Estrogen Enhances Linkage in the Vascular Endothelial Calmodulin Network via a Feedforward Mechanism at the G Protein-coupled Estrogen Receptor 1*

Estrogen exerts many effects on the vascular endothelium. Calmodulin (CaM) is the transducer of Ca2+ signals and is a limiting factor in cardiovascular tissues. It is unknown whether and how estrogen modifies endothelial functions via the network of CaM-dependent proteins. Here we show that 17β-estradiol (E2) up-regulates total CaM level in endothelial cells. Concurrent measurement of Ca2+ and Ca2+-CaM indicated that E2 also increases free Ca2+-CaM. Pharmacological studies, gene silencing, and receptor expression-specific cell studies indicated that the G protein-coupled estrogen receptor 1 (GPER/GPR30) mediates these effects via transactivation of EGFR and subsequent MAPK activation. The outcomes were then examined on four distinct members of the intracellular CaM target network, including GPER/GPR30 itself and estrogen receptor α, the plasma membrane Ca2+-ATPase (PMCA), and endothelial nitric-oxide synthase (eNOS). E2 substantially increases CaM binding to estrogen receptor α and GPER/GPR30. Mutations that reduced CaM binding to GPER/GPR30 in separate binding domains do not affect GPER/GPR30-Gβγ preassociation but decrease GPER/GPR30-mediated ERK1/2 phosphorylation. E2 increases CaM-PMCA association, but the expected stimulation of Ca2+ efflux is reversed by E2-stimulated tyrosine phosphorylation of PMCA. These effects sustain Ca2+ signals and promote Ca2+-dependent CaM interactions with other CaM targets. Consequently, E2 doubles CaM-eNOS interaction and also promotes dual phosphorylation of eNOS at Ser-617 and Ser-1179. Calculations using in-cell and in vitro data revealed substantial individual and combined contribution of these effects to total eNOS activity. Taken together, E2 generates a feedforward loop via GPER/GPR30, which enhances Ca2+/CaM signals and functional linkage in the endothelial CaM target network.

It is not known whether and how estrogen affects endothelial cell functions through the network of CaM-binding proteins. Given the ubiquitous role of CaM in signaling and its limiting nature, factors that regulate CaM expression and linkage among CaM targets probably affect tissue functions substantially. Here, we used multiple approaches to identify the effects of 17␤-estradiol (E 2 ) on total and free cellular CaM levels in endothelial cells, the estrogen receptor responsible for these effects and the underlying mechanisms, the resultant changes in interactions between CaM and four distinct CaM targets, and the associated functional impact. The targets examined include ER␣, the novel CaM target GPER/GPR30, the plasma membrane Ca 2ϩ -ATPase (PMCA), and endothelial nitric-oxide synthase (eNOS). We demonstrate that E 2 generates a feedforward mechanism involving GPER/GPR30 that enhances Ca 2ϩ /CaM signals and functional linkage in the CaM network in vascular endothelial cells.

Experimental Procedures
Cell Culture and Isolation-Primary porcine aortic endothelial cells (PAECs) were obtained as described previously (23)(24)(25)(26). Briefly, the descending portions of thoracic aortas were freshly isolated from a local slaughterhouse. After removal of perivascular adipose tissues, the aortas were dissected, and the intima was mechanically collected using a sterile scalpel. The cell pellets were resuspended in phenol red-free M199 medium (Caisson Laboratories, Logan, UT) containing 20% newborn calf serum (Fisher) and 1% penicillin-streptomycin (MP Biomedicals, Solon, OH) and plated on culture dishes until a monolayer of typical endothelial morphology was obtained. This approach consistently yields highly pure populations of primary endothelial cells, with Ͼ95% purity. In our pilot studies, the use of phenol red-containing medium affected the primary findings, whereas there was no difference between charcoal-stripped and regular sera (newborn calf or fetal bovine sera), despite a much slower growth rate in cells cultured using charcoal-stripped sera. All cells were thus cultured in phenol red-free medium containing regular sera. PAECs were used between passages 1 and 2. Human embryonic kidney 293 (HEK293) and SKBR3 cells (ATCC) were cultured in phenol red-free DMEM containing 10% fetal bovine serum (Fisher). Cells were cultured in a 37°C incubator with 5% CO 2 humidified air. The medium was frequently renewed.
Molecular Biology-Amino acid substitutions in the CaMbinding domains of GPER/GPR30 in SMD2, -3, -4, or a combination thereof (Table 1) were introduced into full-length GPER/ GPR30 using custom gBlock gene fragments (Integrated DNA Technologies, Coralville, IA) between the SbfI and NotI restriction sites in a pEZ plasmid encoding human GPER/GPR30 (Genecopoeia Inc.). The mutant CaM-binding sequences were PCR-amplified and incorporated as inserts between the KpnI and AgeI restriction sites located between a "citrine" version of enhanced yellow fluorescent protein (EYFP C ) and enhanced cyan fluorescent protein (ECFP) framework of previously reported FRET-based biosensors for GPER/GPR30 (27). We term these biosensors BSGPER x , where x denotes the amino acid numbering of the insert sequences. The resultant biosensors were BSGPER 150 -175mut , BSGPER 242-259mut , and BSGPER 330 -351mut , corresponding to the CaM-binding sequences in GPER/GPR30 SMD2, -3, and -4. The primers for these reactions are presented in Table 2. The biosensors were incorporated in a pET bacterial expression vector, as described previously (27). All constructs were verified by DNA sequencing (University of Missouri, Columbia, MO).
Quantitative RT-PCR-Total mRNA was isolated using an RNeasy minikit (catalog no. 74104, Qiagen) and SpinSmart total RNA mini purification kit (catalog no. CM-61050, Denville Scientific). Reverse transcription was performed at 100 ng/l RNA using the High Capacity cDNA reverse transcription kit (catalog no. 4368814, Applied Biosystems). For realtime PCR, each reaction included 5 l of cDNA (1:20 dilution from RT reaction yield), 8.5 l of SYBR Green (catalog no. 1725122, Bio-Rad), 2.5 l of nuclease-free H 2 O, and 1 l of 5 M forward and reverse primer mixture. All reactions were seeded in triplicates for each sample onto a 96-well plate (catalog no. HSP9655, Bio-Rad) sealed with a clear cover (catalog no. MSB1101, Bio-Rad). After brief centrifugation of the plates, quantitative PCR was run on a Chromo4 real-time system (Bio-Rad) with the following sequence: 3-min initiation at 98°C, followed by 60 cycles of 15-s denaturation at 98°C, 30-s annealing at 60°C, and 30-s extension at 72°C. A melt curve was subsequently performed for all reactions to ensure single amplicon yield by the corresponding primers, with a 0.5°C ramp/5 s, from 60 to 98°C. Primers were designed using NCBI Primer Blast. Porcine CALM1 gene was used for CaM, and HMBS (hydroxymethylbilane synthase) was used as the reference gene. The primers and relevant details are listed in Table 3.

