The CNS glycoprotein Shadoo has PrP(C)-like protective properties and displays reduced levels in prion infections.

The cellular prion protein, PrP(C), is neuroprotective in a number of settings and in particular prevents cerebellar degeneration mediated by CNS-expressed Doppel or internally deleted PrP ('DeltaPrP'). This paradigm has facilitated mapping of activity determinants in PrP(C) and implicated a cryptic PrP(C)-like protein, 'pi'. Shadoo (Sho) is a hypothetical GPI-anchored protein encoded by the Sprn gene, exhibiting homology and domain organization similar to the N-terminus of PrP. Here we demonstrate Sprn expression and Sho protein in the adult CNS. Sho expression overlaps PrP(C), but is low in cerebellar granular neurons (CGNs) containing PrP(C) and high in PrP(C)-deficient dendritic processes. In Prnp(0/0) CGNs, Sho transgenes were PrP(C)-like in their ability to counteract neurotoxic effects of either Doppel or DeltaPrP. Additionally, prion-infected mice exhibit a dramatic reduction in endogenous Sho protein. Sho is a candidate for pi, and since it engenders a PrP(C)-like neuroprotective activity, compromised neuroprotective activity resulting from reduced levels may exacerbate damage in prion infections. Sho may prove useful in deciphering several unresolved facets of prion biology.

Introduction

Prions are the causative agents of neurodegenerative diseases, which include bovine spongiform encephalopathy (BSE) in cattle, scrapie in sheep, chronic wasting disease in mule deer and elk, and Creutzfeldt–Jakob Disease (CJD) in humans. The infectious agent is believed to consist of improperly folded forms of a host-encoded protein, the cellular prion protein (PrPC). Conversion of PrPC into the disease-associated isoform, PrPSc, is thought to be the primary pathogenic event, although the mechanisms by which PrPSc causes disease are poorly understood. PrPC is absolutely required for disease progression, as PrP knockout (Prnp0/0) mice do not succumb to disease and do not propagate infectivity following intracerebral challenge with infectious prions (Büeler , 1993).

The mammalian prion protein family currently consists of two proteins: PrPC which is expressed at high levels in the central nervous system (CNS), and Doppel (Dpl), a molecule with a similar three-dimensional structure, whose postnatal expression is normally confined to the testis (Silverman ; Mo ). Whereas a role for Dpl in the proper functioning of the male reproductive system has been confirmed in two lines of Dpl knockout mice (Behrens ; Paisley ), the function of PrPC, a well-conserved neuronal glycoprotein, comprises a conundrum, in part because phenotypic alterations in Prnp0/0 mice have been subtle or disputed. One emerging area of consensus concerns a protective effect of PrPC against neuronal insults. In particular, PrPC is upregulated following ischemic brain damage, in both humans and mice (McLennan ; Weise ). PrPC deficiency in mice increases infarct size following cerebral artery occlusion and increases caspase 3 activation (Spudich ), and PrPC overexpression improves neurological behavior and reduces infarct volume in a rat stroke model (Shyu ). Another widely observed and perhaps related phenomenon involves the ability of PrPC to protect against a variety of proapoptotic stimuli (Kuwahara ; Bounhar ; Chiarini ; Cui ; Sakudo ; Lee ; Novitskaya ). Strong evidence for a neuroprotective activity for PrPC against apoptosis in vivo has come from studies of transgenic (Tg) mice expressing internally deleted forms of PrP or wild-type (wt)Dpl within the CNS. The presence of Dpl in the brain of Prnp0/0 mice leads to a neurodegenerative syndrome characterized by a profound apoptotic loss of cerebellar cells (Nishida ; Moore ; Rossi ). A similar phenotype is observed when N-terminally truncated versions of PrPC (ΔPrP) are expressed in the brain (Shmerling ). Remarkably, both syndromes are abrogated by the coexpression of wt PrPC. Recently, it has been shown that a smaller deletion restricted to the well-conserved central domain of PrP is sufficient to elicit a highly toxic phenotype in Prnp0/0 mice (Baumann ; Li ). The above studies have led to a model in which Dpl and ΔPrP initiate aberrant signaling through a hypothetical prion ligand termed LPrP, a process which is blocked by PrPC binding (Flechsig ). Assuming that the interaction between PrPC and LPrP represents an essential physiological event, the authors also proposed the existence of a PrPC-like protein termed π, which binds to LPrP and is capable of compensating for the absence of PrPC in Prnp0/0 mice. To this date, no candidates for π (or LPrP) have been put forward. However, an open reading frame was discovered which, when translated, exhibits homology to the central hydrophobic domain in PrPC. This gene, denoted Sprn (‘shadow of the prion protein'), is present from zebrafish to humans and is predicted to encode a short protein, Shadoo (Sho) (Premzl ). Sprn is located on chromosome 7 in mice, away from the Prn gene complex on chromosome 2.

Building on the genetic interaction between PrPC and Dpl or ΔPrP, we have established an assay for PrPC activity in primary cultures of cerebellar granule cells (Drisaldi , and references therein). Here, cerebellar granule neurons (CGNs) cultured from Prnp0/0 mice are transfected with plasmids encoding Dpl or PrP alleles of interest and individual apoptotic events scored. This assay recapitulates the phenotypes produced by multiple PrP alleles in Tg mice, including neurotoxicity of both Dpl and ΔPrP, and neuroprotective activity of PrPC against the toxicity elicited by either Dpl or ΔPrP. In conjunction with biochemical and histological analyses, we have used the CGN assay to explore the properties of Sho. We now demonstrate that Sho is a GPI-anchored neuronal glycoprotein present in the CNS from early postnatal life. Not only is Sho PrPC-like in its ability to protect against both Dpl and ΔPrP toxicity in the CGN assay, but it is also strikingly reduced in prion infections.

