Journal of Cell Science partnership with Dryad

Journal of Cell Science makes data accessibility easy with Dryad

Summary

The cellular generation of toxic metabolites and subsequent detoxification failure can cause the uncontrolled accumulation of these metabolites in cells, leading to cellular dysfunction. Amyloid-β protein (Aβ), a normal metabolite of neurons, tends to form toxic oligomeric structures that cause neurodegeneration. It is unclear how healthy neurons control the levels of intracellular oligomeric Aβ in order to avoid neurodegeneration. Using immunochemical and biochemical studies, we show that the Aβ-binding serine protease Omi is a stress-relieving heat-shock protein that protects neurons against neurotoxic oligomeric Aβ. Through its PDZ domain, Omi binds preferentially to neurotoxic oligomeric forms of Aβ rather than non-toxic monomeric forms to detoxify oligomeric Aβ by disaggregation. This specific interaction leads not only to mutual detoxification of the pro-apoptotic activity of Omi and Aβ-induced neurotoxicity, but also to a reduction of neurotoxic-Aβ accumulation. The neuroprotective role of Omi is further supported by its upregulation during normal neurogenesis and neuronal maturation in mice, which could be in response to the increase in the generation of oligomeric Aβ during these processes. These findings provide novel and important insights into the detoxification pathway of intraneuronal oligomeric Aβ in mammals and the protective roles of Omi in neurodegeneration, suggesting a novel therapeutic target in neurodegenerative diseases.

Introduction

Intracellular detoxification is one of the key pathways for cell survival in the face of stress induced by the accumulation and aggregation of protein metabolites. This is especially true in highly differentiated cells, in which the generation of toxic by-products during cellular metabolism is unavoidable, in order to maintain homeostasis. One of these proteolytic by-products, amyloid-β peptide (Aβ), is produced from metabolism of the amyloid precursor protein (APP) through sequential cleavage by β-secretase and γ-secretase in neurons (Selkoe, 2004; Winklhofer et al., 2008). The Aβ that is produced in neurons as a normal metabolite is prone to form toxic oligomeric structures (LaFerla et al., 2007; Selkoe, 2006; Tseng et al., 2004; Wirths et al., 2004). Both in vitro and in vivo studies have confirmed that Aβ oligomerization commences within cells, and that the resultant oligomeric Aβ accumulates intraneuronally in the brain (Lord et al., 2006; Takahashi et al., 2004; Walsh et al., 2000). Extracellular Aβ originates from intracellular pools, resulting in a dynamic equilibrium between these two pools (Mori et al., 2002; Oddo et al., 2006; Turner et al., 1996). In addition, naturally formed intracellular oligomeric Aβ may be secreted from cells (Walsh et al., 2002).

Oligomeric Aβ is not only the most toxic form of Aβ peptide to neurons, but is also the form of the Aβ peptide that correlates the best with the clinical symptoms of Alzheimer's disease (AD) (Billings et al., 2005; Chui et al., 1999; Gong et al., 2006; Gong et al., 2003; Kuo et al., 2001; Walsh et al., 2002; Wirths et al., 2001). Interestingly, oligomeric Aβ induces neuronal apoptosis more easily when it is inside rather than outside of neurons (Kienlen-Campard et al., 2002; Zhang et al., 2002). Therefore, it is crucial that the constant production of neurotoxic Aβ inside neurons is properly controlled, or that the protein is detoxified instantly, in order to prevent death of neuronal cells. Studies in transgenic (Tg) Caenorhabditis elegans have indicated that various heat-shock proteins (HSPs) interact with intracellular Aβ to reduce the toxicity of intracellular oligomeric Aβ (Cohen et al., 2006; Fonte et al., 2002; Fonte et al., 2008). Because HSPs are chaperones that are ubiquitously present in organisms ranging from bacteria to humans and are responsible for the processing or degradation of misfolded proteins, it is reasonable to propose that intracellular-Aβ-induced toxicity is counteracted by HSPs. However, the mammalian mechanism for detoxification of intracellular Aβ remains unclear.

Omi (also known as HtrA2) is a mammalian HSP and a homologue of the Escherichia coli survival factor HtrA (also known as DegP), which protects E. coli against heat stress and acts as a chaperone for unfolded proteins at low temperatures, and acts to refold or degrade misfolded proteins at elevated temperatures (Spiess et al., 1999; Vande Walle et al., 2008). Similar to its bacterial counterpart (HSP), in addition to a serine protease domain, Omi has a PDZ domain that recognizes partially folded or misfolded proteins (Huttunen et al., 2007). Previous studies have focused on the pro-apoptotic function of Omi in non-neuronal somatic cells. Following apoptotic stimuli, Omi has been shown to function as a pro-apoptotic protein; it stimulates apoptosis not only through a caspase-dependent mechanism via its interaction with inhibitors of apoptosis proteins (IAPs), but also through a caspase-independent pathway via cleavage of substrate proteins such as IAPs in the cytoplasm (Hegde et al., 2002; Suzuki et al., 2001; Verhagen et al., 2002; Yang et al., 2003). However, according to genetic studies, the normal physiological function of Omi appears to be anti-apoptotic. Both artificial Omi-knockout mice and natural Omi-inactivated mutant mice exhibit a neurodegenerative disorder, suggesting that Omi plays a neuroprotective role in the central nervous system (CNS) through an as-yet-unidentified mechanism (Jones et al., 2003; Martins et al., 2004). Recent studies indicate that Omi can also prevent lymphocyte apoptosis through suppressing the accumulation of activated Bax (Chao et al., 2008). These results have established the anti-apoptotic role of Omi in vivo, even though some in vitro cell studies have evidence to show that it might also act as a pro-apoptotic protein.

Previously, we identified Omi as one of four Aβ-interacting mitochondrial proteins (the others were ERAB, triose phosphate isomerase and HS1-associated protein X-1) in a yeast two-hybrid screen of a cDNA library of an adult human brain (Liu et al., 2005). Because the interaction of Omi with Aβ could have possible neuroprotective effects, we investigated the consequences of the endogenous interaction of Omi with Aβ in neurons. This interaction led not only to mutual detoxification of both the neurotoxic oligomeric Aβ and the pro-apoptotic Omi, but also to a reduction in the accumulation of toxic Aβ in cell cultures and mouse brains.

Fig. 1.

