Modulation of β-Amyloid Precursor Protein Processing by the Low Density Lipoprotein Receptor-related Protein (LRP)

β-Amyloid peptide (Aβ), which plays a central role in the pathogenesis of Alzheimer's disease, is derived from the transmembrane β-amyloid precursor protein (APP) by proteolytic processing. Although mechanisms associated with Aβ generation are not fully understood, it is known that Aβ can be generated within endosomal compartments upon internalization of APP from the cell surface. The low density lipoprotein receptor-related protein (LRP) was previously shown to mediate the endocytosis of APP isoforms containing the Kunitz proteinase inhibitor domain (Kounnas, M. Z., Moir, R. D., Rebeck, G. W., Bush, A. I., Argraves, W. S., Tanzi, R. E., Hyman, B. T., and Strickland, D. K. (1995)Cell 82, 331–340; Knauer, M. F., Orlando, R. A., and Glabe, C. G. (1996) Brain Res. 740, 6–14). The objective of the current study was to test the hypothesis that LRP-mediated internalization of cell surface APP can modulate APP processing and thereby affect Aβ generation. Here, we show that long term culturing of cells in the presence of the LRP-antagonist RAP leads to increased cell surface levels of APP and a significant reduction in Aβ synthesis. Further, restoring LRP function in LRP-deficient cells results in a substantial increase in Aβ production. These findings demonstrate that LRP contributes to Aβ generation and suggest novel pharmacological approaches to reduce Aβ levels based on selective LRP blockade.

␤-Amyloid peptide (A␤) 1 is the essential component of senile plaques (1,2), which are the major pathological hallmark of Alzheimer's disease (AD) (3). A␤ is derived from proteolytic processing of a ubiquitous transmembrane protein termed ␤-amyloid precursor protein (APP) (4). Although the factors governing production and deposition of A␤ are not fully understood, it has been shown that APP can undergo at least two post-translational processing pathways (3). In a nonamyloidogenic pathway APP is cleaved within the A␤ region by a proteinase activity known as ␣-secretase. This prevents generation of A␤ and gives rise to a soluble form of APP (sAPP␣), which is found in the extracellular milieu and constitutes a marker for nonamyloidogenic processing. Alternatively, APP can undergo proteolytic cleavage by ␤and ␥-secretases to generate A␤. This is referred to as the amyloidogenic pathway of APP processing, which can take place while internalized cell surface APP trafficks through the endocytic pathway (5).
Previous studies have demonstrated that the low density lipoprotein receptor-related protein (LRP), a member of the low density lipoprotein (LDL) receptor family, binds and mediates the endocytosis of soluble (6) as well as cell surface APP (7) isoforms containing a Kunitz proteinase inhibitor (KPI) domain. LRP is a large, multifunctional endocytic receptor abundantly expressed in liver and brain that mediates the hepatic uptake of circulating chylomicron remnants (8), serpin-enzyme complexes (9), and proteinases of the fibrinolytic pathway (10,11). Furthermore, LRP has been shown to regulate the cell surface levels of two receptors, the urokinase receptor (12) and tissue factor (13). LRP is also expressed in fibroblasts, macrophages, smooth muscle cells, neurons, and activated but not resting glial cells (14), suggesting that this receptor is involved in the binding and removal of interstitial ligands (e.g. proteinases and lipoproteins) produced by these cells. Targeted deletion of the LRP gene in mice leads to death of the embryo at day 13.5 (15), demonstrating that LRP plays a critical role during development. A 39-kDa receptor-associated protein (RAP) (16) binds reversibly to LRP and other members of the LDL receptor family such as gp330/megalin (17) and the very low density lipoprotein receptor (18) and inhibits ligand binding (19,20). RAP is found primarily in the endoplasmic reticulum, where it is thought to function as a molecular chaperone by assisting in receptor folding and processing and by preventing the association of newly synthesized receptors with endogenous ligands (21,22). Because of its high affinity for LRP and its ability to antagonize ligand binding, RAP constitutes a powerful tool to study LRP-mediated mechanisms.