Results

Sho protein expression in N2a cells and brain

Motivated by the absence of obvious phenotypic defects in adult Prnp0/0 mice, we considered proteins that might overlap functionally with PrPC. One criterion was evolutionary conservation. In this regard, bioinformatic analyses by Premzl and co-workers have yielded an interesting candidate in the shape of the Sprn open reading frame present in genomic DNA of species from mammals to fish (Premzl , 2004; Miesbauer ) and which, unlike the hypothetical Prnt gene (Makrinou ; Premzl ), is present in the mouse genome. The architecture of the predicted Sho protein loosely resembles the flexibly disordered PrP N-terminus (Figure 1). Analyses of Sprn expression by RT–PCR (Premzl ) and interrogation of expressed sequence tag databases (UniGene) imply expression in embryonic and adult neuronal tissue as well as the retina and visual cortex, but have yet to document spliced mRNAs. Consequently, we evaluated Sho as a putative third member of the prion gene family.

Figure 1

Domain structure of Dpl, PrPΔ32–121, PrP and Sho. α-Helices (A, B and C) are boxed. A basic repeat region in Sho is stippled and the hydrophobic domain in PrP and Sho is striped. Neuroprotective ‘determinants', sequences required for PrP's neuroprotective action and mapped genetically, are shown in boxed numbers, and an expanded view shows the central hydrophobic regions of PrP and Sho aligned with the T-COFFEE algorithm. PrP and Sho deletions used in this study are bracketed alongside the alignment. Two residues that correspond to the N-termini of human PrP ‘C1' fragment are underlined. ‘Neurotoxic' and ‘neuroprotective' refer to assays in cerebellar granule cell cultures or Tg mice (see main text).

To address expression at the protein level, we raised antibodies against Sho. Two antisera (‘04SH-1' and ‘06SH-3') were against a mouse Sho peptide consisting of residues 86–100, whereas a third (‘06rSH-1') was against full-length recombinant mouse Sho(25–122) expressed in Escherichia coli (Figure 2A) and recognizes an N-terminal epitope contained within residues 30–61 (Supplementary Figure S1). Assessed by Western blot of tissue lysates, 06rSH-1 was virtually devoid of cross-reactive species (Supplementary Figure S1). Cross-reactive species of molecular weights incompatible with authentic Sho were present in analyses with antisera 04SH-1 and with 06SH-3, but these had varying intensities and/or different molecular weights for the two antisera. Consequently, the following comments are restricted to signals detected by two or more varieties of α-Sho antibodies.

Figure 2

Analysis of recombinant Sho in E. coli and expression of murine Sho in cultured cells. (A) Schematic representation of the Sho protein. The location of the mapped epitopes for α-Sho peptide antisera (04SH-1 and 06SH-3) and α-recombinant Sho (06rSH-1) are shown. (B) Circular dichroism spectrum of recombinant mouse Sho, rSho(25–122). The spectral trace is consistent with a random coil configuration. (C) Cell surface expression of wt Sho and a mutant Sho allele lacking the hydrophobic tract in non-permeabilized transfected N2a cells as demonstrated by immunocytochemistry. Scale bar, 50 microns. (D) Diminution of Sho signal in the cell lysates of Sho-transfected N2a cells following pretreatment with increasing concentrations of PI-PLC. (E) Western blot showing expression of a wt Sho transgene in N2a cells with or without PNGaseF treatment. A lysate from cells transfected with empty vector is included to show antibody specificity.

Similar to PrPC, murine Sho is revealed as being expressed at the cell surface, N-glycosylated and sensitive to the GPI anchor cleaving enzyme phosphatidylinositol-specific phospholipase C (PI-PLC) in transfected N2a neuroblastoma cells (Figure 2C–E). Following PNGaseF treatment, full-length Sho has a molecular weight of 9.1 kDa as assessed by SDS–PAGE (predicted 9.5 kDa). In addition to the full-length protein, a fraction of the protein in transfected N2a cells has a faster electrophoretic mobility (Figure 2E). In PrP, a well-documented physiological ‘C1' cleavage occurs just before the hydrophobic tract at residues His111 and Met112 (human PrP numbering scheme, underlined; Figure 1) both in cultured cells and in the adult brain (Pan ; Chen ; Vincent ). Since cell lysates for these analyses were prepared in the presence of protease inhibitors, it is possible that an analogous endoproteolytic processing could figure in the biogenesis of Sho. A detailed description of truncated forms of Sho will be presented elsewhere (Coomaraswamy et al, in preparation).

Analysis of mouse tissue by Western blotting defined a predominant glycosylated protein species of similar molecular weight to transfected full-length Sho (approximately 18 kDa), and one that is developmentally regulated, appearing at embryonic day 16 and persisting in early postnatal life and in the brains of adults (Figure 3A and B). PNGaseF-sensitive bands of identical molecular weights were observed with two independent Sho antibodies, confirming the authenticity of the bands and the presence of Sho in the mouse CNS. Sho signal was emphasized in membrane-enriched preparations from adult mouse brains, corroborating its membrane anchorage (Figure 3C). In Western blots performed with α-Sho86–100 antisera, Sho signal increased following treatment with PNGaseF, suggesting that N-glycosylation at Asn107 partially occludes antibody binding at the adjacent epitope for the peptide-directed antiserum.

