Mitotic and neurogenic effects of dehydroepiandrosterone (DHEA) on human neural stem cell cultures derived from the fetal cortex

Masatoshi Suzuki*†, Lynda S. Wright*, Padma Marwah‡, Henry A. Lardy‡, and Clive N. Svendsen*§

*Departments of Anatomy and Neurology and the Waisman Center, University of Wisconsin, 1500 Highland Avenue, Madison, WI 53705-2280; †Department of Veterinary Physiology, University of Tokyo, 1-1-1 Yayoi, Bunkyo-ku, Tokyo 113-8657, Japan; and ‡Department of Biochemistry and Institute for Enzyme Research, University of Wisconsin, 1710 University Avenue, Madison, WI 53726

Contributed by Henry A. Lardy, December 29, 2003 Dehydroepiandrosterone (DHEA) is a neurosteroid with potential very low (6, 11). Thus, there is some question as to the relevance effects on neurogenesis and neuronal survival in humans. How- of the effect of DHEA on rodents and how this relates to the ever, most studies on DHEA have been performed in rodents, and human situation. there is little direct evidence for biological effects on the human Recent studies have shown that neural precursor cells can be nervous system. Furthermore, the mechanism of its action is derived from the developing brain allowing the direct examina- unknown. Here, we show that DHEA significantly increased the tion of proliferation, differentiation, and migration of these cells growth rates of human neural stem cells derived from the fetal in the culture dish (12, 13). These cells can be maintained in cortex and grown with both epidermal growth factor (EGF) and culture as spherical aggregates termed neurospheres (14–16) leukemia inhibitory factor (LIF). However, it had no effect on that consist of both multipotent stem cells and more restricted cultures grown in either factor alone, suggesting a specific action progenitor cells responsive to the mitogens epidermal growth on the EGF͞LIF-responsive cell. Precursors of DHEA such as preg- factor (EGF) and fibroblast growth factor 2 (FGF-2) (17). nenolone or six of its major metabolites, had no significant effect Neurospheres are regionally specified based on the brain region on proliferation rates. DHEA did not alter the small number (<3%) from which they were isolated (18, 19) and also differ between of newly formed neuroblasts or the large number (>95%) of species (20). However, recently, we have grown neurospheres nestin-positive precursors. However, the number of glial fibrillary from the human fetal cortex for extended periods of time in EGF acidic protein-positive cells, its mRNA, and protein were signifi- and LIF and shown that they represent a fairly homogenous cantly increased by DHEA. We found both N-methyl-D-aspartate population of cells immunoreactive for the stem cell marker and sigma 1 antagonists, but not GABA antagonists, could com- nestin (21, 22). After a full genomic analysis of these cells, we pletely eliminate the effects of DHEA on stem cell proliferation. have shown them to be stable in culture over time and between Finally we asked whether the EGF͞LIF͞DHEA-responsive stem cells lines with regard to global gene expression and termed them had an increased potential for neurogenesis and found a 29% long-term human neural stem cells derived from the cortex increase in neuronal production when compared to cultures grown (ltNSCCTX) (22). in EGF͞LIF alone. Together these data suggest that DHEA is in- Given the species differences in DHEA and stem cell biology, volved in the maintenance and division of human neural stem cells. ltNSCCTX provide an ideal and innovative model to understand Given the wide availability of this neurosteroid, this finding has developmental and molecular mechanisms of neurosteroid ac- important implications for future use. tions in the human central nervous system. We show here that DHEA, but not its derivatives, increased the cell proliferation of ehydroepiandrosterone (DHEA) and its sulfate are the ltNSCCTX in vitro. We also established which receptor signaling Dmost abundant steroids in the blood of young adult humans. mediated this DHEA action, and finally showed that ltNSCCTX Levels of DHEA peak at Ϸ20 years of age and then decline to grown in the presence of DHEA underwent increased neuro- reach values of 20–30% at 70–80 years of age (1). Significant genesis. These results have important implications for brain interest has arisen from the hypothesis that declining DHEA repair and are relevant to understanding the actions of DHEA concentrations in adults may serve as an indicator of a number on the human brain. of conditions including the loss of insulin sensitivity, obesity, diabetes, cardiovascular diseases, stress, and aging (2). In clinical Materials and Methods studies, DHEA levels were found to decline in mental illnesses Neuronal Progenitor Cell Culture. Human embryonic tissue was such as major depressive disorder or in systematic diseases that obtained from the Birth Defects Laboratory at the University of respond to DHEA supplementation (3, 4). Washington, Seattle. Neurospheres were generated from three While DHEA represents the most abundant steroid product of cortex samples, M006 (13 weeks postconception male), M007 (18 the adrenal cortex, it has also been identified as a ‘‘neuroste- weeks postconception male), and M009 (16 weeks postconcep- roid.’’ Neurosteroids are synthesized de novo in the central tion male). The method of collection conformed to the guide- nervous system independent, at least in part, of peripheral organ lines recommended by National Institutes of Health for the activity (5, 6). Neurosteroids possess the ability to affect neurons collection of such tissues and set out by the University of ␥ through activation of -aminobutyric acid (GABA)A, N-methyl- Washington and the University of Wisconsin, Madison. Institu- D-aspartate (NMDA), and sigma receptors (5, 6), although the exact role of each with regard to DHEA is not well established. DHEA has been shown to be neuroprotective after oxidative Abbreviations: DHEA, dehydroepiandrosterone; EGF, epidermal growth factor; GABA, ␥-aminobutyric acid; GFAP, glial fibrillary acidic protein; LIF, leukemia inhibitory factor; stress in rat hippocampal neuronal cultures (7) and hippocampal ltNSCCTX, long-term cortical neural stem cells; MK801, (ϩ)-5-methyl-10,11-dihydro-5H- damage induced by NMDA (8). DHEA has also been shown to dibenzo[a,d]cyclohepten-5,10-imine hydrogen malate; NMDA, N-methyl-D-aspartate; be a potential signaling molecule in neuronal differentiation TuJ1, ␤-tubulin III. during development (9) and has recently been shown to increase §To whom correspondence should be addressed at: T611 Waisman Center, 1500 Highland neurogenesis in the adult rodent hippocampus (10). In contrast Avenue, Madison, WI 53705. E-mail: [email protected]. to the human brain, levels of DHEA in the adult rat brain are © 2004 by The National Academy of Sciences of the USA

3202–3207 ͉ PNAS ͉ March 2, 2004 ͉ vol. 101 ͉ no. 9 www.pnas.org͞cgi͞doi͞10.1073͞pnas.0307325101 Downloaded by guest on September 28, 2021 Fig. 1. Schematic describing culture schedules and treatments for ltNSCctx. Neurospheres were generated from three fetal cortex samples, initially grown for 4 weeks in EGF and FGF-2. The cultures were then switched to medium supplemented with EGF alone. After Ͼ10 weeks of growth, the cultures were supplemented with EGF and LIF and maintained for a further 4 weeks. Thirty to 40 neurospheres were cultured in coated T25 flasks with basal medium in the presence or absence of EGF, LIF, and DHEA. After 9 days, neurospheres were pulsed with 0.2 ␮M BrdUrd for 14 h and then dissociated into a single cell suspension with enzyme. Dissociated cells were plated onto poly(L)-lysine͞laminin-coated coverslips for 1 h. To check the number of cells immunostained with BrdUrd or other cell markers, cells were fixed 1 h after plating. For differentiation studies, the plated cells were differentiated for 7 days under serum-free conditions and then visualized with anti-TuJ1, GFAP, and Hoechst-labeled nuclei.

tional Review Board approval was obtained for all of these Systems) were used at a final concentration of 10 ng͞ml. The studies. NMDA receptor antagonist (ϩ)-5-methyl-10,11-dihydro-5H- The schematic illustration of the experimental schedule and dibenzo[a,d]cyclohepten-5,10-imine hydrogen malate (MK801) ␮ treatments is summarized in Fig. 1. Cortical progenitors were and GABAA receptor antagonist bicuculline (10 M each in isolated from embryonic brain and induced to proliferate as medium) were purchased fromR&DSystems. Sigma 1 receptor free-floating neurospheres (16). Freshly isolated tissue was antagonists BD1063 or haloperidol (Tocris Cookson, Ellisville, mechanically chopped into small cubes using a McLwain tissue MO) were used at a final concentration of 3 ␮M. chopper (Mickle Laboratory Engineering, Surrey, England) and seeded into T75 flasks at an approximate density of 200,000 cells Immunocytochemistry. Cultures were fixed in ice cold methanol (eight cubes) per ml of basal medium [DMEM͞Ham’s F-12 (7:3) for 20 min or 4% paraformaldehyde (PFA in PBS) at room containing penicillin͞streptomycin͞amphotericin B (PSA, 1%)] temperature and washed in PBS. The cells were then incubated supplemented with B27 (2%; Life Technologies), EGF (20 in 2 M HCl for 20 min at 37°C. The acid was neutralized with 0.1 ng͞ml; Sigma), and FGF-2 (20 ng͞ml;R&DSystems) with M sodium borate and washed several times in PBS. Fixed cells heparin (5 ␮g͞ml). All cultures were maintained in a humidified were blocked in 5% goat serum with 0.3% Triton X-100 and incubator (37°C) and half the growth medium was replenished incubated with anti-BrdUrd antibodies (rat monoclonal, 1:500, every 4–5 days. For passaging of neurospheres, a chopping Accurate Chemical, Westbury, NY) for 30 min. After incubation method was used that does not require trypsin or mechanical with the primary antibodies, either fluorescein or Cy3 (goat dissociation, and cell͞cell contact was continuously maintained anti-rat IgG, 1:1,000, The Jackson Laboratory) was used to (16). Neurospheres were passaged every 14 days by chopping visualize the signal. Hoechst 33258 (0.5 ␮g͞ml in PBS, Sigma) them into 200-␮m cubes that were seeded into fresh growth was added for 5 min after completion of the secondary antibody medium at a density equivalent to Ϸ100,000 cells per ml. At 2 incubation as a nuclear stain. For labeling of cell markers, fixed weeks after the first passage, cells were switched to basal medium cells were incubated for 30 min with primary antibodies to nestin ͞ ␤ containing N2 supplement (1%; Life Technologies) and 20 ng ml (monoclonal 1:500, Chemicon), -tubulin-III (TuJ1; monoclonal EGF. After 10 weeks of growth, 10 ng͞ml LIF (Chemicon) was IgG2b, 1:500, Sigma), or glial fibrillary acidic protein (GFAP) added to the cultures, which were then grown for a further 4 (rabbit polyclonal, 1:1,000, DAKO) to label neural progenitor weeks before analysis. cells, neurons, or astrocyte͞stem cells, respectively. For double- labeling of BrdUrd in conjunction with either TuJ1 or GFAP, all Cumulative BrdUrd Pulse Labeling. To identify newly dividing cells steps of cell marker staining procedure were completed before in neurospheres, BrdUrd pulse-labeling was performed as de- commencement of staining for BrdUrd. scribed (22). Briefly, neurospheres were collected, medium was removed, and cells were washed three times with DMEM. Thirty Cell Counts. Cell counts were performed by using a Nikon ϫ to 40 neurospheres were transferred to poly(2-hydroxyethyl fluorescence microscope (40 objective) and METAMORPH im- NEUROSCIENCE methacrylate)-coated T25 flasks containing basal medium. After 9 days of culture with feeding every 3 days, neurospheres were Table 1. Effects of DHEA and its derivatives on BrdUrd pulsed with 0.2 ␮M BrdUrd (Sigma) for 14 h, and dispersed into incorporation in human neuropheres a single cell suspension with Accutase (Innovative Cell Tech- nologies, San Diego) for 10 min at 37°C. Dissociated cells (25,000 Common % BrdUrdϩ cells per 50 ␮l) were plated onto poly(L)-lysine͞laminin-coated Steroid name cells glass coverslips in 24-well plates and incubated for 60 min at No steroid 13.8 Ϯ 1.7 37°C. The plating medium consisted of basal medium with B27 3␤-hydroxyandrost-5-en-17-one DHEA 23.6 Ϯ 1.5* (2%; Life Technologies). For differentiation studies, dissociated 4-pregnene-3,20-dione Pregnenolone 17.1 Ϯ 0.8 cells were continuously maintained in the plating medium and Androst 5-ene-3␤,17␤-diol Androstenediol 17.3 Ϯ 0.6 differentiated for 7 days. These serum-free conditions allowed 3␤,7␣-dihydroxyandrost-5-en-17-one 7␣-hydroxy DHEA 15.1 Ϯ 1.7 for differentiation of dissociated cells (23). 3␤,7␤-dihydroxyandrost-5-en-17-one 7␤-hydroxy DHEA 14.0 Ϯ 2.1 DHEA was purchased from Steraloids (Newport, RI). DHEA 3␤-hydroxyandrost-5-ene-7,17-dione 7-oxo-DHEA 12.9 Ϯ 1.7 derivatives listed in Table 1 were synthesized and characterized Androst-5-ene-3␤,7␤,17␤-triol 19.1 Ϯ 2.8 as described (24, 25). Stock solutions (10 mM) were prepared in Androst-5-ene-3␤,16␣,17␤-triol 15.6 Ϯ 2.1 100% ethanol and added to give a final medium concentration of 1 ␮M. Ciliary neurotrophic factor (CNTF) or IL-6 (R & D *P Ͻ 0.05 versus all others.

Suzuki et al. PNAS ͉ March 2, 2004 ͉ vol. 101 ͉ no. 9 ͉ 3203 Downloaded by guest on September 28, 2021 aging software (Universal Imaging, Downingtown, PA). Quan- tification of cells was based on counting the number of Hoechst- stained nuclei and the specific immunostained cells in at least four independent fields (total area Ͼ25 mm2) from a minimum of three coverslips. All cell-counting data were expressed as means Ϯ SEM across three separate cell lines derived from the cortex. One-way functional ANOVA with Newman–Keuls post hoc test was used for analysis. Differences were considered significant when P was Ͻ0.05.