Omi localizes to mitochondria and the ER in neuronal cells. (A-D) Omi localized both within mitochondria and the ER in NT2 and K269 cells. The cells were stained with an anti-Omi antibody together with the subcellular-organelle markers MitoTracker Red or PDI (an ER marker). Images were visualized using a confocal microscope (LSM510, Carl Zeiss). (E-H) Omi localized within both mitochondria and the ER, but not in lysosomes or the Golgi complex in primary mouse cortical neurons. The primary mouse cortical neurons transiently expressing GFP-tagged Omi were stained with the subcellular-organelle markers MitoTracker Red, ERTracker Red or LysoTracker Red. For Golgi staining, a DsRed-Monomer-Golgi fusion protein was coexpressed with GFP-tagged Omi. Images were visualized using a confocal microscope (TCS SP5, Leica). Scale bars: 10 μm.

Results

Omi localizes to mitochondria and the ER in neuronal cells

Omi has a well-documented role as a mitochondrial pro-apoptotic protease (Martins et al., 2004; Suzuki et al., 2001), as well as an endoplasmic reticulum (ER)-localized chaperone-protease (Faccio et al., 2000; Huttunen et al., 2007) involved in the stress response in somatic cells. However, to date, the intraneuronal localization of Omi has been unclear. Confocal immunofluorescence imaging was used to examine the subcellular localization of Omi in a neuronal Ntera-2/D1 (NT2) cell line and in primary mouse cortical neurons. By co-staining with an anti-Omi antibody and organelle markers, Omi was shown to be localized to both the mitochondria and the ER in neuronal NT2 cells, as well as in somatic K269 cells (Fig. 1A-D). In primary mouse cortical neurons transiently expressing GFP-tagged Omi, the pattern of GFP fluorescence overlapped with both MitoTracker Red and ERTracker Red fluorescence, giving rise to a yellow signal (Fig. 1E,F). No colocalization was observed with the punctate, granular staining of lysosomes (Fig. 1G), or with the typical ribbon-like Golgi structure labelled by a DsRed-Monomer-Golgi fusion protein (Fig. 1H). Together, these results indicate that Omi localizes to both the ER and mitochondria in neuronal NT2 cells, as well as in primary cortical neurons, which is similar to its localization in somatic cells.

Omi binds preferentially to oligomeric Aβ rather than APP through its PDZ domain

We demonstrated that Omi is located in both the ER and mitochondria of neurons. Intracellular Aβ was observed to accumulate in both the mitochondria and ER of mouse brain cells (Greenfield et al., 1999; LaFerla et al., 2007; Manczak et al., 2006; Skovronsky et al., 1998), indicating that the location of Omi and Aβ overlap in the ER and/or mitochondria inside neurons. Considering the colocalization pattern of Omi and intraneuronal Aβ, it is possible that Omi interacts with Aβ within neuronal cells. To test this possibility, confocal microscopy was used to determine whether Aβ and Omi interact inside brain neurons. Neuronally differentiated human NT2N cells (clonal human neurons) and brain tissues from the Tg2576 mouse were chosen for immunofluorescent study, using commercially available anti-Omi and anti-Aβ C-terminal (no cross-reactivity with APP) antibodies (supplementary material Table S1). As shown in Fig. 2A, the immunocytochemistry images of NT2N neurons that were probed with anti-Omi and anti-Aβ antibodies overlapped exactly, indicating that Omi and Aβ colocalize inside NT2N cells. The colocalization of Omi and intracellular Aβ in brain tissue from Tg2576 mice were also observed by immunofluorescence histochemistry experiments using anti-Omi and anti-Aβ antibodies (Fig. 2B). Furthermore, no colocalization between Omi and APP was detected by staining with anti-Omi and anti-APP antibodies (supplementary material Fig. S1), indicating that Omi specifically colocalizes with intracellular Aβ, but not APP, in brain neurons from Tg2576 mice. Consistent with the confocal-fluorescence-microscopy data, western blot analysis of brain mitochondrial protein extracts (Fig. 2C) also confirmed that intracellular monomeric and oligomeric forms of Aβ exist in mitochondria that were isolated from the brains of Tg2576 mice.

Fig. 2.

Omi binds preferentially to oligomeric Aβ rather than to APP. (A) Omi and Aβ colocalized in NT2N cells, which are neuronally differentiated human NT2 cells. Primary rabbit anti-Omi and mouse anti-Aβ (specific to Aβ C-terminal and no cross-reactivity to APP) antibodies, in addition to secondary goat anti-rabbit IgG-FITC and horse anti-mouse IgG–Texas-Red antibodies, were used to detect intracellular Omi (green) and Aβ (red), respectively. Images were visualized using a confocal microscope (LSM510, Carl Zeiss). Scale bars: 50 μm. (B) Omi and Aβ colocalized in Tg2576 mouse brain tissues. Primary mouse anti-Omi and rabbit anti-Aβ (specific to Aβ C-terminal and no cross-reactivity to APP) antibodies, in addition to secondary Alexa-Fluor-488-conjugated goat anti-mouse IgG and Alexa-Fluor-568-conjugated goat anti-rabbit IgG antibodies, were used to detect intracellular Omi (green) and Aβ (red), respectively. Confocal images were obtained with a Carl Zeiss LSM510 Meta microscope. Scale bars: 20 μm. (C) Both monomeric and oligomeric forms of Aβ were found in Tg2576 mouse brain mitochondria, and Omi was also located in the mitochondria. WT, wild-type littermate; FL-Omi, full-length Omi without processing into mature Omi; Cyt C, the mitochondrial protein cytochrome C. Arrow indicates oligomeric Aβ. Asterisks indicate non-specific bands from the mouse brain mitochondria. (D,E) Omi and Aβ formed complexes inside K269 cells and Tg2576 mouse brain mitochondria. Omi-Aβ complexes were detected by reciprocal immunoprecipitation of K269 cell protein extracts and Tg2576 mouse brain mitochondria with either anti-Omi or anti-Aβ (6E10). rIgG and mIgG, nonimmune rabbit and mouse IgG, respectively. (F) Omi bound selectively to oligomeric Aβ in a dose-dependent fashion in an immunoblot binding assay. The same amounts of reverse Aβ (Aβ42-1), monomeric Aβ or oligomeric Aβ were spotted on the membranes as indicated. The upper immunoblot was initially hybridized with the Omi protein and then probed with an anti-Omi antibody. The lower immunoblots were directly detected with an anti-oligomer (A11) antibody specific to the oligomeric Aβ forms or an anti-Aβ (6E10) antibody that recognizes both monomeric and oligomeric Aβ. One representative experiment out of three independent experiments is shown.

