Ca²⁺ oscillations induced by testosterone enhance neurite outgrowth

The steroid hormone testosterone controls a vast number of cellular processes including cell growth and differentiation (Beato, 1989; Mooradian et al., 1987). In neurons, this hormone can induce changes at the cellular level, leading to changes in behavior (Kelly et al., 1999). As a neurosteroid, testosterone can influence sleep, the reaction to stress, mood and memory (McEwen, 1991; Naghdi and Asadollahi, 2004). The genomic responses to testosterone are mediated through the intracellular androgen receptor (iAR), a 110 kDa protein with domains for androgen binding, nuclear localization, DNA binding and transactivation (Mooradian et al., 1987). These responses occur following a delay measured in hours. By contrast, several reports indicate that androgens are capable of producing rapid, seconds to minutes, non-genomic effects (Benten et al., 1999a; Lieberherr and Grosse, 1994). The non-genomic actions of androgens are diverse, but common to these early effects is a rapid intracellular Ca²⁺ increase and subsequent activation of Ca²⁺-dependent signaling cascades (Benten et al., 1999b; Estrada et al., 2000; Lieberherr and Grosse, 1994).

By altering the intracellular Ca²⁺ concentration in characteristic ways the cell can use the same signaling molecule to specifically regulate different cellular functions (Carafoli et al., 2001). For example, the rapid turn on and off of Ca²⁺ signals often produces oscillations. The key quantifiable properties of oscillations are frequency and amplitude, and modulation of either property can control cellular processes (Aizman et al., 2001; Dolmetsch et al., 1998; Li et al., 1998). Initial reports showed that the rapid, steroid-induced Ca²⁺ increases were single Ca²⁺ transients (Losel and Wehling, 2003). More recently, it has been shown in skeletal muscle cells that testosterone induces intracellular Ca²⁺ oscillations (Estrada et al., 2005; Estrada et al., 2003; Estrada et al., 2000), which may encode more precise information than alterations in the amplitude of a single Ca²⁺ transient.

An important component of the Ca²⁺ signaling toolkit which is essential for regulation of Ca²⁺ oscillations is the inositol 1,4,5-trisphosphate receptor [Ins(1,4,5)P₃R], a Ca²⁺ permeable channel located in the membrane of the endoplasmic reticulum. The Ins(1,4,5)P₃R has several properties that make it ideal for modulating oscillatory patterns: Ins(1,4,5)P₃-gated channel activity can be regulated by Ca²⁺ with a bell-shaped dependence (Bezprozvanny et al., 1991), by a number of Ca²⁺ binding proteins (Berridge et al., 2003; Choe et al., 2004) and by phosphorylation (Tang et al., 2003). Interestingly, the Ins(1,4,5)P₃R is also found inside the nucleus, associated with the nucleoplasmic reticulum (Echevarria et al., 2003), and this Ins(1,4,5)P₃R can release Ca²⁺ independently of cytoplasmic receptors (Echevarria et al., 2003; Leite et al., 2003; Pusl et al., 2002). It has been suggested that Ca²⁺ levels in the nucleus are responsible for processes related to gene transcription and cell growth whereas changes in cytoplasmic Ca²⁺ are used for processes such as secretion and motility.

In this study we found a new pathway for testosterone activity that has a rapid onset and leads to the generation of long lasting Ca²⁺ oscillations. Within seconds after its addition, testosterone induces intracellular Ca²⁺ increases in neuroblastoma cells, which begin as Ca²⁺ transients initiated in the cytosol and propagate as waves of Ca²⁺ in the cytoplasm and nucleus. These complex Ca²⁺ signals depend on an interplay between Ins(1,4,5)P₃-sensitive stores and influx from the extracellular medium. The Ca²⁺ transients develop into oscillatory patterns which have been shown to activate specific downstream pathways (Dolmetsch et al., 1998; Li et al., 1998). The Ca²⁺ signals are independent of the iAR and are mediated via a pertussis toxin-sensitive plasma membrane receptor. In addition, we found that testosterone enhanced the neurite outgrowth and an increase in cytosolic Ca²⁺ was shown to be required for this effect. These results demonstrate an important physiological mechanism for the action of testosterone in parallel, or prior, to androgen receptor-mediated genomic events in neurons.

Intracellular Ca²⁺ changes were monitored in SH-SY5Y neuroblastoma cells loaded with Fluo-4/AM, a Ca²⁺-senstive, cell permeable fluorescent dye. Addition of testosterone evoked a local intracellular Ca²⁺ increase and then a propagating wave within the cell. The initial Ca²⁺ rise resolved and was followed by subsequent Ca²⁺ increases creating an oscillatory pattern (Fig. 1). Typically, the first increase in intracellular Ca²⁺ occurred at the extremity of the cell, approximately 30 seconds after addition of testosterone (Fig. 1A). A Ca²⁺ wave then spread into the cytosol along the axis of the cell, invaded the nucleus and reached the opposite extremity of the cell. An oscillatory Ca²⁺ response was defined when intracellular Ca²⁺ increases were repeated more than three times within a single cell. In some rare cases, Ca²⁺ oscillations were observed without propagating Ca²⁺ waves. The response to testosterone was concentration dependent. We characterized the testosterone-induce Ca²⁺ response to different levels of the hormone by several parameters: percentage of cells responding, frequency of Ca²⁺ oscillations, and the amplitude of the Ca²⁺ peaks. Concentrations in the 10 nM range evoked Ca²⁺ signals in 30% of the cells (14 of 48 cells from six different cultures; Fig. 1B). When the testosterone concentration was increased to 100 nM, the number of responsive cells increased to 68% (87 of 128 cells, from 16 different cultures) as well as the frequency of oscillatory Ca²⁺ response (Fig. 1C). A similar intracellular Ca²⁺ response has previously been observed in neuroblastoma cells using agonists known to activate G-protein coupled receptors (Tovey et al., 2001).

