The Fate of PrP GPI-Anchor Signal Peptide is Modulated by P238S Pathogenic Mutation

Authors


Abstract

Glycosylphosphatidylinositol (GPI)-anchored proteins are localized to the plasma membrane via a C-terminally linked GPI anchor. The GPI anchor is added concomitantly to the cleavage of the carboxy-terminal GPI-anchor signal sequence, thereby causing the release of a C-terminal hydrophobic peptide, whose fate has not yet been investigated. Here we followed the fate of the GPI-attachment signal of the prion protein (PrP), a protein implicated in various types of transmissible neurodegenerative spongiform encephalopathies (TSE). The PrP GPI-anchor signal sequence shows a remarkable and unusual degree of conservation across the species and contains two point mutations (M232R/T and P238S) that are responsible for genetic forms of prion disorders. We show that the PrP GPI-anchor signal peptide (SP), but not the one from an unrelated GPI-anchored protein (folate receptor), undergoes degradation via the proteasome. Moreover, the P238S point mutation partially protects the PrP GPI-anchor SP from degradation. Our data provide the first attempt to address the fate of a GPI-anchor SP and identify a role for the P238S mutation, suggesting the possibility that the PrP GPI-anchor SP could play a role in neurodegenerative prion diseases.

image

There are over 100 GPI-anchored proteins in humans, including enzymes, surface receptors and adhesion molecules [1, 2]. Proteins destined to be GPI-anchored are translated with cleavable N- and C-terminal signal sequences. The N-terminal signal sequence is cleaved after directing the nascent polypeptide to the endoplasmic reticulum (ER) [3]. Within the lumen of the ER the pre-assembled GPI anchor is transferred to the acceptor amino acid (termed ω site), concomitant with the cleavage of the C-terminal GPI-anchor signal sequence (GPI-SS) by the action of a GPI transamidase [4]. As a result, a peptide (GPI-anchor signal peptide or GPI-SP) is released after each GPI-anchor attachment.

GPI-SS lack a clear consensus, although some characteristic features are conserved [5]. These include (i) the residue to which the anchor is attached (the ω site), which always has a small side chain, (ii) a hydrophilic spacer region (5–10 amino acids) of moderately polar amino acids and (iii) a C-terminal hydrophobic sequence of variable length (15–20 amino acids). GPI-SS have a remarkable degree of flexibility [6-8], to the point that a GPI-SS can be completely recreated by the synthetic polymer Ser[3]-Thr[8]-Leu[14], placed at the C-terminus of a protein [9]. The ability to tolerate a high level of diversity could make GPI-SS a suitable target for evolution. For example, mutations that do not disrupt the GPI-anchoring function but that are useful for some other purpose could arise and be maintained by selection over time. The idea that GPI-SPs, once released from the protein ectodomain, could carry on an independent function is quite intriguing, also considering that small peptides have been shown to perform a variety of actions, including working as hormones, neurotransmitters and antimicrobial peptides [10, 11].

Prion diseases are fatal neurodegenerative diseases of infectious, sporadic or inherited origin, affecting humans and other animals. The disease is caused by the conversion of the normal cellular prion protein (PrPC) into the infectious abnormal prion protein (PrPSc) [12]. The prion protein is a glycoprotein that localizes to the plasma membrane via a C-terminally linked glycosylphosphatidylinositol (GPI) anchor [13]. Approximately 15% of human prion diseases are associated with autosomal dominant pathogenic mutations in the prion protein (PrP) [14-17]. In most cases the mutation is thought to increase the tendency of PrPC to convert into the pathogenic form PrPSc. Over 30 autosomal dominant pathogenic PrP mutations have been described [18]. While the large majority map to the region coding for the PrP ectodomain, two amino acid substitutions, the M232R/T and P238S [19, 20] localize within the GPI-anchor signal sequence (GPI-SS). These mutations are particularly perplexing since the GPI-SS is replaced by a GPI anchor and is absent in the mature PrP.

With the exception of the variant surface glycoproteins (VSG) of Trypanosoma brucei [21], GPI-SSs from homologous genes have been shown to have a low degree of conservation across species. Interestingly, when we compared the 24 amino acids of PrP GPI-SS from different species, we observed a remarkable degree of conservation (see Scheme 1A). Because of the unusual level of conservation of its GPI-SS and because of the presence of unexplained autosomal dominant pathogenic mutations in this sequence, PrP was chosen as the model protein to address the fate of the GPI-SP after its release following GPI-anchor attachment.

Scheme 1.

