Clustering of cortical dynein regulates the mechanics of spindle orientation in human mitotic cells

Abstract

The forces which orient the spindle in human cells remain poorly understood due to a lack of direct mechanical measurements in mammalian systems. We use magnetic tweezers to measure the force on human mitotic spindles. Combining the spindle's measured resistance to rotation, the speed it rotates after laser ablating astral microtubules, and estimates of the number of ablated microtubules reveals that each microtubule contacting the cell cortex is subject to ~1 pN of pulling force, suggesting that each is pulled on by an individual dynein motor. We find that the concentration of dynein at the cell cortex and extent of dynein clustering are key determinants of the spindle's resistance to rotation, with little contribution from cytoplasmic viscosity, which we explain using a biophysically based mathematical model. This work reveals how pulling forces on astral microtubules determine the mechanics of spindle orientation and demonstrates the central role of cortical dynein clustering.

Keywords: Mitosis; clustering; dynein; forces; mechanics; metaphase; microtubules; orientation; spindle.

Fig 1.. Magnetic tweezers exert calibrated forces directly on the spindle.

(A) Experimental setup: an intracellular magnetic bead pushes directly on the mitotic spindle with force magnitude F_mag and angle relative to the spindle φ. Brightfield and fluorescence images of tweezer tip next to U2OS cell (GFP-hCentrin2, green; GFP-CENP-A, green; mCherry-alpha Tubulin, pink) with magnetic bead (647-N biotin, cyan). (All scale bars in this figure are 10 μm. Shaded error bars are SEM.) (B) The spindle rotates in response to applied force. Spindle length (L_sp) and axis angle (θ) marked in panels ii and iii, respectively. Initial (grey) and final (blue) axis angles marked. (C) Spindle axis angle and calibrated force magnitude of B (matching timepoints circled) with angular velocity (ω) and tweezer force magnitude $\left(\Vert {\overrightarrow{ℱ}}_{mag}\Vert \right)$ shown. (D) Average spindle rotation response after normalizing to force. The spindle’s resistance to rotation, H_WT, is calculated using ω fit to 0–0.4 min (n=14).

Fig 2.. Cytoplasmic viscosity does not determine the spindle’s resistance to rotation.

(A) Experimental setup using the magnetic tweezer system and viscosity calculation. (B) A bead (cyan, 647-N biotin) moves in a U2OS metaphase cell (initial position, white). Scale bar is 10 μm. (C) Bead movement and tweezer force from B (matching timepoints circled). (D) The predicted rotational drag coefficient from cytoplasmic fluid drag alone, H_cyto, is lower than the measured rotational drag coefficient, H_WT. (E) The cytoplasmic viscosity of WT U2OS cells (n=65) is increased ~3-fold in ΔCTDNEP1 U2OS cells (n=39, p<0.0001, ΔCTDNEP1 data also presented in Merta et al., 2021). (F) There is no significant difference between the rotational drag in WT and ΔCTDNEP1 (n=10, p=0.9) cells, and both measurements are lower than the predicted coefficient in ΔCTDNEP1 due to fluid drag.

Fig 3.. Astral pulling forces strongly contribute to the spindle’s resistance to rotation.

