β-Adrenergic receptors suppress Rap1B prenylation and promote the metastatic phenotype in breast cancer cells

Abstract

A greater understanding of the molecular basis of breast cancer metastasis will lead to identification of novel therapeutic targets and better treatments. Rap1B is a small GTPase that suppresses the metastasis of breast cancer cells by increasing cell-cell adhesion. In breast cancer, a decrease in Rap1B prenylation and subsequent loss of Rap1B at the plasma membrane decreases cell-cell adhesion and increases cell scattering, which promotes the metastatic phenotype. Protein kinase A (PKA) was recently found to phosphorylate Rap1B and inhibit its prenylation. PKA is activated by G protein-coupled receptors (GPCR) that stimulate Gα_s. In this study, we investigated whether the general Gα_s activator, cholera toxin, and agonists of the β-adrenergic receptor (βAR), which is a Gα_s-coupled GPCR, promote Rap1B phosphorylation and inhibit its prenylation. We show here that cholera toxin and βAR activation phosphorylate Rap1B and inhibit its prenylation and membrane localization, reducing cell-cell adhesion and promoting cell scattering. Furthermore, we report that breast cancer cell migration is decreased by the FDA-approved β-blocker, propranolol. Pharmacological targeting of GPCRs, especially those such as the βAR that are regulated by FDA-approved drugs, to increase cell adhesion and decrease cell scattering could provide a promising therapeutic approach to reduce breast cancer metastasis.

Abbreviations

Adenosine 2B Receptor

(A₂BR)

β-adrenergic receptor

βAR)

Bay 60-6583

Bay)

Cholera toxin

Ctx)

half time

T_1/2)

Isoproterenol

Iso)

G protein-coupled receptors

GPCR)

Phosphate buffered saline

PBS)

Protein kinase A

PKA)

Propranolol

Prop).

Introduction

Metastasis remains one of the leading causes of death for breast cancer patients. In order for metastasis to occur, cells must lose their cell-cell adhesion and become more motile to travel into the blood stream and disseminate throughout the body. ¹ Rap1B is a small GTPase that promotes cell-cell adhesion,^2-4 and thus plays an important role in metastasis.

The cysteine of the C-terminal CAAX motif of Rap1B can be prenylated, which is the addition of an isoprenoid moiety that assists in membrane localization.⁵ We recently found that prenylation of newly synthesized Rap1B is inhibited by phosphorylation of serines 179 and 180 in the Rap1B polybasic region (PBR), which consists of basic amino acids adjacent to the CAAX motif.⁶ Inhibition of Rap1B prenylation significantly alters cell functions, because Rap1B must be prenylated to localize at the plasma membrane and promote cell-cell adhesion. Loss of Rap1B at the plasma membrane due to loss of prenylation can lead to increased cell scattering, which is one of the first steps in breast cancer metastasis.⁶ Therefore, understanding the signaling pathways that promote prenylation of Rap1B and thereby increase cell adhesion could lead to better therapeutics to decrease metastasis.

Our laboratory recently reported that increased adenosine in the hypoxic tumor environment might increase metastasis through activation of the adenosine A₂B receptor (A₂BR), which is a Gα_s-coupled GPCR.⁶ Through activation of the A₂BR, adenylyl cyclase is activated, increasing cAMP levels, which activates protein kinase A (PKA). PKA phosphorylates serines 179 and 180 in the PBR of Rap1B, preventing newly synthesized Rap1B from being prenylated. Phosphorylated Rap1B stays in the non-prenylated state in the cytosol, unable to stabilize adherens junctions, leading to cell scattering and promoting the metastatic phenotype.⁶ Furthermore, A₂BR activation has been identified as a potential promoter of migration, invasion and metastatic formation in vivo,^6-12 indicating the A₂BR is a potential target for inhibiting breast cancer metastasis.

The adenosine receptor antagonist theophylline is already in clinical use for asthma, suggesting that theophylline could be repurposed to prevent cancer metastasis by inhibiting A₂BR activity. However, the multiple side effects induced by theophylline, which are potentially caused by its actions as a cAMP phosphodiesterase inhibitor, ¹³ make theophylline a less than ideal drug for repurposing for cancer treatment. Due to the limited pharmacological options available for targeting the A₂BR, there is a need to determine whether other GPCRs can be therapeutically targeted to regulate the participation of prenylated Rap1B in cancer.

Here, we examine whether this signaling pathway that inhibits Rap1B prenylation is specific to the A₂BR, or whether other GPCRs coupled to Gα_s can also inhibit prenylation of Rap1B, providing better targets for decreasing breast cancer metastasis. Perhaps the most interesting potential target is the β-adrenergic receptor (βAR), which has already been shown to increase metastatic potential in breast cancer cells.^14-19 Furthermore, β-blockers (inhibitors of β-adrenergic receptors) are a common treatment choice for hypertension and cardiac dysfunction. Notably, retrospective studies have shown that breast cancer patients taking a β-blocker for other purposes have decreased cancer reoccurrence and increased survival.^15,20-29 Clinical trials have been initiated to examine whether treatment of breast cancer patients with β-blockers decreases the chance of reoccurrence.^30,31 A greater understanding of the mechanism through which the inhibition of βAR decreases reoccurrence will lead to increased success in treating patients with aggressive tumors.