RT-PCR. Total RNA was isolated from neurospheres by using TRIzol reagent (Life Technologies). For RT-PCR, we used a GeneAmp RNA PCR kit (Roche Molecular Systems). All procedures for total RNA isolation and RT-PCR were per- formed according to the procedure suggested by the manufac- turers. The following primer pairs were used: human GFAP (26), NMDA receptor subunits NR1 and NR2A-D (27), and sigma 1 receptor (28). The primers of cyclophylin were designed on the basis of the sequences of human cDNA (GenBank accession no. Y00052; forward, 5Ј-GTTTGCAGACAAGGTCCCAAA-3Ј; reverse, 5Ј-TGATCTTCTTGCTGGTCTTG-3Ј), and the pre- dicted size of PCR product is 397 bp. The amplification profile was 94°C for 0.5 min, 58°C for 1 min, and 72°C for 1 min. For PCR of cyclophylin and the others, samples were incubated for 25 and 32 cycles, respectively. PCR products were electropho- Fig. 2. DHEA increases the number of BrdUrdϩ cells in ltNSCctx in the SCION IMAGE resed and analyzed with densitometry by using presence of EGF and LIF. When EGF was withdrawn for 9 days, BrdUrdϩ cells software (Scion, Frederick, MD). were not altered by the addition of LIF and DHEA (a). DHEA significantly increased the number of BrdUrdϩ cells in the presence of EGF and LIF, al- Western Blotting. Preparation of cell lysates and Western blotting though no significant effect was observed with DHEA in the absence of LIF (b). were performed on neurospheres, using monoclonal anti-GFAP Photomicrographs represent BrdUrd immunoreactivity in cells grown in EGF (Transduction Laboratories, Lexington, KY) or NR1 (Chemi- (c)orEGF͞LIF͞DHEA (d) for 9 days. (Inset) Arrows designate BrdUrd-labeled cells at higher magnification. (Scale bar, 20 ␮m.) con), as described (22). Blots were quantitated by SCION IMAGE. Results and Discussion Other Cytokines Cannot Substitute for LIF in Making EGF Cells Re- ͞ DHEA Significantly Increases Proliferation of EGF LIF Responsive Cells sponsive to DHEA. LIF is a member of a family of IL-6 cytokines ctx Within ItNSC Cultures. The three lines generated for this study that also includes ciliary neurotrophic factor (CNTF), oncosta- grew as neurospheres as described in detail previously (16, 22). tin, and cardiotrophin 1 (29). These factors signal through a As predicted the cultures became responsive to LIF after 10 ͞ receptor complex made up of unique components specific for the weeks and from this point were maintained in EGF LIF. To particular cytokines and a common receptor subunit, gp130, but confirm the differentiation potential of these lines, individual do not have the same effect as LIF on the proliferation of ͞ neurospheres were dissociated and plated to laminin poly(L)- EGF-responsive neural stem cells (22). The addition of CNTF or lysine coated coverslips for 7 days under differentiating condi- IL-6 could not mimic the effect of LIF with regard to enabling tions. After fixation and immunostaining for the neuronal the EGF-responsive cells to respond to DHEA (see Fig. 6, which ͞ marker TuJ1 and astrocyte stem cell marker GFAP, these is published as supporting information on the PNAS web site). cultures were able to produce large numbers of well differenti- ated neurons and astrocyte͞stem cells with a more fibroblast-like Other Active Components of the DHEA Pathway Do Not Affect the morphology (data not shown). Growth Rate of ltNSCctx. DHEA is readily converted metabolically We first asked what role DHEA may play in the growth of to derivatives that are more active in some assays (24, 30, 31). In these ltNSCctx cultures. To establish whether DHEA could rat liver homogenate, the sequence is DHEA 3 7␣-OH-DHEA enhance growth rates in the absence of EGF or LIF, we designed 3 7-oxo-DHEA 3 7␤-OH-DHEA 3 androst-5-en-3␤,7␤,17␤- a withdrawal experiment (Fig. 1). EGF, LIF or both factors were triol. In the human brain, DHEA is synthesized from preg- withdrawn from the medium in the presence or absence of nenolone and metabolized mainly to 7␣-OH-DHEA and an- DHEA for 9 days. In the complete absence of EGF, the numbers drost-5-ene-3␤,17␤-diol (androstenediol) (6, 32). To determine ϩ of BrdUrd cells tended to be increased in the presence of either whether DHEA per se or one of its metabolites stimulated cell LIF or DHEA, although this did not reach significance (Fig. 2a). growth and incorporation of BrdUrd, neurospheres were cul- This increase may have been due to trace amounts of EGF tured in the continual presence of individual steroids (1 ␮M; remaining associated with membranes. In the presence of EGF, listed in Table. 1) with basal medium containing EGF͞LIF for LIF significantly increased BrdUrd incorporation (Fig. 2b)aswe 9 days as described previously. The cultures were then pulse- have shown previously (22), whereas DHEA again tended to labeled with BrdUrd, dissociated, plated for 1 h, and stained with increase BrdUrd incorporation but did not reach significance BrdUrd antibody. Neither pregnenolone nor any of the six .(Fig. 2b). However, in the presence of EGF͞LIF, DHEA sig- metabolites affected the number of BrdUrd؉ cells (Table 1) ϩ nificantly increased the number of BrdUrd cells from Ϸ14% to These experiments suggest that the effects reported here are due 24% (P Ͻ 0.05; Fig. 2 b–d). These data show DHEA selectively to the DHEA introduced into the culture medium, not to the increases the division of the EGF͞LIF-responsive cell within the metabolism of DHEA, because no derivative tested in this neurospheres. experiment was more effective.