To measure the regularity of the testosterone-induced Ca²⁺ oscillations we used power spectral analysis (Fig. 1D). This computational method reduces a complex signal into the different sine wave components that contribute to form the complex signal and ranks the components by power so that the most dominant frequencies in the complex signal can be identified (Uhlen, 2004). An added advantage of spectral analysis is that irregular and stochastic contributions to the signal are minimized. We found that the Ca²⁺ oscillations could be described adequately with one major peak (Fig. 1D), indicating a highly regular oscillatory response. Spectral analysis of testosterone-induced Ca²⁺ oscillations generated an average frequency of 18±2 mHz, n=7 (periodicity of 56±5 seconds) and 22±1 mHz, n=34 (periodicity of 46±3 seconds), for 10 nM and 100 nM of testosterone, respectively. The concentration dependence of the amplitude of the first peak showed that elevating concentrations of testosterone significantly increased the Ca²⁺ peaks, reaching a maximum response at 100 nM (Fig. 1E). These concentrations of testosterone are in the range found in normal human males (Kelly et al., 1999; Mooradian et al., 1987). In all subsequent experiments, 100 nM testosterone was used.

The specificity of the response to testosterone was investigated by applying several other steroids. Cortisol, progesterone, or dexamethasone (all at 100 nM) had no effect on intracellular Ca²⁺ levels (data not shown). Interestingly, 100 nM 17β-estradiol (Fig. 1F, red line; n=58 of 86 cells, four independent cultures) produced Ca²⁺ transients whereas at the same concentration the biologically inactive isomer 17α-estradiol (Fig. 1F, green line; n=46 of 46 cells, three independent cultures) did not induce any changes in intracellular Ca²⁺ levels. To eliminate the possibility that the effects of testosterone were due to the activation of an estrogen receptor pathway, cells were pre-incubated with 1 mM tamoxifen, an antagonist of the estrogen receptor, prior to addition of testosterone. Under these conditions the testosterone-induced Ca²⁺ oscillations were the same as in the presence of testosterone alone (Fig. 1F, black line, n=28 of 37 cells, three independent cultures), indicating that a functional estrogen receptor is not needed.

Typically the Ca²⁺ response was initiated in one discrete zone at the tip of the neurite. Then, the Ca²⁺ transient spread from the point of origin and propagated along the axis of the cell (Fig. 1A). To improve the time resolution of this signaling event we performed linescan experiments (Fig. 1G). The fluorescence intensity was recorded in different regions of the cytosol and nucleus (Fig. 1G). Transient rises in Ca²⁺ were observed along the dendrite, producing a cytosolic Ca²⁺ wave that propagated towards the nucleus, into the nucleus, and often to the tip of the opposite dendrite (Fig. 1G,H). The propagation rate of the Ca²⁺ wave was 11.2±1.8 μm/second (Fig. 1G). For comparison, the diffusion constant (D) of Ca²⁺ ions in cytosol is estimated to be 13 μm²/second and that for Ins(1,4,5)P₃ is 280 μm²/second (Allbritton et al., 1992).

In order to determine the contribution of cytosolic and nuclear Ca²⁺ to the global Ca²⁺ response, Ca²⁺ was buffered in each compartment separately. To accomplish compartmentalized buffering, we heterologously expressed a fusion protein that contained the Ca²⁺ buffer parvalbumin (PV) tagged with a red fluorescent protein (DSR) and signal sequences which targeted the fusion protein either to the cytosol or nucleus (Pusl et al., 2002). Cells were transfected with PV-DSR targeted to nucleus (PV-NLS-DSR), a Ca²⁺ insensitive form of PV-DSR targeted to nucleus (PV-NLS-CDEF-DSR), or PV-DSR targeted to cytosol (PV-NES-DSR). Western blot analysis of whole cell lysates show the increase in the amount of PV in all three conditions (Fig. 2A-C). The location of DSR and a nuclear stain was used to show that the fusion proteins were expressed in the expected subcellular location (Fig. 2D-F). Basal levels of Ca²⁺ were not changed in cells expressing PV-NLS-DSR or PV-NES-DSR, relative to control cells, as described previously (Pusl et al., 2002). Transfection with an empty vector containing DRS did not modify the effects of testosterone (data not shown). Cells transfected with the PV-NLS-DSR did not exhibit any nuclear Ca²⁺ increase after testosterone exposure, but the cells showed normal cytosolic Ca²⁺ oscillations (Fig. 2G, n=21 of 21 cells, four independent cultures). To test whether the inhibition of the nuclear Ca²⁺ signal was due to the Ca²⁺-chelating properties of PV, cells were transfected with PV-NLS-CDEF-DSR, which has a double mutated EF-hand motif that prevents the protein from binding Ca²⁺ (Fig. 2H). In cells expressing PV-NLS-CDEF-DSR, the addition of testosterone induced cytosolic and nuclear Ca²⁺ oscillations (Fig. 2H, n=27 of 27 cells, four independent cultures), showing that the Ca²⁺-buffering capacity of the PV is required. By contrast, cells expressing PV-NES-DSR did not exhibit any Ca²⁺ increase in either the cytosol or nucleus (Fig. 2I; n=31 of 31, three independent cultures), indicating that the testosterone-induced Ca²⁺ increase began in the cytosol and then propagated into the nucleus.

