Gelsolin Modulates Phospholipase C Activity In Vivo through Phospholipid Binding

NIH3T3 cell lines stably overexpressing human gelsolin were gifts of Drs. D.J. Kwiatkowski and C. Cunningham (Brigham and Women's Hospital, Boston, MA). They were obtained by transfection of a human cytoplasmic gelsolin expression vector and clonally selected with neomycin (9). Early passage cells that had been directly quantitated for gelsolin content were used, and their gelsolin expression was reconfirmed in our laboratory (see below). CapG overexpression cells were obtained as described in reference 31. Control (Ctrl) cells were transfected with the vector without a cDNA insert. Cells were maintained in DME with 10% FCS.

For inositol 1,4,5-trisphosphate (IP₃) and intracellular Ca²⁺ measurements, cell monolayers were starved in DME/0.2% serum for 24 h. They were switched to a completely defined 1:1 DME/F12 medium supplemented with 20 mM Hepes, pH 7.4, 0.5 mg/ml BSA, 1 μg/ml insulin, and 1 μg/ml transferrin (Q-medium) for an additional 18 h before agonist application. In some cases, cells were treated with 100 ng/ml pertussis toxin (List Biological Lab., Inc., Campbell, CA) overnight. Bradykinin and PDGF (AB chains) were from Sigma Chemical Co. (St. Louis, MO) and Upstate Biotechnology, Inc. (Lake Placid, NY), respectively.

Cells were trypsinized and resuspended in Chelex 100–treated buffer containing 100 mM KCl, 20 mM NaCl, 20 mM Hepes, pH 7.4. 5 × 10⁶ cells were permeabilized in 0.4–0.6 U/ml streptolysin O (SLO) (Murex Diagnostics Limited, Dartford, England) with 1 mg/ml creatinine phosphate, 0.1 mg/ml creatinine phosphokinase, 5 mM MgCl₂, and 3 mM ATP (21). Cells were usually permeabilized for 4 min at 37°C before agonist application.

Gelsolin NH₂-terminal half domains containing zero to three PIP₂-binding sites were expressed in Escherichia coli and isolated as described (35, 37). The final products were >99% pure, based on Coomassie blue staining of protein bands in sodium dodecyl sulfate polyacrylamide gels. Protein concentrations were determined by the method of Bradford (5). P2, a 22-mer synthetic peptide encompassing gelsolin residues 150–169 (CKHVVPNEVVVQRLFQVKGRR, amino-terminal cysteine added) (16), was synthesized by solid phase methods. CapG and thymosin β4 were expressed in E. coli and purified as described previously (35, 38).

Serum-deprived cells were washed and incubated with a buffer containing Fura 2 (2.5 μM), 1 mg/ml BSA, and 2 mg/ml glucose for 1 h at room temperature (31). Cells were trypsinized briefly and resuspended in a buffer containing BSA and glucose and warmed to 37°C. Fluorescence measurements were performed with 2 × 10⁶ cells in 1.5 ml with slow stirring at 37°C, at excitation and emission wavelengths of 340 and 500 nm, respectively. [Ca²⁺] was calibrated by adding 25 μg/ml digitonin and 0.25 mM MnCl₂ to obtain F_max and F_min, respectively. A K_dCa²⁺ of 224 nM was used for calculation.

Cell monolayers starved in Q-medium were stimulated with 5 μM bradykinin for 7 or 15 s, and the reaction was stopped by adding an equal volume of a 5% perchloric acid solution containing 10 mM EDTA and 1 μM ATP. The soluble extract was neutralized with 1:1 freon/octylamine, and IP₃ content was assayed by competition with exogenous [³H]IP₃ to bind calf cerebellum microsomes (31). IP₃ standards were extracted and assayed under identical conditions. The IP₃ content of permeabilized cells was determined with cells in suspension under conditions identical to those used for Ca²⁺ measurements.

