sgk, a primary glucocorticoid-induced gene, facilitates memory consolidation of spatial learning in rats Kuen J. Tsai*, Shau K. Chen†, Yun L. Ma†, Wei L. Hsu†, and Eminy H. Y. Lee†‡ *Graduate Institute of Life Sciences, National Defense Medical Center, and †Institute of Biomedical Sciences, Academia Sinica, Taipei 115, Taiwan, Republic of China Edited by Richard F. Thompson, University of Southern California, Los Angeles, CA, and approved January 25, 2002 (received for review August 1, 2001)

By using differential display PCR, we have identified 98 cDNA fragments from the rat dorsal hippocampus that are expressed differentially between the fast learners and slow learners in the water maze learning task. One of these cDNA fragments encodes the rat serum- and glucocorticoid-inducible kinase (sgk) gene. Northern blot analysis revealed that the sgk mRNA level was approximately 4-fold higher in the hippocampus of fast learners than slow learners. In situ hybridization results indicated that sgk mRNA level was increased markedly in CA1, CA3, and dentate gyrus of hippocampus in fast learners. Transient transfection of the sgk mutant DNA to the CA1 area impaired, whereas transfection of the sgk wild-type DNA facilitated water maze performance in rats. These results provide direct evidence that enhanced sgk expression facilitates memory consolidation of spatial learning in rats. These results also elucidate the molecular mechanism of glucocorticoidinduced memory facilitation in mammals.

I

t is well accepted that long-term memory formation requires de novo RNA and protein synthesis. Evidence supporting this notion comes from the results that inhibition of mRNA and protein synthesis impairs long-term memory formation in various behavioral tasks in rats (1–3). This evidence suggests that neural activities associated with learning leads to the expression of various genes, and the protein products of these genes play important roles in the process of memory formation. Extensive efforts have been made to identify genes that are associated specifically with certain forms of learning and memory. For example, by using two-dimensional gel analysis, several candidate proteins have been identified that are related to long-term sensitization of the gill-withdrawal reflex in Aplysia (4). Screening in Drosophila mutants has yielded approximately 10 genes that are associated with the process of olfactory learning and memory (5). Further, by using a double-labeling method, proteins that show increased glycosylation as a result of training were identified in rats (6). The methods used in the above studies are effective; however, identifying and characterizing these genes takes a long time when using these methods. In addition, most of these studies were carried out in the invertebrate, in which the neuronal circuits and genes involved in memory formation probably are different from that in the vertebrate. In a more recent study, by using differential display–PCR (DD-PCR), we have successfully identified genes that are associated specifically with memory formation of one-way inhibitory avoidance learning in rats (7, 8). Some of these results were confirmed further by a gene-knockout study (9). These results suggest that DD-PCR is an effective method in identifying genes that are involved in complex forms of learning and memory in the vertebrate. Therefore, we have adopted the same strategy in the present study to identify genes that are associated specifically with memory formation of spatial learning in rats. The Morris water maze was used as the behavioral paradigm, and the dorsal hippocampus was the target of interest because this area is implicated in place navigation in mammals (10, 11). Upon identification of the candidate genes from DD-PCR, we confirmed it further by Northern blot and in situ hybridization analyses. We then transfected the dominant negative mutant of

3990 –3995 兩 PNAS 兩 March 19, 2002 兩 vol. 99 兩 no. 6

the gene to the CA1 area and examined its effect on water maze performance. Our results demonstrated that the serum- and glucocorticoid-inducible kinase (sgk) gene is one of these genes that is causally involved in memory consolidation of spatial learning in rats. Materials and Methods Animals. Adult male Sprague–Dawley rats bred at the Animal

Facility of the Institute of Biomedical Sciences (Academia Sinica, Taiwan) were used in the present study. Animals were housed in a room maintained on a 12-h兾12-h light兾dark cycle (light on at 6:30 a.m.) with food and water continuously available. Experimental procedures follow the Guidelines of Animal Use and Care of the National Institutes of Health. Water Maze and Spatial Learning Procedure. The water maze used

was a plastic, circular pool, 2.0 m in diameter and 0.6 m in height, that was filled with water (25 ⫾ 2°C) to a depth of 20 cm. A circular platform (8 cm in diameter) was placed at a specific location away from the edge of the pool. The top of the platform was submerged 1.5 cm below the water surface. Water was made cloudy by the addition of milk powder. Distinctive, visual cues were set on the wall. For spatial learning, animals were subjected to four trials a session, two sessions a day, with one given in the morning and the other given in the afternoon. A total of eight sessions were given for screening the fast and slow learners, and four sessions were given for the sgk DNA transfection experiment. This 2-day interval was chosen because transient DNA transfection was shown to have an optimal efficiency between 48 and 72 h after injection (12). In each of the four trials, animals were placed at four different starting positions equally spaced around the perimeter of the pool in a random order. Animals were allowed to find the platform in 120 s. If an animal could not find the platform in 120 s, it was guided to the platform. After mounting the platform, animals were allowed to stay there for 20 s. The time each animal spent to reach the platform was recorded as the escape latency. In the experiment to select the fast learners and slow learners, a total of 114 rats were screened. The rats that found the hidden platform within 30 s before the end of the third session were assigned as the fast learners (i.e., fewer trials were needed for memory consolidation). Animals that did not find the platform within 30 s until session 7 or 8 were assigned as the slow learners (i.e., more trials were needed for memory consolidation). Animals that fell in between were assigned as the controls.