TABLE 1 Wild-type and mutant sequences of the GPER/GPR30 CaM-binding sequences in SMD2, -3, and -4 to change hydrophobicity and charges
Substitutions are in italic type and underlined.  Efficiencies for both primer pairs were determined from standard curves generated by quantitative PCR. Average data from triplicate curves up to 20% of fractional fluorescence were fit to the exponential equation.
Crossing points (CPs) were determined as x values corresponding to y ϭ 0.1 on fitted curves. ⌬CP values for CaM were corrected for corresponding values of HMBS, and induction of CaM mRNA was expressed as -fold change based on the mathematical model by Pfaffl (28), where R is the -fold change in CaM mRNA, and E CaM and E HMBS are efficiencies of the primer pairs for CaM and HMBS, respectively. Transfection-Transfection of GPER/GPR30 antisense and scrambled oligonucleotides was carried out using the Transit-Oligo transfection kit (Mirus LLC). pEZ plasmids encoding wild-type and mutant GPER/GPR30 containing mutations in one or more of its CaM-binding domains were transfected into HEK293 cells using the Transit-LT1 transfection kit (Mirus LLC). Transfection of the pCDNA3.1 plasmid encoding the CaM biosensor BSCaM 2 (17, 18, 20, 29 -31) into primary PAECs was performed using a Nucleofector II-S transfection system (Lonza Inc., Walkersville, MD).
GPER/GPR30 Gene Silencing-GPER/GPR30 knockdown was performed as described previously (23), using an antisense oligonucleotide directed against the porcine GPER/GPR30 sequence. A control oligonucleotide was randomly scrambled from the antisense sequence. The efficiency of this approach has been carefully validated previously using immunoblotting of total cell lysate and immunofluorescence in single cell imaging experiments (23).
Simultaneous Measurement of Free Intracellular Ca 2ϩ and Ca 2ϩ -CaM Concentrations-Primary PAECs transfected with BSCaM 2 were treated with or without E 2 for 48 h. Cells were then loaded with 4 M indo-1/AM for 30 min at 37°C. BSCaM 2 and indo-1 excitation was alternated between 430 nm and 340 nm for 50 ms per 500-ms cycle through sputtered excitation filters housed in a Lambda-DG4 illuminator (Sutter Instrument Co.) with a 1-ms switching interval between wavelengths. Emission lights were passed through a custom 410/30M/460LP polychroic beam splitter (Chroma Technology Inc.) and selected through sputtered bandpass filters of 480 and 535 nm for BSCaM 2 , and 405 and 485 nm for indo-1, housed in a Lambda-10 filter switcher (Sutter Instrument) with switching intervals of 40 ms between wavelengths. Filtered emission lights were then collected via an Andor DU-885 EMCCD camera. Data were processed by Imaging Workbench version 6.1 software (INDEC Biosystems). R min values for indo-1 were obtained similarly to fura-2 described earlier (23). R min values for BSCaM 2 were obtained similarly, because the interaction between CaM and BSCaM 2 is strictly Ca 2ϩ -dependent. To obtain R max values of both indo-1 and BSCaM 2 , 50 M ␤-escin was added at the end of each experiment to permeabilize cells in the presence of saturating concentrations of Ca 2ϩ and CaM. Using these values, free Ca 2ϩ and Ca 2ϩ -CaM concentrations were calculated as described earlier (17,18).
Co-immunoprecipitation and Western Blotting-Following treatment and cell lysis, protein content was determined using BCA assay (Pierce) performed in triplicate for all samples, with a standard curve generated for every determination, and the protein content for each sample was determined based on the average of the triplicate. Co-immunoprecipitation was performed using protein A/G beads as per the manufacturer's instructions (Fisher). Three milligrams of cellular proteins were then rocked with protein A/G beads overnight at 4°C prior to centrifugation and collection of supernatant. For regular Western blotting, all lanes in each gel were loaded with equal amounts of total proteins, adjusted for detectability with each antibody. After electrophoresis and transfer, membranes from coimmunoprecipitation samples were separated between the levels of the bait and the prey proteins prior to incubation with specific primary antibodies. For immunoblots assessing protein expression levels, when possible, antibodies recognizing the protein of interest and the housekeeping protein were mixed on the same membrane fragment in a "double primary" approach to show both proteins on the same development. Otherwise, membranes were cut between the levels of the protein of interest and the housekeeping protein, and fragments of the same membranes were probed separately with the respective antibodies. Following secondary antibody treatment, membrane fragments were developed using enhanced chemiluminescence (Thermo Scientific) and detected with film (Hyblot, Denville Scientific). To ensure linearity of the protein bands in analyses, multiple film exposures were taken to avoid either overexposure or underexposure. For co-immunoprecipitation data, densitometric values of the baits were corrected for those of the preys calculated using the rectangular volume tool in the Image Lab version 5.0 software (Bio-Rad). For protein expression blots, densitometric values of the protein of interest were used and normalized against control values, based on equal BCAbased inputs at the front end; nevertheless, matching blots of a housekeeping protein were shown alongside in all cases. Measurement of PMCA Activity-PMCA activity was measured in PAECs as described previously (23), with slight modifications. Cells were plated on number 1.5 glass coverslips and grown to subconfluence. Following treatment, cells were loaded with 4 M fura-2/AM for 30 min at 37°C. After removal of fura-2/AM, cells were equilibrated at room temperature in modified Tyrode's buffer (150 mM NaCl, 2.7 mM KCl, 1.2 mM KH 2 PO 4 , 1.2 mM MgSO 4 , 10 mM HEPES, 1 mM CaCl 2 , pH 7.4) for 15 min. Fura-2 was excited at 340 nm and 380 nm for 50 ms per 500-ms cycle using the system described under "Simultaneous Measurement of Free Intracellular Ca 2ϩ and Ca 2ϩ -CaM Concentrations." The ratio of emission intensities following excitation at 340 nm and 380 nm was used to calculate the free Ca 2ϩ concentration using the equation, where the K d value is 224 nM; R is the observed ratio fluorescent signal during the experiment. S f and S b are the emission intensities collected at 510 nm corresponding to the Ca 2ϩ -free and Ca 2ϩ -bound states, respectively. To minimize interexperiment variance, R max and R min were obtained for individual cells in every experiment, as described in detail previously (23). After post-loading equilibration, thapsigargin (1 M) was added in Ca 2ϩ -free buffer to deplete the estrogen receptor. Extracellular Na ϩ was then replaced by equimolar N-methyl-glucamine to prevent the Na ϩ -Ca 2ϩ exchanger. Ca 2ϩ entry was then initiated by adding 1.5 mM CaCl 2 . When peak influx was reached, extracellular Ca 2ϩ was removed in the presence of 5 mM BAPTA. The rate of reduction in intracellular Ca 2ϩ concentration now mainly reflected Ca 2ϩ extrusion via the PMCA and was determined by fitting the time course of the apparent free Ca 2ϩ concentration to the monoexponential equation, y ϭ y 0 ϩ Ae Ϫx/t . Extrusion rates of Ca 2ϩ are presented as relaxation times (). Because the activity of PMCA is Ca 2ϩ -dependent, comparisons were made only among cells in which free Ca 2ϩ concentrations at the beginning of the Ca 2ϩ extrusion time course were in the same ranges (17,18,23). Relative PMCA activity was expressed as the inverse of the relaxation time.