In this study, we use several drugs (or toxins) targeting Gα_s-coupled receptors. Cholera toxin (Ctx) is a general Gα_s protein activator.³² 5′-N-Ethylcarboxamidoadenosine (NECA) is a general adenosine receptor agonist,⁶ while Bay 60-6583 (Bay) is a specific A₂BR agonist.³³ Isoproterenol (Iso) is a general βAR agonist, while propranolol (Prop) is a general βAR antagonist. ²¹ KT-5720 is a protein kinase A (PKA) inhibitor. ³⁴

We show here that βAR activation inhibits Rap1B prenylation, providing a greater understanding of how the βAR increases breast cancer metastasis. Furthermore, we show that Rap1B prenylation is also inhibited by the general Gα_s activator Ctx. We show that the β-blocker Prop decreases migration in a concentration-dependent manner. These findings provide support for the repurposing of FDA-approved drugs, such as β-blockers, to target the GPCR-dependent signaling pathways that regulate Rap1B prenylation in breast cancer.

Results

Activation of Gα_s protein-coupled receptors inhibits Rap1B prenylation

Two Rap1B mutants were used to examine the phosphorylation status of Rap1B; a phosphodeficient mutant with serines 179 and 180 mutated to alanines (Rap1B-AA) and a phosphomimetic mutant with serines 179 and 180 mutated to glutamate (Rap1B-EE). A prenylation deficient Rap1B mutant was created (Rap1B-SAAX), with the normally prenylated cysteine mutated to serine, resisting prenylation. To first determine how phosphorylation of Rap1B affects its prenylation status in breast cancer cells, MDA-MB-231 cells were transfected with cDNAs encoding either HA-tagged wildtype Rap1B (HA-Rap1B-WT), phosphodeficient HA-Rap1B-AA, phosphomimetic HA-Rap1B-EE, or prenylation deficient HA-Rap1B-SAAX. Prenylated Rap1B can be identified by shifts in migration that occurred during SDS-PAGE, with prenylated Rap1B migrating more quickly than non-prenylated Rap1B, as previously described. ⁶ As shown in Figure 1, both HA-Rap1B-WT and HA-Rap1B-AA migrate as doublets, with the faster migrating form representing prenylated Rap1B and the slower migrating form representing non-prenylated Rap1B. Phosphodeficient HA-Rap1B-AA has a higher ratio of the prenylated to non-prenylated form compared to HA-Rap1B-WT. The slower migrating non-prenylated HA-Rap1B-WT and -AA bands run with the prenylation deficient -SAAX mutant. Phosphomimetic HA-Rap1B-EE also migrates with the slower migrating non-prenylated HA-Rap1B-SAAX mutant, suggesting that phosphorylated Rap1B is not prenylated in MDA-MB-231 breast cancer cells.

Figure 1.

Rap1B phosphomutants in MDA-MB-231 breast cancer cells. MDA-MB-231 cells were transfected with cDNAs encoding the indicated HA-tagged wildtype or mutant Rap1B protein, and lysed 24 hr later. Cell lysates were immunoblotted using HA antibody. In each lane, the faster migrating protein represents prenylated Rap1B, whereas the slower migrating protein represents non-prenylated Rap1B, as indicated by the “p” and “np” symbols respectively, at the right of the blot. The immunoblot is representative of 3 independent experiments. The # symbol on the immunoblot indicates the molecular weight marker of 37 kDa, whereas the * symbol indicates the molecular weight marker of 26 kDa.

To examine the rate of prenylation of Rap1B, a mevastatin-block-and-release assay was used. In this assay, mevastatin inhibits isoprenoid synthesis so prenylation cannot occur, causing newly synthesized GTPases to accumulate in the non-prenylated state. When the mevastatin is washed away, isoprenoid synthesis resumes, which initiates prenylation of the GTPases that were synthesized but not prenylated when mevastatin was present. The rate of prenylation can be determined by Western blotting of the cell lysates collected at time points after removing the mevastatin. The ratio of prenylated Rap1B to total Rap1B can be calculated by densitometry of the differently migrating forms of Rap1B detected in the immunoblots. Using this assay, we first examined the rate of prenylation of Rap1B in MDA-MB-231 breast cancer cells transfected with HA-tagged Rap1B and HA-tagged Rap1B mutants. The general Gα_s activator Ctx inhibits prenylation of HA-Rap1B-WT (Fig. 2A). Additionally, treatment of cells with the βAR agonist Iso inhibits prenylation of HA-Rap1B-WT (Fig. 2B). Furthermore, examination of a more epithelial breast cancer line, MDA-MB-468, determined that Iso treatment inhibited prenylation of HA-Rap1B-WT in that cell line as well (Fig. 2C). Phosphomimetic HA-Rap1B-EE is not detectably prenylated (Figs. 2A and B). There is a small decrease in prenylation of HA-Rap1B-AA induced by both Ctx and Iso, but analysis using 2-way ANOVA indicated that Ctx and Iso did not significantly alter HA-Rap1B-AA prenylation (Figs. 2A and B). No significant differences in prenylation were detected due to drug treatments of MDA-MB-468 cells transfected with the HA-Rap1B phosphomutants (data not shown), similar to the MDA-MB-231 cells. Taken together, these data indicate that the Ctx- and Iso-induced inhibition of prenylation is through phosphorylation of serines 179 and 180 in the PBR of Rap1B.