A) Degree of conservation between the GPI-SS of PrP and FR across species; identical amino acids are marked with a red asterisk. B) General design of the fusion constructs: GFP (green box) is linked to the N-terminal signal sequence (ss) (gray box) and to the GPI-SS at the C-terminus (red box), which, in turn, is fused to the MH tag (blue box) via a flexible Gly/Ser linker (black line). This construct is processed within the ER to produce GFP linked to the GPI-anchor and the GPI-SP linked to MH tag. C) Particulars of the construct GFP-PrP-MH, containing PrP GPI-SS (top) and GFP-FR-MH, containing FR GPI-SS (bottom); the hydrophilic segment is in blue letters, while the hydrophobic region is in red.

Results and discussion

Characterization of GFP-GPI-SS-Tag constructs

Scheme 1 shows the comparison between the GPI-SS of PrP and the one of an unrelated GPI-anchored protein, the FR [22]. As shown in this figure, FR was chosen because its GPI-SS has similar characteristics to that of PrP. Not only do they have the same length (24 amino acids), but even an identical size of both the hydrophilic (9 amino acids) and the hydrophobic regions (14 amino acids); both also contain a Ser residue as the ω-site. While in PrP GPI-SS 18 of 24 amino acids are identical across different species, the GPI-SS of FR only contains two identical amino acids. This is also the case for other GPI-anchored proteins analyzed, like placental alkaline phosphatase (PLAP) [23], which has 4 of 29 identical amino acids.

To be able to follow the GPI-SP after its release from the protein ectodomain, we assembled DNA constructs containing the PrP N-terminal signal sequence (for the insertion in the ER) attached to the green fluorescent protein (GFP), whose C-terminus was fused to the GPI-SS of either PrP or FR. This, in turn, was linked to the double Myc-His tag (MH) (see Scheme 1B). A Gly-Ser flexible linker (Gly-Gly-Gly-Ser-Gly-Gly-Ser) was inserted as a spacer between the GPI-SS and the MH tag, to obtain the constructs GFP-PrP-MH and GFP-FR-MH (Scheme 1C). The Gly-Ser linker was inserted to prevent the MH tag from interfering with the GPI-SS. The Gly-Ser linker has been shown to adopt a random-coil structure when fused between two protein domains, allowing the protein domains to rotate freely around the end of the linker [24]. The position of the double tag, placed at the C-terminus of the GPI-SS, was not expected to block the process of GPI-anchor attachment, as internal GPI-SS have been shown to be able, though less efficiently, to work as GPI-attachment signals [25]. Upon entering the ER, the constructs are expected to undergo GPI-anchor attachment to produce the fully processed GFP-GPI and the released GPI-SP-MH (Scheme 1B). Scheme 1C shows the final design of the constructs containing the human GPI-SS of PrP and the human GPI-SS of FR.

The constructs were transfected in HeLa cells and the localization of both the ectodomain (GFP) and the GPI-SP (via the MH tag) was analyzed by immunofluorescence (IF). HeLa cells have been extensively used to study the process of GPI-anchor attachment [26], from the steps in GPI-anchor assembly [27, 28] to its transfer to the target protein by the GPI transamidase [29, 30], and to study their intracellular trafficking [31]. HeLa cells were also used to compare the efficiency of C-terminal signals of native GPI-anchored proteins in conferring GPI-anchoring [32]. Finally, different aspects of PrP biology, such as the biosynthesis, maturation, processing, trafficking and metabolism [33-39], as well as the metabolism and processing of disease-causing mutant PrPs [40], have been investigated in HeLa cells.

Cells expressing GFP-PrP-MH and GFP-FR-MH constructs were analyzed by IF (Figure 1A-I) and western blot (WB) (Figure 1j,K) to visualize processed and unprocessed proteins as well as the respective GPI-SPs. Qualitative data by IF analysis confirmed, as expected [41], that the GFP ectodomains of both constructs were able to reach the perinuclear Golgi area and the plasma membrane (PM) (GFP-PrP-MH is shown in Figure 1A–C and GFP-FR-MH is shown in Figure 1D–F), together with some level of ER accumulation, likely due to the presence of unprocessed proteins in the ER. The released GPI-SP as well as the unprocessed protein (indistinguishable by IF) localized in a perinuclear reticular domain which proved to be the endoplasmic reticulum (ER) upon colocalization with the ER marker Calnexin (Figure 1G–I).

Figure 1.

Characterization of GFP-GPI-SS-MH constructs. IF of HeLa cells expressing GFP-PrP-MH (A–C) and GFP-FR-MH (D–F). Processed proteins, visualized via anti-GFP antibody, localize mostly in the perinuclear Golgi and at the plasma membrane (B, E), while unprocessed proteins (visualized via anti-His antibody) localizes in the ER (C, F). Confirmation of ER localization is shown in (G–I), where unprocessed protein (I) co-localizes with the ER marker Calnexin (H). Non-transfected cells (negative control), visible via DAPI nuclear staining, are marked by an asterisk (*). Scale bar is 10 µm. J) WB analysis of unprocessed and processed proteins (left side, visualized by anti-GFP antibody) and unprocessed and GPI-SP (right side, visualized by anti-His antibody). K) Quantification of the ratio between unprocessed (blue), processed (green) and GPI-SP (red). The average data were calculated based on six independent experiments and represented as mean ± SD. Significant at *p < 0.05; **p < 0.01.