Expression and Purification of Proteins-pET plasmids encoding FRET-based biosensors and CaM were expressed in BL21(DE3) competent cells. Biosensors and CaM were purified using affinity chromatography as described in detail previously (27).
Determination of CaM Binding Affinities of Mutant Biosensors-Biosensors containing mutations in the CaM-binding domains of GPER were tested for CaM binding at 22°C using a QuantaMaster TM -40 spectrofluorometer (Photon Technology International Inc.) as described previously (27) with modifications for mutant biosensors that require large amounts of CaM for saturation. Mutant BSGPER x (0.5 M) was mixed in a quartz cuvette (Hellma Analytics) containing a titration buffer (25 mM Tris, 100 mM KCl, pH 7.5, 0.1 mg/ml BSA) and an appropriate amount of Ca 2ϩ sufficient to saturate the maximal amount of Ca 2ϩ binding sites on titrated CaM. CaM was titrated in the mixture as biosensor emission spectra were monitored from 460 to 560 nm in response to ECFP excitation at 430 nm. We defined positive interaction between CaM and BSGPER x based on three diagnostic criteria: 1) an increase in ECFP emission, 2) a corresponding decrease in EYFP C emission, and 3) a crossing of emission spectra from CaM-free versus CaM-added conditions at the isoemissive point of 510 nm. From these spectral data, fractional biosensor responses (BS fract ) upon CaM addition were calculated using the formula, where R min and R max are the ratios between emission intensities at 476 and 528 nm (F476/F528) when the biosensor is in unbound and maximally bound states, respectively. For reactions that required a large concentration of CaM to saturate biosensor responses, to avoid dilution effects, titrations were successively carried out in identical single cuvettes containing initially a fixed concentration of biosensor in titration buffer and a different concentration of CaM in the presence of an appropriate concentration of 1,2-bis(o-aminophenoxy)ethane N,N,NЈ,NЈ-tetraacetic acid to essentially prevent any Ca 2ϩ -dependent CaM interaction with the biosensor. Following measurement of biosensor emission spectra in the Ca 2ϩ -free condition, reactions were initiated by the addition of saturating Ca 2ϩ (sufficient to saturate all Ca 2ϩ binding sites in the amount of CaM added) using high concentration Ca 2ϩ stock. Biosensor ratios from individual measurements upon the addition of saturating Ca 2ϩ were normalized against the respective initial biosensor ratio obtained prior to the addition of Ca 2ϩ . Normalized biosensor ratios were then used to obtain biosensor fractional responses, using Equation 1. Because most mutant biosensors bind CaM with fairly low affinities, apparent K d values for mutant biosensor and CaM interactions were obtained by fitting the biosensor's fractional responses as a function of CaM to the quadratic equation,

Estrogen, GPER/GPR30, and Endothelial Calmodulin Network
where BS fract , [BS], and [CaM] are the fractional response of BSGPER x and total concentration of biosensor and CaM in the mixture, respectively.
Ca 2ϩ sensitivity of biosensor-CaM interactions was determined as described previously (27), with Mag-indo1 used in place of XRhod-5F for a number of reactions where the EC 50 (Ca 2ϩ ) values of CaM-biosensor interactions were initially observed to be substantially higher than the K d values for XRhod-5F. EC 50 (Ca 2ϩ ) values for biosensor-CaM interactions were obtained by fits of fractional saturation-free Ca 2ϩ relationships to Equation 4, where n is the Hill coefficient.
Statistical Analysis-Data are expressed as means Ϯ S.E. Statistical analysis was done using Student's t test, assuming unequal variances between control and treated groups. Statistical significance was determined at p Ͻ 0.05.

