Subplasmalemmal hydrogen peroxide triggers calcium influx in gonadotropes

Gonadotropin-releasing hormone (GnRH) stimulation of its eponymous receptor on the surface of endocrine anterior pituitary gonadotrope cells (gonadotropes) initiates multiple signaling cascades that culminate in the secretion of luteinizing and follicle-stimulating hormones, which have critical roles in fertility and reproduction. Enhanced luteinizing hormone biosynthesis, a necessary event for ovulation, requires a signaling pathway characterized by calcium influx through L-type calcium channels and subsequent activation of the mitogen-activated protein kinase extracellular signal-regulated kinase (ERK). We previously reported that highly localized subplasmalemmal calcium microdomains produced by L-type calcium channels (calcium sparklets) play an essential part in GnRH-dependent ERK activation. Similar to calcium, reactive oxygen species (ROS) are ubiquitous intracellular signaling molecules whose subcellular localization determines their specificity. To investigate the potential influence of oxidant signaling in gonadotropes, here we examined the impact of ROS generation on L-type calcium channel function. Total internal reflection fluorescence (TIRF) microscopy revealed that GnRH induces spatially restricted sites of ROS generation in gonadotrope-derived αT3-1 cells. Furthermore, GnRH-dependent stimulation of L-type calcium channels required intracellular hydrogen peroxide signaling in these cells and in primary mouse gonadotropes. NADPH oxidase and mitochondrial ROS generation were each necessary for GnRH-mediated stimulation of L-type calcium channels. Congruently, GnRH increased oxidation within subplasmalemmal mitochondria, and L-type calcium channel activity correlated strongly with the presence of adjacent mitochondria. Collectively, our results provide compelling evidence that NADPH oxidase activity and mitochondria-derived hydrogen peroxide signaling play a fundamental role in GnRH-dependent stimulation of L-type calcium channels in anterior pituitary gonadotropes.

The hypothalamic-pituitary-gonadal axis orchestrates reproductive function. Hypothalamic-pituitary-gonadal axis signaling begins with the release of gonadotropin-releasing hormone (GnRH), 2 a hypothalamic neuropeptide, into the hypophyseal portal system. Subsequent activation of GnRH receptors on anterior pituitary gonadotropes leads to increased synthesis and secretion of luteinizing hormone (LH) and follicle-stimulating hormone (FSH). In females, a dramatic gonadotrope GnRH receptor-dependent rise in LH is critical for ovulation and as such is a mandatory event for reproduction and fertility. Thus, the biological importance of molecular mechanisms regulating LH synthesis cannot be understated.
Activation of the GnRH receptor leads to canonical G␣ q protein signaling where phospholipase C cleaves phosphatidylinositol 4 -5-bisphosphate to generate the classic second messengers inositol 1,4,5-trisphosphate (IP 3 ) and diacylglycerol (1). IP 3 subsequently promotes calcium (Ca 2ϩ ) release from the endoplasmic reticulum, whereas diacylglycerol stimulates protein kinase C (PKC), which ultimately increases Ca 2ϩ influx through voltage-dependent L-type Ca 2ϩ channels (2). Each of these GnRH receptor-dependent Ca 2ϩ signals participates in the activation of mitogen-activated protein kinase signaling pathways, which lead to the increased transcriptional expression of LH and FSH. Experimental evidence suggests that Jun N-terminal kinase signaling, which is essential for up-regulation of FSH expression, depends on IP 3 -mediated Ca 2ϩ release from the endoplasmic reticulum. In contrast, extracellular signal-regulated kinase (ERK) signaling, which is essential for upregulation of LH expression, requires Ca 2ϩ influx through L-type Ca 2ϩ channels located in the plasma membrane (2,3). The divergence of Ca 2ϩ -dependent regulation of two distinct mitogen-activated protein kinase signaling pathways by two distinct Ca 2ϩ signals in gonadotropes exquisitely demonstrates the functional importance of spatially restricted Ca 2ϩ signaling. 2 The abbreviations used are: GnRH, gonadotropin-releasing hormone; [ cro ARTICLE Using a combination of electrophysiology and total internal reflection fluorescence (TIRF) microscopy, we reported that GnRH application induces discrete sites of Ca 2ϩ influx through L-type Ca 2ϩ channels (Ca 2ϩ sparklets) in gonadotrope-derived ␣T3-1 cells (4). Consistent with and expanding on prior work (2, 3), we found GnRH-dependent stimulation of L-type Ca 2ϩ channel sparklets to be dependent on PKC activity. Importantly, our data also indicate that highly localized PKC-dependent L-type Ca 2ϩ channel sparklet activity is necessary and sufficient for ERK activation (4). G␣ q protein-coupled receptor signaling in many cells stimulates NADPH oxidase enzyme complexes, leading to the generation of reactive oxygen species (ROS) (5). Intriguingly, recent evidence suggests that, in gonadotropes exposed to GnRH, Ca 2ϩ -and PKC-dependent NADPH oxidase activity increases ERK signaling and gonadotropin gene expression (6). Together, these observations evoke the hypothesis of functional coupling between GnRH-dependent ROS generation and Ca 2ϩ signaling in gonadotropes.
In contrast to their well-documented role in cell damage and dysfunction, ROS also operate as purposeful signaling molecules in the context of normal cell activity (7). Two major sources of ROS generation include the plasmalemmal NADPH oxidase family of enzyme complexes and the mitochondrial electron transport chain. In this study, we tested the hypothesis that intracellular ROS signaling contributes to GnRH-dependent stimulation of L-type Ca 2ϩ channels in gonadotropes.
Using a TIRF imaging-based approach, we found that GnRH induced highly localized, punctate sites of ROS generation in the subplasmalemmal space of gonadotropes. GnRH also increased oxidation within mitochondria near the cell periphery. ROS production in response to GnRH required contributions from NADPH oxidase and mitochondria. Consistent with our overall hypothesis, exogenous hydrogen peroxide (H 2 O 2 ) increased L-type Ca 2ϩ channel activity, whereas removal of endogenous H 2 O 2 with catalase abolished GnRH-dependent stimulation. Furthermore, L-type Ca 2ϩ channel activity observed in response to GnRH occurred predominantly at subplasmalemmal sites enriched with mitochondria. In sum, our data support a model where GnRH-dependent stimulation of L-type Ca 2ϩ channels incorporates a localized H 2 O 2 signaling mechanism generated by NADPH oxidase and subplasmalemmal mitochondria.