WB analysis was then performed to establish the ratio between (i) unprocessed protein, containing GFP still linked to the GPI-SS and the MH tag, (ii) fully processed protein (i.e. having undergone addition of the GPI-anchor); and (iii) the released GPI-SP, also linked to the MH tag. As an additional control for processing, we used untagged GFP-PrP and GFP-FR (indicated as PrP-wt and FR-wt in Figure 1J), which have been previously used in our lab and are known to be GPI-anchored [41]. As shown in Figure 1J, left panel, processed proteins (GPI-linked) from both untagged and tagged (MH) constructs run at the same level. Moreover, no unprocessed bands can be seen in the untagged constructs, suggesting that the processing is more efficient when no additional tag is present. This confirms the evidence [25] that the presence of a C-terminal extension to the GPI-SS (in our case the MH tag), although not blocking per se the GPI-attachment process, does induce a decrease in the rate of processing. The ratio between unprocessed and processed proteins (visualized by the anti-GFP antibody) can be used to gain information about the efficiency of each GPI-SS to be cleaved and replaced by a GPI-anchor (Figure 1J, left). For the construct GFP-FR-MH, the processed protein is almost 1.5 times the level of the unprocessed, while in the case of GFP-PrP-MH, the processed protein represents 75% of the amount of the unprocessed one. This suggests that FR GPI-SS is somehow more efficiently cleaved and replaced by the GPI-anchor (Figure 1K). Differences in the efficiency of GPI-anchor addition for various signal sequences had already been reported, showing an 8-fold range of efficiencies for the six different native SPs studied [42]. Altogether this confirms that different GPI-SPs are differently processed.

We then compared the ratio between unprocessed protein and released GPI-SP (Figure 1J, right). Since both proteins contain MH tag, they were detected in the total cell extract of transfected cells with an anti-His antibody. For the GFP-FR-MH construct, the amount of GPI-SP was about 1.5 times the amount of the unprocessed protein. Combining this data with the one measuring the ratio between unprocessed and processed (GFP-GPI), we conclude that, in the case of GFP-FR-MH, there are comparable amounts of processed protein and GPI-SP. This is indeed expected, as processed GFP-GPI and GPI-SP are both the products of the same GPI-anchoring reaction. When we looked at GFP-PrP-MH, we were surprised to be unable to detect the GPI-SP, indicating that PrP GPI-SP, once released from the protein ectodomain, was no longer present within the cell.

To exclude that these findings could be specific to HeLa cells, GFP-PrP-MH and GFP-FR-MH were also expressed in neuronal CAD cells. As shown in Figure S1, Supporting Information, the localization and the relative processing of both constructs was comparable to that observed in HeLa cells. We therefore concluded that the differential processing of PrP GPI-SP is not cell type specific.

Because the tag is removed from the protein during the GPI-attachment processing, we would expect the tag to affect the rate of processing (as is indeed the case), but not the behavior of the processed proteins, which no longer contain the tag. However, to rule out the possibility that the presence of the tag could impair the trafficking or the stability of the processed proteins, we proceeded to compare GFP-PrP-MH and GFP-FR-MH with their untagged counterparts (GFP-PrP-wt and GFP-FR-wt).

To assess the extent of impairment in trafficking imposed by the tag, we expressed tagged and untagged constructs and visualized their localization by IF. As shown in Figure 2, tagged proteins (A,B and D,E) localized at the perinuclear Golgi and at the plasma membrane, similarly to the untagged constructs (C,F). This indicates that, as expected [41], once the GPI-SP (and the linked tag) is removed, the GFP-GPI processed proteins behave exactly as the untagged counterparts. However, while untagged proteins did not show ER accumulation, a portion of the tagged constructs was always present in the ER (as indicated by the staining of the perinuclear rim, Figure 2B,E). This ER accumulation is likely the consequence of the decrease in the rate of processing originated by the presence of the tag, as expected based on the analysis of internally positioned GPI-SS [25].

Figure 2.