This paper was submitted directly (Track II) to the PNAS office. Abbreviations: SGK, serum- and glucocorticoid-inducible kinase; DD, differential display; GR, glucocorticoid receptor; HA, hemagglutinin; RT, reverse transcription; HPRT, hypoxanthine phosphoribosyl transferase. Data deposition: The sequence reported in this paper has been deposited in the GenBank database (accession no. NM㛭019232). ‡To

whom reprint requests should be addressed. E-mail: [email protected].

The publication costs of this article were defrayed in part by page charge payment. This article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

www.pnas.org兾cgi兾doi兾10.1073兾pnas.062405399

DD-PCR. DD-PCR was performed according to that described

previously (8), with slight modifications. Random primers (RNAimage Kit) were purchased from GenHunter (Nashville, TN). Briefly, three separate reaction mixtures were prepared, each containing 2 ␮l of one-base-anchored primers (2 ␮M) (H-T11A, H-T11C, and H-T11G) for reverse transcription (RT). These three primers were incubated with 2 ␮l of 0.1 ␮g兾␮l total RNA, 1.6 ␮l of 250 ␮M dNTP, 4 ␮l of 5⫻ RT buffer, and 9.4 ␮l of diethyl pyrocarbonate H2O, separately, to a final volume of 19 ␮l, and incubated at 65°C for 5 min then at 37°C for 60 min. The tubes stayed at 37°C for 10 min, and 1 ␮l of mouse murine leukemia virus reverse transcriptase (100 units兾␮l) (Life Technologies, Gaithersburg, MD) was added. The reaction was terminated by heating the sample at 75°C for 5 min. After RT, the reverse-transcribed products were aliquoted into fresh tubes and subjected to different amplification reactions by using 80 arbitrary primers (H-AP1⬃H-AP80). Two microliters of the RT product were added to the PCR mixture containing 2 ␮l of 2 ␮M H-AP primer, 2 ␮l of 2 ␮M H-T11M primer (M may be A, C, or G), 1.6 ␮l of 25 ␮M dNTP, 2 ␮l of 10⫻ PCR buffer, 1 ␮l of [␣-35S]dATP (1,000 Ci兾mmol) (Amersham Pharmacia), 1 unit of Taq DNA polymerase (Perkin–Elmer), and 9.2 ␮l of distilled water to a total volume of 20 ␮l. The PCR parameters were 94°C for 30 s, 40°C for 2 min, and 72°C for 30 s for 40 cycles followed by 72°C for 5 min. All 240 reactions were performed in duplicate. Differentially expressed cDNA fragments were recovered from the sequencing gels and cloned into the PCR 2.1 TA vector (Invitrogen). Northern Blot. Total RNA was fractionated by electrophoresis

through 1.0% agarose兾formaldehyde gels and transferred overnight on Hybond-XL (Amersham Pharmacia) in a solution of 20⫻ SSC (1 SSC ⫽ 0.15 M sodium chloride兾0.015 M sodium citrate, pH 7). Total RNA was fixed to the membrane by exposure to UV irradiation. The membrane was prehybridized for 4 h at 65°C in hybridization buffer containing 6⫻ SSC, 5⫻ Denhardt’s solution (0.5% Ficoll兾0.5% polyvinylpyrrolidone兾 0.5% BSA), 0.5% SDS, and 100 ␮g兾ml heat-denatured salmon sperm DNA. The membrane was hybridized overnight at 65°C in hybridization buffer with 5 ng兾ml 32P-labeled cDNA probe. A cDNA probe synthesized from 608 bp of the sgk cDNA (nucleotide 1828–2435) was used to hybridize with a single mRNA species of about 2.5 kb (13). The rat glyceraldehyde-3-phosphate dehydrogenase cDNA probe that hybridizes a 1.3-kb band was used as the internal control. After two washes in 2⫻ SSC兾0.1% SDS at 42°C for 10 min, followed by two washes at 65°C for 10 min in 0.1⫻ SSC兾0.1% SDS, the membrane was dried and autoradiographed. In Situ Hybridization. In situ hybridization was performed with a

45-base synthetic oligonucleotide (5⬘-GCG GAG ATC CCT CTT AGA CCT GCA TCT TCC TTC TCA CTG AGA CCA-3⬘) (14). The antisense probe (complementary to the cloned rat sgk cDNA, nucleotide 1654–1698) and the sense probe (5⬘-TGG TCT CAG TGA GAA GGA AGA TGC AGG TCT AAG AGG GAT CTC CGC-3⬘) were synthesized and purified by Genosys (The Woodlands, TX). The probes (15 pmol兾␮l) were 3⬘ endlabeled by incubating at 37°C for 15 min with [␣-35S]dATP and terminal deoxynucleotidyl transferase (25 units; Roche Molecular Biochemicals). Prehybridization treatment of tissue conTsai et al.