Figure 2.

For figure legend, see page .

Next we examined the prenylation of endogenous Rap1B. We compared the effects of Ctx, Iso, and Bay on the prenylation of endogenous Rap1B. Bay is a specific A₂BR agonist that was tested for two reasons. First, even though a previous study using a general adenosine receptor agonist suggested that Rap1B prenylation is regulated by the A₂B subtype of adenosine receptor, ⁶ the ability of specific A₂BR agonists to regulate this pathway has not been formally tested. Secondly, by comparing the effects of Bay, Iso, and Ctx, we could compare how Rap1B prenylation is regulated by A₂BR activation, βAR activation, and activation of all Gα_s subunits, respectively. We found that prenylation of endogenous Rap1B is most effectively inhibited by activating all Gα_s subunits with Ctx (Figs. 3B and D). Prenylation of endogenous Rap1B is also inhibited by Iso and Bay (Fig. 3A, C-D). Interestingly, prenylation is delayed more by treatment with Ctx or Iso than by treatment with Bay. This conclusion is based on our finding that 6 hours after removal of mevastatin, there is significantly less prenylation of Rap1B in Ctx- and Iso-treated cells, but not in Bay cells, compared to untreated cells (Fig. 3D). These results suggest that Rap1B prenylation is more effectively inhibited by activating βAR or all Gα_s subunits than by activating A₂BR in these assay conditions.

Figure 3.

Prenylation of endogenous Rap1B in MDA-MB-231 cells is inhibited by activating Gα_s with Ctx, activating the A₂BR with Bay, or activating the βAR with Iso. Untransfected MDA-MB-231 cells were incubated with mevastatin for 22-26 hrs, washed, and then incubated in the absence or presence of Bay (0.1 µM) (A), Ctx (0.1μg/ml) (B), or Iso (0.1 μM) (C). Cell lysates were collected at the indicated times and immunoblotted using Rap1B antibody. (D) The graphs were generated as in Figure 2, and show the mean ± SEM, with n=3. A two-way ANOVA with Bonferroni multiple comparisons was performed with the comparison to no drug indicated below the time point analyzed. The two-way ANOVA had a p-value of p <0.0001 when looking at differences due to the treatment/drug. ** p<0.01, *** p< 0.001, **** p<0.0001.

Interestingly, we observed that transfected HA-Rap1B-WT is completely non-prenylated after the cells were incubated with mevastatin (0 timepoint, Fig. 2), while a small proportion of endogenous Rap1B remains prenylated at the end of the mevastatin block (0 timepoint, Fig. 3D). HA-Rap1B-WT might be completely non-prenylated because the cells were treated with mevastatin shortly after they were transfected with the cDNA encoding HA-Rap1B-WT, which would cause HA-Rap1B-WT to be synthesized in the absence of any isoprenoid production. In contrast, endogenous Rap1B would be prenylated in the cells before the addition of mevastatin. Although most of this endogenous, prenylated Rap1B would be degraded during the 20-24 hour mevastatin incubation, the small fraction of previously prenylated Rap1B remaining in the cells would be detectable at the end of the mevastatin block. When mevastatin was washed away at time zero, treatment of these non-transfected cells with Ctx, Bay, or Iso would promote phosphorylation of newly synthesized, endogenous Rap1B and inhibit its prenylation. During this time, the previously synthesized and prenylated Rap1B will continue to be degraded, while the more recently synthesized and unprenylated Rap1B will remain, leading to the slight decrease in the ratio of prenylated to total endogenous Rap1B observed during the first few hours after release from the mevastatin block (time 0 - 2 hours, Fig. 3D).

Having shown that Gα_s activation can inhibit prenylation of Rap1B, we used the PKA inhibitor KT-5720 to examine whether the inhibition of Rap1B prenylation occurs through PKA mediated-phosphorylation of Rap1B. After the wash to remove mevastatin, the cells were pre-incubated with KT-5720 before adding the Ctx. As shown in Figure 4A and B, pretreatment of MDA-MB-231 cells with KT-5720 before and during Ctx treatment mitigates the effects of Ctx on prenylation. The cells at the 4 and 6 hour timepoints treated with KT-5720 and Ctx (KT-5720 + Ctx) have increased prenylation of Rap1B compared to cells treated only with Ctx. These data support our model, which shows that PKA-mediated phosphorylation of Rap1B inhibits prenylation. KT-5720 by itself has no significant effect on prenylation. Similar to the results shown in Figure 3, cells treated with Ctx exhibited a slight increase in Rap1B prenylation at the earlier timepoints, followed by a decrease, as the older, previously prenylated Rap1B was degraded and the Ctx continued to inhibit prenylation of newly synthesized Rap1B.

