A Novel Interaction Between Aging and ER Overload in a Protein Conformational Dementia

Abstract

Intraneuronal deposition of aggregated proteins in tauopathies, Parkinson disease, or familial encephalopathy with neuroserpin inclusion bodies (FENIB) leads to impaired protein homeostasis (proteostasis). FENIB represents a conformational dementia, caused by intraneuronal polymerization of mutant variants of the serine protease inhibitor neuroserpin. In contrast to the aggregation process, the kinetic relationship between neuronal proteostasis and aggregation are poorly understood. To address aggregate formation dynamics, we studied FENIB in Caenorhabditis elegans and mice. Point mutations causing FENIB also result in aggregation of the neuroserpin homolog SRP-2 most likely within the ER lumen in worms, recapitulating morphological and biochemical features of the human disease. Intriguingly, we identified conserved protein quality control pathways to modulate protein aggregation both in worms and mice. Specifically, downregulation of the unfolded protein response (UPR) pathways in the worm favors mutant SRP-2 accumulation, while mice overexpressing a polymerizing mutant of neuroserpin undergo transient induction of the UPR in young but not in aged mice. Thus, we find that perturbations of proteostasis through impairment of the heat shock response or altered UPR signaling enhance neuroserpin accumulation in vivo. Moreover, accumulation of neuroserpin polymers in mice is associated with an age-related induction of the UPR suggesting a novel interaction between aging and ER overload. These data suggest that targets aimed at increasing UPR capacity in neurons are valuable tools for therapeutic intervention.

CONFORMATIONAL disorders are mainly caused by deposition of proteins with conformational changes reducing the function and increasing formation of protein aggregates (Lomas and Carrell 2002; Kikis et al. 2010). Despite the fact that protein aggregation occurs in different tissues, neurons are especially sensitive for accumulation of proteins, which could lead to neurodegeneration. While tauopathies, Parkinson disease, and familial encephalopathy with neuroserpin inclusion bodies (FENIB) are commonly characterized by deposition of aggregated proteins in neurons, FENIB specifically exhibits conspicuous protein aggregates in the endoplasmatic reticulum (ER) (Lomas and Carrell 2002).

Serpinopathies, including α₁-antitrypsin deficiency and FENIB, are based on intracellular polymer formation between mutant variants carrying destabilizing point mutations (Lomas and Carrell 2002). Neuroserpin, an axonally secreted protein acts neuroprotectively and plays a role in synaptic plasticity (Borges et al. 2010; Wu et al. 2010). Serpinopathies are caused by intracellular polymer formation between mutant variants carrying destabilizing mutations that enable insertion of the reactive center loop of one serpin molecule into the β-sheet of another (Lomas and Carrell 2002). For example, FENIB is characterized by intraneuronal accumulation of neuroserpin polymers (Davis et al. 2002). To date, six autosomal dominantly inherited mutations have been described (Hagen et al. 2011), with the S49P mutation causing older age of onset and milder clinical course as compared to the H338R mutation.

Proteostasis, which is disturbed in dementias, encompasses the interplay of translational regulation with protein folding, trafficking, and processing but also assembly or disassembly and degradation of proteins (Kikis et al. 2010). Protein degradation, which is essential for proteostasis, is executed by the ubiquitin/proteasome system (UPS) and the autophagy–lysosome pathway (Rubinsztein 2006). Secretory proteins that accumulate in the ER are retrotranslocated into the cytosol and degraded via ER-associated protein degradation (ERAD) (Rubinsztein 2006), which can be upregulated by the UPR (Ron and Walter 2007). The UPR is a signaling program executed by ER transmembrane receptors activating transcription factor 6 (ATF6), inositol requiring kinase 1 (IRE1), and double-stranded RNA-activated protein kinase (PKR)-like endoplasmic reticulum kinase (PERK), aimed at restoring ER homeostasis (Chakrabarti et al. 2011). ERAD is essential for degradation of monomeric mutant neuroserpin (Kroeger et al. 2009). In cultured cells, accumulation of polymeric neuroserpin leads to activation of NF-κB by the ER overload pathway in the absence of the UPR (Davies et al. 2009; Kroeger et al. 2009).