E 2 Increases Total CaM and Free Ca 2ϩ -CaM in Endothelial
Cells-CaM expression is limiting in the endothelium and other cardiovascular tissues. To examine the effect of E 2 on CaM expression, we used primary porcine aortic endothelial cells (PAECs) freshly isolated and cultured until subconfluence. These cells have been used frequently to investigate endothelial biology (23,26,(32)(33)(34)(35). Primary PAECs were treated with vehicle or various doses of E 2 for 48 h, followed by immunoblotting for ␤-actin and CaM. E 2 dose-dependently increased CaM expression, with maximal effect at 5 nM E 2 (Fig. 1A). Because most cellular CaM is engaged in inseparable interactions (15), up-regulated total CaM may not be associated with increases in free Ca 2ϩ -CaM available for dynamic interactions. Because Ca 2ϩ -CaM signals are dictated by both the initial Ca 2ϩ signals and available CaM, it was important to measure simultaneously both free Ca 2ϩ and Ca 2ϩ -CaM levels. To assess the effect of E 2 on free Ca 2ϩ -CaM levels, we transfected PAECs with BSCaM 2 , a genetically coded FRET-based biosensor for Ca 2ϩ -CaM (17,18,20,30,31). Cells were then exposed to 5 nM E 2 for 48 h, followed by loading with 4 M indo-1/AM. Fig. 1B shows indo-1 fluorescence in a group of PAECs (top left), with ϳ40% expressing BSCaM 2 (top right). Only cells expressing BSCaM2 and loaded with indo ( Fig. 1B, merged image, bottom) were chosen for dual imaging. Free Ca 2ϩ and Ca 2ϩ -CaM signals were then concurrently monitored, with a typical time course shown in Fig. 1, C (indo-1) and D (BSCaM 2 ) (see description under "Experimental Procedures"). To avoid interexperimental variability in BSCaM 2 expression levels, which could complicate the interpretation due to buffering effects, care was taken to compare cells with similar BSCaM 2 fluorescent intensities in unstimulated, basal conditions. E 2 treatment significantly increased the average free Ca 2ϩ -CaM signal stimulated by ionomycin (Fig. 1E). The plot of free Ca 2ϩ -CaM as a function of Ca 2ϩ during the influx and extrusion phases shows increased Ca 2ϩ -CaM complexes in E 2 -treated cells (Fig. 1F).

GPER/GPR30 Mediates the Effect of E 2 to Up-regulate CaM
Expression-To date there are three known receptors for estrogen: ER␣, ER␤, and GPER/GPR30. To begin identifying the specific receptor(s) responsible for the E 2 effect of up-regulating CaM, primary PAECs were exposed to different doses of PPT, DPN, and G-1 for 48 h to specifically stimulate ER␣ (EC 50 ϭ 200 pM), ER␤ (EC 50 ϭ 0.85 nM), and GPER/GPR30 (EC 50 ϭ 11 nM), respectively. ICI182,780 is a nonselective ER␣/ER␤ antagonist (IC 50 0.29 nM) and is also a GPER/GPR30 agonist (4,9), which represents a useful tool to identify the responsible receptor. Fig.  2 shows the total CaM expression levels from cells treated with PPT ( Fig. 2A), DPN (Fig. 2B), G-1 (Fig. 2C), and ICI182,780 (Fig. 2D). Only G-1 up-regulated total CaM. Interestingly, ICI182,780 also increased CaM expression, with the maximal effect at 20 M even higher than that produced by 5 nM E 2 (Fig.  2D). These data suggest that GPER/GPR30 may mediate the E 2 effect of up-regulating CaM expression. To assess the specificity of the effect of G-1 on CaM, effects of G-1 were examined on two other Ca 2ϩ -binding proteins, including annexin V and calreticulin. Treatment with different doses of G-1 for 48 h did not affect expression levels of either protein (Fig. 2, E and F).
To more specifically confirm the role of GPER/GPR30, we tested the effect of E 2 on CaM expression in PAECs in which GPER/GPR30 has been knocked down. Primary PAECs were transfected with scrambled or antisense oligonucleotides directed against porcine GPER/GPR30 sequence for 32 h as described previously (23), followed by 48-h treatment with vehicle or 5 nM E 2 . Fig. 3A shows the effects of scrambled and antisense oligonucleotides on GPER/GPR30 expression, with ϳ65% silencing efficiency. Treatment with E 2 for 48 h increased total CaM expression by ϳ5.5-fold in PAECs transfected with scrambled oligonucleotide, consistent with the data in Fig. 1A, but only increased CaM expression in GPER/GPR30 antisensetransfected cells by 1.8-fold, equivalent to a 68% reduction in the effect to up-regulate CaM expression (Fig. 3B).
To further confirm the role of GPER/GPR30 in mediating the effect of E 2 to up-regulate CaM expression, we tested the effect of E 2 on SKBr3 cells, a cell type known to express only GPER/ GPR30 and not ER␣ or ER␤ (9,36). Lysate of PAECs and SKBr3 cells were first probed for all three receptors to verify their expression profiles. Primary PAECs expressed all three receptors, whereas SKBr3 cells only expressed GPER/GPR30 (Fig.  3C). SKBr3 cells were then subjected to E 2 treatment for 48 h, which dose-dependently up-regulated total CaM expression (Fig. 3D). Together, these data strongly indicate that GPER/ GPR30 mediates the E 2 effect of up-regulating CaM expression.
GPER/GPR30 Up-regulates CaM mRNA via Transactivation of EGFR and MEK1-Having demonstrated that GPER/GPR30 is the key receptor responsible for the effect of E 2 to up-regulate CaM, we sought to examine the underlying mechanism. GPER/ GPR30 activation has been shown to transactivate EGFR via the release of the membrane-tethered HB-EGF and subsequent activation of MAPK (37,38). This mechanism was shown to be responsible for a number of genomic effects of GPER/GPR30 activation, such as up-regulating the expression of c-fos, cyclin A and D1, and fatty acid synthase (5)(6)(7)(8). We hypothesized that a similar mechanism could account for the effects of GPER/ GPR30 to up-regulate CaM expression at the transcriptional level. We thus tested the effects of AG-1478 and PD-98059, specific inhibitors of EGFR and MEK1, on the effect of G-1 to induce CaM mRNA, using quantitative RT-PCR. Primary PAECs were treated with vehicle; various doses of AG-1478, PD-98059, and G-1 alone or in combination for 48 h prior to lysis; and total RNA extraction. Fig. 4A shows the fractional fluorescence of the titration curves for the primers of the porcine CALM1 gene (see "Experimental Procedures" and Table   3). Fig. 4 (B and C) shows exponential fits of up to 20% of fractional data in Fig. 4A and the residual of a representative fit. Fig.  4D shows the average melt curves following the experiment in Fig. 4A, demonstrating a single product using the CaM primer set. Similar data were obtained for the reference gene HMBS (not shown). Using this system, inhibition of EGFR with AG-1478 caused insignificant changes in CaM mRNA. Stimulation of GPER/GPR30 for 48 h was associated with a ϳ2-fold up-regulation in CaM mRNA, and AG-1478 prevented the effect of G-1 (Fig. 4E). Similarly, inhibition of MEK1 using PD-98059 caused insignificant changes in CaM mRNA at the doses tested but significantly prevented the up-regulation in CaM mRNA induced by treatment with G-1 (Fig. 4F). These data are consistent with the idea that GPER/GPR30 activation up-regulates CaM expression at the transcriptional level via transactivation of EGFR and subsequent activation of the MAPK cascade in vascular endothelial cells.  (17,18), the observed effect of E 2 to up-regulate total and free CaM is probably of broad impact on many CaM target proteins. We thus decided to test this idea by examining a number of CaM targets with different affinities for CaM. Among the known receptors for estrogen, ER␤ does not interact with CaM, whereas estrogen receptor ER␣ is well known to be regulated by Ca 2ϩ -dependent CaM interaction with nanomolar affinity Primary PAECs were treated with the specified doses of the respective agonists for 48 h prior to lysis, followed by simultaneous immunoblotting of ␤-actin and CaM on the same membranes. SDS-PAGE membranes for annexin V and calreticulin were cut between the levels for these proteins and vinculin, followed by simultaneous immunoblotting as indicated. Data are averages from n ϭ 6 for A-D and n ϭ 3 for E-F; *, p Ͻ 0.05. (39). CaM binding to ER␣ enhances the receptor's stability, dimerization, and transcriptional activities (40 -42). We recently showed that CaM interacts with GPER/GPR30 at four sites located separately in the receptor's four submembrane domains (SMDs) with disparate affinities and Ca 2ϩ sensitivities (27). In biosensor formats, these domains interact with CaM with submicromolar or low micromolar affinities. In-cell studies and Ca 2ϩ sensitivity data indicated that GPER/GPR30-CaM interaction can occur at resting Ca 2ϩ levels in cells (27). It is completely unknown whether E 2 can affect the activities of its own receptors via changes in CaM binding. Thus, primary PAECs were exposed to vehicle or 5 nM E 2 for 48 h. Cells were subsequently treated with 2 M ionomycin in the presence of 1 mM extracellular Ca 2ϩ for 3 min prior to lysis for analyses of receptor expression and interaction with CaM. The use of ionomycin was to guarantee that the Ca 2ϩ signals produced in cells were not influenced by potential effects on agonist-induced Ca 2ϩ entry that E 2 might have. E 2 did not change the expression of either receptor (Fig. 5, A and B) but significantly enhanced the association of CaM with both ER␣ (Fig. 5C) and GPER/ GPR30 (Fig. 5D). CaM immunoblots from the same lysate confirmed CaM up-regulation in cells exposed to E 2 (Fig. 5E).
CaM Binding Promotes GPER/GPR30 Function-For the two estrogen receptors that interact with CaM, ER␣, and GPER/ GPR30, the functional role of ER␣-CaM interaction has been well documented through many studies (40,(42)(43)(44); however, the functional effect of CaM-GPER/GPR30 association (27) is not known. The likelihood that CaM promotes GPER/GPR30 function is important in that GPER/GPR30 itself mediates the E 2 effect of up-regulating total and free CaM expression (Figs.  1-3). Because CaM interacts with GPER/GPR30 at four separate domains on the receptor, we generated mutations in the higher affinity CaM-binding domains, amino acids 150 -175, 242-259, and 330 -351, located on GPER/GPR30 submembrane domains 2, 3, and 4, to alter their hydrophobicity and charge properties (Table 1). We recently developed a biosensor approach to identify these CaM-binding sites and characterized their binding properties (27). We decided to utilize this technique to verify modifications in CaM binding properties of the mutant sequences. Fig. 6A encapsulates the biosensor principle. A CaM-binding domain is inserted between donor ECFP and acceptor EYFP C . In the unbound state, fluorescent resonance energy transfer (FRET) is robust between ECFP and EYFP C due to the proximity between the donor and acceptor fluorophores (Fig. 6A (i)). Upon interaction with CaM, conformational changes in the insert sequence disrupt FRET, increasing ECFP emission while decreasing that of EYFP C (Fig. 6A (ii)). Fig. 6A (iii) shows typical biosensor responses in the unbound and bound states. We used three diagnostic criteria for positive CaM binding: 1) increased ECFP emission, 2) a reciprocal decrease in EYFP C emission, and 3) crossing of the spectra at the isoemissive point around 510 nm. The emission ratio between ECFP and EYFP C then allowed quantitation of CaM binding properties with high precision (27). These biosensors are termed BSGPER x , with x denoting the amino acid numbering of the insert sequences. Ca 2ϩ -saturated CaM was titrated into purified mutant biosensors to obtain binding affinities. Panels i in Fig. 6, B-D, show biosensor spectral changes upon Ca 2ϩ -CaM titration to BSGPER 150 -175mut , BSGPER 242-259mut , and BSGPER 330 -351mut , respectively. All mutant biosensors still bind CaM, based on the diagnostic criteria. However, plots of biosensor fractional saturation as a function of CaM (panels ii in Fig. 6, B-D) showed significant reductions in binding affinities compared with previously determined wild-type responses (27). To assess the Ca 2ϩ sensitivities of interactions between CaM and the mutant biosensors, we concurrently measured the responses of biosensor and Ca 2ϩ indicator Mag-Indo1 (panel iii in Fig. 6B) or X-Rhod5F (panels iii in Fig. 6, C and D) as Ca 2ϩ was titrated into a mixture of biosensor and saturating CaM concentration. Fractional saturation-Ca 2ϩ relationships showed reduced Ca 2ϩ sensitivi- ties for the interactions between CaM and all mutant biosensors versus wild-type counterparts (panels iii in Fig. 6, B-D). Fits of biosensor responses and corresponding Ca 2ϩ concentrations to Equation 4 gave EC 50 (Ca 2ϩ ) values for CaM interactions with BSGPER 150 -175mut , BSGPER 242-259mut , and BSGPER 330 -351mut . All mutant biosensors have significantly reduced Ca 2ϩ sensitivities in their interactions with CaM (Table 4).
