Prion diseases are a group of invariably fatal neurodegenerative disorders that include Creutzfeldt–Jakob disease in humans, scrapie in sheep and goats, and bovine spongiform encephalopathy in cattle. The infectious agent or prion is largely composed of an abnormal isoform (PrPSc) of a host encoded normal cellular protein (PrPc). The conversion of PrPc to PrPSc is a dynamic process and, for reasons that are not clear, the distribution of spongiform change and PrPSc deposition varies among prion strains. An obvious explanation for this would be that the transformation efficiency in any given brain region depends on favourable interactions between conformations of PrPc and the prion strain being propagated within it. However, identification of specific PrPc conformations has until now been hampered by a lack of suitable panels of antibodies that discriminate PrPc subspecies under native conditions. In this study, we show that monoclonal antibodies raised against recombinant human prion protein folded into α or β conformations exhibit striking heterogeneity in their specificity for truncations and glycoforms of mouse brain PrPc. We then show that some of these PrPc isoforms are expressed differentially in certain mouse brain regions. This suggests that variation in the expression of PrPc conformations in different brain regions may dictate the pattern of PrPSc deposition and vacuolation, characteristic for different prion strains.
Prion diseases are invariably fatal transmissible neurodegenerative disorders including scrapie in sheep and goats, bovine spongiform encephalopathy (BSE) in cattle, and Creutzfeldt–Jakob disease (CJD) in humans. The infectious agent or prion is mainly composed of a detergent‐insoluble and protease‐resistant isoform (PrPSc) of the cellular prion protein (PrPc) (Prusiner, 1982). The acquisition of protease resistance is explicable by the post‐translational and autocatalytic conversion of PrPc from a largely α‐helical conformation into one rich in β‐sheet (PrPSc). PrPc is almost ubiquitously expressed and conserved in several mammalian species (Oesch et al., 1991). Although the highest levels are found in neurons (Bendheim et al., 1992), its precise physiological role remains unknown (Büeler et al., 1992). Two sites of non‐obligatory Asn‐linked glycosylation at residues 180 and 196 and a disulphide bond between cysteine residues at 178 and 213 have been identified (Caughey, 1993). Mature and fully glycosylated mouse PrPc migrates at 33–35 kDa on electrophoretic gels and its unglycosylated counterpart at 27 kDa (Haraguchi et al., 1989). In human brain, two other amino‐terminal truncated prion proteins have also been identified, resulting from endogenous proteolytic cleavage. Their unglycosylated forms migrate at 18 and 21–22 kDa, respectively (Jimenez‐Huete et al., 1998). PrPc is anchored at the cell surface by a carboxy‐terminal glycosylphosphatidylinositol moiety (Stahl et al., 1990).
Within species, the disease phenotype is not uniform and prion ‘strains’ can be differentiated on the basis of the incubation period and the neuropathological changes they induce in experimentally infected inbred mouse lines. The prevailing view is that strain diversity is determined by variations in PrPSc conformation or glycoform composition (Bessen and Marsh, 1994; Collinge et al., 1996; Telling et al., 1996). Neuropathologically, the precise regional variation in vacuolation and PrPSc deposition suggests strain‐specific targeting of particular neuronal populations (Bruce et al., 1994a). However, it remains to be explained how, during prion propagation, PrPSc selectively accumulates in some brain regions and not in others. An attractive hypothesis would be that alternative PrPSc conformations interact more or less efficiently with subspecies or isoforms of PrPc differentially expressed in certain brain regions or neuronal populations. The aim of this study therefore was to determine if such anatomical variation in the CNS expression of PrPc isoforms exists.
Using a new panel of monoclonal antibodies (mAbs) that exhibit differential affinity for truncated and glycosylated forms of native PrPc, we now show that there are indeed qualitative differences in PrPc expression in normal brain.