(A) Laser ablation removed one quadrant of astral microtubules using a curved plane (2 μm radius, 5 μm height, 0–60° off spindle axis). (B) Spindles rotated after astral microtubules were cut (initial angle in grey, final in blue, cut in pink). All scale bars in this figure are 10 μm. (C) Spindle axis angle after cut (matching timepoints circled). (D) Astral cuts in WT cells (n=31) result in strong rotation away from cut site, compared to no rotation after cuts in the cytoplasm (n=13). (Cut locations and angle sign convention shown in inset) (E) EB1-GFP comets at the midline and cortex (Fig S3E) were used to determine the number of microtubules in contact with the cortex. (F) We calculate the force exerted by each astral microtubule by combining the spindle’s rotational resistance from magnetic tweezer experiments, the spindle’s rotational velocity after astral ablation from laser ablation experiments, and the number of ablated astral microtubules from EB1 measurements. (G) WT U2OS cells treated with LGN siRNA had no midline dynein signal (tdTomato-DYNH2), as measured by integrated dynein intensity I_DYN (WT 5.26 ± 0.85, n=30; LGN 2.26 ± 0.49, n=24, p=0.007). (H) U2OS cells treated with LGN siRNA also did not rotate after astral microtubule ablation (n=15), in contrast to WT (n=31). (I) LGN RNAi treatment (n=10) resulted in more spindle rotation during magnetic tweezer experiments. (J) LGN siRNA treatment significantly reduced the spindle’s rotational drag coefficient by almost a factor of 5 (p=0.0004). (K) LGN siRNA treatment removes cortical dynein and reduces cortically-generated force.

Fig 4.. Changing the amount and extent of clustering of dynein at the cortex changes the spindle’s resistance to rotation.

(A) siRNA treatments significantly changed the spindle’s resistance to rotation. RNAi of gα decreased the spindle’s rotational drag (n=8, p=0.035), as did RNAi of DLG1 (n=12, p=0.033). MARK2 siRNA treatment significantly increased the spindle’s rotational drag (n=9, p=0.007). (B) All siRNA conditions showed decreased rotation after astral microtubule ablation compared to WT (WT n = 31, gα n=10, DLG1 n=13, MARK2 n=18). (C) Cortical intensity line scans of WT U2OS cells expressing GFP-LGN show a midline crescent phenotype (n=56) with an integrated LGN intensity I_LGN of 16.06 ± 2.09. Scale bars in figure are 10 μm unless otherwise noted. (D) Imaging of the cortical coverslip plane of WT U2OS cells expressing GFP-LGN shows spots of LGN. Scale bar is 5 μm. (E) We quantified the intensity of LGN cortical spots relative to background, and the average distance to the five closest neighbors. (F) Cortical dynein and LGN signals in U2OS cells treated with gα siRNA. (G) gα RNAi did not affect cortical dynein localization or intensity (I_DYN = 7.44 ± 0.9, p=0.087 (n=26). (H) gα RNAi resulted in uniform, decreased cortical distribution of LGN in the midline plane with intensity I_LGN = 1.12 ± 0.59, p<1e-5 (n=48). (I) These results suggest that gα RNAi cold affect the activation state of dynein through LGN but does not change the midline crescent phenotype. (J) Cortical dynein and LGN signals in U2OS cells treated with DLG1 siRNA. (K) DLG1 RNAi did not affect cortical dynein localization or intensity (I_DYN = 7.34 ± 1.14, p=0.15) (n=32). (L) LGN midline crescent shape and intensity (I_LGN = 15.05 ± 1.52, p=0.71) were not changed with DLG1 siRNA (n=46). (M) Under DLG1 siRNA treatment, both LGN spot intensity and the distance between spots significantly increased (p=0.013 and p<0.0001). (N) These results suggest that DLG1 siRNA increases LGN clustering. (O) Cortical dynein and LGN signals in U2OS cells treated with MARK2 siRNA. (P) MARK2 RNAi resulted in uniform cortical distribution of dynein (n=70), however there was no change in cortical dynein intensity (I_DYN = 4.75 ± 0.62, p=0.626). (Q) Midline LGN crescents were unchanged with MARK2 siRNA treatment (I_LGN = 19.48 ± 3.58, p=0.38, n=38). (R) MARK2 siRNA treatment resulted in a decreased cortical LGN spot intensity (p<0.00001), but the distance between spots stayed relatively similar (p=0.003). (S) MARK2 siRNA delocalized cortical dynein and decreases dynein clustering within the crescents.

Fig 5.. A three-dimensional mean field shows that changes in dynein clustering are sufficient to explain changes in the spindle’s resistance to rotation.