3204 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0307325101 Suzuki et al. Downloaded by guest on September 28, 2021 80% to 90% (Fig. 3a; significantly different from EGF͞LIF at P Ͻ 0.05). To further determine whether newly dividing cells expressed GFAP, we performed double labeling with GFAP and BrdUrd antibody (Fig. 3 b–d). In the presence of LIF, DHEA significantly increased the number of GFAPϩ͞BrdUrdϩ double- positive cells by Ͼ30% (P Ͻ 0.05; Fig. 3b). Furthermore, we found that Ͼ95% of BrdUrdϩ cells were also immunostained by GFAP antibody, suggesting that in the presence of EGF͞LIF͞ DHEA the majority of cells dividing within the neurospheres were GFAP-positive. We next determined absolute RNA and protein levels of GFAP in response to LIF and DHEA by using RT-PCR and Western blot. ltNSCctx were grown in basal medium with EGF in the presence or absence of LIF and DHEA for 9 days, followed by total RNA and protein isolation. Semiquantitative analysis of RT-PCR for GFAP revealed that LIF͞DHEA treatment in- creased GFAP mRNA (Fig. 3e). Densitometric analysis showed that the level of GFAP expression was 1.3 times higher in LIF͞DHEA͞EGF compared to LIF͞EGF alone (Fig. 3e). West- ern blot analysis confirmed the effect of DHEA on GFAP protein expression (Fig. 3f). Using the densitometric analysis of each band, we found that DHEA could increase GFAP protein by Ϸ40% when compared to LIF͞EGF alone (Fig. 3f). We have previously shown that LIF also has powerful effects on GFAP expression in ltNSCctx (22), presumably through the gp130 receptor system and the Janus kinase-signal transducer and activators of transcription (JAK-STAT) signaling pathway (33, 34). In the current study we confirm these findings and show that although DHEA had no direct effect on EGF-treated cultures, it increased both GFAP expression and the number of GFAP͞BrdUrd-positive cells when compared to EGF͞LIF. Fig. 3. DHEA dynamically regulates GFAP expression. (a) The numbers of GFAPϩ cells treated with LIF (P Ͻ 0.05) and LIF͞DHEA (P Ͻ 0.01) were signif- Interestingly, neural stem cells capable of forming large numbers icantly increased compared to that in the absence of both LIF and DHEA. (b) of neurons in adult mammals have been identified as astrocyte- The percentages of double-staining with GFAP and BrdUrd antibody after resembling cells that are GFAP- and nestin-positive (35, 36). 9-day culture with LIF and DHEA. Photomicrographs of cells cultured in the Given that the DHEA cultures showed increased GFAP pro- absence (c) and presence (d) of LIF͞DHEA for 9 days show a difference in the duction and neurogenesis, our results lend further support to the number and intensity of GFAPϩ͞BrdUrdϩ cells. (Scale bar, 15 ␮m.) (e) Semi- idea that an ‘‘astrocyte-like’’ cell may act as a stem cell in the quantitative RT-PCR revealed that LIF͞DHEA treatment increased GFAP mRNA human central nervous system (36–39). compared to LIF alone. P, a positive control using total RNA isolated from human embryonic brain; N, a negative control, reverse transcriptase with- DHEA Acts Through both NMDA and Sigma Receptors in ltNSCctx. We drawn during RT step. (f) Western blot analysis confirmed the effect of DHEA on GFAP protein expression. next asked which receptors mediated the effect of DHEA in ltNSCctx. It has been reported that neurosteroids are able to modulate NMDA receptor functions through non-genomic ac- The DHEA Responsive Cell Expresses High Levels of GFAP. Although tion (9, 40). NR1 is common to functional NMDA receptors (41) ͞ under our culture conditions neurospheres contain Ͼ95% nes- and was clearly detected in LIF and LIF DHEA treatments, tin-positive precursor cells we consider stem cells, there remain whereas it was barely detectable in the absence of LIF (Fig. 4a). In contrast, NR2B mRNA was detected in all cultures and not a small number of neuroblasts that spontaneously arise within ͞ these cultures (22). To determine whether the ratio of neuro- differentially expressed in the presence or absence of LIF blasts to stem cells was affected by DHEA treatment (i.e., which DHEA, whereas no NR2A, NR2C, or NR2D mRNA was detected in any of the neurosphere cultures (Fig. 4a). When type of cells were proliferating in response to DHEA), acutely

