Coordinated integrin activation by actin-dependent force during T-cell migration

Abstract

For a cell to move forward it must convert chemical energy into mechanical propulsion. Force produced by actin polymerization can generate traction across the plasma membrane by transmission through integrins to their ligands. However, the role this force plays in integrin activation is unknown. Here we show that integrin activity and cytoskeletal dynamics are reciprocally linked, where actin-dependent force itself appears to regulate integrin activity. We generated fluorescent tension-sensing constructs of integrin αLβ2 (LFA-1) to visualize intramolecular tension during cell migration. Using quantitative imaging of migrating T cells, we correlate tension in the αL or β2 subunit with cell and actin dynamics. We find that actin engagement produces tension within the β2 subunit to induce and stabilize an active integrin conformational state and that this requires intact talin and kindlin motifs. This supports a general mechanism where localized actin polymerization can coordinate activation of the complex machinery required for cell migration.

Figure 1. Tensile force is transmitted through the integrin β-subunit.

(a) Integrin structure and conformational states. (b) Current model of actin-dependent integrin activation and illustration of the tension sensor module. The same conformations are shown as in a. The tension sensor (TS) consists of two FRET-compatible fluorescent proteins, monomeric teal (mTFP) and venus (mVenus), linked together with a repeating sequence that can be elongated by tensile force (GPGGA)₈. Black arrows represent force applied by the actin cytoskeleton and resisted by bound ligand. (c,d) Tension sensor insertion positions where the preceding 4-residue sequences (XXXX) are being repeated except for the C-terminal constructs. Important interaction sites are highlighted in magenta. Linker details are in d. (e) Representative FRET images from movies of migrating Jurkat T cells expressing indicated tension sensor constructs. Arrow indicates direction of cell movement. Scale bar, 5 μm. (f) Whole-cell average FRET in migrating Jurkat T cells with 100 ng ml⁻¹ SDF-1α. Circles represent individual cells from three independent experiments with median shown as a line. ****: Kruskal–Wallis with Dunn's multiple comparison test of differences between β2-TS3 and all other TS had P values <0.0001. (g) Acceptor photobleaching of live Jurkat T cells. Curves depict donor (mTFP) intensity levels and FRET ratios before and after photobleaching of acceptor (mVenus). Representative curves out of 10 cells.

Figure 2. β2 tension sensors placed N-terminal to actin adaptor-binding sites respond to force.

(a) Transient expression of β2 tension sensors in 293T cells. Each sensor was transfected together with wild type αL. LFA-1 was quantitated by flow cytometry with the heterodimer-specific β2 monoclonal antibody TS1/18. Average values±s.e.m. from three independent transfections. (b) Migration of Jurkat cells on 20 μg ml⁻¹ ICAM-1 with or without 100 ng ml⁻¹ SDF-1α. Each β2 construct was introduced by lentiviral transduction without addition of αL to allow pairing with wild type αL. WT is non-transduced Jurkat cells. From left to right: N=166/187; 177/293; 102/121; 355/317; 215/350; 315/409; 186/233; 169/164; 348/159. Average values with±s.e.m. are shown. Data are from three independent experiments. (c) Representative time-lapse DIC images of cells used for quantification of (b). Scale bars, 50 μm. (d) Representative cell tracks of cells used in (b) with all starting points shifted to the center. (e) Representative FRET images from fixed Jurkat T cells. Scale bar, 5 μm. Images are from 4 to 5 independent experiments. (f) Quantification of whole-cell FRET levels in fixed Jurkat T cells that were migrating on either ICAM-1 or fibronectin. Values are shown as mean +−s.e.m. Kruskal–Wallis with Dunn's multiple comparison test for β2-cTerm on ICAM-1 and the others yielded P-values: ***P<0.001, **P<0.01, *P<0.05. No significant differences found when comparing cells on fibronectin.

Figure 3. Integrin tension depends on interaction with specific ligand.