In order to confirm that Omi directly interacts with Aβ, a co-immunoprecipitation–western-blot (IP/WB) assay was performed. Aβ-producing K269 cells and brain mitochondria isolated from Tg2576 mice were used to further confirm the endogenous interaction between the two proteins by reciprocal immunoprecipitation with anti-Omi and anti-Aβ antibodies, and the immunoprecipitates were subsequently immunoblotted (Fig. 2D,E). Omi-Aβ complexes were detected in immunoprecipitates from the protein extracts of both K269 cells and Tg2576-mouse brain mitochondria using either anti-Omi or anti-Aβ (6E10) (Fig. 2D,E). Non-immune IgG did not detect the Omi-Aβ complexes. According to the co-immunoprecipitation assay, Omi and Aβ formed complexes inside the mitochondria. The detected Omi-bound Aβ appeared to be a low-oligomeric form, which is the most typically observed neurotoxic form of Aβ in clinical AD brains (Walsh et al., 2000).

The detection of an endogenous interaction between intracellular oligomeric Aβ and Omi poses the following question: because the oligomeric forms of Aβ predominate within cells, can Omi selectively bind to oligomeric Aβ? In order to investigate this issue, a specific immunoblot-binding assay was performed (Fig. 2F). An immunoblot membrane was prepared by spotting equal amounts of reverse Aβ (Aβ42-1), monomeric Aβ or oligomeric Aβ, and then hybridizing the membrane to the mature Omi protein (aa 134-458) followed by an anti-Omi antibody. The results of this analysis clearly showed that Omi preferentially interacts with oligomeric Aβ in a dose-dependent fashion, and the blot-binding signal of monomeric Aβ or the control peptide Aβ42-1 was negligible (Fig. 2F, upper panel). The oligomeric conformation of Aβ used in this experiment was confirmed by the fact that the anti-oligomeric antibody, A11, only recognized the oligomeric form of Aβ, whereas the 6E10 antibody recognized both the monomeric and oligomeric forms of Aβ (Fig. 2F, lower panel).

Fig. 3.

Interaction of Omi with oligomeric Aβ leads to mutual detoxification. (A-C) Oligomeric Aβ reduced the apoptotic activity of Omi. (A) Caspase-3 activation by Omi was weakened by pre-incubation with oligomeric Aβ (lane 8 versus 6). Lane 1, cell extracts alone (as a non-activating procaspase-3 control); lane 2, caspase-3 activation induced by adding cyt-c/dATP into the cell extracts; lanes 3-5, cyt-c/dATP-activated cell extracts plus control peptide, monomeric Aβ and oligomeric Aβ (indicated as C, M and O, respectively); lanes 6-8, cyt-c/dATP-activated cell extracts plus Omi pre-incubated with C, M and O. The active caspase-3 levels in replicate experiments were quantified by densitometry and expressed as the band intensity relative to lane 2 (right panel). ***P<0.001, lane 6 versus 2, lane 8 versus 6. A two-tailed Student's t-test was used (mean ± s.d., n=3). (B) Formation of active Omi was affected by pre-incubation with oligomeric Aβ. Lanes 1-3, Omi incubated with C, M or O. (C) The proteolytic activity of Omi was inhibited by pre-incubation with oligomeric Aβ in a dose-dependent pattern. Lane 1, XIAP; lane 2, XIAP plus Omi; lanes 3-11, XIAP plus Omi pre-incubated with C, M or O at the indicated concentrations. (D-G) Omi reduced Aβ neurotoxicity by disaggregating toxic oligomeric Aβ. (D) Exogenous Omi reduced the neurotoxicity of oligomeric Aβ, increasing the viability of NT2 cells. ***P<0.001, versus 0 nM Omi. (E) The reduction of endogenous Omi augmented the neurotoxicity of oligomeric Aβ. *P<0.05; ***P<0.001, versus 0 nM siRNA-Omi. A two-tailed Student's t-test was used (mean ± s.d., n=4). (F) Omi disaggregated the toxic oligomeric Aβ, especially the trimeric and tetrameric forms of Aβ. Omi (0-0.5 μM) was incubated with 0.5 μM (lanes 1-3) or 1 μM (lanes 4-6) of oligomeric Aβ for 1 hour. (G) Omi prevented the formation of toxic oligomeric Aβ. Monomeric Aβ (1 μM) was incubated without (lane 1) or with (lane 2) 0.5 μM of Omi under oligomerization conditions for 24 hours. One representative experiment out of three independent experiments is shown.

It has been reported that Omi binds to monomeric Aβ via the PDZ domain of Omi (aa 364-445) (Park et al., 2004). To verify whether this domain is also responsible for the interaction with oligomeric Aβ, the membrane was hybridized with an inactivated mature Omi mutant (aa 134-458, OmiS306A) or PDZ-domain-truncated mature Omi protein (aa 134-364, OmiΔPDZ), and then detected by an anti-Omi antibody that can recognize both OmiS306A and OmiΔPDZ in a western blot assay (data not shown). As shown in supplementary material Fig. S2, mutation of the serine-protease site (changing Ser306 to Ala) inactivated the protease, but did not result in the alteration of its Aβ-binding ability. Consistent with the previous result, OmiΔPDZ showed strong β-casein-hydrolysis activity, but could bind neither monomeric nor oligomeric Aβ (Fig. S2C,D). These results clearly demonstrate that Omi binds preferentially to oligomeric Aβ, the most neurotoxic form of Aβ inside neurons, through its PDZ domain.