Material and methods
Production of the ICSM antibodies
All the experiments with mice have been performed in compliance with our institutional and national guidelines. FVB/N PrP null mice [FVB/N Prnp0/0 or Zurich I mice (Büeler et al., 1992)] were immunized subcutaneously with 50–100 µg of human recombinant PrP91–231 folded either into α (to produce ICSM 1–26) or β (to produce ICSM 35) conformation (Jackson et al., 1999b) in adjuvant on days 0, 21 and 42, and then finally boosted intraperitoneally on day 50 with 50 µg in phosphate‐buffered saline (PBS). Three days later, the mice were culled and single cell suspensions of splenocytes were cryopreserved. These were later thawed and then fused with non‐secreting NS0 cells using conventional technology, and hybridomas were subsequently screened for reactivity to α‐ or β‐PrP and to native PrP. Positive hybridomas were cloned repeatedly until stable.
Peptide enzyme‐linked immunosorbent assay (ELISA)
High binding ELISA plates were coated with 50 µl of a 10 µg/ml solution of overlapping 15‐ to 20‐mer mouse and human PrP peptides in ELISA coating buffer (35 mM sodium bicarbonate, 15 mM sodium carbonate, pH 9.4). The plates were incubated for 1 h at 37°C and then washed three times with PBS–0.05% Tween. After blocking with RPMI media supplemented with 10% fetal calf serum, 50 µl of the relevant mAb (as culture supernatant) was added for 1 h at 37°C. After three washes in PBS–Tween, a 1/5000 dilution of a horseradish‐peroxidase conjugated anti‐mouse immunoglobulin G (IgG) (Sigma, Poole, UK) was added for 30 min at 37°C and washed a further three times. The plate was then developed with ortho‐phenylene diamine (OPD) buffer and the reaction was stopped with 3M sulphuric acid prior to spectrophotometric analysis.
Immunoprecipitation of murine PrPc
Brain tissues from three FVB/N and FVB/N Prnp0/0 (Zurich I) mice were homogenized (10% w/v. in PBS) with a Dounce homogenizer, and centrifuged at 1000 g. The supernatants were stored at –80°C until further use. For immunoprecipitation, brain homogenates were diluted to 0.5% in lysis buffer (20 mM Tris–HCl pH 7.5, 150 mM sodium chloride, 1% Nonidet P40, 0.5% sodium deoxycholate) with a cocktail of protease inhibitors (Roche Biochemicals, Lewis, UK). The solution was then incubated (1:1 dilution) with 10 µg/ml purified ICSM mAbs in PBS or with neat hybridoma supernatant for 2 h at 4°C on a rotator. Negative controls consisted of omitting the capture mAb or IgG1 (Ozato et al., 1980) and IgG2b isotype controls (Avent et al., 1988). The immune complexes were then adsorbed overnight to protein G‐agarose beads (Roche Biomedicals) at 4°C on a rotator. The beads were then washed with high and low salt buffers according to the manufacturer’s recommendation. After the final wash, the beads were resuspended in Laemmli buffer (Laemmli, 1970), heated at 100°C for 5 min to detach/denature the bound protein.
Enzymatic deglycosylation of immunoprecipitated PrPc
A 10–20 µl aliquot of the immunoprecipitated and subsequently denatured PrPc was digested with 1000 U of recombinant PNGase (New England Biolabs, Hitchin, UK) for 2 h at 37°C in 1% Nonidet P40 and the proprietary buffer. The deglycosylated proteins were then precipitated in three volumes of cold acetone and resuspended in 10–20 µl Laemmli buffer.
Immunoprecipitated protein (deglycosylated or not) was run on 12% polyacrylamide gels, electrotransferred onto PVDF membranes (Millipore, Watford, UK) and immunoblotted with 0.2 µg/ml of biotinylated ICSM 18 avoiding detection of the immunoprecipitating antibody. After several washes with PBS–Tween, a 1/10 000 dilution of streptavidin–horseradish peroxidase (Sigma) was added. Immunoreactivity was visualized with an enhanced chemiluminescence kit on autoradiographic films (ECL+; Amersham, Biosciences, Chalfont St Giles, UK). Biotinylated molecular weight markers (Amersham) were used to correlate the electrophoretic mobility of the immunoprecipitate accurately to its molecular weight.