Caenorhabditis elegans models expressing aggregation-prone proteins such as polyQ-GFP or fragments of metastable human proteins have helped in deciphering molecular details of protein aggregation and proteostasis (Nollen et al. 2004; Garcia et al. 2007; van Ham et al. 2010). To study mechanisms of protein aggregation in FENIB, we created a disease model in C. elegans by introducing FENIB-associated mutations in the neuroserpin homolog SRP-2. Our data suggest that activation of either heat shock response or UPR influences aggregation of mutant SRP-2 in worms, while in FENIB mice, transient activation of the UPR diminishes with age and correlates with an increase in polymeric neuroserpin. These data support a crucial role for neuronal proteostasis mechanisms in conformational dementias.

Materials and Methods

Mouse experiments

Experiments were carried out in accordance with institutional guidelines of the University Medical Centre Hamburg-Eppendorf. Mice overexpressing mutant (TgNS^S49P) and wild-type [Tg(NS)] human neuroserpin (Galliciotti et al. 2007) were used. Unless stated otherwise, three mice per genotype were used at 8, 12, 20, 34, 45, 64, and >80 weeks of age.

Nematode strains and maintenance

Standard conditions were used for C. elegans propagation and crosses at 20° (Brenner 1974). Mutations and transgenes used in this study are listed as follows: hsf-1(sy441)I, ire-1(v33)II, unc-119(ed3)III, atf-6(ok551)X, and pek-1(ok275)X. Bristol N2 was used as a wild-type strain. nhx-2::mCherry::lgg-1; myo-2::GFP was provided by Cliff J. Luke.

Generation of transgenic nematodes

For srp-2::srp-2::YFP, the 7.1-kb fragment containing the promoter, full-length srp-2, and 3′-UTR was cloned into the YFP expression vector pLN022yfp. For unc-54::srp-2::YFP, a truncated unc-54 promoter of pPD88.27 and full-length unc-54 3′-UTR were fused to full-length (1.6-kb) srp-2 and then cloned into pLN022yfp. srp-2::srp-2^H302R::YFP and unc-54::srp-2^H302R::YFP were created by site-directed mutagenesis using the Quikchange Lightning kit according to the manufacturer’s instruction (Agilent technologies). Transgenic worms were generated by microparticle bombardment of 7 µg DNA using the Biolistic PDS-1000/HE (Bio-Rad) gene gun (Praitis et al. 2001).

RNAi

RNA interference was performed at 20° using the feeding method (Kamath et al. 2001). Synchronized L1 larvae were placed on IPTG-containing NGM plates (1 mM) seeded with Escherichia coli [HT115(DE3)] expressing double-stranded RNA as described (Timmons and Fire 1998).

Imaging and fluorescence recovery after photobleaching analysis

For C. elegans immunostaining, synchronized young adults were fixed, permeabilized, and stained with rhodamine-phalloidin (Molecular Probes) or DAPI (Sigma-Aldrich) as described (McIntire et al. 1992; Nonet et al. 1993). Worms were imaged using an Axioimager (Zeiss) or an Ultraview Vox spinning disc confocal (Perkin Elmer) microscope and analyzed with either Axiovision 4.7 (Zeiss) or Volocity 5 (Perkin Elmer) software. Fluorescence recovery after photobleaching (FRAP) analysis was performed as previously described (Brignull et al. 2006).

For immunohistochemistry of mice, brains were processed into 3-µm thin paraffin sections and stained using antibodies against NeuN (Chemicon), activated caspase-3 (R&D Bisosystems), and neuroserpin (Santa Cruz). Immunosignals for above-mentioned antibodies were quantified in the entire CA1 and CA2 hippocampal region (Galliciotti et al. 2007) at three anatomically matched levels with an Axioscope bright field microscope and Axiovision 4.1 (Zeiss) software (Sepulveda-Falla et al. 2011).

Electron microscopy

For preembedding Diaminobenzidin (DAB) analysis, 150-µm-thick sagittal vibratome sections were incubated with antineuroserpin antibody (1:2000) (Miranda et al. 2004). Sections were further processed according to Lappe-Siefke et al. (2009). For postembedding immunogold labeling, small pieces of cryoprotected cortices were mounted on specimen holders and immersed in liquid nitrogen. Ultrathin sections (80 nm) were cut and labeled according to Slot and Geuze (2007). Imaging was done with a Zeiss EM 902.