Cells were lysed in cold RIPA buffer containing a protease inhibitor cocktail as described previously (31). Lysate protein concentrations were determined by the micro-BCA method (Pierce Chemical Co., Rockford, IL). Proteins were electrophoresed on 5–10% polyacrylamide, discontinuous pH slab gels in the presence of SDS and transferred to nitrocellulose for Western blotting. Blots were probed with rabbit anti–mouse gelsolin, which recognizes the endogenous gelsolin (31), and/or a monoclonal anti– human gelsolin (clone 2C4), which recognizes human but not mouse gelsolin (6). Endogenous mouse and overexpressed human gelsolin were quantitated by comparing the intensity of gelsolin bands in cell lysates with that of purified mouse and human plasma gelsolins, respectively, as described in reference 9. In leak-out experiments, ctrl cells were permeabilized with streptolysin O at 37°C and centrifuged at 1,500 rpm in an Eppendorf microcentrifuge for 1 min at room temperature. Samples were blotted with rabbit anti–mouse gelsolin, rabbit anti-PLC_β3 (gift of P. Sternweis, University of Texas Southwestern Medical Center, Dallas, TX) (11), and monoclonal antiactin (clone C4; Boehringer Mannheim Corp., Indianapolis, IN).

Recombinant PLC_β1 expressed in SF9 cells and purified to homogeneity was a gift of Dr. E. Ross (University of Texas Southwestern Medical Center) (4). Vesicles containing 100 μM phosphatidylethanolamine, 10 μM phosphatidylserine, 10 μM PIP₂, and 0.028 μM [³H] PIP₂ were made by sonicating in the PLC buffer (25 mM Hepes, 80 mM KCl, 3 mM EGTA, 0.5 mM DTT, pH 7.0). Vesicles were preincubated with gelsolin domains in the presence of 11 μM free [Ca²⁺] at 4°C for 10 min, and enzyme reaction was initiated by adding 4 ng (per 60 μl) PLC_β1 and warming to 37°C. Reactions were terminated after 1–4 min by adding TCA and 3 mg/ml BSA. Precipitated proteins were removed by centrifugation, and radioactive IP₃ in the TCA-soluble supernatant was determined by scintillation counting. Triplicate samples were assayed per condition. PIP₂ hydrolysis was linear for 15 min under the conditions used. In some experiments, PIP₂, ranging from 5–200 μM, was presented as micelles.

Phospholipase C_γ1, purified from bovine brain (26), was a gift of Dr. S.G. Rhee (National Institute of Health, Bethesda, MD). A preparation enriched in EGF receptor tyrosine kinase was prepared from A431 plasma membrane by Triton X-100 extraction and passaged through a wheat germ agglutinin column with N-acetyl-d glucosamine (32). The eluted receptor preparation was dialyzed and used at 3 μg/ml to phosphorylate 0.6 μg/ml PLC_γ1 , in the presence of 20 μM ATP and 3 μM EGF at 4°C for 1 h (22). Its activity was assessed by incubating PLC_γ1 with [³²P]ATP to detect EGF- dependent ³²P incorporation. PLC_γ1 activity was assayed by using 6–12.5 ng PLC_γ1/60 μl reaction mixture, and with vesicles containing 10 μM PIP₂/0.02 μM [³H]PIP₂/100 μM PE. The buffer conditions were as described above. The reaction was linear for at least 15 min at 37°C. Unphosphorylated PLC_γ1 control was incubated with the kinase preparation in the absence of ATP and EGF.

We showed previously that CapG overexpression increased Ca²⁺ and IP₃ response to PDGF (31). These results suggest that CapG enhanced PLC_γ in vivo, although gelsolin (1) and CapG (see below) are predominantly inhibitory in vitro. We therefore examined the effect of overexpression on PLC_β, which is activated through the bradykinin/heterotrimeric G protein pathway. 5 μM bradykinin stimulated ctrl cells to increase cytosolic Ca²⁺ in a sharp peak (Fig. 1,A). CapG clones had a reduced Ca²⁺ response (Fig. 1,A and Table I). Likewise, the IP₃ response was also attenuated. Bradykinin-stimulated ctrl cells generated IP₃ to a maximal level at 7 s (Fig. 1,B). The CapG clones generated less IP₃ at 7 and 15 s. Ca²⁺ signal and IP₃ response were inhibited to a similar extent, suggesting that CapG inhibited PLC_β and Ca²⁺ signaling by a similar mechanism (see Fig. 3 B).