using densitometric analysis of each band, the levels of NR1 NEUROSCIENCE dissociated neurospheres were immunostained for the neuronal mRNA were found to be Ͼ10-fold greater in the presence of marker TuJ1 and the precursor cell marker nestin (see Fig. 7, LIF͞DHEA when compared to EGF alone (data not shown). which is published as supporting information on the PNAS web Furthermore, Western blot analysis using anti-NR1 antibody site). The number of TuJ1ϩ cells seen at this time (Ͻ3% of the ϩ Ͼ revealed that NR1 proteins were strongly induced by LIF and population) and nestin cells ( 95% of the population) were not LIF͞DHEA treatments (Fig. 4b). Thus, the functional subunits affected by LIF or DHEA, suggesting that DHEA did not induce of NMDA receptor on cells within ltNSCctx could be induced by the division of a cell that spontaneously generated neuroblasts LIF and LIF͞DHEA. within the neurospheres. Rather, it was likely that the DHEA- We next determined whether inhibition of NMDA receptors responsive cell remained as a stem cell under these conditions. using the antagonist MK801 could block the effects of DHEA on Our previous results also indicated that the expression of proliferation. The addition of 10 ␮M MK801 in the medium with GFAP was positively regulated by LIF in ltNSCctx (22). We EGF͞LIF for 9 days did not change the basal levels of BrdUrdϩ hypothesized that DHEA treatment with LIF might further cell number (Fig. 4c). However, the increased proliferation enhance GFAP expression in ltNSCctx. As expected, LIF in- induced by DHEA͞LIF based on BrdUrdϩ cell number could be creased the number of GFAP-positive cells over EGF alone (Fig. completely blocked by MK801 (Fig. 4c). These effects of MK801 3a; P Ͻ 0.05). DHEA alone or with EGF had no effect on the suggested that glutamate receptor activation with either NMDA number of GFAP-positive cells, but in combination with LIF it or glutamate may also be able to enhance proliferation of cells further increased the total number of GFAP-positive cells from within these cultures. To test this, we added glutamate agonists

Suzuki et al. PNAS ͉ March 2, 2004 ͉ vol. 101 ͉ no. 9 ͉ 3205 Downloaded by guest on September 28, 2021 Fig. 5. ltNSCCTX grown in the presence of DHEA generated more maturing neurons. Neurospheres were cultured in the presence or absence of EGF, LIF, and DHEA for 9 days, pulsed with BrdUrd, plated, and immunostained with anti-TuJ1, anti-GFAP, and Hoechst-labeled nuclei. (a) TuJ1ϩ cells were signif- icantly increased by simultaneous treatment of LIF and DHEA, although the numbers of TuJ1ϩ cells were not altered by either DHEA or LIF alone. Pho- Fig. 4. NMDA and sigma 1 receptor signaling are involved in the effect of tomicrographs represent TuJ1 immunoreactivity (red) in differentiated cells DHEA. (a) Gene expression of NMDA receptor subunits. Positive (P) and grown in the absence (d) or presence (e)ofLIF͞DHEA for 9 days. (b) There were negative (N) controls were prepared similarly to the protocol mentioned in no significant differences in the percentage of GFAP-labeled cells. (c) The Fig. 3. (b) Western blot analysis confirmed the effect of LIF on NR1 protein numbers of TuJ1ϩ͞BrdUrdϩ double-positive cells were not affected by either expression in ltNSCctx.(c) Specific NMDA receptor antagonist MK-801 (10 ␮M) DHEA or LIF alone. However, TuJ1ϩ͞BrdUrdϩ double-positive cells were in- significantly blocks the effect of DHEA. (d) Sigma 1 receptors are also involved creased by LIF͞DHEA. (f) After 7 days of differentiation, many TuJ1 (green) in the effect of DHEA. Neurospheres were cultured for 9 days in basal medium and BrdUrd (red) double-positive cells (designated by arrows) were found in with EGF, LIF, and DHEA, plus 3 ␮M BD1063 or haloperidol. (e) The blockade cells grown in the presence of LIF͞DHEA for 9 days. *, P Ͻ 0.01 versus LIF or ctx Ͻ ␮ of GABAA receptors has no effect in ltNSC . *, P 0.05 versus LIF without DHEA alone. (Scale bar, 40 m.) DHEA. NS, not significant.

effects of DHEA reported here in vitro, although we cannot to the culture medium. Glutamate (0.1 ␮M) and NMDA were exclude its importance in other species or in vivo. both found to significantly increase the number of BrdUrdϩ cells, suggesting a direct effect on stem cell proliferation (see Fig. ltNSCctx Grown in DHEA Undergo a Greater Level of Neurogenesis 8, which is published as supporting information on the PNAS After Differentiation. We hypothesized that LIF͞DHEA treat- web site). ment may alter the potential of ltNSCctx to generate neurons. To Sigma receptors (a type of opiate receptor) may be able to test this possibility, neurospheres in the continual presence or modify NMDA receptor signaling for cellular effects of neuro- absence of LIF or LIF͞DHEA were cultured in basal medium steroids (42, 43). To test this possibility, we first checked sigma containing EGF for 9 days. After dissociation and plating onto 1 receptor expression in ltNSCctx by using RT-PCR. Sigma 1 laminin-coated coverslips, the cultures were allowed to differ- receptor