The Gtx homeodomain transcription factor exerts neuroprotection using its homeodomain.

Certain cases of familial Alzheimer's disease are caused by mutants of amyloid-beta precursor protein (AbetaPP), including V642I-AbetaPP, K595N/M596L-AbetaPP (NL-AbetaPP), A617G-AbetaPP, and L648P-AbetaPP. By using an unbiased functional screening with transfection and expression of a human brain cDNA library, we searched for genes that protect neuronal cells from toxicity by V642I-AbetaPP. One protective clone was identical to the human GTX, a neuronal homeobox gene. Human Gtx (hGtx) inhibited caspase inhibitor-sensitive neuronal cell death not only by V642I-AbetaPP but also by L648P-, NL-, A617G-AbetaPP, apolipoprotein E4, and Abeta. The region of hGtx responsible for this rescue function was specified to be its homeodomain (Lys148-His207). The rescue function was shared by DLX4, a distal-less family gene with a homeodomain only 38.3% homologous to that of hGtx, suggesting that this function would be generally shared by homeodomains. The neuroprotective function of hGtx was attributable to hGtx-stimulated production and secretion of insulin-like growth factor-I. This study provides molecular clues to understand how neuronal cells developmentally regulate themselves against cell death as well as to develop reagents effective in curative therapeutics of Alzheimer's disease.

GTX, a neuronal homeobox gene, has been isolated from the postnatal mouse brain (1) and is expressed in differentiated oligodendrocytes with its potential role for myelinization (2,3). It is also termed as Nkx6.2. Gtx is also expressed in neural progenitors in ventral spinal cord and inhibits V0 neurons from being generated from the progenitors (4). Human GTX has also been cloned and mapped to chromosome 10q26, a region where frequent loss of heterozygosity has been observed in various malignant brain tumors (5). Mice lacking Gtx normally develop without apparent changes in the central nervous systems (6). Although information about the function of GTX is thus limited, it is accepted that homeobox genes encode regulatory proteins relevant to nuclear transcription, and that their coordinated activation is fundamental for temporal and spatial pattern formation in embryonic development. Molecular con-trol of differentiation depends on lineage-restricted transcription factors that regulate expression of tissue-specific genes. Among the genes governing the development of Drosophila, sequential activation of homeotic and segmentation genes controls the identity, polarity, and number of segments. Many such genes, including the Antennapedia (Antp), Engrailed (En), and Paired (Prd) families, contain a characteristic ϳ60-amino acid sequence motif, termed a homeobox (7). Consistent with the notion that these genes act as sequence-specific transcription factors, Gtx binds to (A/T)TTAATGA as an optimal binding sequence and represses the transcription of the downstream genes (8).
Alzheimer's disease (AD) 1 is the most prevalent neurodegenerative disease, for which a curative therapy has not yet been developed. Three kinds of known mutant genes cause early onset familial AD (FAD): mutants of amyloid-␤ precursor protein (A␤PP), presenilin (PS)1, and PS2, each of which has been shown to cause neuronal cell death (9 -15). Investigations using an ecdysone (EcD)-inducible system (16) revealed that four different FAD mutants (V642I-A␤PP, K595N/M596L-A␤PP (NL-A␤PP), M146L-PS1, and N141I-PS2) trigger different combinations of distinct toxic mechanisms in an expression-dependent manner (13)(14)(15). Considering that these FAD mutants cause only a small fraction of total AD cases, such mechanistic complexity suggests that it may be difficult to identify a single anti-AD neuroprotective reagent solely based on the underlying mechanisms.
We thus took advantage of a new approach, called "death trap" screening developed by D'Adamio et al. (17), as an unbiased functional screening of molecules that allow dying cells to survive. We applied this method to V642I A␤PP-inducible neuronal cells (18) using an expression cDNA library constructed from an occipital lobe of an AD patient brain. Here we report that one of the obtained genes was human GTX (hGtx), which has inhibitory activity against neuronal cell death by AD-relevant insults through its homeodomain.

MATERIALS AND METHODS
Cell and Genes-V642I-A␤PP, NL-A␤PP, A617G-A␤PP, L648P-A␤PP, M146L-PS1, and N141I-PS2 cDNAs, subcloned in pcDNA or pIND plasmids, were described previously (9, 13, 19 -22). G85R superoxide dismutase 1 (G85R-SOD1) cDNA in pEF-BOS plasmid, kindly provided by Drs. Takashi Koide and Shoji Tsuji, was described previously (20). F11 cells, the hybrid cells of a rat embryonic primary neuron and a mouse neuroblastoma (23), were grown in Ham's F-12 (Invitrogen) plus 18% fetal bovine serum (FBS; Hyclone, Logan, UT) and . The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
§ Both authors contributed equally to this work. ʈ To whom correspondence may be addressed. antibiotics. F11/EcR cells stably overexpressing both the EcD receptor (EcR) and the retinoid X receptor were established using the co-expression vector pVgRXR and Zeocin selection (Invitrogen), as described previously (13). Full-length DLX4 cDNA was obtained from Invitrogen.
