Lethal recessive myelin toxicity of prion protein lacking its central domain

Frank Baumann, Markus Tolnay, Christine Brabeck, Jens Pahnke, Ulrich Kloz, Hartmut H Niemann, Mathias Heikenwalder, Thomas Rülicke, Alexander Bürkle, Adriano Aguzzi

Author Affiliations

  1. Frank Baumann1,
  2. Markus Tolnay1,
  3. Christine Brabeck2,,
  4. Jens Pahnke1,,
  5. Ulrich Kloz3,
  6. Hartmut H Niemann4,
  7. Mathias Heikenwalder1,
  8. Thomas Rülicke5,
  9. Alexander Bürkle2 and
  10. Adriano Aguzzi*,1
  1. 1 Institute of Neuropathology, University Hospital of Zurich, Zurich, Switzerland
  2. 2 Molecular Toxicology Group, Department of Biology, University of Konstanz, Konstanz, Germany
  3. 3 Transgenic Core Facility, German Cancer Research Center, Heidelberg, Germany
  4. 4 Strukturbiologie Helmholtz‐Zentrum für Infektionsforschung GmbH, Braunschweig, Germany
  5. 5 Institute of Laboratory Animal Science and Research Center Biomodels Austria University of Veterinary Medicine Vienna, Vienna, Austria
  1. *Corresponding author. Department of Pathology, Institute of Neuropathology, University Hospital of Zurich, Schmelzbergstrasse 12, 8091 Zurich, Switzerland. Tel.: +41 1 255 2107; Fax: +41 1 255 4402; E‐mail: adriano.aguzzi{at}
  • Present address: Institute for Pathology, Staedt. Klin. Brandenburg GmbH, 14770 Brandenburg, Germany

  • Present address: University of Rostock, Neurological Clinic, 18147 Rostock, Germany


PrPC‐deficient mice expressing prion protein variants with large amino‐proximal deletions (termed PrPΔF) suffer from neurodegeneration, which is rescued by full‐length PrPC. We now report that expression of PrPΔCD, a PrP variant lacking 40 central residues (94–134), induces a rapidly progressive, lethal phenotype with extensive central and peripheral myelin degeneration. This phenotype was rescued dose‐dependently by coexpression of full‐length PrPC or PrPC lacking all octarepeats. Expression of a PrPC variant lacking eight residues (114–121) was innocuous in the presence or absence of full‐length PrPC, yet enhanced the toxicity of PrPΔCD and diminished that of PrPΔF. Therefore, deletion of the entire central domain generates a strong recessive‐negative mutant of PrPC, whereas removal of residues 114–121 creates a partial agonist with context‐dependent action. These findings suggest that myelin integrity is maintained by a constitutively active neurotrophic protein complex involving PrPC, whose effector domain encompasses residues 94–134.


PrPSc is the only known constituent of mammalian prions (Prusiner, 1982), the infectious agents causing transmissible spongiform encephalopathies. PrPSc is an isoform of the cellular prion protein PrPC (Oesch et al, 1985), which is expressed in a broad range of vertebrate tissues (Bendheim et al, 1992). PrPC is secreted into the endoplasmic reticulum and processed by removal of its amino‐terminal signal peptide, N‐linked glycosylation at two sites, and replacement of its carboxy‐terminal tail by a glycosyl phosphatidyl inositol (GPI) group, which anchors it to the cell surface. Ablation of Prnp, the gene encoding PrPC, abrogates prion replication (Bueler et al, 1993) and scrapie pathogenesis (Brandner et al, 1996).

The physiological function of PrPC is unclear (Aguzzi and Polymenidou, 2004). Prnpo/o mice enjoy a normal life expectancy (Bueler et al, 1992). Changes in circadian rhythms, alterations of hippocampal function, and behavioral abnormalities were reported in Prnpo/o mice, but their molecular basis has not been clarified. PrPC may have antiapoptotic (Kuwahara et al, 1999; Bounhar et al, 2001; Chiarini et al, 2002; Zanata et al, 2002) or proapoptotic properties (Paitel et al, 2002, 2003, 2004). A possible superoxide dismutase activity of PrPC (Brown et al, 1997) has not been confirmed in vivo (Hutter et al, 2003). Unsurprisingly, a unified view of the above observations is still missing.

Transgenic expression of PrP variants carrying amino‐proximal interstitial deletions (Δ32–121, termed PrPΔE, and Δ32–134, termed PrPΔF) causes ataxia, progressive cerebellar granule cell degeneration, leukoencephalopathy, and death of Prnpo/o mice at 3–4 months of age (Shmerling et al, 1998; Radovanovic et al, 2005). Toxicity was associated only with PrP versions whose deletions extended into the highly conserved hydrophobic core (HC) domain (112–133), which is situated between the flexible amino‐proximal tail of PrPC and its globular carboxy‐proximal domain (Wuthrich and Riek, 2001). A similar phenotype was detected in mice overexpressing Dpl (Moore et al, 1999; Weissmann and Aguzzi, 1999; Rossi et al, 2001), whose structure resembles that of PrPΔF. In each case, the pathological phenotype was rescued by substoichiometric coexpression of full‐length PrPC, either from the endogenous Prnp locus or as a transgene driven by various promoters. Those results suggested that PrPΔF and Dpl interfere with a physiological function of PrPC in the brain.