Figure 3

Analysis of mouse Sho in tissue preparations. (A) Expression of Sho in whole mouse embryos assessed by Western blotting using a 12% Tris–glycine gel. Sho is expressed beginning at embryonic day 16. (B) Postnatal expression of Sho in neonatal and adult mouse brains with or without PNGaseF treatment assessed by Western blotting using 14% Tris–glycine (α-rSho) or 4–12% NuPAGE (α-Sho(86–100) peptide and α-PrP) gels. Full-length species before and after enzymatic treatment are bracketed. (C) Sho is present in a membrane-enriched fraction prepared from mouse brains. Membrane preparations with or without PNGaseF treatment were analyzed on 4–12% NuPAGE gels by Western blotting. Sho was detected using either α-Sho peptide polyclonal 04SH-1 (A–C) or α-recombinant Sho polyclonal 06rSH-1 (B), whereas PrP was probed with single-chain antibodies D13 (A–B) or D18 (C).

Overlapping and complementary PrPC/Sho expression

In situ hybridization using antisense-strand riboprobes prepared against the mouse Sprn open reading frame (but not sense-strand controls) yielded signals in the adult mouse CNS. Analyses of the hippocampus and cerebellum revealed prominent signals in the cell bodies of pyramidal cells and Purkinje cells, respectively (Figure 4B and J). By way of comparison, Prnp has a broader pattern of neuronal expression (Kretzschmar ; Taraboulos ). Immunohistochemistry for Sho protein yielded prominent signals in the same cell types defined by in situ hybridization (Figure 4D and L), that is, hippocampal neurons and cerebellar Purkinje cells. In the case of antisera 04SH-1, these signals were absent when antibodies were preincubated in a solution containing the Sho86–100 peptide immunogen (Figure 4C and K). Besides defining Sho as the ‘second' cellular prion protein present in neurons of the adult CNS, these data define intracellular transport phenomena, as immunohistochemical signals were present in cell processes in addition to the cell bodies detected by antisense riboprobes (i.e., predicted to contain Sho mRNA). In the case of Purkinje cells, immunostaining was present not only in cell bodies but also prominent in their processes, specifically in the dendritic arborizations present within the molecular layer of the cerebellum (Figures 4L and 5F–H, signals detected with all three antisera). A related phenomenon was observed in the hippocampus, notably in CA1 pyramidal neurons. Here, Sho immunoreactivity was absent from axonal projections (with all three α-Sho antibodies), present in cell bodies (seen by all three α-Sho antibodies), and notable in the apical dendritic processes located in the stratum radiatum of the hippocampus (strong signals with 04SH-1 and 06SH-3, and a less intense signal with 06rSH-1) (Figures 4D and 5A–C).

Figure 4ah

Localization of Sprn mRNA and Sho protein in the adult mouse brain. (A–H) The hippocampus, and (I–P) the cerebellum. wt C57/B6 mice are presented in all sections, with the exception of B6 congenic Prnp0/0 (E, M) and Tg(SHaPrP)7 mice (H, P). Panels A, C, E, G, I, K, M and O in the left-hand columns comprise negative controls (i.e., sense-strand riboprobes, peptide-directed antisera preincubated with a soluble form of the antigenic peptide, Prnp0/0 mice for PrP-directed antibodies) for analyses presented in the right-hand columns. In situ hybridization: panels A, B, and I, J represent hybridizations with Sho sense-strand (A, I) or antisense (B, J) cRNA probe. Sections are not counterstained and blue staining from NBT/BCIP substrate represents hybridization to Sprn mRNA. Immunohistochemistry: all other panels of mouse brains with genotypes as noted above. Anti-Sho antibody 04SH-1 (α-Sho) antibody was used with (C, K) or without (D, L) preincubation with Sho(86–100) peptide. Antibodies 7A12 and 3F4 were used for the detection of mouse PrP (E, F, M, N) and hamster PrP (G, H, O, P), respectively. Note the Sho staining of CA1 apical dendritic processes (D, black bracket) and Purkinje cell layer (L, white arrow), and absence of Purkinje cell-body staining with α-moPrP (N) and relative paucity of staining with α-HaPrP antibody (P, black arrow). Scale bar in panels A–H, and I–P, 100 μm.

Figure 4ip

Continued.

Figure 5

Reciprocal and overlapping expression of Sho and PrPC in the CNS. (A–E) CA1 hippocampal neurons of adult mice probed as indicated. The signal for Sho immunohistochemistry in apical dendrites of wt mice (A–C) has an equivalent in a ‘negative image' (white brackets) in the molecular layer of the neuropil imaged for either mouse PrP (D) or hamster PrP in Tg(SHaPrP)7 mice (E). Apical dendrites were less intensely stained with the α-Sho 06rSH-1 antibody (C, black arrowhead) Analogous analyses are presented for Purkinje cells (F–J). Somatodendritic localization of Sho in Purkinje cell bodies and dendritic arborizations (F–H) is contrasted by a reciprocal ‘negative image' in molecular layer of the neuropil imaged for either mouse PrP (I) or hamster PrP in Tg(SHaPrP)7 mice (J). Note the complete absence of MoPrP staining in cell bodies (I, black arrowheads). Somatodendritic localization of calbindin in Purkinje cells of wt mice (K) is shown for comparative purposes. (L–N) The cerebral cortex probed simultaneously with α-Sho (06rSH-1, green) and α-PrP (8H4, red) antibodies. Overlapping PrPC and Sho expression is observed in neurons of the cerebral cortex, including colocalization in cell bodies (N, white arrowheads). Scale bar, 25 μm (A–E), 10 μm (F–K) or 20 μm (L–N).

