Transforming Growth Factor-β1 Potentiates Amyloid-β Generation in Astrocytes and in Transgenic Mice*

Accumulation of the amyloid-β peptide (Aβ) in the brain is crucial for development of Alzheimer's disease. Expression of transforming growth factor-β1 (TGF-β1), an immunosuppressive cytokine, has been correlated in vivowith Aβ accumulation in transgenic mice and recently with Aβ clearance by activated microglia. Here, we demonstrate that TGF-β1 drives the production of Aβ40/42 by astrocytes leading to Aβ production in TGF-β1 transgenic mice. First, TGF-β1 induces the overexpression of the amyloid precursor protein (APP) in astrocytes but not in neurons, involving a highly conserved TGF-β1-responsive element in the 5′-untranslated region (+54/+74) of the APP promoter. Second, we demonstrated an increased release of soluble APP-β which led to TGF-β1-induced Aβ generation in both murine and human astrocytes. These results demonstrate that TGF-β1 potentiates Aβ production in human astrocytes and may enhance the formation of plaques burden in the brain of Alzheimer's disease patients.

GTGGAGAG-3Ј, 340 bp. The conditions of amplification were 30 s at 95°C, 30 s at 58°C, and 1 min at 72°C for 30 cycles, corresponding to the 50% of the saturation curve of the PCR product.
Nuclear Extracts-Nuclear extracts were prepared from control and TGF-␤1-treated astrocytes and neurons. Cells were harvested 1 h after treatment and processed as described previously (14).
Electrophoretic Mobility Shift Assay-Oligonucleotides were end labeled with [␣-32 P]dCTP using the Klenow fragment of DNA polymerase. Binding reactions, containing 10 g of nuclear extracts and 2 ng of labeled oligonucleotides, were performed for 20 min at 37°C in appropriate binding buffer (14). The sequence of the double stranded oligonucleotide used as probe was: 5Ј-CGGGAGACGGCGGCGG-3Ј. Protein-DNA complexes were resolved in 5% polyacrylamide gels containing 0.5 ϫ TBE.
Primary Cell Cultures-Mouse cortical cultures of neurons were prepared from 14 -15-day-old embryos as described previously (15). After 3 days in vitro, neurons were treated with 10 M cytosine arabinoside to inhibit the proliferation of astrocytes. Experiments were realized on pure neuronal cultures (Ͼ 98% of microtubule-associated protein-2positive cells) after 12-14 days in vitro.
Murine cortical cultures of astrocytes were prepared from 1-3 day postnatal mice (15). Experiments were performed on confluent cultured cortical astrocytes after 10 days in vitro. Primary cultures of cortical astrocytes were washed three times in PBS each 3 days in vitro to prevent the adhesion of microglial cells onto the monolayer of astrocytes until use (10 -12 days in vitro). This protocol leads to fewer than 1% of CD11b-positive cells.
Primary cultures of human neurons and astrocytes were established from brain tissues of therapeutically aborted human brain fetuses after 10 weeks of gestation. The protocol for tissues so obtained complied with institutional and national guidelines.
Transgenic Animals-All transgenic mice were of BALB/c ϫ SJL background and heterozygous for the respective transgene. Nontransgenic littermates served as controls. Glial fibrillary acidic protein (GFAP)-TGF-␤1 line T65 has been generated similarly to lines 64 and 115, which have been described previously (16). Briefly, a 1.35-kb porcine TGF-␤1 cDNA was inserted into the first exon of a modified mouse GFAP gene. This cDNA had been mutated to allow secreted TGF-␤1 to be functional once released into the extracellular space. Homozygous TGF-␤1 transgenic mice develop communicating hydrocephalus (16); however, for this study we used heterozygous TGF-␤1 mice, which do not develop this complication. Identification of transgenic mice was performed by analysis of tail genomic DNA and from cerebral total RNAs.
Western Blotting Experiments-Cells were harvested in a lysis solution containing 50 mM Tris-HCl (pH 7.6), 1% Nonidet P-40 (Sigma), 150 mM NaCl, 2 mM EDTA, with 100 M phenylmethylsulfonyl fluoride in the presence of a protease inhibitor mixture (Sigma). Equally conditioned media were harvested in the presence of 100 M phenylmethylsulfonyl fluoride and protease inhibitor mixture prior vacuum concentration (10-fold). Electrophoreses were done on 8% SDS-polyacrylamide Tris-glycine gels or 16.5% Tris-Tricine gels containing 8 M urea. Gels were transferred to a polyvinylidene difluoride membrane (Poly-Screen ® , PerkinElmer Life Sciences), membranes were blocked in nonfat milk and probed with appropriate antiserum. Blots were finally developed with an enhanced chemiluminescence Western blotting detection system (PerkinElmer Life Sciences).