We also tested the effects of the β-blocker Prop on prenylation of Rap1B in MDA-MB-231 cells, as the therapeutic goal of these studies is to decrease the metastatic potential of breast cancer cells through increased Rap1B at the plasma membrane. Prop increased the rate of prenylation of endogenous Rap1B, which was most noticeable at 4 - 5 hours after release from the mevastatin block (Fig. 4C and D). Our findings that the βAR agonist Iso suppresses Rap1B prenylation (Fig. 3D), while the βAR antagonist Prop has the opposite effect (Fig. 4D), supports our model that βAR activation negatively regulates Rap1B prenylation.

Figure 4.

The PKA inhibitor KT-5720 mitigates Ctx-mediated inhibition of endogenous Rap1B prenylation, and Prop increases prenylation of endogenous Rap1B. (A, B) Untransfected MDA-MB-231 cells were incubated with mevastatin for 22-26 hrs. After washing the cells to remove the mevastatin, the cells were preincubated in the absence or presence of KT-5720 (10 μM) for 15 min, and then incubated up to 6 hours in the absence or presence of KT-5720 (10 μM) and/or Ctx (0.1μg/ml), as indicated. Cell lysates were collected at the indicated times and immunoblotted using Rap1B antibody. The graphs were generated as in Figure 3, and show the mean ± SEM, with n=3. A two-way ANOVA with Bonferroni multiple comparisons was performed, with the comparison to no drug indicated below the graph at the time point analyzed. A comparison between Ctx and KT-5720 + Ctx is shown on the graph, with significance only at 6 hours. The two-way ANOVA had a p-value of p <0.0001 when comparing overall differences due to the treatment/drug. * p<0.05, ** p<0.01, *** p< 0.001, **** p<0.0001 (C, D) Cells were treated as in A and B, but the cells were preincubated in the absence or presence of 25 μM Prop for 30 min before removing the mevastatin, and then incubated up to 6 hours in the presence or absence of 25 μM Prop after removal of the mevastatin. The graph was generated as in B, and shows the mean ± SEM, with n=4. A two-way ANOVA with Bonferroni multiple comparisons was performed. The two-way ANOVA indicated a p-value of 0.2138 when comparing Rap1B prenylation in Prop-treated cells vs. untreated cells over the time course of the 4 experiments.

Gα_s activation leads to phosphorylation of Rap1B

The results we obtained using the Rap1B phosphomutants indicate that Ctx or Iso inhibits Rap1B prenylation through phosphorylation of serines 179 and 180 (Figs. 2-4). To confirm that these agonists promote Rap1B phosphorylation, we examined whether Ctx and Iso induce ³²P-labeling of serines 179 and 180 in Rap1B. HEK293T cells were transfected with cDNAs encoding myc-Rap1B-WT, phosphodeficient myc-Rap1B-AA, phosphomimetic myc-Rap1B-EE, or prenylation deficient myc-Rap1B-SAAX. After incubating the cells with media containing ³²P-orthophosphate with or without Ctx or Iso, the myc-tagged Rap1B proteins were immunoprecipitated and examined for phosphorylation. Immunoblotting of the immunoprecipitates corroborated the results obtained in Figure 2; Iso and Ctx both inhibit prenylation of myc-Rap1B-WT, as shown by reduced detection of the rapidly migrating prenylated form of Rap1B (Fig. 5A, lanes 1-3). Phosphodeficient myc-Rap1B-AA is prenylated more effectively than wildtype Rap1B (Fig. 5A, lane 4 versus lane 1), whereas phosphomimetic myc-Rap1B-EE and prenylation deficient myc-Rap1B-SAAX are not detectably prenylated irrespective of drug treatment (Fig. 5A, lanes 7-12).

Figure 5.

Treatment with Ctx or Iso promotes phosphorylation of Rap1B at serines 179 and 180. (A) HEK293T cells expressing myc-tagged wildtype or mutant Rap1B proteins were pulsed with ³²P orthophosphate for 3 hours in the presence or absence of Ctx (0.1 μg/ml) or Iso (10 μM). The myc-Rap1B was immunoprecipitated and immunoblotted using myc antibody. (B) The immunoblot shown in A was analyzed by phosphoimaging to detect the ³²P-labeled proteins in the immunoblot. Densitometry was performed on the phosphoimage using ImageJ.

Examination of the phosphoimage indicates phosphorylation of both myc-Rap1B-WT (Fig. 5B, lane 1) and the prenylation deficient myc-Rap1B-SAAX mutant (Fig. 5B, lane 10), indicating that Rap1B phosphorylation can occur before Rap1B is prenylated. Most notably, neither phosphodeficient myc-Rap1B-AA nor phosphomimetic myc-Rap1B-EE exhibits phosphorylation (Fig. 5B, lanes 4 - 9), indicating that phosphorylation occurs on serines 179 and 180. Phosphorylation of myc-Rap1B-WT and myc-Rap1B-SAAX is increased more by treatment with Ctx (Fig. 5B, lanes 2 and 11) than by treatment with Iso (Fig. 5B, lane 3 and 12). These results support our model that activation of all Gα_s in the cells, or activation of only those Gα_s that are coupled to βAR, inhibits prenylation of Rap1B through phosphorylation of serines 179 and 180.