We and others have demonstrated that G-1-induced ERK1/2 phosphorylation is an indicator of GPER/GPR30 activity (9,23,45). To test the impact of CaM binding on GPER/GPR30 functions, we introduced the mutations verified in Fig. 6 into the full-length receptor and expressed them in HEK293 cells to test the effects of reduced CaM binding on GPER/GPR30-mediated ERK1/2 phosphorylation. HEK293 cells were used in these experiments to ensure high transfection efficiency. GPER/ GPR30-mediated ERK1/2 phosphorylation occurs via a G ␤␥ -dependent pathway (9). Therefore, it was necessary to first rule out if the mutations themselves affected GPER/GPR30 preassociation with G ␤␥ . HEK293 cells were transfected with wildtype or mutant GPER/GPR30 plasmids. After 24 h, G ␤␥ was immunoprecipitated from the lysate of unstimulated cells, followed by immunoblotting for GPER/GPR30 (Fig. 7A, bottom  immunoblot) in the pull-down fractions. The same membrane fragments were subsequently stripped and reprobed for the G ␤␥ inputs (Fig. 7A, top immunoblot). The histogram (Fig. 7A) represents average relative GPER/GPR30-G ␤␥ preassociation in cells expressing wild-type and mutant GPER/GPR30. Having confirmed that the CaM binding-reducing mutations did not affect the preassociation between GPER/GPR30 and G ␤␥ , we next tested the effect of these mutations on GPER/GPR30-mediated ERK1/2 phosphorylation. Stimulation of mock-transfected HEK293 cells with GPER/GPR30 agonist G-1 stimulated ERK1/2 phosphorylation due to activation of endogenous GPER/GPR30 as we have observed previously (23). Overexpression of the wild-type receptor enhanced this effect by 4-fold, and CaM binding-reducing mutations in SMD2, -3, -4, or a combination thereof individually and additively reduced GPER/GPR30-mediated ERK1/2 phosphorylation (Fig. 7B). These data indicate that CaM binding to SMD2, -3, and -4 of GPER/GPR30 promotes the receptor's function. Taken together with data from Figs. 1-3, this indicates a feedforward loop at GPER/GPR30 that increases CaM expression in endothelial cells.

E 2 Enhances CaM-PMCA Interaction, yet the Stimulatory Effect on Ca 2ϩ Extrusion Is Masked by PMCA Phosphorylation-Having
identified that E 2 indeed promotes CaM binding to its own receptors, we decided to examine two additional CaM targets that previously have been shown to functionally affect each other via competition for CaM: PMCA and eNOS. PMCA represents a major Ca 2ϩ extrusion mechanism in many cell types. CaM binding increases the affinity of PMCA for Ca 2ϩ Ͼ20-fold (46,47). We previously showed that, due to limiting CaM, acute increases in CaM binding to eNOS, a high affinity and abundant CaM target in endothelial cells, are associated with reduced CaM-dependent activities of PMCA (17,18). We recently showed that GPER/GPR30 forms a hetero-oligomeric complex with PMCA4b, which inhibits Ca 2ϩ extrusion, and acute GPER/GPR30 activation further inhibits this process by promoting PMCA tyrosine phosphorylation (23). The data presented so far led to the possibility that increases in CaM availability and binding to its targets as a result of long term treatment with E 2 could reverse the inhibitory effect of acute GPER/GPR30 activation on PMCA activity due to potential increased PMCA-CaM interaction. To test this, primary PAECs were subjected to 5 nM E 2 or vehicle for 48 h, followed by lysis at the peak of Ca 2ϩ entry induced by 1 M thapsigargin. PMCA was then immunoprecipitated from the lysate, followed by immunoblotting for CaM and PMCA in the pull-down fraction. E 2 increased PMCA-CaM interaction ϳ2.7-fold (Fig. 8A). Simultaneous probing for ␤-actin and CaM in the same lysate (Fig. 8B) confirmed increased CaM expression. To test the functional impact of the E 2 effect of increasing PMCA-CaM interaction, we used a specific protocol to mea-   MAY 13, 2016 • VOLUME 291 • NUMBER 20

Estrogen, GPER/GPR30, and Endothelial Calmodulin Network
sure Ca 2ϩ extrusion via the PMCA (17,18,23,48). Fig. 8C provides a typical time course of this protocol (see "Experimental Procedures" for description). Primary PAECs were pre-treated for 48 h with various doses of E 2 , followed by measurement of PMCA activity using this protocol. Fig. 8D shows average Ca 2ϩ extrusion time courses in cells with similar Ca 2ϩ concentrations at the beginning of the extrusion phase. Fig. 8E shows average corresponding PMCA activities. E 2 apparently did not increase PMCA activity as expected but dose-dependently inhibited it. PMCA expression did not change by 48-h E 2 treatment (Fig. 8F). The reduced PMCA activity was at odds with a 2.7-fold increase in PMCA-CaM interaction in cells exposed to E 2 . Based on our recent findings that GPER/GPR30 forms a heterooligomeric complex with PMCA4b, which inhibits the pump's activity, and acute GPER/GPR30 activation by agonist G-1 further inhibits the pump by stimulating PMCA tyrosine phosphorylation (23), we considered the possibility that the effect of enhanced PMCA-CaM interaction was masked by E 2 -induced tyrosine phosphorylation of PMCA. To verify this possibility, PAECs were exposed to vehicle or 5 nM E 2 for 48 h, followed by treatment with thapsigargin for 2 min before immunoprecipitation of PMCA and subsequent immunoblotting of pull-down fractions for phosphotyrosine and CaM (Fig. 9A, middle and bottom immunoblots). The phosphotyrosine immunoblots were then stripped and reprobed for PMCA (Fig. 9A, top). E 2 clearly increased both PMCA-CaM interaction and PMCA phosphorylation. We further reasoned that if the stimulatory effect of enhanced PMCA-CaM interaction was masked by the inhibitory effect of tyrosine phosphorylation, prevention of the latter should reveal the former. PAECs were exposed to vehicle or 5 nM E 2 for 48 h, followed by 30-min treatment with 10 M PP2, an Src kinase inhibitor that effectively prevents PMCA tyrosine phosphorylation stimulated by thrombin or GPER/ GPR30 agonist G-1 (23,49). The increase in PMCA tyrosine phosphorylation in E 2 -treated PAECs was now no longer seen, whereas the increased PMCA-CaM interaction still was obvious (Fig. 9B). PMCA activity was next tested on PAECs treated with similar paradigms. Fig. 9C shows the average Ca 2ϩ extrusion time courses from these cells. E 2 inhibited PMCA activity (Fig. 9D), and as expected, inhibition of tyrosine phosphorylation now revealed an overshoot in PMCA activity in E 2 -treated cells (Fig. 9E). These data indicate that E 2 up-regulates CaM and enhances PMCA-CaM interaction, but the associated stimulatory effect is masked by E 2 -mediated tyrosine phosphorylation. The outcome of these combined effects is prolongation of cytoplasmic Ca 2ϩ signals (Fig. 9C). (Figs. 1 and 2) and prolonged Ca 2ϩ signals (Figs. 8D and 9C) together explained the increased Ca 2ϩ -CaM complex formation in E 2 -treated cells (Fig. 1F). These combined effects are potentially of pervasive impact on the network of CaM-binding proteins. We have now observed increases in CaM binding to ER␣, PMCA, and GPER/GPR30, which are CaM targets with nanomolar to micromolar affinities. Endothelial NOS binds CaM (50) with picomolar affinity (18) and is capable of binding up to 25% of total CaM in endothelial cells (17). To test the effect of E 2 on eNOS-CaM association, PAECs were treated with and without 5 nM E 2 for 48 h and exposed to 2 M ionomycin for 3 min prior to lysis for co-immunoprecipitation. E 2