Oligomeric Aβ reduces the apoptotic activity of Omi

Because both oligomeric Aβ and Omi are known to activate apoptotic pathways, we investigated the consequences of the Omi-Aβ interaction upon neuronal cytotoxic pathways. We used an in vitro cell-free caspase-3-activating assay to test whether the interaction of Aβ with Omi affected the ability of these proteins to induce apoptosis. As shown in Fig. 3A, cytochrome C (cyt c)/dATP-induced caspase-3 activation (lane 2) was not affected by the control peptide (lane 3) or monomeric Aβ (lane 4), but increased slightly in the presence of oligomeric Aβ (lane 5). As expected, the presence of Omi significantly potentiated caspase-3 activation (lane 6). However, it should be noted that pre-incubation of Omi with oligomeric Aβ did not strengthen but actually weakened the ability of oligomeric Aβ to activate caspase-3 (lane 8 versus 6). Therefore, it is likely that the interaction between oligomeric Aβ and Omi inhibits the caspase-3-dependent apoptosis pathway. In addition, we found that pre-incubation of Omi with oligomeric Aβ resulted in a reduction in the active dimeric and trimeric forms of Omi, but pre-incubation with monomeric Aβ or control Aβ42-1 peptide did not have the same effect (Fig. 3B). Along with destabilization of the active Omi trimer by oligomeric Aβ, auto-proteolysis of Omi was slightly accelerated in the presence of the oligomeric form (Fig. 3B). As a consequence, the proteolytic activity of Omi towards its endogenous substrate protein, X-chromosome-linked IAP (XIAP), was clearly inhibited in a dose-dependent fashion by oligomeric Aβ, but not by monomeric Aβ (Fig. 3C, lanes 8-11 versus lanes 4-7). These results imply that oligomeric Aβ might inhibit the caspase-3-independent apoptosis pathway that is mediated by Omi by selectively reducing the active form of Omi. Taken together, these data suggest that Omi inhibits both the caspase-3-dependent and -independent apoptosis pathways by specifically interacting with oligomeric Aβ.

Omi reduces neurotoxicity of Aβ via the disaggregation of toxic oligomeric Aβ

MTT reduction assays were performed to directly assess the consequences of the interaction between oligomeric Aβ and Omi on Aβ neurotoxicity in neurons. Consistent with previous results (Kim et al., 2003), even nanomolar concentrations of oligomeric Aβ were highly toxic to NT2 cells, whereas Omi was only slightly toxic, even at the highest concentration tested (Fig. 3D). However, Aβ cytotoxicity was significantly attenuated by incubation with Omi in a dose-dependent manner at all tested concentrations. This suggests that the interaction of exogenous Omi with Aβ can reduce the neurotoxicity of oligomeric Aβ in cultured neuronal cells, indicating that the interaction between Omi and intracellular Aβ might have a potential detoxification role. In order to examine whether endogenous Omi can also detoxify oligomeric Aβ, we examined the effects of intraneuronal Omi on oligomeric-Aβ toxicity using primary mouse cortical neurons. The expression levels of intraneuronal Omi were significantly reduced by treatment with Omi siRNA (Fig. 3E, insert panel). As expected, this reduction in Omi expression significantly increased the toxicity of oligomeric Aβ in a dose-dependent fashion at all tested concentrations (Fig. 3E). These results suggest that endogenous Omi can detoxify oligomeric Aβ in mouse cortical neurons.

To determine how the interaction of Omi with oligomeric Aβ could reduce Aβ-induced neurotoxicity, the incubated mixtures of Omi and oligomeric Aβ were analyzed by immunoblot assays (Fig. 3F). Similar to the disaggregating effects of worm HSP upon oligomeric Aβ (Cohen et al., 2006), Omi dose-dependently disaggregated toxic oligomeric Aβ, particularly the trimeric and tetrameric forms, through its interactions with oligomeric Aβ. To further confirm the disaggregating effect of oligomeric Aβ by Omi, Omi was co-incubated with monomeric Aβ during the incubation for Aβ oligomerization. Immunoblot analysis of the oligomeric-Aβ species resulting from this incubation demonstrated that the formation of dimeric, trimeric and tetrameric Aβ was clearly inhibited by incubation with Omi (Fig. 3G). Although Omi disaggregated oligomeric Aβ and inhibited Aβ oligomerization, it did not directly degrade Aβ, which is similar to the observations that the worm HSP suppresses Aβ toxicity without directly degrading Aβ (Fonte et al., 2008). Therefore, these findings indicate that Omi acts as a mammalian version of HSP to reduce Aβ toxicity to neuronal cells by either disaggregating toxic oligomeric Aβ, or by preventing the formation of toxic oligomeric Aβ.

Omi expression is upregulated during neurogenesis and neuronal maturation, along with an increase in the generation of oligomeric Aβ

It has been shown that the expression of HSP is induced by the accumulation of intracellular Aβ in Tg worms (Fonte et al., 2008). Because the generation of Aβ and the accumulation of oligomeric Aβ increase continuously with neuronal age, it is of interest to study the regulation of Omi expression in neurons. In order to study Omi expression in an endogenous environment, an in vitro neuronal differentiation and maturation system was used initially. NT2 cells were treated with retinoic acid (RA) to induce differentiation, and were subsequently manipulated to yield pure cultures of NT2N neurons (Fig. 4A). During the manipulation, extracellular Aβ levels were upregulated, as detected by immunoprecipitation and western blot assays (Fig. 4B). Along with neuronal differentiation in NT2 cells, as well as the increase in the number of neuronal processes in NT2N cells with age, the expression levels of Omi clearly increased (Fig. 4C), which was similar to the pattern of generation of neurotoxic Aβ. Furthermore, the correlation between Omi expression levels with neuronal development was confirmed in mouse brain cortex and cerebellum (Fig. 4D). Although Omi was upregulated during neurogenesis and neuronal maturation, somatic cells and non-neuronal tissues exhibited constant levels of Omi, as expected (Fig. 4C, EcR293/RA; and 4D, mouse liver). Taken together, these data show that the inducible expression of Omi correlates with generation of Aβ along with neurogenesis, neuronal maturation and mouse brain development.

Omi expression affects cellular Aβ release, thus reducing Aβ-induced stress

In addition to the detoxifying effects of the interaction between Omi and intracellular Aβ, we examined whether Omi affects cellular Aβ release, thus reducing Aβ stress in the brain. Using quantitative IP/WB assays and ELISA, we found that reducing the Omi expression levels in K269 cells by using Omi siRNA clearly enhanced Aβ secretion (Fig. 4E,F). The extracellular Aβ levels in cells with reduced Omi expression increased 59%, from 70±14 to 111±15 pg/ml, as measured by ELISA. In accordance with these data, overexpression of Omi significantly reduced Aβ secretion (Fig. 4G,H), and the levels of extracellular Aβ decreased 37%, from 98±13 to 62±9 pg/ml, as measured by ELISA. These results indicate that the expression of Omi is inversely correlated with Aβ secretion, and this might result in a reduction of Aβ-induced stress in the brain.