The iAR is present in neuroblastoma cells suggesting that these cells are a target for the physiological action of testosterone (Maggi et al., 1998; Yerramilli-Rao et al., 1995). Using an antibody directed against the N-terminal domain of the human iAR, the iAR was found throughout the cell, but with highest expression in the nucleus (Fig. 3A). In some cells the highest expression of the iAR was at the extreme tip of the neurites (Fig. 3A, upper panel). Upon stimulation with testosterone, the cytosolic iAR translocated to the nucleus after 1 hour, showing that the system was functional (Fig. 3A, lower panel). To evaluate whether the rapid Ca²⁺ response seen after the addition of testosterone was independent of the iAR, we used both a pharmacological blocker and genetic knock-down of the iAR. Cells transfected with small interfering RNA for the iAR (iAR-siRNA) showed a significantly reduced protein expression. Western blot analysis revealed that iAR-siRNA downregulated the protein by 80% (Fig. 3B,C). Neither the knock-down (Fig. 3D, n=25 of 25 cells, five independent cultures) nor the treatment with the specific inhibitor of the iAR, cyproterone (Fig. 3E, n=29 of 35, four independent cultures), modified the ability of testosterone to induce intracellular Ca²⁺ oscillations. To evaluate whether the effect of the hormone was mediated by an extracellular membrane receptor, the effect of testosterone covalently bound to albumin (T-BSA) was tested. This compound does not cross the plasma membrane making it unable to act on iAR. The response was similar when cells were exposed to testosterone or T-BSA (100 nM; Fig. 3F, n=32 of 40 cells, four independent cultures). Albumin alone (0.1%) does not produce any change in the intracellular Ca²⁺ (data not shown). In addition, the peak frequencies calculated from spectral analysis of the Ca²⁺ oscillations in all three test conditions and untreated cells were similar. Taken together these results indicate that testosterone-induced intracellular Ca²⁺ oscillations were independent of the iAR-mediated genomic response.

Experiments were done to determine whether the action of testosterone on intracellular Ca²⁺ was due to an influx of Ca²⁺ from the extracellular medium and/or mobilization of Ca²⁺ from intracellular stores. Pre-treatment of the cells with nifedipine (5 μM), a L-type voltage-dependent Ca²⁺ channel blocker, did not modify the ability of testosterone to induce Ca²⁺ oscillations (Fig. 4A, n=16 of 19 cells, three independent cultures). When cells were incubated in Ca²⁺-free medium for 5 minutes prior to testosterone stimulation, the hormone induced a single Ca²⁺ transient after a delay of 33±14 seconds (n=38 of 72 cells from 12 different cultures). No oscillations were produced under these conditions (Fig. 4B). The response in Ca²⁺-free medium resembled the first Ca²⁺ peak generated in Ca²⁺-containing medium with a similar delay and magnitude of response (compare Fig. 1C with Fig. 4B, 3.2±0.6 and 3.0±0.4 F/F₀, Ca²⁺-containing or Ca²⁺-free medium, respectively). These results suggest that the testosterone-induced regenerative Ca²⁺ oscillations require both extracellular and intracellular Ca²⁺ mobilization. In control experiments, the basal levels of intracellular Ca²⁺ in cells not treated with the hormone were unaffected by the removal of extracellular Ca²⁺ during the time of data acquisition (data not shown). In addition, pre-incubation of the cells with Gd³⁺ (1 μM), a nonspecific plasma membrane Ca²⁺ channel blocker, in Ca²⁺ containing medium, suppressed the Ca²⁺ oscillations without affecting the first Ca²⁺ peak (Fig. 4C, n=28 of 35, four independent cultures). These results indicate that the extracellular Ca²⁺ influx was due to activation of plasma membrane Ca²⁺ channels that are distinct from the L-type voltage-dependent Ca²⁺ channels. More importantly, extracellular Ca²⁺ influx is necessary to maintain Ca²⁺ oscillations in response to testosterone.