The objective of the present investigation was to test the hypothesis that LRP plays a role in the pathobiology of AD by facilitating delivery of cell surface APP to endosomal compartments, where the A␤ peptide can be generated. To this end, we employed two main experimental strategies. The first one involved culturing cells in the presence of recombinant RAP as a means of preventing cell surface LRP-APP interactions. The results of these experiments demonstrated that long term incubation with RAP results in increased levels of both cell surface APP and sAPP␣, along with a significant decrease in A␤ production. As a second strategy the LRP gene was restored in LRP-deficient cells, and the consequences on APP processing were analyzed. These experiments revealed that LRP expression favors the amyloidogenic pathway of APP processing, leading to decreased sAPP␣ levels along with a significant increase in A␤ production.

MATERIALS AND METHODS
Proteins and Antibodies-Human RAP was expressed in bacteria as a fusion protein with glutathione S-transferase as described previously (20). Cleavage with thrombin and purification of recombinant RAP was carried out as described (20). Monoclonal antibody 5A6, which reacts with the LRP ␤ subunit (85 kDa), and monoclonal antibody 8G1, which reacts with the LRP ␣ subunit (515 kDa), have been described previously (23). The rabbit polyclonal IgG R777, which reacts with the LRP ␣ subunit, was prepared as described (17). Monoclonal antibody 6E10 (and biotinylated 6E10) raised against residues 1-17 of A␤ (specific for ␣-secretase cleavage prosite), and monoclonal antibody 4G8, specific for residues 17 to 24 of A␤ were purchased from Senetek PLC (St. Louis, MO). A rabbit polyclonal anti-A␤ that recognizes the C terminus of the peptide, was obtained from BIOSOURCE International (Camarillow, CA). Monoclonal antibody 22C11, raised against an epitope within the N terminus of APP, was purchased from Roche Molecular Biochemicals.
Cell Culture Conditions and RAP Treatment-H4 cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% heat-inactivated fetal bovine serum, 2 mM L-glutamine, 100 units/ml of penicillin, and 100 g/ml of streptomycin. For stably transfected H4/APP cells, the medium was supplemented with G418 (200 g/ml). 13-5-1 cells were grown in Ham's F-12 medium, supplemented with 5% heatinactivated fetal calf serum optimized for Chinese hamster ovary cells, 2 mM L-glutamine, 100 units/ml of penicillin, and 100 g/ml of streptomycin. In the case of 13-5-1/APP cells, 200 g/ml zeocine was added to the medium. For the experiments involving RAP treatment, subconfluent cells were cultured in the absence (control) or presence of 500 nM recombinant RAP (RAP-treated) for 5 days in serum-containing medium and then for 24 h in serum-free medium supplemented with Nutridoma®-NS medium supplement (Roche Molecular Biochemicals). Before changing to serum-free medium the cells were thoroughly washed with 1ϫ Dulbecco's phosphate-buffered saline to rid them of serum proteins. This 24 h conditioned medium was then used to detect both sAPP␣ and secreted A␤. During the 5-day period, cell were kept at Յ80% confluency.
Transfections-13-5-1 cells were stably co-transfected with the cDNA for human APP751 (cloned into pHD plasmid) (24) and pSec-Tag plasmid, which contained the zeocine resistance gene. 48 h later the selection process was initiated by adding 600 g/ml zeocine to the growth medium. 13-5-1/APP cells were transiently transfected with the cDNA for full-length human LRP (cloned into a pcDNA3.1 plasmid vector) or with vector alone (mock transfectants), using FuGene TM 6 transfection reagent (Roche Molecular Biochemicals) and following manufacturer's instructions.
LRP Functional Assays-Cellular uptake of 125 I-labeled ␣ 2 -macroglobulin activated with methylamine ( 125 I-␣ 2 M*) by 13-5-1/APP cells transiently transfected with LRP or vector alone was measured as described previously (25). Briefly, cells were plated onto 12-well plates at ϳ1 ϫ 10 5 cells/well and grown overnight at 37°C with 5% CO 2 . Cells were then washed twice with assay medium (Dulbecco's modified Eagle's medium containing Nutridoma medium supplement; 20 mM HEPES, pH 7.4, and 1.5% bovine serum albumin) and incubated in this medium for 1 h at 37°C. The cells were then incubated with assay medium containing 9 nM 125 I-␣ 2 M* for 5 h at 37°C. Following incubation, the cells were washed three times with assay medium and then detached with 0.5 ml of trypsin/proteinase K (to dissociate LRP-125 I-␣ 2 M* complexes from the cell surface) and then pelleted by centrifugation (3000 rpm for 5 min). The amount of internalized 125 I-␣ 2 M* was determined upon measuring the radioactivity associated with the cell pellet in a ␥-counter. Nonspecific uptake was assessed by measuring 125 I-␣ 2 M* internalization in the presence of excess amounts (1 M) of the LRP antagonist RAP.