To explore the basis for the opposite effects of CapG overexpression on PDGF- vs. bradykinin-activated PLC, we felt that studying cells with a wider range of overexpression and comparing cells overexpressing related proteins would be important. In spite of repeated attempts, we were not able to isolate stable CapG clones with higher overexpression. We therefore directed our effort to gelsolin-overexpressing clones, which have a wider range of overexpression (9).

Western blotting with human gelsolin–specific antibody, which recognizes the transfected human gelsolin but not endogenous mouse gelsolin, confirmed that the clones had a range of gelsolin overexpression (Fig. 2, bottom), as described previously (9). C2 was not positive, indicating that it had lost the expression vector, and was not studied further. Blotting with anti–mouse gelsolin, which cross-reacts poorly with human gelsolin, showed that the endogenous gelsolin level in overexpressing cells was comparable to that of the ctrl cells (Fig. 2, top). Therefore, there was no compensatory change in endogenous gelsolin expression. Total gelsolin level, determined by quantitative Western blotting using purified human and mouse gelsolin standards, confirmed that the levels of expression were similar to that shown previously (9) (Table I).

Gelsolin overexpression modified the Ca²⁺ response to bradykinin. C3 that has 1.3 times as much gelsolin as ctrl was slightly more responsive to bradykinin and PDGF (Table I). In contrast, C6, C1, C4, and C5 that have higher overexpression had a striking decrease in bradykinin- induced Ca²⁺ signal (Fig. 3, Table I). The PDGF response was also reduced, but to a lesser extent (Table I). Since inhibition was observed with multiple independently isolated clones, the phenotype was attributed to gelsolin overexpression and was not an artifact of clonal selection. The stimulation we observed with the lowest overexpressing clone, C3, suggests that gelsolin may also increase signaling. However, because only one such clone was available, ruling out clonal variations unrelated to gelsolin was not possible.

The effect of gelsolin on PLC_β activity per se was confirmed by quantitating IP₃. Ctrl and overexpressing clones had similar basal IP₃ levels (∼10 pmol/mg cell protein). Gelsolin clones had reduced IP₃ response (Fig. 3 B), and the extent of inhibition correlated with that of the Ca²⁺ response. Overall, our results are consistent with the hypothesis that the gelsolin- and CapG-induced change in Ca²⁺ signal was due to PLC_β inhibition.

Since the bradykinin response was strikingly inhibited, we will focus on this aspect in the rest of the paper. Bradykinin stimulation of Ca²⁺ release is a multistep process that involves bradykinin binding to receptors, activation of heterotrimeric G proteins, PLC_β hydrolysis of PIP₂, and IP₃-induced release of Ca²⁺ from intracellular stores. Ctrl cells responded to as low as 0.1 μM bradykinin, and Ca²⁺ release was maximally stimulated at 1 μM (data not shown). C5 had a very small Ca²⁺ peak, even at 50 μM bradykinin, suggesting that the problem was unlikely to be due to decreased receptor affinity. To identify the step compromised by gelsolin overexpression, NaF was used to activate G proteins directly, bypassing receptor involvement. NaF induced a slow Ca²⁺ rise (Fig. 3,A), which had a smaller amplitude than the bradykinin peak (320 ± 50 nM). Gelsolin overexpression reduced the NaF response, suggesting that it acted at or downstream of the heterotrimeric G protein activation step (Fig. 3,A). However, gelsolin did not inhibit the NaF response to as great an extent as the bradykinin response (Fig. 3 B). The Ca²⁺ peak for C4 was reduced to 1.8 and 47.8% of ctrl after bradykinin and NaF stimulation, respectively. The difference may be due to activation of different PLCs. NaF activates all heterotrimeric G proteins, while bradykinin activates the alpha subunit (Gα_q/11) (11, 29). This was confirmed with pertussis toxin that inhibits G_i but not Gα_q/11. In ctrl cells, pertussis toxin reduced the NaF response by 45.7% but had minimal effect on the bradykinin response (data not shown). Our result suggested that Gα_q/11-activated PLCs were more sensitive to gelsolin inhibition than Gα_i-activated enzymes.