Double Fluorescent Immunocytochemistry-Cultured murine cortical astrocytes were gently washed in PBS and fixed with ice-cold 4% paraformaldehyde. Cells were washed, and nonantigenic sites were blocked in PBS plus 4% bovine serum albumin and 0.1% Tween 20 (Sigma) and incubated overnight at 4°C with the primary antibody raised against GFAP in PBS, 1% BSA, and 0.1% Tween 20. Cells were then washed and incubated for 1 h with the appropriate secondary Alexa Fluor ® 488-conjugated antibody (Molecular Probes). Thereafter, astrocytes were incubated with the OX42 monoclonal antibody as de-scribed above. The appropriate secondary biotin-conjugated antibody was used, and antibody-antigen complexes were revealed with streptavidin Alexa Fluor ® 555-conjugate (Molecular Probes). Cells were finally counterstained with DAPI-containing PBS and 0.1% Tween 20.
A␤ ELISAs-A␤ peptides were captured with a monoclonal antibody raised against the N-terminal domain of A␤ (8 -17) and revealed using a secondary antibody against the C-terminal extremity (40 or 42 end) of A␤ (BioSource Europe, Nivelles, Belgium). Either recombinant rodent or human A␤s (Calbiochem) were used as controls.
mRNA Decay Experiments-Mouse cortical astrocytes were exposed to 10 g/ml actinomycin D for 0 -16 h after 24-h treatment with recombinant human TGF-␤1 (R&D Systems Europe) before semiquantitative RT-PCR analysis. Gels were scanned and quantified by densitometry.
Transfection Protocols-Cells were transiently transfected with the constructs indicated using the Transfast ® Transfection Reagent (Promega) as described by the manufacturer. For each transfection experiment, sister culture dishes were used to control the efficiency of transfection using an enhanced green fluorescent protein (EGFP)-containing plasmid driven by cytomegalovirus promoter (pEGFP-C1 vector) (BD Biosciences-Clontech). All transfections were performed with the empty vector (pcDNA3.1, Invitrogen) corresponding to the experimental plasmid used (pcDNA3.1-APPtre-luc).
Transfection efficiency was determined by counting total cells and EGFP-positive astrocytes. In our hands, we reached the ratio of 71 Ϯ 8% of transfected astrocytes. Potential toxicity was estimated by examination of the cultures under phase microscopy and quantified by measurement of the activity of the cytosolic enzyme lactate dehydrogenase released by damaged cells into the bathing medium as described previously (21). No differences were observed between untransfected and transfected cells (data not shown). Moreover, the pRL-TK values provided by the Dual-Luciferase ® Reporter Assay System did not differ from controls (data not shown).
Reporter Gene Assay-Two days after transfection, cells were treated and luciferase activities (firefly luciferase and Renilla luciferase) were evaluated after 1 day using the Dual Luciferase ® Reporter Assay System (Promega). Values were normalized to the Renilla luciferase activity (Promega). The Dual-Luciferase ® Reporter Assay System refers to the simultaneous expression and measurement of two individual reporter enzymes within a single system. Typically, the "experimental" reporter (firefly luciferase) is correlated with the effect of specific experimental conditions (e.g. TGF-␤1 treatment), whereas the activity of the cotransfected "control" (Renilla luciferase) reporter provides an internal control, which serves as the base line of the response. Indeed, the pRL-TK control vector contains the herpes simplex virus thymidine kinase promoter region upstream of Renilla luciferase. It provides low level and constitutive expression in transfected cells. Normalizing the activity of the experimental reporter to the activity of the internal control minimizes experimental variability caused by differences in cell toxicity, transfection efficiency and proliferation.
APP Promoter Constructs-pGL 3 -APPluc constructs were prepared into the pGL 3 -basic vector (Promega) containing a major late promoter sequence (pGL 3 -MLP) (14). The Ϫ1104/ϩ104 fragment of the Rhesus monkey APP promoter (GenBank accession number AF 067971) was provided by Dr. Lahiri (22). The Ϫ309/ϩ104 and the Ϫ201/ϩ104 fragments were PCR amplified, digested by KpnI and BglII restriction enzymes, and ligated into pGL 3 -MLP. The Ϫ309/ϩ104 luciferase vector was digested by XhoI and BglII restriction enzymes to obtain the Ϫ75/ϩ104 fragment. The ϩ54/ϩ74 luciferase reporter vector was obtained by inserting into pGL 3 -MLP a double stranded oligonucleotide corresponding to the ϩ54/ϩ74 region of the human APP promoter and flanked by two XhoI restriction sites.