C. elegans motility assay

To measure motility, staged young adult nematodes were placed individually in a drop of M9 buffer, following 1-min acclimatization; completed body bends were counted for 1 min with at least 15 animals per experiment.

C. elegans generation time

Fifteen adult hermaphrodites were allowed to lay eggs for 4 hr at 20° to obtain a population of synchronized embryos. After removal of the adult worm, progenies were monitored for their developmental stage after 65 or 70 hr, respectively. At least three independent experiments were conducted.

Autophagy

Synchronized L1 worms were exposed for 48 hr to RNAi specific for genes required for autophagy, bec-1, unc-51, lgg-1, or to the empty RNAi feeding vector pPD129.36 (control). Subsequently the amount of aggregates in the head region of 10 worms was counted. As a control for efficient inhibition of the autophagy pathway, autophagosome formation of nhx-2::mCherry::lgg-1; myo-2::GFP worms was visualized upon RNAi treatment following 4 hr starvation as previously described (Miedel et al. 2012).

SDS–PAGE and immunoblotting

Tissue from mice or worms was lysed with either SDS sample buffer, SDS containing and SDS-free RIPA, or worm lysis buffer (Gidalevitz et al. 2009). Western blotting was performed under denaturing and nondenaturing (Miranda et al. 2004) conditions using antibodies against GFP (Clonetch), α-tubulin (Sigma-Aldrich), ATF6 (Imgenex), eIF2α/eIF2αP (Cell Signaling), neuroserpin (Miranda et al. 2008), LC3-I/-II (Thelen et al. 2012), and actin (Sigma-Aldrich). Visualization and quantification was done with either Imager Gel Doc System and Quantity One software (Bio-Rad) or Typhoon Trio scanner with ImageQuant software (GE Healthcare). Changes in protein expression of ATF6, LC3-I, LC3-II, or eIF2αP were calculated by normalization to β-actin or eI2Fα in three independent experiments, where wild-type values were set to one (Glatzel et al. 2001; Thelen et al. 2012).

Analysis of XBP1 mRNA splicing

RNA extraction and cDNA synthesis was performed with RNA Miniprep (Stratagene) and RevertAid (Fermentas) kits. PCR and restriction digestion with PstI, discriminating spliced and unspliced forms of XBP1, was performed as described before (Calfon et al. 2002).

ELISA

The ELISA was performed according to published protocols (Miranda et al. 2008). Three mice per genotype and timepoint were analyzed.

Data analysis

For all experiments, mean ±SEM is indicated. Statistical comparison among groups was calculated using two-tailed Student’s t-test with a significance value of *P ≤ 0.05 or **P ≤ 0.01.

Results

Accumulation of polymeric neuroserpin in the ER correlates with neurodegeneration

To study the relation between deposition of polymeric neuroserpin and neurodegeneration, we performed time course experiments with mice expressing the S49P mutant of neuroserpin, Tg(NS^S49P) or wild-type neuroserpin Tg(NS). Previous studies suggested that mutant neuroserpin accumulates intraneuronally in perinuclear compartments compatible with ER localization (Davis et al. 1999; Galliciotti et al. 2007). Using conventional and immunoelectron microscopy, we demonstrate progressive accumulation of mutant neuroserpin (Figure 1, A and B) and ER localization of inclusion bodies immunolabeled for neuroserpin in Tg(NS^S49P) mice (Figure 1C). Using nondenaturing gels, the polymeric nature of these aggregates became apparent (Figure 1D). Interestingly, the amount of neuroserpin deposits gradually raised, reaching a first plateau between weeks 34 and 45, followed by a sharp increase to week 64 and a second peak around week 80 (Figure 1B). Since it has been suggested that accumulation of polymeric proteins correlates with neurodegeneration (Davis et al. 2002; Lomas and Carrell 2002), we assessed neuronal loss and apoptotic cell death morphologically. Tg(NS^S49P) but not Tg(NS) mice display neuronal loss as quantified by immunostaining for the neuronal marker NeuN, labeling neuronal nuclei, in the CA1 and CA2 region of the hippocampus at 45 and 80 weeks of age (Figure 1, E and F). This was accompanied by a mild induction of apoptosis as assayed by positive staining for activated caspase-3 (Figure 1, G and H). Thus, we demonstrate that FENIB mice show considerable neurodegeneration and mild induction of apoptotic cell death. Although the correlation between increase in aggregated neuroserpin within the ER and apoptosis is not strict, it may contribute to neurodegeneration in FENIB mice.