Depletion of intracellular Ca²⁺ stores by pre-treating the cells with the endoplasmic reticulum Ca²⁺-ATPase inhibitor thapsigargin, completely blocked the hormone-triggered Ca²⁺ response (data not shown, n=25 of 25 cells) indicating that the Ca²⁺ increase produced by testosterone also required thapsigargin-sensitive intracellular Ca²⁺ stores. Pre-treatment of neuroblastoma cells with ryanodine, at 20 μM, a concentration known to inhibit the RyR channel (Ehrlich et al., 1994), had no effect on testosterone-induced Ca²⁺ oscillations, suggesting that Ca²⁺ mobilization did not require the RyR (Fig. 4D, n=23 of 23, from four different cultures). By contrast, the testosterone-induced Ca²⁺ response was completely abolished by the phospholipase C (PLC) inhibitor, U-73122 (Fig. 4E, n=26 of 28, four independent cultures) showing that generation of Ins(1,4,5)P₃ is necessary. The amount of Ca²⁺ in the intracellular stores was not affected by U-73122, as determined by the magnitude of the release after addition of thapsigargin (Fig. 4E). Cells pre-treated with 2-aminoethyl diphenylborate (2-APB; 10 μM), a plasma membrane permeable inhibitor of the Ins(1,4,5)P₃R, did not respond to testosterone (data not shown, n=20). To confirm the participation of the Ins(1,4,5)P₃R in the testosterone-induced Ca²⁺ oscillations, two additional types of experiments were done. First, 2-APB was added to testosterone-treated cells that were in the process of producing Ca²⁺ oscillations. Immediately after 2-APB was applied to the cells, the Ca²⁺ oscillations were abolished (Fig. 4F, n=12 of 16 cells, from three independent cultures). Second, the amount of Ins(1,4,5)P₃R was knocked-down by transiently transfecting the cells with siRNA for the Ins(1,4,5)P₃R type 1. This isoform was chosen because the cell line used in these studies (SH-SY5Y) expresses mainly the Ins(1,4,5)P₃R type 1 (Wojcikiewicz, 1995). There was a significant reduction in the imunosignal for the Ins(1,4,5)P₃R type 1 in transfected cells (∼80%, P<0.05; Fig. 5A,B). Interestingly, we observed two types of Ca²⁺ response to testosterone in the Ins(1,4,5)P₃R type 1 knocked-down cells (Fig. 5C). In one sub-group (n=26 of 46 cells, from five different cultures), the testosterone response was completely abolished. In the other sub-group (n=20 of 46 cells), cells exhibited a fast and transient Ca²⁺ increase, similar to the peak Ca²⁺ increase observed in the Ca²⁺-free medium. However, this peak response was reduced by 68% (P<0.05 compared with the control), as compared with the initial Ca²⁺ peak in non-transfected cells (Fig. 5D). It is important to note that no Ca²⁺ oscillations were observed in Ins(1,4,5)P₃R type 1 knock-down cells. These results suggest that the loss of functional Ins(1,4,5)P₃R impedes the ability of the initial Ca²⁺ response to trigger a regenerative Ca²⁺ oscillation. Taken together, these results imply that testosterone-induced intracellular Ca²⁺ oscillations are the result of the coordinated actions of Ca²⁺ mobilization from Ins(1,4,5)P₃-sensitive Ca²⁺ stores and Ca²⁺ influx through plasma membrane Ca²⁺ channels.

Activation of PLC at the plasma membrane could involve either tyrosine kinase or G-protein coupled receptors. To investigate the early events involved in generating the Ca²⁺ signals produced by addition of testosterone, cells were incubated with genistein (50 μM, 20 minutes), a tyrosine kinase receptor inhibitor. Genistein modified neither the initial intracellular Ca²⁺ increase nor the Ca²⁺ oscillations induced by testosterone (Fig. 6A, n=16 of 16, two independent cultures). To test for the involvement of a G-protein-coupled receptor, cells were permeabilized for 5 minutes with saponin in the presence of guanosine 5′-O-(2-thiodiphosphate) (GDPβS; 100 nM), a non-hydrolysable analog of GTP. Permeabilization did not modify testosterone-induced Ca²⁺ responses (Fig. 6B, dashed line; n=12 of 16 cells, two independent cultures), whereas GDPβS did suppress the Ca²⁺ increases (Fig. 6B, solid line; n=21 of 21; four different cultures). When cells were pre-incubated with pertussis toxin (PTX, 1 μg/ml, 6 hours) before the addition of testosterone there was no Ca²⁺ signal produced by the hormone (Fig. 6C n=38 of 38 cells; six different cultures). These results suggest that testosterone action requires PTX-sensitive G-protein coupled receptors to activate PLC and generate Ins(1,4,5)P₃ to produce Ca²⁺ signals.

Neuroblastoma cells grown under control conditions have short neurites. To determine the effect of testosterone on morphology, cells were incubated with the hormone for 3 days. To visualize the morphological changes, cells were loaded with Cell Tracker (Fig. 7A), a fluorescent dye that does not affect the neuronal viability (Ang et al., 2003). For the analysis of neurite outgrowth, cells with a neurite that was longer than the soma were included. To normalize for variations in cell size, neurite outgrowth was calculated as the ratio of the neurite length to soma length in the neurite projection. When neuroblastoma cells were exposed to testosterone, smooth and elongated neurites were observed (Fig. 7B). Testosterone treatment increased the neurite outgrowth 2- to 2.5-fold (n=108; P<0.01) when compared with untreated cells (n=99) (Fig. 7C). Next, we examined the Ca²⁺ requirements for this testosterone-induced phenomenon using the parvalbumin constructs targeted to cytosol (PV-NES-DSR) or nucleus (PV-NLS-DSR). Expression of the parvalbumin proteins did not modify the normal growth or cell morphology of the control cells. Testosterone-induced neurite outgrowth was inhibited in cells expressing the cytosolic Ca²⁺ buffer protein, PV-NES-DSR (Fig. 7C; n=78 of 78 transfected cells; P<0.01). When the nuclear pool of Ca²⁺ was buffered with PV-NLS-DSR, an increase in the neurite elongation produced by testosterone was observed (Fig. 7C; n=86 of 86 transfected cells, P<0.05). This enhanced neurite outgrowth, however, was reduced by 40% with respect to non-transfected stimulated cells. Cells expressing the nuclear localized mutated parvalbumin (PV-NLS-CDEF-DSR) showed a rise in the neurite outgrowth in response to testosterone as observed in control cells (n=36 of 36 transfected cells, P<0.01). These data demonstrate that targeted cytosolic expression of parvalbumin inhibits testosterone-induced neurite outgrowth and that a nuclear Ca²⁺ increase is also needed for the neurite elongation. Under conditions that block the iAR, the increase in neurite outgrowth was smaller than the response obtained with the addition of testosterone alone (Fig. 7D). To determine the contribution of the genomic pathway in testosterone-induced neurite elongation, we silenced the iAR by use of siRNA-AR or by treatment with the iAR antagonist cyproterone. Cells were co-transfected with siRNA-AR and DSRed-pCMV so that transfected cells (those with DSRed) could be visualized and compared with non-transfected cells in the same experiment. After these cells were stimulated with testosterone (100 nM) for 24 hours, the testosterone-mediated neurite outgrowth was decreased 70% relative to non-transfected cells in the siRNA-iAR treated cells (Fig. 7D; n=78 of 78 cells, P<0.05, four independent cultures) or 59% in cyproterone-treated cells (Fig. 7D; n=46 of 46, P<0.05, three independent cultures). Treatment with the membrane-delimited testosterone (T-BSA) decreased the neurite elongation by 60% (Fig. 7D; n=36 of 36 cells, P<0.05, three independent cultures). In all these comparisons, the magnitude of the neurite outgrowth in the absence of testosterone is defined as control. The responses after silencing the iAR show testosterone-induced neurite outgrowth and activation of the iAR allows an enhancement of this response. Together these results suggest that full activation of the testosterone-induced neurite outgrowth requires both a Ca²⁺-mediated signaling pathway as well as the transcriptional activity associated with the iAR.