Results
To test our hypothesis that ROS signaling contributes to localized Ca 2ϩ influx through L-type Ca 2ϩ channels in gonadotropes exposed to GnRH, we formulated four requisite experimental criteria. 1) Exogenous ROS must stimulate localized Ca 2ϩ influx through L-type Ca 2ϩ channels. 2) GnRH must increase ROS generation in the subplasmalemmal space. 3) Inhibition of endogenous ROS signaling must decrease GnRH-mediated L-type channel Ca 2ϩ influx. 4) GnRH-dependent sites of Ca 2ϩ influx through L-type channels must be associated with sites of ROS generation.

Primary mouse gonadotropes produce L-type Ca 2؉ channel sparklets in response to GnRH and hydrogen peroxide
To continue our investigation of L-type Ca 2ϩ channel signaling in gonadotropes, we once again used a combination of electrophysiology and TIRF microscopy to visualize subplasmalemmal Ca 2ϩ influx with high temporal and spatial resolution (4). To this point, we have limited our investigation of L-type Ca 2ϩ channel signaling in gonadotropes to immortalized mouse gonadotrope-derived ␣T3-1 cells (4). ␣T3-1 cells were carefully selected as this cell line faithfully recapitulates proximal GnRH receptor signaling mechanisms observed in native gonadotropes (2)(3)(4)8). However, to directly extend our findings in ␣T3-1 cells to native cells, we examined the effects of GnRH and H 2 O 2 on L-type Ca 2ϩ channel sparklets in genetically labeled primary mouse gonadotrope cells (9).
Similar to ␣T3-1 cells (4), incubating primary gonadotropes with the L-type Ca 2ϩ channel antagonist nicardipine (10 M for 10 min) abolished GnRH-dependent Ca 2ϩ influx ( Fig. 1B; n ϭ  19). These data demonstrate that GnRH promotes localized Ca 2ϩ influx through L-type Ca 2ϩ channels in primary gonadotropes by increasing the number of active Ca 2ϩ sparklet sites and by increasing the activity of those sites.
The L-type Ca 2ϩ channel sparklets observed in primary gonadotropes were qualitatively similar to those observed in ␣T3-1 cells (e.g. nP s values, bimodal activity distribution, the quantal amplitude of single-channel Ca 2ϩ influx events, and size of the Ca 2ϩ sparklet itself) (4). Primary cells, however, appeared to have an average L-type Ca 2ϩ channel sparklet site density ϳ1.5-fold greater than that of ␣T3-1 cells. To the best of our knowledge, these data represent the first evidence of localized L-type Ca 2ϩ channel signaling in native gonadotropes and validate our use of ␣T3-1 cells to investigate localized Ca 2ϩ signaling in gonadotropes.

ROS promote colocalized L-type calcium channel sparklets
Similar to GnRH, incubating primary gonadotropes with the L-type Ca 2ϩ channel antagonist nicardipine (10 M for 10 min) abolished H 2 O 2 -dependent Ca 2ϩ influx ( Fig. 1D; n ϭ 21). Importantly, the time course of increased L-type Ca 2ϩ channel sparklet activity (within 5 min) and the average nP s and density (Ca 2ϩ sparklet sites/m 2 ) in response to H 2 O 2 were similar to our prior observations in ␣T3-1 cells exposed to GnRH (4). These data suggest that GnRH receptor stimulation and exogenous H 2 O 2 application could be regulating L-type Ca 2ϩ channels through a common mechanism.

Hydrogen peroxide also evokes localized L-type channel Ca 2؉ influx in immortalized gonadotrope-derived ␣T3-1 cells
Confirming our results in primary cells, exposing single gonadotrope-derived ␣T3-1 cells to exogenous H 2 O 2 (100 M) increased Ca 2ϩ sparklet site activity compared with control conditions (Fig. 2, A and B; control median nP s ϭ 0.010, IQR ϭ 0.016; H 2 O 2 median nP s ϭ 0.25, IQR ϭ 0.41; p Ͻ 0.05, n ϭ 9). H 2 O 2 also increased Ca 2ϩ sparklet site density (Fig. 2, A and B; control density ϭ 0.0024 Ϯ 0.0008 sites/m 2 ; H 2 O 2 density ϭ 0.0110 Ϯ 0.0008 sites/m 2 ; p Ͻ 0.05, n ϭ 9). We verified that the Ca 2ϩ sparklets visualized in response to H 2 O 2 were produced by L-type Ca 2ϩ channels by pretreating cells with nicardipine (10 M for 10 min). In the presence of L-type Ca 2ϩ channel blockade, H 2 O 2 exposure had no demonstrable effect on Ca 2ϩ sparklet activity in ␣T3-1 cells ( Fig. 2D; p Ͼ 0.05, n ϭ 12). These data demonstrate that, as in primary mouse gonadotropes, H 2 O 2 promotes localized Ca 2ϩ influx in ␣T3-1 cells by increasing the number of active L-type Ca 2ϩ channel sparklet sites and by increasing the activity of those sites.

GnRH induces punctate ROS formation in the subplasmalemmal space of ␣T3-1 cells
Next, we used TIRF microscopy to visualize changes in subplasmalemmal oxidation in ␣T3-1 cells loaded with the cellpermeant fluorescent ROS indicator 2Ј,7Ј-dichlorodihydrofluorescein diacetate (DCF). External application of a physiologically relevant concentration of exogenous H 2 O 2 (100 M) (13,14) increased DCF fluorescence throughout the cell footprint visible in the TIRF field (Fig. 3A). Although the