Immunohistochemical studies were performed on five FVB/N and FVB/N Prnp0/0 (Zurich I) mice killed with an overdose of pentobarbital and from which the brains were rapidly removed, embedded in OCT compound and frozen on dry ice. Cryostat sections (8 µm) were cut, fixed in acetone for 10 min and air‐dried. Endogenous peroxidase was inactivated for 30 min with 0.3% H2O2 solution in methanol. After washing in PBS, non‐specific antibody binding was blocked with normal goat serum for 30 min. The sections were then stained for 1 h with either 10 µg/ml ICSM mAbs or with neat hybridoma supernatant; conditions optimized for specific binding using equivalent PrP null mouse sections. After washing in PBS, a 1/100 dilution of horseradish peroxidase‐conjugated anti‐mouse IgG (Sigma) was added for 45 min. Peroxidase activity was revealed with 3,3′‐diaminobenzidine tetrahydrochloride for 3–10 min (Sigma). Sections were counter‐stained with haematoxylin (Rosslab, Macclesfield, UK), mounted, and covered for microscopic observation.
Numbering of PrP residues corresponds to mouse PrP throughout the study.
Epitope mapping of the ICSM monoclonal antibodies
Initially 26 ICSM mAbs were produced from FVB/N Prnp0/0 mice immunized with human recombinant PrP91–231 produced in E.coli and refolded into a predominantly α‐helical PrPc‐like conformation (Jackson et al., 1999a). ICSM 35 was obtained after immunization with human recombinant PrP91–231 refolded into a β conformation (beta‐PrP) (Jackson et al., 1999a). Antibody epitopes were defined by peptide ELISA using overlapping mouse and human 15‐ to 20‐mer peptides covering the codon 91–231 sequence. ICSM 18 bound strongly to peptide 146–159, a central region encompassing the first α helix of PrPc (Riek et al., 1996). ICSM 17 recognized a similar region, between residues 140 and 159. ICSM 35 recognized a peptide between residues 96 and 109. None of the other ICSM mAbs recognized synthetic peptides absorbed to ELISA plates or inhibited mAb binding to recombinant protein in competition assays (data not shown), suggesting that their epitopes were conformation‐dependent.
Immunoprecipitation of native forms of PrPc
PrPc was immunoprecipitated from murine brain homogenates to identify native PrPc isoforms. Optimal conditions were determined using ICSM 35. Titration experiments showed that 5–10 µg/ml permitted maximal specific binding of PrPc, allowing immunoprecipitation from as little as 5 µg of mouse brain (data not shown). 250 µg of brain was used in all subsequent biochemical studies. We then studied PrPc immunoreactivity of the ICSM mAbs used either as purified antibody or culture supernatant (10 µg/ml). None reacted with brain homogenates from Prnp0/0 mice, confirming their strict specificity for PrPc (data not shown). ICSM 35 immunoprecipitated all the PrPc bands except for the 18 kDa truncated isoform (subsequently referred to as the C1 fragment, Fig. 1A,B). This was not unexpected given that its epitope lies N‐terminal to the C1 fragment cleavage site at residue 110/111. ICSM 18 apparently recognized all the native glycoforms of full‐length and truncated PrPc visualizable on 12% polyacrylamide gels (Fig. 1A–D). With other mAbs, the influence of Asn‐linked glycosylation on their PrPc reactivity was striking. ICSM 4 was uniquely specific for unglycosylated full‐length and C1 fragment isoforms (Fig. 1A,B). ICSM 3 and ICSM 1 immunoprecipitated a 29 kDa band in addition to those visualized with ICSM 4 that disappeared after PNGase treatment (Fig. 1A,B). This band represented full‐length PrPc monoglycoforms rather than the diglycoform of the C1 fragment, since it was not immunoprecipitated by the mAbs when the brain homogenate was first depleted of full length PrPc with ICSM 35 (data not shown). Thus ICSM 1 and 3 bind unglycosylated and monoglycosylated forms of full‐length and, to a lesser extent, 18 kDa‐truncated PrPc (Fig. 1C), but do not recognize diglycosylated PrPc. The fact that the same reactivity was seen when PrPc was denatured suggests that ICSM 4, 1 and 3 have epitopes at, or near, one or both of the sites of Asn‐linked glycosylation at residues 180 or 196. ICSM 10 had almost the same specificity, except that the sizes of the presumed monoglycoforms were larger than those recognized by ICSM 1 and 3, at around 31 kDa (Fig. 1A) and 25 kDa, respectively (Fig. 1C). This suggested that either ICSM 3, 1 and 10 binding is inhibited by a subset of glycans distributed across both sites of attachment (e.g. by virtue of size or charge) or that their epitopes are sterically hindered by all the glycans attached to one site and not the other. We performed immunoprecipitation from smaller quantities of material and demonstrated that the two bands of monoglycoforms immunoprecipitated by ICSM 1 and 10 represent all the full‐length monoglycoforms, immunoprecipitated by ICSM 18 or 14 (Fig. 1D).