The goal of this study was to investigate the mechanisms used in the generation of rapid testosterone-induced Ca²⁺ signals in neurons. Rapid, non-genomic effects of androgens have been described (Benten et al., 1999b; Estrada et al., 2000; Lieberherr and Grosse, 1994), but their biological implication was unresolved. To investigate these effects of testosterone, we monitored the temporal and spatial characteristics of testosterone-induced Ca²⁺ signals and began a dissection of the molecular basis of the signals. Our data show that testosterone induces rapid intracellular Ca²⁺ increases, which begin as discrete Ca²⁺ transients and develop into propagating waves in the cytosol, a process that repeats to become an oscillatory pattern. These Ca²⁺ signals are mediated via a plasma membrane receptor rather than the iAR, and they depend on an interplay between Ins(1,4,5)P₃-sensitive Ca²⁺ stores and influx from the extracellular medium. This new pathway for the action of testosterone leads to the modulation of the neurite outgrowth. These rapid changes in Ca²⁺ precede the traditional genomic effects, but most probably they are critical components of the overall physiological response to this hormone.

Intracellular Ca²⁺ oscillations are a common signaling event observed in many cell types (Berridge et al., 2003). These signaling cascades are not random events, but rather specific signals that depend on the stimuli and the cell type (Aizman et al., 2001; Dolmetsch et al., 1998; Estrada et al., 2005). For example, in the cells we investigated, the initial Ca²⁺ release never began in the nucleus, implying that the signal needs to propagate through the cell to reach the nucleus. Ca²⁺ ions have been shown to travel slower than Ins(1,4,5)P₃ and generally, Ca²⁺ does not diffuse more than 1 μm before it is sequestered in the cell (Allbritton et al., 1992). The distance between the place where the Ca²⁺ transient began and the nucleus in these cells is approximately 10 μm and hence too long for Ca²⁺ to travel by simple diffusion. It is also unlikely that the propagated Ca²⁺ wave in testosterone-treated neuroblastoma cells can be explained purely by diffusion of Ins(1,4,5)P₃ because the signal observed traveled 10 μm/second, which is considerably slower than the diffusion constant for Ins(1,4,5)P₃ (280 μm²/second). These data suggest that the testosterone-induced Ca²⁺ wave observed in these experiments is a dynamic consequence of the diffusion of both Ca²⁺ and Ins(1,4,5)P₃.

We found that the oscillatory pattern induced by testosterone exhibits a constant frequency of ∼20 mHz. The requirement for oscillations, rather than single Ca²⁺ transients or prolonged elevations in intracellular Ca²⁺, to encode information has been reported and specific frequencies were shown to activate specific genes (Dolmetsch et al., 1998; Li et al., 1998; Sneyd et al., 2004). Basically, by exploiting the two key features of oscillatory signals - frequency and amplitude - the cell can use Ca²⁺ as a second messenger to generate a large variety of intracellular signals. This is an efficient way to use the same second messenger to activate many different processes. The testosterone-induced Ca²⁺ increase was evident in both cytosol and nucleus of neuroblastoma cells. Nuclear Ca²⁺ signals can directly modify gene expression, fertilization, meiosis and apoptosis (Hardingham et al., 1997), representing a pivotal connection point between extracellular and intracellular stimuli. Several reports show that nuclear Ca²⁺ elevations are due to changes in the cytosolic Ca²⁺ concentration (Genka et al., 1999) whereas other reports indicate that the nuclear Ca²⁺ changes can be regulated independently of cytosolic Ca²⁺ (Leite et al., 2003). In order to choose between these possibilities we selectively inhibited the Ca²⁺ signaling in either the nucleus or cytosol by expressing location-specific Ca²⁺ buffer protein, parvalbumin. Although we cannot rule out testosterone-induced intranuclear Ca²⁺ release, our data suggests that the response to testosterone is initiated in the cytoplasm and that subsequent Ca²⁺ signals are generated by diffusion of Ca²⁺ and Ins(1,4,5)P₃ in the cytoplasm and into the nucleus. In addition, aromatase, an enzyme that converts testosterone into the estrogen (17β-estradiol) has been reported in the CNS and neuroblastoma cells (Wozniak et al., 1998). We found that 17β-estradiol but not 17α-estradiol induced Ca²⁺ signals. The estrogen-induced Ca²⁺ rises have a different temporal pattern than those measured after addition of testosterone. To rule out the possibility that testosterone acts via the estrogen pathway, experiments were done in the presence of tamoxifen, an antagonist of the intracellular estrogen receptor. As shown in Fig. 1F (black line), the inhibitor did not have any effect on the testosterone-induced Ca²⁺ oscillations, suggesting that the response was specific for a direct testosterone action and not due to its metabolization to estrogen.