Loss of Rap1B prenylation leads to loss of Rap1B localization at the plasma membrane and reduced cell-cell adhesion

Prenylation is required for membrane localization of Rap1B. Therefore, as we have shown that Ctx and Iso inhibit prenylation, we used live-cell imaging to examine whether these drugs also affect the subcellular localization of wildtype or mutant Rap1B proteins tagged with green fluorescent protein (GFP). We found that GFP-Rap1B-WT has perinuclear and membrane localization in MDA-MB-231 breast cancer cells (Fig. 6B). Treatment with Ctx or Iso diminishes the membrane localization of GFP-Rap1B-WT, causing GFP-Rap1B-WT to accumulate in both the cytosol and nucleus (Fig. 6F and J). In contrast, treatment with Ctx or Iso does not alter the membrane- and perinuclear-localization of GFP-Rap1B-AA (Figs. 6G and K), whereas GFP-Rap1B-EE accumulates in the cytosol and nucleus even in the absence of drugs (Fig. 6D). These findings indicate that when cells are treated with Ctx or Iso, the increase in Rap1B phosphorylation and subsequent suppression of Rap1B prenylation inhibits the ability of Rap1B to localize at the plasma membrane.

Figure 6.

Rap1B localizes in the cytosol and nucleus of MDA-MB-231 cells when βAR are activated by Iso, or when Gα_s is activated by Ctx, and this localization is controlled by phosphorylation of serines 179 and 180 in Rap1B. MDA-MB-231 cells were transfected with cDNAs encoding the indicated GFP-tagged wildtype or mutant Rap1B proteins, and 90 minutes after transfection, the cells were exposed to no drug (A-D), Iso (0.1 μM, E-H), or Ctx (0.1 μg/m, I-L). The cells were imaged by confocal fluorescence microscopy after culturing for an additional 18 hours in the presence or absence of the drugs. All images are at the same magnification and are representative of at least 3 independent experiments.

Because MDA-MB-231 cells already have a very mesenchymal and therefore non-adherent phenotype, localization of GFP-Rap1B was examined in the more epithelial MDA-MB-468 breast cancer cells as well. MDA-MB-468 cells stably expressing GFP-vector, GFP-Rap1B-WT, -AA, or -EE were imaged and are shown in Figure 7. Similar to MDA-MB-231 cells, both GFP-Rap1B-WT and phosphodeficient GFP-Rap1B-AA localize at membranes and cell junctions of MDA-MB-431 cells, while phosphomimetic GFP-Rap1B-EE has almost no membrane or junctional localization (Fig. 7).

Figure 7.

Rap1B-WT and phosphodeficient Rap1B-AA exhibit more membrane and junctional localization than phosphomimetic Rap1B-EE in MDA-MB-468 cells. MDA-MB-468 cells were stably transfected with cDNAs encoding the indicated GFP-tagged wildtype or mutant Rap1B proteins. The cells were imaged by confocal fluorescence microscopy. All images are at the same magnification and are representative of at least 2 independent experiments.

Changes in Rap1B prenylation through Gα_s activation leads to changes in breast cancer cell migration

To examine migration of MDA-MB-231 breast cancer cells, a scratch wound assay was performed using the Incucyte Live Cell Imager, which obtains images of the migrating cells every two hours. The Incucyte software calculates the relative wound density, which is the percent of the original scratched wound covered by the migrating cells at a given time point. Cells treated with the A₂BR agonist Bay or with Ctx exhibited an increase in migration, as well as an increased rate constant and decreased half time (T_1/2) to wound closure (Fig. 8A). Representative images of the scratched wounds after ten hours (Fig. 8A) indicate that cells treated with Ctx or Bay had migrated fast enough to completely repopulate the scratched wound, while the non-treated cells had not yet completely populated this area.

Figure 8.

Activation of GPCRs coupled to Gα_s alters the migration but not the proliferation of MDA-MB-231 cells. (A, B) Confluent MDA-MB-231 cells were scratched to generate a region devoid of cells, treated with the indicated drugs, and then imaged every 2 hours using the Incucyte Live Cell Imager. The relative wound density represents the percent of the original scratched area that is repopulated by the migrating cells at a given time. Data are the mean ± SEM of at least 3 independent experiments, with at least 5 wells per treatment per experiment. The best-fit line (One phase association, Prism) for each individual experiment was determined, and the constants were averaged to obtain the values shown in the tables. The T_1/2 is half the time until the scratched wound is completely filled by migrating cells. A one-way ANOVA with Dunnet's Multiple Comparisons to no drug was performed on all of the rate constants, with ** p <0.01 and *** p<0.001. Representative images of the cells collected at 10 hours (A) and 14 hours (B) are shown at the right of the graphs. In these images, the black color depicts the area populated by the confluent cells at the time when the scratch was made, the gray color depicts the area of the scratch that was re-populated by the migrating cells at 10 hours (A) and 14 hours (B), and the blue color depicts the area that remained cell-free at these times. (C) MDA-MB-231 cells were cultured in the absence or presence of the indicated drugs for 72 hours, and cell proliferation was determined by measuring uptake of [³H]thymidine. Values were normalized to [³H]thymidine uptake occurring in control cells treated without drugs. The values are the mean ± SEM of 3 independent experiments. The data were analyzed using a one-way ANOVA with Dunnet's post-test resulting in no significant differences in cell proliferation. In addition, each column was examined using a one-sample t-test with the theoretical mean 1.000 (as they are all normalized to no drug, which was set to 1), and no significant differences in cell proliferation were found.