Activity of eNOS is influenced by multisite phosphorylation, including Ser-116, -617, -635, and -1179; Thr-497; and Tyr-85 (bovine sequence) (51). Of all of these sites, only phosphorylation at Thr-497, Ser-617, and Ser-1179 affects CaM binding to eNOS. Thr-497 is located in the CaM-binding domain of eNOS, and its phosphorylation status directly affects CaM binding to eNOS; dephosphorylation at Thr-497 promotes CaM binding, whereas phosphorylation at this site prevents CaM-eNOS interaction (35). E 2 was shown to promote phosphorylation at Ser-1179 (52) before the identification of Ser-617 and Ser-635 as phosphorylation sites (53). We previously demonstrated that biphosphorylation at Ser-617 and Ser-1179 substantially increases the Ca 2ϩ sensitivity for the eNOS-CaM interaction, leading to synthase activation at much lower Ca 2ϩ levels (54,55). We thus tested if E 2 promotes phosphorylation at these sites. Primary PAECs were exposed to vehicle or 5 nM E 2 for 48 h, followed by immunoblotting for total eNOS and phosphorylation at Thr-497, Ser-617, and Ser-1179. E 2 did not affect phosphorylation at Thr-497 but clearly promoted phosphorylation at both Ser-617 and Ser-1179 (Fig. 10C).
Proportional Contributions of the E 2 Effect on Ca 2ϩ , CaM, and eNOS Biphosphorylation at Ser-617 and Ser-1179 to eNOS Activity-Given the multifaceted effects of E 2 to prolong Ca 2ϩ signals, enhance eNOS-CaM interaction, and promote phos- phorylation at Ser-617 and Ser-1179, it was necessary to quantitate the contribution of each factor in the overall increase in eNOS activity. It would be challenging to do this based solely on cellular data due to the complexity of the system; however, incell and in vitro data are now available to do this. We have shown previously that eNOS-CaM interaction and activation follows a sequential model (55), where E is eNOS and E* represents CaM-activated eNOS. The initial step in this model represents CaM binding, determined by the equation, where K 1 K 2 is the product of the Ca 2ϩ binding constants for one of the EF-hand pairs in CaM (55). The second step represents CaM-dependent activation of eNOS, which requires Ca 2ϩ binding to both EF-hand pairs. The fractional eNOS activation is described by the equation, where K 3 K 4 is the product of the Ca 2ϩ binding constants of the remaining pair of Ca 2ϩ binding sites on CaM (55). This model has also been applied to eNOS-CaM binding and activation data to determine K 1 K 2 and K 3 K 4 values for wild-type eNOS and eNOS biphosphorylated at Ser-617 and Ser-1179 (54). Because the experiments so far in the present study have demonstrated the effects of E 2 to prolong cytoplasmic Ca 2ϩ signals, increase eNOS-CaM interaction, and promote phosphorylation at Ser-617 and Ser-1179, we now have sufficient data to utilize this model to delineate the proportional contribution of each effect to the overall increase in total eNOS activity in endothelial cells. Fig. 11A shows the average time courses of Ca 2ϩ extrusion measured in primary PAECs exposed to vehicle or 5 nM E 2 for 48 h. We now know that E 2 enhances eNOS-CaM interaction and phosphorylation at Ser-617 and Ser-1179, both by a factor of ϳ2 (Fig. 10, B and C). In addition, eNOS-CaM binding linearly correlates with eNOS activity (50,56). Applying to Equation 8 intracellular Ca 2ϩ values from Fig. 11A and previously determined values of K 1 K 2 and K 3 K 4 for wild-type eNOS (54) and adding a factor of 2 for the effect of increased eNOS-CaM binding on eNOS activity following E 2 treatment (Fig. 10B), we were able to calculate specific eNOS point activity corresponding to each Ca 2ϩ value in Fig. 11A. This process enabled us to calculate the portion of enhancement in eNOS activity due to the effect of E 2 to prolong Ca 2ϩ signals alone (Figs. 9C and 11A) or in combination with the increased CaM expression (Fig. 10B). Similarly, applying to Equation 8 values of K 1 K 2 and K 3 K 4 for eNOS biphosphorylated at Ser-617 and Ser-1179 (54) and intracellular Ca 2ϩ data from Fig. 11A and adding a factor of 2 for synthase activity from the effect of E 2 on eNOS biphosphorylation at Ser-617 and Ser-1179 (Fig. 10C), we were able to calculate E 2 -induced changes in eNOS point activity as a result of phosphorylation alone, combined effects on Ca 2ϩ and phosphorylation, and combined effects on Ca 2ϩ , CaM binding, and phosphorylation (Fig. 10C). The parameters used for these calculations are shown in Table 5. Fig. 11D shows the average relative eNOS point activity at the end of the Ca 2ϩ extrusion time courses, when intracellular concentration in untreated cells was only ϳ50 nM, and the individual (inset) or combined effects of E 2 on Ca 2ϩ , CaM, and eNOS biphosphorylation at Ser-617 and Ser-1179. The data show that the increases in Ca 2ϩ , Ca 2ϩ -CaM, and eNOS phosphorylation each contributed substantially to the total increase in eNOS activity in endothelial cells exposed to E 2 .

Discussion
The present study demonstrated a feedforward loop involving GPER/GPR30 that increases CaM expression, which, together with the inhibition of Ca 2ϩ efflux via the PMCA, results in prolonged Ca 2ϩ -CaM signals produced in cells for interactions with different targets and modulation of their activities (Fig. 12). The proteins examined here represent disparate categories of CaM targets in endothelial cells. Endothelial NOS is an abundant and high affinity CaM target, able to bind up to 25% of total CaM in endothelial cells (17) with picomolar affinity (18). PMCA and ER␣ bind CaM with low nanomolar affinity, whereas the four CaM-binding sequences in GPER/GPR30, when in the isolated biosensor format, bind CaM with submicromolar to low micromolar affinities (27). Our current data showed that E 2 enhances CaM binding to all four proteins. These results once again confirm the limited expression of CaM for its numerous targets and demonstrate that E 2 can improve endothelial functions through these pervasive effects.