Since PrP has been reported to possess an unusual property for a GPI-linked protein, the ability to undergo basolateral (dendritic) sorting in polarized cells in distinction to the more typical apical (axonal) sorting, parallel analyses were undertaken for PrPC and Sho. PrPC was examined at basal levels from the (endogenous) Prnp gene, or expressed from a cosmid transgene of the hamster PrP gene including 25 kb of sequences 5′ to the transcriptional start-site (Scott ; Prusiner ). In the case of wt mouse PrP, analyses were performed with 7A12 antibody (Li ) using Prnp0/0 mice as negative controls (Figure 4E and M), whereas non-Tg wt mice served as negative controls for the use of the hamster PrP transgene-specific 3F4 antibody (Figure 4G and O). Besides the anticipated differences in signal levels between wt and Tg(SHaPrP)7 mice, the PrPC-directed antibodies yielded similar expression patterns, with widespread staining throughout the brain. PrPC was underrepresented in the cell bodies of the pyramidal neurons of the hippocampus but abundantly present in the stratum oriens containing the axonal projections, and also in the radiatum, lacunosum moleculare and molecular layers (Figure 4F and H). In the cerebellum, mouse PrPC was absent in wt Purkinje cells (as described previously (Liu ; Baumann )) and hamster PrPC was represented in some but not all Purkinje cells of Tg(SHaPrP)7 mice, and at a level lower than that of the surrounding neuropil (Figure 4N and P). On the other hand, PrPC was present in the granule cell layer and abundant in the molecular layer. Strikingly, the abundant signal present in the radiatum layer was not totally uniform, and—by virtue of a negative staining effect—revealed the outlines of Purkinje cell dendritic arborizations in the cerebellum and the apical dendritic of CA1 neurons (Figure 5D–E, and I–J): these are structures with marked immunostaining with α-Sho86–100 (04SH-1 and 06SH-3) and α-rSho (06-rSH1); Figure 5A–C and F–H). These data therefore define a complementary ‘interlocking' aspect to PrPC/Sho protein expression in the cerebellum and suggest a similar effect in the hippocampus.

Although the most prominent Sho signals were obtained in the hippocampus and cerebellum (Figure 4), signals for both Sprn mRNA and Sho protein were also present in other areas of the brain including the cerebral cortex, the thalamus, and the medulla (Figure 5L–N; Supplementary Figure S2). Coincidence with neurofilament staining indicates that neurons are the prominent site of Sho expression (Supplementary Figure S2). PrPC signal was abundant in these regions (Figure 5, and data not shown), confirming that Sho and PrPC expression profiles overlap in certain areas of the brain. A further example of this phenomenon is apparent in the adult retina (Premzl ; Chishti et al, in preparation). In overview, Sho expression within the adult mouse brain is either less widespread than PrPC, or basal Sho levels in certain areas fall below the detection threshold of our current procedures. Lastly, to assess whether there is a physiological cross-regulation effect between PrPC and Sho, we performed a number of analyses on Prnp0/0 mice (Supplementary Figure S4). These failed to reveal a clear upregulation of Sho in response to PrPC deficiency.

Neuroprotective activity and the central region of PrP

Two N-terminal PrP modules, the charged region and the copper-binding octarepeats, contribute to neuroprotective activity in a neuronal assay (Drisaldi ). However, while others analyzed the same N-terminal modules, they highlighted a trafficking effect (Sunyach ), so we sought determinants beyond these potential delivery signals. The C-terminus of PrPC comprises a three-helix bundle (Riek ) similar to Dpl (Mo ), and in the form of alleles such as PrPΔ32–121 (Shmerling ), exhibits Dpl-like toxicity. We therefore focused on the ‘remaining' area of PrP, residues 91–121 between the octarepeats and the structured domain (Figure 1) containing the hydrophobic domain (HD) with Sho homology. The deletion alleles were tested for their ability to engender stable forms of PrP. As anticipated from prior analyses of in-frame deletions (Holscher ; Shmerling ; Hegde ), these forms of PrP were synthesized at levels similar to wt PrP, and also underwent similar maturation (Baumann ; Li ) (Supplementary Figure S3).

Residues 100–105 are poorly conserved (Figure 1) and in the cerebellar granule neuron (CGN) cellular assay (Drisaldi ), a PrPΔ100–105 allele exhibited protective activity like wt PrP (P<0.001) when measured against Dpl. Conversely, PrPΔ91–99 and PrPΔ106–112 alleles had decreased protective activity (neither is significantly different from Dpl alone; Figure 6B). PrPΔ106–121 transfected alone exhibited partial neurotoxic activity approaching that of a toxic PrPΔ32–121 control allele (Shmerling ), but different from that of wt PrP (P<0.01) (Figure 6A). These data offer a parallel to assays in Tg mice. Here, deletions invading the PrP central region from residues 91 to 121 produce proteins that are progressively toxic: no significant pathological abnormalities are noted in Tg mice expressing PrP Δ32–80, Δ32–93, and Δ32–106, but mice expressing Δ32–121 (or Δ32–134) exhibit profound loss of cerebellar cells (Shmerling ; Flechsig ), as do mice expressing PrPΔ94–134 or Δ105–125 (Baumann ; Li ). From this we infer that a third determinant of neuroprotection lies between PrP residues 91 and 99, and a complex or modular activity determinant (‘determinant 4') lies between residues 106–121, whereby a Δ106–112 deletion loses protective activity and a larger Δ106–121 deletion exhibits partial neurotoxic activity. These data extend observations made in hippocampal neurons, where PrP residues 95–124 were implicated in neuroprotective action versus Dpl (Lee ).