PrPΔF lacks the octarepeat region (OR) and a central domain (CD) that comprises a charge cluster (CC, residues 95–110) and a hydrophobic core (HC, residues 112–134) (Figure 1A and B). To analyze the relative contribution of these domains to the phenotype, we generated Prnp transgenes lacking the entire CD (henceforth termed PrPΔCD) or part of the HC domain (termed PrPΔpHC). Whereas PrPΔpHC was innocuous even in the absence of full‐length PrPC, expression of PrPΔCD caused a drastic neuropathological phenotype and death within 20–30 days. This phenotype was partially counteracted by high levels of PrPC and, to a lesser extent, by a PrP mutant lacking the octarepeats. While PrPΔpHC behaved neutrally in the absence of PrPC, it partially rescued the toxic effects of PrPΔF, whereas it accentuated those of PrPΔCD.

Figure 1.

(A) Hydrophobicity plot of the full‐length murine prion protein. SP: secretory signal peptide, cleaved after sorting of the precursor to endoplasmic reticulum; repeats: five repeats of eight amino acids; CC: charge cluster; HC: hydrophobic core; CD: central domain; H1, H2, H3: α‐helices 1, 2, 3 of the globular carboxy‐proximal domain; MA: membrane anchor of precursor protein, replaced during maturation with glycosyl phosphatidyl inositol anchor. (B) Scheme of wild‐type prion protein (WT PrP) and deletion mutants utilized for generation of transgenic mice. Antibody POM1 recognizes a carboxy‐proximal epitope, whereas POM11 binds to octarepeats, and POM3 binds to the charge cluster. (C) Southern blot analysis of PrPΔCD (lines Tg1046, Tg1047, and Tg1050) and PrPΔpHC transgenic mice (line Tg902). Genomic DNA was digested with EcoRI before loading. PrPΔCD transgenic band: 2916 bp; PrPpHC: 3015 bp; Prnpo: 2747 bp; Prnp+: 2148 bp. The lanes denoted phgΔCD and MoPrP‐ΔH1 contain the plasmid constructs used for pronuclear injection. Numbers below the respective lanes indicate the numbers of transgenic copies per haploid genome, as determined by quantitation of 32P Southern blotting signals against the respective Prnp+ genomic band. Line Tg1050 contained the largest number of transgenic copies. (D) Transgenic PrP transcription. RNA was isolated from PrPwt, PrPΔCD+/o (Tg1047), and PrPΔpHCo/o (Tg902) brains, reverse‐transcribed, and PrP‐containing cDNA was amplified with specific primers. Wild‐type brain and original plasmid served as molecular weight standards. DNase‐treated, non‐retrotranscribed samples were loaded for control.


Characterization of interstitially deleted PrP species in transgenic mice

All PrP mutants described below are based on the ‘half‐genomic’ pPrPHG backbone containing a redacted murine Prnp gene from which intron #2 was deleted, and flanked by 6 and 2.2 kb of 5′ and 3′ genomic regions, respectively (Borchelt et al, 1996; Fischer et al, 1996). PrPΔC lacks residues 32–93 within the OR region, and PrPΔC mice do not display any pathological phenotype (Shmerling et al, 1998). The deletions in PrPΔE and PrPΔF extend into the HC domain (aa 32–121 and 32–134; Figure 1A and B) and induce neuropathological lesions in Prnpo/o mice. These lesions may be caused by the ablation of CD, by the combined ablation of octarepeat and CD, or by the disruption of HC. We tested these alternatives by creating additional PrP mutants with interstitial deletions of amino acids 94–134 (termed PrPΔCD) and 114–121 (termed PrPΔpHC; Figure 1B).

PrPΔCD was injected into Prnp+/o zygotes derived from crosses between Prnpo/o and C57BL/6 mice. Three of 35 offspring were found to carry the transgene. Surprisingly, all three pups died between 10 and 23 days of age, suggesting perinatal toxicity. Further microinjections into hybrid B6D2F1 Prnp+/+ zygotes resulted in eight PrPΔCD and four PrPΔpHC founders carrying diploid Prnp alleles (henceforth termed PrPΔCD+/+ and PrPΔpHC+/+, with superscripts defining the Prnp allelic status and subscripts denoting the respective hemizygous transgenes).

PrPΔCD founder mice 1046, 1047, 1048, and 1050, as well as PrPΔpHC founder mice 630, 642, and 902 exhibited undistorted Mendelian transmission of the transgene. Four PrPΔCD lines (Tg1046, Tg1047, Tg1048, Tg1050) and one PrPΔpHC line (Tg902) were chosen for further analyses. Quantitative Southern blotting revealed that the number of transgenic copies varied between 1 and 5 per haploid genome (Figure 1C). Each line was crossed twice to Prnpo/o mice, and F2 offspring (PrPΔCDo/o and PrPΔpHCo/o) were analyzed for transgenic transcription (Figure 1D) and protein expression (Figure 2).

Figure 2.

Expression of transgenic proteins. (A) Similar glycosylation patterns of full‐length PrP, PrPΔCD, and PrPΔpHC. Brain homogenates were subjected to PNGase F treatment as indicated and analyzed by Western blotting using anti‐PrP antibody POM1 (Polymenidou et al, 2005). Arrowhead: C1 fragment. (B) Density gradient analysis of detergent‐resistant membrane (DRM) preparations from the brains of wild‐type, PrPΔCD, and PrPΔpHC transgenic mice. Fractions 4–16 of the step gradient were analyzed by Western blotting. Full‐length PrP, PrPΔCD, and PrPΔpHC were detectable to a similar extent in the upper fractions, indicating that they all localize to DRMs similar to flotillin (48 kDa). Non‐DRM contaminations were controlled for by probing blots with antibodies to GAPDH (lowest panel, 35 kDa). Non‐buoyant PrP may indicate raft disruption or may represent immature protein fractions. (C, D) Expression levels of PrPΔCD and PrPΔpHC. Total brain homogenates were digested with PNGase F treatment and analyzed using the monoclonal antibody POM11 (*PrPwt band. **PrPΔCD band). (D) Densitometric quantification of protein expression after Western blotting. Each column represents the average of at least three individual mice from each transgenic strain coexpressing wild‐type PrP from one haploid genomic allele. (E–S) POM1‐immunostained brain sections. (E–I) Overview; (J–N) higher magnification of hippocampus and corpus callosum; (O–S), cerebellum. Similar spatial expression patterns for PrPC and its variants in wild‐type, PrPΔCD , PrPΔpHC, and PrPΔF mice. No immunoreactivity was visible in Prnpo/o mice. Scale bar: (E–I) 10 mm; (J–N) 100 μm; (O–S) 10 μm.