Transfection Protocols (F11 Cells)-For transient transfection of F11 cells with pcDNA, pEF-BOS, or pEF4/His plasmids, F11 cells were seeded at 7 ϫ 10 4 cells/well in a 6-well plate and cultured in Ham's F-12 medium supplemented with 18% FBS for 12-16 h, and were transfected with the plasmids (0.75 g of pcDNA plasmid, 0.75 g of pEF plasmid, 3 l of LipofectAMINE, and 6 l of LipofectAMINE PLUS for each well; Invitrogen) in the absence of serum for 3 h. After subsequent 2-h incubations with Ham's F-12 plus 18% FBS, cells were cultured with Ham's F-12 plus 10% FBS. Cell mortality was measured by trypan blue exclusion at 48 h after the onset of transfection (1.5 g of protocol). Some experiments were performed with 0.5 g of pcDNA, 0.5 g of pEF4-hGtx, 2 l of LipofectAMINE, and 4 l of LipofectAMINE PLUS and cell mortality measurement 72 h after transfection (0.5-g protocol). In multiple studies (13-15, 20, 22, 24), it has been indicated that the 0.5-g protocol with F11 cells mimics expression in the F11/EcR cell system of the coding protein with 1 g of pIND transfection and 10 M EcD treatment; and the 1.5-g protocol with F11 cells mimics expression of the coding protein with 1 g of pIND transfection and 40 M EcD treatment in F11/EcR cells.
Death Trap Screening-F11/EcR/V642I cells, in which V642I-A␤PP is overexpressed by treatment with EcD, were described previously (18). By using these cells, in which death occurred in a vast majority of EcD-treated cells for a few days (18), we performed the death trap screening with our modifications. We first transfected F11/EcR/V642I cells with a mammalian expression cDNA library in pEF-BOS constructed from an occipital cortex of a necropsy-diagnosed AD patient, treated cells with EcD for 72 h, and recovered the plasmids from surviving cells. The recovered plasmids were again transfected to F11/ EcR/V642I cells, which were then treated with EcD for 72 h. This procedure was repeated for three rounds, and we randomly picked 250 clones of plasmids. Using dot blot hybridization with randomly selected clones, we classified them into 36 independent groups.
Subcloning of hGtx and Constructions of Deletion Mutants-To construct the tagged hGtx open reading frame (ORF) cDNA, we digested the DT236 by EcoRI and XbaI; the EcoRI site is 39 nucleotides upstream from the translation start codon ATG, and the XbaI site is in pEF-BOS vector, downstream from DT236 ORF. We subcloned the entire ORF of hGtx into pEF4-hGtx, with subsequent confirmation of the sequences. To construct pEF4-hGtx, pEF4 hGtxNT, and pEF4-hGtxCT, we digested DT236 and obtained two fragments corresponding residues at position 1-94 (plus 5Ј-untranslated upstream region of hGtx) and residues at position 95-277 (plus 5Ј-untranslated upstream region of hGtx), and subcloned each fragment into pEF4/His. To construct pEF4-hGtx⌬P, we first performed PCR (using Pfu turbo-polymerase from Stratagene, La Jolla, CA) with the following primers: the sense primer including the EcoRI site, 39 nucleotides upstream from the start codon of hGtx, was 5Ј-TTTTTTGTGTGGTGGAATTCCGGAT-GGGAG-3Ј; the antisense primer for hGtx⌬P was 5Ј-TTTTCTAGAT-CAGTGCCGCTTGCGCCACTTG-3Ј. The sense primer for pEF4-hGtxH was 5Ј-TTTTTTGAATTCAAGAAGCACTCGCGCCCGACCTTC-3Ј, and the antisense primer was the same as the antisense primer for hGtx⌬P. Each amplified fragment was subcloned into pEF4/His at the EcoRI site and the XbaI site. The sequences of deletion mutants were confirmed, and expression of each mutant was detected by anti-(His) 6 antibody (Santa Cruz Biotechnology, Santa Cruz, CA).
DNA Sequence and Homology Search-Sequencing of the cloned cDNA fragments was performed using an automated sequencer (ABI-PRISM310 Genetic Analyzer, Applied Biosystems, PerkinElmer Life Sciences). Analysis of DNA homology was performed using BLAST (National Center for Biotechnology Information).
Cytotoxicity and Viability Assays-Cytotoxicity was assessed by trypan blue exclusion assay, as described previously (13-15, 20 -22, 25). In brief, cells were suspended by pipetting gently, and 50 l of 0.4% trypan blue solution (Sigma) was mixed with 200 l of cell suspension (final trypan blue concentration was 0.08%) at room temperature.
Stained cells were counted within 3 min after being mixed with the trypan blue solution. The mortality of cells was then determined as a percentage of trypan blue-stained cells in total cells. The cell mortality thus assessed represents the population of dead cells in total cells, including both adhesive and floating cells, at the termination of experiments. Cell viability assay (Cell Counting Kit-8, Wako Pure Chemical Industries, Osaka, Japan) was performed using WST-8 ([2-(2-methoxy-4 nitrophenyl)-3-(4-nitrophenyl)-5-(2,4-disulfophenyl)-2H-tetrazolium, monosodium salt), as described previously (25). It has been demonstrated that in the F11 systems, the cell mortality assessed by trypan blue exclusion assay correlates reciprocally with the cell viability assessed by WST-8 assay (25). ␣IR3 was purchased from Oncogene Science (Cambridge, MA), and anti-IGF-I-specific neutralizing antibody was from R & D systems (Minneapolis, MN).
Statistical Analysis-All experiments described in this study were repeated at least three times with independent transfections. Statistical analysis was performed with one-way analysis of variance followed by post hoc test, in which p Ͻ 0.05 was assessed as significant.