ROS promote colocalized L-type calcium channel sparklets
increase in fluorescence was not entirely uniform, H 2 O 2 increased the average DCF fluorescence ϳ2-fold ( Fig. 3B; p Ͻ 0.05, n ϭ 7). Under control conditions (i.e. baseline), our TIRF images also contained sites of highly localized DCF fluorescence (ROS puncta; Fig. 3A, circled yellow) (12,15,16). Following H 2 O 2 application, the increase in global fluorescence often obscured the identification of designatable ROS puncta (see "Experimental procedures"). Indeed, the amplitude of the average cell fluorescence following H 2 O 2 was not different from that of ROS puncta observed under control conditions or in the presence of H 2 O 2 ( Fig. 3B; p Ͼ 0.05, n ϭ 7). Consistent with these observations, the ROS puncta density (ROS puncta sites/m 2 ) before and after H 2 O 2 was not different ( Fig. 3C; p Ͼ 0.05, n ϭ 7).
In contrast to application of exogenous H 2 O 2 , exposing cells to GnRH (10 nM) in the presence of external Ca 2ϩ (2 mM) had no effect on the average DCF fluorescence (Fig. 3, D and E; p Ͼ 0.05, n ϭ 17). However, consistent with a localized ROS signaling mechanism, GnRH did increase the occurrence of ROS puncta ϳ4-fold (Fig. 3, D and F; density before GnRH ϭ 0.0021 Ϯ 0.00047 ROS puncta/m 2 ; density after GnRH ϭ 0.0088 Ϯ 0.0011 ROS puncta/m 2 ; p Ͻ 0.05, n ϭ 17). Interestingly, in the absence of external Ca 2ϩ (nominally Ca 2ϩ -free), GnRH produced no observable effect on the average DCF fluorescence or the occurrence of ROS puncta (Fig. 3, D, E, and F; p Ͼ 0.05, n ϭ 8 cells). For our GnRH experiments (before and after) in the presence (n ϭ 17) or absence (n ϭ 8) of external Ca 2ϩ , average DCF and ROS puncta fluorescence intensities were not different across groups ( Fig. 3E; p Ͼ 0.05). These data demonstrate that, analogous to L-type Ca 2ϩ channel signaling (see Ref. 4 and below), GnRH induces localized subplasmalemmal ROS signaling in gonadotrope-derived ␣T3-1 cells.

NADPH oxidase contributes to localized H 2 O 2 generation in the subplasmalemmal space of ␣T3-1 cells
NADPH oxidase enzyme complexes initiate oxidant signaling cascades in response to activation of G␣ q protein-coupled receptors (5). Therefore, we examined whether stimulation of GnRH receptors, which couple to G␣ q proteins, promote NADPH oxidase-dependent ROS generation in ␣T3-1 cells. GnRH alone increased ROS puncta densities ϳ4-fold (see   Superoxide (O 2 . ) is the de novo ROS product of NADPH oxidase. However, the half-life of superoxide is extremely short due to rapid conversion to H 2 O 2 and oxygen by the action of superoxide dismutase or spontaneous dismutation (7). Accordingly, we assessed the contribution of H 2 O 2 to GnRH-dependent ROS puncta formation. To do so, we incubated ␣T3-1 cells with a membrane-permeable PEG conjugate of the H 2 O 2 -decomposing enzyme catalase (PEG-catalase). Following a 10-min incubation with PEG-catalase (500 units/ ml), GnRH (10 nM) failed to increase the number of observable ROS puncta (Fig. 4, D and F; p Ͼ 0.05, n ϭ 14). PEG-catalase did not change the baseline ROS puncta density (i.e. before GnRH) and did not affect the average cell DCF fluorescence or the fluorescence of detected ROS puncta ( Fig. 4E; p Ͼ 0.05). These observations suggest a potential contribution of localized H 2 O 2 signaling in the subplasmalemmal space following GnRH receptor activation. Because of the promiscuity of DCF as an ROS indicator, the observed PEG-catalase sensitivity does not provide sufficient evidence to support the conclusion that H 2 O 2 is directly responsible for DCF ROS puncta formation in response to GnRH. These data are, however, consistent with a recent report describing the importance of NADPH oxidase activity to the function of gonadotropes (6) and suggest that H 2 O 2 produced by NADPH oxidase could contribute to L-type Ca 2ϩ channel stimulation by GnRH.

NADPH oxidase-derived H 2 O 2 is necessary for GnRH-dependent stimulation of L-type channel Ca 2؉ influx in ␣T3-1 cells
To assess the role of NADPH oxidase in stimulating L-type Ca 2ϩ channels in gonadotropes, we measured Ca 2ϩ sparklet activity in response to GnRH following inhibition of NADPH oxidase with the nonselective inhibitor apocynin and the Nox1selective inhibitor ML171 (17,18). Note that NADPH oxidase catalytic isoforms known to be expressed explicitly in gonadotropes include Nox1, Nox2, and the dual oxidases Duox1 and Duox2 (6). Consistent with our previous results (4), GnRH (10 nM) increased ␣T3-1 cell L-type Ca 2ϩ channel activity by increasing the number of active Ca 2ϩ sparklet sites and by increasing the activity at those sites (Fig. 5 Application of exogenous H 2 O 2 peroxide promotes localized L-type channel Ca 2ϩ influx (see Figs. 1 and 2) and removal of endogenous H 2 O 2 with PEG-catalase decreases GnRH-dependent ROS puncta formation (see Fig. 4). To examine the necessity of endogenous H 2 O 2 for GnRH-dependent stimulation of L-type Ca 2ϩ channels, we tested the effect of catalase on GnRH-dependent Ca 2ϩ influx. Intracellular dialysis of cell-impermeant unmodified catalase into the cytosol, achieved by inclusion of the enzyme (500 units/ml) in our internal pipette solution, abolished the stimulatory effect of GnRH on Ca 2ϩ influx (Fig. 5C). Indeed, in the presence of intracellular catalase, GnRH did not change nP s or density (Ca 2ϩ sparklet sites/m 2 ) compared with control (Fig. 5, D and E; p Ͻ 0.05, n ϭ 19). These data provide further evidence that GnRH-dependent stimulation of Ca 2ϩ influx in gonadotropes involves localized generation of H 2 O 2 to promote localized Ca 2ϩ influx through L-type Ca 2ϩ channels.