In all transgenic lines, brain expression of PrP mutants was lower than that of endogenous PrPC (Figure 2C) and paralleled the number of integrated transgenic copies. Tg1046 PrPΔCD and Tg902 PrPΔpHC expressed <20% of Prnp+/o, whereas Tg1047, Tg1048, and Tg1050 expressed ca. 35, 80, and 75% of Prnp+/o, respectively (Figure 2D). The anatomical distribution of the transgenic proteins resembled that of PrPΔF (Figure 2E–S). No pathological aggregates were detected by PrP immunohistochemistry with antibody SAF84 (data not shown).

Upon PNGase‐F treatment, PrPΔCD, PrPΔpHC, and PrPC displayed enhanced electrophoretic motility similar to PrPC (Figure 2A), suggesting that each protein was N‐glycosylated to a similar extent. PrPC displayed a significant amount of C1 fragment (Figure 2A, arrowhead), yet C1 formation was reduced in PrPΔpHC and absent from PrPΔCD, suggesting abnormal endoproteolysis of these proteins. We then analyzed detergent‐resistant membranes from PrPC, PrPΔCD, and PrPΔpHC brains by step density‐gradient centrifugation. The buoyancy of PrPΔCD and PrPΔpHC was similar to that of PrPC and flotillin (Figure 2B), suggesting that they all reside in similar membrane microdomains. Therefore, most aspects of PrPΔCD and PrPΔpHC biogenesis appear similar to those of PrPC.

Pathological phenotypes of mice expressing interstitially deleted PrP species

All PrPΔCD transgenic lines (Tg1046, Tg1047, Tg1048, and Tg1050) were maintained in the Prnp+/+ background. Founders Tg1046, Tg1047, and Tg1050 remained healthy, whereas founder Tg1048, which displayed the highest expression of the transgene, developed behavioral abnormalities at 120 days of age (Supplementary movie). However, its entire transgenic PrPΔCD+/o offspring died at 20 days of age. As incipient paraparesis of founder Tg1048 precluded natural breeding, sperm was used for in vitro fertilization of Prnp+/+ C57BL/6 oocytes. This resulted in a total of 117 Prnp+/+ pups, of which 46 carried the PrPΔCD transgene. The gender distribution of pups was mendelian (58 ♂; 59 ♀), but transgenic males were marginally underrepresented (17 ♂; 29 ♀; χ2 P<0.03). By 50 days of age, all (46/46) Tg1048 PrPΔCD+/+ animals had developed ataxia, which progressed over 20–30 days, eventually leading to the inability to stride on cage grids. By the age of 80 days, 40/46 animals had developed spastic paraparesis, which by 95 days progressed to full paraplegia. A four‐degree clinical score was developed for monitoring and animal welfare purposes (Table I). Mice were euthanized within 3 days of reaching a score of 3.5, or within 1 day of developing a score of 4 (Figure 3E and Supplementary Figure 1). By the age of 99 days all 46 animals had to be euthanized (Figure 3A). All transgenic Tg1050 PrPΔCD+/+ mice developed a disease similar to that of Tg1048, but progression was slower. First signs became noticeable at 110 days and progressed to terminal disease at 160 days (Figure 3B). Tg1046 and Tg1047 PrPΔCD+/+ mice, which expressed lower levels of transgenic protein, did not experience spontaneous disease before 450 days of age.

Figure 3.

(A–D) Survival of PrPΔCD mice. Longevity was dependent on their Prnp status, with Prnp+/+ transgenic mice surviving longer than Prnp+/o mice and Prnpo/o mice showing the shortest survival times. Each line represents ⩾5 individuals. (E) Typical clinical phenotypes at terminal stage: kyphosis (left), ataxia (middle), and clasping when held head‐down at the tail (right). See also Supplement movie online. (F) Several litters from a Tg1047 breeding pair were monitored for their weight. Transgenic mice (PrPΔCD+/o) were constantly underweight as compared to their non‐transgenic littermates (PrP+/o). (G) The ratio between the expression of PrPΔCD and full‐length PrPC determines the life expectancy of mice. The black bars depict the ratios of steady‐state level PrPΔCD/PrPwt, as determined by quantitative Western blot analysis of brain homogenates of at least four transgene‐positive individuals of the indicated line. The gray bars represent the mean life expectancy of ⩾4 transgenic individuals of each line.

View this table:
Table 1. Scoring of neurological signs in mice expressing PrP variants

Genetic interactions between PrPC and mutated PrP

In order to probe genetic interaction between PrPΔCD and PrPC, we removed Prnp through sequential crosses of each transgenic line with Prnpo/o mice. Tg1050 PrPΔCD+/o offspring developed clinical signs by 29 days and reached clinical scores of >3 by 45 days (Figure 3B). Similarly, Tg1047 PrPΔCD+/o mice developed neurological disease at 75–80 days (Figure 3C) and were consistently underweight despite facilitated access to food and water (Figure 3F). Only 20% of Tg1046 PrPΔCD+/o offspring developed disease within 500 days (Figure 3D), yet PrPΔCDo/o mice developed clinical signs at 17–21 days and progressed to terminal disease by 23–28 days. The most dramatic phenotype was seen in Tg1050 PrPΔCDo/o mice, which developed perinatal ataxia and terminal disease before 10 days of age (Figure 3B). We conclude that reduction or removal of PrPC results in greatly exacerbated phenotypes.