We quantitated the differential immunoreactivity by densitometry (Fig. 2). ICSM 18 and 35 reacted similarly with either full‐length or the 21–22 fragment of PrPc, although ICSM 35 did not bind the C1 fragment. ICSM 4 reacted with ∼5% of full‐length PrPc or its C1 fragment, representing the majority (75%) of the unglycosylated molecules recognized by ICSM 18.
Distribution of native PrPc in the mouse brain
We then looked for regional differences in PrPc expression in fresh frozen coronal and sagittal sections of normal mouse brain and observed, using ICSM 18, an antero‐posterior gradient of PrPc expression from all five FVB/N mice studied. On anterior coronal sections, PrPc immunoreactivity was pronounced in the cerebral cortex (particularly the pyriform and cingulate cortex), the striatum, the ventral pallidum and the fibre bundles of the genu of the corpus callosum (Fig. 3). The olfactory bulb was also strongly positive (see Fig. 5). More centrally, the hippocampus was strongly immunoreactive for PrPc. Dense neuropil PrPc immunoreactivity was observed in the oriens and stratum radiatum synaptic layers. The molecular layer (synaptic region), the lacunosum moleculare and the polymorph layer from the dendate gyrus were also positive. In contrast, the pyramidal cell layer of the hippocampus and the granular cell layer of the dendate gyrus were negative (Fig. 3). In the midbrain and brain stem, PrPc was detected in the substantia nigra (Fig. 4), but not in the surrounding cerebral peduncle and expressed very weakly in the medulla oblongata (data not shown). Although the molecular cell layer of the cerebellum was positive, PrPc was not detected in Purkinje cells or in the cerebellar granular cell layer (Fig. 4). We then compared ICSM 18 staining of these regions with several other mAbs.
Under‐representation of unglycosylated and monoglycosylated forms of PrPc
ICSM 4, which recognizes the major proportion of unglycosylated full‐length (Fig. 2) and 18 kDa‐truncated PrPc (Fig. 1) showed weak staining in the alveus and the external capsule at the border of the CA1 region of the hippocampus (Fig. 4), other brain regions being negative. By comparing ICSM 4 staining with ICSM 1 and ISCM 10, we found PrPc monoglycoforms to be expressed in the oriens layer of the hippocampus, the substantia nigra and the fimbria (Fig. 4). The striatum, the corpus callosum and the anterior commissure were also slightly positive (data not shown). Interestingly, the region surrounded by the granular cell layer of the cerebellum was also positive with ICSM 10 (Fig. 4). It is noteworthy that PrPc monoglycoforms were not detectable in regions where ICSM 18 staining was particularly strong such as the olfactory bulb, the cerebral cortex, the ventral pallidum, the hippocampus and the outside part of the molecular cell layer of the cerebellum (data not shown), suggesting that in these regions diglycosylated PrPc predominates.
Differential distribution of full‐length and truncated PrPc
As ICSM 18 and ICSM 35 bound PrPc molecules with similar efficiency except for the C1 fragment (Figs 1 and 2), we compared their staining to map the distribution of C1. As for whole brain immunoprecipitation, ICSM 35 and ICSM 18 exhibited on all sections similar immunoreactivity in the majority of the brain regions, such as the cortex, the hippocampus, and the olfactory bulb (Fig. 5 and data not shown). In some areas such as the CA3 region of the hippocampus, ICSM 35 staining was even stronger and more granular (data not shown). In contrast, staining in the dorsal hippocampal commissure, in the substantia nigra and in the cerebellum was much weaker (Fig. 5). Comparative western blots (with ICSM 18) of the cerebellum and the cerebrum were also performed and confirmed that C1 was more abundant in the cerebellum and full‐length PrPc was predominant in the cerebrum (Fig. 6).