(a) Representative FRET images over 60 s of migrating Jurkat T cells expressing indicated tension sensor constructs. Images are corrected for photobleaching. Images are from three independent experiments. (b) Whole-cell FRET levels in migrating Jurkat T cells with 100 ng ml⁻¹ SDF-1α. Circles represent individual cells from three independent experiments with median shown as a line. Kruskal–Wallis with Dunn's multiple comparison test of differences between indicated conditions yielded P-values: ****P<0.0001, ***P≤0.001, **P≤0.01. (c) Representative images of focal plane in TIRF z-stacks (50 nm step) of Lifeact-mCherry and TS-mVenus with distribution of fluorescence variance and Gaussian fits with indicated relative difference of the whole-cell distribution. The dotted lines across the submicron bead (100 nm) images are shown as side views at the top. Scale bar, 5 μm. (d) Relative differences of integrin z center to actin after chromatic aberration correction (N=15,13,15,16,20,22,16,16,11). Box plots show the 5–95 percentile range (whiskers) of observations with median as line, mean as plus sign, and 25–75 percentile range boxed. Mean±s.e.m. from left: −42.9±12.4, −40.8±11.4, −57.2±12.7, 121.3±24.0, 114.5±17.5, 102.8±20.33, 130.8±20.3, 88.1±23.7, 120.5±32.2. Kruskal–Wallis with Dunn's multiple comparison test of differences between indicated conditions yielded P-values: ****P<0.0001, ***P≤0.001, **P≤0.01.

Figure 4. Integrin tension depends on both NPxF motifs.

(a) Representative FRET images taken from movies of migrating Jurkat T cells. Images are corrected for photobleaching. Images are from three independent experiments. (b) Whole-cell FRET levels in migrating Jurkat T cells with 100 ng ml⁻¹ SDF-1α. Individual cells from three independent experiments are shown as circles with median as line. Kruskal–Wallis with Dunn's multiple comparison test of differences between β2-TS5 and the other conditions all had P values<0.0001. (c) Average FRET efficiency of migrating cells treated with 250 nM cytochalasin D at time zero. Each point indicates one cell, imaged only once to avoid complications from photobleaching. Data from two independent experiments are fit to single exponentials. (d) Representative images of focal plane in TIRF z-stacks (50 nm step) of Lifeact-mCherry and TS-mVenus with distribution of fluorescence variance and Gaussian fits with indicated relative difference of the whole-cell distribution. Scale bar, 5 μm. (e) Relative differences of integrin z center to actin after chromatic aberration correction (N=13, 16 and 11). Box plots show the 5–95 percentile range (whiskers) of observations with median as line, mean as plus sign, and 25–75 percentile range boxed. Mean±s.e.m. from left: −68.0±17.1, 161.8±28.9, 84.1±20.5. Kruskal–Wallis with Dunn's multiple comparison test of differences between indicated conditions yielded P values: P<0.0001 and P=0.0013.

Figure 5. Integrin tension correlates with active conformational state.

(a) Integrin conformational states showing binding sites for antibodies used. (b) Representative TIRF images of migrating Jurkat T cells on ICAM-1 expressing indicated tension sensor constructs and stained with indicated antibodies. Scale bar, 5 μm. Images are from five independent experiments. (c–e) Correlation of antibody fluorescence intensity with inverted FRET ratio. Circles represent Pearson's correlation coefficient from individual cells from five independent experiments with median shown as a line. Mean correlation values from left, TS1/18: −0.017, 0.001, −0.059, −0.032; KIM127: −0.304, 0.061, 0.057, −0.010; m24: −0.127, −0.004, −0.025, −0.039. Kruskal–Wallis with Dunn's multiple comparison test of differences between β2-TS5 and other constructs were not significant for TS1/18, all P values <0.0001 for KIM127, and for m24 the P values were P=0.0065, P<0.0001 and P=0.0062.

Figure 6. Strong actin retrograde flow in conditions with low integrin tension.