We then tested whether overexpression of PrPC antagonizes PrPΔCD toxicity. We crossed Tg1050 PrPΔCDo/o to tga20 mice, which overexpress PrPC six‐fold. In order to circumvent the early lethality of Tg1050 PrPΔCDo/o mice, we first introduced the tga20 allele into PrPΔCD+/+ mice and subsequently removed the Prnp wild‐type alleles by crossing to Prnpo/o mice. The resulting PrPΔCDtga20o/o mice stayed healthy for >300 days (Figure 3B). Therefore, the severity of the disease is attenuated by overexpression of PrPC (Figure 3G). As these phenomena were confirmed in four independently generated lines of transgenic mice, we conclude that both the PrPC‐dependent attenuation and the Prnpo‐dependent exacerbation of the phenotype are independent of the site of transgene integration.

Tg902 PrPΔpHC+/+ mice developed normally and transmitted the transgene in an undistorted Mendelian manner (data not shown), although PrPΔpHC expression was similar to that of PrPΔCD in Tg1046 mice. Tg902 mice were then crossed to Prnpo/o mice. Over 150 Tg902 PrPΔpHCo/o mice were monitored, yet did not develop disease during an observation period of >450 days. Therefore, expression of PrPΔpHC is not intrinsically toxic in the absence of PrPC.

Neuropathological changes

Histological examination of the brains from terminally sick Tg1047 PrPΔCD+/o mice revealed vacuolar degeneration and astrogliosis of several white matter tracts, most pronouncedly in the internal capsule and the pyramidal tracts (Figure 4B), the white matter areas of the brainstem, and, to a lesser degree, in the cerebellum. No degeneration, however, was present in the cerebellar cortex including the cerebellar granule cell layer (Figure 4D). There were no changes in the cerebral cortex or in the basal ganglia. No changes were found in age‐matched non‐transgenic littermates (Figure 4A and C).

Figure 4.

White matter pathology and peripheral neuropathy in terminally sick PrPΔCD mice. Age‐matched non‐transgenic Prnp+/+ littermates served as controls for all histological investigations. GFAP‐stained pyramidal tract showed vacuolar degeneration and associated astrogliosis in a sick PrPΔCD mouse (B) but not in a control mouse (A). GFAP‐stained cerebellum in PrPΔCD mice revealed moderate astrogliosis but preserved cerebellar granule cell layer (D). The same region in an age‐matched control littermate (C). Luxol‐Nissl‐stained cross‐section through the mid‐thoracic spinal cord showed coarse vacuolar degeneration in descending and ascending fiber tracts (E, G). No such changes were observed in control mice (F). Paraphenylene diamine‐stained semithin section of a descending myelinated fiber tract in a PrPΔCD mouse revealed axonal loss together with some enlarged ‘vacuolated’ myelinated fibers with thinning of their myelin sheaths (asterisk) (I). Age‐matched control littermate (H). Luxol‐Nissl‐stained longitudinal section through the sciatic nerve of a control (J) and a PrPΔCD mouse (K). Axonal breakdown and segmentation of myelin into digestion chambers (arrows). Axonal loss, hypomyelinated fibers, and degeneration of myelinated fibers with collapsed myelin sheaths (arrow) were observed in paraphenylene diamine‐stained semithin sections of the sciatic nerve in PrPΔCD mice (M) but not in control mice (L). Electron microscopic analysis of the sciatic nerve showed demyelinated fibers (N) associated with macrophages filled with myelin debris (O). Some fibers showed collapsed axons surrounded by abnormal condensed multilamellar myelin sheaths (P). Normal myelinated fibers in a control mouse (Q). Scale bar: (A–E) 100 μm; (F, G, J, K) 20 μm; (H, I, L, M) 10 μm; (N–Q) 2 μm.

Coarse vacuolar degeneration was found in both descending and ascending fiber tracts of the spinal cord (Figure 4E and G). In contrast to non‐transgenic Prnp+/o littermates (Figure 4F and H), myelinated fiber tracts in the spinal cord displayed axonal loss, which was associated with enlarged degenerating fibers with thinning of their myelin sheaths (Figure 4I). There was no onion bulb formation. Evidence for degeneration of myelinated fibers, including axonal breakdown and segmentation of myelin into ellipsoids (‘digestion chambers’), was obtained in the sciatic nerve (Figure 4K). Analysis of semithin sections revealed axonal loss, enlarged ‘vacuolated’ fibers, as well as degenerating myelinated fibers (Figure 4M). No such changes were found in sciatic nerves from control littermates (Figure 4J and L). Electron microscopy revealed axonal breakdown and hypomyelinated fibers, as well as demyelinated fibers and myelin‐digesting macrophages (Figure 4N–P). These findings are indicative of widespread axomyelinic degeneration with minor neuronal degeneration. Because axons build functional units with their myelin sheaths, and both compartments underwent degeneration, it is difficult to establish whether primary damage was to axons or to myelin sheaths. We proposed earlier that myelin toxicity is the primary cause of damage, as the white matter disease in Shmerling mice was rescued by oligodendrocyte‐restricted (but not neuron‐restricted) expression of PrPC (Radovanovic et al, 2005).