Figure 1.

Accumulation of mutant neuroserpin in the ER correlates with neurodegeneration in FENIB mice. (A) Immunohistochemical staining for neuroserpin in CA1 and CA2 hippocampal regions of aged (8, 12, 20, 34, 45, 64, and >80 weeks) mice (n = 4). Neuroserpin accumulates progressively in Tg(NS^S49P) but not in Tg(NS) over time. (B) Quantification of sections shown in A. S49P neuroserpin aggregation reaches a temporary plateau from weeks 34 to 45 with an exponential rise at a later timepoint. (C) Electron micrographs from cortical neurons of >70-week-old Tg(NS^S49P) mice (left). Inclusion bodies (IBs) are found in the cytoplasm (Nuc, nucleus). IBs are positive in immunogold labeling for neuroserpin (IB, middle). In preembedding DAB-labeled tissue, neuroserpin positive IBs are found within rough ER (ER, right arrows). (D) Brain homogenates of 45-week-old Tg(NS^S49P) and Tg(NS) mice, analyzed by nondenaturating PAGE show mutant neuroserpin as a polymer ladder, wild-type neuroserpin as monomers. (E and G) Immunohistochemical stainings for NeuN and caspase-3 in the hippocampus of mice. (F and H) Quantification of sections shown in E and G. (F) NeuN positive neurons show substantial loss of neurons only in aged (45 and 80 weeks) Tg(NS^S49P) mice but not in Tg(NS) mice when compared to younger littermates (12 weeks). (H) Caspase-3 positive neurons show enhanced apoptosis in Tg(NS^S49P) mice. Bars, (A, E, and G) 50 µm; (C, left to right) 5 µm, 500 nm, and 1 µm. Statistical significance *P ≤ 0.05; NS P ≥ 0.05.

Analysis of SRP-2 aggregation in C. elegans

To further address molecular mechanisms underlying aggregation in serpinopathies, we generated a C. elegans model for serpin aggregation. Worms do not encode a direct homolog of neuroserpin, but express nine serpin-like proteins, which show highly conserved structural characteristics (Pak et al. 2004; Luke et al. 2006). Carrying out in silico analysis, we identified SRP-2 as the closest homolog to mammalian neuroserpin based on protein sequence similarity and conservation of critical domains (Pak et al. 2004) (Figure 2A and Supporting Information, Figure S1). This bioinformatic approach identified codon 302 of C. elegans srp-2 corresponding to codon 338 of human neuroserpin, with a histidine-to-arginine (H302R) exchange as the best modification to induce aggregation of SRP-2^H302R (Figure 2B). Another reason for choosing the disruptive H302R mutation was that milder neuroserpin mutations such as the S49P mutation develop FENIB much later in life, with detection of aggregates potentially exceeding life spans of short-lived organisms such as C. elegans. For visualization of the transgenic protein, YFP was fused to the C terminus of SRP-2. Transgenic lines expressing wild-type SRP-2 or SRP-2^H302R under the endogenous (srp-2::srp-2::YFP; srp-2::srp-2^H302R::YFP) or a muscle-specific (unc-54::srp-2::YFP; unc-54::srp-2^H302R::YFP) promoter were generated. These worms expressing SRP-2^H302R showed fluorescently labeled protein aggregates in the corresponding tissues (intestine and phasmid neurons or striated muscle) in an age-dependent manner (Figure 3A, Figure S2 and Figure S3). Aggregates are first detectable at L2 stage with most pronounced SRP-2 accumulation in young adults (Figure S1). Unlike previously described by Pak et al. (2004), we did not detect growth defects or developmental problems (Figure S4). This might be due to the lower copy number of our SRP-2 transgenes, which were integrated by biolistic transformation avoiding excessive overexpression (Praitis et al. 2001). To further define the nature of SRP-2 aggregates, we performed FRAP experiments for SRP-2^H302R, YFP, and polyQ-YFP expressed in body wall muscle cells (Morley et al. 2002; Brignull et al. 2006). In contrast to the wild-type protein, no obvious recovery was detectable for the mutant form (Figure 3, B and C), indicating that SRP-2^H302R forms insoluble polymers within cellular compartments. Polymer formation was confirmed by nondenaturing PAGE. Whereas wild-type SRP-2 runs predominantely as a single band, laddering of mutant SRP-2^H302R expressed from independent transgenic lines reflects polymeric aggregates of different size. We show this for two integrated (Is) and one extrachromosomal (Ex) line of either wild-type or mutant SRP-2 (Figure 3D).