Densitometric Analyses-Agarose gels or blots from three independent experiments were acquired by a CCD camera and saved as a resolution of 600 dpi for software analysis. PCR products and Western blot signals were quantified by two-dimension densitometric analysis using the OptiQuant ® software (Packard Instruments Inc.).
Statistical Analysis-Results are expressed as the mean Ϯ S.D. Statistical analyses were performed with StatView (Abacus, Berkeley, CA) by one-way variance analysis (ANOVA) followed by the Bonferroni-Dunn test or Student's t test.

Endogenous Increase of APP and A␤ Expression in TGF-␤1
Transgenic Mice-To understand the role of TGF-␤1 in A␤ deposition, the influence of TGF-␤1 on APP and A␤ expression was investigated in 6-month-old transgenic mice overexpressing TGF-␤1 (line T65) through the use of semiquantitative RT-PCR, Western blotting analysis, and A␤ ELISAs. Transgenic mice were generated by Prof. Lennart Mucke's laboratory similarly to the previously described low expressing GFAP-TGF-␤1 transgenic lines, T64 and T115 (16). These mice express large amounts of the transgene associated with enhanced expression of endogenous TGF-␤1 as determined either by RT-PCR (Fig. 1A) or by immunoblotting (Fig. 1B). In cortices of TGF-␤1 mice, we observed a significant up-regulation of the mRNAs encoding APP isoforms compared with wild-type mice ( Fig. 1, C and D). These data were confirmed after immunoblotting performed from the same brain extracts (Fig. 1E). When revealed with 22C11 antibody, membrane-bound APP proteins were significantly overexpressed in TGF-␤1 mice compared with wild-type mice (Fig. 1E). Because TGF-␤1 possesses an amyloidogenic role (8), we determined A␤ production by ELISAs in our T65 transgenic line and observed a significant increase of soluble A␤42 (7.41 Ϯ 0.48 in wild-type versus 13.07 Ϯ 1.88 pg/mg in transgenic; p ϭ 0.04) and A␤40 (32.91 Ϯ 1.36 in wild-type versus 59.12 Ϯ 3.57 pg/mg in transgenic; p ϭ 0.01) without modifying the A␤42:A␤40 ratio (0.225 Ϯ 0.06 in wild-type versus 0.221 Ϯ 0.11 in transgenic) (Fig. 1F).
TGF-␤1-dependent Transcription of APP in Astrocytes-To determine which cell type was involved in the increased expression of APP in TGF-␤1 mice, we exposed primary cultures of mouse cortical neurons or astrocytes to exogenous recombinant TGF-␤1. As reported previously (23), the three APP isoforms are expressed in astrocytes with a predominant expression of APP 770 and APP 751 , whereas cortical neurons abundantly express APP 695 mRNA but only small amounts of APP 770 and APP 751 mRNAs (Fig. 2, A and C). Although 1 ng/ml TGF-␤1 exposure failed to modify the expression pattern of APP mRNAs in cortical neurons (Fig. 2, A and B), TGF-␤1 markedly FIG. 1. TGF-␤1 transgenic mice display enhanced A␤ generation. A, expression of the GFAP-TGF-␤1 transgene, endogenous TGF-␤1 mRNA in the brain of 6-month-old mice overexpressing TGF-␤1 in astrocytes (T65, TGF-␤1) was determined by semiquantitative RT-PCR analysis. Experiments were performed in triplicate. ␤-Actin was used as a housekeeping gene. B, Western blotting analysis of endogenous TGF-␤1 in the brain of 6-month-old mice overexpressing TGF-␤1 in astrocytes (T65, TGF-␤1) revealed using the sc-146 antibody (n ϭ 3). The membrane was reprobed to determine the expression of actin to estimate the homogeneity of proteins loaded. C, expression of APP mRNAs in the brain of 6-month-old mice overexpressing TGF-␤1 in astrocytes (T65, TGF-␤1) was determined by semiquantitative RT-PCR analysis. Experiments were performed in triplicate. ␤-Actin was used as a housekeeping gene. D, densitometric quantification of brain relative expression of APP mRNA isoforms. Results are the mean Ϯ S.D. of experiments performed in triplicate. Asterisk, p Ͻ 0.01, Student's t test. E, Western blotting analysis of total APP proteins in the brain of 6-month-old mice overexpressing TGF-␤1 in astrocytes (T65, TGF-␤1) revealed using the 22C11 antibody (n ϭ 3). The membrane was reprobed to determine the expression of actin to estimate the homogeneity of proteins loaded. F, A␤ ELISAs in the brain of 6-month-old mice overexpressing TGF-␤1 in astrocytes (T65, TGF-␤1). Amounts of respective A␤ species are indicated in the left panel, and rationalized expressions are shown in the right panel. Results are the mean Ϯ S.E. of six independent experiments. Asterisk, p Ͻ 0.05, Student's t test). enhanced the expression of the mRNAs encoding the three isoforms of APP in astrocytes (Fig. 2, C and D).