Figure 6

Neuroprotective activity and Sho expression in CGN cells. Test for toxicity of PrP mutant alleles versus the performance of neurotoxic wtDpl and PrPΔ32–121 controls in single transfections (A), results of cotransfections of PrP deletion mutants, wt Sho and pBUD.GFP empty vector controls with wtDpl plasmid (B), and results of cotransfections to assess putative protective effects of Sho and internally deleted Sho versus a toxic PrPΔ32–121 allele (C). Determinations (% cells±s.e.m undergoing apoptosis) reflect the results of two or more triplicate transfections of independent batches of Prnp0/0 CGNs scored by independent observers. Whereas PrPΔ100–105 and wt Sho had potent neuroprotective activity (both P<0.001, n=6, n=11, respectively) not significantly different from that of wt PrP, PrPΔ91–99 and PrPΔ106–112 exhibited a tendency for reduced neuroprotective activity against Dpl (neither cotransfection is significantly different than Dpl single transfections). ShoΔ62–77 cotransfections were not significantly different from Dpl or PrPΔ32–121 single transfections. PrPΔ106–121 (n=11) transfected alone was significantly different from wt PrP (P<0.01; n=15) and all other PrP alleles (P<0.001), with the exception of the previously documented toxic PrPΔ32–121 allele (Shmerling ): PrPΔ32–121 (n=7) versus wt PrP, P<0.001. Cerebellar granule cell assays for bioactivity, creation of plasmid constructs and verification of protein expression were performed as described previously (Drisaldi ; Supplementary Figure 3) **P<0.01, ***P<0.001. (D) Western blot analysis using 06rSH-1 detects Sho in mouse brain lysates but not in a normalized loading of lysate derived from mouse CGNs. ‘ns'=non-specific bands detected by the antibody. (E) Immunocytochemistry on non-permeabilized CGNs cotransfected with non-fluorescent Sho and PrPΔ32–121 plasmids (3:1 ratio). Sho was detected with 06rSH-1 and PrPΔ32–121 with 8H4. Both Sho and PrPΔ32–121 are expressed in a single cell and colocalization is observed in the cell body. Scale bar, 15 μm.

Sho has PrPC-like neuroprotective activity in a functional assay

Sequence identity between Sho and PrP occurs in the central hydrophobic region. Since this area is highly conserved in PrP (Wopfner ) and is predicted to overlap or be immediately adjacent to determinants of neuroprotective action (Figures 1 and 6B), we tested Sho in a functional assay in Prnp0/0 CGN cultures. Remarkably, a wt mouse Sho cDNA was PrPC-like in its ability to block the proapoptotic action of cotransfected Dpl (Figure 6B) (P<0.001 versus Dpl alone, no significant difference between wt PrP alone versus Dpl+Sho). Since the putative PrPC-like activity called π was inferred from the action of ΔPrP, rather than the action of Dpl (Shmerling ), further experiments were carried out to assess the effect of Sho upon the toxic action of a ΔPrP allele (Figure 6C). In these cotransfections, wt Sho reduced the toxic activity of singly transfected PrPΔ32–121 to a baseline defined by ‘empty vector' (pBUD.GFP) controls (P<0.001), and overlapping expression and colocalization was apparent in transfected neurons (Figure 6E). Besides adding to the similarities between Sho and π, these data underscore the notion that ΔPrP and Dpl exert toxicity through the same pathway.

Given the similarity between Sho and PrP in the HD, this domain was a strong candidate for the ‘active site' of Sho. Consequently, we assessed the protective activity of Sho bearing a precise HD deletion. Here, the cotransfected ShoΔ62–77 allele was not protective against Dpl (not significant versus Dpl alone) (Figure 6B). wt and Δ62–77 Sho alleles are expressed at similar levels in bulk-transfected cells, and with immunoreactivity present upon the surface of transfected N2a cells (Figure 2C; Supplementary Figure S1). In contrast to a wt control allele, the ShoΔ62–77 allele also did not exert significant protection against a toxic PrPΔ32–121 allele (P<0.001 versus baseline of pBUD.GFP empty vector; Figure 6C).

As an extrapolation from results presented above and from prior analyses (Nishida ; Moore ; Rossi ), the ability of Dpl to cause cell death in Prnp0/0 CGNs implies lack of a compensatory neuroprotective activity (e.g., Sho) in these cells. To test this inference, and as an additional control for immunohistochemical analyses presented in Figures 4 and 5, we assessed Sho expression in the lysates of purified CGNs. Although Sho was readily detected in brain lysate, the analysis failed to detect a comparable signal in a normalized loading of CGN lysate (Figure 6D).

Sho expression in prion-infected brains

To assess Sho in a disease setting, Western blots were performed on brain homogenates from wt mice infected with the RML isolate of mouse-adapted prions (Carlson ). Two independent cohorts produced in different laboratories were used for this analysis. Assessed in the clinical phase of disease, animals from both cohorts demonstrated a striking reduction in Sho protein levels (Figure 7A and E). Notably, this decrease in signal is apparent using two distinct Sho antibodies with independent epitopes. Normalized against Thy-1, another GPI-anchored neuronal protein, Sho signals of RML-inoculated brain extracts versus brain extracts from healthy controls were reduced eight-fold (Figure 7B). A variety of other neuronal proteins were probed as internal controls (Figure 7C). Although mild loss of signal was apparent for some of these—perhaps a direct consequence of neuronal damage and death—in no instance did these reductions approach those seen for Sho. As a further control we assessed Sho levels in a widely used animal model of Alzheimer's Disease, TgCRND8 mice expressing a mutant β-amyloid precursor protein transgene (Chishti ). Using mice in clinical phase of the disease (marked by overt amyloid deposits, normalized Aβ levels equaling those of sporadic Alzheimer's disease, and profound cognitive impairment), no difference was apparent between TgCRND8 and non-Tg mice with regards to Sho levels (Figure 7D). Thus, depletion of Sho is absent in non-infectious CNS amyloidosis.