Figure 2.

Comparison of neuroserpin and SRP-2. (A) Alignment of human (hNS) and mouse (mNS) neuroserpin with SRP-2. Sites causing FENIB are shadowed in red and amino acids forming a hydrophobic pocket in green (*, identical amino acids; : and ., amino acids with similar characteristics). (B) Structure of mouse neuroserpin showing close proximity of amino acids with disease-causing mutations (in red).

Figure 3.

SRP-2^H302R forms immobile protein aggregates in muscle cells leading to movement defects of transgenic worms. (A) In contrast to SRP-2 (wild-type), SRP-2^H302R forms protein aggregates when expressed under its own and muscle-specific promoter. (B) Fluorescent images of representative FRAP experiments (B) and their graphic demonstration (C) show efficient fluorescence recovery of SRP-2 but not SRP-2^H302R. SRP-2 (light green) shows slightly less fluorescence recovery than Q0::YFP (as control lacking inclusions, blue), whereas FRAP of SRP-2^H302R (purple) is comparable to Q40::YFP (as control expressing 40 repeats of fluorescently tagged polyQ proteins with considerable inclusions, red). The traces are averages of five or more replicates (n = 7). (D, upper panel) 7.5 % native PAGE of extracts from indicated strains. SRP-2^H302R (two independent integrated lines, hhls76, hhls77, and one extrachromosomal line hhEx39) forms high molecular weight aggregates (lanes 4 to 6) whereas SRP-2 (two independent integrated lines, hhls74, hhls75, and one extrachromosomal line, hhEx39, lanes 1 to 3) runs as a single band. Lower panel: SDS-Page to quantify tubulin for equal loading. (E) Confocal projections of transgenic worms showing the distribution of YFP–SRP-2 (green) and rhodamine-phalloidin that stained myofilaments (red). SRP-2 colocalizes with F-actin protein as shown by scatterplot, while SRP-2^H302R deposits as discrete fluorescent foci that do not disrupt the myofilament structure. (F) Body bends of worms (lines as described in D) per minute were measured. (n = 30) Bars, (A, B, and E) 50 µm. Statistical significance *P ≤ 0.05; NS P ≥ 0.05.

To further investigate whether these aggregates interfere with cellular integrity, we compared the localization of fluorescently tagged SRP-2 with F-actin of muscle fibers. While wild-type SRP-2 colocalized with actin-myosin filaments, SRP-2^H302R accumulated adjacent to actin-myosin filaments (Figure 3E). Aggregation of mutant SRP-2^H302R was associated with decreased muscle motility as measured by counting body bends of worms (Figure 3F). These movement defects can be reduced by RNAi-mediated depletion of the YFP fusion protein, thus SRP-2^H302R aggregation seems to affect muscle function (Figure 3F). As a control for the experiments we use the established lines for YFP and Q40::YFP as previously described (Gidalevitz et al. 2009).