Ca²⁺ oscillations induced by testosterone only occurred in the presence of extracellular Ca²⁺. Stimulation of cells in Ca²⁺-free conditions evoked a localized Ca²⁺ increase which did not propagate throughout the cell, but resembled the initial peak in Ca²⁺-containing conditions. Similar results were obtained when Ca²⁺ was blocked with Gd³⁺. These findings suggest that the initial Ca²⁺ rise requires mobilization from internal stores, but Ca²⁺ influx is required to generate a propagated Ca²⁺ wave and subsequent Ca²⁺ oscillations. In several cell models, Ca²⁺ oscillations have been reported to be initiated by Ins(1,4,5)P₃-induced release of Ca²⁺ from intracellular Ca²⁺ stores (Aizman et al., 2001; Berridge et al., 2003; Estrada et al., 2005). Maintenance of these waves, however, required Ca²⁺ influx through Ca²⁺ channels in the plasma membrane (Sneyd et al., 2004).

Testosterone exerts its genomic effects through binding to, and activation of, a iAR which translocates to the nucleus and functions as a transcription factor (Beato, 1989). The iAR translocation from the cytosol to nucleus upon hormone binding is necessary for its activation and subsequent action on transcriptional machinery (Lucas and Granner, 1992). We found that testosterone-evoked Ca²⁺ oscillations were not blunted by cyproterone, an antagonist of the iAR, the membrane-impermeant testosterone conjugate (T-BSA) induced effects that were similar to the free hormone, and knock-down of the iAR did not modify the oscillatory pattern induced by testosterone. Taken together, these results suggest that the rapid effects of testosterone are mediated by a receptor restricted to the plasma membrane. Recently, several reports have shown that androgens can activate PTX-sensitive G proteins (Benten et al., 1999b; Estrada et al., 2003). We used a pharmacological approach to determine whether testosterone activated G proteins and found evidence to support the involvement of a plasma membrane receptor in this signaling event. The presence of membrane binding sites for androgens has been previously suggested (Estrada et al., 2003; Lieberherr and Grosse, 1994), even in macrophages, which lack a classical iAR (Benten et al., 1999b). Our results expand this concept to a neuronal cell line. Thus, a transient increase in Ca²⁺ appears to be a response used by many cell types to directly alter cellular processes through a nongenomic pathway, without the need for slower, genomic processes.

A major question about the rapid effects of steroid hormones is whether there is a physiological role for these signals. In neurons, Ca²⁺ oscillations have been shown to be essential to migration (Spitzer et al., 2000), differentiation and neurite outgrowth (Gomez and Spitzer, 1999; Lautermilch and Spitzer, 2000). Our observations are in agreement with previous studies showing the ability of androgens to increase neurite outgrowth in PC12 cells (Lustig, 1994) and in motor neuron cells (Marron et al., 2005). These effects were assumed to be a consequence of activation of the iAR, but the participation of second messenger cascades was not clear. Cytosolic kinase activity, such as PKA and PI 3-kinase, have been shown to be important for the initial steps of neurite elongation in SH-SY5Y cells directly stimulated by addition of a membrane permeable activator of PKA, cyclic AMP (Sanchez et al., 2001). In addition, it has been reported that stimulation of ERK pathways regulates neuronal growth and survival (Bonni et al., 1999). Interestingly, in neurons, these kinase pathways can be activated and modulated by intracellular Ca²⁺ signaling (Doherty et al., 2000). Thus, Ca²⁺ oscillations induced by testosterone could represent an early key regulator of these events. To investigate this suggestion, we buffered intracellular Ca²⁺ signaling in specific locations in the cell. We found that the testosterone-induced elongations of the neurites were inhibited in cells in which the cytosolic Ca²⁺ was buffered, whereas buffering the nuclear Ca²⁺ pool inhibited the neurite outgrowth by only 40%. Our data strongly suggest that cytosolic and nuclear Ca²⁺ increases are critical mechanisms controlling the neurite outgrowth induced by testosterone. Androgens stimulate the differentiation of different neuronal cell types via the activation of the iAR and gene expression (Kelly et al., 1999). We found that the magnitude of the neurite elongation was less in cells that had been modified to inactivate the iAR pathway than in cells stimulated with testosterone alone. In both cases the neurite elongation was significantly more than in cells that had not been treated with testosterone. These data show that the initial Ca²⁺ signaling as well as an intact iAR are required to maximize neurite elongation. Nongenomic actions can play a physiological role in events prior to iAR activation. At the cytosolic level, Ca²⁺ increases stimulate the binding of androgens to their receptors (Cabeza et al., 2004). Moreover, androgens can activate Ca²⁺-dependent kinase pathways, such as ERK, PI 3-kinase or Src (Estrada et al., 2003; Migliaccio et al., 2000), which could phosphorylate the iAR and enhance its activity. It has also been suggested that both the temporal and spatial changes of the nuclear Ca²⁺ signal (Echevarria et al., 2003; Leite et al., 2003) and cytosolic Ca²⁺ signal (Dolmetsch et al., 1998; Li et al., 1998) are involved in the control of gene expression. Our results lead to the conclusion that the genomic and non-genomic pathways of testosterone action are inter-linked and that a concerted action is required for normal cell function.