sisted of rinsing in 20⫻ SSC for 10 min at room temperature. For hybridization, the labeled sgk oligonucleotide probe (1 ⫻ 106 cpm兾slide) in 100 ␮g兾ml yeast transfer RNA, 500 ␮g兾ml salmon sperm DNA, and Denhardt’s solution was applied to each slide. Slides were coverslipped with parafilm, and hybridization proceeded at 42°C for 24 h. Coverslips then were removed, and sections were rinsed in 2⫻ SSC and then in 1⫻ SSC containing 1.0 M DTT (0.1%), followed by a 30-min wash in 0.5⫻ SSC containing 1.0 M DTT at 47°C. A final wash in 0.5⫻ SSC containing 1.0 M DTT was performed at room temperature for 30 min. Slides were dehydrated through a series of ethanol bath (50%, 75%, 95%, and 100%) and exposed to Hyperfilm MP for 2 weeks. Signals from in situ hybridization were quantified by measuring the optic density of the relevant fields by using the National Institutes of Health IMAGE program. Plasmid DNA Construction and DNA兾Polyethyleneimine (PEI) Complexes Preparation. For construction of the hemagglutinin (HA)

epitope-tagged DNA plasmid (HA-SGK), full-length sgk was cloned by amplifying the rat hippocampal sgk cDNA with primers 5⬘-CGG AAT TCA CCG TCA AAA CCG AGG CTC G-3⬘ and 5⬘-GCT CTA GAT CAG AGG AAG GAG TCC ATA GG-3⬘. The PCR product was subcloned between the EcoRI and XbaI sites of the mammalian expression vector pcDNA3-HA. The mutant at Ser-422 (kinase-deficient HA-SGK S422A) (15) was generated by PCR with primers 5⬘-CGG AAT TCA CCG TCA AAA CCG AGG CTC G-3⬘ and 5⬘-GCT CTA GAT CAG AGG AAG GAG TCC ATA GGA GGG GCA TAG GCG AAG-3⬘, with HA-SGK as template, and inserted into the pcDNA3-HA expression vector. The efficiency of sgk mutant DNA transfection was confirmed by a decreased sgk activity in HEK 293 cells (15). Before injection, plasmid DNA was diluted in 5% glucose to 5 ␮g兾␮l concentration. Linearized PEI of 22 kDa (Sigma) was diluted to 0.1 M in 5% glucose and added to the DNA solution. The mixture was vortexed for 30 s and allowed to equilibrate at room temperature for 10 min before injection. Intrahippocampal Gene Transfection. Animals were anesthetized

with pentobarbital (40 mg兾kg, i.p.) and subjected to stereotaxic surgery. Two 23-gauge, stainless-steel, thin-wall cannulae were implanted bilaterally to the CA1 area at the following coordinates: 3.5 mm posterior to the bregma, 2.5 mm lateral to the midline, and 3.4 mm ventral to the skull surface. After recovery from the surgery, 1 ␮l of 5% glucose solution containing 2 ␮g of plasmid DNA complexes with 10 PEI equivalents (12) was injected into the CA1 area (0.5 ␮l兾min). The injection needle was left in place for 5 min to limit the diffusion of the injected DNA. Animals were subjected to the spatial learning task 48 h later. The area of transfection in CA1 then was examined and quantified by using the same National Institutes of Health IMAGE program according to the criterion of Amaral and Witter (16). Immunohistochemistry. Brain sections were rinsed with 1⫻ PBS

for 10 min at room temperature and permeabilized with pre-cold EtOH兾CH3COOH (95%:5%) for 10 min, followed by 1⫻ PBS for 10 min three times. The sections were preincubated in a blocking solution containing 3% normal goat serum, 3% BSA, and 0.2% Triton X-100 in 1⫻ PBS for 2 h at room temperature, followed by 1⫻ PBS for 10 min three times. For immunofluorescence analysis, tissue sections were incubated with a mouse monoclonal anti-HA antibody (dilution, 1:100; Roche Molecular Biochemicals) in blocking buffer at 4°C overnight. Sections were washed three times in 1⫻ PBS and incubated in the secondary antibody as follows: goat anti-mouse FITC-conjugated IgG antibody (dilution, 1:100; Sigma) in 1⫻ PBS for 1 h at room temperature. Sections were washed three times in 1⫻ PBS and mounted with mounting medium. PNAS 兩 March 19, 2002 兩 vol. 99 兩 no. 6 兩 3991

NEUROBIOLOGY

Among the 114 rats screened, 21 rats were classified as the fast learners and 21 rats were classified as the slow learners. The remaining (n ⫽ 72) were the control rats (Fig. 1 A). In the visible platform experiment, a flag was mounted on the platform and the platform was 2.5 cm above the water surface. In addition, milk powder was not added to the swimming pool so that the animals can visualize the location of the platform.