Figure 7

Reduced Sho levels in clinically ill prion-infected mice. (A) Western blot of homogenates prepared from the brains of non-inoculated or clinically ill (average of 172 days post-inoculation) RML prion-inoculated mice (C3H/C57BL6 background). Sho protein levels are notably reduced in prion-infected brains. Levels of the GPI-anchored protein Thy-1 are shown for comparison purposes. (B) Quantitation of Sho (06rSH-1) and Thy-1 blot signals in panel A by densitometry. Sho levels in prion-infected brains (normalized against Thy-1 signal) are reduced to 12.1±2.8% (P<0.001) the levels observed in non-inoculated mice. ***P<0.001. (C) Expression of neuronal markers in prion-infected and control mouse brains as assessed by Western blotting. No change in neuron-specific enolase (NSE) or calbindin levels, and only a moderate decrease in synaptophysin levels are observed in prion-infected brains. (D) No change in Sho levels are observed in brain homogenates prepared from clinically ill (8 months old) Tg mice (TgCRND8) expressing a familial Alzheimer's disease-associated variant of the amyloid precursor protein and control non-Tg littermates. (E) Reduced Sho expression in normalized brain homogenates in a second cohort of RML-inoculated mice versus control mice injected with a brain homogenate from healthy mice (C57BL6 background, 154 days post inoculation).

Discussion

Our first conclusion is that Sho is a bona fide neuronal glycoprotein and the third member of the mammalian prion protein family. Beyond mere sequence homology, Sho demonstrates a number of biochemical and cell biological properties also exhibited by PrPC. These include addition of a GPI anchor, N-glycosylation at one or two sites and a cleavage event likely positioned N-terminal to the hydrophobic tract (Figures 1 and 2C–E; Supplementary Figure S1). Furthermore, there are functional redundancies between PrPC and Sho measured in CGN cells. Although Sho and PrPC have dissimilar N-terminal regions that may tailor their activities in different neuroanatomic sites, we infer that the center of both PrPC and Sho comprises a functionally conserved and ancient activity domain contributing to neuroprotection. In PrPC-deficient areas of the brain, Sho may be particularly important in filling-in a PrP-like activity. Although the exact relevance of PrPC/Sho neuroprotection against exogenous stimuli versus in vivo brain physiology remains to be deciphered, a role against neuronal insults such as ischemic damage seems plausible. Indeed, increased PrPC expression following stroke has been observed in both humans and mice (McLennan ; Weise ).

A second conclusion is that the study of Sho may help decipher the molecular mechanisms underlying physiological signaling from the prion protein family. Here, some initial hints can be inferred from mutational analysis. As deletions invade the central region of PrPC, not only is protective activity lost, but the mutant PrPs gain spontaneous proapoptotic activity, as shown by loss of the granule cell layer in Tg mice (Flechsig ), and echoed here by the divergent performance of PrPΔ106–112 and Δ106–121 alleles in the CGN assay. It is possible that a gain of proapoptotic function for PrP within CA1 neurons when it is sterically blocked and dimerized by antibodies against residues 95–105 proceeds by a related mechanism (Solforosi ). In any event, the PrPC central region defined by genetic mapping (analyzed by CGN assays in this paper, see also Lee ) is natively unstructured in NMR studies (Donne ; Riek ) and the homologous region in Sho is likely to be natively unstructured as well (Figure 2B). This region may serve as a site for protein–protein interactions, and in this regard, our results parallel and extend a model for signaling from PrPC protein complexes. Besides PrPC, this model involves a hypothetical ligand, LPrP, and a conjectural PrPC-like protein denoted ‘π' (Shmerling ; Flechsig ). Two docking sites are posited between LPrP and PrPC, and one docking site between Dpl or ΔPrP and LPrP. Based on sequence similarity and domain structure (Figure 1), biochemical properties (Figure 2), a low level expression of Sho in the cerebellar granule cells (Figures 4L and 6D), and similar effects of wt PrP or Sho versus Dpl or PrPΔ32–121 in CGN assays (Drisaldi ) (Figure 6B), Sho can be seen to resemble π. In this scheme, sequences within or around the hydrophobic domain of Sho or PrPC would comprise a docking site for LPrP, a site associated with a protective effect. In PrPC, partial loss of this site may confer spontaneous neurotoxic behavior, consistent with the enhanced toxicity observed in Tg mice expressing PrP with HD deletions (Baumann ; Li ). A second site lying within the α-helical domain present in PrP, ΔPrP and Dpl is associated with a toxic effect (i.e., as per ΔPrP and Dpl) unless the first site is also present (as in wt PrPC) (Shmerling ; Cui ; Drisaldi ). Even though the original LPrP–PrPC–π model may prove incorrect in a cell biological sense (i.e., that signaling may have more to do with protection from toxic stimuli rather than regulation of endogenous trophic signals), it provides a useful framework for deciphering molecular interactions. Certainly the ability to now perform studies with PrP, ΔPrP, Dpl and Sho and assess responses within a cellular assay should aid the identification of authentic PrP interactors.