To determine if the change in PLC_β activity is directly attributable to gelsolin, a gelsolin NH₂-terminal half polypeptide (45 kD, spanning gelsolin residues 1–406 [GS1-406]) that contains at least two PIP₂-binding sites (17, 35) was added to SLO-permeabilized ctrl cells. Under the permeabilizing conditions used, cells retained agonist responsiveness, as demonstrated previously for another type of cell (21). Fig. 4 showed the result of a representative experiment. The mean IP₃ response in the presence of 4 and 32 μM GS1-406 were 46.5 ± 4.0% (n = 8) and 6.4 ± 1.1% (n = 3) of control, respectively. Thus, exogenous gelsolin added to permeabilized cells mimicked the inhibitory effects of gelsolin overexpression.

Exogenous CapG also inhibited IP₃ rise, while thymosin β4, an actin-binding protein that does not interact with PIP₂ (38), had no effect at 10 μM. (IP₃ response was 91.7 and 101.8% of control after bradykinin stimulation in two experiments.) These results suggested that inhibition of IP₃ generation in cells was linked to an increase in PIP₂-binding proteins.

Our data showed that gelsolin, either overexpressed in intact cells or added to semiintact cells, reduced bradykinin responsiveness. Removal of gelsolin should restore the bradykinin response. Cells were permeabilized for progressively longer intervals before bradykinin stimulation to allow gelsolin and other cytosolic proteins to leak out. After 3 min, gelsolin and actin were detected in the medium (Fig. 5 A). More proteins leaked out with longer treatment. In contrast, PLC_β3, the predominant PLC in NIH3T3 cells, did not leak out to a significant extent, as would be consistent with its stable plasma membrane association. There was no difference in the leakage of PLC_β3 between ctrl and C5.

The effect of cytosol leakage on IP₃ generation was determined by stimulating cells permeabilized for different intervals with bradykinin (Fig. 5,B). Both intact and permeabilized ctrl cells increased IP₃ after bradykinin stimulation. The semiintact cell's response was more robust, suggesting that soluble inhibitors had leaked out or that other clamping mechanisms were disrupted. As shown above, intact C5 had a much lower IP₃ response than ctrl cells. At 2 min after SLO treatment, C5 was still inhibited relative to ctrl. Between 2 to 5 min, C5 response increased to a level comparable to that of ctrl. With longer permeabilization, both ctrl and C5 were less active, probably due to loss of membrane integrity and depletion of soluble cofactors such as GTP. The convergence of the IP₃ response between ctrl and C5 was not due to differential leakout of PLC_β3 (Fig. 5,A). Recovery of bradykinin-stimulated PLC activity indicates that reduced bradykinin responsiveness in intact overexpressing cells is unlikely to be due to bradykinin receptor defects. A similar conclusion was reached using NaF to stimulate G proteins directly (Fig. 3 A). Therefore, we conclude that gelsolin depresses PLC activity in intact C5, and reducing gelsolin concentration relieves the inhibition.

It has been reported that gelsolin inhibits several PLC isozymes in vitro (1). In light of our in vivo results, we reexamined this issue to determine if gelsolin and CapG inhibit PLC_β and PLC_γ differentially, and whether inhibition is due to PIP₂ binding. Gelsolin GS1-406 that contains two identified PIP₂-binding sites increased Ca²⁺-activated PLC_β at 1 μM and inhibited at higher concentrations (Fig. 6,A). A biphasic effect has been reported previously (1), although the stimulatory phase was ignored. CapG stimulated and inhibited with a comparable dose response, consistent with its similar affinity as gelsolin for PIP₂ (18). Gelsolin also had a biphasic effect on PLC_γ phosphorylated with EGF receptor kinase. In contrast, gelsolin did not increase the activity of nonphosphorylated PLC_γ (Fig. 6 B) and inhibited it more strongly. The differential effects on phosphorylated and unphosphorylated PLC_γ, including activation of the phosphorylated enzyme at low concentration, were similar to that reported previously for profilin (10). A direct comparison between gelsolin and profilin showed that gelsolin was a stronger inhibitor (data not shown), consistent with their different affinity for PIP₂ (18, 20). Although gelsolin was reported to bind PLC_γ (1), we were not able to detect such an interaction either with purified proteins or in cell lysates (data not shown). Therefore, gelsolin is not likely to act by binding PLC directly.