mRNA was detected in all neurosphere cultures and entiate over 7 days. During this time new neurons are born was not seen to be affected by either LIF or DHEA (data not through the proliferation of neuronal progenitors (T. Ostenfeld shown). To establish whether these receptors were important for and C.N.S., unpublished observations). The differentiated cells the effects of DHEA, neurospheres were cultured for 9 days in were immunostained for TuJ1 (Fig. 5 a, d, and e) and GFAP (Fig. basal medium with EGF͞LIF or EGF͞LIF͞DHEA with or 5b). When the cultures were maintained in the absence of without 3 ␮M BD1063 or haloperidol, which specifically antag- LIF͞DHEA for 9 days, 10% of total cells were TuJ1ϩ after onize sigma 1 receptor functions (42). The blockade of sigma differentiation for 7 days (Fig. 5a). DHEA alone during the receptor again completely abolished the effects of DHEA on growth phase had no effect, whereas LIF treatment tended to proliferation (Fig. 4d). These experiments suggested that increase the number of TuJ1ϩ cells (P ϭ 0.07). In contrast, NMDA receptor signaling after sigma 1 receptor-activation combined LIF͞DHEA treatment significantly increased the might involve DHEA actions on these cells. number of TuJ1ϩ cells after differentiation (P Ͻ 0.05 over EGF ϩ ϩ Previous reports indicate that GABAA receptors are another alone; Fig. 5a). In contrast to TuJ1 cells, the number of GFAP possible candidate other than the NMDA receptor to mediate cells was not significantly different between the groups (Fig. 5b). various functions of DHEA in the central nervous system (44, This finding suggests that the increases seen in GFAP levels at 45). We next tested the role of GABAA receptors in the cell 1 h postplating (Fig. 3) in response to DHEA disappear when the proliferation induced by LIF͞DHEA. Neurospheres were cul- cells exit proliferation and are allowed to differentiate. To tured for 9 days in basal medium with EGF͞LIF or EGF͞LIF͞ determine whether the increased number of new neurons were DHEA with or without the GABAA receptor antagonist bicu- derived from cells induced by DHEA just before plating, neu- culline. Bicuculline (10 ␮M) did not reduce the LIF͞DHEA- rospheres were pulse-labeled with BrdUrd for 14 h under the mediated proliferation (Fig. 4e). Moreover, the treatment of 10 various conditions before plating (Fig. 5 c and f). At 7 days of or 0.1 ␮M GABA for 9 days had no effect on the number of differentiation, the numbers of TuJ1ϩ͞BrdUrdϩ-positive cells BrdUrdϩ cells in ltNSCctx (data not shown). From these results, were not changed by adding DHEA or LIF in the medium before we conclude that GABA receptor signaling is not involved in the plating (Fig. 5c). However, the number of double-positive cells

3206 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0307325101 Suzuki et al. Downloaded by guest on September 28, 2021 was clearly increased after EGF͞LIF͞DHEA treatment com- recently been shown to be neuroprotective (8) and increase pared to EGF͞LIF alone (P Ͻ 0.05; Fig. 5c). These results neurogenesis (10) in the adult rodent hippocampus. Although it indicate that DHEA positively regulates the number of neurons is impossible to test the effects of DHEA on the adult human produced by these cultures. brain in vivo, our results show an effect on human neural stem cells, and this may be relevant to some of the beneficial effects Conclusion of DHEA in clinical studies (3, 4). More importantly, these direct In this study, we show that DHEA can increase proliferation of effects of DHEA (which is currently available as a health human neural stem cells and positively regulate the number of supplement in the United States) on human cells warn against its neurons produced by these cultures. It is known that DHEA long-term use until the consequences of its actions have been amounts fall progressively during aging, and reduced levels of further investigated. DHEA have been reported in both adolescents and adults with major depressive disorders. Interestingly, the adult human brain This work was supported by funds from the University of Wisconsin also has neural stem cells within specific regions that contribute Foundation and by Japan Society for the Promotion of Science Post- to divide and make new neurons (46). Furthermore, DHEA has doctoral Fellowships for Research Abroad (to M.S.).