In neurons, androgens can induce changes at the cellular level, which can lead to changes in behavior such as sleep, the reaction to stress, mood and memory (Kelly et al., 1999). Although testosterone is an important neurotrophic and neuroprotective agent, which protects cells against death and injury, it also has its negative side. The concentration of testosterone used in this study (100 nM) is on the high end of the normal range measured in human males. However, a number of physiological circumstances including age, sex, or physical condition, can cause plasma levels to increase, reaching values similar to those studied here (Kelly et al., 1999; Mooradian et al., 1987). It is likely then, that the signals we report here correspond to normal responses of the neuronal cells to transient levels of these hormones that may be reached under some particular physiological condition. Testosterone increases aggressive behavior (Kelly et al., 1999) and high levels have been linked to neuronal apoptosis and cell death (M.E. and B.E.E., unpublished observations). These negative effects of androgens are not limited to motor neurons but extend to many cell types. Surprisingly, these harmful effects of high levels of testosterone can occur rapidly, and may be initiated by a dysregulation of the Ca²⁺ signals described in this report. With a more global view, these effects of testosterone observed at the single cell level can be shown to have long-term effects at the organ level.

In this study, we found that testosterone induces rapid intracellular Ca²⁺ increases, which are characterized as propagated Ca²⁺ waves in the cytosol and nucleus generating oscillatory Ca²⁺ patterns. The oscillatory nature of the Ca²⁺ signals depends on a concerted action between the Ins(1,4,5)P₃-sensitive intracellular Ca²⁺ stores and Ca²⁺ influx from the extracellular medium. Kinetic analysis and nuclear/cytosol-targeted parvalbumin fusion proteins showed that the Ca²⁺ rise is initiated in the cytosol, principally in the tip of the neurite, and then is propagated to the nucleus producing regenerative Ca²⁺ oscillations. Moreover, these effects were independent of the iAR and were mediated by a plasma membrane G protein-coupled receptor. The testosterone-induced neurite outgrowth required changes in intracellular Ca²⁺ signaling in the cytosol, presumably to initiate translocation of factors that activate nuclear events. These data support the conclusion that there is cross-talk in the pathways used for neuronal cell differentiation involving rapid non-genomic effects, such as Ca²⁺ oscillations, and the classical genomic actions of androgens, suggesting a novel role for testosterone.

The nuclear-targeted parvalbumin (PV) expression vector was constructed as previously described (Pusl et al., 2002), using the full-length parvalbumin gene, which was sub-cloned into the pCMV-Myc-Nuc vector (Invitrogen). DsRed (DSR) was substituted for the original GFP in each construct. The DSR coding sequence from pDsRed2-N1 (Clontech) was PCR-amplified to introduce 5′ and 3′ NotI sites. The DSR PCR product was digested with NotI and subcloned into each PV construct. For the cytosolic-targeted PV expression vector, the specific sub-cellular localization of the fusion protein was achieved by introducing the nuclear exclusion signal (NES) sequence derived from MKK1. The NES sequence derived from MKK1 encodes a short stretch of amino acids that represent residues 32-44 of MEK1, which does not contain the putative ERK binding site. These primers also introduced SalI and XbaI sites, which were used to subclone this DNA fragment into pCMV-Myc-Cyto to generate PV-NES-DSR. The final resulting fusion proteins were designated PV-NLS-DSR with nuclear localization and PV-NES-DSR with cytosolic localization. A site-directed mutagenesis was carried out with the nuclear PV to generate the mutant parvalbumin PV-NLS-CDEF coding for protein in which both functional Ca²⁺-binding sites (CD and EF domain) were inactivated by substituting a glutamate for a valine residue at position 12 of each Ca²⁺-binding loop. This new plasmid is designated as PV-NLS-CDEF-DSR. The mutant PV was used as a control for the Ca²⁺-buffering capacity of PV.

The human neuroblastoma cell line (SH-SY5Y; ATCC) was cultured in 1:1 DMEM:Ham's F12 medium supplemented with 10% (v/v) heat-inactivated fetal bovine serum, 5% non-essential amino acids, 100 IU penicillin and 50 μg/ml streptomycin, in a 95% air-5% CO₂ humidified atmosphere in an incubator at 37°C. Cells were grown on 22-mm gelatin-coated glass coverslips for the Ca²⁺ measurements or 60-mm Petri dishes for biochemical assays. Cultured cells were washed once with PBS before stimulation with testosterone (from concentrated stocks made in ethanol). The final ethanol concentration (<0.01%) had no effect on intracellular Ca²⁺ concentration or biochemical determinations. For transient transfections, neuroblastoma cells were grown in 60-mm dishes to 60% confluence. Before transfection, the cells were washed to remove serum and were transiently transfected with empty vector: PV-NES-DSR, PV-NLS-DSR or PV-CDEF-DSR (2-4 μg) using Lipofectamine (Invitrogen) in Opti-MEM (Invitrogen) for 4 hours following the manufacturer's instructions. A single plate of transfected cells was then used to set-up the experimental cultures required for each assay, ensuring equal transfection efficiencies between different treatments, and cells were cultured for an additional 24 hours before treatment with testosterone.