Direct comparisons showed similar myelin damage in PrPΔCD and PrPΔF mice. However, PrPΔCD mice lacked the extensive cerebellar granule cell degeneration found in PrPΔF mice (Shmerling et al, 1998; Radovanovic et al, 2005). Myelin damage in PrnpNsk/Nsk and Prnpo/o mice was associated with concentric ‘onion bulb’ figures indicative of multiple episodes of demyelination and remyelination, which was not a feature in PrPΔCD mice. The latter discrepancy may be related to the higher toxicity of PrPΔCD, which may preclude reparative remyelination.

Genetic interactions between PrPΔF and PrPΔCD

The toxicity of both PrPΔF and PrPΔCD is counteracted by overexpression of PrPC and enhanced by its absence, suggesting that both toxic mutants act through the same molecular mechanism. If so, expression of PrPΔF should not rescue the toxicity of PrPΔCD, and/or vice versa. We tested this prediction by crossing Tg1046 PrPΔCD and TgF35 PrPΔF mice (n=33 mice). All PrPΔCDΔFo/o mice developed a phenotype identical to that of parental PrPΔCD mice. Coexpression of PrPΔF and PrPΔCD in Prnp+/o mice (PrPΔCDΔF+/o) resulted in fully penetrant neurological disease starting at 70 days of age (n=9), and by 300 days 50% had reached a clinical score of ⩾3.5 (Figure 5A). When PrPΔCD and PrPΔF were coexpressed in the presence of diploid Prnp (PrPΔCDΔF+/+), only two of six aged mice developed disease (Figure 5A). We conclude that neither PrPΔF nor PrPΔCD antagonizes the toxicity of each other. This reinforces the suggestion that PrPΔF and PrPΔCD compete with PrPC by the same molecular mechanism. Interestingly, PrPΔCD fully imposed its phenotype onto PrPΔCDΔF mice, although PrPΔF was much more abundant than PrPΔCD (Figure 5E).

Figure 5.

Survival of compound transgenic mice. (A–D) Survival curves of mice expressing PrPΔCD (Tg1046), PrPΔF (TgF35), or both, in the absence or presence of full‐length PrPC. Each line comprises ⩾5 individuals. (B) Mice expressing PrPΔCD (Tg1047) or PrPΔC (TgC4), or both, in the presence of one wild‐type Prnp allele. Each line comprises ⩾7 individuals. (C) Mice expressing PrPΔpHC (Tg902), PrPΔF (TgF35), or both, in the absence or presence of PrPC. Each line represents ⩾3 individual mice. (D) Mice expressing PrPΔCD (Tg1047) or PrPΔpHC (Tg902), or both, in the presence of one wild‐type Prnp allele. Each line represents ⩾7 individuals. (E–G) Brain expression of PrPC and transgenic PrP. (E) PrPΔF has a higher steady‐state level than PrPwt and PrPΔCD. Brain homogenates were optionally treated with PNGase F and Western blots were decorated with the antibody POM1. The arrows refer to the molecular weight of each protein after deglycosylation. (F) PrPΔC expression is stronger than that of PrPwt and PrPΔCD. Brain homogenates were treated with PNGase F and Western blots were decorated with antibodies POM1, POM3, and POM11 recognizing different domains of PrP. (G) PrPΔpHC expression is lower than that of PrPC and PrPΔF, but similar to that of some affected PrPΔCD lines. All homogenates were treated with PNGase F.

Tg1047 or Tg1050 PrPΔCD+/+ mice were then mated to PrPΔC mice, which express a PrP mutant devoid of all ORs (line TgC4; Shmerling et al, 1998). The longevity of PrPΔCDΔC+/o offspring (n=37) was 290 and 120 days of age, respectively, which is three‐fold more than PrPΔCD+/o mice from the same lines (n=22; Figure 5B). As PrPΔC does not contain any OR, we conclude that the rescuing ability of PrPC does not require this structural motif, as was shown in coexpression experiments of PrPΔC and PrPΔF (Flechsig et al, 2000; Flechsig and Weissmann, 2004). PrPΔC expression was higher than that of PrPC and PrPΔCD (Figure 5F), suggesting that the rescuing potency of PrPC was lower than that of PrPC.

PrPΔpHC did not elicit any pathological phenotype in Prnpo/o mice. However, Western blots (Figure 5G) revealed that PrPΔpHC was ca. 20% endogenous PrPC. We therefore reasoned that any potential effect of PrPΔpHC may become apparent in intercrosses to PrPΔF mice. Whereas TgF35 PrPΔFo/o mice developed progressive neurological signs and terminal disease at 106±23 days (Figure 5C), we found that double transgenic PrPΔFΔpHCo/o mice survived to 157±27 days. Therefore, PrPΔpHC expression extends the lifespan of PrPΔF mice (n=4; Mann–Whitney test: P=0.064).

We then tested whether PrPΔpHC rescues the toxicity of PrPΔCD by crossing Tg1047 or Tg1050 PrPΔCD+/+ to Tg902 PrPΔpHCo/o mice. Unexpectedly, we observed significantly reduced latency of first clinical signs and reduced longevity in PrPΔCDΔpHC+/o mice (Figure 5D). Tg1047 PrPΔCDΔpHC+/o mice (n=19) developed a clinical phenotype 15–20 days earlier and had to be euthanized 35 days earlier than their PrPΔCD+/o littermates (n=24). Similarly, crossing of Tg902 PrPΔpHCo/o and Tg1050 PrPΔCD+/+ mice (n=10) resulted in a reduced life expectancy (PrPΔCDΔpHC+/o 30 days versus PrPΔCD+/o 40 days, data not shown). Therefore, although non‐toxic to Prnpo/o and wild‐type mice, PrPΔpHC mitigates the toxic effect of PrPΔF and enhances the toxicity of PrPΔCD.