Direct interaction between Omi and intracellular Aβ leads to a reduction of Aβ-induced stress

Huttunen et al. also reported that Aβ secretion was increased in cells lacking Omi (Huttunen et al., 2007). They believed that the increase in Aβ secretion resulted from the APP-binding properties of Omi. However, our confocal-laser-scanning analyses showed that Omi clearly colocalized with intracellular Aβ (Fig. 2A,B), whereas the locations of Omi and APP did not overlap each other in brain neurons (supplementary material Fig. S1). To further test whether the reduction in Aβ release from cells overexpressing Omi is a result of a direct interaction between Omi and intracellular Aβ or APP-Omi interaction, we constructed a plasmid expressing an inactivated Omi mutant, which lacked any protease activity. The catalytic Ser306 residue in Omi was replaced by alanine using overlapping PCR (OmiS306A). As shown in supplementary material Fig. S2, OmiS306A lost its protease activity, but still preferentially bound to oligomeric Aβ, thus affecting the neurotoxic effects of Aβ. Overexpression of OmiS306A in K269 cells caused no detectable cell death (data not shown), but still resulted in a significant reduction in extracellular Aβ levels when compared to the vector control (Fig. 5A). Under the same conditions, overexpression of wild-type Omi led to a slightly greater reduction in Aβ secretion than did OmiS306A (Fig. 5A). Moreover, we examined the effects of Omi on the cleavage of APP by secretases. Overexpression of both wild-type Omi and OmiS306A caused no detectable effect on the cleavage of either mature or immature APP, and no detectable change of the APP fragments produced by the secretases was observed (data not shown). These results indicate that the reduction in Aβ secretion from cells overexpressing Omi is caused by the direct interaction between Omi and intracellular Aβ, and not by interactions between APP and Omi.

Fig. 4.

Omi expression is correlated with Aβ secretion. (A-D) Omi expression was upregulated along with neurogenesis and neuronal maturation. (A) Images of retinoic acid (RA)-induced NT2 differentiation and postmitotic NT2N process formation (visualized by a Nikon ECLIPSE TE2000-S microscope). Scale bars: 20 μm. (B) Aβ production increased along with neuronal differentiation in NT2/RA and NT2N cells, and Omi showed a similar expression pattern. Extracellular Aβ was immunoprecipitated by addition of the 6E10 antibody to media conditioned for 24 hours from NT2/RA and NT2N cells at the indicated time points, and then analyzed by an immunoblot assay. (C) Omi expression was upregulated along with neuronal differentiation (NT2/RA) and neuronal process formation (NT2N), but was not upregulated in a somatic cell line (EcR293/RA). Whole-cell lysates were used to analyze the expression levels of Omi by a western blot assay. (D) Omi expression was upregulated along with the development of brain tissues (cortex and cerebellum), whereas constant levels of expression were maintained in somatic tissues (liver). One representative experiment out of three independent experiments is shown. (E-H) Omi expression affected cellular Aβ release, thus reducing Aβ-induced stress. The conditioned growth media were collected and analyzed for levels of secreted Aβ with both a quantitative IP/WB assay (E,G) and ELISA (F,H). Aβ levels in replicate experiments of IP/WB assay (E and G, left panel) were quantified by densitometry and expressed as the band intensity relative to the siRNA-Ctrl (E, right panel) or vector control (G, right panel). (E,F) The reduction of Omi in K269 cells clearly increased Aβ secretion. The level of Omi was reduced by Omi siRNA (66 nM) when compared with controls (siRNA-Ctrl) (E). (G,H) Overexpression of Omi in K269 cells clearly reduced Aβ secretion. *P<0.05; **P<0.01; ***P<0.001, versus siRNA-Ctrl or the vector control. A two-tailed Student's t-test was used (mean ± s.d., n=3 or 4). One representative experiment out of three independent experiments is shown.

Absence of wild-type Omi causes accumulation of Aβ in the brain

Further in vivo studies of the relationship between Omi and Aβ were carried out using two different model mice: Omi-inactivated homozygous MND2 mice, which usually die by the age of 40 days (Jones et al., 2003), and Omi-knockout homozygous mice, which usually die around 30 days of age (Martins et al., 2004). In terms of Omi-Aβ interactions, MND2 mice and Omi-knockout mice have no wild-type Omi protein available to bind Aβ in the brain (Fig. 5B, upper panel). At 30-days old, the levels of Aβ in the brain were quantified in both mouse models using ELISA. As expected, the brain Aβ levels were significantly increased by up to 79% (Aβ40) and 90% (Aβ42) in Omi-knockout mice, and up to 49% (Aβ40) and 58% (Aβ42) in MND2 mice, compared with the wild-type controls (Fig. 5B, middle and lower panels). In the brain tissues of these two model mice, there was still no clear change in the total level of either mature APP or immature APP holoprotein, as well as in the levels of the APP fragments produced by the secretases (data not shown). The MND2 mouse has an Omi-inactivating missense mutation, whereas the Omi-knockout mouse completely lacks Omi. Therefore, we expected that the presence of inactivated Omi protein in the brains of the MND2 mice would lead to the accumulation of much less Aβ in the brains of these mice than in Omi-knockout mice. These in vivo data indicate that Omi influences Aβ accumulation in the mouse brain via its interaction with Aβ.

Fig. 5.

A direct interaction between Omi and intracellular Aβ leads to the reduction of Aβ stress both in vitro and in vivo. (A) Overexpression of inactivated OmiS306A clearly reduced Aβ secretion. Extracellular Aβ was immunoprecipitated from the conditioned growth media using the 6E10 antibody, and then analyzed by a western blot assay. FL-Omi, full-length Omi without processing into mature Omi. (B) The absence of wild-type Omi caused the accumulation of Aβ in the brain of MND2 and Omi-knockout (KO) mice. The total Aβ in transgenic mice hemi-brains was extracted by 70% formic acid, followed by neutralization with 2 M Tris base and analysis with an ELISA kit specific for mouse Aβ40 or Aβ42. Sample Aβ values were normalized to the protein concentration of the brain extracts and expressed as pg of Aβ per mg of protein. *P<0.05; **P<0.01; ***P<0.001. A two-tailed Student's t-test was used (mean ± s.d., n=4 of wild-type mice, and n=3 of MND2 or Omi-KO mice).