Cells treated with the βAR agonist Iso also had an increase in migration, with an increased rate constant and a decreased half time to wound closure (Fig. 8B). Additionally, the β-blocker Prop decreased migration in a concentration dependent manner, shown graphically and by a significantly decreased plateau in wound closure by cells treated with 25 μM and 50 μM Prop (Fig. 8B). The plateau is the value in the rate equation that indicates the greatest relative wound density the cells will achieve, or the most the wound will actually close. A full closure is a relative wound density of 100. The only plateaus below 100 are those exhibited by cells treated with 25 μM and 50 μM Prop (Fig. 8B). A one-way ANOVA with Dunnet's Multiple Comparisons to no drug was performed on all of the rate constants, with significance seen only for the plateau values of cells treated with 25 μM and 50 μM Prop (Fig. 8A, B). Representative figures of the scratched wounds at 14 hours (Fig. 8B) indicate that Iso-treated cells had closed the wound completely, while untreated cells had not yet completely closed the wound. Treatment with 25 μM Prop slows the ability of the migrating cells to completely repopulate the scratched area and close the wound, and treatment with 50 μM Prop slows wound closure even further. These drug treatments did not inhibit the proliferation of the cells (Fig. 8C), indicating that the effects of the drugs on the abilities of the cells to repopulate the scratched areas were due to changes in the migratory behavior of the cells.

To statistically analyze the effects of the drugs on cells repopulating the scratched areas, a conditional growth curve analysis was conducted. Conditional growth curves consider both inter- and intra-group variation to account for systematic differences in the outcome caused by the treatment while also controlling for between-unit heterogeneity and the effects of the testing environment. In the case of the cells migrating in the wells of a 96-well plate (Fig. 8), the analysis models growth trajectory at the individual-well level, but controls for changes in wound healing due to potential irregularities in cell plating, cell passage, and any other differences based on the day of the trial. These models offer a number of benefits relative to traditional ANOVA by investigating growth trajectory throughout the trials, not just at the end, modeling non-linear growth, calculating growth rates, and increasing statistical power. Conditional growth curve models offer the additional benefit of capturing variation in treatment effects over time. For example, a given treatment's effect on wound healing may be strongest during the initial periods, but diminish as relative wound density approaches 100. Traditional analyses often obscure this substantively interesting and important variation, leading to Type-I error.

The analysis confirms that cells treated with Bay or Ctx re-populated the scratch wound faster and more completely than the control group. Marginal effects for Ctx and Bay displayed in Table S1 (column 3-4) show greater initial (0-2 hr) repopulation rates (β_Bay = 3.119, β_Ctx = 4.80) elative to the control group. However, these rates diminish relative to the control group over time, as indicated by the decreasing marginal effects values as time increases. Cells treated with Bay repopulate significantly faster than the control group for approximately fourteen hours while those treated with Ctx maintain a faster growth rate throughout the entire timeframe. This is depicted in Figure S1, which shows the average marginal effects with 95% confidence intervals for Bay and Ctx. Each line depicts the predicted effect of the treatment/drug on relative wound density relative to the control group. If the control group was depicted, it would be a straight line at y=0. The lines can be interpreted as testing the hypothesis, β_treatment – β_control = 0 Therefore, where the 95% confidence interval is not crossing 0, the relative wound density at that time point is significantly different than the control group. Note that the 95% confidence intervals for Bay and Ctx overlap, indicating that their influences on relative wound density cannot be distinguished from one another.

Table S1 also presents similar models for the effects of Prop and Iso on relative wound closure relative to the control group. Iso (Table S1, column 2) plays a role similar those of Bay and Ctx. Cells treated with Iso migrated significantly faster than untreated cells for approximately 10 hours (see Table S1, column 2). The analysis suggests a dose-dependent negative effect on relative wound density for Prop. Figure S2 best demonstrates this, showing the average marginal effects with 95% confidence intervals for Iso and Prop, similar to Figure S1. Note that all three Prop doses, 10, 25, and 50 μM, begin with statistically similar effects on the outcome. By hour 4, the 50 μM dose has an appreciably distinct inhibiting effect, followed by the 25 μM dose after about 12 hours. All three doses of Prop produce statistically different effects in the cells compared to the untreated cells, indicating that cell migration is inhibited by inactivating the βAR with increasing concentrations of the β-blocker Prop.

Discussion

Taken together, these data indicate that Gα_s protein-coupled receptors, specifically the βAR, may be potential targets to decrease breast cancer metastasis. We show that both Ctx and the βAR agonist Iso inhibit prenylation of Rap1B due to the phosphorylation of serines 179 and 180, resulting in decreased membrane localization of Rap1B and reduced cell-cell adhesion, which promotes cell scattering. Our observation that cell migration is diminished by inhibiting the βAR with the β-blocker Prop supports further investigation of this pathway as a therapeutic target in cancer.