Disease-based Death Trap
Screening-F11/EcR/V642I cells, in which V642I-A␤PP was inducibly overexpressed by EcD treatment, were described previously (18). By using these cells, in which death occurred in a vast majority of EcD-treated cells within a few days (18), we performed the death trap screening with a unique modification, termed disease-based death trap screening. In the original screening by Vito et al. (26), a normal tissue cDNA library was transfected to Jurkat cells, which were killed by T cell receptor antibody; library fragments were then recovered from surviving cells. In contrast, our approach was unique in that we employed an expression cDNA library constructed from an occipital lobe of an autopsy-diagnosed AD patient brain, as we reasoned that neuroprotective genes must be expressed in an occipital lobe of the AD brain. We first transfected F11/EcR/V642I cells with AD-1 library (human AD occipital lobe cDNA library in pEF-BOS), treated cells with EcD for 72 h, and recovered the plasmids from surviving cells. This procedure was repeated for three rounds, and we randomly picked 250 clones from the recovered plasmids. We classified all 250 clones into 36 independent groups by dot blot hybridization. One such gene was DT236.
By using a simple transient transfection system with F11 neurohybrid cells (9), we confirmed the antagonistic activity of DT236 against neuronal cell death by V642I-A␤PP (Table I). When F11 cells were transfected with V642I-A␤PP cDNA in pcDNA, 50 -60% of cells underwent death by 72 h. As the transfection efficiency was constantly 65-70% under the employed conditions (13), it followed that most cells expressing this FAD gene were dead 72 h after the onset of transfection. When F11 cells were co-transfected with DT236, cell death by V642I A␤PP was significantly suppressed. This result suggests that DT236 does inhibit neuronal cell death by V642I-A␤PP.
The DT236 cDNA sequence had an ORF encoding a polypeptide consisting of 277 amino acids. This ORF sequence turned out to be identical to human GTX (GenBank TM accession number AAK13251) and showed an extreme homology to murine Gtx (mGtx, GenBank TM accession number L08074). Comparing human and murine GTX genes, the homology was 88.6% in the nucleotide level and 97.5% in the amino acid level. The hGtx homeodomain exhibited high similarity to that of Nkx6.1 (27), indicating that Nkx6.1 is a close relative of Gtx (Fig. 1). Gen-erally, the hGtx homeodomain also showed a limited homology (30 -45%) to most homeodomain sequences (Fig. 1). The homology of the homeodomain of hGTX with that of Nkx6.1 or DLX4, a distal-less X4 gene (28,29), was 97 or 38.3%, respectively.
The C-terminal portion of hGtx showed structural similarity to the helix-turn-helix DNA-binding domain of some regulatory proteins, and the N-terminal arm and the third helix of the homeodomain interact specifically with bases in the minor and major grooves of DNA, respectively (30 -32). This information suggests that hGtx functions as a nuclear transcription factor, not only as a homeobox protein but also as other types of DNA-binding proteins.
hGtx Inhibits Neuronal Cell Death by Overexpression of V642I-A␤PP-We constructed pEF4-hGtx by inserting the ORF of hGtx into the polycloning site in pEF4/His. We examined the effect of pEF4-hGtx on death of F11 neurohybrid cells caused by either V642I-A␤PP or NL-A␤PP in pcDNA using the 1.5-g protocol ( Fig. 2A). Neuronal cell death by V642I-A␤PP was significantly suppressed by pEF4-hGtx but not by the empty pEF vector ( Fig. 2A). Co-transfection of pEF4-hGtx did not critically affect expression of V642I-A␤PP, as compared with the case that co-transfected with empty pEF4/His ( Fig.  2A, inset). Although cytomegalovirus (CMV) promoter-driven expression of V642I-A␤PP was inhibited by hGtx to 70 -80% of its expression with pEF4/His co-transfection, the attenuated level of V642I-A␤PP expression was still 6 -7 times higher than the expression level of endogenous A␤PP. Thus, it was highly likely that neurotoxicity inhibition by hGtx was not due to the inhibition of V642I-A␤PP expression. In further support, neu-ronal cell death by NL-A␤PP was not suppressed by hGtx when hGtx similarly inhibited the expression level of NL-A␤PP (see below).
hGtx Inhibits Neuronal Cell Death by Antibody-To confirm that the cytoprotective effect of hGtx was not due to hGtxmediated inhibition of V642I-A␤PP expression, we examined whether hGtx can also inhibit F11 neurohybrid cell death caused by extracellularly added A␤1-43. A␤ is the major constituent of senile plaques, extracellular deposits characteristic for AD, and has been implicated in the pathogenesis of AD. Whereas F11 cells per se were not killed by A␤1-43, as reported previously (9,19), cell death was induced by A␤1-43 when F11 cells were transfected with the p75 neurotrophin receptor p75 NTR (Table II). This is in agreement with the notion that p75 NTR mediates A␤ neurotoxicity, as has been shown by multiple studies (33)(34)(35). When p75 NTR -transfected F11 cells were co-transfected with hGtx, cell death by 25 M A␤1-43 was drastically suppressed (Table II). Expression of p75 NTR was not inhibited by co-transfection of hGtx (data not shown). hGtx could thus inhibit A␤ neurotoxicity in F11 cells, suggesting that neuroprotection by hGtx would not be due to the inhibition of FAD gene expression.