Because it could be advanced that the up-regulation of APP observed after TGF-␤1 exposure could result from the secondary activation of astrocytes by proinflammatory cytokines such as interleukin-1␤ released by microglia cells, we have estimated the amount of microglia in our cultures of astrocytes by performing double immunocytochemical labeling against the GFAP, an astrocytic marker, and CD11b, a specific marker of microglial cells. Less than 1% of microglial contamination in our primary cultures of cortical astrocytes was evidenced (Fig. 2E).
TGF-␤1 elicits its biological effects through a heteromeric complex of transmembrane serine/threonine kinase receptors, cloned as type I and type II receptors. After receptor-dependent phosphorylation, Smad2 or Smad3 interacts with the common mediator Smad4 to form a heteromeric complex that translocates to the nucleus to initiate TGF-␤-dependent transcriptional activity. This complex binds DNA sequences termed Smad-binding elements that contain a minimal four-nucleotide domain AGAC also called "CAGA box" (24). Accordingly, we Results are the mean of three experiments. Asterisk, p Ͻ 0.01, Student's t test. C, expression of APP mRNA in cultured astrocytes was determined by semiquantitative RT-PCR analysis at the indicated time after treatment in the presence of 1 ng/ml TGF-␤1 (24 and 72 h) (n ϭ 3). ␤-Actin was used as a housekeeping gene. All extracts elicited similar levels for ␤-actin. D, densitometric quantification of the experiments presented in C 24 h after treatment. White bars represent sham wash control astrocytes, and dark bars represent TGF-␤1-treated astrocytes. Results are the mean of three experiments. Asterisk, p Ͻ 0.03, Student's t test. E, astrocytes were stained with an antibody raised against the astrocytic marker GFAP (green) and the microglial marker CD11b (red). Finally, cells were counterstained with 1 g/ml DAPI (blue). Overlaid images are presented in the bottom right panel. have transiently transfected astrocytes with expression vectors encoding constitutively activated versions of the TGF-␤ type I receptor (Alks) (25) or with an expression vector encoding the transcription factor Smad3 (Fig. 3). To validate the efficiency of transfection in our system, murine cultured astrocytes were transfected with the EGFP-C1 vector encoding EGFP using cationic lipids. 24 h later, astrocytes were immunolabeled for GFAP and counterstained with DAPI before determination of transfection efficiency (ϳ70%) (Fig. 3A). Cell viability did not differ from untransfected astrocytes as estimated by phase microscopic observation and lactate dehydrogenase activity (data not shown).
Expression of either Alk-5 or Alk-4 led to an enhanced transcription of all isoforms of APP, whereas Alks linked to other members of the TGF-␤ superfamily did not enhance APP transcription (Fig. 3, B and C). Indeed, transfecting astrocytes with Alk-3 unexpectedly lowered APP mRNA expression. Accordingly, Smad3 overexpression potentiated APP mRNA expression (Fig. 3, D and E). As expected, overexpression of Smad7, a physiological dominant negative Smad, prevented TGF-␤1-in-duced expression of APP in cultured astrocytes (Fig. 3, D and  E). Thus, TGF-␤1 signaling machinery controls APP transcription in cultured astrocytes by a Smad3-dependent mechanism.