We observed that phosphorylation of Rap1B was induced both by Ctx and, to a lesser extent, by Iso. Ctx would be expected to induce a much greater increase in phosphorylation than that induced by Iso, because Ctx ADP-ribosylates Gα_s which locks Gα_s into an active formation that cannot cycle, causing continuous activation of PKA by ADP-ribosylated Gα_s.³² On the other hand, when Iso increases Gα_s activity through stimulation of βAR, the activated Gα_s cycles on and off as GTP is hydrolyzed; it does not stay locked in an active formation. Furthermore, Ctx will activate all of the Gα_s subunits in a cell, whereas Iso will only activate the Gα_s subunits associated with βAR.

All of our data indicate that the changes in Rap1B occurring due to Gα_s activation, including Rap1B phosphorylation, prenylation, and localization, are due to phosphorylation of serine 179 and 180 on Rap1B, because Gα_s activation significantly affects Rap1B-WT, but not the phosphomutants. However, we do see a slight, although not significant, decrease in prenylation of Rap1B-AA with Iso or Ctx treatment, suggesting that Gα_s activation may inhibit prenylation of Rap1B through additional mechanisms that occur independently of Rap1B phosphorylation. It is interesting to speculate that the activation of cAMP-PKA-mediated signaling may also inhibit the activity of the prenyltransferase, specifically geranylgeranyltransferase.

According to our model, PKA must phosphorylate newly synthesized Rap1B before Rap1B enters the prenylation pathway. This model is supported by our observations that treatment with Ctx or Iso phosphorylates the non-prenylated Rap1B-SAAX mutant (Fig. 5) and slows the initiation of Rap1B prenylation when mevastatin-treated cells are washed to remove the mevastatin (Fig. 2). An important aspect of this model is that PKA continuously phosphorylates Rap1B as it is constantly synthesized in the cell. The βAR is an ideal GPCR to promote this continuous phosphorylation of newly synthesized Rap1B, because βAR are activated by the catecholamine stress hormones, epinephrine and norepinephrine, which are chronically elevated in patients with cancer. ³⁵ Thus, newly synthesized Rap1B might be continuously phosphorylated and inhibited from entering the prenylation pathway in tumors that have βAR, which are chronically activated by the catecholamine stress hormones, or in tumors that have A₂BR, which are chronically activated by adenosine generated by hypoxic tumors. ⁶

Prop and other β-blockers that target βAR are especially promising drugs for reducing breast cancer metastasis and reoccurrence for several reasons. First, they are already in use in clinic for several cardiovascular diseases and therefore may be more easily repurposed for breast cancer than trying to get a new drug FDA-approved. Second, there are already several retrospective studies that indicate breast cancer patients taking β-blockers have increased survival and decreased reoccurrence. ^{15,20,21,23-27,29} Lastly, the βAR has been found to be important in regulating metastasis and/or migration in many other cancers, including pancreatic,^36-38 ovarian,³⁵ colorectal, ³⁹ lung, ⁴⁰ prostate, ^18,36 and leukemia.^41,42 It will be interesting to investigate whether additional receptors are also promising targets to reduce breast cancer metastasis, and possibly metastasis of other solid cancers.

Many previous studies examining the roles of β-blockers in cancer focus on the use of β-blockers in combination with Cox inhibitors perioperatively (especially pre-operatively).^{19,22,40,42,43} The results of these previous studies indicate a maximal decrease in cancer reoccurrence or metastatic formation with treatment with both a β-blocker and a Cox inhibitor, although several studies indicate a positive outcome with just one treatment as well.^{19,22,40,42,43} Interestingly, Cox enzymes synthesize prostaglandins, which are agonists for several Gα_s protein-coupled receptors. Thus, it is possible that Cox inhibitors suppress prostaglandin-dependent Gα_s signaling.

The rationale for using β-blockers and Cox inhibitors perioperatively is based on the clinical observation that surgery increases metastasis, through many mechanisms including immunosuppression and dissemination of tumor cells into the body. ^44,45 Surgery puts a great stress on the body, and thereby leads to increased stress hormones (such as the βAR agonists epinephrine and norepinephrine) and increased prostaglandins, both of which have an immunosuppressive effect. ⁴⁵ It is possible that the elevation in catecholamines (βAR agonists) and prostaglandins that occurs perioperatively not only increases metastasis through immune suppression, as previously shown, but also increases metastasis through activating Gα_s protein-coupled receptors that decrease Rap1B prenylation and therefore allow for dissemination of the tumor cells throughout the body. Our results support this hypothesis, providing another mechanism for β-blockers (and potentially Cox inhibitors) to decrease breast cancer metastasis and reoccurrence post-surgery of the primary tumor. Future studies should explore mechanisms to inhibit Rap1B phosphorylation and promote its prenylation as a novel therapeutic approach in cancer.

Experimental Procedures

Cell culture and transfection

MDA-MB-231 breast cancer cells and HEK293T cells were cultured in Dulbecco's modified Eagles medium (DMEM) with 10% heat inactivated fetal bovine serum, sodium pyruvate, and antibiotics (penicillin/streptomycin). MDA-MB-468 breast cancer cells were cultured in RPMI-1640 medium with L-glutamine, 10% heat inactivated fetal bovine serum, and antibiotics. Complementary DNAs (cDNAs) were synthesized previously and transfected using Lipofectamine 2000 (Invitrogen) according to the manufacturer's protocol.