A third conclusion is that Sho may open a window to pathological signaling events in prion disease. Studies of scrapie-infected rodents support the concept of neuroanatomic ‘target areas' that give rise to the clinical features of disease (Kimberlin and Walker, 1986). Thus far, the mechanisms by which disease associated forms of PrP (PrPSc or ‘PrPD') might cause dysfunction in these target areas have proven elusive (Flechsig ). For example, whereas a requirement for PrPC was inferred from grafting tissue from mice overexpressing PrPC into Prnp0/0 mouse brains and performing prion challenge of chimeric mice (Brandner ), a genetic reconstitution experiment employing an astrocyte-specific hamster PrP transgene demonstrated an opposite effect, namely neuronal damage in PrPC-deficient neurons upon prion challenge (Jeffrey ). The concept of pathogenesis in trans, that is, independent of PrPC expression, is underlined by prior studies and by experiments herein, which demonstrate that the dendrites of CA1 neurons, an early ‘clinical target area' in prion infections, are largely devoid of PrPC (Figure 5D and E). Susceptibility of dendritic structures in this neuroanatomic area has been demonstrated by several techniques in different models of infectious prion disease (Hogan ; Johnston ; Brown ). We note that these structures are strongly labeled by both Sho peptide directed antisera in wt mice (Figure 5A and B), and are distorted and attenuated in prion-infected mice (not shown), as already described by others (Hogan ; Johnston ; Brown ).

Strikingly, Western blot analyses of brains from clinically ill prion-infected mice revealed a dramatic reduction of Sho protein (Figure 7). Because Sho has neuroprotective properties (Figure 6), it seems reasonable to consider whether a loss of Sho may underlie some of the clinical or pathological features of prion disease. In this regard, future studies utilizing Sprn0/0 mice should prove informative. The mechanism governing the loss of Sho protein during prion infections remains to be determined. While a transcriptional effect seems unlikely (not shown), protein structure may play a role. Since Sho's conserved region coincides with the epicenter of PrPC misfolding to disease-associated forms such as PrPSc (Peretz , 2001; Beringue ), conformational alterations in unstructured, cell-anchored Sho (or in Sho shed into the parenchyma, like other GPI-anchored proteins; Parkin ) may occur upon interaction with PrPSc. However, this does not explain how altered Sho proteins, should they exist, might be more prone to degradation. Expanding the view to include the concept of aberrant signaling may provide a more useful take on this problem. Assuming that Sho exerts a neuroprotective function, perturbations in this protective pathway could destabilize Sho or lead to a loss of Sho-expressing cells and a commensurate drop in protein level. According to this hypothesis, parenchymal PrPSc might interfere with the physiological protective activity of Sho in trans by virtue of both its aberrant fold and cell biological resemblance to Sho. While new experiments will be required to probe the validity of these concepts and the mechanisms controlling Sho expression, we suggest that Sho is a potential modulator for the biological actions of normal and abnormal PrP.

Materials and methods

Bioinformatics and statistics

Protein alignments were created using the T-COFFEE algorithm and EST searches conducted using UniGene. Data sets for all transfected constructs were assessed for normality (P>0.10) by the Kolmogorov–Smirnov test and analyzed by one-way ANOVA, Tukey pairwise comparisons, with significance set at P<0.05, using Prism software (v4.0c, GraphPad Inc.).

Mouse lines, husbandry and inoculations

Zrch I Prnp0/0 mice and littermates were maintained on a C57/B6 congenic background. Mice for inoculations, either C3H/B6 hybrids (Toronto) or B6 inbred (McLaughlin Research Institute) were inoculated intracerebrally with 30 μl of 0.1% RML-infected brain homogenate diluted in phosphate-buffered saline (PBS) containing BSA (50 mg/ml), penicillin (0.5 U/ml) and streptomycin (0.5 μg/ml). Age-matched non-inoculated or mice inoculated with normal brain comprised negative controls. Clinically ill mice were killed and 10% brain homogenates in PBS were prepared. Proteinase K digestions (50 μg/ml) were performed at 37°C for 1 h. Sho levels were calculated by densitometry (Scion Image) and comparison to a standard curve created from serial dilutions of non-infected brain homogenate. TgCRND8 mice (Chishti ) on a C3H/C57BL6 outbred background were killed at clinical illness (∼8 months) and 10% brain homogenates prepared in 0.32 M sucrose containing protease inhibitors.

Plasmid generation and recombinant protein

The Sho open reading frame was amplified from mouse genomic DNA and inserted between the HindIII and XbaI sites of either pcDNA3 or pBUD.CE4. GFP (Invitrogen). PrP and Sho mutants were generated using methods previously described (Drisaldi ), with constructs verified by DNA sequencing (details available on request). cDNA encoding residues 25–122 of the mouse protein was amplified from pcDNA3.Sho and inserted between the NdeI and BamHI sites of the pET-19b vector (Novagen) and expressed in E. coli BL21(DE3)/RIL (Stratagene). Following induction and lysis, filtered lysate was subjected to Ni2+ affinity chromatography (Qiagen), eluted with 250 mM imidazole and dialyzed into 20 mM Tns–Hcl pH8.0, 100 mM NaCl. The polyhistidine tag was removed with enterokinase (Roche). Purity was assessed by SDS–PAGE and concentration estimated using bovine serum albumin standards. Far-UV circular dichroism spectra of recombinant Sho in PBS pH 7.0 and a protein concentration of 0.1 mg/ml were collected at 22°C using a Jasco J-715 spectropolarimeter.