Discussion

One key characteristic of neurons is that their proteolytic processing of APP constantly generates Aβ as a normal metabolite, and this intracellular Aβ is prone to form oligomers (Casas et al., 2004; Knobloch et al., 2007; Oakley et al., 2006; Walsh et al., 2000). Because oligomeric Aβ is the most neurotoxic form of the various Aβ conformations (Kim et al., 2003; Kuo et al., 1996; Lue et al., 1999; McLean et al., 1999), neurons must have a mechanism to detoxify this toxic metabolite. In Tg worms, various HSPs contribute to cellular survival through their interactions with intracellular Aβ and subsequent reduction of intracellular-Aβ-induced toxicity (Cohen et al., 2006; Fonte et al., 2002; Fonte et al., 2008). It is reasonable to postulate that mammals also have a similar HSP-mediated detoxification pathway that plays a crucial role in the neuronal detoxification of intracellular Aβ. In neurons, ER- and mitochondria-localized Omi is a mammalian version of HSP that can bind and detoxify intracellular toxic Aβ for the following reasons: first, Omi is a neuroprotective homologue of the bacterial survival factor HtrA, the major HSP that protects bacteria from heat stress (Martins et al., 2004; Spiess et al., 1999). Second, Omi expression is upregulated during neurogenesis and neuronal maturation, as well as during mouse brain development, probably representing a protective response to the accumulation of toxic metabolites, and this expression is similar to the inducible HSP expression that protects Tg worms against intracellular-Aβ accumulation (Fonte et al., 2008). Finally, similar to HSP in Tg worms (Cohen et al., 2006), Omi does not directly degrade Aβ but clearly reduces the toxicity of Aβ in neuronal cells by disaggregating toxic oligomeric Aβ. Because the major function of bacterial HSPs is to refold rather than to proteolytically degrade denatured proteins, it is therefore not surprising that Omi reduces the neurotoxicity of Aβ by disaggregating oligomeric Aβ, rather than by directly proteolytically degrading toxic oligomeric Aβ.

The interaction between Omi and intracellular Aβ also inhibits the pro-apoptotic activity of Omi, as well as detoxifying oligomeric Aβ. Omi is a pro-apoptotic serine protease that is involved in caspase-dependent as well as caspase-independent cell death (Suzuki et al., 2001; Yang et al., 2003). It is well known that the formation of a homotrimer is necessary for Omi to exhibit proteolytic activity towards its substrate proteins, leading to apoptosis (Li et al., 2002). In this study, we showed that the Omi-induced apoptotic pathways (through the formation of active trimeric Omi and its proteolytic activity towards substrate proteins, e.g. XIAP) were strongly inhibited by the interaction of Omi with oligomeric Aβ. Thus, the possibility of causing neuronal apoptosis by upregulating the expression of the pro-apoptotic protease Omi can be efficiently avoided in neurons as a result of the interaction between Omi and intraneuronal oligomeric Aβ. These unforeseen effects of the specific interaction between Omi and Aβ provide crucial evidence for mutual detoxification of these two toxic molecules in neurons.

Biochemical and electron-microscopy studies revealed that both the monomeric and oligomeric forms of Aβ were located in the inner membrane of mitochondria (Caspersen et al., 2005; Manczak et al., 2006), and Omi was also found to be located in this region (Martins et al., 2004; Suzuki et al., 2001). In addition to their mitochondrial locations, Omi and Aβ were also observed in the ER (Faccio et al., 2000; Greenfield et al., 1999; Huttunen et al., 2007; Skovronsky et al., 1998). In this study, the interaction between Omi and intracellular Aβ in neurons was confirmed by immunofluorescent colocalization of these two proteins by confocal microscopy, indicating that these proteins interact in the mitochondria and ER. The interaction between Omi and intracellular Aβ was further confirmed by reciprocal co-immunoprecipitation experiments, as well as by immunoblot binding assays. The specific interaction of intracellular Aβ with Omi leads to the reduction of Aβ secretion and accumulation. As a consequence, a significant reduction in the extracellular levels of Aβ would be expected to lead to the relief of extracellular-Aβ-induced stress in the brain.

Investigation of two different Tg mouse models, Omi-inactivated MND2 mice and Omi-knockout mice, presented strong evidence for the protective role of the interaction between Omi and intracellular Aβ in mouse brain. Owing to the absence of wild-type Omi protein available for binding with Aβ, the brains of these two mouse models accumulated much higher levels of Aβ than wild-type mouse brains. Mice in the former model lack Omi protease activity, but still express Omi, whereas those from latter model have no Omi expression at all (Jones et al., 2003; Martins et al., 2004). According to our quantitative analysis of the levels of Aβ in the brains of these mice, this difference led to significantly different levels of Aβ accumulation at the same age in the two model mice. Because the inactivated Omi protein might still be able to bind to and detoxify intraneuronal Aβ, MND2 mice accumulated much less Aβ, and survived significantly longer (about 25% of their normal lifespan) than the Omi-knockout mice. Taken together, these data imply a direct link between the intraneuronal Omi-Aβ interaction and the levels of Aβ in the brain, as well as disease severity.

Similar to Aβ, α-synuclein can also form intraneuronal oligomers that are extremely toxic to neurons (Dawson and Dawson, 2003). It is known that different types of soluble amyloid oligomers, such as Aβ, α-synuclein and prion protein, have a common structure and possibly share a common mechanism of toxicity (Kayed et al., 2003). The accumulation and aggregation of these toxic proteins is observed predominantly in CNS-related neurons, and results in serious oxidative and ER stress, and proteosomal and mitochondrial dysfunction, leading to neuronal death (Winklhofer et al., 2008). Our recent observations show that Omi also interacts with intracellular oligomeric α-synuclein to avoid neuronal cell death (M.-L.L. and S.-T.H., unpublished data), suggesting a common role for Omi in protecting neurons against these toxic intracellular oligomeric proteins that share a common structure and toxic function. Similar to its role in controlling the accumulation of activated Bax in peripheral lymphocytes (Chao et al., 2008), this neuronal role of Omi might be a crucial protective pathway for relieving proteolytic accumulation and aggregation-mediated stress in neurons, and might provide mechanistic insights into the neuroprotective functions of Omi.