Defects in the UPR cause enhanced SRP-2 aggregation in worms

Since mutant neuroserpin accumulates in the ER of transgenic mice, we were interested to see whether SRP-2^H302R aggregates similarly in C. elegans (Galliciotti et al. 2007). Given the motility defect of worms expressing mutant SRP-2^H302R under the muscle-specific promoter, these transgenic worms were used for further analysis. Indeed, aggregated SRP-2^H302R accumulates around the nucleus, suggesting a related localization in the ER as shown for mutant neuroserpin in mice (Figure 1C and Figure 4A) (Galliciotti et al. 2007). It is known that ER accumulation of misfolded proteins leads to ER stress and induction of the UPR, although for serpins this requires a second insult (Hidvegi et al. 2005; Marciniak and Ron 2006; Davies et al. 2009). Therefore, we investigated the importance of different stress response pathways and monitored SRP-2^H302R aggregation in the absence of the central transcription factor HSF-1, regulating the heat shock response or one of the three different pathways of the UPR that require IRE-1, ATF-6, and PEK-1 (PERK in mammals), respectively. Genetic depletion of HSF-1 led to strong increase in aggregate formation of SRP-2^H302R, while inactivation of either branch of the ER stress response led to significant but less severe aggregation, with pek-1 showing the most pronounced contribution (Figure 4, B and C). To further confirm these findings we used denaturing (Figure 4D, bottom) and nondenaturing (Figure 4D, top) PAGE to detect changes in the total level of mutant SRP-2 and the polymeric fraction. In line with the increased aggregate formation in muscle cells (Figure 4, B and C), the biochemical analysis of whole worm lysates suggests that loss of HSF-1 or PEK-1 activity causes an increase of polymeric SRP-2^H302R (Figure 4D, top), which is further reflected by increased total amounts (Figure 4D, bottom). Interestingly, we observed additive effects upon codepletion of hsf-1 with pek-1 or ire-1, resulting in significantly enhanced aggregation of SRP-2^H302R (Figure 4, E and F and Table S1). In contrast, ATF-6 activity seems to overlap with PEK-1, IRE-1, and HSF-1 function (Figure S6). Together, these results suggest that both cytosolic and ER quality control pathways might cooperate in proteostasis, modulating the aggregation of SRP-2^H302R.

Figure 4.

Genetic deletion of HSF-1 and UPR pathways increase the amount of SRP-2^H302R aggregates. (A) Fluorescent image of SRP-2^H302R shows perinuclear localization consistent with ER localization (DAPI staining; right picture is a close-up of the merged image). (B) Representative fluorescent images of mutants lacking hsf-1, ire-1, atf-6, and pek-1. (C) Quantification of aggregates of SRP-2^H302R worms described in B (n = 15). (D) Native and denaturing PAGE of extracts from indicated strains. Top native gel shows polymeric forms of SRP-2^H302R; bottom denaturing gel shows the total level of SRP-2^H302R probed with α-YFP antibodies. Tubulin was used as loading control. Integrated transgenic line hhIs76 was used for B–D. (E) Representative fluorescent images of mutant worms lacking either pek-1 or ire-1 after inactivation of hsf-1 for 48 hr by RNAi. The empty RNAi feeding vector pPD129.36 served as negative control. (F) Quantification of SRP-2^H302R aggregates described in E (n = 25). Bars, (A) 10 μm, (B and E) 50 μm. Statistical significance *P ≤ 0.05; **P ≤ 0.01 (two-tailed Student’s t-test).

Neuroserpin aggregation activates the UPR

As our analysis in worms provides evidence for the importance of ER stress pathways in modifying the formation of protein aggregates, we were interested in testing whether similar mechanisms exist in mammals. Therefore, we tested whether neuroserpin aggregation might correlate with the induction of ER stress markers in mice. ER stress causes activation of the PERK-, IRE1-, and ATF6-dependent pathways, which attenuate protein translation and induce chaperone expression to restore ER function (Ron and Walter 2007). Thus, we analyzed activation of the IRE1-mediated cleavage of XBP1 mRNA. We could detect processed forms of XBP1 mRNA in Tg(NS^S49P) and Tg(NS) mice at the assessed timepoints (Figure 5A). Since this assay does not allow for quantitative analysis of XBP1 splicing, subtle differences pointing to a mutation-specific induction of this branch of the UPR could not be detected and presence of cleaved XBP1 mRNA in Tg(NS) mice is most likely due to neuroserpin overexpression (Calfon et al. 2002; Galliciotti et al. 2007). We next measured the presence of activated or cleaved ATF6 fragments and the phosphorylated form of eukaryotic initiation factor 2 alpha (eIF2alphaP, as a PERK-dependent signal). Expression of cleaved ATF6 peaked in mutant neuroserpin mice between 34 and 45 weeks (Figure 5, B and C), whereas eIF2alphaP was highest in Tg(NS^S49P) mice at 20 weeks of age when compared to wild-type mice (Figure 5, D and E). This argues for an age-dependent selective and temporally limited activation of two different ER stress pathways in Tg(NS^S49P). The availability of an ELISA specific for polymeric neuroserpin (Miranda et al. 2008) allowed us to investigate the temporal distribution of polymeric neuroserpin in Tg(NS^S49P) and Tg(NS) mice. This analysis revealed a temporary decline of polymeric neuroserpin between weeks 20 and 45 and an exponential increase from week 64 onwards in Tg(NS^S49P), whereas polymeric neuroserpin could not be detected in Tg(NS) at any timepoint (Figure 5F). This is partially paralleled by a slight decrease of total neuroserpin from week 12 to week 34 and a subsequent exponential increase in total neuroserpin in Tg(NS^S49P), with a constant and low level of total neuroserpin in Tg(NS) (Figure 5G).