Sho antibody production

The peptide CRRTSGPGELGLEDDE (Sho residues 86–100 plus an N-terminal cysteine) was conjugated to maleimide-activated KLH (Pierce) and injected into New Zealand White rabbits. Alternatively, recombinant Sho was conjugated to KLH using EDC (1-ethyl-3-[3-dimethylaminopropyl] carbodiimide hydrochloride) chemistry (Pierce) and injected into rabbits. Antibodies were precipitated from serum using ammonium sulfate and then affinity purified using the respective peptide (SulfoLink column) or polypeptide (AminoLink Plus Coupling Gel column) immunogens. Epitope positions are shown in Figure 2.

Cell culture and PI-PLC treatment

N2a cells, HEK293 and CGN cells were cultured, transfected and lysed as described previously (Drisaldi ), except that staining for an activated caspase 3 neo-epitope (Cell Signaling Technology) was used to quantify apoptotic events and conditioned medium from adjacent wells replaced medium on transfected cells. CGN assays represent two or more independent triplicate transfections into Prnp0/0 cells, with observers being blind to the genotype of transfected test plasmids. For PIPLC treatment, 24 h post-transfection, Sho-transfected N2a cells were washed three times with PBS, and then treated with PI-PLC (Invitrogen) diluted in PBS for 40 min at 4°C. Cells were then washed twice and lysed as above.

Immunocytochemistry and immunohistochemistry

N2a or HEK293 cells 24 h post-transfection were washed with PBS, fixed with 4% paraformaldehyde, washed with PBS, blocked with 2% goat serum, then incubated with either Sho antibody (04SH-1, 6 μg/ml) or PrP antibody (8H4, 1 μg/ml), overnight at 4°C. Following PBS washes, cells were incubated with AlexaFluor488- or AlexaFluor594-conjugated secondary antibodies (Invitrogen; 1:300) for 2 h and then washed three times with PBS. For tissue analysis, mice were saline-perfused. Brains were fixed in Carnoy's fixative (10% glacial acetic acid, 60% methanol, 30% chloroform) bisected and processed. Six-micrometer sections were dried, deparaffinized and taken to TBS (pH 7.2) for incubation with primary antibody at 4°C (anti-Sho 04SH-1 (1:100, this paper), anti-Sho 06SH-3 (1:200, this paper), anti-Sho 06rSH-1 (1:50, this paper), anti-PrP 3F4 (1:20 000; Signet), anti-PrP 7A12 (1:10 000; Li ), and anti-calbindin D-28K (1:2000; Chemicon). Other antibodies used include anti-Thy-1, R194 (a generous gift from Roger Morris); anti-NSE, rabbit polyclonal from Polysciences Inc.; anti-synaptophysin, SY38 from Chemicon; anti-actin, rabbit polyclonal from Sigma (20–33); and anti-neurofilament H, SMI-32 from Sternberger Monoclonals Inc. For peptide blocking of antibody reactivity, peptide was added in four-fold mass excess to Sho antibody and incubated overnight at 4°C before use. Sections were rinsed in TBS and processed in DakoCytomation EnVision Labelled Polymer, visualized with DAB (3,3′-diaminobenzidine) and counterstained with Harris' hematoxylin. Images were obtained with a Leica DM6000B microscope using a Micropublisher 3.3RTV camera (Q Imaging Inc.) and OpenLab4.0.4 software (Improvision). Images in Figures 5L, M and 6E were de-convoluted using the iterative restoration function in Volocity. For fluorescent double labeling experiments, the secondary antibodies used were AlexaFluor-488 (green) and AlexaFluor-594 (red)-conjugated antibodies (Invitrogen). For immunocytochemistry on CGNs (Figure 6E), cells were fixed 24 h post-transfection in ice-cold methanol for 5 min at −20°C, and then processed and stained.

In situ hybridization

pcDNA3.Sho plasmid was linearized with HindIII or BglII to generate antisense and sense probes, respectively, using the DIG RNA labeling kit (Roche). Sections (6 μm) of formalin-fixed brains were cut using an RNase-free blade, mounted and dried overnight at 63°C. Tissue processing and in situ hybridization was performed as described by Peterson . DIG-labeled RNA probes were diluted 1/200–1/400 in hybridization buffer. Alkaline phosphate substrate NBT/BCIP was added, color development performed for 16 h in the dark and slides mounted without counterstaining.

Brain preparation and Western blotting

Mouse whole-brain homogenates were prepared in 0.32 M sucrose containing complete protease inhibitor cocktail (Invitrogen). For preparation of crude membranes, homogenates were spun at 700 g for 10 min at 4°C, pellets washed with 1 volume of homogenization buffer, spun again as above, supernatants were pooled then spun at 100 000 g for 1 h at 4°C and pellets resuspended in cold 50 mM Tris–HCl, pH 7.4, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100. After spinning (100 000 g, 1 h, 4°C) the supernatant was stored at −80°C. Protein was separated on either NuPAGE 4–12% gels (Invitrogen) using the MES buffer system, or by conventional SDS–PAGE using 14% polyacrylamide gels and transferred to either nitrocellulose (PrP blots, 3% BSA block) or PVDF (Sho blots, 5% non-fat skim milk block). Blots were incubated overnight with primary antibodies (D13 or D18 at 0.5 μg/ml for PrP (InPro Inc.), or 04SH-1 or 06rSH-1 Sho polyclonals at 1:500 or 1:2000, respectively, incubated with HRP-conjugated secondary antibody and developed using ‘Western Lightning' ECL (Perkin-Elmer). The embryo blot was supplied from Zyagen (San Diego, CA), probed with 04SH-1 and then stripped using 0.2 M glycine pH 2.2, 1% Tween-20, 0.1% SDS, followed by reprobing with D13 antibody

Supplementary Figures