The heterogeneity of normal cellular prion protein under denaturing conditions has long been recognized. We now show for the first time that similar heterogeneity exists for PrPc isoforms in their native state. Although the highest levels of PrPc are found in neurons (Bendheim et al., 1992), its precise physiological role remains unknown and transgenic mice devoid of PrPc (Prnp0/0) have little phenotypic abnormality (Büeler et al., 1992). PrPc may have a physiological role in neuronal differentiation (Wion et al., 1988), synaptic transmission (Collinge et al., 1994) and copper binding (Jackson et al., 2001), but the role of heterogeneous PrPc subspecies, particularly the role of differential glycosylation, remains obscure. Very few other CNS proteins have been found thus far with as many isoforms (Liesi et al., 2001; Matthews et al., 2002) and further heterogeneity based on protein conformation alone is likely. Intriguingly, we found that the epitopes of some antibodies were sterically hindered by either one (ICSM 1, 3, 10) or both (ICSM 4) sites of Asn‐linked glycosylation. Furthermore, ICSM 1 and ICSM 10 recognized monoglycoforms with differing electrophoretic mobility. The fact that the epitope of ICSM 4 was sterically hindered by glycans attached to either site indicates that both sites of attachment are very close in PrPc, and raises the possibility that ICSM 1 and 10 recognize a subset of glycans attached at both sites. It has been reported that glycans attached at Asn 196 have a higher proportion of tri‐ and tetra‐antennary glycans than those attached to Asn 180 (Stimson et al., 1999) and, since monoglycoforms recognized by ICSM 10 run more slowly on polyacrylamide gels, its epitope is likely to be in the vicinity of Asn 180 with the epitope of ICSM 1 at, or near to, Asn 196. Although none of these mAbs recognized synthetic peptides, they showed a similar pattern of specificity for denatured un‐ and mono‐glycosylated PrPc in western blots (data not shown), suggesting that their epitopes were relatively insensitive to changes in protein conformation. ICSM 35 has a definable epitope between residues 96 and 109 and, although it did not immunoprecipitate the 18 kDa‐fragment starting at residue 110–111 (Chen et al., 1995), it readily immunoprecipitated the different glycoforms of both full‐length PrPc and the 21–22 kDa fragment.
Regional distribution of glycoforms and truncations of PrPc in the mouse brain
Detection of denatured PrPc has been reported on membranes by histoblot (Taraboulos et al., 1992) and on fixed sections (Manson et al., 1992), and we now show that PrPc in its native state has highest expression in anterior brain regions (Fig. 3). We also found that synaptic layers of the hippocampus and the dendate gyrus expressed high levels of PrPc, although the cell bodies were negative; this is consistent with the view that the normal prion protein plays a physiological role at the synaptic level (Collinge et al., 1994). In the cerebellum, the molecular cell layer was positive and the Purkinje cell layer was negative confirming recent studies on fixed sections of mouse brain (Liu et al., 2001).
Clearly unglycosylated full‐length and 18 kDa‐truncated forms of prion protein are less abundant in the mouse brain, expression being 5% of total mouse PrP (Fig. 2). However, this is not due to a global reduction in unglycosylated PrP levels as expression was mainly restricted to fibre bundles of the alveus and to the external capsule at the border of the hippocampal CA1 region. Small amounts of monoglycosylated PrPc were present in the corpus callosum, the striatum, the anterior commissure, the alveus/external capsule region and the cerebellum. Given that ICSM 3, 1 or 10 immunoreactivity was absent in the regions where ICSM 18 was strongly detected, we infer that diglycosylated PrPc is over‐expressed mainly in the cortex, the ventral pallidum, the hippocampus, the olfactory bulb and part of the cerebellum.