Comparison between tagged and untagged constructs. Cells expressing GFP-PrP-MH (A and B) and GFP-FR-MH (D and E) are compared by IF with cells expressing GFP-PrP (C) and GFP-FR (F). Processed proteins, visualized via anti-GFP antibody, localize mostly in the perinuclear Golgi and at the plasma membrane; ER localization, visible as a perinuclear rim, is observed for the tagged constructs (B, E). Non-transfected cells (negative control), visible via DAPI nuclear staining, are marked by an asterisk (*). Scale bar is 10 µm. G) WB analysis to measure the stability of GFP-GPI originated from tagged and untagged constructs (visualized by anti-GFP antibody). Tubulin (Tub) is used as loading control; untransfected cells (cntr) are also loaded as control. H) Quantification of the levels of processed proteins originated from untagged and tagged constructs. I) Quantification of the levels of unprocessed and processed proteins originated from the tagged constructs. Error bars represent SD from three independent experiments. Significant at *p < 0.05; **p < 0.01.

Additionally, we measured the stability of GFP-GPI from tagged and untagged constructs. HeLa cells were transfected with GFP-FR-wt, GFP-PrP-wt, GFP-FR-MH and GFP-PrP-MH. Twenty-four hours after transfections, each group of cells was split into four dishes and incubated with cycloheximide (100 µm), an inhibitor of protein synthesis, for 0 h–4 h–8 h–24 h (Figure 2G).

By WB (Figure 2G–I), we observed that while levels of GFP-GPI from untagged constructs steadily decreased after 4 h and 8 h treatment, the levels of GFP-GPI originating from the tagged constructs remained constant for up to 8 h (quantification in Figure 2H). At the same time the levels of unprocessed proteins from the tagged constructs decreased (quantification in Figure 2I). These data show that the decreased rate of processing induced by the tag creates a reservoir of unprocessed protein which accumulates within the lumen of the ER (as shown by IF). The unprocessed proteins are eventually processed, replenishing the pool of processed proteins, as shown by WB in Figure 2G–I.

We concluded that the presence of the Myc-His tag, despite causing a reduction in the efficiency of GPI-processing, does not interfere significantly with the cellular distribution or with the stability of the processed protein. Having validated our tagging approach, we sought to understand the reason behind the disappearance of PrP's GPI-SP.

PrP GPI-SP is degraded by the proteasome

To address the mechanism by which PrP GPI-SP was cleared from the cell, we pharmacologically inhibited different degradation pathways: autophagosomes, lysosomes and the proteasome were blocked while monitoring the levels of processed, unprocessed and GPI-SP by WB.

To inhibit autophagosome degradation, HeLa cells expressing GFP-PrP-MH were treated with 3-MA, a specific inhibitor of autophagy, which has been shown to specifically impair the formation of autophagosomes [43]. To inhibit lysosomal function we used NH4Cl, which affects lysosomal proteases by increasing the intra-lysosomal pH [44, 45] without affecting total protein synthesis [46].

Finally, to inhibit proteasomal degradation, cells expressing GFP-PrP-MH were treated with 1 µm epoxomicin [47], a selective and irreversible proteasome inhibitor, for 6 h. After the pharmacological treatment, the cells were processed for WB. Figure 3 shows that 3-MA and NH4Cl treatment were not able to restore the levels of PrP GPI-SP (Figure 3A). Epoxomicin treatment, however, induced a significant recovery of PrP GPI-SP, indicating that the disappearance of PrP GPI-SP could be ascribed to proteasome clearance. Untreated GFP-PrP-MH and untreated GFP-FR-MH were used as controls; specifically, GFP-PrP-MH was used to test the effect of the different treatments on the processing of PrP; GFP-FR-MH was used to compare the size of the respective GPI-SP, confirming that the after proteasomal inhibition PrP GPI-SP was at the expected MW.

Figure 3.

Pharmacological inhibition of different degradation pathways. A) Cells treated with the indicated compounds were analyzed to reveal the presence of unprocessed and processed proteins (left, anti-GFP antibody) and also unprocessed protein and GPI-SP (right, anti-His antibody). The diagram shows the ratio between unprocessed (blue), processed (green) and GPI-SP (red). The average data were calculated based on three independent experiments and represented as mean ± SD. Significant at **p < 0.01. B–J) Immunofluorescence analysis: cells treated with 3-MA (B–D), NH4Cl (E–G) and epoxomicin (H–J) are processed for IF to reveal the localization of processed protein (C–F–I) and unprocessed/GPI-SP (D–G–J). Non-transfected cells (negative control), visible via DAPI nuclear staining, are marked by an asterisk (*). Scale bar is 10 µm.