An interesting finding here is that GPER/GPR30 mediates the E 2 effect of up-regulating CaM, which in turn is important for the receptor's functions. This role of GPER/GPR30 is supported by the specific agonist effect, gene silencing data, and studies in cells that only express this receptor. The finding that ICI182,780, a dual ER␣/ER␤ antagonist and agonist for GPER/ GPR30, generates the same effect as E 2 on CaM expression further strengthens this conclusion. The effect of GPER/GPR30 activation on expression of CaM as a Ca 2ϩ -binding protein appears to be specific, because G-1 treatment did not affect expression levels of either annexin V or calreticulin in primary endothelial cells. Our data are quite consistent with findings that E 2 administration to ovariectomized mice up-regulates CaM mRNA in the uterus independently of ER␣ and ER␤ (57). Mechanistically, our data indicate that GPER/GPR30 up-regulates CaM expression at the transcriptional level via transactivation of EGFR, with downstream activation of the MAPK cascade. This is consistent with previous studies documenting that GPER/GPR30 activation causes EGFR transactivation and MAPK activation (9, 37) as well as the role of this cascade in up-regulating a number of genes, including c-fos, cyclin A and D1 (6,58), connective tissue growth factor (59), fatty acid synthase (8), and transcription factor SF-1 (60). Our results also indicate that CaM binding to SMD2, -3, and -4 of GPER/GPR30 individually and synergistically promotes the receptor's function (Fig. 7B). Among the domains, CaM interaction with amino acids 330 -351 on SMD4 apparently has the largest impact on GPER/GPR30 function. Previous data have suggested a role for CaM in regulating G ␤␥ interactions with the metabotropic glutamate receptors (61,62) and the angiotensin II type 1A receptor (63). It was unclear, however, whether CaM binding prevents G ␤␥ association or facilitates G ␤␥ dissociation. Because GPER/GPR30-mediated phosphorylation of ERK1/2 is G ␤␥ -dependent (9), the findings that CaM binding-reducing mutations did not affect GPER/GPR30-G ␤␥ preassociation but decreased GPER/GPR30-mediated phosphorylation of ERK1/2 confirmed that the latter was due to reduction in CaM binding to these domains. Speculatively, these data are also consistent with the possibility that CaM binding to these domains facilitates dissociation of G ␤␥ , therefore enhancing GPER/GPR30mediated signaling. However, this possibility requires further investigation.
From a functional standpoint, E 2 , via activation of GPER/ GPR30, affects the CaM network in vascular endothelial cells by both an increase in CaM expression and a decrease in Ca 2ϩ efflux, which prolongs cytoplasmic Ca 2ϩ -CaM signals. It is now clear that estrogen exerts multifaceted effects on Ca 2ϩ efflux via the PMCA. We recently showed that GPER/GPR30 forms a constitutive hetero-oligomeric complex with PMCA4b, the predominant endothelial PMCA isoform (64), via their C-terminal PDZ-binding motifs, an interaction that inhibits PMCA activity while promoting GPER/GPR30 activity; and GPER/ GPR30 activation adds to the constitutive inhibition by promoting PMCA tyrosine phosphorylation (23). Data in the present study are consistent with these results and earlier reports on the effect of tyrosine phosphorylation on PMCA activity (49) and further demonstrate that long term E 2 treatment enhances PMCA-CaM interaction, but this is masked by the inhibitory effect of tyrosine phosphorylation. The ultimate outcome is prolongation of Ca 2ϩ signals.
Our data indicate that eNOS, despite its picomolar K d for CaM (18), does not have sufficient CaM to saturate its binding sites in primary cultured, untreated endothelial cells and that E 2  treatment up-regulates CaM for further eNOS activation. Using a combination of in-cell and in vitro data, we were able to dissect the contribution of the many effects that E 2 exerts to increase eNOS activity. At the end of the Ca 2ϩ extrusion time courses, when cytoplasmic Ca 2ϩ in control PAECs is only ϳ50 nM, E 2 inhibits Ca 2ϩ extrusion to keep Ca 2ϩ at ϳ150 nM (Fig.  11A). This effect alone increases eNOS point activity by ϳ80fold and, when combined with the 2-fold increase in CaM binding, increases eNOS activity by ϳ160-fold (Fig. 11D, inset). Assuming no change in Ca 2ϩ or CaM binding, the effect of dual phosphorylation at Ser-617 and Ser-1179 alone increases point synthase activity 175-fold (Fig. 11D, inset). It is noteworthy that the ultimate effect of combined phosphorylation at Ser-617 and Ser-1179 is due per se to an increase in the Ca 2ϩ sensitivity for eNOS-CaM interaction and eNOS activation (54,55). Due to these reciprocal relationships, the effects of increases in Ca 2ϩ and CaM added to those of eNOS phosphorylation resulted in the drastic 3,163-and 6,327-fold increases in point eNOS activity in cells exposed to E 2 at the end of the Ca 2ϩ extrusion time course (Fig. 11D). The increases in Ca 2ϩ and CaM levels thus contribute substantially to the promotion of eNOS activity by E 2 across the time course of Ca 2ϩ signals in cells and should contribute significantly to the control of vascular tone. These data delineate the various mechanisms whereby estrogen improves vascular tone via enhancement of eNOS activity. The combined effects of increased CaM expression and prolonged cytoplasmic Ca 2ϩ signals are predicted to have a pervasive impact on the activities of CaM-dependent proteins. The examples of the network members presented here indicate that the up-regulation in CaM availability, although potentially of stimulatory impact on most targets, will have to be considered in conjunction with other factors that E 2 also modulates, such as various phosphorylation events. Thus, for GPER/GPR30 and eNOS, E 2 increases CaM binding and activities, whereas for PMCA, the effect of enhanced CaM binding is masked by inhibitory tyrosine phosphorylation. Interestingly, in this specific case, this inhibitory effect turns out to have a stimulatory effect on other targets via prolongation of cytoplasmic Ca 2ϩ signals.