We found that prion proteins lacking solely a highly conserved sequence of 40 amino acids (the ‘central domain’, CD) are potent neurotoxins. In the absence of full‐length PrPC, even small amounts of PrPΔCD induce early postnatal death. Toxicity was strictly dose‐dependent, with doubling of expression resulting in a decrease of longevity by 10 days (Figures 2D and 3A–D).

The clinical signs of PrPΔCD mice were associated with widespread central and peripheral neuropathy similar to mice overexpressing Dpl or PrPΔF (Shmerling et al, 1998; Rossi et al, 2001). Myelin damage, rather than neuronal loss, was the primary cause of lethality in these mice (Radovanovic et al, 2005). A similar, but milder, myelin phenotype was described in aged Zrch‐I Prnpo/o mice (Nishida et al, 1999). Therefore, rather than inducing an entirely artificial disease, PrPΔCD, PrPΔE, PrPΔF, and Dpl appear to greatly aggravate a pathology caused by a defect in PrPC function. This intriguing observation suggests that maintenance of myelin integrity represents a major physiological role of PrPC.

What may be the mechanism of PrPΔCD neurotoxicity? Any interstitial deletions may destabilize proteins, resulting in cytotoxicity. However, as most of the CD domain is intrinsically unstructured, its deletion is unlikely to dramatically destabilize PrP. Secondly, toxicity may result from subcellular mislocalization (Kaneko et al, 1997; Ma et al, 2002) or inappropriate membrane topology of PrP (Hegde et al, 1998). However, these scenarios are unlikely as (1) PrPΔCD undergoes glycosylation, processing, and transport to detergent‐resistant membrane microdomains, similar to full‐length PrP, and (2) full‐length PrPC does not antagonize the toxicity of cytosolic PrP (Ma et al, 2002).

We tested whether the ΔCD, ΔE, and ΔF deletions increase the aggregation propensity of PrPC. This is unlikely, as PrPΔCD was fully sensitive to proteinase K digestion and did not form immunohistochemically detectable aggregates in transgenic mice (data not shown). Finally, formation of the carboxy‐proximal C1 fragment of PrP was significantly impaired in PrPΔCD mice, suggesting abnormal endoproteolysis of PrPΔCD. Although the molecular basis of this phenomenon is unclear, it is unlikely to account for toxicity, as healthy PrPΔpHC mice also displayed a similarly reduced C1 fragment.

The pathogenetic mechanisms of many human hereditary diseases, as well as many bacterial toxins, include inappropriate signaling events. As PrPC may transduce signals (Mouillet‐Richard et al, 2000; Mattei et al, 2004; Solforosi et al, 2004), the neurotoxicity of PrP mutants may rely on similar mechanisms. The pathological phenotype of PrPΔCD, PrPΔE, and PrPΔF mice was counteracted by overexpression of full‐length PrPC (Shmerling et al, 1998) and exacerbated by removal of the endogenous Prnp gene, suggesting that PrPΔCD, PrPΔE, and PrPΔF compete with PrPC for a common signaling molecule. This suggestion is strengthened by crossing experiments showing that neither mutant protein ameliorates the phenotype induced by the other. The phenotype was solely determined by the stoichiometric ratio of full‐length versus redacted PrP, regardless of whether PrPC was expressed from its endogenous Prnp locus or as a transgene.

All toxic mutants displayed disruption of the charge cluster (CC, residues 95–110) and a part of the hydrophobic core (HC, residues 111–121) of PrPC. Toxicity was ameliorated by coexpressing PrP variants with intact CC and HC, even if these variants—as in the case of PrPΔC—lacked the OR. We therefore posit that PrPC exerts a myelin‐maintaining activity by signaling through the CD domain to an unknown receptor (tentatively termed PrPR). In all paradigms investigated, the phenotype was determined by the stoichiometry of mutant to full‐length PrP, strongly suggesting that PrPC, the various mutant PrP moieties, and PrPR form hetero‐oligomeric complexes (Figure 6). The mild pathological phenotype of Prnpo/o mice suggests that myelin integrity is supported by residual PrPR activity, whereas disruption of the CD domain sequesters PrPR in a dominant‐negative state. Complex stability could be influenced by domains distinct from those involved in executing signal transduction: the context‐dependent toxicity of PrPΔpHC implicates the OR as one such domain (Figure 6). Although deletion of 40 amino acids produced a powerfully toxic molecule, ablation of eight amino acids within this domain (PrPΔpHC) was innocuous to both wild‐type and Prnpo/o mice. Crossing experiments show that PrPΔpHC is not functionally equivalent to PrPC. The toxicity of PrPΔF was diminished, yet that of PrPΔCD was augmented by coexpression of PrPΔpHC. In the frame of the signaling model, the deletion in PrPΔpHC may affect the interaction between PrPC and PrPR.

Figure 6.

A model for the positive and negative effects of PrPC and its variants. (A) PrP (black) consists of a globular C‐terminal domain (hexagon) and an N‐terminal flexible tail (arch) encompassing the ORs (circle). The model rests on the following assumptions: (1) PrPC activates a hitherto unidentified receptor (PrPR), which transmits myelin maintenance signals (flashes); (2) in the absence of PrPC, PrPR exerts some residual activity, either constitutively or by recruiting a surrogate ligand; (3) the activity of PrPC and its mutants requires homo‐ or heterodimerization, and induces dimerization of PrPR; and (4) PrP dimers containing PrPΔCD or PrPΔCD trap PrPR in an inactive dominant‐negative state. Finally, (5) the OR region stabilizes the interaction between PrP and PrPR, but does not contribute directly to signaling. (B) By increasing complex affinity to PrPR through its ORs, PrPΔpHC may displace PrPΔF homodimers, thereby ameliorating the phenotype of compound PrPΔFΔpHC transgenic mice. However, as PrPΔCD also contains ORs, it may more efficiently compete with PrPΔpHC for PrPR and thus prevents any phenotypic improvement.