hGtx Inhibits Neuronal Cell Death by Low Expression of NL-A␤PP-With the 1.5-g protocol used here, 0.75 g of FAD gene (in pcDNA) with 0.75 g of pEF4-hGtx (in pEF4/His) was transfected to F11 neurohybrid cells, and cell mortality was measured 48 h after transfection. In the pIND system, 1 g of FAD gene (in pIND) with 1 g of pEF4-hGtx (in pEF4/His) was transfected to F11/EcR cells; cells were then treated with EcD;  The lower sequences indicate similarities of the hGtx homeodomain to other homeodomains. The residue identical to that of hGtx is indicated by a dash. The percent identity with hGtx is given at the right. The consensus (CONS.) is derived from known homeoboxes (30). Symbols at the top stand for the putative interaction sites with DNA bases (closed circles) and with the DNA sugar phosphate backbone (open circles), based on the crystal structure of the DNA homeodomain complex of En and MATa2 (31,32). a.a., amino acids. The boundaries of three helices are also assumed, based on the reported crystal structure of the homeodomain. and cell mortality was measured 72 h after the onset of transfection. The neurotoxicity by V642I-A␤PP and NL-A␤PP in pcDNA transfected by the 1.5-g protocol was quantitatively comparable with neurotoxicity by these genes in pIND trans-fected by the pIND protocol (1 g of pIND plasmid transfection and 40 M EcD treatment) in the pIND system, as shown in multiple studies (13-15, 20 -22, 24). In this pIND system, high induction of NL-A␤PP causes DEVD-resistant non-apoptotic FIG. 2. Effect of pEF4-hGtx on neuronal cell death by V642I-A␤PP or NL-A␤PP. A, effect of pEF4-hGtx on F11 cell death by transfection with V642I-A␤PP or NL-A␤PP cDNA by the 1.5-g protocol. F11 cells were transfected without (pcDNA) or with either V642I-A␤PP or NL-A␤PP (each in pcDNA), with pEF4-hGtx (hGtx) or empty pEF4/His (pEF/His). Cell mortality was measured by trypan blue exclusion assay 48 h after transfection. No T in the figures in this study indicates the cases without transfection. All values in this study indicate the means Ϯ S.D. of at least three independent experiments. Inset, effect of pEF4-hGtx on expression of V642I-A␤PP or NL-A␤PP. Using the same 1.5-g protocol, F11 cells were transfected with either V642I-A␤PP or NL-A␤PP (each in pcDNA) with pEF4-hGtx (hGtx) or empty pEF4/His (pEF/His). Cellular expression of A␤PP was measured by immunoblot analysis 48 h after transfection. Arrowhead indicates the holoproteins of expressed A␤PPs. n.s., not significant. B and C, effect of pEF4-hGtx on F11/EcR cell death by EcD-induced expression of V642I-A␤PP (B) or NL-A␤PP (C) in the pIND system. F11/EcR cells were transfected with pIND-encoded V642I-A␤PP or NL-A␤PP with pEF4-hGtx or empty pEF4/His. Nearly 1 day after transfection, cells were treated with or without increasing concentrations of EcD. Because transfection was performed nearly 1 day before the onset of EcD treatment, EcD treatment could not affect transfection efficiency. In all experiments in this study, transfection efficiency was constantly 65-70%, as described previously (13). pIND is an EcD-inducible plasmid. Cell mortality was measured by trypan blue exclusion 72 h after EcD treatment. cell death different from DEVD-sensitive apoptotic cell death by its low induction, whereas low to high induction of V642I-A␤PP solely causes DEVD-sensitive cell death (13). Given that hGtx inhibited DEVD-sensitive cell death by V642I-A␤PP, it was possible that hGtx could also inhibit neuronal cell death by low expression of NL-A␤PP.
We thus examined the effect of hGtx on neuronal cell death by low expression of NL-A␤PP using the pIND system, in which cell death by low and high induction is best analyzed. In this system, F11/EcR cells were transfected with NL-A␤PP cDNA in EcD-inducible pIND plasmid with or without co-transfection of pEF4-hGtx and treated with various concentrations of EcD. Hashimoto et al. (13)(14)(15) demonstrated the following: (i) enhanced green fluorescent protein cDNAs in pIND express the enhanced green fluorescent protein in a manner proportional to EcD doses; (ii) wild-type (wt)A␤PP is expressed in a relationship linear to treated EcD doses; and (iii) V642I-A␤PP or NL-A␤PP in the presence of DEVD or glutathione-ethyl-ester (GEE) is expressed in almost the same linear EcD dependence as that observed for wtA␤PP. In F11/EcR cells transfected with pIND-wtA␤PP cDNA, 10 and 40 M EcD caused ϳ1.5 and ϳ6-fold expression of transfected wtA␤PP relative to the expression of endogenous A␤PP (total amounts of A␤PP were ϳ2.5and ϳ7-fold that of endogenous A␤PP, respectively), as described previously (13). Expression of V642I-A␤PP or NL-A␤PP in the absence of DEVD reached saturation at 10 M EcD, and its expression at Ͼ10 M EcD stayed at the level of expression observed at 10 M EcD. This feature of the expression profile is due to the feedback loop comprised of mutant A␤PP-induced activation of DEVD caspases and the DEVDcaspase-induced proteolysis of the expressed A␤PP holoprotein (36), as discussed in detail previously (13). The vehicle of EcD was ethanol, which provided no induction of A␤PP proteins. Hence, the observed effects of treated EcD were attributed to the expressed A␤PP proteins. In this pIND system, NL-A␤PP neurotoxicity elicited by 10 M EcD was almost fully suppressed by co-expression of hGtx (Fig. 2B). In contrast, NL-A␤PP neurotoxicity elicited by 40 M EcD was not affected by hGtx. The curve of NL-A␤PP neurotoxicity in the presence of 100 M Z-VAD-fmk virtually overlapped its curve in the presence of co-expression of hGtx (Fig.  2B). Co-expression of hGtx did not inhibit expression of NL-A␤PP encoded in pIND (Fig. 2B, inset). As noted above and reported previously (13), final expression of NL-A␤PP reached saturation at 10 M EcD, and its expression at Ͼ10 M EcD stayed at the level of expression observed at 10 M EcD. In contrast, expression of NL-A␤PP almost linearly increased in the presence of co-expressed hGtx. This is consistent with the previous report (13) that expression of NL-A␤PP linearly increases when intracellular caspase activity is suppressed.
