGFP vector (enhanced green fluorescent protein) (0.4 μg) was cotransfected with 2 μg pCEP-Hip-1 and/or pCEP-Hippi and/or empty pCEP, as indicated. The CaPO4–DNA precipitate was left for 1 h then washed twice with PBS before adding back neurobasal medium. After 48 h, EGFP-positive neurons remaining on the dish were counted by confocal microscopy.

Immunocytochemistry on neuronal NT2 cells and primary rat striatal neurons

Cells were plated on Lab-Tek chamber slide covers (Nalge Nunc International), precoated with Cell-Tak cell adhesive (Collaborative Biomedical Products), and grown overnight in normal DMEM medium. Cells attached to the bottom of chamber were washed twice with PBS and fixed in 3% paraformaldehyde–PBS solution (pH 7.5) for 30 min at room temperature, and then washed twice for 10 min in PBS supplemented with 10 mM glycine. Cells were permeabilized by incubation in 1% Triton-PBS solution for 5 min at room temperature, followed by two washes in PBS supplemented with 10 mM glycine, and were stained for 1 h at 4 °C with affinity-purified rabbit polyclonal anti-Hippi and/or mouse monoclonal anti-Hip-1 antibodies followed by two washes in PBS–glycine buffer. Cells were then incubated for another hour in the dark at 4 °C with secondary antibodies (Alexa 594 goat-anti-rabbit IgG and Alexa 488 goat-anti-mouse IgG or Alexa 594 goat-anti mouse IgG (Molecular Probes) at a 1:200 dilution). Cells were given a final round of two washes in PBS–glycine before mounting with ProLong Antifade medium (Molecular Probes). For colocalization with the Golgi apparatus, just before mounting cells were incubated with C5-ceramide (BODIPY FL C5-ceramide; Molecular Probes) at 1 μm in PBS–glycine for 30 min at room temperature. C5-ceramide has been shown to localize to the trans-Golgi41,42. This was followed by three washes in PBS–glycine buffer. Images were scanned with a Zeiss LSM-510 confocal microscope. Similar procedures were used for primary rat striatal neurons isolated from E18 embryos. Cells were obtained by trypsinization and tritriation and cultured for 7–10 days before fixation and confocal microscopy as described above.