Verification of the model presented above requires the physical identification of PrPR. Towards that goal, it will be crucial to enumerate the cellular constituents—which may not necessarily all consist of protein—binding differentially to PrPC and to the various mutants described here. Functional assays of such constituents may finally shed light on the physiological function of the prion protein, and may eventually help understanding neurodegeneration in transmissible spongiform encephalopathies.

Materials and methods

Construction of transgenes

The coding sequence of the murine Prnp gene was analyzed using DNAMAN software (Lynnon BioSoft, Canada) and a hydrophobicity plot was generated using a window of nine amino‐acid residues. The regions identified in this plot were used to define the CC, CD, and HC domains. Based on pPrPHG (Fischer et al, 1996), a PmeI/NheI subclone was generated in the pMECA backbone. To create the PrnpΔCD cDNA, this plasmid was used as template to obtain two PCR fragments with primer sets RAMP fw (5′‐CTA TCA GTC ATC ATG GCG AAC C)/RAMP 93 ex‐rc (5′‐CCA AAA TGG ATC ATG GGC CTA CCC CCT CCT TGG CCC CAT C) and RAMP 93–134 fw (5′‐GAT GGG GCC AAG GAG GGG GTA GGC CCA TGA TCC ATT TTG G)/RAMP rc (5′‐ CAT CAT CTT CAC ATC GGT CTC G). The resulting PCR products were purified, mixed in stoichiometric amounts, and annealed by slowly decreasing the temperature from 95 to 40°C. The flanking primers RAMP fw and RAMP rc were then added and the fusion product was amplified. This product was purified, digested with XmaI/BstEII, and cloned into the respective sites of the pMECA PrP subcloning vector. The subcloned construct was then cloned into the PmeI/NheI sites of the pPrPHG plasmid. Plasmid MoPrP‐ΔH1 coding for PrPΔpHC was constructed by inserting the cDNA encoding moPrPΔ114–121, recovered from plasmid pUC‐PrPΔ114–121 (Holscher et al, 1998) by restriction with AatII and SacI, followed by blunting into plasmid MoPrP.XhoI, a derivative of pPrPHG (Borchelt et al, 1996; Fischer et al, 1996) upon restriction with XhoI and blunting.

Generation, identification, and maintenance of transgenic mice

For construction of PrPΔCD and PrPpHC, the pPrPHG plasmids containing the PrP coding sequences were propagated in Escherichia coli XL1 blue, the minigene excised with NotI and SalI, processed as described (Fischer et al, 1996), and injected into fertilized Prnp+/+ oocytes (C57BL/6 × DBA/2 cross and C3H × C57BL6) by standard procedures (Rulicke, 2004). Transgenes on a Prnpo/o background were identified by PCR using the exon 2 primer pE2+ (5‐CAA CCG AGC TGA AGC ATT CTG CCT) and the exon 3 primer Mut217 (5‐CCT GGG ACT CCT TCT GGT ACC GGG TGA CGC). In order to outbreed the Prnp+ allele, PCR analysis was carried out using primers P10 (Prnp exon 3, 5′‐GTA CCC ATA ATC AGT GGA ACA AGC CCA GC), 3′NC (non‐coding region at 3′ of exon 3, 5′‐CCC TCC CCC AGC CTA GAC CAC GA), and P3 (neoR gene, 5′‐ATT CGC AGC GCA TCG CCT TCT ATC GCC); P10 and 3′NC gave a 560‐bp signal for the Prnp+ allele and P3 and 3′NC gave a 362‐bp product for the Prnp0 allele. Alternatively, to test for the presence or absence of the Prnp+ allele, an additional PCR was performed using primers P2 (Prnp int 2, 5′‐ATA CTG GGC ACT GAT ACC TTG TTC CTC AT) and P10rev (reverse complementary of P10 5′‐GCT GGG CTT GTT CCA CTG ATT ATG GGT AC) giving a product of 352 bp for the Prnp+ allele.

Southern blot analysis

Mouse tail biopsies were kept on ice and digested in lysis buffer (100 mM Tris–HCl (pH 8.5), 5 mM EDTA, 0.2% SDS, 20 mM NaCl, and 100 μg/ml Proteinase K) overnight at 55°C under constant agitation. Purified genomic DNA (20 μg) was digested overnight (EcoRI, 100 U and separated on 0.7% agarose gels (GIBCO BRL ultra Pure agarose). Completeness of the restriction digest and DNA separation was checked by ethidium bromide staining. Genomic DNA was transferred by vacuum blotting at 50–55 mbar with a Vacu Gene Pump and a Gene XL blotting apparatus (Pharmacia) using solutions according to the manufacturer's manual (Pharmacia). The blot was marked, washed for 5 min in 2 × SSC, UV‐crosslinked and prehybridized according to Church and Gilbert (1984). After hybridization (overnight at 65°C) with a radioactively labeled probe (PrP open reading frame), the blot was washed and membranes were subjected to autoradiography using a PhosphoImager (Molecular Dynamics). Images were scanned and signals were quantified with Aida 2.41 imaging software.