Table 1. Identification and classification of cDNA fragments that are differentially expressed in the dorsal hippocampus between fast and slow learners Classification Unknown Mitochondrial protein Transmembrane protein Intracellular protein Nuclear protein Total number of cDNA fragments

41 52 2 2 1 98

Quantitative RT-PCR. The RT reaction was performed as described above. The rat hypoxanthine phosphoribosyl transferase (HPRT) mRNA was used as an internal control template. Synthetic primers 5⬘-CTC TGT GTG CTG AAG GGG GG-3⬘ and 5⬘-GGG ACG CAG CAA CAG ACA TT-3⬘ were used to detect the HPRT mRNA, which yielded a PCR product of 625 bp in length. Synthetic primers 5⬘-TTT TTT TTC CCA ACC CTT GC-3⬘ and 5⬘-AAT GAA CAA AGG TTG GGG GG-3⬘ were used to detect the SGK mRNA, which yielded a PCR product of 390 bp in length. The parameters used were: 94°C for 30 s, 57°C for 30 s, 72°C for 40 s for 26 cycles, followed by a final elongation at 72°C for 7 min. The PCR product was analyzed on a native 8.0% polyacrylamide gel and autoradiographed for PhosphorImager analysis (Molecular Dynamics).

Results DD-PCR. Three fast learners and three slow learners were chosen randomly among 21 rats in each group, and the total RNA extracted from their dorsal hippocampus was subjected to DDPCR analysis. Other fast learners and slow learners were subjected to Northern blot analysis, in situ hybridization, and spatial learning test with visible platform, respectively. Three 3⬘ end primers (H-T11A, H-T11C, and H-T11G) in combination with 80 5⬘ end arbitrary primers (H-AP1⬃H-AP80) (240 primer pairs in total) were used for screening, and approximately 15,000 DNA fragments were yielded. Among these cDNA fragments, there were 98 cDNA bands that are consistently and differentially expressed between the fast learners and slow learners based on an ‘‘all-or-none’’ criterion. Further cloning and sequencing analyses revealed that there were 41 cDNA bands that are unknown genes. There were 52 cDNA bands that encode genes of mitochondrial proteins. Two of these 98 cDNA bands were transmembrane protein genes, and another two encode genes of intracellular proteins. The last one encodes the gene of a nucleus protein. Identification and characterization of these genes are summarized in Table 1. When the primer set H-AP48 (5⬘ primer with oligonucleotide sequence as 5⬘-AAGCTTGCGGTGA-3⬘) and H-T11A (3⬘ primer with oligonucleotide sequence as 5⬘AAGCTTTTTTTTTTTA-3⬘) was used, one identified cDNA fragment that was 402 bp in length showed 96% sequence homology to the 3⬘ end region of the rat sgk gene (Fig. 1D). The expression level of this gene is much higher in the dorsal hippocampus of fast learners than slow learners (Fig. 1B). Northern Blot Analysis. Northern blot experiments were carried

out in seven fast learners and seven slow learners to confirm the results obtained from DD-PCR. Representative autoradiographs showing the SGK and glyceraldehyde-3-phosphate dehydrogenase cDNA bands are illustrated in Fig. 1C. Statistical analysis revealed that the sgk mRNA level is approximately 4-fold higher in the dorsal hippocampus of fast learners than slow learners [t(1,12) ⫽ 8.01, P ⫽ 0.01].

3992 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.062405399

Fig. 1. (A) Water maze performance of fast learners (that found the hidden platform within 30 s before the third session; n ⫽ 21), slow learners (that found the hidden platform within 30 s at session 7 or 8; n ⫽ 21), and control rats (animals in between fast and slow; n ⫽ 72). Data are expressed as means ⫾ SE. FL, fast learner; SL, slow learner. (B) DD-PCR of hippocampal RNA associated with memory consolidation of water maze learning in rats. Total RNA extracted from the dorsal hippocampus of three slow learners and three fast learners was subjected to DD-PCR. When the primer set H-AP48 and H-T11A was used, a cDNA fragment that encodes the rat sgk gene was differentially expressed between the slow learners and fast learners. (C) Representative autoradiograph showing the SGK and glyceraldehyde-3-phosphate dehydrogenase cDNA bands from four slow learners and four fast learners in Northern blot analysis. Statistical analysis revealed that the hippocampal sgk mRNA level was about 4-fold higher in fast learners than slow learners. n ⫽ 7 in each group. Data are expressed as means ⫾ SE; **, P ⫽ 0.01 by Student’s t test. (D) Alignment of the sequence of A48 –1-6 with rat sgk. The numbers correspond to the cDNA sequence according to Webster et al. (17). Vertical lines indicate identity. The bold text indicates the sequence and location of the arbitrary primer used in DD-PCR.