(A) Overview of geometry and variables for mean field theory showing a cell with radius R, containing microtubules (purple), force generators (orange), and the two centrosomes. (See Supplementary Information) (B) Microtubules nucleate from each centrosome with rate γ, grow at rate V_g, and spontaneously depolymerize with rate λ. (C) Force generators are located at the cell cortex, each covering an area S_i equal to the intersection of the cell boundary with a sphere of radius r. Force generators exert force f₀ on bound microtubules, and microtubules can detach with rate κ. (D) The probability of a force generator being bound is calculated for the entire surface of the spherical cortex. A midline slice at t = 0 seconds (top) shows a symmetrical probability distribution with motors most likely to be bound around the spindle poles. After a 60 second push (bottom, push location denoted by black arrow), the bound motor probability distribution shifts as the spindle moves closer to one part of the cortex. (E) Experimental setup to compare spindle translation and rotation. (F) The theory quantitatively matches the experimental results where pushing at the pole results in a large rotational response and very little translation (n=26). (G) Increasing the number of cortical force generators results in a linear increase of the spindle’s rotational resistance. (H) Increasing the force exerted by individual force generators also results in a linear increase of the spindle’s rotational resistance. (I) Increasing motor cluster size results in a non-linear decrease in the spindle’s rotational resistance, consistent with experimental data. Smaller individual force generators result in larger effects from clustering.

Fig 6.. Dynein clustering and activation determine the force on the spindle.

(A) Varying the number of cortical motors (M) and their clustering (N) is predicted to change the spindle’s rotational drag coefficient. By decreasing motor number from WT, the theory predicts a decrease in rotational drag (grey to green). An increase in clustering (grey to orange) results in a decreased rotational drag coefficient, while a decrease in clustering (grey to purple) results in an increased rotational drag coefficient. (B) Using M=5000 and N=5, we find quantitative agreement between the measured rotational drag and theory prediction (p=0.06). Under LGN siRNA treatment, we assume M drops to 2000 resulting in a predicted rotational drag coefficient of 7.3 pN min μm / degree (experiment: 3.4 ± 0.6 pN min μm / degree) (p=0.000091). Increasing motor clustering to N=20 predicts the decreased rotational drag coefficient in the DLG1 siRNA condition (theory: 5.6 pN min μm / degree, experiment: 4.5 ± 1.5 pN min μm / degree; p=0.08). If MARK2 siRNA results in decreased motor clustering and setting N=1, the theory predicts the resulting increased rotational drag coefficient (theory: 30.3 pN min μm / degree, experiment: 38.0 ± 9.9 pN min μm / degree; p=0.30). (C) We modeled the spindle’s response to changing M and N at the same time. (D, i) U2OS cells treated with both LGN and DLG1 siRNAs showed no difference in the spindle’s rotational drag (1.8 ± 0.6 pN min μm / degree, n=17) compared with RNAi of LGN or DLG1 alone (p=0.22 and p=0.11, respectively). All 3 conditions had rotational drag coefficients significantly decreased compared to WT (LGN/DLG1, p=0.01; LGN, p=0.00041; DLG1 p=0.03). This quantitatively matches the theory prediction of 3.4 pN min μm / degree, if a double siRNA treatment of LGN and DLG1 would result in reduced M and increasing N (M=2000, N=20; p=0.24). (ii) Treatment with both LGN and MARK2 siRNAs rescued the drag coefficient decrease caused by LGN RNAi (p=0.00032) and the increased caused by MARK2 RNAi alone (p=0.011), resulting in a drag coefficient similar to WT (18.6 ± 2.6 pN min μm / degree, n=16, p=0.39). Decreasing both M and N (M=2000, N=1) results in a theoretical prediction of 13.7 pN min μm / degree, quantitatively matching our experimental results and showing the return to a WT phenotype (p=0.009). (E) Cortical force generation is regulated in multiple ways, including the cortical recruitment and clustering of dynein.