(a) TIRF images of Lifeact-mCherry in migrating Jurkat T cells with 100 ng ml⁻¹ SDF-1α. Red lines show regions from which kymographs (right) were generated. Images are from 3 to 6 independent experiments. Scale bar, 5 μm. (b,c) Actin retrograde flow and protrusion velocities from kymographs. Gaussian fits to the data are displayed (N=64, 55, 72, 68, 60, 45, 38, 72, 42 kymographs analysed from 15–25 cells per condition). Values are shown as mean±s.d. Kruskal–Wallis with Dunn's multiple comparison test of differences between WT cells and other conditions yielded P values: ****P<0.0001. (d) Representative FRET (colour) and actin (grey scale) images before start of high-speed actin movies quantified in b. The arrows indicate the direction of cell movement. Scale bar, 5 μm. Images are from three independent experiments. (e) FRET and force versus actin retrograde flow. FRET data from Figs 3b and 4b are shown with retrograde flow data from d as mean values±s.d.

Figure 7. Protrusion can be powered by low force on integrins.

(a) Representative FRET and actin image series on migrating Jurkat T cell taken every 3 s (from >10 independent experiments). Arrowheads point to high force regions that correspond with protrusion in the following frame. Scale bar, 5 μm. (b) Image series from morphodynamics analysis. Each window (gray polygons) is tracked over time. Red arrows show edge velocity vectors. Force scale as in a. (c,d) Example of morphodynamics analysis of protrusion activity (c) and FRET (d). The windows are organized in bands along the cell periphery, with band 1 being all windows along the cell edge, band 2 being the row behind, and so on. Window 55 is near the front of the migrating cell and windows 1 and 110 are at the cell rear. (e) Correlation of instantaneous force versus protrusion over time in TS5 cells. In band 1 there is a weak but significant inverse correlation at zero lag. (f) Correlation of force vs protrusion localization. By averaging windows across the entire movie, a peak correlation value is found at 2.4 μm (band 3) from protruding edge. (g) Distribution of mean force vs mean cell edge velocity over time with Pearson correlation (r and P value shown) and linear regression plotted as solid line with 95% confidence interval shown as dashed lines. N=324 protrusion events. (h) Leading edge quantification. The average force at the leading edge of migrating cells was measured to be 1.5 pN for TS5 cells (N=122) and 0 pN for cTerm cells (N=105). A two-tailed Mann–Whitney test yielded a P value<0.0001. (i) Representative images from migrating TS5 cells showing peak force localization behind the leading edge (from >10 independent experiments). Force scale as in a. Arrows indicate edge movement in next frame.

Figure 8. Model for actin-dependent integrin activation in cell migration.

(1) Inactive LFA-1 is neither bound to ICAM-1 on the extracellular side nor linked to actin on the cytoplasmic side. (2) Talin and kindlin link actin to the cytoplasmic tail of the β2 subunit. (3) When LFA-1 binds ICAM-1 and the actin cytoskeleton simultaneously, tensile force exerted across LFA-1 stabilizes the high-affinity conformation of the αI domain bound through its internal ligand to the βI domain with the open headpiece conformation and the hybrid domain swung out. (4–5) When the intracellular link to actin is lost, the open headpiece conformation is no longer stabilized, and the integrin returns to its lower energy, bent-closed conformation. The stability of the states is likely more relevant than the order of binding and conformational change. Previous work has shown that LFA-1 has 1,000-fold lower affinity for ICAM-1 in the bent-closed and extended-closed conformations than in the extended-open conformation. The bent-closed conformation is more populated on the cell surface than the other states under resting conditions, but rapid kinetic exchange between the three conformational states at thermal equilibrium likely results in significant population of extended-closed and extended-open states in basal conditions. Under basal conditions, individual integrin molecules may exchange thousands of times per second between conformational states, enabling them to sample CT binding to adapters and ectodomain binding to ICAM-1. These processes would enable a pre-existing extended integrin conformation to bind ICAM-1 and actin adaptors at the same time. Once pulling on the integrin by the cytoskeleton meets resistance from ICAM-1 anchored on the surface of another cell, tensile force exerted on the integrin between the ligand anchor point and polymerized actin stabilizes the open, high-affinity integrin conformation. The molecules in the model are not drawn to scale.