Gelsolin inhibited Ca²⁺-activated PLC_β more than phosphorylated PLC_γ. 2 μM gelsolin stimulated phosphorylated PLC_γ but was located on the inhibitory arm of the PLC_β dose-response curve (Fig. 6, A and B). Stimulation of phosphorylated PLC_γ may contribute to the increased PDGF response in clones overexpressing low levels of CapG and in the single low-level gelsolin-overexpressing clone (C3). The higher sensitivity of PLC_β to inhibition may explain why PLC_β and PLC_γ were affected in opposite directions in CapG clones and why the gelsolin clones showed more inhibition of PLC_β than PLC_γ. Thus, the differential inhibitory effects of gelsolin on PLC_β and PLC_γ in vitro recapitulated the in vivo effects of overexpression.

The inhibitory effect of GS1-406 was overcome by increasing PIP₂ concentration. As expected, PLC_β generated more IP₃ as the substrate concentration increased. 5 μM GS1-406 completely inhibited PLC_β activity at 5 and 10 μM PIP₂ but was progressively less effective at higher PIP₂ concentration (Fig. 6 C). No inhibition was observed at PIP₂ concentrations above 120 μM. The simplest interpretation was that gelsolin inhibited PLC_β by substrate competition, as suggested for PLC_γ in an earlier study (1).

Experiments using gelsolin domains with and without PIP₂-binding sites confirmed that gelsolin altered PLC activity through PIP₂ binding. GS1-406, encompassing gelsolin segments 1–3, contains at least two (and possibly three) PIP₂-binding sites (Fig. 7,A). The S1 and S2 PIP₂-binding sites have been mapped precisely by deletion analyses (37) and peptide analog studies (16, 37). The existence of a third site is inferred from previous experiments after deleting the S2 site and is predicted to be in S3 (37). GS1-406, which binds PIP₂ with an estimated K_d of 2.9 μM and a stoichiometry of ∼3 (18), had the most activating and inhibitory effects (Fig. 7 B). GS150-406 (S2-3), which binds with a similar K_d but with a stoichiometry of ∼2, was slightly less effective. GS150-267 that contains a single PIP₂-binding site was markedly less effective, while GS171-267 with no PIP₂-binding site had no effect. These results showed that gelsolin altered PLC activity through PIP₂ binding, and maximal effect was observed with the participation of two to three PIP₂-binding sites. Binding to multiple PIP₂ may form a more stable complex that is less likely to be displaced by PLC, and the larger polypeptides are likely to be more effective in sterically hindering PLC access to its substrate.

P2, the PIP₂-binding peptide derived from gelsolin S2 (16), had no effect at low concentration and stimulated PLC_β by more than threefold at 53 μM (Fig.7 C). We were not able to use more P2 to determine if higher concentrations were inhibitory. P2 may stimulate by organizing PIP₂ in a more favorable conformation for PLC hydrolysis. It may be more stimulatory than the parent S2 domain because its positive effects are not neutralized by steric hindrance to block PLC access. However, in contrast to the marked stimulation in vitro, P2 had minimal effect on PLC_β activity in permeabilized cells. Semiintact cells treated with 100 μM P2 showed 72.5 ± 6.7% (n = 9) of the bradykinin-induced IP₃ response of untreated cells. No stimulation was observed between 5–100 μM (data not shown). We cannot explain why P2 had divergent effects in vitro and in vivo.

In this paper, we examined the ability of gelsolin and CapG to compete with PLC for PIP₂ in live cells, in semiintact cells, and in vitro. The various approaches help to define the mechanism by which gelsolin and CapG modulate PLC activity in two independent signaling pathways. Our results suggest that these cytoskeletal proteins may participate in signaling by controlling the availability of PIP₂ substrates.