In conclusion, we showed that Omi binds preferentially to neurotoxic oligomeric forms rather than non-toxic monomeric forms of Aβ. There are at least three very important physiological consequences of this specific interaction: (1) detoxification of Aβ-induced neurotoxicity through the disaggregation of oligomeric Aβ, (2) an inhibition of the apoptotic activity of Omi by both caspase-3-dependent and -independent pathways, and (3) a reduction of Aβ accumulation both in cell culture and mouse brains. This study demonstrates the existence of a unique mutual-detoxification mechanism between neurotoxic oligomeric Aβ and pro-apoptotic Omi by specific interactions between these two proteins, which underlies an Omi-dependent neuroprotective response against protein accumulation and aggregation-mediated stress in neurons. It is highly desirable that the toxic effects of oligomeric Aβ can be ameliorated by the specific interaction of Omi with oligomeric Aβ through reducing Aβ accumulation, inhibiting Aβ oligomerization and disaggregating oligomeric Aβ. Therefore, the Omi-dependent detoxification pathway could be a prospective target for the development of amyloid-directed therapies for neurodegenerative disorders, including AD and Parkinson's disease.

Materials and Methods

Animals

Animal studies followed protocols approved by the Animal Care and Use Committee of the Chonbuk National University Medical School. Tg2576 mice were from Taconic Animal Models. C57BL/6J-MND2 mice were from Jackson Laboratory. C57BL/6J-Omi-knockout mice were from the Cancer Research UK London Research Institute. Homozygous and wild-type mice were obtained by crossing heterozygous mice, so that the ages and background strains of the mice were comparable. Genotypes were identified by PCR (Jones et al., 2003; Martins et al., 2004).

Animal-tissue preparation

To assay Omi expression levels, ICR mouse tissues (including cortex, cerebellum and liver) were isolated from mice at foetal day 14 (F14d), postnatal day 1 (P1d), 1 week (P1w), 8 weeks (P8w) and 12 months (P12m), and lysed with buffer A (20 mM Tris-HCl buffer, pH 7.5, containing: 150 mM NaCl; 2 μM leupeptin, aprotinin, pepstatin and E64; 2 mM AEBSF; 1 mM EDTA; 1% Triton X-100; and 1% NP40). After a BCA assay (Sigma), 30-60 μg of the total protein was loaded for SDS-PAGE and western blot assays.

To assay the Aβ levels, C57BL/6J (wild type, homozygous MND2, homozygous Omi-knockout) mouse hemi-brains were homogenized in seven volumes of 70% formic acid (FA). The homogenates were sonicated until clear (3×10 seconds, each with a 30-second break on ice). After centrifugation at 28,000 g for 2 hours, the supernatants were aliquoted into new tubes and stored at –80°C until they were used for ELISA. The total-protein levels in the FA extracts were quantified using a Bradford Assay Kit (Pierce).

Cell culture and differentiation

Primary cortical-neuron cultures were prepared from embryos at gestation-day 15-17 of ICR mice as described previously (Kao et al., 2004). NT2 cells (Stratagene) and the differentiated neurons (NT2N) were maintained as described previously (Pleasure et al., 1992). NT2 differentiation was induced by treatment with 10 μM all-trans RA (Sigma) for 4 weeks. To measure Omi expression levels, NT2 cells treated with RA and NT2N cells were harvested by scraping with a scraper at each indicated time point. As a control, somatic EcR-293 cells (Invitrogen) were treated with RA for 0-7 days. A HEK 293 cell line stably expressing human wild-type APP695 protein (K269 cell, provided by Dennis J. Selkoe, Harvard Medical School, MA) was maintained in DMEM/F-12 medium supplemented with 10% FBS, an additional 2 mM glutamine, 1% penicillin/streptomycin and 0.4 mg/ml G418.

Immunocytochemistry

In order to examine the subcellular localization of Omi, NT2 and K269 cells at 60% confluency were fixed by 4% paraformaldehyde and stained with anti-Omi antibody together with the subcellular-organelle markers MitoTracker Red or PDI (an ER marker). The anti-Omi and anti-PDI antibodies were then detected with secondary antibodies conjugated with FITC and Texas Red, respectively. Primary mouse cortical neurons at 3 days were transiently transfected with a plasmid expressing GFP-tagged Omi for about 16 hours, and then stained with the subcellular-organelle markers MitoTracker Red, ERTracker Red or LysoTracker Red for live cells according to the manufacturer's protocol (Molecular Probes). For Golgi staining, a DsRed-Monomer-Golgi fusion protein (Clontech) was co-expressed with GFP-tagged Omi. To detect the intracellular colocalization of Omi and Aβ, differentiated NT2N neurons at 8 weeks were double-stained with anti-Omi and anti-Aβ C-terminal (no cross-reactivity with APP) antibodies, followed by double secondary antibodies conjugated with either FITC (for Omi) or Texas Red (for PDI). Confocal images were obtained either with a Carl Zeiss LSM510 microscope (for NT2, K269 and NT2N) or a Leica TCS SP5 microscope (for primary mouse cortical neurons), using the vendor-provided software (LSM510 or Leica LAS AF 1.7.0).

Immunohistochemistry

In order to investigate the intracellular distribution of Omi, Aβ and APP in brain tissues, 12-month-old Tg2576 mice and age-matched littermates were fixed by transcardial perfusion with ice-cold PBS and then 4% paraformaldehyde. Brains were removed, post-fixed in the same fixative for 4 hours at 4°C, and incubated overnight at 4°C in 30% sucrose. After sectioning in the coronal plane on a freezing microtome, floating brain sections (40 μm) were treated with 70% formic acid for 10 minutes and blocked with 10% normal goat serum in PBS containing 0.25% Triton X-100 for 1 hour. The sections were then double-stained overnight at 4°C with mouse anti-Omi/rabbit anti-Aβ C-terminal (no cross-reactivity with APP), mouse anti-APP/rabbit anti-Omi, and mouse anti-Aβ and APP (6E10)/rabbit anti-Omi, respectively, using PBS containing 0.25% Triton X-100 and 2% normal goat serum for dilution. The secondary Alexa-Fluor-488-conjugated goat anti-mouse IgG and Alexa-Fluor-568-conjugated goat anti-rabbit IgG antibodies were used for detection. Confocal images were obtained with a Carl Zeiss LSM510 Meta microscope.

Isolation of mitochondria

Cortical mitochondria from Tg2576 mice and age-matched littermates were isolated as previously described (Caspersen et al., 2005; Manczak et al., 2006). The total mitochondrial protein was extracted with buffer A for co-immunoprecipitation and western blot assays.

