Gβ Regulates Coupling between Actin Oscillators for Cell Polarity and Directional Migration

Abstract

For directional movement, eukaryotic cells depend on the proper organization of their actin cytoskeleton. This engine of motility is made up of highly dynamic nonequilibrium actin structures such as flashes, oscillations, and traveling waves. In Dictyostelium, oscillatory actin foci interact with signals such as Ras and phosphatidylinositol 3,4,5-trisphosphate (PIP3) to form protrusions. However, how signaling cues tame actin dynamics to produce a pseudopod and guide cellular motility is a critical open question in eukaryotic chemotaxis. Here, we demonstrate that the strength of coupling between individual actin oscillators controls cell polarization and directional movement. We implement an inducible sequestration system to inactivate the heterotrimeric G protein subunit Gβ and find that this acute perturbation triggers persistent, high-amplitude cortical oscillations of F-actin. Actin oscillators that are normally weakly coupled to one another in wild-type cells become strongly synchronized following acute inactivation of Gβ. This global coupling impairs sensing of internal cues during spontaneous polarization and sensing of external cues during directional motility. A simple mathematical model of coupled actin oscillators reveals the importance of appropriate coupling strength for chemotaxis: moderate coupling can increase sensitivity to noisy inputs. Taken together, our data suggest that Gβ regulates the strength of coupling between actin oscillators for efficient polarity and directional migration. As these observations are only possible following acute inhibition of Gβ and are masked by slow compensation in genetic knockouts, our work also shows that acute loss-of-function approaches can complement and extend the reach of classical genetics in Dictyostelium and likely other systems as well.

Fig 1. Inducible protein sequestration as a method to acutely inactivate Gβ.

(A) Inducible sequestration can be exploited to inactivate a protein of interest. Using the small molecule rapamycin (RAP), FRB-tagged Gβ can be recruited to an FKBP-tagged “anchor” at the endoplasmic reticulum (ER). Addition of RAP sequesters Gβ from its normal site of action at the plasma membrane and prevents it from activating downstream effectors. (B) Timecourse of Gβ sequestration. In cells lacking endogenous Gβ, but expressing FRB-RFP-Gβ and calnexinA-YFP-FKBP as an anchor at the ER, the speed and extent of sequestration were assayed by measuring the spatial correlation between YFP and RFP signals. For the highest dose of RAP, half-maximal heterodimerization is achieved within 5.6 min. To keep cells immobile, the experiment was performed in the presence of 10 μM latrunculinA. The spatial correlations between fluorescence signals from Gβ and anchor are plotted (n ≥ 20 cells per condition; mean +/- standard error of the mean [SEM]). Raw data can be found in S1 Data. (C) Representative images from a Gβ sequestration timecourse described in (B). Scale bar = 10 μm. (D) Gβ sequestration recapitulates Gβ-null phenotypes for receptor-stimulated signaling. Timecourses of chemoattractant stimulation (cyclic-AMP [cAMP]; 10 μM) are shown in four strains: wild-type (wt), Gβ-null, and cells expressing one or both components of the Gβ sequestration system. Each strain was stimulated in the presence and absence of rapamycin (5 μM; > 20 min incubation). Blot shows phosphorylation of PKBR1 (T309); Ras is used as a loading control. Schematic indicates the localization of Gβ (in orange) for each condition in test strains and the published localization for wt and Gβ-null cells. Further examples of signaling events blocked after Gβ sequestration can be found in S1 Fig.

Fig 2. Gβ sequestration impairs directional migration.