These data indicate that hGtx exerts an effect similar to that of Z-VAD-fmk and inhibits apoptotic component of NL-A␤PP neurotoxicity.
We also examined whether and how hGtx inhibits V642I-A␤PP neurotoxicity by low (10 M) or higher (40 M) EcD induction (Fig. 2C). As expected from pcDNA data, co-expres-sion of hGtx inhibited V642I-A␤PP neurotoxicity elicited by both 10 and 40 M EcD (Fig. 2C). The curve of V642I-A␤PP neurotoxicity in the presence of co-expression of hGtx became virtually identical to the V642I-A␤PP neurotoxicity curve in the presence of 100 M Z-VAD-fmk. As was the case with expression of NL-A␤PP, expression of V642I-A␤PP was not attenuated by co-expression of hGtx in the pIND system (Fig.  2C, inset). Although hGtx appeared to exert a slight but significant negative effect on the expression of the constructs in pcDNA, a plasmid driven by a CMV promoter, hGtx, had no inhibitory effect on the expression of the pIND constructs driven by the heat shock protein (HSP) promoter. It is therefore reasonable to assume that EcR expression potentially inhibited by expressed hGtx (EcR of pVgRXR is driven by the CMV promoter) was sufficient to mediate the EcD effect on the expression of the pIND constructs. These data indicate, again, that hGtx suppresses caspase inhibitor-sensitive apoptotic neuronal cell death by V642I-A␤PP, which concurs with the notion that hGtx effectively suppresses the apoptotic component of neurotoxicity by FAD mutants of A␤PP.
hGtx Inhibits Cell Death by L648P-A␤PP and Low Expression of A617G-A␤PP-A617G-A␤PP is known as a Flemish mutant of FAD (37) and L648P-A␤PP as an Australian mutant (38), both of which have been shown to cause neuronal cell death in vitro (22,38). By using the 1.5-g protocol allowing for overexpression of FAD genes, we examined the effect of hGtx on neuronal cell death by A617G-A␤PP and by L648P-A␤PP. Fig. 3A shows that overexpression of either FAD mutant caused neuronal cell death as efficaciously as V642I-A␤PP and NL-A␤PP. hGtx inhibited neurotoxicity by L648P-A␤PP to the level of wtA␤PP-induced toxicity or to the level of V642I-A␤PPinduced toxicity in the presence of hGtx co-expression (Fig. 3A). In contrast, hGtx did not at all inhibit neuronal cell death by overexpressed A617G-A␤PP. Under the same condition, hGtx did not critically affect expression of A617G-or L648P-A␤PP (Fig. 3A, inset). These data suggest that hGtx inhibits neurotoxicity by overexpressed L648P-A␤PP, but not by overexpressed A617G-A␤PP, and that these two FAD mutants cause cell death through different mechanisms, as is the case with V642I-A␤PP and NL-A␤PP (13).
It should also be noted that in the pIND system, low induction of NL-A␤PP caused DEVD/hGtx-sensitive neuronal cell death, whereas high induction of NL-A␤PP additionally caused DEVD/hGtx-resistant cell death (Fig. 2B). We thus reasoned that there is a possibility that neuronal cell death by low expression of A617G-A␤PP and L648P-A␤PP is inhibited by hGtx. We examined this possibility by using the 0.5-g protocol, which allows for low expression of the FAD genes. Transfection of F11 cells with wtA␤PP cDNA in pcDNA by this protocol resulted in 1-1.5 times higher expression of wtA␤PP than expression of endogenous A␤PP (Fig. 3A, inset; the total A␤PP expression resulted in 2-2.5 times higher than that of endogenous A␤PP), which was equivalent to the expression of wtA␤PP in F11/EcR cells transfected with wtA␤PP cDNA (in pIND) by the pIND protocol and treated with 10 M EcD for 72 h. Both A617G-A␤PP and L648P-A␤PP, transfected by the  3. Effect of pEF4-hGtx on neuronal cytotoxicity by A617G-or L648P-A␤PP. A, effect of pEF4-hGtx on neuronal cytotoxicity by A617G-A␤PP or L648P-A␤PP by the 1.5-g protocol. F11 cells were transfected with empty pcDNA (pcDNA) or each of pcDNA-encoded V642I-A␤PP, A617G-A␤PP, or L648P-A␤PP with pEF4-hGtx (hGtx) or empty pEF4/His (pEF). Cell mortality was measured 48 h after transfection. *, significantly different. n.s., no significant difference. Inset, effect of pEF4-hGtx on expression of A617G-A␤PP or L648P-A␤PP. Using the same 0.5-g protocol, caused neurotoxicity as high as the neurotoxicity by similarly transfected V642I-A␤PP and NL-A␤PP (Fig.  3B, left panels). Co-transfection with hGtx greatly inhibited neuronal cell death by low expression of A617G-A␤PP or L648P-A␤PP (Fig. 3B, left panels) without critically affecting expression of A617G-or L648P-A␤PP as compared with their expression with pEF4/His co-transfection (data not shown). This was also the case, when cell death was reciprocally assessed by cell viability (Fig. 3B, right panels). Therefore, hGtx suppressed neurotoxicity by low expression of A617Gand L648P-A␤PP.