Figure 5.

Transient and exhaustible induction of the UPR in Tg(NS^S49P) mice negatively correlates with amounts of polymeric neuroserpin. (A) Reverse transcriptase–PCR from Tg(NS^S49P) and Tg(NS) aged 12, 20, and 34 weeks and tunicamycin-treated (T⁺) or untreated (T⁻) cells, to evaluate the levels of XBP1 (representative gel from three independent experiments). PstI digest distinguishes between processed and unprocessed XBP1. (B and D) Mutant Tg(NS^S49P) and Tg(NS) mice (n = 3) were analyzed at 12, 20, 34, 45, 64, and >80 weeks by Western blot for presence of cleaved ATF6 (B) and phosphorylated eIF2α (D), with actin and eIF2α as loading controls. (C and E) Densitometric analysis of Western blots from B and D. Wild-type levels were set to one. (C) Tg(NS^S49P) mice show increase of cleaved ATF6 with a maximum between 34 and 45 weeks. (E) Expression of phosphorylated eIF2α peaked in Tg(NS^S49P) mice at 20 weeks. (F and G) Tg(NS^S49P) and Tg(NS) mice (each n = 3) aged 8, 12, 20, 34, 45, 64, and >80 weeks were assessed for total and polymeric neuroserpin by ELISA. (F) Tg(NS) mice show no polymeric neuroserpin at any of the investigated timepoints. Polymeric neuroserpin was readily detectable in Tg(NS^S49P) at 8 weeks of age. Surprisingly, levels of polymeric neuroserpin drop at 34 weeks. With disease progression, levels of polymeric neuroserpin increased exponentially. (G) Tg(NS) mice show constant low levels of total neuroserpin, whereas in Tg(NS^S49P), neuroserpin plateaus between weeks 12 and 34; at later timepoints total neuroserpin increased exponentially. Statistical significance *P ≤ 0.05; NS P ≥ 0.05 (two-tailed Student’s t-test).

In summary, our data support the idea that the accumulation of mutant SRP-2 is sensitive to perturbation of cytosolic or ER proteostasis in the worm, which nicely correlates with the observation that partial activation of the UPR reduced the formation of mutant neuroserpin polymers in FENIB mice.