We were also able to map truncated and full‐length forms of PrPc in the brain by comparing the immunoreactivity of ICSM 35 (reactive with all but the C1 fragment) with ICSM 18. These studies showed that the cerebellum, the substantia nigra and, to a lesser extent, the dorsal hippocampal commissure were enriched in C1, confirmed for the cerebellum by western blot (Fig. 6), and similar to the findings of Liu et al. (2001). Interestingly, the generation of the C1 fragment is thought to be under the control of matrix metalloproteinases (Jimenez‐Huete et al., 1998; Vincent et al., 2001) and the highest expression of one of the these, ADAM 10, is in the cerebellum (Karkkainen et al., 2000), where we have found the highest expression of truncated PrPc. Our study suggests, therefore, that there is considerable variability in the production and/or degradation of PrPc isoforms in different brain regions.
Regional variability of brain PrPc expression may participate in the neuropathological and biochemical phenotype of prion strains
It is tempting to speculate that the heterogeneity of PrPc expression observed here could be fundamental to the differential propagation of prion strains. In the same mouse genotype such as the FVB/N mouse, several prion strains can be differentiated on the basis of the unique neuropathological and biochemical changes they induce. Indeed, for each strain, PrPSc deposition and vacuolation are targeted precisely to particular sets of neurons (Bruce et al., 1994b; DeArmond et al., 1997). The wealth of information from studies of prion transmission indicates that the nature of PrPSc synthesized in the infected host is dictated by inoculated PrPSc conformation (Bessen and Marsh, 1994; Collinge et al., 1996; Telling et al., 1996). Given the findings reported here, an attractive hypothesis would be that combinations of PrPc fragments and glycoforms expressed within specific brain regions might favourably interact with exogenous prions of one strain and not others. Thereafter, prion replication at these sites might incorporate all, or a subset of, the available PrPc molecules into the enlarging protein aggregate. Thus the strain is defined both anatomically and biochemically, with the clinical phenotype being defined by the anatomical distribution of disease. Interestingly, the regional distribution of PrPc obtained with ICSM 18 is consistent with that described using a carboxy‐terminal anti‐PrPc antibody on fixed sections (Liu et al., 2001). Moreover, the overall distribution, the gradient expression and the predominance of diglycosylated PrPc we observed have been found to be similar in man and primate brain (Sales et al., 1998; Moya et al., 2000). It is therefore possible that some prion strains target regions in diverse species expressing similar PrPc isoforms. This is generally the case for bovine spongiform encephalopathy prions, as they usually produce a similar biochemical phenotype after crossing the species barrier into mice, humans and primates (Collinge et al., 1996; Lasmezas et al., 2001). However, the recent demonstration that they may also produce two distinct PrPSc patterns in several strains of wild‐type mice (Asante et al., 2002) suggests that conformations of PrPc specific of these animal strains may interact with PrPSc. Finally, the PrPSc distribution may be heterogeneous in the brain of a single prion‐infected individual (Hainfellner et al., 1999), suggesting also a potential role of PrPc. This far more complex situation has been recently highlighted in vitro, where it was shown that strain‐specific PrPSc glycoforms were the consequence of a complex interaction between PrPSc, the glycosylation state of PrPc and the cellular compartment where the conversion occurred (Vorberg and Priola, 2002). Regional differences in protein glycosylation may not only be restricted to constituent proteins such as PrPc, as considerable heterogeneity in glycosylation of a viral protein eluted from different brain regions has been described (Lynch and Sharpe, 2000). Perhaps many proteins undergo differential post‐translational modification within distinct CNS regions.
In summary, we have shown that PrPc isoforms (including its different truncations and glycosylations) have a differential anatomical distribution in the mouse brain. This raises the possibility that exogenous PrPSc (or PrPSc exogenous to the region in question) targets and accumulates in neurons expressing compatible PrPc isoforms, which then synthesize new PrPSc molecules thereby producing a pattern anatomically distinct for each prion strain.
We wish to thank Charles Weissmann for helpful discussions, Jackie Linehan for technical assistance and Ray Young for graphic editing. This work was supported by grants from the Medical Research Council (London, UK), the Fondation pour la Recherche Médicale and NATO (Paris, France). V.B. is the recipient of a Marie Curie fellowship.