We then looked at the cellular distribution of PrP after the same pharmacological treatment. As shown in Figure 3, neither 3-MA (Figure 3B-D) nor NH4Cl (Figure 3E–G) affected the localization of PrP. Under these conditions, GFP-PrP-MH localized at the perinuclear Golgi and plasma membrane (Figure 3C,F) and the unprocessed protein remained within the lumen of the ER (Figure 3D–G). However, under proteasome inhibition, PrP GPI-SP showed an overall diffuse localization (Figure 3J), indicating cytosolic redistribution. The epoxomicin treatment also induced a certain extent of vacuolization, visible in Figure 3H–L, which could derive from ER stress; however, this did not seem to have an effect on the processing of our control (GFP-FR-MH); indeed, under the same epoxomicin treatment (where a similar effect of vacuolization was also noted), the GPI-SP of FR remained restricted to the perinuclear ER (see Figure S2). Other proteasome inhibitors tested, including MG132 and ALLN [48], confirmed the results obtained with epoxomicin; interestingly, when different proteasome inhibitors were compared, their efficiency (measured by the enrichment of poly-ubiquitinated bands by WB), matched their ability to restore the presence of PrP's GPI-SP (see Figure S2).

We also considered the possibility that the degradation of PrP GPI-SP could be ascribed to the action of signal sequence peptidases (which degrade N-terminal leader sequences) or other intramembrane proteases. However, such enzymes specifically hydrolyze transmembrane substrates [49-51], while GPI-APs (and therefore GPI-SPs) have been shown to be fully released within the lumen of the ER [52, 53]. To validate this assumption, we performed a trypsin proteolysis assay after selective permeabilization on cells expressing GFP-PrP-MH. We were able to confirm that GFP-PrP-MH, as the other GPI-APs studied, is fully released within the ER before the cleavage of the GPI-SP and the addition of the GPI-anchor (see Figure S3). The GPI-SP should therefore be an unsuitable target for the activity of signal sequence peptidases or other intramembrane proteases.

The hydrophilic region of PrP GPI-SP is necessary for its clearance

To understand what part of the sequence of PrP GPI-SP rendered the peptide able to be redirected to the cytosol toward proteasomal degradation, we created two hybrid constructs containing the hydrophilic region of one GPI-SP linked to the hydrophobic part of the other, namely GFP-FR/PrP-MH and GFP-PrP/FR-MH. Figure 4A illustrates the design of the constructs.

Figure 4.

Characterization of GFP-FR/PrP-MH and GFP-PrP/FR-MH hybrid constructs. A) Design of the constructs. B–G) IF of HeLa cells expressing GFP-FR/PrP-MH (B–D) and GFP-PrP/FR-MH (E–G). Processed proteins, visualized via anti-GFP antibody, localize mostly at the plasma membrane (C, F), while unprocessed proteins (visualized via anti-His antibody) localize in the ER (D, G). Non-transfected cells (negative control), visible via DAPI nuclear staining, are marked by an asterisk (*). Scale bar is 10 µm. H) Top: WB; unprocessed and processed proteins are visualized with anti-GFP antibody (left), while unprocessed and GPI-SP are visualized with anti-His antibody (right). Bottom: quantification of the ratio between unprocessed (GFP-GPI-SP, blue), processed (GFP-GPI, green) and GPI-SP (red). The average data were calculated based on 3 independent experiments and represented as mean ± SD. Significant at **p < 0.01.

When GFP-FR/PrP-MH and GFP-PrP/FR-MH were expressed in HeLa cells, qualitative IF analysis showed that both ectodomains localized at the Golgi and plasma membrane (Figure 4C and F, respectively), while unprocessed proteins (together with GPI-SP) were retained in the ER (Figure 4D,G). By WB both constructs showed almost an equal ratio between processed and unprocessed proteins, indicating that the hybrid constructs were processed at a comparable rate to PrP-MH (Figure 4H). When we looked at the presence of GPI-SP, we noticed that only the construct containing the hydrophilic region of FR (GFP-FR/PrP-MH) showed the presence of the peptide. FR/PrP GPI-SP was present in an almost equal ratio to the processed protein, as expected if no degradation would take place. The hybrid construct containing the hydrophilic portion of PrP, instead, behaved like GFP-PrP-MH, showing almost no trace of GPI-SP. These results suggest that the hydrophilic region of PrP GPI-SP is necessary for its clearance.

The pathogenic mutation P238S protects PrP GPI-SP from degradation

The fact that the hydrophilic sequence of PrP GPI-SP was sufficient to induce its disappearance indicates that this region must contain the signal responsible for its degradation. It is interesting to note that this small stretch of nine amino acids contains the only two point mutations in human genetic prion diseases known to map to PrP GPI-SS. Because the GPI-SP is replaced by a GPI anchor and is not included in the mature PrP, it is unclear how mutations in this sequence might lead to neurotoxicity. While the mutation M232R was shown to cause PrP to be post-translationally translocated to the plasma membrane in a potentially neurotoxic C-transmembrane orientation [54], the pathogenesis of mutation P238S has not yet been explained.

Although a likely scenario is that it could interfere with the process of GPI-anchor attachment [55], it has been shown that P238S allows cleavage of the GPI-SP and the addition of the GPI anchor [56].