A TGF-␤-responsive Element Mediates TGF-␤1-dependent Expression of APP-We determined whether the induction of the APP mRNAs by TGF-␤1 in astrocytes was the result of a stabilization of the transcripts. In the presence of actinomycin D, the relative amount of APP and ␤-actin mRNAs was reduced in a time-dependent manner. The administration of TGF-␤1 failed to modify the stability of APP transcripts (Fig. 4A). Considering the TGF-␤1-dependent overexpression of the three APP isoforms mRNAs, we investigated whether TGF-␤1 modulates the activity of the APP promoter. Thus, different regions (Ϫ309/ϩ104, Ϫ201/ϩ104, and Ϫ75/ϩ104) of the Rhesus monkey APP promoter were subcloned into the pGL 3 -basic luciferase reporter vector containing a MLP. When transfected in the mink lung epithelial cell line Mv1Lu, previously characterized as a TGF-␤-responsive cell line (26,27), these entire promoter constructs were activated by TGF-␤1 treatment or Smad3 transfection (Fig. 4B). By performing sequence alignments  12). Asterisk, p Ͻ 0.01, ANOVA followed by the Bonferroni-Dunn test. G, immunoblot analyses of full-length APP proteins associated with the cell monolayer of cultured murine cortical astrocytes treated (T) or not (C) in the presence of TGF-␤1 for 24 h were performed using the 22C11 antibody. To determine the involvement of endogenous TGF-␤1 in controlling APP expression, parallel experiments were done in the presence of soluble TGF-␤ type II receptor (sT␤RII) as a TGF-␤ antagonist (n ϭ 2). (Fig. 4C), we noticed that a conserved sequence, located between ϩ56 and ϩ71 within the 5Ј-untranslated region, was retrieved in the cloned APP promoter from mouse, rat, Rhesus monkey, and man (5Ј-CGGGAGACGGCGGCGG-3Ј). Although the empty vector was nonresponsive to TGF-␤1 treatment or Smad3 transfection, it became sensitive to TGF-␤1 and Smad3 when five copies of the ϩ54/ϩ74 sequence of the human APP promoter (named APP TGF-␤-responsive element or APPtre) were inserted (Fig. 5, A and B). Results were similar both in Mv1Lu cells (Fig. 5A) and in primary cultures of murine astrocytes (Fig. 5B). Transfection of cultured astrocytes with expression vectors encoding Alk-4 and Alk-5 resulted in a marked increase in luciferase activities confirming the necessity of TGF-␤ receptor activation to induce its response (Fig. 5C). Moreover, overexpression of Smad7 led to 90% inhibition of TGF-␤1-induced activation of the APPtre reporter construct (Fig. 5D). In MDA-MB468 cells deficient for endogenous Smad4 expression (28,29), TGF-␤1 or Smad3 had no effect, whereas transfection of Smad4 rescued TGF-␤1-induced activation of the APPtre reporter construct (Fig. 5E). These data demonstrates that a Smad3⅐Smad4 complex is necessary to mediate the TGF-␤1-induced transcription of APP.
Finally, we mutated the AGAC sequence required for Smad binding to the DNA into an ACAT sequence, previously reported to abolish Smad binding to the DNA (14), within the Ϫ204/ϩ104 gene promoter construct (Fig. 5F). This construct also provides evidences of the influence of this element in a similar context to the full-length APP promoter. Results obtained by Dual-Luciferase ® Reporter Assay System clearly indicate that this double site mutation abolishes the ability of this construct to respond to TGF-␤1 treatment and to Smad3 transfection. Thus, our findings demonstrate the absolute necessity of this sequence for TGF-␤1-induced transcriptional activity of the APP promoter.
To characterize further the involvement of TGF-␤1 in the control of APP expression in cultured astrocytes, we have performed experiments using a soluble TGF-␤ type II receptor as a TGF-␤ antagonist (30,31). Addition of the this receptor in the media of cultured astrocytes prevented TGF-␤1-induced APP overexpression. Moreover, it is interesting to note that treatment in the presence of the soluble TGF-␤ type II receptor also decreased the basal expression of APP, confirming the presence of endogenous TGF-␤1 in our model of cultured astrocytes as demonstrated previously (32). Thus, endogenous TGF-␤1 regulates APP basal expression and can enhance APP transcription.
To determine whether astrocytes and neurons were TGF-␤1responsive, we performed an electrophoretic mobility shift assay using astrocytic and neuronal nuclear extracts in an attempt to characterize the DNA binding activity on the TGF-␤responsive APP sequence containing the previously characterized TGF-␤-responsive AGAC sequence (14). Although TGF-␤1 failed to induce formation of complexes in cultured neurons, increased binding complexes were observed in TGF-␤1-treated astrocytes (Fig. 6, center lanes). To confirm the specificity of the assay, a 50-fold excess of unlabeled probe was added to nuclear extracts from TGF-␤1-treated cells, which totally prevented the formation of complexes (Fig. 6,  right lanes). These data provide evidence that primary cultured mature cortical neurons (14 days in vitro) are not capable of mediating Smad-dependent TGF-␤1 signaling in the APP promoter.