GnRH increases oxidation within subplasmalemmal mitochondria in ␣T3-1 cells
In addition to NADPH oxidase, the mitochondrial electron transport chain is a major source of ROS generation. Mitochondria generate ROS as a consequence of cellular respiration, and ROS production by mitochondria can change in response to cellular activity and function (19 -21). To determine whether mito-chondria participate in oxidant-dependent stimulation of L-type Ca 2ϩ channels in gonadotropes, we examined the effects of GnRH on mitochondrial oxidative status. To do so, we incubated ␣T3-1 cells with the mitochondrial matrix-targeted fluorescent oxidant probe MitoSOX-Red (1 M for 30 min) and visualized the oxidative status of subplasmalemmal mitochondria with TIRF microscopy. Indicative of increased mitochondrial oxidation, GnRH (10 nM) increased MitoSOX-Red fluorescence at discrete sites throughout the visible subplasmalemmal space of ␣T3-1 cells (Fig. 6, A and F; p Ͻ 0.05, n ϭ 9).
We performed two control experiments to confirm that the observed MitoSOX-Red fluorescence involved a change in mitochondrial ROS production. For a positive control, we provoked mitochondrial ROS production with the electron transport chain complex III inhibitor antimycin (500 nM) (22,23). We found that antimycin produced an increase in MitoSOX-Red fluorescence that was qualitatively similar to GnRH and not statistically significantly different from GnRH regarding overall fluorescence intensity (Fig. 6, B and F; p Ͼ 0.05, n ϭ 7 cells). As a negative control, we disrupted mitochondrial ROS production by collapsing the inner mitochondrial membrane potential with carbonyl cyanide m-chlorophenylhydrazone (CCCP). Following incubation with CCCP (1 M for 10 min), we found that neither GnRH (n ϭ 6) nor antimycin (n ϭ 5) produced an increase in MitoSOX-Red fluorescence (

ROS promote colocalized L-type calcium channel sparklets
dria-targeted antioxidant mitoTEMPO (25 nM for 10 min) (24) prevented GnRH-dependent changes in MitoSOX-Red fluorescence (n ϭ 12; Fig. 6, D and F). Together, these results support the conclusion that increased MitoSOX-Red fluorescence observed following GnRH exposure resulted from an increase in mitochondrial ROS production. We found that inhibition of NADPH oxidase with apocynin abolished GnRH-dependent ROS puncta formation (see Fig. 4). Accordingly, to investigate the importance of NADPH oxidase on GnRH-dependent mitochondrial ROS production, we examined the effect of GnRH on MitoSOX-Red fluorescence in the presence of apocynin. Similar to our results with the mitochondrial uncoupler CCCP, applying GnRH (10 nM) to ␣T3-1 cells incubated with apocynin (25 M for 10 min) produced no change in MitoSOX-Red fluorescence (Fig. 6, E and F; p Ͼ 0.05, n ϭ 5). We suggest that, in ␣T3-1 cells, NADPH oxidase activity is necessary for GnRH-dependent induction of mitochondrial ROS generation. This hypothesis is consistent with analogous ROS-induced ROS release mechanisms described in other cell types (16,22,25). Importantly, we did not observe significant changes in the distribution or shape of mitochondria in the subplasmalemmal space using MitoSOX-Red across all experimental groups (p Ͼ 0.05). Thus, our data suggest that GnRHdependent stimulation of subplasmalemmal mitochondrial ROS generation does not involve mechanisms associated with alterations in mitochondrial morphology or mitochondrial recruitment to specific subplasmalemmal sites.

Mitochondria-derived ROS are necessary for GnRH-dependent stimulation of L-type channel Ca 2؉ influx in ␣T3-1 cells
Having shown that GnRH induces oxidation within mitochondria located in the subplasmalemmal space of ␣T3-1 cells,

ROS promote colocalized L-type calcium channel sparklets
we examined the relationship between mitochondrial ROS generation and L-type Ca 2ϩ channel activity. First, we incubated ␣T3-1 cells with the mitochondria-targeted antioxidant mito-TEMPO. Evidence suggests that mitoTEMPO attenuates mitochondrial ROS production in response to oxidant exposure (i.e. mitochondrial ROS-induced ROS release) (16,24,26,27). Thus, we hypothesized that attenuation of mitochondrial ROS generation with mitoTEMPO should decrease ROS puncta formation in response to GnRH. Despite a high degree of variability, we found on average that incubating ␣T3-1 cells with mito-TEMPO (25 nM for 10 min) attenuated GnRH-dependent formation of ROS puncta (Fig. 7, A and B; p Ͼ 0.05, n ϭ 12). These data suggest that, similar to NADPH oxidase inhibition, attenuation of mitochondrial ROS formation inhibits GnRHdependent ROS microdomain signaling.
Consistent with our ROS puncta observations (Fig. 7A), we found that incubating ␣T3-1 cells with mitoTEMPO (25 nM for 10 min) abolished the stimulatory effect of GnRH on L-type Ca 2ϩ channel sparklet activity (Fig. 7, D and E; p Ͻ 0.05, n ϭ  14). MitoTEMPO had no effect on baseline Ca 2ϩ sparklet activity in the absence of GnRH (p Ͼ 0.05 relative to control, n ϭ 14). From these observations, we conclude that, in ␣T3-1 cells, mitochondrial ROS production is necessary for L-type Ca 2ϩ channel stimulation by GnRH (mitoTEMPO findings) and sufficient for oxidant-dependent stimulation (antimycin findings).

GnRH-dependent L-type channel Ca 2؉ influx occurs near subplasmalemmal mitochondria
Our data indicate that GnRH promotes localized H 2 O 2 generation by an NADPH oxidase/mitochondria-dependent mechanism. If H 2 O 2 produced by this mechanism regulates ) in ␣T3-1 cells before and after GnRH (n ϭ 9), before and after antimycin (n ϭ 7), before and after GnRH (n ϭ 6) and antimycin (representative data not shown, n ϭ 5) in the continuous presence of CCCP, and before and after GnRH in the continuous presence of mitoTEMPO (n ϭ 12) and apocynin (n ϭ 5). Error bars represent S.E. *, p Ͻ 0.05; ns, not significantly different.