In Situ Hybridization. Results of in situ hybridization analyses from

three fast learners and three slow learners revealed that the sgk mRNA is expressed mainly in the dentate gyrus of hippocampus in slow learners, although the distribution of sgk mRNA in the pyramidal cell layer also can be visualized (Fig. 2A). However, an apparent increase in the sgk mRNA level was observed in both the pyramidal cell layer and dentate gyrus of the hippocampus Tsai et al.

in fast learners (Fig. 2B). Further analyses revealed that the sgk mRNA level was increased markedly in CA1, CA3, and the dentate gyrus [t(1,4) ⫽ 33.85, 39.37, and 63.83, respectively, all P ⬍ 0.01] (Fig. 2D). When the sense probe was used as the control, no specific labeling was observed (Fig. 2C).

approximately 22.4% for the SGKS422A group (22.4 ⫾ 2.2) and 23.1% for the SGKWT group (23.1 ⫾ 1.8).

Effects of sgk DNA Transfection on Water Maze Performance in Rats.

Sgk mRNA Expression Is Induced upon Learning. Results from the

To elucidate the cause–effect relationship between sgk mRNA expression and memory consolidation, we examined the effects of sgk DNA transfection on water maze performance in rats. Results revealed that there was an overall significant effect of sgk DNA transfection on spatial learning in rats [F(2,21) ⫽ 17.06, P ⬍ 0.001] (Fig. 3A). Further analyses revealed that sgk mutant DNA transfection to CA1 significantly impaired spatial learning performance [F(2,21) ⫽ 6.9, P ⬍ 0.05], whereas sgk wild-type DNA transfection markedly enhanced spatial learning performance in rats [F(2,21) ⫽ 10.32, P ⬍ 0.05]. The first trial performance was not markedly different [F(2,21) ⫽ 1.00, P ⬎ 0.05]. To confirm the efficiency of sgk DNA transfection, we conducted an immunohistochemistry experiment by using antibody against the HA-tagged protein and FITC-conjugated IgG secondary antibody. Results revealed that when the non-HA vector was transfected, no fluorescence labeling was observed (Fig. 3B). However, when the HA-SGKS422A mutant DNA was transfected to the CA1 area, fluorescence labeling was clearly visualized (Fig. 3C). At a higher magnification, it shows that the DNA construct indeed was transfected into neurons in the CA1 area (Fig. 3E). The average area of transfection in CA1 was Tsai et al.

above experiment demonstrated that enhanced sgk mRNA expression facilitates water maze performance in rats. However, one question that remains is whether sgk mRNA expression is induced upon learning or whether the fast learners may have a constitutively higher level of sgk. We addressed this issue in the present study. Because hippocampal sgk mRNA level is almost nondetectable in control animals when using Northern blot analysis (Fig. 1B), we have adopted the quantitative RT-PCR method to analyze the sgk mRNA level in 10 randomly chosen, naı¨ve rats. A linear relationship between the amount of total RNA (25, 50, 75, and 100 ng) and optical density of sgk and HPRT cDNA bands as well as between the PCR cycle number (24, 26, 28, and 30) and optical density of sgk and HPRT cDNA bands was demonstrated (data not shown). Thus, 50 ng of total RNA and 26 cycles were used for this experiment. Representative autoradiographs showing the SGK and HPRT cDNA bands in the dorsal hippocampus of naı¨ve rats are shown in Fig. 4A. Quantitative analyses revealed that the hippocampal sgk mRNA levels differ within 37% among these individual animals. Water Maze Performance of Fast Learners and Slow Learners as Well as of Sgk DNA-Transfected Animals with Visible Platform. The above

results have demonstrated the relationship between sgk expresPNAS 兩 March 19, 2002 兩 vol. 99 兩 no. 6 兩 3993

NEUROBIOLOGY

Fig. 2. In situ hybridization of sgk mRNA distribution in the hippocampus. Coronal sections through the dorsal hippocampus from representative slow learners (n ⫽ 3) (A) and fast learners (n ⫽ 3) (B) were subjected to in situ hybridization. (C) Hippocampal tissue section hybridized with the sense probe to serve as the control. CA1, CA1 cell layer; CA3, CA3 cell layer; DG, dentate gyrus. (Bar ⫽ 500 ␮m.) (D) Statistical analyses revealed a significant increase in the optical density of sgk mRNA in CA1, CA3, and DG of fast learners (all P ⬍ 0.01) by Student’s t test. Data are expressed as means ⫾ SE.

Fig. 3. Effects of sgk DNA transfection on water maze performance in rats. (A) Animals were transfected with pcDNA3 vector only, pcDNA3 construct containing the sgk mutant DNA (SGKS422A) and HA-tagged DNA, and pcDNA3 construct containing the sgk wild-type DNA and HA-tagged DNA, respectively. Statistical analysis revealed that sgk mutant DNA transfection significantly impaired, whereas sgk wild-type DNA transfection markedly facilitated, water maze performance in rats (two-way ANOVA, both P ⬍ 0.05). n ⫽ 8 in each group. Data are expressed as means ⫾ SE. (B and C) Representative immunohistochemical staining showing transfection of vector only (B) and transfection of the sgk mutant DNA construct to the CA1 area of the hippocampus (C). The FITC-conjugated IgG secondary antibody and anti-HA primary antibody were used. Arrows indicate the areas of transfection. (Bar ⫽ 240 ␮m.) (D) A higher magnification of B at the arrows. (E) A higher magnification of C at the arrows. (Bar ⫽ 30 ␮m.)