A␤ Production Is Enhanced by TGF-␤1 Signaling-As observed at the mRNA level, TGF-␤1 failed to modify the expression of either full-length or derivatives of APP in cortical neurons (Fig. 7). In contrast, immunoblotting performed from cell extracts of murine astrocytes showed that membrane-bound APP proteins were increased markedly in the presence of TGF-␤1 over 24 -72 h, whereas actin remained unchanged (Fig.  7, A and B). Using the R7 antiserum targeted against the KPI domain of KPI-APPs, i.e. APP 770 and APP 751 , we confirmed the overexpression of APP 770 and APP 751 at the membrane of astrocytes stimulated by TGF-␤1 (Fig. 7, A and B). Furthermore, we evidenced that the release of sAPP derivatives was increased in the media of TGF-␤1-treated astrocytes (Fig. 7C). Similarly, we showed that sAPP-␤ was increased in the conditioned media of astrocytes treated by TGF-␤1 (for 24 or 72 h) (Fig. 7C). Because ␤-secretase activity is necessary for A␤ production, incubation with TGF-␤1 for 24 or 72 h induced a marked secretion of the 4-kDa A␤ species in the bathing medium of cultured astrocytes (Fig. 7D). Moreover, our results revealed that the overexpression of Smad3 led to an accumulation of A␤ as obtained previously after TGF-␤1 treatment (Fig. 7E). In contrast, overexpression of Smad7 prevented the TGF-␤-induced accumulation of A␤ (Fig. 7E). Using FCA3340 and FCA3542 antisera raised against A␤40 or A␤42, we observed an accumulation of A␤40 and A␤42 under TGF-␤1 treatment, whereas A␤ was not be detected in control conditions (Fig. 7F). Because we were unable to detect basal secretion of astrocytic A␤ using immunoblotting (Fig. 7, D-F), we have finally estimated A␤ release by quantitative ELISAs and confirmed a ϳ2-fold elevation of A␤ loads (Fig. 7G). This 2-fold increase in A␤ secretion (ϫ1.9 for A␤42 and ϫ1.8 for A␤40) can be directly associated with the 2-fold up-regulation of astrocytic APP transcription. Nevertheless, we cannot exclude the possibility of indirect effects of TGF-␤1 on either the APP processing or the A␤ degradation. Taken overall, these experiments evidence that TGF-␤1 promotes A␤ generation in astrocytes.
TGF-␤1 Promotes A␤ in Cultured Human Astrocytes-Because rodents do not display amyloid deposits, we performed a FIG. 6. Direct binding of TGF-␤1-induced transcription factors to the APPtre sequence containing the AGAC. An electrophoretic mobility shift assay was performed using a 32 P-labeled probe containing the AGAC sequence and nuclear extract from neurons or astrocytes induced for 1h by TGF-␤1 or not induced. Bands corresponding to specific TGF-␤1-induced complexes are indicated. 50 molar excesses of nonradiolabeled oligonucleotide were added as competitor (n ϭ 3).
set of experiments in primary cultures of human astrocytes and neurons. As observed previously in murine cultures, TGF-␤1 treatment failed to influence either transcription or accumulation of APP derivatives in cultured human neurons (Fig. 8).
However, the addition of exogenous TGF-␤1 to human astrocytes induced an increased expression of APP (Fig. 8A) as well as the activation of the APPtre luciferase reporter vector (Fig.  8B). These transcriptional data were confirmed at the protein level. Although TGF-␤1 failed to influence APP expression in neurons, it increased the amounts of full-length APP and sAPP-␤ associated with either the plasma membrane or the conditioned media of cultured human astrocytes (Fig. 8, C and  D). Moreover, TGF-␤1 (Fig. 8E) or transfection with the Smad3 vector (Fig. 8F), respectively, led to a 1.5-2.5 increase in A␤ content in the conditioned media of human astrocytes. Results are the mean of three experiments. Asterisk, p Ͻ 0.01, Student's t test. N represents neurons, and A, astrocytes. C, immunoblot analyses of total sAPP and sAPP-␤ from conditioned media of cultured murine cortical neurons or astrocytes treated (T) or not (C) in the presence of TGF-␤1 for 24 h were performed using either the 22C11 antibody or the 192wt antiserum. The membrane was stained by naphthol blue to confirm the homogeneity of loaded amounts of proteins (data not shown). Experiments were performed in triplicate. D, conditioned media of astrocytes treated in the presence of 1 ng/ml TGF-␤1 for 72 h were concentrated as described under "Experimental Procedures," and Western blot analyses were performed using the R600 polyclonal antiserum raised against the N terminus (residues 1-10) of the A␤ peptide. Experiments were performed in triplicate. E, Western analysis of A␤ were performed using the R600 antiserum from concentrated conditioned media of primary cultures of cortical astrocytes previously transfected either with the expression vector encoding Smad3 or the expression vector encoding Smad7 or with the empty vector in the presence or not of TGF-␤1 for 72 h. Asterisks indicate nonspecific labeling. F, to discriminate A␤ species, Western analyses of A␤ were performed from concentrated conditioned media of astrocytes incubated with 1 ng/ml TGF-␤1 for 72 h and probed with a polyclonal antiserum raised against A␤ 1-40 , FCA3340, or A␤ 1-42 , FCA3542. Asterisks indicate nonspecific labeling. The experiments were performed in triplicate. G, in parallel, A␤ ELISAs were