Effects of hGtx on Neurotoxicity by M146L-PS1 and N141I-PS2-We next examined whether hGtx inhibits neuronal cell death by low and high induction of M146L-PS1 or N141I-PS2 in the pIND system. Both FAD mutants cause or enhance neuronal cell death in various cell systems, including F11 neurohybrid cells (10 -12, 14, 15, 20, 39 -43). As shown in Fig. 4, A and  B, co-expression of hGtx had no inhibitory effect on neuronal cell death by low to higher induction of M146L-PS1. In contrast, co-expression of hGtx, but not with empty pEF4/His, drastically inhibited neuronal cell death by low induction of N141I-PS2, without effect on neuronal cell death by higher induction of N141I-PS2. No effect of hGtx on neurotoxicity by high induction of M146L-PS1 or N141I-PS2 is consistent with the F11 cell transfection data using the 1.5-g protocol (data not shown). These results concur with the notion that hGtx selectively inhibits DEVD-sensitive apoptotic neurotoxicity by mutant PS1/2, because (i) low induction of N141I-PS2 (achieved by the pIND protocol transfection and 10 M EcD treatment) induces DEVD-sensitive apoptotic neurotoxicity; (ii) high induction (achieved by the pIND protocol transfection and 40 M EcD treatment) of N141I-PS2 additionally causes DEVD-resistant non-apoptotic neurotoxicity; and (iii) low to high induction of M146L-PS1 solely causes DEVD-resistant non-apoptotic neurotoxicity (14,15).
Again using the 1.5-g protocol, we transfected F11 neurohybrid cells with each of these deletion mutants with V642I-A␤PP cDNA (in pcDNA), and we assessed whether cell death was inhibited. Both hGtxCT and hGtx⌬P inhibited neuronal cell death by V642I-A␤PP as significantly as full-length hGtx. In contrast, hGtxNT did not inhibit it at all. hGtxCT, hGtx⌬P, and hGtxNT were expressed similarly to hGtx (Fig. 5B, upper  panel). Hence, the homeodomain and/or the Pro-rich domain would be responsible for the neuroprotective action of hGtx. We therefore examined the effect of hGtxH, the homeodomain (residues 148 -207) of hGtx. hGtxH inhibited neuronal cell death by V642I-A␤PP and by L648P-A␤PP as significantly as fulllength hGtx did (Fig. 5D). Expression of these FAD mutant 1.5-g protocol, F11 cells were transfected with either A617G-A␤PP or L648P-A␤PP (each in pcDNA) with pEF4-hGtx (hGtx) or empty pEF4/His (pEF/His). Cellular expression of A␤PP was measured by immunoblot analysis 48 h after transfection. An arrowhead indicates the holoproteins of expressed A␤PPs. B, effect of pEF4-hGtx on neuronal cytotoxicity by V642I-A␤PP, L648P-A␤PP, NL-A␤PP, or A617G-A␤PP by the 0.5-g protocol. F11 cells were transfected with pEF4-hGtx (hGtx) or empty pEF4/His (pEF/His). At 72 h after transfection, cell mortality and viability were measured by trypan blue exclusion (left panels) and WST-8 assay (right panels), respectively. As negative controls, transfection with pcDNA alone (vec) or wtA␤PP cDNA in pcDNA (wtA␤PP) was performed in parallel. *, significantly different. n.s., no significant difference. A␤PPs in cells co-transfected with hGtxCT, hGtx⌬P, hGtxNT, or hGtxH was not significantly inhibited, as compared with the expression of the A␤PP mutants in cells co-transfected with empty pEF4/His (Fig. 5B, lower panel). Therefore, it was most likely that the neuroprotective function of hGtx against FAD mutants of A␤PP is commonly located in its homeodomain.
Effect of DLX4 on Neuronal Cell Death by V642I-A␤PP-We further tested whether other homeodomain-containing proteins inhibit neuronal cell death by overexpression of V642I-A␤PP. For this purpose, we examined the effect of DLX4. Because the homeodomain of DLX4 is only 38.3% homologous to that of hGtx, the neuroprotective function of the hGtx homeodomain would be common for other homeodomain-containing proteins, if DLX4 inhibits neuronal cell death by V642I-A␤PP. As shown in Fig. 5E, co-expression of DLX4 inhibited neurotoxicity by V642I-A␤PP as significantly as hGtxH and fulllength hGtx did. Immunoblot analysis with (His) 6 tag antibody revealed that DLX4 was expressed similarly to hGtxH (Fig. 5B,  upper panel). CMV-promoter-driven expression of V642I-A␤PP was virtually the same with hGtxH co-transfection and with DLX4 co-transfection (Fig. 5B, lower panel). It was therefore likely that homeodomains would be able to cause neuroprotection against FAD mutant A␤PP.
Effect of hGtx Homeodomain on A␤ Neurotoxicity-We examined whether the homeodomain of hGtx suppresses A␤-induced neurotoxicity. As noted above, F11 neurohybrid cells became sensitive to A␤ neurotoxicity when cells were transfected with p75 NTR cDNA (Table II). As shown in Table II, not only hGtxCT but also hGtxH inhibited neuronal cell death by A␤ as strongly as full-length hGtx, whereas hGtxNT could not inhibit A␤ neurotoxicity. hGtx constructs were expressed similarly (data not shown). Expression of p75 NTR was not affected by co-transfection of hGtx and its fragments (data not shown). Again, these data confirm that (i) hGtx can suppress A␤ neurotoxicity; and (ii) the homeodomain of hGtx is responsible for the neuroprotective function of hGtx. Table II also indicates that A␤induced neurotoxicity was DEVD-sensitive apoptosis, as has been reported by multiple groups (44 -47). This finding is in good agreement with the notion that hGtx is effective in inhibiting apoptotic neuronal cell death.