ROS promote colocalized L-type calcium channel sparklets
L-type Ca 2ϩ channel activity, then Ca 2ϩ influx through these channels must occur near sites of H 2 O 2 generation such as subplasmalemmal mitochondria. To visualize the subcellular distribution of mitochondria in ␣T3-1 cells, we labeled mitochondria with fluorescent mitochondrial marker MitoTracker Green (200 nM for 15 min), marked the plasma membrane with a fluorescent wheat germ agglutinin-Alexa Fluor 555 conjugate (Alexa 555-WGA; 5 g/ml for 5 min), and imaged the resulting fluorescence with confocal microscopy. Our image stacks show that mitochondria occupied 9.86 Ϯ 0.88% of the total cell volume in ␣T3-1 cells (Fig. 8, A and B; n ϭ 6 cells) with most of the MitoTracker signal occurring distal to the plasma membrane (Ͼ 0.5 m; Fig. 8A, panel 3, green pixels). However, we did find that 28.38 Ϯ 6.36% of the mitochondrial volume resided in the subplasmalemmal space of these cells (Fig. 8, A, panel 3, yellow pixels, and C; n ϭ 6 cells). These data indicate that, in ␣T3-1 cells, the subplasmalemmal space contains a sufficient quantity of mitochondria to support ROS-dependent stimulation of L-type Ca 2ϩ channels. We then used TIRF microscopy to visualize the spatial relationship between subplasmalemmal mitochondria and L-type Ca 2ϩ channel activity. To do so, we imaged L-type Ca 2ϩ channel sparklets (with fluo-5F as before) in voltage-clamped ␣T3-1 cells loaded with MitoTracker Green (200 nM for 15 min). Our TIRF images revealed a scattered population of subplasmalemmal mitochondria (Fig. 9A, panel 1). Using thresholded Mito-Tracker Green images to establish clear mitochondrial boundaries, we calculated that these mitochondria occupied 13.34 Ϯ 2.20% of the visible subplasmalemmal space (Fig. 9A, panel 2; n ϭ 13). Exposing these cells to GnRH (10 nM) evoked L-type

ROS promote colocalized L-type calcium channel sparklets
Ca 2ϩ channel sparklets (Fig. 9A, panel 3) that were similar (regarding activity and density) to those observed in our previous experiments using GnRH (see Fig. 5

and Ref. 4).
To quantify the spatial relationship between subplasmalemmal mitochondria and L-type Ca 2ϩ channel sparklets, we overlaid our thresholded MitoTracker Green and fluo-5F images (Fig. 9A, panel 4) and measured the distance between the Ca 2ϩ sparklet site peaks (pixels of highest intensity) and the edge of the nearest thresholded MitoTracker signal. Mitochondria-associated L-type Ca 2ϩ channel sparklet sites were defined a priori as those sites with peaks Յ0.5 m from the edge of the nearest thresholded MitoTracker signal. This boundary is represented by the vertical dashed gray line in Fig. 9B. By plotting the cumulative values of these distances, we found that GnRHdependent L-type Ca 2ϩ channel sparklets associate with subplasmalemmal mitochondria (Fig. 9B; n ϭ 13 cells). Indeed, the half-distance of Ca 2ϩ sparklet sites observed (n ϭ 26 sites) to the nearest mitochondria was less than that of 130 randomly selected points within the visible plasma membrane (0.64 m, 95% confidence interval (0.54, 0.74) for observed Ca 2ϩ sparklet sites; 2.78 m, 95% confidence interval (2.69, 2.87) for 130 random points).
To further quantify the relationship between subplasmalemmal mitochondria and L-type Ca 2ϩ channel function, we com-pared the nP s of GnRH-dependent Ca 2ϩ sparklet sites associated with mitochondria (peak distance Յ 0.5 m) with those not associated with mitochondria (peak distance Ͼ0.5 m). We found that GnRH-dependent L-type Ca 2ϩ channel sparklet activity at mitochondria-associated sites (median nP s ϭ 0.72, IQR ϭ 0.88) was ϳ3-fold greater than that of Ca 2ϩ sparklet sites not associated with mitochondria (median nP s ϭ 0.24, IQR ϭ 0.20; Fig. 9C; n ϭ 13 cells, p Ͻ 0.05). From these data, we conclude that, in ␣T3-1 cells, the spatial distribution of GnRHdependent L-type Ca 2ϩ channel activity correlates strongly with the presence of subplasmalemmal mitochondria. This conclusion supports our overall hypothesis that localized ROS generation in response to GnRH receptor stimulation promotes the opening of nearby L-type Ca 2ϩ channels in gonadotropes.

Discussion
Herein, we investigated the hypothesis that localized L-type channel Ca 2ϩ influx in GnRH-stimulated gonadotropes involves an intracellular oxidant signaling mechanism. Our key observations in support of this supposition are: 1) exposing primary and immortalized mouse gonadotropes to exogenous ROS (H 2 O 2 ) provoked localized sites of L-type channel Ca 2ϩ influx, 2) applying GnRH induced discrete sites of ROS generation in the subplasmalemmal space, 3) GnRH induced oxidation within subplasmalemmal mitochondria, 4) inhibiting NADPH oxidase and mitochondria-dependent H 2 O 2 generation abolished GnRH-dependent stimulation of L-type Ca 2ϩ channels, and 5) the spatial distribution of GnRH-dependent L-type Ca 2ϩ channel activity correlated with the observed incidence of subplasmalemmal mitochondria. Considering these and other findings, we conclude that in mouse gonadotropes GnRH-dependent Ca 2ϩ influx through L-type Ca 2ϩ channels entails a localized H 2 O 2 -generating mechanism with synergistic contributions from NADPH oxidase and subplasmalemmal mitochondria (Fig. 10).
Immortalized ␣T3-1 cells recapitulate many properties attributed to primary gonadotropes (8,28,29). However, it is possible that the localized L-type Ca 2ϩ channel function observed in these cells (see Ref. (4) is nonrepresentative of L-type Ca 2ϩ channel function in native cells. We addressed this important issue directly by examining the effect of GnRH on Ca 2ϩ influx in genetically labeled primary mouse gonadotropes (9). Similar to ␣T3-1 cells, GnRH induced discrete sites of Ca 2ϩ influx through L-type Ca 2ϩ channels (Ca 2ϩ sparklets). Importantly, other than a higher overall occurrence, the L-type Ca 2ϩ channel sparklets we observed in primary gonadotropes were indistinguishable from those observed in ␣T3-1 cells.
With a sampling frequency of 50 Hz, short-lived Ca 2ϩ sparklet events are likely under-represented in our recordings. However, the degree of Ca 2ϩ influx associated with these brief events is relatively minor compared with that associated with the longer-lasting Ca 2ϩ sparklets observed following cell stimulation (see Figs. 1 and 2 and Refs. 4 and 11). Indeed, we previously established that evoked L-type Ca 2ϩ channel sparklets are biologically relevant Ca 2ϩ signals in ␣T3-1 gonadotropes by showing a strong correlation between the Ca 2ϩ sparklet activity observed (at 50 Hz) and ERK activation (4,30). Thus, we suggest