realized from the same extracts used for immunoblotting experiments. White bars correspond to untreated astrocytes and dark bars to TGF-␤1 treated astrocytes. Asterisk, p Ͻ 0.001, Student's t test. (Fig. 6G), whereas mice cultured astrocytes did not display any difference, suggesting slightly different mechanisms in A␤ generation. DISCUSSION A␤ accumulation into amyloid plaques in the brain is one of the two histopathological hallmarks of AD. The process that regulates the deposition of A␤ in the brain is still under investigation. A better understanding of the mechanism leading to A␤ production would facilitate the development of treatments for AD. We document here that the overexpression of the antiinflammatory cytokine TGF-␤1 in transgenic mice induces higher expression of endogenous APP isoforms and increased A␤ generation in cerebral tissues. Furthermore, we demonstrate that exogenous TGF-␤1 enhances APP synthesis in astrocytes and leads to A␤ generation in vitro.
Although neurons are known to be the major source of A␤ in AD (33)(34)(35), the contribution of astrocytes to amyloidogenic processes has never been clearly established. Studies have suggested that cultured astrocytes could generate modest amounts of A␤ compared with neurons (36 -39). Although TGF-␤1 induced the overexpression of APP in astrocytes by involving a TGF-␤-responsive element, we observed no effect of TGF-␤1 in primary cultures of mature neurons (14 days in vitro). This absence of a TGF-␤1 response is puzzling but has been described previously (40). In addition, contradictory data report the neuronal expression of type II TGF-␤ receptor (32,41), which binds the ligand and then activates the TGF-␤ signaling intracellular pathway. Finally, these data are in agreement with a previous study (42), demonstrating that although TGF-␤ induced transcription of plasminogen activator inhibitor-1 in cultured astrocytes, it failed to mediate this response in cultured neurons. To understand better why TGF-␤1 did not activate APP transcription in mature cultured neurons (14 days in vitro), we have performed an electrophoretic mobility shift assay and observed that TGF-␤1 is unable to activate the Smad pathway in mature neurons.
However, an astrocytic gliosis is always observed in brains of FIG. 8. TGF-␤1 promotes A␤ generation in human cultured astrocytes. A, expression of APP mRNAs in primary cultures of human astrocytes was determined by RT-PCR after treatment in the presence of 1 ng/ml TGF-␤1 for 24 h (n ϭ 3). ␤-Actin was used as a housekeeping gene. All extracts elicited similar levels of ␤-actin. B, mean Ϯ S.D. of the luciferase activity of cultured human astrocytes transiently transfected with a luciferase reporter construct containing five copies of the ϩ54/ ϩ74 sequence of the human APP promoter inserted 5Ј of the MLP luciferase reporter gene. For each condition, transfected cells were treated (T) or not (C) in the presence of TGF-␤1 (n ϭ 12). Asterisk, p Ͻ 0.01, ANOVA followed by the Bonferroni-Dunn test. C, immunoblot analysis of full-length APP proteins associated with the cell monolayer of cultured human cortical neurons or astrocytes treated (T) or not (C) in the presence of TGF-␤1 for 24 h using the R7 antibody. The same immunoblot was reprobed with actin antibody as control. Experiments were performed in triplicate. The histogram illustrates densitometric quantification of the experiments presented in C 24 h after treatment. White bars represent sham wash control cells, and dark bars represent TGF-␤1-treated cells. Results are the mean of three experiments. Asterisk, p Ͻ 0.01, Student's t test. D, immunoblot analysis of sAPP-␤ was performed from the conditioned media of cultured human cortical neurons or astrocytes treated (T) or not (C) in the presence of TGF-␤1 for 24 h using the 192wt antiserum. The histogram illustrates densitometric quantification of the experiments presented in D 24 h after treatment. White bars represent sham wash control cells, and dark bars represent TGF-␤1-treated cells. Results are the mean of three experiments. Asterisk, p Ͻ 0.01, Student's t test. E, immunoblot analyses of A␤ were performed using the R600 antiserum from the conditioned media of cultured human cortical neurons or astrocytes treated (T) or not (C) in the presence of TGF-␤1 for 24 h. Experiments were performed in triplicate. Densitometry analysis of A␤ production in the media of cultured human cortical neurons or astrocytes treated (black bars) or not (white bars) in the presence of TGF-␤1 for 72 h was performed. Results are the mean of three independent experiments. Asterisk, p Ͻ 0.03, Student's t test. F, Western analyses of A␤ were performed using the R600 antiserum from concentrated conditioned media of primary cultures of cortical astrocytes previously transfected with the expression vector encoding Smad3 or with the empty vector in the presence or not of TGF-␤1 for 72 h. G, A␤ ELISAs were performed from human astrocytes treated with TGF-␤1 for 24 h. The amounts of respective A␤ species are indicated in the left panel, and rationalized expressions are shown in the right panel. Results are the mean Ϯ S.E. of eight independent experiments. Asterisk, p Ͻ 0.05, Student's t test.