Cells were loaded with 5 μM Fluo-4/AM at 37°C for 30 minutes in a standard solution (in mM): 135 NaCl, 5 KCl, 2 MgCl₂, 2 CaCl₂, 10 Hepes, 5.6 glucose, pH 7.4. After loading, cells were washed twice with the standard solution and placed on an inverted microscope connected to a laser scanning imaging system (LSM 510 META, Zeiss). Cells were stimulated with the hormone diluted in the standard solution. A 488 nm excitation wavelength focused through a 40× Neo-Fluor objective lens (NA 1.4; Zeiss) was provided by an argon laser. In order to identify the cells expressing the PV construct with the DSR protein, transfected cells were examined using fluorescence excitation at 568 nm. Regions of interest (ROI) were monitored within the nucleus and cytoplasm of the same cell; transfected and non-transfected cells in the same field were followed over time. Each experiment involved a single independent cell and whole cell fluorescence was measured. A given cell was considered to oscillate when at least three Ca²⁺ transients were recorded over the time monitored, usually 10 minutes. ROIs with the same pixel dimensions, were identified and analyzed using ImageJ software (NIH, Bethesda, MA, USA). The inhibitors were added during the dye incubation; times and concentrations are indicated in the Results. To assess the role of G proteins, cells were incubated either with 1 μg/ml PTX for 6 hours or in a permeabilization solution [100 mM KCl, 20 mM NaCl, 5 mM MgSO₄, 1 mM NaH₂PO₄, 25 mM NaHCO₃, 3 mM EGTA, 1 mM CaCl₂, 20 mM Tris-HCl (pH 7.4), 0.1% BSA, 1 mM ATP, 0.1% glucose and 40 mg/ml saponin] for 5 minutes in the presence or absence of 100 nM GDPβS, a nonhydrolyzable analog of GDP. After permeabilization, but before hormone stimulation, the cells were incubated for 1 hour in the imaging solution buffer to re-stabilize the membrane integrity (Estrada et al., 2003). Ca²⁺-induced fluorescence intensity ratio (F/F_o) was plotted as a function of time. For line scanning, a single line (shown in the cell images as a solid white line) was chosen from the entire confocal section and repeatedly scanned every 20 mseconds. The successive lines were stacked horizontally to compile an image where time increased from left to right, and the spatial dimension was preserved in the vertical axis. To perform power spectrum analysis, we used an algorithm written in MATLAB as described previously (Uhlen, 2004).

We used siRNA for Ins(1,4,5)P₃R type 1 in order to reduce the levels of the receptor (provided by F. Leite, Federal University of Minas Gerais, Belo Horizonte, Brazil). The siRNA template was obtained from Ambion and the sequence was AAAGCACCAGCAGCTACAACTCCTGTCTC of the human Ins(1,4,5)P₃R type 1. The siRNA for the androgen receptor was obtained from Santa Cruz Biotechnology (cat. # sc-29204). Transfection with siRNA was performed using RNAiFect (Qiagen) and down-regulation of either Ins(1,4,5)P₃R or iAR was confirmed by immunofluorescence and western blot analysis.

Neuroblastoma cells grown on coverslips were fixed in ice-cold methanol for 15 minutes, blocked in PBS containing 1% BSA for 60 minutes, and incubated with primary antibodies at 4°C overnight. Primary antibody against androgen receptor (N-20; Santa Cruz Biotechnology) was used at 1:250. The cells were then washed five times with PBS/BSA and incubated with the appropriate Alexa-conjugated goat anti-rabbit secondary antibody (Molecular Probes) for 1 hour at room temperature (1:5000). After three washes, the coverslips were mounted in Prolong Antifade (Molecular Probes) to retard photo-bleaching. The samples were evaluated using confocal microscopy (LSM 510 META, Zeiss). To compare the confocal images of control and stimulated cells, the settings for the data acquisition and analysis were standardized.

Cells lysates containing 40 μg of protein were separated by SDS-PAGE in a 4-20% linear gradient for Ins(1,4,5)P₃R-1 and iAR or 10% gel for parvalbumin protein followed by electrophoretic transfer onto PVDF membranes for 2 hours at 400 mA. The following primary antibodies and their respective dilutions were used: anti-iAR (1:1000; Santa Cruz), anti-Ins(1,4,5)P₃R type 1 (1:2000) (Choe et al., 2004), anti-parvalbumin (1:1000; Sigma), anti-actin (1:1000; Santa Cruz). Membranes were incubated with primary antibodies overnight at 4°C. After incubation with horseradish peroxidase-conjugated secondary antibodies (1:5000) for 2 hours at room temperature, the bands were visualized by an enhanced chemiluminescence system (Pierce). Membranes were stripped and re-probed with β-actin antibody in order to control the protein loaded. Blots were quantified by scanning densitometry.

Neuroblastoma cells were cultured in growth medium for 24 hours. Next, the medium was replaced with a medium supplemented with or without testosterone and cultured for an additional 3 days. To determine whether Ca²⁺ signaling was necessary for neurite outgrowth, cells were transfected with parvalbumin plasmids that were engineered to express either in the cytoplasm (PV-NES-DSR) or nucleus (PV-NLS-DSR or PV-NLS-DSR). Intracellular androgen receptor participation was determined using siRNA-AR, cyproterone and T-BSA, as described above. To quantify the morphological changes induced by testosterone, cells were incubated with the fluorescent dye Cell-Tracker (Molecular Probes) for 45 min. Cells were visualized by confocal microscopy (LSM 510 META, Zeiss), and the acquired fluorescence images were analyzed and compared with LSM Image Browser (Zeiss). For the analysis of neurite outgrowth, only cells with a neurite that was longer than the soma were included. To normalize for various cell size, neurite outgrowth was calculated as the ratio of the neurite length to soma length in the neurite projection.

Data are expressed as mean ± s.e.m. or as representative traces. Statistical analysis of the differences between groups was performed using ANOVA, followed by the Bonferroni post-test. P<0.05 was considered statistically significant.

This work was supported by NIH grants GM63496 and DK61747 (B.E.E.), a Grass Foundation Fellowship (M.E.), and a grant from Vetenskapsrådet - the Swedish Research Council (P.U.). We are grateful to A. Bennett and F. Leite for providing critical reagents and to M. Nathanson, P. Correa, A. Varshney and C. Gibson for thoughtful discussions and comments on the manuscript.