Fig. 4. (A) Quantitative RT-PCR analysis of sgk mRNA level in rat hippocampus. Representative autoradiograph showing the sgk cDNA band from the hippocampus of randomly chosen, naı¨ve rats. The ratio of SGK兾HPRT mRNA level in each individual rat is also shown. (B) Water maze performance of fast learners and slow learners with visible platform. Student’s t test revealed a nonsignificant difference in escape latency between these two groups of rats. (C) Water maze performance of rats transfected with vector only and with SGKS422A DNA construct under visible platform. No significant difference was found between these two groups. n ⫽ 8 in each group. Data are expressed as means ⫾ SE for B and C.

sion and memory consolidation. However, it is not known whether the visual discrimination ability, motor coordination, and motivational state are different between the fast learners and slow learners, and that may contribute to the difference in their water maze performance. We tested this hypothesis in this experiment. Different batches of fast learners (n ⫽ 8) and slow learners (n ⫽ 8) were subjected to the same water maze learning task but with a visual cue (flag) attached to the platform. Results revealed that both the fast learners and slow learners reached the platform successfully during trials in the first session. Statistical analysis indicated that there was not a significant difference in water maze performance between these two groups of rats for the entire session [F(1,14) ⫽ 0.34, P ⬎ 0.05] and for the first trial only [F(1,14) ⫽ 3.03, P ⬎ 0.05] (Fig. 4B). We examined further whether sgk is involved in sensory and motor functions and, therefore, affects spatial learning in rats. Results from Fig. 4C revealed that animals transfected with the sgk mutant DNA (n ⫽ 8) performed similarly to the vector transfection group (n ⫽ 8) under visible platform learning [F(1,14) ⫽ 1.05, P ⬎ 0.05]. In addition, no significant difference was found for the first trial [F(1,14) ⫽ 0.37, P ⬎ 0.05]. Discussion The sgk gene originally was identified as a member of the serine兾threonine protein kinase gene family that is induced 3994 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.062405399

transcriptionally by glucocorticoid and serum (17). Sgk is also known as a primary glucocorticoid-induced gene in several cell lines studied in humans (18). In studying the genes involved in the stimulatory effect of aldosterone on sodium reabsorption in renal epithelia, sgk was demonstrated to be an important gene mediating the early phase of aldosterone effect on sodium transport (19, 20). The sgk mRNA expression also was shown to be sensitive to anisotonic and isotonic alterations of a cell, suggesting that it is involved in the regulation of cell volume (21). In addition, through the study of heparin-regulated genes, sgk was found to be associated with the proliferation of vascular smooth muscle cells (22). All of the above studies examine the role of sgk in peripheral cells, whereas relatively less is known about the role of sgk in the central nervous system. In an earlier study, sgk mRNA level was found to be increased at the lesion site after brain injury, suggesting that sgk probably is involved in axonal regeneration (13). Transcriptionally regulated sgk also shows a tissue- and stage-specific expression pattern during embryogenesis and postnatal development (13, 23). In the present study, we first have demonstrated that sgk plays an important role in memory consolidation of spatial learning in rats. Northern blot analysis revealed that hippocampal sgk mRNA level was approximately 4-fold higher in fast learners than slow learners in the Morris water maze task. In situ hybridization results further revealed significant increases in sgk mRNA level in CA1, CA3, and the dentate gyrus in fast learners. These results parallel earlier findings, which indicated that single-unit activity in the dorsal hippocampus correlates with spatial learning (24) and that complex-spike cells in the CA1 area have a high specificity to place兾direction during radial arm maze learning in rats (25). The increase of sgk mRNA level in the dentate gyrus is also consistent with the report concluding that spatial training enhances neurogenesis in rat dentate gyrus (26). Further, we have demonstrated that transient transfection of the dominant negative sgk mutant DNA to CA1 impaired, whereas transfection of the sgk wild-type DNA enhanced, water maze performance in rats. Transfection of sgk mutant DNA did not alter spatial learning performance under visible platform, suggesting that sgk did not affect sensory and motor functions in rats. These results not only demonstrate the importance and specificity of sgk expression involved in memory consolidation but also support the notion that the dorsal hippocampus plays a critical role in spatial learning and memory (11, 27). In the present study, the averaged area of transfection in CA1 is approximately 20% viewed from a single plane and much less than 20% of total CA1 neurons in the dorsal hippocampus. Moser et al. (27) previously have reported that a minimum volume of 20% lesion in the dorsal hippocampus is needed to produce spatial learning deficit in rats. How does sgk DNA transfection in such a small group of neurons produce more significant behavioral changes than lesion? One possible explanation is that, in the case of lesion, the remaining 80% of neurons compensates for the lost neurons and maintains the neural network for memory encoding. In the case of sgk mutant DNA-transfected neurons, memory encoding still may take place at steps upstream of SGK, but the molecular alteration of sgk may prevent the downstream signaling pathway that is necessary to stabilize and maintain memory. This explanation is supported by our results that, in SGKS422A-transfected animals, spatial learning performance is similar to that of control animals for the first several trials but is impaired significantly afterward (Fig. 3A). The network somehow is tricked to use the defective neurons that were sufficient to degrade the memory trace to a greater extent than if the neurons were not present. The same situation might happen in another learning task in which NMDA receptor blockade does not prevent the beneficial effect of pretraining on subsequent spatial learning (28). One possible Tsai et al.