ROS promote colocalized L-type calcium channel sparklets
that although a portion of the shorter-lived Ca 2ϩ influx events may not be apparent in our recordings, this will minimally impact interpretation of our data due to the small contribution of missed events to total Ca 2ϩ influx. Our data show that inhibition of ROS production by either NADPH oxidase (with apocynin or ML171) or by mitochondria (with mitoTEMPO) is in each case sufficient for preventing ROS puncta formation or L-type Ca 2ϩ channel stimulation in response to GnRH. From an unbiased teleological perspective, either source alone or both in tandem could initiate ROS generation following GnRH receptor stimulation. However, studies on the angiotensin II type 1 receptor, which also couples to G␣ q proteins, conclude that ROS generation by NADPH oxidase precedes and induces subsequent mitochondrial ROS generation through a mechanism descriptively termed ROS-induced ROS release (4,9,20,21). Accordingly, our group found that, in arterial smooth muscle cells exposed to angiotensin II, subplasmalemmal mitochondria function as amplifiers of localized oxidant signaling microdomains initiated by NADPH oxidase (16). Although data presented here are consistent with an analogous ROS-induced ROS release mechanism in gonadotropes, additional experimentation is necessary to confirm the importance of ROS-induced ROS release in gonadotropes.
Similar to our findings with GnRH, we showed that localized angiotensin II-dependent ROS generation promoted colocalized Ca 2ϩ influx through L-type Ca 2ϩ channels (12, 15, 16). The Our data, in conjunction with previous reports (2-4), support a localized mechanism of oxidant-dependent L-type Ca 2ϩ channel signaling in gonadotropes. GnRH receptor stimulation promotes the generation of NADPH oxidase-dependent H 2 O 2 microdomains functionally coupled (via PKC) to L-type Ca 2ϩ channels. Colocalized L-type channel Ca 2ϩ influx, in turn, promotes ERK activation and ultimately an increase in gonadotropin biosynthesis. Furthermore, our data suggest that mitochondria-dependent ROS-induced ROS release (RIRR) serves as an associated amplification mechanism that is necessary for the generation of functionally relevant colocalized Ca 2ϩ and H 2 O 2 signaling microdomains. GnRHR, gonadotropin releasing hormone receptor; LTCC, L-type Ca 2ϩ channel; ETC, mitochondrial electron transport chain. See text for further details. stimulatory effect of angiotensin II required oxidant-dependent activation of PKC␣ (12,(31)(32)(33). Our prior work on gonadotropes revealed that GnRH-dependent stimulation of L-type Ca 2ϩ channels also required PKC with the novel ␦ and ⑀ isoforms identified as likely candidates (4). Importantly, similar to the conventional isoforms (e.g. PKC␣), the two pairs of zinc fingers located in the regulatory domains of novel PKC isoforms are subject to oxidation, thus rendering these kinases susceptible to activation by ROS (32, 34). Thus, given our findings that PKC activation and ROS generation are each required for GnRH-dependent stimulation of L-type Ca 2ϩ channels in gonadotropes, it is conceivable, if not likely, that ROS-dependent regulation of PKC activity occurs in these cells. Future experimentation is necessary to confirm the occurrence and significance of ROS-dependent regulation of PKC activity in gonadotropes.
Analogous to our angiotensin II-related findings in arterial smooth muscle (12,15,16), the observed incidence of punctate DCF fluorescence in gonadotropes occurs only after cell stimulation (e.g. GnRH). Similarly, ROS puncta generation in response to GnRH was ablated by inhibiting endogenous ROS generation by NADPH oxidase (with apocynin and ML171) and mitochondria (with mitoTEMPO) and by disrupting H 2 O 2 signaling with exogenous catalase. Together, these findings indicate that our results obtained with DCF require biologically relevant ROS generation and that some of the properties of localized ROS signaling in these two disparate cell types are shared.
However, in contrast to our findings in smooth muscle where the absence of external Ca 2ϩ did not influence ROS puncta formation, our results from gonadotropes indicate that GnRHdependent ROS generation requires external Ca 2ϩ . The underlying basis for the difference in Ca 2ϩ dependence is unclear. The mechanisms by which Ca 2ϩ contributes to GnRH-dependent ROS generation are, in general, not understood. Evidence suggests that Ca 2ϩ increases ROS signaling in gonadotropes by activating NADPH oxidase complexes containing Ca 2ϩ -responsive Duox catalytic subunits (6). However, indicating a role for Nox1-containing complexes, we found that inhibition of NADPH oxidase with the Nox1-selective inhibitor ML171 prevented GnRH-dependent stimulation of L-type Ca 2ϩ channels. Thus, further examination of NADPH oxidase signaling complexes in gonadotropes is warranted. Ca 2ϩ influx could also increase ROS generation by NADPH oxidase complexes by activating PKC (5). Alternatively, elevated Ca 2ϩ within the mitochondrial matrix promotes ROS generation by increasing the activity of citric acid cycle dehydrogenases and possibly the respiratory chain directly (35,36). Irrespective of mechanism, our data showing ROS-dependent Ca 2ϩ influx and Ca 2ϩ -dependent ROS generation in gonadotropes indicate that GnRH induces a self-amplifying signaling unit via a reciprocal coupling mechanism with two outputs (i.e. Ca 2ϩ and ROS).
ROS generation by NADPH oxidase and mitochondria essentially begins with the formation of superoxide (O 2 . ) (5,7,25 To conclude, our data support a conceptual model in gonadotropes where GnRH receptor activation promotes a localized subplasmalemmal ROS signaling mechanism mediated by H 2 O 2 , initiated by NADPH oxidase, and amplified by mitochondria that in turn stimulates localized Ca 2ϩ influx through L-type Ca 2ϩ channels (Fig. 10). Although several mechanistic uncertainties require further investigation, our functionally coupled Ca 2ϩ and ROS microdomain model implies that reciprocal modulation of L-type Ca 2ϩ channels and NADPH oxidase and mitochondrial ROS generation extend beyond simple colocalization of two signaling mechanisms. For example, our model suggests potential perturbation of numerous cellular processes by conditions such as metabolic dysfunction (e.g. obesity and diabetes) by influencing the availability of the reducing equivalent NADPH or by altering the bioenergetic activity of subplasmalemmal mitochondria. Thus, continued characterization of Ca 2ϩ and ROS signaling mechanisms will impact the discovery of novel therapeutic strategies to manipulate GnRH receptor function in the management of fertility, obesity-related endocrine dysfunction (37), polycystic ovarian syndrome (38,39), and GnRH-sensitive carcinomas (40 -42).