Effect of hGtx on ApoE4-induced Neurotoxicity-To further confirm that neuroprotection by hGtx is not due to inhibition of the FAD gene expression, we examined whether hGtx inhibits F11 neurohybrid cell death caused by extracellularly added apolipoprotein E4 (apoE4). ApoE4 is the established insults relevant to AD and causes apoptotic cell death, when F11 neurohybrid cells are treated with apoE4 (21). As shown in Fig.  6A, apoE4 robustly induced cell death, as assessed with both trypan blue exclusion mortality assay and WST-8 cell viability assay. When cells had been transfected with hGtx, apoE4induced toxicity was greatly suppressed, as assessed by both assays. Furthermore, hGtx-mediated suppression of apoE4induced neuronal cell death was reproduced by hGtxH but not by hGtxNT. These Gtx constructs were expressed similarly Lower panel, effect of pEF4-hGtx on expression of V642I-A␤PP or L648P-A␤PP. By using the same 1.5-g protocol, F11 cells were transfected with either V642I-A␤PP or L648P-A␤PP (each in pcDNA) with empty pEF4/His, pEF4-hGtx, hGtx mutant, or DLX4. Cellular expression of A␤PP was measured by immunoblot analysis 48 h after transfection. C, effects of deletion mutants of hGtx on neuronal cytotoxicity by V642I-A␤PP and L648P-A␤PP. F11 cells were transfected with empty vector (pcDNA) or pcDNA-encoded V642I-A␤PP or L648P-A␤PP with pEF4/His vector (pEF) or either hGtx construct (pEF4-hGtx, pEF-hGtxNT, pEF-hGtxCT, pEF-hGtx⌬P) using the 1.5-g protocol. Cell mortality was measured 48 h after transfection. D, effect of hGtxH on neuronal cytotoxicity by V642I-or L648P-A␤PP. Under the same conditions as in C, F11 cells were transfected with empty vector (pcDNA) or pcDNA-encoded V642I-or L648P-A␤PP with empty pEF4/His vector (pEF) or either pEF-hGtx or pEF-hGtxH. Cell mortality was similarly measured. E, effect of DLX4 on neuronal cytotoxicity by V642I-A␤PP. Under the same conditions as in C, F11 cells were transfected with empty vector (pcDNA) or pcDNA-encoded V642I-A␤PP with either empty pEF4/His vector (pEF), pEF4-hGtxH, or DLX4. Cell mortality was similarly measured. C-D, * indicates cytotoxicity significantly lower than that by V642I-A␤PP plus pEF4/His; and ** indicates cytotoxicity not significantly differing from that by V642I-A␤PP plus pEF4/His. q indicates cytotoxicity by V642I-A␤PP plus pEF4/His, which is significantly augmented relative to the cytotoxicity by pcDNA plus pEF4/His.
(data not shown). These data provide definite support to the notion that hGtx inhibition of AD insult-induced neurotoxicity is due to direct suppression of the relevant toxic pathways.
Mediation by IGF-I of hGtx-induced Neuroprotection-We noticed that hGtx transfection totally suppressed apoE4-induced cell death, despite the fact that the transfection effi-FIG. 6. Effects of hGtx on neuronal cytotoxicity by apoE4 and on IGF-I production. A, effect of hGtx on neuronal cytotoxicity by apoE4. F11 cells were transfected without (no T) or with empty pEF4/His (pEF) or either hGtx construct (pEF4-hGtx (hGtx), pEF4-hGtxNT (NT), or pEF-hGtxH (hGtxH)) and treated with or without 30 g/ml apoE4. Seventy-two h after apoE4 treatment, cell mortality and viability were measured by trypan blue exclusion (left panel) and WST-8 assay (right panel and bottom photograph), respectively. q indicates cytotoxicity significantly differing from cytotoxicity without apoE4. * indicates cytotoxicity significantly different from that of nontransfected cells treated with apoE4. B: upper panel, effect of anti-IGF-I neutralizing antibody on the neuroprotective activity of hGtx. F11 cells were transfected with empty pEF4/His (pEF) or hGtx construct (pEF4-hGtx (hGtx)) and treated with or without 30 g/ml apoE4 in the presence or absence of 50 nM anti-IGF-I antibody. Seventy-two h after apoE4 treatment, cell mortality and viability were measured by trypan blue exclusion test. * indicates cytotoxicity significantly different from that of pEFtransfected cells treated without apoE4. Lower panel, immunoblot analysis of expressed hGtx. Lysates of F11 cells transfected with hGtx or its deletion mutants were submitted to SDS-PAGE and immunoblot analysis with anti-(His) 6 antibody. C, effect of anti-IGF-I neutralizing antibody and ␣IR3, which is a neutralizing antibody against IGF-I receptor, on the neuroprotective activities of hGtx, HN, or DLX4. F11 cells were transfected without (no T) or either construct (pEF4-hGtx (hGtx), pFLAG-HN (HN), or pEF-DLX4-(DLX4)) and treated with or without 30 g/ml apoE4 and 50 nM IgG, 50 nM anti-IGF-I (␣IGF), or 50 nM ␣IR3. Seventy-two h after apoE4 treatment, cell mortality and viability were measured by trypan blue exclusion test. * indicates cytotoxicity significantly different from that of nontransfected cells treated without apoE4. ** indicates no difference of cell mortality. D, effect of hGtx, hGtxNT, hGtxH, and DLX4 on IGF-I production and secretion. F11 cells were transfected without (no T) or either construct (pEF4-hGtx (hGtx), pEF4-hGtxNT (hGtxNT), pEF4-hGtxH (hGtxH), or pEF-DLX4-(DLX4)). Cellular expression (in cell lysate) and secretion (in cultured medium) of IGF-I were detected by immunoblot analysis 72 h after transfection. ciency was ϳ65-70%. It was therefore reasonable to assume that a soluble rescue factor is secreted from hGtx-transfected cells. We have shown so far that IGF-I inhibits apoptotic neuronal cell death by V642I-A␤PP (48) and does not affect DEVDresistant cell death by NL-A␤PP, N141I-PS2, or M141L-PS1 (22). As shown in Fig. 6B, IGF-I inhibited the F11 cell death by apoE4. Under the same conditions, the neuroprotective activity of hGtx was virtually antagonized by the addition of anti-IGF-I-specific antibody. The expression of hGtx did not change in the presence or absence of anti-IGF-I antibody. This spectrum of the effectiveness of IGF-I and its functional specificity against an apoptotic component of cell death resembles such a soluble rescue factor potentially secreted from hGtx-transfected cells.