SDS-PAGE and Immunoblotting-For detection of intracellular proteins, whole cell extracts were prepared using an SDS-free lysis buffer (0.5 M NaCl, 50 mM HEPES, 1% Triton X-100, 0.05% Tween-20) sup-plemented with proteinase inhibitors. To detect sAPP␣ and secreted A␤, 24 h conditioned media were concentrated 5-10ϫ either by using Centricon spin columns (3 kDa molecular mass cut-off) or by lyophilization and resuspension in water. Extracts and conditioned media were subjected to SDS-PAGE using pre-cast Tris-glycine gels from NOVEX (San Diego, CA). In all cases, gel loading was normalized to total protein content in the cell extract (or the corresponding cell extract when medium samples were used). Proteins were transferred onto nitrocellulose membranes, which were then blocked with Tris-buffered saline containing 3% (w/v) dry skim milk for at least 1 h at room temperature. The membranes were then incubated with primary antibody followed by the appropriate horseradish peroxidase-conjugated secondary antibody for at least 1 h at room temperature. Both antibodies were diluted in 3% milk/Tris-buffered saline containing 0.05% Tween-20. Immunoblotted proteins were detected using an enhanced chemiluminescence kit (Pierce).
Detection of Secreted A␤-A␤ present in conditioned medium was detected by sandwich ELISA, using antibody 4G8 (Senetek PLC) to capture A␤ and biotinylated antibody 6E10 (Senetek PLC) to detect it. Alternatively, antibody 6E10 was used to capture A␤ and rabbit polyclonal anti-A␤ (BIOSOURCE International) used for detection. Both assays gave similar results. 100-l aliquots of concentrated conditioned medium were incubated in 96-well plates previously coated with 100 l of capturing antibody (4 g/ml) and blocked with bovine serum albumin-containing buffer, for 2 days at 4°C. At the completion of this time, the wells were incubated overnight at 4°C with 100 l of the appropriate detecting antibody i.e. biotinylated 6E10 or polyclonal anti-A␤, used at a final concentration of 1 and 0.5 g/ml, respectively. Finally, depending on whether biotinylated 6E10 or polyclonal anti-A␤ was used as the detecting antibody, the wells were incubated for 2-4 h at room temperature with streptavidin-horseradish peroxidase conjugate or with goat anti-rabbit IgG-alkaline phosphatase conjugate, respectively. A␤ levels were quantified using a standard curve prepared with synthetic A␤ 1-40 (from BIOSOURCE International); control experiments showed that RAP did not interfere with A␤ detection. The data were normalized to mg of protein in the corresponding cell extract and statistically analyzed using the unpaired version of the Student's t test.
Quantification of Cell Surface APP-Monoclonal antibody 7H5 raised against the Kunitz protease inhibitor domain of APP751/770 (26) was 125 I-labeled using Iodogen reagent and used to detect surface APP751 in H4-APP cells that had been grown in the absence or presence of RAP (see RAP treatment methods above). The surface binding assay was performed as described previously (20). Briefly, cells were washed twice with assay medium (Dulbecco's modified Eagle's medium containing Nutridoma medium supplement, 20 mM HEPES, pH 7.4, and 1.5% bovine serum albumin) and then incubated in this medium for 1 h at 37°C. The cells were then chilled by placing them on ice for 1 h (to minimize endocytosis) and incubated with assay medium containing 10 nM 125 I-labeled 7H5 for 2 h at 4°C. Nonspecific binding was assessed by incubating cells with 10 nM 125 I-labeled 7H5 and excess (250 nM) unlabeled 7H5. In addition, the assay medium contained 250 nM mouse IgG to minimize binding to nonspecific IgG-binding sites. The cells were then washed three times with cold 1ϫ Dulbecco's phosphate-buffered saline and removed from the wells by treating with trypsin-versene. The radioactivity associated with the cells was measured in a ␥-counter.