However, based on our results, it is also possible that these mutations could have an effect on the processing of PrP GPI-SP. P238S mutation was particularly interesting because Pro238 is present in the PrP all across the species. A likely possibility based on our results, is that this mutation could have an effect on the processing of PrP GPI-SP. To address this question we introduced in the GFP-PrP-MH construct the point mutations M232R and P238S (indicated as M/R and P/S, respectively in Figure 5). Additionally, we also introduced the double mutation P238S-P239V (indicated as PP/SV in Figure 5). Although there is no evidence of both prolines being mutated in any case of genetic prion diseases, we decided none the less to test the effect of a double-Pro mutant. Proline 239 was replaced with valine, based on the fact that in PrP's sequence P239 is followed by V240. The three constructs were then transfected in HeLa cells and analyzed by IF and WB (Figure 5).

Figure 5.

Characterization of constructs harboring point mutations in the GPI-SP. IF of HeLa cells expressing M232R-GPI-SP (A–C), P238S-GPI-SP (D–F) and PP/SV-GPI-SP (G–I). GFP staining is shown in (B, E, H), while His staining is shown in (C, F, I). Non-transfected cells (negative control), visible via DAPI nuclear staining, are marked by an asterisk (*).Scale bar is 10 µm. K) WB analysis of unprocessed/processed proteins (left side, anti-GFP antibody) and quantification of the ratio between unprocessed (blue) and processed (green) proteins. The average data were calculated based on five independent experiments and represented as mean ± SD. Significant at ***p < 0.001; *p < 0.05; **p < 0.01.

Upon expression, the construct containing M232R mutation localized to the plasma membrane (Figure 5A); however, the IF staining of the processed protein (by anti-GFP Ab; Figure 5B) completely co-localized with the unprocessed protein (by anti-His Ab, Figure 5C).

This suggested that M232R mutation abolished the cleavage of GPI-SP and the attachment of the GPI-anchor. The M/R mutation has been shown to alter the membrane orientation of full length PrP, where a potential transmembrane domain within PrP sequence would be responsible for its altered C-transmembrane orientation. Because in our construct GFP replaces PrP's ectodomain, we do not expect the overall topology to be altered. Since GFP reaches the PM (with an uncleaved GPI-SP), we speculate that the protein may be retained at the PM due to the hydrophobic nature of PrP's GPI-SP. The constructs where either one or both Pro had been mutated were on the other hand fully processed, with processed proteins localizing at the plasma membrane (Figure 5E and H, respectively) and the unprocessed/GPI-SP proteins being restricted to the ER (Figure 5F and I, respectively).

The constructs were then analyzed by WB to compare the efficiency of processing with respect to the non-mutated GRP-PrP-MH construct. As shown in Figure 5J, the M232R mutation was confirmed to impair GPI-anchoring (Figure 5J). P238S and PP/SV mutants appeared to be correctly processed, showing both unprocessed and processed bands (Figure 5J). PrP-MH and the PP/SV mutant showed a similar processing efficiency: unprocessed and processed proteins each represented roughly 50% of the total protein level. However, in the case of the P/S mutant, the processed protein represented almost twice the amount of the unprocessed protein. This suggests that the GPI-SP harboring the P238S mutation is more efficiently cleaved compared to the wild type SP of PrP (Figure 5J).

We then compared the ratio between unprocessed and released GPI-SP (Figure 6); because PrP's SP is rapidly degraded, we considered the GPI-SP of FR as mean of comparison. As expected, the M232R mutant showed no GPI-SP, confirming that this mutation impairs the processing of the protein. Surprisingly, however, in the construct containing P238S mutation, the GPI-SP was clearly visible, and amounted to 30% of the amount of the unprocessed protein. Combining this data with the measure of the ratio between unprocessed and processed proteins, we concluded that 15% of GPI-SP was recovered. The result was even more significant when considering the PP/SV mutant: in this case, the recovery of GPI-SP was close to 100%, indicating that the majority of GPI-SP had been spared from degradation. This offers the first cellular phenotypes for the genetic mutation P238S: first, it renders the GPI-signal sequence of PrP more efficiently cleaved; second, the released GPI-SP accumulates within the cell.

Figure 6.

Recovery of GPI-SP in constructs harboring point mutations in the GPI-SP. Top: WB analysis of unprocessed/processed proteins (left side, anti-GFP antibody), and unprocessed/GPI-SP (right side, anti-His antibody). Bottom: quantification of the ratio between unprocessed (blue), processed (green) and GPI-SP (red). Error bars represent SD from three independent experiments. Significant at *p < 0.05; **p < 0.01.