1. Orentreich, N., Brind, J. L., Vogelman, J. H., Andres, R. & Baldwin, H. (1992) 26. Sanchez-Ramos, J. R., Song, S., Kamath, S. G., Zigova, T., Willing, A., J. Clin. Endocrinol. Metab. 75, 1002–1004. Cardozo-Pelaez, F., Stedeford, T., Chopp, M. & Sanberg, P. R. (2001) Exp. 2. Celec, P. & Starka, L. (2003) Physiol. Res. 52, 397–407. Neurol. 171, 109–115. 3. Bloch, M., Schmidt, P. J., Danaceau, M. A., Adams, L. F. & Rubinow, D. R. 27. Yoshioka, A., Ikegaki, N., Williams, M. & Pleasure, D. (1996) J. Neurosci. Res. (1999) Biol. Psychiatry 45, 1533–1541. 46, 164–178. 4. Roshan, S., Nader, S. & Orlander, P. (1999) Eur. J. Clin. Invest. 29, 210–213. 28. Shamsul, O. M., Moore, P., El Sherbeny, A., Roon, P., Agarwal, N., Sarthy, V. P., 5. Baulieu, E. E. (1997) Recent Prog. Horm. Res. 52, 1–32. Casellas, P., Ganapathy, V. & Smith, S. B. (2001) Mol. Brain Res. 95, 86–95. 6. Baulieu, E. E. (1998) Psychoneuroendocrinology 23, 963–987. 29. Auernhammer, C. J. & Melmed, S. (2000) Endocr. Rev. 21, 313–345. 7. Bastianetto, S., Ramassamy, C., Poirier, J. & Quirion, R. (1999) Mol. Brain Res. 30. Lardy, H., Marwah, A. & Marwah, P. (2002) Lipids 37, 1187–1191. 66, 35–41. 31. Shi, J., Schulze, S. & Lardy, H. A. (2000) Steroids 65, 124–129. 8. Kimonides, V. G., Khatibi, N. H., Svendsen, C. N., Sofroniew, M. V. & Herbert, 32. Weill-Engerer, S., David, J. P., Sazdovitch, V., Liere, P., Schumacher, M., J. (1998) Proc. Natl. Acad. Sci. USA 95, 1852–1857. Delacourte, A., Baulieu, E. E. & Akwa, Y. (2003) Brain Res. 969, 117–125. 9. Compagnone, N. A. & Mellon, S. H. (1998) Proc. Natl. Acad. Sci. USA 95, 33. Bonni, A., Sun, Y., Nadal-Vicens, M., Bhatt, A., Frank, D. A., Rozovsky, I., 4678–4683. Stahl, N., Yancopoulos, G. D. & Greenberg, M. E. (1997) Science 278, 477–483. 10. Karishma, K. K. & Herbert, J. (2002) Eur. J. Neurosci. 16, 445–453. 34. Nakashima, K., Wiese, S., Yanagisawa, M., Arakawa, H., Kimura, N., Hisat- 11. Corpechot, C., Robel, P., Axelson, M., Sjovall, J. & Baulieu, E. E. (1981) Proc. sune, T., Yoshida, K., Kishimoto, T., Sendtner, M. & Taga, T. (1999) Natl. Acad. Sci. USA 78, 4704–4707. J. Neurosci. 19, 5429–5434. 12. McKay, R. (1997) Science 276, 66–71. 35. Garcia-Verdugo, J., Doetsch, F., Wichterle, H., Lim, D. A. & Alvarez-Buylla, 13. Gage, F. H. (2000) Science 287, 1433–1438. A. (1998) J. Neurobiol. 36, 234–248. 14. Chalmers-Redman, R. M., Priestley, T., Kemp, J. A. & Fine, A. (1997) 36. Doetsch, F., Caille, I., Lim, D. A., Garcia-Verdugo, J. M. & Alvarez-Buylla, A. Neuroscience 76, 1121–1128. (1999) Cell 97, 703–716. 15. Reynolds, B. A., Tetzlaff, W. & Weiss, S. (1992) J. Neurosci. 12, 37. Laywell, E. D., Rakic, P., Kukekov, V. G., Holland, E. C. & Steindler, D. A. 4565–4574. (2000) Proc. Natl. Acad. Sci. USA 97, 13883–13888. 16. Svendsen, C. N., ter Borg, M. G., Armstrong, R. J., Rosser, A. E., Chandran, 38. Skogh, C., Eriksson, C., Kokaia, M., Meijer, X. C., Wahlberg, L. U., Wictorin, S., Ostenfeld, T. & Caldwell, M. A. (1998) J. Neurosci. Methods 85, 141–152. K. & Campbell, K. (2001) Mol. Cell. Neurosci. 17, 811–820. 17. Ciccolini, F. & Svendsen, C. N. (1998) J. Neurosci. 18, 7869–7880. 39. Imura, T., Kornblum, H. I. & Sofroniew, M. V. (2003) J. Neurosci. 23, 18. Hitoshi, S., Tropepe, V., Ekker, M. & van der Kooy, D. (2002) Development 2824–2832. (Cambridge, U.K.) 129, 233–244. 40. Wu, F. S., Gibbs, T. T. & Farb, D. H. (1991) Mol. Pharmacol. 40, 333–336. 19. Ostenfeld, T., Joly, E., Tai, Y. T., Peters, A., Caldwell, M., Jauniaux, E. & 41. Monyer, H., Burnashev, N., Laurie, D. J., Sakmann, B. & Seeburg, P. H. (1994) Svendsen, C. N. (2002) Dev. Brain Res. 134, 43–55. Neuron 12, 529–540. 20. Svendsen, C. N., Skepper, J., Rosser, A. E., ter Borg, M. G., Tyres, P. & Ryken, 42. Monnet, F. P., Mahe, V., Robel, P. & Baulieu, E. E. (1995) Proc. Natl. Acad. T. (1997) Dev. Brain Res. 99, 253–258. Sci. USA 92, 3774–3778. 21. Lendahl, U., Zimmerman, L. B. & McKay, R. D. (1990) Cell 60, 585–595. 43. Maurice, T., Phan, V. L., Urani, A., Kamei, H., Noda, Y. & Nabeshima, T. 22. Wright, L. S., Li, J., Caldwell, M. A., Wallace, K., Johnson, J. A. & Svendsen, (1999) Jpn. J. Pharmacol. 81, 125–155. C. N. (2003) J. Neurochem. 86, 179–195. 44. Majewska, M. D., Demirgoren, S., Spivak, C. E. & London, E. D. (1990) Brain 23. Caldwell, M. A., He, X. L., Wilkie, N., Pollack, S., Marshall, G., Wafford, K. A. Res. 526, 143–146. & Svendsen, C. N. (2001) Nat. Biotechnol. 19, 475–479. 45. Demirgoren, S., Majewska, M. D., Spivak, C. E. & London, E. D. (1991) 24. Lardy, H., Kneer, N., Wei, Y., Partridge, B. & Marwah, P. (1998) Steroids 63, Neuroscience 45, 127–135. 158–165. 46. Eriksson, P. S., Perfilieva, E., Bjork-Eriksson, T., Alborn, A. M., Nordborg, C., 25. Marwah, P., Marwah, A., Kneer, N. & Lardy, H. (2001) Steroids 66, 581–595. Peterson, D. A. & Gage, F. H. (1998) Nat. Med. 4, 1313–1317. NEUROSCIENCE

Suzuki et al. PNAS ͉ March 2, 2004 ͉ vol. 101 ͉ no. 9 ͉ 3207 Downloaded by guest on September 28, 2021