Mouse pituitary cell dissociation and gonadotrope isolation
Primary cells were from transgenic mice with genetically labeled fluorescent gonadotropes produced by a cross of GnRH receptor promoter-driven Cre recombinase (GRIC) mice with ROSA26-YFP or ROSA-tdTomato mice (9). Sexually mature ROS promote colocalized L-type calcium channel sparklets (6 -12 weeks of age) male and female mice were euthanized with sodium pentobarbital (200 mg/kg intraperitoneally) in strict accordance with institutional guidelines and with approval by the Institutional Animal Care and Use Committee of Colorado State University. Removed pituitaries were enzymatically digested in Ca 2ϩ -free buffer containing papain (10 units/ml) and DTT (1 mg/ml) for 15 min at 37°C followed by a second incubation (15 min at 37°C) in Ca 2ϩ -free buffer supplemented with collagenase (300 units/ml, type II). Digested pituitary tissue was then washed with and placed in high-glucose DMEM for 30 min after which trituration with a fire-polished Pasteur pipette was used to create a cell suspension for plating. Cells plated on poly-L-lysinecoated MatTek dishes were incubated overnight at 37°C in 5% CO 2 humidified air and used for experimentation within 24 -36 h. YFP-and tdTomato-labeled gonadotrope knockin mice were used to ensure consistency among mouse lines. Significant differences between groups were not observed; thus, data from the two groups were combined for analysis.

Detection of ROS generation
TIRF microscopy was used to visualize subplasmalemmal ROS generation as described previously (12,15). Briefly, we incubated ␣T3-1 cells in Ca 2ϩ -free buffer supplemented with the cell-permeant ROS indicator DCF (1 M) for 30 min at room temperature (22-25°C). Cells were then placed in either Ca 2ϩ -free or Ca 2ϩ -containing (2 mM) buffer for experimentation following removal of excess DCF. A 491 nm laser provided excitation of oxidized DCF; appropriate filters separated excitation and emission light. Due to the lack of specificity of DCF as an ROS indicator, increases in DCF fluorescence are interpreted only as an indication of intracellular oxidation without reference to specific oxidants (e.g. H 2 O 2 ).
For an area of elevated DCF fluorescence to be considered a site of increased ROS generation (a ROS "punctum"), a grid of 3 ϫ 3 contiguous pixels had to have a fluorescence amplitude equal to or larger than the mean basal DCF fluorescence plus 3 times the standard deviation (10,12). We calculated the density of ROS puncta (ROS puncta/m 2 ) by dividing the number of sites detected by the area of cell membrane visible in the TIRF images. For consistency and to ensure proper cell loading with DCF, cells chosen for analysis had one to two visible ROS puncta at baseline and showed changes in DCF fluorescence only in response to cell stimulation. We calculated changes in DCF fluorescence (⌬DCF) from the mean pixel intensities of the total intracellular submembranous slice visible in the TIRF images (average ⌬DCF) and the areas confined to identified ROS puncta (puncta ⌬DCF).

Electrophysiology and Ca 2؉ imaging
Simultaneous electrophysiology and TIRF Ca 2ϩ imaging experiments were carried out using the conventional dialyzed whole-cell patch clamp technique as described previously (4,10,12,15,43). Briefly, we imaged Ca 2ϩ influx with a TILL Photonics (FEI, Munich, Germany) through-the-lens TIRF built around an inverted Olympus IX-71 microscope (Center Valley, PA) equipped with a 100ϫ TIRF oil-immersion objective (numerical aperture, 1.45) and an iXON electron-multiplying charge-coupled device camera (Andor Technology, South Windsor, CT). Ca 2ϩ influx was visualized in gonadotropes using the fluorescent Ca 2ϩ indicator fluo-5F (200 M) and an excess of EGTA (10 mM) introduced via the patch pipette. An Axopatch 200B amplifier (Molecular Devices, Sunnyvale, CA) was used to control membrane potential, a 491 nm laser provided fluo-5F excitation, and appropriate filters separated excitation and emission light. Ca 2ϩ influx was recorded in the presence of 2 mM external Ca 2ϩ at a frame rate of 50 Hz and holding membrane potential of Ϫ70 mV to increase the driving force for Ca 2ϩ entry. To preclude potential contaminating Ca 2ϩ release events from the endoplasmic reticulum, the Ca 2ϩ -ATPase inhibitor thapsigargin (1 M) was present in all experiments. We imaged cells for 2 min before applying GnRH (10 nM), H 2 O 2 (100 M), nicardipine (10 M), apocynin (25 M), catalase (500 units/ml), antimycin A (500 nM), or mitoTEMPO (25 nM). Imaging of all treatment conditions continued for a minimum of 10 min. We performed vehicle control experiments as appropriate; all experiments were carried out at room temperature (22-25°C).