We thus examined the effect of specific neutralizing anti-IGF-I antibody added to the culture medium. In fact, IGF-I antibody was able to completely inhibit hGtx-mediated suppression of apoE4-induced neuronal apoptosis without downregulating expression of hGtx (Fig. 6B). In addition, as shown in Fig. 6C, not only anti-IGF-I neutralizing antibody but also ␣IR3, which is a neutralizing antibody against IGF-I receptor (48), antagonized this hGtx-induced neuroprotective effect. In the case of Humanin (HN), which is an anti-AD peptide (20 -22), both neutralizing antibodies had no effect on neuroprotective activity of HN. The neuroprotective activity of DLX4 was also antagonized by both anti-IGF-I antibody and ␣IR3 (Fig.  6C). Finally, we confirmed that expression of IGF-I was actually induced in F11 cells and secreted into culture medium by transfection of hGtx, hGtxH, and DLX4 but not by transfection of hGtxNT (Fig. 6D). Taken these findings together, we concluded that IGF-I is the actual mediator for hGtx-induced neuroprotection.
To determine whether Gtx and DLX4 can actually enhance the secretion of IGF-I, we performed immunoblotting. As shown in Fig. 6D, hGtx, hGtxH, and DLX4 increased the production and the secretion of IGF-I. From these results, there are some possibilities that homeobox-containing genes can enhance the production of growth factors to modulate development, cell death, and survival. DISCUSSION We have shown here that hGtx has a novel function that inhibits neuronal cell death by FAD mutants of A␤PP. In multiple systems, hGtx significantly inhibited neurotoxicity by A␤, by low and high expression of V642I-and L648P-A␤PP, and by low expression of NL-A␤PP, A617G-A␤PP, and N141I-PS2. On the contrary, hGtx did not significantly inhibit neuronal cell death by M146L-PS1, by high expression of NL-A␤PP, A617G-A␤PP, or N141I-PS2. Neurotoxicity by G85R-SOD1, a familial amyotrophic lateral sclerosis-linked mutant SOD1, is not suppressed by Gtx. 2 The former hGtx-suppressive neurotoxicity is sensitive to caspase inhibitors, and the latter hGtx-resistant neurotoxicity is caspase-inhibitor-insensitive, indicating that hGtx specifically inhibits apoptotic death. This study has further specified the neuroprotective region of hGtx to be its homeodomain. As DLX4, which has a homeodomain only 38.3% homologous to the homeodomain of hGtx, has a similar cytoprotective function, it is conceivable that such a function is broadly associated with homeodomains.
The function of hGtx to promote neuronal cell survival is novel but not surprising, because (i) mGtx does not bind to the sequences recognized by Antp-, Eve-, En-, or Prd-type homeobox proteins but binds to the MEF-2-binding sequence (1); and (ii) neuronal activity-dependent cell survival is mediated by MEF-2 (49). Whereas the relevance of this novel function of hGtx is not clear at present, this study does indicate interesting cross-talk between developmentally regulated machinery and AD-causative mechanisms. Some functional linkage has so far been postulated between them, based on the regulatory roles of PS proteins in activation of Notch (50 -52) as well as the genetargeting studies showing the essential roles of PS1 in skeletal and neuronal development (53,54) and PS2 in pulmonary development (55). Inappropriate expression of developmentally regulated genes in AD has also been reported (56,57). Hong et al. (58) reported that FAD-associated mutant PS1 plays a role in the differentiation of neuronal precursors. It has also been reported that mouse homeobox gene Hox-3.1 negatively regulates the A␤PP mRNA level (59).
In this study, we identified IGF-I to be the mediator for hGtx-mediated neuroprotection. Expression of Gtx up-regulates expression of IGF-I. It has been known that Nkx family transcription factors including Gtx act as transcriptional repressors (1). Therefore, there is a possibility that Gtx downregulates expression of the unknown genes that are suppressors or silencers for the IGF-I gene, whereas the precise mechanism of Gtx-mediated induction of IGF-I expression remains still unclear at this moment.
As a neuroprotective factor, IGF-I has been already known to antagonize apoptotic neuronal cell death by V642I-A␤PP (48). We additionally obtained two important insights regarding biological and therapeutic roles of IGF-I by this study. First, some homeobox-containing transcription factors such as Gtx and DLX4 regulate expression of IGF-I, indicating that IGF-I plays a important role not only in neuronal cell protection but also in embryonic development. Second, as a therapeutic agent for a certain type of AD, IGF-I seems to be more promising than previously because it is considered to be a putative physiological mediator for Gtx-mediated neuroprotection from A␤PPrelevant insults.
In summary, this study provides a molecular clue toward understanding neuroprotection against A␤PP-relevant insults and suggests a novel linkage between neural development and inhibition of AD mechanisms. This study may open a new horizon in the functions of homeodomain-containing proteins.