Cells Cultured in the Presence of RAP Produce Less A␤ Than
Control Cells-To test the hypothesis that LRP-mediated internalization of cell surface APP alters the processing of APP, we sought to prevent the association of APP and LRP on the cell surface and study the consequences on APP metabolism and A␤ generation. To this end, we used RAP, a potent LRP antagonist that disrupts interaction of this receptor with all known ligands (20), including APP. H4 human neuroglioma cells stably transfected with human APP751 (H4-APP cells) were chosen for these experiments as they express significant amounts of LRP (Fig. 1A, right panel) and had previously been used to study APP processing (27). Expression of APP and production of A␤ by H4-APP cells was confirmed by immunoblotting (Fig. 1A, left panel) and sandwich ELISA, respectively (Fig. 1B). H4-APP cells were cultured for 5 days in the absence (control) or presence of recombinant RAP (RAP-treated). Immunoblotting of cell extracts showed that RAP treatment did not alter the total amount of APP present in cells (Fig. 2A).
Additionally, RAP treatment had no noticeable effect on the total levels of LRP as assessed by immunoblotting (data not shown). However, surface binding of an 125 I-labeled monoclonal antibody specific for the extracellular domain of APP (monoclonal antibody 7H5) revealed that RAP treatment caused a significant increase (ϳ3-4-fold) in cell surface APP levels (Fig.  2B). Moreover, immunoblotting analysis showed that the conditioned medium from RAP-treated cells contained substantially more sAPP␣ than that from control cells (Fig. 3A). We next analyzed the effect of RAP treatment on the amount of A␤ secreted by H4-APP cells. Fig. 3B shows that RAP-treated cells produced and secreted substantially less (ϳ4 -5-fold) A␤ than control cells. Control experiments confirmed that RAP does not interfere with detection of A␤ by ELISA. The effect of RAP treatment on APP metabolism (i.e. surface levels and A␤ generation) was not immediate and required at least 3 days in culture to observe the maximal change. Although it is not clear why prolonged treatment with RAP is required, other studies have found that prolonged treatment of cells with RAP is also necessary to increase cell surface levels of the urokinase receptor (28), a receptor system that can also be internalized by LRP.
In summary, these data show that blocking LRP function by prolonged treatment of cells with the LRP antagonist RAP favors the nonamyloidogenic pathway of APP processing, resulting in increased levels of both cell surface APP and sAPP␣, along with a significant decrease in A␤ production.
Expression of Human LRP in LRP-deficient Cells-Although RAP is a potent LRP antagonist, it is also known to bind to other members of the LDL receptor family (e.g. megalin/gp330 (17) and the very low density lipoprotein receptor (18)). To confirm the specificity of an LRP-mediated effect on APP processing, we sought to express the human LRP gene in LRPdeficient cells and analyze the consequences on APP/A␤ metabolism. We predicted that if LRP facilitates the amyloidogenic processing of APP as our experiments with RAP suggested, restoring LRP function in LRP-deficient cells should result in decreased levels of sAPP␣ and increased generation/secretion of A␤. To date, only a mouse fibroblast (29) and a Chinese hamster ovary cell line, termed 13-5-1 (25), have been developed that are genetically deficient in LRP. For these studies we used the Chinese hamster ovary cell line, because these cells are relatively easy to transfect and have been extensively used  /ml). B, A␤ in 24 h conditioned medium from parental H4 and H4-APP cells was detected by sandwich ELISA using 4G8 to capture and biotinylated antibody 6E10 to detect (see "Materials and Methods"). A␤ levels in conditioned medium were normalized to total protein content in the corresponding cell extract.

FIG. 2. RAP treatment causes accumulation of APP on the cell surface.
Subconfluent H4-APP cells were cultured in the absence or presence of 500 nM RAP for 5 days in serum-containing medium, followed by 24 h in serum-free medium. A, immunoblotting of cellular APP in extracts from H4-APP cells cultured in the absence (control) or presence of RAP (RAP-treated) using monoclonal antibody 22C11 (0.5 g/ml). Whole cell extracts were subjected to nonreducing 4 -12% SDS-PAGE (25 g of protein were loaded per lane). B, cell surface APP levels in control and RAP-treated H4-APP cells. To quantify surface APP levels, cells were chilled to 4°C (to prevent endocytosis), and binding of 125 I-labeled monoclonal antibody 7H5 (specific for the ectodomain of APP) was measured as described under "Materials and Methods." Nonspecific surface binding was minimized by including excess mouse IgG in the assay medium. In addition, the extent of 7H5 nonspecific binding was assessed by measuring surface binding of 125 I-labeled 7H5 in the presence of excess unlabeled 7H5. In all cases, nonspecific binding was subtracted from the total binding. *, p Ͻ 0.05 compared with controls.