(A) Cells of the Gβ sequestration strain were incubated with rapamycin and exposed to a gradient of folate. Based on the expression of sequestration components, different subpopulations were identified, and directionality was measured after 30 min of migration. Plotted are the means (+/- S.E.M) of wild-type (wt): Gβ+/anchor- cells (n = 30, red); Gβ-null (Gβ-): Gβ-/anchor- cells (n = 48, green); and Gβ-sequestered: Gβ+/anchor+ cells (n = 31, yellow). ** indicates a highly significant p-value of < 0.02; n.s. indicates a not-significant p-value of > 0.05 (Student’s two tailed t test). Data are derived from five videos in two independent experiments. For comparison, directedness of wt (DH1) and Gβ- cells (n = 97 and n = 98, data from two videos in single experiments, respectively) is shown in light and dark grey bars. Raw data can be found in S1 Data. (B) Cells of the Gβ sequestration strain were incubated with rapamycin and exposed to an electrical field. Based on the expression of sequestration components, different subpopulations were identified, and directionality was measured after 30 min of migration. Plotted are the means (+/- stdev) of wt: Gβ+/anchor- cells (n = 33, red), Gβ-null: Gβ-/anchor- cells (n = 34, green); and Gβ-sequestered: Gβ+/anchor+ cells (n = 34, yellow). ** indicates a highly significant p-value of < 0.01; n.s. indicates a not-significant p-value of > 0.05 (Student’s two tailed t test). Data are combined from several fields of view of movies recorded on two separate days. A movie corresponding to the stills in Fig 2B is included as S2 Movie. Raw data can be found in S1 Data.

Fig 3. Acute Gβ sequestration leads to oscillations in cortical F-actin.

(A) Acute sequestration of Gβ induces cytoplasmic oscillations of the F-actin reporter LimE-GFP. Cells were treated with 1 μM rapamycin, and LimE-GFP (upper panels) was imaged over time. Graphs show cytoplasmic LimE-GFP quantified from individual cells. Lower panels: FRB-RFP-Gβ and calnexinA-YFP-FKBP images show colocalization (sequestration) of Gβ at the anchor. Scale bar = 5 μm. Numbers indicate time in seconds. Corresponding oscillations at the cortex can be seen in S2 Fig. (B) Behavior of the F-actin reporter in Gβ unsequestered cells. LimE-GFP (upper panels) was imaged over time. Graphs show cytoplasmic LimE-GFP quantified from individual cells. Lower panels: FRB-RFP-Gβ and calnexinA-CFP-FKBP images show distinct Gβ and anchor localization. Scale bar = 5 μm. Numbers indicate time in seconds. (C) The percentage of oscillating cells was quantified from cells expressing LimE-GFP and FRB-RFP-Gβ, either in the presence (+ anchor) or absence (–anchor) of calnexinA-YFP-FKBP. In both strains, cells were left untreated (–Rap) or incubated with 1 μM rapamycin (+Rap) for at least 20 min (n ≥ 150 cells per condition from three independent days; plotted are means +/- SEM). Raw data can be found in S1 Data. (D) The oscillatory phenotype is rescued by performing Gβ sequestration in the presence of wild-type Gβ. This indicates that sequestration of Gβ induces a loss-of-function phenotype. Wild-type cells expressing FRB-RFP-Gβ were incubated with 1 μM rapamycin (+Rap) for at least 20 min. LimE-GFP oscillations were compared between cells that co-expressed the anchor (calnexinA-CFP-FKBP) or lacked the anchor. (- anchor: n = 91; mean +/- stdev; + anchor: n = 23; mean +/- stdev). Further experiments presented as supplement: Oscillations of LimE are due to loss, and not gain, of Gβ function (S3 Fig). The computational pipeline used to analyze oscillations is presented in S4 Fig. In some cases, Gβ sequestration also induces waves of actin polymerization that travel around the cell perimeter (S5 Fig). Oscillations of LimE start rapidly after Gβ is sequestered (S6 Fig). Raw data can be found in S1 Data.

Fig 4. Actin oscillations depend on the extent and timing of Gβ sequestration.