The accumulation of PrP's GPI-SP in the constructs harboring the P/S and PP/SV mutations could not be ascribed to a more efficient processing (which would produce more GPI-anchored protein and an equal amount of GPI-SP). This is particularly evident in the case of mutation PP/SV, which is not processed more efficiently than PrP-wt (see Figures 5J and 6) and yet it clearly shows the presence of the GPI-SP.

We further analyzed the effect of P/S and PP/SV mutation by comparing the stability of the processed proteins originated from the mutant constructs with the one originating from PrP-wt (Figure 7A). Cells were transfected with GFP-PrP-MH, GFP-PrP(P/S)-MH and GFP-PrP(PP/SV)-MH; after 24 h expression, the cells were divided into 4 groups and treated with cycloheximide for 0 h–4 h–8 h–24 h. The cells were then collected and the amount of unprocessed and processed proteins was measured.

Figure 7.

Analysis of the stability of processed proteins and GPI-SP originating from P/S and PP/SV mutant constructs. A) WB analysis to measure the stability of processed and unprocessed proteins (visualized by anti-GFP antibody) upon cycloheximide treatment. P/S and PP/SV mutant constructs are compared to PrP-MH. B) WB analysis to measure the stability of the GPI-SP (visualized by anti-His antibody) upon cycloheximide treatment. P/S and PP/SV mutant constructs are compared to FR-MH. Error bars represent SD from three independent experiments. Significant at **p < 0.01.

At time 0 h, the respective levels of processed and unprocessed proteins were both fixed at 100%. Figure 7A shows that, for all three constructs, the level of processed proteins remained constant up to 8 h treatment, while the levels of unprocessed proteins steadily decreased, reproducing that described in Figure 2G for GFP-FR-MH. Interestingly, in the case of P/S mutant the amount of unprocessed protein dropped very rapidly after 4 h, confirming that the processing for this construct is more efficient than for the wt GFP-PrP-MH.

We then measured the stability of P/S and PP/SV GPI-SPs by comparing it with the GPI-SP of FR-MH, our control protein (Figure 7B). Cells were transfected with GFP-FR-MH, GFP-PrP(P/S)-MH and PrP(PP/SV)-MH and treated with cycloheximide for 0 h–4 h–8 h–24 h. The cells were then collected and the amount of unprocessed proteins and GPI-SP were measured. Figure 7B shows that the stability of the GPI-SPs from P/S and PP/SV mutants matched the one of the GPI-SP from FR-MH.

In conclusion, our results represent the first attempt in addressing the metabolism of peptides released after GPI-anchor addition. Our approach consisted in linking the C-terminal end of PrP and FR GPI-SP to a Myc-His tag and then following the fate of the released GPI-SP. This C-terminal tagging approach had been previously shown not to block the process of GPI-anchor addition, despite slowing down this process. We validate our approach by showing that our chimeric constructs are correctly processed, and that the resulting proteins behave as their untagged counterparts. We were then able to show that GPI-SP from PrP and FR are subjected to different fates. While the GPI-SP of FR is retained within the lumen of the ER, the highly conserved GPI-SP of PrP undergoes degradation via the proteasome. A hybrid GPI-SP containing the hydrophilic portion of PrP GPI-SS and the hydrophobic stretch of FR was also subjected to degradation, indicating that the clearance of PrP GPI-SP is due to the presence of key amino acids in the hydrophilic segment. This seemed quite unexpected, as while the hydrophobic region of PrP shows 100% amino acid identity across species, only the second half of the hydrophilic domain is completely conserved. Interestingly, among the conserved amino acids are two proline residues, one of which is mutated to Ser in a genetic case of Prion disease. When this mutation was introduced in PrP GPI-SS, we observed a 15% recovery of PrP GPI-SP. Moreover, when both Pro were mutated, the recovered GPI-SP raised to 80%. Connecting the mutation P238S to the accumulation of PrP GPI-SP has the potential to establish a link between the accumulation of PrP GPI-SP and PrP induced neurodegeneration and opens the intriguing possibility of an involvement of PrP GPI-SP in prion diseases.

Materials and Methods

Cell lines and transfections

HeLa cells (obtained by Dr. Cossart P., Institut Pasteur, Paris, France) were maintained in Dulbecco's modified Eagle's medium (DMEM, Invitrogen) supplemented with 10% of fetal calf serum (FCS) in a 5% CO2 incubator at 37°C. Cells were transfected at 70% confluence using FuGENE6 (Roche Diagnostic) according to manufacturer's protocol.