to study APP processing (30,31). 13-5-1 cells were stably transfected with human APP751 (13-5-1/APP cells); expression of APP was confirmed by immunoblotting (Fig. 4A), and A␤ production in control and transfected cells was quantified by sandwich ELISA (Fig. 4B). 13-5-1/APP cells were then transiently transfected with the human LRP gene. LRP is synthesized as a single chain molecule of 600 kDa that is then processed in the trans-Golgi into its ␣ (515 kDa) and ␤ (85 kDa) subunits (32). Immunoblotting with antibodies specific for the ␣ and ␤ subunits showed that LRP was expressed and appropriately processed in the transiently transfected 13-5-1/APP cells (Fig. 5A). Normal function of the transfected receptor was confirmed by measuring cellular uptake of 125 I-␣ 2 M*, a major LRP ligand (23). As expected, only the LRP-transfected and not the mock transfected cells were able to internalize 125 I-␣ 2 M* (Fig. 5B), and this uptake was inhibited in the presence of excess RAP.
LRP Function Favors the Amyloidogenic Pathway of APP Processing-Once expression and function of transfected LRP was confirmed, we studied the impact of restoring LRP function on APP processing. The experiments revealed that although LRP expression did not significantly alter total cellular APP levels (Fig. 6A), the conditioned medium from LRP-transfected cells contained substantially less sAPP␣ than that from mock transfected cells (Fig. 6B). This finding supported the prediction that LRP function favors the amyloidogenic processing of APP suggested by our RAP treatment experiments. We then investigated whether restoring LRP function had also affected the amount of A␤ produced and secreted by cells. When conditioned medium from mock transfected and LRP-transfected 13-5-1/APP cells was assayed by sandwich ELISA to detect A␤, it was found that the medium from LRP-transfected cells contained approximately three to four times more A␤ than that of mock transfected cells (Fig. 6C). In summary, these results indicate that the presence of functional LRP favors the amyloidogenic pathway of APP processing, resulting in decreased levels of sAPP␣ and a dramatic increase in A␤ production.

DISCUSSION
The objective of the current investigation was to determine whether the multifunctional endocytic receptor LRP can modulate APP metabolism and in particular, the production of A␤ peptide. Two different experimental approaches were employed to test this hypothesis. In the first series of experiments we utilized RAP, a potent LRP antagonist, to prevent the association of APP and LRP on the cell surface. Long term culture of cells in the presence of RAP resulted in a significant reduction in the amount of A␤ peptide secreted into the medium. This decrease was associated with an increase in the steady-state level of cell surface APP and an increase in the amount of sAPP␣ present in the media. Although the need for prolonged treatment with RAP to observe maximal changes in APP metabolism is not understood at this point, it is possible that only a small fraction of the APP present on the cell surface interacts with LRP because APP-proteinase complexes are the preferred ligand for LRP (7). Formation of such complexes are likely to be limited by the availability of the proteinases. In fact, studies on the urokinase receptor, another receptor system that interacts with LRP, show that surface levels of this receptor increase significantly only after cells are cultured for at least 3 days in the presence of RAP (28). Although we have interpreted the effect of RAP treatment as blocking APP/LRP interactions on the cell surface, it cannot be ruled out that exogenously added RAP might also affect interaction of these molecules within intracellular compartments. However, the facts that both of these cell lines express endogenous RAP, which is localized intracellularly, and that RAP treatment caused an increase in cell surface levels of APP are more consistent with the hypothesis that RAP is acting on the cell surface.
To further confirm the role of LRP in APP processing, a second series of experiments were performed in which the human LRP gene was introduced into LRP-deficient cells stably expressing human APP. The results indicated that restoring expression and function of LRP in these cells resulted in decreased levels of sAPP␣ in the medium and a substantial in-crease in the amount of A␤ produced and secreted by the cells.