For probe labeling, template DNA (50 ng) was diluted in sterile water and random primers (nonamers or hexamers) were added (Prime‐IT® II; Random Primer labeling kit cat# 300385; Stratagene). After gentle mixing, incubation was performed at 95°C for 5 min. Brief centrifugation was followed by 3 min incubation on ice. This was followed by addition of 5 × primer buffer (−dCTP), radioactively labeled [α−32]P‐dCTP (corresponding to 25 μCi; Amersham Biosciences), and Klenow 5 U/μl (Prime‐IT® II; Random Primer labeling kit cat# 300385). For primer elongation, the probe was incubated at 37°C for 75 min. For purification, the QIAquick PCR purification kit was used according to the manufacturer's manual (Quiagen).

mRNA analysis

Total brain RNA was isolated in Trizol (Life Technologies), purified, and DNase‐treated according to the manufacturer's manual (Roche). After reverse transcription (Geneamp; Roche), cDNA was used for PCR using primers Taq5′PrPa (5′‐CAA CCG AGC TGA AGC ATT CTG CCT) and Taq3′PrPa (5′‐GAT CTT CTC CCG TCG TAA T). As control for the PCR procedure and possible DNA contamination, DNase‐treated RNA from wild type and transgenic mice that had not been reversely transcribed was used.

Western blot analysis

Brain hemispheres were homogenized in 7 vol PBS, 0.5% Nonidet P‐40, and 0.5% deoxycholate and the solution was centrifuged for 5 min in an Eppendorf centrifuge. For deglycosylation, up to 50 μg of denatured total protein was incubated at 37°C for 4 h with 500 U PNGase F (New England Biolabs) according to the manufacturer's instructions. The protease inhibitors Pefabloc (1 mg/ml), leupeptin (10 μg/ml), pepstatin (10 μg/ml), aprotinin (1 μg/ml) (all from Boehringer, Mannheim), and 0.5 mg/ml EDTA were added. After electrophoresis of protein samples through 12% SDS–polyacrylamide gels, samples were transferred to nitrocellulose membranes (Schleicher & Schuell) and incubated with mouse monoclonal anti‐PrP antibodies POM1, POM3, and POM11 (Polymenidou et al, 2005), followed by incubation with peroxidase‐labeled anti‐mouse antiserum (1:2500; Amersham) and developed with the ECL detection system (Pierce). Antibody incubations were performed in 1% Top Block (Juro) in Tris‐buffered saline‐Tween (TBS‐T) for 1 h at room temperature or overnight at 4°C.

Flotation assays

Flotation of detergent insoluble complexes was performed as described (Naslavsky et al, 1997). Appropriate brain homogenates were extracted for 2 h on ice in cold lysis buffer (150 mM NaCl, 25 mM Tris–HCl, pH 7.5, 5 mM EDTA, 1% Triton X‐100; total protein: 1 mg in 1.6 ml). Extracts were mixed with two volumes (3.2 ml) of 60% Optiprep (Nycomed) to reach a final concentration of 40%. All lysates were loaded at the bottom of Beckman ultracentrifuge tubes. A 5–30% Optiprep® step gradient in TNE (150 mM NaCl, 25 mM Tris–HCl, pH 7.5, 5 mM EDTA) was then overlaid onto the lysate (8.4 ml of 30% Optiprep® and 3.6 ml of 5% Optiprep®). Tubes were centrifuged for 24 h at 4°C in a TLS55 Beckman rotor at 100 000 g. Fractions (1 ml) were collected from the top of the tube and processed for immunoblotting and visualization with anti‐PrP antibody POM1 (Polymenidou et al, 2005), anti‐flotillin 1, and anti‐GAPDH antibody (both from BD Transduction Laboratories).

Clinical scoring and observation

Mice were examined once weekly for clinical signs, including weight loss, kyphosis, rough hair coat, paresis, ataxia, and decreased motor activity. In addition, mice were placed on a metal grid and monitored for 5 min for their walking proficiency. The incidence of slipping between bars was assessed and integrated with all parameters mentioned above into a clinical score (Table I). Mice were euthanized when they reached a score of 3.5 or higher. Statistical significance was assessed using Mann–Whitney test.

Morphological analysis

Brains, spinal cords, and sciatic nerves were removed and fixed in 4% formaldehyde in PBS, pH 7.5, paraffin embedded, and cut into 2–4 μm sections. Sections were stained with hematoxylin and eosin (H&E), Luxol‐Nissl (myelin and neurons), and commercial antibodies to GFAP (glial fibrillary acidic protein; activated astrocytes), MBP (myelin basic protein), NF200 (neurofilament 200), IBA1 (microglia), SAF84 (PrPSc aggregates), and POM1 (soluble cellular PrP). For semi‐thin sections and electron microscopy, mice were perfused with ice‐cold 4% PFA/2% glutaraldehyde. Spinal cord and sciatic nerve tissues were removed, immersed in the same solutions, and kept at 4°C until processing. Tissues were embedded in Epon, and semi‐thin sections were stained with toluidine blue and paraphenylene diamine or used for electron microscopy.

Supplementary Information

Supplementary movie [emboj7601510-sup-0001.mpg]


We thank DR Borchelt for plasmid MoPrP.XhoI and Petra Schwarz, Marianne König, Andrea Schifferli, Li‐Chun Infanger, and Katharina Hüttner for technical assistance. FB was a postdoctoral fellow of the Deutsche Forschungsgemeinschaft (BA2257/1‐1). This work was supported by the DFG (TransRegio‐Sonderforschungsbereich 11 Konstanz‐Zürich) and the EU Commission (QLRT‐2000‐01924 to AB, TSEUR to AA, and APOPIS to AA and FB).