The CapG and gelsolin overexpressing cell lines were isolated independently by two different laboratories, using the same parent NIH3T3 cell line and expression vector. Their remarkably similar effects on PLC_β strongly support the conclusion that gelsolin and CapG bind PIP₂ in vivo to inhibit PLC_β. There are, however, differences with respect to PLC_γ. At an apparently similar level of overexpression, CapG clone gK8 and gelsolin clone C6 had increased and decreased response to PDGF, respectively. The divergent effects of gelsolin and CapG may be due to multiple factors. CapG has one PIP₂-binding site, while gelsolin has two (37). CapG is half the size of gelsolin (36). Thus, CapG may be more readily displaced by activated PLC_γ and may not sterically block PLC_γ access to PIP₂ as effectively. CapG is phosphorylated (23), and phospho-CapG may have a different affinity for PIP₂ (Lu, P., A. Hsu, D. Wong, H. Yan, H.L. Yin, and C. Chen, manuscript submitted for publication). Another possibility may be different intracellular localization. CapG is a nuclear and cytosolic protein, while gelsolin is excluded from the nucleus (24). There is emerging evidence for phosphoinositide signaling pathways in nuclei (Lu, P., A. Hsu, D. Wong, H. Yan, H.L. Yin, and C. Chen, manuscript submitted for publication), and CapG may therefore act in the nucleus.

Another feature that emerges from these studies is that there is a difference in the sensitivity of PLC_β and PLC_γ towards gelsolin and CapG. This may be explained in several ways. Phosphorylated PLC_γ may have a higher affinity for PIP₂ than PLC_β and is therefore better able to compete with gelsolin for PIP₂. Differential sensitivity may account for the paradoxical finding that CapG-overexpressing cells have increased PLC_γ (31) but reduced PLC_β activity. Another possibility is that PLC isozymes may use different pools of PIP₂, and gelsolin/CapG have differential access to these pools. PIP₂ pools with distinct turnover characteristics have been identified (25), although their relation to the different PLC isozymes, gelsolin and CapG, has not been examined.

Inhibition of PLC is most likely due to competition of gelsolin and CapG with PLC for PIP₂. Steric hindrance may also be a contributing factor. The mechanism by which PLC is enhanced by a small increase in CapG or gelsolin is not known. Since only gelsolin domains with PIP₂-binding sites stimulate, this effect depends on PIP₂ binding. The simplest hypothesis is that gelsolin and CapG improve the presentation of PIP₂ to PLC by altering the phospholipid conformation. Among the gelsolin domains tested, P2, the small synthetic peptide containing a single PIP₂-binding site, is most stimulatory in vitro. However, it does not have any effect on PLC in semiintact cells. This peptide has been used by others to examine the relation between gelsolin and PIP₂ in cells (8, 13). P2 added to semiintact platelets blocks Rac1-induced uncapping of actin filament + ends, and this effect is overcome by adding PIP₂ (13). Overexpression of P2 in a permanently transfected cell line reduces gelsolin association with the plasma membrane (8). Both experiments indicate that P2 competes with gelsolin for PIP₂, although effects on other PIP₂ -requiring processes have not been ruled out. Our result suggests that PLC was not affected by increasing P2 levels in cells. Additional experiments will be required to determine if other parameters are altered by P2.

In conclusion, we show that gelsolin and CapG interact with PIP₂ in vivo, and a modest increase in concentration has profound effects on two PIP₂ signaling pathways. Overexpression is used as a tool to probe intracellular function. Our overexpression results have physiological relevance. While total gelsolin and CapG content do not change per se, cytosolic concentration and membrane binding are expected to fluctuate during signaling. Gelsolin and CapG bind to and are released from filament ends and the membrane PIP₂ content changes due to hydrolysis and resynthesis. Ca²⁺ increases gelsolin and CapG affinity for PIP₂ (18), enhancing their ability to compete with other PIP₂-binding proteins at a time when membrane PIP₂ content decreases. CapG and gelsolin can provide positive and negative inputs on PLC signaling pathways, and these pathways are modulated selectively. The data presented here indicate that there is significant cross talk between components of the transmembrane signal machinery and actin cytoskeleton at multiple levels, including that of the generation of important second messengers.

We are grateful to David Kwiatkowksi and Casey Cunningham for the gelsolin overexpressing cell lines, Sue Goo Rhee for PLC_γ1 and Paul Sternweis for anti-PLC_β3. We thank Elliot Ross and Gloria Biddlecome (University of Texas Southwestern Medical Center) for initiating the PLC_β1 assays and providing for recombinant PLC_β1.

This work was supported by the National Institutes of Health grant GM51112.

C

gelsolin-overexpressing lines

ctrl

control clones

IP₃

inositol 1,4,5-trisphosphate

PIP₂

phosphatidylinositol 4,5-bisphosphate

PLC

phospholipase C

SLO

streptolysin O