Construction, expression and purification of Omi

In order to construct the plasmid for expressing GFP-fused Omi in cortical neurons, a cDNA coding for the full-length human Omi protein (pCMV.SPORT6.Omi, Image 97002RG) was subcloned into the pEGFP-N1 (Clontech) vector, upstream of the GFP coding sequence. The expression of the fusion protein in cortical neurons was confirmed by western blotting with both anti-GFP and anti-Omi antibodies. The mature Omi (Omi134-458) cDNA was introduced into the pET28a (Novagen) vector, and recombinant Omi (Omi134-458) was expressed in BL21 (DE3) pLys E. coli cells and purified as previously described (Liu et al., 2005). The OmiS306A and OmiΔPDZ mutants were created by PCR technology, and purified as described for wild-type Omi.

Preparation of Aβ42 assemblies and immunoblot binding assay

Synthetic Aβ42 (American Peptides Company, Sunnyvale, CA) was used to prepare oligomeric Aβ according to a previously described method (Kim et al., 2003). For monomeric Aβ, it was necessary to treat the peptide with 1,1,1,3,3,3-hexafluoro-2-propanol (HFIP, Sigma) immediately before each experiment, because storage even after HFIP treatment resulted in the formation of dimeric Aβ, which was cytotoxic and could bind Omi. HFIP was removed from the dissolved peptide by evaporation just before use. Equal amounts of reverse Aβ (Aβ42-1), monomeric Aβ or oligomeric Aβ were spotted onto the immunoblot membranes. The blots were hybridized with Omi and then probed with anti-Omi, or they were directly detected with an anti-oligomeric antibody (A11), which was specific to the oligomeric Aβ, or with anti-Aβ (6E10), which recognized both the monomeric and oligomeric forms of Aβ.

IP/WB assay

A highly sensitive IP/WB protocol was used that was able to detect as little as 200 pg of endogenously secreted Aβ, as previously reported by Walsh and colleagues (Walsh et al., 2000). Conditioned growth media from NT2/RA and NT2N cells, as well as K269 cells subjected to Omi knockdown or Omi overexpression, were evaluated using immunoprecipitation. The 6E10 or 4G8 antibodies were used to visualize the levels of immunoprecipitated Aβ on western blots. Whole-cell lysates from NT2/RA, NT2N, EcR293/RA and K269 cells were prepared by lysing with buffer A. After measuring the protein concentration with a BCA kit (Sigma), the lysates were loaded to measure Omi levels by western blotting.

For IP/WB assaying of the Omi-Aβ complex, K269 cell lysates or mouse brain mitochondrial protein extracts were diluted using the same buffer. An IP matrix (Santa Cruz Biotechnology) and rabbit anti-Omi (Alexis), mouse anti-Omi (R&D Systems) or 6E10 (BioSource) antibodies were used to pull down the Omi-Aβ complex from 0.5 ml diluted lysate, using non-immune IgG as a control.

In vitro cell-free caspase-3-activating assay

EcR293 cell extracts were prepared for a caspase-3-activating assay according to previously described procedures (Suzuki et al., 2001). For the caspase-3-activating assay, 5 μM Aβ42-1 (control peptide), mono-Aβ or oligo-Aβ was pre-incubated with or without 0.5 μM Omi for 4 hours, prior to mixing with cell extracts (about 50 μg protein), 2 μM horse heart cytochrome C (Sigma) and 1 mM dATP in a final volume of 30 μl. Activation of caspase-3 was allowed to proceed for 1 hour, and was stopped by adding 10 μl of 4×LDS sample buffer (containing 8% β-mercaptoethanol) and boiling for 5 minutes. After centrifugation at 22,250 g for 10 minutes, 15 μl of the supernatant was loaded for SDS-PAGE analysis. Immunoblotting was performed using rabbit anti-caspase-3 (BioSource).

Omi-proteolysis-activity assay

For the protease-activity assay, 15 μM of Aβ42-1 and 1.25-15 μM of mono-Aβ or oligo-Aβ were pre-incubated with 0.5 μM Omi at 37°C for 3 hours and then incubated with 0.2 μM human recombinant XIAP (R&D Systems) at 37°C for 2 hours in a total volume of 30 μl. The proteolytic reaction was stopped as described above. The effect of mono-Aβ or oligo-Aβ on XIAP degradation by Omi was detected by immunoblotting using a goat anti-XIAP antibody (R&D Systems).

Transfection

K269 cells were transfected using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions for plasmid DNA or siRNA, and the media was changed 5 hours later. The conditioned media was collected and analyzed for the levels of secreted Aβ using ELISA or IP/WB assays. The control (luciferase GL2) siRNA and Omi siRNA were from QIAGEN and had the same sequences as previously described (Martins et al., 2002). The siRNA transfection efficiency by Lipofectamine 2000 for primary cortical neurons has been reported to be up to 90%, and the reported method was followed with several modifications (Kao et al., 2004). Briefly, the primary cortical neurons were counted and diluted into culture medium that contained no penicillin/streptomycin sulphate. The diluted cortical neurons were mixed with a transfection pre-mixture of siRNA/Lipofectamine 2000 and immediately plated onto poly-D-lysine/mouse laminin-precoated coverslips or plates. Following these procedures, the neurons grew normally during the tested periods, and more than 90% of the neurons were fluorescein-positive when transfected with BLOCK-iT Fluorescent Oligo (Invitrogen) as an indicator of transfection (supplementary material Fig. S3). The overexpression or knockdown of Omi was demonstrated by immunoblot assay.

Cytotoxicity assay

Prior to the treatment of NT2 cells, oligomeric Aβ or an equivalent amount of F12 medium was incubated with Omi at 37°C for 1 hour. The incubated mixtures were diluted using fresh culture medium to the final concentrations as indicated, and then added to cells after removing the culture media. Primary mouse cortical neurons were transfected with siRNA and cultured as above for 48 hours, then treated with 0-100 nM oligomeric Aβ. At 24 hours after treatment, cell viability was tested by the MTT assay according to previously described methods (Kim et al., 2003). Data were expressed as a percentage of the absorbance readings from the control group.

ELISA for Aβ

The levels of Aβ in the conditioned growth medium from transfected K269 cells or in the mouse-brain FA extracts were assayed by the BioSource ELISA kits specific for human Aβ42 or specific for mouse Aβ40 and Aβ42 according to the manufacturer's protocols, and calculated using a corresponding standard curve that was generated in parallel.

Footnotes

  • Supplementary material available online at http://jcs.biologists.org/cgi/content/full/122/11/1917/DC1

  • This work was supported by the Regional Research Centers Program of the Korean Ministry of Education and Human Resources Development through the Center for Healthcare Technology Development.

  • Accepted February 23, 2009.

References

View Abstract