patients with AD (43). These data led us to reconsider the participation of astrocytes in the amyloidogenic process. Indeed, recent work performed from transgenic mice (Tg2576) exhibiting amyloid plaques evidenced that astrocyte-derived A␤ participate in plaque formation and maturation at later stages than neuronal A␤ (44). Here, we demonstrate that TGF-␤1 potentiates A␤ production in astrocytes.
Our data are in agreement with previous findings demonstrating that the expression of TGF-␤1 in the brain parenchyma induces cerebrovascular and meningeal A␤ deposition at 12-18 months of age in TGF-␤1 transgenic mice (line T64 or T115) or at 2-3 months of age in hAPP/TGF-␤1 biogenic mice expressing the human APP and TGF-␤1 (line T64 or T115) (8). In addition to its proamyloidogenic effect, TGF-␤1 may exert a more complex role because 12-15-month-old hAPP/TGF-␤1 double transgenic mice (10) displayed plaque burden reduction associated with increased microglia activation and increased clearance of A␤ compared with hAPP mice. In the present study, we report that 6-month-old transgenic mice overexpressing TGF-␤1 (line T65) display increased endogenous APP expression and A␤ production in vivo. Moreover, we postulate that this effect is sustained by a transcriptional activation of APP in both murine and human astrocytes. Although AD cannot be equated with a simple unique gene deregulation, this is, to our knowledge, the only transgenic animal overexpressing a brain-derived cytokine that promotes A␤ deposition in the brain. Nevertheless, we cannot exclude the possibility of indirect effects of TGF-␤1 on either the APP processing or the A␤ degradation. These data are strengthened by several publications. First, TGF-␤ immunoreactivity has been found within the plaques of AD (5). Second, increased TGF-␤2 levels in reactive astrocytes are associated with AD (7). Third, aged transgenic mice containing the Swedish double mutation of APP 695 display TGF-␤1 immunoreactive astrocytes found in close proximity to A␤ deposits (44). We demonstrate that TGF-␤1-induced APP overexpression leads to an enhanced A␤ production. Similarly, the increase in APP expression because of duplication of the 21 chromosome in Down's syndrome results in an overproduction of A␤ peptides leading to the appearance of AD-type brain lesions with associated microgliosis and astrogliosis (45).
In vitro, an overexpression of APP mRNAs induced by TGF-␤1 has been reported previously in astrocytes or in astrocytoma cell line (46,47). Amara et al. (47) have described an increased half-life of the APP transcripts induced by TGF-␤1. By performing mRNA decay experiments after 24 h of TGF-␤1 treatment, we demonstrated that in our hands the half-life of APP mRNA was not stabilized by a TGF-␤1 treatment. In contrast, we evidenced that TGF-␤-induced up-regulation of APP mRNA expression involved the activation of a TGF-␤ϭresponsive element within the ϩ54/ϩ74 region of the APP promoter. Additional data provided evidences that the 5Ј-untranslated region of the APP promoter is crucial for driving APP expression. Indeed, the Ϫ75/ϩ104 region has been shown to mediate up to 40% of the promoter activity compared with the full-length promoter-driven activity (Ϫ7900/ϩ104) (48).
Overall, these data reveal a molecular mechanism through which TGF-␤1 promotes A␤ generation and underline the critical role that astrocytes could hold in AD pathogenesis.