Discussion

Dementias are typically associated with the deposition of nonnative aggregated proteins (Lomas and Carrell 2002). Usually, protein aggregation is caused by aberrant protein interactions resulting in intracellular or extracellular deposits. Recently, the concept of neuronal proteostasis has been postulated, which explains the integrity of intracellular proteins regulated by quality control pathways that influence protein synthesis, folding, disaggregation, and degradation (Garcia et al. 2007; Balch et al. 2008). For example, secretory and ER resident proteins are specifically controlled by the UPR and ERAD (Schroder and Kaufman 2005). In certain dementias, intracellular protein accumulation seems to be involved in disease pathophysiology, yet the role of the UPR in protein aggregation and disease progression is unclear. Here we studied the importance of proteostasis in C. elegans and murine models of the conformational dementia FENIB, where aberrantly folded neuroserpin accumulates within the ER lumen causing neurodegeneration. FENIB is well suited as a model disease to study how ER dysfunction, ERAD, and protein aggregation are linked, since the entire disease pathophysiology is dictated by mutation-specific accumulation of a single protein in the ER, and mouse models fully recapitulate the human disease (Galliciotti et al. 2007; Takasawa et al. 2008). Our work extends these analyses and clearly shows ER localization of mutant neuroserpin and prominent neurodegeneration with the involvement of apoptotic pathways. The C. elegans model we established here differs from others, where chimeric polyQ stretches or human aggregation-prone proteins are expressed, by choosing an endogenous homolog to introduce disease-causing mutations (Link 1995; Cooper et al. 2006; van Ham et al. 2008; Brandt et al. 2009; Markaki and Tavernarakis 2010). Besides using the well-characterized promoter unc-54 (Morley et al. 2002) leading to expression in striated muscle, we have additionally expressed mutant SRP-2^H302R and wild-type SRP-2 under its endogenous promoter. Morphological and behavioral analysis of transgenic worms demonstrated aged-dependent deposition of SRP-2^H302R causing movement defects. Interestingly, these defects are associated with the presence of polymeric forms of SRP-2 in intracellular aggregates. Thus, our model nicely extends the existing repertoire of C. elegans aggregation approaches and shows that a carefully chosen single point mutation in a worm-encoded gene can phenocopy aggregation of the human homolog. Unlike other proteins that spontaneously accumulate in aged C. elegans (i.e., KIN-19 or RHO-1), SRP-2 only accumulates when specific disease-associated mutations are introduced (David et al. 2010). Since neuroserpin and SRP-2 aggregate within the ER lumen, we explored the relationship between protein aggregation and the UPR response. Interestingly, genetic disruption of all three branches of the UPR or central chaperones led to increase in both deposited and polymeric forms of SRP-2 in worms. However, an even more dramatic effect was seen following perturbation of the cytosolic heat shock response, arguing that serpin accumulation is sensitive to defects in general proteostasis. Having shown that impaired proteostasis can exacerbate protein deposition in worms, we looked for activation of the UPR in our murine FENIB model. These data demonstrate an age-dependent activation of ATF6 and PERK, whereas we could not detect IRE1 induction in mice expressing mutant neuroserpin. Beneficial effects of active ATF6 in the clearance of serpin polymers have recently been shown in cells thus supporting our finding of selective UPR activity (Smith et al. 2011). The induction of all three branches of the UPR is controversial for serpinopathies, with several reports showing absence (Davies et al. 2009; Kroeger et al. 2009) or presence of the UPR (Lawless et al. 2004). Instead, an ER stress response distinct from classical UPR signaling has been reported (Davies et al. 2009). The transient, age-dependent induction of the UPR suggests the necessity of a second insult, possibly aging, to mount an UPR response. This finding unifies current concepts of disturbed proteostasis in FENIB (Lawless et al. 2004; Hidvegi et al. 2005; Davies et al. 2009). Since IRE1 signaling has been linked to autophagy, the relatively low induction of IRE1 signaling in our murine FENIB model may help to explain the fact that autophagy is only mildly increased in FENIB (Hetz et al. 2009; Kroeger et al. 2009). Other than the serpinopathies, ER stress has been shown to be prominently involved in the pathophysiology of other dementias with intracellular protein accumulation (Hoozemans et al. 2009; Salminen et al. 2009). ERAD and the UPR are coregulated and the recently demonstrated involvement of ERAD in the degradation of mutant neuroserpin further underscores the tight integration of the proteostasis machinery (Kroeger et al. 2009; Ying et al. 2011). The mechanistic importance of ERAD is further extended by our data showing a fundamental role of HSF-1 in controlling SRP-2^H302R accumulation, further supporting the idea that HSF-1 might be linked to ERAD and UPR (Liu and Chang 2008). Intriguingly, our data demonstrate that transient stimulation of the UPR decreases neuroserpin polymer formation only in younger mice and probably age-related decline of ER quality control results in an exponential increase of neuroserpin polymers causing neurodegeneration (Figure 6). The fact that dementias show age-specific distribution and exponential rise of incidence with age argues in favor of age-specific phenomena perturbing neuronal protein homeostasis (Rubinsztein 2006; Salminen et al. 2010). Our data suggest that protein aggregation is tightly regulated by proteostasis pathways such as the UPR and ERAD, which seem to be modified by additional stressors accumulating with age. Future studies will further clarify whether ER stress pathways are specific for dementias with intracellular protein depositions, such as FENIB, or are also generally important for extracellular protein aggregation.

Figure 6.

Hypothetical model how transient UPR induction and aggregation-prone proteins are interconnected in conformational dementia. In our proposed model, transient induction of the UPR in young mice limits polymer formation. With increasing age, these mechanisms cease resulting in an exponential increase of protein aggregates. Increase in aggregates (present as polymers and potentially in equilibrium with oligomers) cosegregates with neurodegeneration and clinical disease develops once the functional reserve capacity of the CNS is exhausted.