1. Davis, H. P. & Squire, L. R. (1984) Psychol. Bull. 96, 518–559. 2. Matthies, H. (1989) Annu. Rev. Psychol. 40, 381–404. 3. Lee, E. H. Y., Hung, H. C., Lu, K. T., Chen, W. H. & Chen, H. Y. (1992) Peptides 13, 927–937. 4. Castelluci, V. F., Kennedy, T. E., Kandel, E. R. & Goelet, P. (1988) Neuron 1, 321–328. 5. Tully, T. (1996) Proc. Natl. Acad. Sci. USA 93, 13460–13467. 6. Popov, N., Schulzeck, S., Pohle, W. & Matthies, H. (1980) Neuroscience 5, 161–167. 7. Huang, A. M. & Lee, E. H. Y. (1997) NeuroReport 8, 1619–1624. 8. Huang, A. M., Wang, H. L., Tang, Y. P. & Lee, E. H. Y. (1998) J. Neurosci. 18, 4305–4313. 9. Chang, H. P., Lindberg, F. P., Wang, H. L., Huang, A. M. & Lee, E. H. Y. (1999) Learn. Mem. 6, 448–457. 10. Morris, R. G. M., Garrud, P., Rawlins, J. N. P. & O’Keefe, J. (1982) Nature (London) 297, 681–683. 11. Moser, M. B., Moser, E. I., Forrest, E., Andersen, P. & Morris, R. G. M. (1995) Proc. Natl. Acad. Sci. USA 92, 9697–9701. 12. Abdallah, B., Hassan, A., Benoist, C., Goula, D., Behr, J. P. & Demeneix, B. A. (1996) Human Gene Ther. 7, 1947–1954. 13. Imaizumi, K., Tsuda, M., Wanaka, A., Tohyama, M. & Takagi, T. (1994) Mol. Brain Res. 26, 189–196. 14. Hollister, R. D., Page, K. J. & Hyman, B. T. (1997) Neuroscience 79, 1111–1119. 15. Park, J., Leong, M. L., Buse, P., Maiyer, A. C., Firestone, G. L. & Hemmings, B. A. (1999) EMBO J. 18, 3024–3033. 16. Amaral, D. G. & Witter, M. P. (1995) in The Rat Nervous System, ed. Paxinos, G. (Academic, New York), 2nd Ed., pp. 443–493. 17. Webster, M. K., Goya, L., Ge, Y., Maiyar, A. C. & Firestone, G. L. (1993) Mol. Cell. Biol. 13, 2031–2040. 18. Naray-Fejes-Toth, A., Fejes-Toth, G., Volk, K. A. & Stokes, J. B. (2000) J. Steroid Biochem. Mol. Biol. 75, 51–56. 19. Naray-Fejes-Toth, A., Canessa, C., Cleaveland, E. S., Aldrich, G. & Fejes-Toth, G. (1999) J. Biol. Chem. 274, 16973–16978. 20. Pearce, D., Verrey, F., Chen, S. Y., Mastroberardino, L., Meijer, O. C., Wang, J. & Bhargava, A. (2000) Kidney Inter. 57, 1283–1289. 21. Waldegger, S., Barth, P., Raber, G. & Lang, F. (1997) Proc. Natl. Acad. Sci. USA 94, 4440–4445.

Tsai et al.

induced gene (18), plays a critical role in the process of memory consolidation of spatial learning in rats. These results have provided direct evidence that enhanced sgk expression facilitates memory consolidation in mammals. These results are also in agreement with reports showing abundant expressions of GR (38, 39) and GR mRNA in the hippocampus (40). Other supports come from evidence showing that corticosterone enhances cognitive function in animals when GR and mineralocorticoid receptor (MR) are activated in a balanced way (41) and that corticosterone up-regulates sgk mRNA expression in vivo (42). On the other hand, short-term blockade of GR impairs, whereas continuous blockade of GR facilitates, spatial learning and memory in rats (43). These findings are not inconsistent with the present results because the duration of behavioral training adopted in the present study parallels the phase of short-term GR activation in another study. However, the present results are incongruent with the observation that GR activation suppresses long-term potentiation in the CA1 area of the hippocampus (44). This discrepancy remains to be clarified. Accumulative evidence has shown that moderate concentrations of glucocorticoid enhance memory consolidation in a variety of learning tasks in animals. Sgk is known as a primary glucocorticoid-induced gene. In the present study, we have demonstrated that enhanced sgk expression facilitates memory consolidation of spatial learning in rats. These results bridge the relationships among glucocorticoid, sgk expression, and memory consolidation. This work was supported by Grant NSC 89-2321-B-001–007 from the National Science Council of Taiwan, Republic of China. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34.