(A) Higher levels of sequestration (lower concentration of active Gβ) result in a larger fraction of oscillating cells. To achieve stable, intermediate levels of Gβ sequestration, cells were cotreated with 300 nM rapamycin and 0, 5, 10, 20 or 40 μM FK506, a competitive inhibitor of rapamycin. The correlation between Gβ and the anchor signal was extracted from single cells of all treatment conditions, and cells with similar levels of correlation were analyzed together (see S7 Fig). Gβ-unsequestered cells (wt) were analyzed for comparison. Negative correlation values indicate anticorrelation of Gβ and anchor in the unsequestered state. The mean of at least 20 cells in each bin is shown. Raw data can be found in S1 Data. (B) Higher levels of sequestration (a lower concentration of active Gβ) do not affect the period of oscillation. Cells and treatment conditions are the same as analyzed in (A). (n ≥ 20 cells per sequestration bin; plotted are means +/- stdev). The amplitude of actin oscillations is also not affected by sequestration of Gβ (S8 Fig). Raw data can be found in S1 Data. (C) The percentage of oscillating cells decreases over time during Gβ sequestration and approaches the terminal Gβ-null state. Cells were incubated with 1 μM rapamycin, and the fraction of oscillating cells was determined at the timepoints indicated (n > 25 Gβ-sequestered cells per condition). Raw data can be found in S1 Data. (D) Acute inhibition via rapamycin mediated protein sequestration can reveal phenotypes that are not accessible through classic genetic perturbations. First, it can reveal consequences of protein depletion to intermediate levels, such as the gradual or all-or-none emergence of phenotypes (axis 1). Second, rapid inactivation can reveal immediate phenotypes that are not accessible to slower methods of gene inactivation (axis 2).

Fig 5. Gβ regulates coupling between individual actin oscillators.

(A) Method of analysis. Cells expressing LimE-GFP were imaged by confocal microscopy in the plane where cells make contact with the coverslip. At each timepoint, the cell periphery was divided computationally into 36 sectors. For each sector, the intensity of LimE-GFP was quantified over time. The graph shows the trace for a single sector. (B) The temporal order of actin regulators at peripheral actin foci. The indicated actin reporters showed pulsatile behavior at the cell’s periphery. We measured their appearance, relative to LimE, in the same sectors in double-labeled cells. Oscillators from several cells (HSPC-300 n = 9; ABD n = 11; Arp2 n = 4; Coronin n = 4) were analyzed (plotted are means +/- SEM). An example for a dual color sequence of LimE and Coronin oscillations is shown in S10 Fig. Raw data can be found in S1 Data. (C) Traces for all 36 sectors of one control cell in which Gβ was not sequestered. Individual sectors oscillate, but the overall average does not. (D) Traces for all 36 sectors of one cell in which Gβ sequestration was induced by treatment with 1 μM rapamycin. Individual sectors oscillate, and so does the overall average. (E) Analysis of oscillation parameters. Each point represents the average of all 36 sectors of one cell. Amplitude and period of oscillations are similar in unsequestered (Gβ+), Gβ-sequestered (Gβ-sequ.), and Gβ-null cells. In contrast, synchrony of oscillations is increased in acutely sequestered cells. Raw data can be found in S1 Data. Further data are presented as supplements: a histogram of oscillation periods of individual sectors for one cell for each condition is shown in S9 Fig. To extract and compare phase information, we used the Hilbert transform, as shown in S11 Fig. A histogram of phase distributions for one cell each is shown in S12 Fig. Membership of individual oscillating sectors with the phase locked consensus is fluid as shown in S13 Fig.

Fig 6. Gβ-mediated coupling bypasses established signaling pathways.

(A) Signaling activities upstream of F-actin formation do not show global oscillations upon sequestration of Gβ. Representative traces (changes in intensity of reporter constructs in the cytoplasm) of several movies are shown for Ras activity (Ras*) visualized with YFP-RBD(PI3K1), PIP₃ visualized with PhdA-GFP, and Rac activity (Rac*) visualized with GBD(PAK)-YFP. (B) Perturbations of core chemotactic regulatory pathways do not induce global oscillations of LimE. Neither induced expression of a constitutively active version of Rac1A (GFP-Rac1A-V12) nor a triple drug cocktail (BEL|LY294002|pp242) inhibiting PLA2-, PI3K-, and TORC2-mediated signaling induces global oscillations of LimE-RFP in wild-type (Ax2) cells. Similarly, increasing the intracellular concentration of Ca²⁺ has no effect. Data represent the means of more than 35 cells from at least 2 d for each condition (+/- stdev). Raw data can be found in S1 Data. (C) Inhibition of core chemotactic regulatory pathways does not abolish Gβ mediated global oscillations of LimE-GFP. Gβ-sequestered, oscillating cells were treated with various drugs to determine their effect on oscillatory behavior. Neither a triple-drug cocktail (BEL|LY294002|pp242) that simultaneously blocks PLA2-, PI3K-, and TORC2-mediated signaling, nor unbalancing Ca²⁺ levels (blocking PLC with U73122, supplying Ca²⁺ or chelating any Ca²⁺ present in the buffer with EGTA) had a significant effect on the presence of global LimE-GFP oscillations. Data represent the means of more than 25 cells from at least 2 d for each condition (+/- stdev). Raw data can be found in S1 Data. (D) Gβ appears to bypass established signaling pathways to regulate the spatial range of coupling. S19 Fig shows data to validate the use of pp242 as an inhibitor for TORC2 mediated signaling in Dictyostelium cells.