Mevastatin-block-and-release assay

MDA-MB-231 cells were transfected (except when examining endogenous Rap1B) with cDNAs encoding HA-tagged Rap1B, cultured for 4 hours, and then treated with 10 μM mevastatin, which blocks the prenylation of newly synthesized Rap1B. After culturing for 18 - 26 hours (as indicated in figure legends), the cells were washed with phosphate buffered saline (PBS), which removes the mevastatin and releases the blockade of prenylation. ⁶ When examining the effects of Prop, it was added to the mevastatin-treated cells 30 minutes before the cells were washed and trypsinized. The cells were then treated with the indicated drug and re-plated, and lysed at the indicated time after removal of mevastatin. Samples were subjected to SDS-PAGE and then transferred to PVDF membranes. Western blotting of the cell lysates was performed using a mouse-anti-HA antibody (Covance No. MMS-101P) and a rabbit-anti-Rap1B antibody (Cell Signaling No. 2326S). Changes in Rap1B prenylation were identified by differences in the migration of Rap1B in the immunoblots, which occurs due to prenylated Rap1B migrating more rapidly than non-prenylated Rap1B during SDS-PAGE, as previously described. ⁶

Analysis of Rap1B phosphorylation by in vivo ³²P labeling

This ³²P labeling assay was performed as described previously. ⁶ Briefly, HEK293T cells were transfected with cDNA encoding myc-tagged Rap1B and incubated overnight. The media was then replaced with phosphate-free DMEM containing 200 μCi/ml of ³²P-orthophosphate with or without 10 μM isoproterenol or 0.1 μg/ml Ctx and the cells were incubated for 3 hours. The cells were then washed twice in PBS, lysed in NP-40 lysis buffer with phosphatase and protease inhibitors, and immunoprecipitated with rabbit-anti-Myc antibody (Sigma No. C3956) and Protein A agarose beads (Invitrogen No. 15918-014). Samples were subjected to SDS-PAGE and transferred to PVDF membranes. After immunoblotting using a mouse-anti-Myc antibody (Abcam No. 18185), phosphoimages of the immunoblots were obtained using a Storm 820 Phosphoimager.

Fluorescence imaging

HEK293T, MDA-MB-468, or MDA-MB-231 cells were plated on 35-mm glass bottom culture dishes (MatTek Corp.) and cultured for 48 hours. Cells were then transfected with cDNAs encoding green fluorescent protein (GFP)-tagged Rap1B and treated with the indicated drugs. The next day, cells were imaged using a Nikon A1-R Confocal microscope.

Thymidine proliferation assays

Thymidine proliferation assays were performed as described previously. ⁴⁶ Briefly, MDA-MB-231 cells were plated in a 96-well plate at a concentration of 2.5 × 10³ cells/well, incubated overnight, and then treated with the indicated drug. After three days, ³H thymidine was added to each well and cells were incubated for 3 hours as previously described. ⁴⁶ The cells were then harvested using a Scantron Cell Harvester and the thymidine uptake in the cells was measured by β-scintillation counting.

Migration assays

MDA-MB-231 cells were plated to confluence in a 96-well plate, and the cells were treated with the indicated drug overnight before a uniform wound was made in each well using a 96-well Wound Maker (Incucyte). The cells were then washed twice with PBS and fresh media containing the indicated drug was added. The plates were then placed in the Essen Incucyte Cell Imager, which collected images automatically every two hours for up to two days. The relative wound density was calculated by the imager software and indicates the percent of the original wound covered by the migrating cells. In this analysis, a relative wound density of 0% indicates that no cell migration occurred, whereas a relative wound density of 100% indicates that the migrating cells completely populated the cell-free area. The one phase association curve (Prism) was calculated for each experiment and the rate constant, half time, and plateau were all averaged to obtain the values in the tables.

Statistical Analysis

A two-way analysis of variance (ANOVA) with Bonferroni multiple comparisons was used to examine differences in prenylation rate in the mevastatin-block-and-release assay, examining the effects of time and drug/treatment. A one-way ANOVA with Dunnet's multiple comparisons compared to no drug was used to examine differences in thymidine proliferation. Additionally, for the thymidine proliferation assay, each column/drug treatment was compared to the value 1.000 using a student's t-test, to evaluate differences from the normalized no drug value of 1.000.

Conditional growth curve analysis

Data from the Essen Incuctye Cell Imager was analyzed using STATA v13.1 statistical software. Conditional growth curves were estimated using STATA's xtmixed package suite. Likelihood ratio-tests suggested a quartic functional form with nested effects based on well and trial date, as well as random coefficients for time. Two separate models were estimated for Ctx/Bay and Iso/Prop, though a pooled analysis produces similar findings. Marginal effects over time were calculated using STATAs margins post-estimation command. Subsequent figures were produced using marginsplot. All data and replication .do files, with code, are available upon request.

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.

Supplemental Material

Supplemental data for this article can be accessed on the publisher's website.

Funding

This research is supported by funding from the NIH (R01 CA136799 and R01 CA188871), the Rock River Cancer Research Foundation, WBCS Inc., and by The Nancy Laning Sobczak, Ph.D., Breast Cancer Research Award. Jessica Wilson is a member of the Medical Scientist Training Program at MCW, which is partially supported by a training grant from NIGMS T32-GM080202.