35. 36. 37. 38. 39. 40. 41. 42. 43. 44.

Delmolino, L. M. & Castellot, J. J., Jr. (1997) J. Cell. Physiol. 173, 371–379. Lee, E., Lein, E. S. & Firestone, G. L. (2001) Mech. Dev. 103, 177–181. Olton, D. S., Branch, M. & Best, P. J. (1978) Exp. Neurol. 58, 387–409. McNaughton, B. L., Barnes, C. A. & O’Keefe, J. (1983) Exp. Brain Res. 52, 41–49. Gould, E., Beylin, A. V., Tanapat, P., Reeves, A. & Shors, T. J. (1999) Nat. Neurosci. 2, 260–265. Moser, E., Moser, M. B. & Andersen, P. (1993) J. Neurosci. 13, 3916–3925. Bannerman, D., Good, M. A., Butcher, S. P., Ramsay, M. & Morris, R. G. M. (1995) Nature (London) 378, 182–186. Moser, M. B. & Moser, E. I. (2000) Neuron 26, 559–561. Sandi, C. (1998) Neural Plast. 6, 41–52. Roozendaal, B. (2000) Psychoneuroendocrinology 25, 213–238. Oitzl, M. S. & de Kloet, E. R. (1992) Behav. Neurosci. 108, 62–71. Roozendaal, B., Portillo-Marquez, G. & McGaugh, J. L. (1996) Behav. Neurosci. 110, 1074–1083. Cole, T. J., Blendy, J. A., Monaghan, A. P., Krieglstein, K., Schmid, W., Aguzzi, A., Fantuzzi, G., Hummler, E., Unsicker, K. & Schu ¨tz, G. (1995) Genes Dev. 9, 1608–1621. Oitzl, M. S., de Kloet, E. R., Joels, M., Schmid, W. & Cole, T. J. (1997) Eur. J. Neurosci. 9, 2284–2296. Oitzl, M. S., Reichardt, H. M., Joels, M. & de Kloet, E. R. (2001) Proc. Natl. Acad. Sci. USA 98, 12790–12795. Rousse, I., Beaulieu, S., Rowe, W., Meaney, M. J., Barden, N. & Rochford, J. (1997) NeuroReport 8, 841–845. Reul, J. M. H. M. & de Kloet, E. R. (1985) Endocrinology 117, 2505–2512. van Steensel, B., van Binnendijk, E. P., Hornsby, C. D., van der Voort, H. T., Krozowski, Z. S., de Kloet, E. R. & van Driel, R. (1996) J. Cell Sci. 109, 787–792. Morimoto, M., Morita, N., Ozawa, H., Yokoyama, K. & Kawata, M. (1996) Neurosci. Res. 26, 235–269. de Kloet, E. R., Oitzl, M. S. & Joels, M. (1999) Trends Neurosci. 22, 422–426. Brennan, F. E. & Fuller, P. J. (2000) Mol. Cell Endocrinol. 166, 129–136. Oitzl, M. S., Fluttert, M., Sutanto, W. & de Kloet, E. R. (1998) Eur. J. Neurosci. 10, 3759–3766. Pavlides, C., Ogawa, S., Kimura, A. & McEwen, B. S. (1996) Brain Res. 738, 229–235.

PNAS 兩 March 19, 2002 兩 vol. 99 兩 no. 6 兩 3995

NEUROBIOLOGY

explanation is that the NMDA receptor blockade is incomplete and the few unblocked receptors are sufficient for synaptic plasticity and behavioral plasticity to occur (29). Alternatively, the defective neurons may convey the wrong signaling information to other neurons through crosstalk and, consequently, may amplify the effect of sgk mutant DNA. However, this does not occur in the case of lesion. The exact mechanism through which transfection to a small group of neurons affects learning remains to be elucidated. Low to moderate levels of glucocorticoid are known to facilitate memory consolidation and memory storage in a variety of learning tasks in chicks, mice, and rats (30, 31). Removal of endogenous corticosterone by adrenalectomy, on the other hand, impairs memory consolidation in water maze learning (32, 33). This behavioral deficit can be reversed by the administration of the synthetic glucocorticoid dexamethasone (33), suggesting that endogenously released glucocorticoid plays an important role in regulating the memory consolidation process. In more recent years, through the generation of glucocorticoid receptor (GR)-deficient mice (34), Oitzl et al. have demonstrated an impairment in spatial learning and memory in these GRdeficient animals (35). In mice with point mutation of GR, spatial memory also was impaired (36). In another study, transgenic mice bearing the GR antisense construct showed impairment in both water maze and radial arm maze spatial learning tasks (37). These results suggest that memory consolidation depends on the activation of the GR-signaling pathway. However, the molecular mechanism underlying GR-mediated memory facilitation is unknown. In the present study, we have demonstrated that the sgk gene, a primary glucocorticoid-