Fig 7. A hypercoupled cytoskeleton competes with establishment of cell polarity.

(A) In Gβ unsequestered (wt) cells, phases of polarization, characterized by low cytoplasmic LimE-GFP intensity, alternate with apolar phases. A confocal slice from the middle of an unsequestered cell is stacked into a kymograph (t-stack), where the y-axis represents time and the x-axis represents intensity along the cell’s lateral surface. Continuous bright areas (white oval) indicate LimE-GFP accumulation in a pseudopod. The corresponding trace of cytoplasmic LimE-GFP intensity on the right shows that phases of polarity (pink shading; see S14 Fig) coincide with low levels of cytoplasmic reporter (and, therefore, higher levels of polymerized actin at the periphery). A total of 28 unsequestered cells were analyzed with similar results. (B) Whole-field LimE-GFP oscillations are restricted to apolar phases in Gβ-sequestered cells. One representative t-stack and corresponding trace is shown. A total of 46 Gβ-sequestered cells were analyzed with similar results. S15 Fig shows that the strength of polarization is similar between Gβ-sequestered and Gβ-unsequestered cells. (C) The lifetime of poles in Gβ-unsequestered cells (grey; n = 28) and Gβ-sequestered cells (orange; n = 46) is similar (p = 0.51 for their difference, Student’s two-tailed t test; plotted are means +/- SEM). Raw data can be found in S1 Data. (D) The frequency at which poles are established in Gβ-sequestered cells is reduced compared to Gβ-unsequestered cells (the same set of cells as in Fig 4C is analyzed; plotted are means +/- SEM; p < 10^-4, Student’s two-tailed t test). Raw data can be found in S1 Data. (E) Cellular translocation is slowed down for Gβ-sequestered cells (mean +/- SEM = 0.59 +/- 0.14 μm²/min, n = 46) compared to Gβ-unsequestered cells (mean +/- SEM = 1.43 +/- 0.35 μm²/min, n = 28; p < 0.003, Student’s two-tailed t test). Mean squared displacement is a suitable metric for cell migration over short periods of time (S20 Fig). Raw data can be found in S1 Data. S18 Fig shows that oscillating actin foci are suppressed during cell polarization in wild-type cells. Abbreviations used: SEM = standard error of the mean; n.s. = not significant, p-value > 0.05; ** indicates a highly significant p-value of < 0.01.

Fig 8. Acute loss of Gβ induces a hypercoupled cytoskeleton.

By synchronizing weakly coupled peripheral oscillators, a hypercoupled state is induced that is apparent as whole-field oscillations. This pathologic state is less permissive to the establishment of cell polarity and continuous realignment of polarity in a gradient. We suggest that this hypercoupled state prevents oscillators from becoming patterned by upstream signaling cues from inside or outside the cell.

Fig 9. A mathematical model demonstrates that intermediate oscillator coupling is sufficient to increase sensitivity to noisy inputs.

Wild-type cells may be in a range of optimal coupling between cytoskeletal oscillators to facilitate entrainment by signaling cues. To investigate how coupling strength influences signal detection, we have built a simple model in which sectors all around a circle couple at a strength k ₂. The test area couples locally with a strength k ₁, and entrainment to an external input of strength k _IN (left panel) is assessed. At intermediate values for oscillator coupling, k ₁ entrainment to the input is optimal (middle panel). Examples for oscillator entrainment at different values of k ₁ are shown (right panel).