Together, the results of these experiments show that the lack of LRP activity correlates with a decline in A␤ production, whereas the presence of functional LRP is associated with increased generation and secretion of A␤. We propose that the interaction of APP with LRP on the cell surface modulates APP trafficking within the endocytic pathway such that its amyloidogenic processing is favored. This could occur if association of APP with LRP increased the resident time of APP within endosomes, where A␤ can be generated (5). Alternatively, binding of APP to LRP could result in conformational changes such that the ␤and/or ␥-secretase cleavage sites on APP become more accessible.
APP exists as multiple isoforms resulting from alternate splicing of its mRNA. The longer isoforms (APP770 and APP751) contain a KPI domain capable of inhibiting proteinases such as factors XIa and IXa (33) and are expressed in neurons as well as other cell types. Studies suggest that the majority of APP in the central nervous system contains the KPI domain (34). Moreover, KPI-containing APP isoforms seem to be more amyloidogenic (35), and their relative abundance in the brain increases with AD (36). The fact that the presence of a KPI domain on the extracellular region of APP is required for binding to LRP (6) may represent another clue to why the interaction of LRP with APP on the surface of neurons may contribute to the pathological processes leading to AD. Further, adaptor proteins such as FE65, whose protein interaction domains bind the cytoplasmic portions of both LRP and APP (37), may also be involved in linking these two molecules.
The dramatic loss of A␤ synthesis observed when LRP function is ablated either pharmacologically or genetically is consistent with a large body of data implicating LRP in the pathogenesis of AD. First, genetic studies reveal that a silent polymorphism in the LRP gene is associated with increased risk for AD (38 -41). Moreover, LRP expression is up-regulated in both reactive astrocytes and activated microglia (42), two components of mature amyloid plaques, and LRP itself is found in senile plaques along with many of its ligands (14,43). In addition to binding APP, LRP is the primary receptor in the brain for apoE and activated ␣ 2 M, two proteins shown to form complexes with A␤ (44,45). Thus, LRP is a converging point for the metabolism of three proteins (APP, apoE, and ␣ 2 M) genetically and biochemically associated with AD. The direct link we now demonstrate between LRP activity and A␤ production provides strong biochemical evidence for the involvement of LRP in the pathobiology of AD and suggests new approaches for reducing A␤ levels by preventing the association of fulllength APP with LRP. FIG. 5. Expression of human LRP in LRP-deficient 13-5-1/APP cells. A, LRP-deficient 13-5-1/APP cells were transiently transfected with vector alone (lanes 1) or the cDNA for human LRP (lanes 2). Whole cell extracts were subjected to nonreducing 4 -12% SDS-PAGE (20 g of protein/lane) and analyzed by immunoblotting for the LRP ␣ subunit (left panel) using R777 and the ␤ subunit (right panel) using monoclonal IgG 5A6. B, cellular uptake of 125 I-␣ 2 M* by mock transfected and LRP-transfected 13-5-1/ APP cells was assessed by incubating cells at 37°C for 5 h in the presence of 125 I-labeld ␣ 2 M* and then measuring the extent of internalized 125 I-labeld ␣ 2 M* as described under "Materials and Methods." Nonspecific uptake was assessed by measuring internalization of ␣ 2 M* in the presence of excess amounts of the LRP antagonist RAP.
FIG. 6. Restoring LRP in LRP-deficient cells facilitates the amyloidogenic pathway of APP processing. A, cellular APP was detected in LRP-deficient 13-5-1/APP cells transiently transfected with human LRP cDNA (LRP) or vector alone (mock). Whole cell extracts were subjected to nonreducing 4 -12% SDS-PAGE (20 g of protein/ lane) and to immunoblotting, using monoclonal antibody 22C11 (0.5 g/ml). B, detection of sAPP␣ in conditioned medium from mock transfected and LRP-transfected 13-5-1/APP cells by immunoblotting using monoclonal antibody 6E10 (1 g/ml). Gel loading was normalized to protein content in the corresponding cell extract. C, A␤ levels in conditioned medium from LRP-transfected and mock transfected cells were quantified by sandwich ELISA using monoclonal antibody 6E10 to capture and rabbit anti A␤ to detect (see "Materials and Methods"). A␤ levels were normalized to total protein content in the corresponding cell extract. The data shown are representative of three experiments. *, p Ͻ 0.05 compared with controls.