Antibodies and reagents

Primary antibodies: anti-His (rabbit) was a gift from Dr. Dautry, Institut Pasteur. Anti-GFP3E6 (mouse) was from Molecular Probes. Anti-calnexin (mouse) was from AbCam. Secondary antibodies for IF: Alexa fluor 488 goat anti mouse (1:500) and Alexa Fluor 594 goat anti rabbit (1:500) from Molecular Probes. 3-Methyladenine (6-amino-3-methyl-purine) was purchased from Fluka and was used at 10 mm concentration; NH4Cl (Sigma) was used at 20 mm; Epoxomicin (Sigma) was used at 1 µm for 6 h; Carbobenzoxyl-leucinyl-leucinyl-leucinal (MG132, Sigma) was used at 10 µm for 6 h; N-Acetyl-leucyl-leucyl-norleucinal (ALLN, Sigma) was used at 10 µm for 6 h. Cycloheximide (Sigma) was used at 100 µm for 0 h–4 h–8 h–24 h.

DNA constructs and site-directed mutagenesis

All constructs were designed by inserting into pcDNA3.1MycHis expression vector (Invitrogen) the N-terminal sequence of PrP (NheI/XhoI) linked to GFP (XhoI/EcoRI) and the GPI-anchor signal sequence and Gly/Ser linker (EcoRI/BamHI), followed by the Myc-His tag. Hybrid constructs were obtained via three sets of polymerase chain reaction (PCR) reactions: the first to amplify the hydrophilic parts of FR (5′-AAG GAG ATG AGG AGG ATG ACA GG CCA GGC TGC CCA GGG CCC AGC-3′) and PrP (5′-GGG CCA GGC TAA GCA GGA AAG GAG GGG AGG AGA AAA GCA CGG TG-3′); the second to amplify the hydrophobic portions of FR (5′-GGC CCT GGG CAG CCT GGC CTG TCA TCC TCC TCA TCT CCT TCC TC-3′) and PrP (5′-GCT TTT CTC CTC CCC TCC TTT CCT GCT TAG CCT GGC CCT AAT GC-3′); the third PCR used PCR-1 and PCR-2 as template to obtain the hybrid constructs. Mutants were obtained by site-directed mutagenesis using the QuickChange II XL site-directed mutagenesis kit (Stratagene). The oligonucleotides used to obtain the M/R mutant were: Eco-M/R-Fw (5′-CTG TAC AAG GAA TTC GAC GGG AGA AGA TCC AGC AGC AGA GTG CTT TTC TCC TCC-3′) and BGH-Rv (5′-TAG AAG GCA CAG TCG AGG-3′). For the P/S mutant: Eco-P/S-Fw (5′-CTG TAC AAG GAA TTC GAC GGG AGA AGA TCC AGC AGC ACC GTG CTT TTC TCC TCC TCT CCT GTC ATC CTC CTC-3′) and BGH-Rv. For the double mutant PP/SV: Eco-PP/SV-Fw: 5′-CTG TAC AAG GAA TTC GAC GGG AGA AGA TCC AGC AGC ACC GTG CTT TTC TCC TCC TCT GTG GTC ATC CTC CTC-3′) and BGH-Rv. All oligonucleotides were synthesized and purified by MWG-Biotech.

Immunofluorescence

HeLa cells plated on 12 mm glass coverslips were fixed with 4% formaldehyde, permeabilized with 0.1% TritonX100 in PBS for 5 min, then incubated in blocking buffer (PBS containing 10% FCS) for 30 min at room temperature. The cells were then incubated with primary and secondary antibody diluted in blocking buffer. Hoescht (1:100 000, Molecular Probes) was used to stain DNA. Coverslips were then mounted onto glass slides with Aqua/Poly Mount (Polysciences) and visualized using a Zeiss Observer Z1 156 inverted microscope with 63× objective controlled by AxioVision software (Zeiss).

Western blot

Cells were lysed with PAGE loading buffer (60 mm Tris, pH 6.8, 5% 2-mercaptoethanol, 2% SDS, 0.01% Bromophenol blue and 10% glycerol). Proteins in the lysate were separated by SDS–PAGE using a 4–20% gradient Tris-Tricine gel (Mini-Protean Tris-Tricine Precast Gels, BioRad). Proteins were transferred on Whatman Protran nitrocellulose membranes (0.1 µm pore size, Sigma). The membrane was kept in blocking buffer (50 mm Tris, pH 7.5, 150 mm NaCl, 0.05% Tween 20 and 3% BSA) and then incubated with the primary antibody diluted in blocking buffer. HRP-conjugated secondary antibodies and ECL™ reagents from Amersham (GE Healthcare) were used for detection on Kodak Biomax films.

Acknowledgments

The work in C. Z. lab was supported by fellowships from the Federation of European Biochemical Societies (FEBS), the Fondation pour la Recherche Médicale (FRM), the European Union (grant number FP7-KBBE-2007-2A-22287), the ANR (Agence Nationale de la Recherche) (grant number ANR-09-BLAN-0122) and the Pasteur–Weizmann Foundation (2010–2012). We also thank Dr. Guiliana Victoria for helpful discussion and comments.

Ancillary