L-type Ca 2؉ channel sparklet analysis
Following background subtraction, we converted fluo-5F fluorescence signal image stacks to intracellular Ca 2ϩ concentrations ([Ca 2ϩ ] i ) and analyzed these data with custom software as described previously (4,12,15,43). Briefly, we quantified L-type Ca 2ϩ channel sparklet activity by calculating the nP s of each site where n is the number of quantal levels detected and P s is the probability that the site is active. nP s values were obtained using pCLAMP 10.0 (Molecular Devices) on imported [Ca 2ϩ ] i time course records using an initial unitary [Ca 2ϩ ] i elevation of Ϸ20 nM as determined empirically. For an elevation in [Ca 2ϩ ] i to be considered an L-type Ca 2ϩ channel sparklet event, a grid of 3 ϫ 3 contiguous pixels had to have a [Ca 2ϩ ] i amplitude equal to or larger than the mean basal [Ca 2ϩ ] i plus 3 times the standard deviation. Consistent with previous reports (10 -12, 43), we observed a bimodal distribution of L-type Ca 2ϩ channel sparklet activity in gonadotropes with sites of low activity (nP s between 0 and 0.2) and high activity (nP s greater than 0.2). We calculated active L-type Ca 2ϩ channel sparklet site densities (Ca 2ϩ sparklet sites/m 2 ) by dividing the number of active sites by the area of cell membrane visible in the TIRF images. Image stacks of Ca 2ϩ sparklet activity were analyzed after a minimum of 5 min and before a maximum of 10 min following each manipulation in all experiments to maintain consistency and enhance reproducibility.

Imaging of subplasmalemmal mitochondria
We used TIRF microscopy to image subplasmalemmal mitochondria in ␣T3-1 cells. Briefly, cells were incubated with Mito-Tracker Green (200 nM) for 15 min in Ca 2ϩ -free buffer at 37°C. Subplasmalemmal MitoTracker Green was excited with a 491 nm laser; appropriate filters separated excitation and emission light. For an area of elevated MitoTracker fluorescence to be considered indicative of subplasmalemmal mitochondria, the fluorescence amplitude had to be equal to or larger than the mean basal fluorescence plus 3 times the standard deviation. Using this criterion, we generated thresholded MitoTracker ROS promote colocalized L-type calcium channel sparklets TIRF images to establish clear mitochondrial boundaries. We calculated the percentage of plasma membrane associated with subplasmalemmal mitochondria by dividing the area of mitochondria-associated membrane by the area of plasma membrane visible in the TIRF field.
For the experiments where we imaged subplasmalemmal mitochondria (first) and Ca 2ϩ influx (second), we defined mitochondria-associated L-type Ca 2ϩ channel sparklet sites as those sites with peaks (pixels of highest intensity) less than or equal to 0.5 m from the edge of the nearest thresholded MitoTracker signal. Euclidean distance mapping analysis was used to quantitate the distance of observed Ca 2ϩ sparklet site peaks from the nearest thresholded MitoTracker signal and 130 randomly distributed points located within the visible TIRF footprint of the cells analyzed. Each cumulative distribution was fit with a single exponential function, Y ϭ Y 0 ϩ (plateau Ϫ Y 0 ) ϫ [1 Ϫ exp(Ϫ(ln2/X 0.5 ) ϫ X)] where Y is the cumulative frequency, Y 0 is the Y value when X (distance) is zero, plateau is the Y value at infinite times, and X 0.5 (half-distance) is X where 50% of the Y values are distal to X ϭ zero.

Mitochondrial oxidation measurements
TIRF microscopy was used to assess the relative oxidative status within subplasmalemmal mitochondria in ␣T3-1 cells using the cell-permeant fluorescent mitochondrial indicator MitoSOX-Red. We incubated ␣T3-1 cells in Ca 2ϩ -free buffer supplemented with MitoSOX-Red (1 M) for 30 min at 37°C. A 491 nm laser provided MitoSOX-Red excitation; appropriate filters separated excitation and emission light. Cells were exposed to GnRH (10 nM) or antimycin A (500 nM) with or without pretreatment (10 min) with CCCP (1 M) or apocynin (25 M). We analyzed MitoSOX-Red fluorescence intensity (indicative of mitochondrial oxidation) on background-subtracted images. For an area of elevated MitoSOX-Red fluorescence to be considered indicative of oxidation within subplasmalemmal mitochondria, the fluorescence amplitude had to be equal to or larger than the mean basal fluorescence plus 3 times the standard deviation. Using this criterion, we identified normalized regions of interest to compare mitochondrial oxidation before and after treatments by comparing the fluorescence intensity. The cell outline (white dotted line) was generated by the custom software used to identify cellular borders visible in the TIRF field for Ca 2ϩ sparklet analyses (4,12,15,43).

Confocal microscopy
We used laser-scanning confocal microscopy to image plasma membranes and mitochondria of ␣T3-1 cells. The extracellular face of the plasma membrane was marked with Alexa 555-WGA (5 g/ml) in Ca 2ϩ -free buffer for 5 min at room temperature. Mitochondria were labeled with Mito-Tracker Green (200 nM for 15 min) in Ca 2ϩ -free buffer at 37°C. Alexa Fluor 555 was excited with a 543 nm laser, and Mito-Tracker Green was excited with a 488 nm laser; appropriate filters separated excitation and emission light. We analyzed data with Volocity 3D Image Analysis Software (PerkinElmer Life Sciences). Mitochondria-associated fluorescence located Յ0.5 m from the center of the Alexa 555-WGA signal was designated as peripheral.

Statistics
Statistical analyses were performed using GraphPad Prism 6 software. We have presented normally distributed data as the mean Ϯ S.E. Two-sample comparisons of these data were performed using either a paired or unpaired (as appropriate) twotailed Student's t test, and comparisons between more than two groups were performed using a one-way analysis of variance with Tukey's multiple comparison post-test. L-type nP s data sets were bimodally distributed (10,12,15). Thus, we performed two-sample comparisons of nP s data using the nonparametric Wilcoxon-Mann-Whitney test (two-tailed), whereas comparisons between more than two groups were performed using the nonparametric Friedman test with Dunn's multiple comparison post-test. Arithmetic means of nP s data sets are indicated in the figures (solid gray horizontal lines) for nonstatistical visual purposes, and dashed gray lines mark the threshold for high-activity Ca 2ϩ sparklet sites (nP s Ն 0.2) (10,12,43). A p value of Ͻ0.05 was considered significant, and asterisks (*) used in the figures are included to indicate significance, ns indicates not significantly different, and n indicates the number of independent experiments performed on single cells unless stated otherwise.