Dynamic force-induced direct dissociation of protein complexes in a nuclear body in living cells

A local surface force directly dissociates coilin from SMN

To investigate whether a surface force applied via integrins could directly deform nuclear proteins in the CBs, we transiently co-transfected HeLa cells with two major essential CB components critical for its structural integrity, CFP-SMN and YFP-coilin¹⁶ (Fig. 1a). We quantified their displacements that were synchronized with the oscillatory loading applied via an RGD-coated magnetic bead (Fig. 1b). We employed an established synchronous detection approach that could detect external-stress-induced nanoscale displacements at ~4 nm resolution while filtering out spontaneous movements^{23, 24}. Both CFP-SMN and YFP-coilin in the same CB were directly displaced by the external dynamic force applied on the cell surface but SMN exhibited greater displacements in response to the same force (Fig. 1c), suggesting that differential displacements of coilin and SMN might lead to dissociation of the protein-protein complex, which is known to interact with each other inside the CB²⁵. To further determine if the applied surface force had any direct impact on the dissociation of the coilin from the SMN in the CB, we applied a quick instant force (a step-function) with the magnetic bead via the integrins and quantified the FRET changes of the CFP-SMN (the donor) and YFP-coilin (the receptor) protein pairs. The FRET ratio of CFP-SMN to YFP-coilin rapidly increased within 350 ms in response to the applied stress (Fig. 1d). The emission intensity of the protein pairs was anti-correlated: CFP-SMN intensity increased while YFP-coilin intensity decreased (Fig. 1e). This is a strong evidence for a FRET change, suggesting that SMN were dissociated away from coilin by force. Varying force magnitudes revealed that the threshold stress for the FRET change was ~14–17 Pa and that the FRET change reached a plateau by ~350 ms (Fig. 1f). Turning off the applied force did not return the FRET ratio of CFP-SMN to YFP-coilin back to the pre-force level, suggesting that the force-induced structural changes to the protein pairs were “plastic” or “permanent”, which lasted at least 10 s (Fig. 1g). To further examine the plasticity of dissociation of the protein complex, we quantified mean square displacements (MSDs) of both proteins before, during and after a cyclic stress application. Both coilin and SMN inside the same CB exhibited constrained diffusion behaviors and had almost identical spontaneous MSDs before stress application (Fig. 1h), suggesting that they were tightly associated and moved as a single entity. Both responded synchronously to the cyclic stress (Fig. 1h), suggesting that the proteins were directly displaced by the applied stress, but coilin had much greater spontaneous MSDs than SMN after stress application (Fig. 1h), indicating that coilin was tethered less tightly than SMN after stress application. Taken together, these data strongly suggest that an external force on the cell surface induced direct differential displacements between SMN and coilin, which resulted in the increased distances and thus the FRET change between these two proteins.

We next examined if force-induced displacements of coilin or SMN in the nucleus were dependent on the distance between the site of force application and the CB. The magnitudes of SMN or coilin displacements decreased only slowly as the distance increased (Supplementary Information, Fig. S1); there were significant force-induced displacements of both proteins even when the local force (the bead) was placed >27 µm away from the CB, suggesting the force transmission pathway from the cell surface to the CB was concentrated along the force-bearing cytoskeletal-nuclear structures and that stress field did not follow the decay pattern from the homogeneous material prediction²⁴.

Structural basis for dissociation of coilin from SMN by force

To further investigate the structural basis of stress transmission into the nucleus, we quantified changes of FRET ratio of CFP-SMN to YFP-coilin when different cytoskeletal or nuclear structural components were specifically disrupted or knocked down. Treatments of Latrunculin A, an F-actin disrupter, or blebbistatin, a myosin II inhibitor, but not colchicine, a microtubule-disrupting agent, prevented the force-induced FRET changes (Fig. 2a; Supplementary Information, Fig. S2), suggesting that intact F-actin and myosin II, but not microtubules, are indispensable for force transmission from the cell surface to the CB in the nucleus. Knocking out Lamin A/C, a critical nuclear filament underneath the nuclear envelope, also abolished force-induced FRET changes (Fig. 2a), indicating the essential role of Lamin A/C in force transmission to the CB. Consistent with the necessary role of Lamin A/C in force transmission into the nucleus, using mCherry-Lamin A as a marker of the nuclear envelope, we observed deformation of the nuclear envelope in response to the local force on the cell surface (Fig. 2b, 2c; Supplementary Movie 1). Knocking out plectin, a linking protein connecting intermediate filaments to the nuclear envelope, did not have any effect on stress-induced FRET change (Fig. 2a; Supplementary Information, Fig. S2), suggesting that plectin does not play an essential role in mediating force to the CB. Taken together, these results suggest that the applied stress was directly transmitted into the CB across the nuclear envelope via F-actin-LINC (linker of nucleoskeleton and cytoskeleton)-nuclear lamina structural pathways.

It is known that substrate rigidity has profound influence on cellular functions²⁶. Force-induced FRET changes were observed in HeLa cells plated on 2 kPa and 8 kPa substrates, similar to those FRET changes observed in cells plated on rigid dishes, but not in cells plated on 0.6 kPa substrates (Fig. 2d). HeLa cells have a maximum stiffness (cell stiffness measured on a rigid dish) of ~1.51 kPa (Fig. 2e) and express ample actin bundles and high tractions on 2 and 8 kPa but not on 0.6 kPa substrate (Fig. 2f, 2g). Together with the above data after F-actin disruption and myosin II inhibition, these data suggest that substrate rigidity regulates force-induced nuclear protein dissociation via controlling tension of the actin bundles in the cytoskeleton for long distance force propagation.

Other protein pairs exhibit different force-induced changes

Next, we analyzed how the local surface force affects interactions between the major interacting partners of coilin and SMN in CBs. Coilin and SMN are self-interacting proteins and their interaction is mediated by WRAP53 which is also able to self-interact²⁷. Importantly, both coilin and SMN interact with spliceosomal snRNPs²⁸. In addition, coilin interacts with the nucleolar chaperone Nopp140 on its N-terminus²⁹ and with the U4/U6 snRNP assembly factor SART3³⁰. In the absence of external stress, spliceosomal snRNP core SmE-SmG proteins had lowest baseline CFP/YFP emission ratios (~0.3), followed by coilin-coilin, coilin-SmE, coilin-Nopp140, WRAP53-WRAP53, coilin-WARP53, coilin-SART3, SMN-WRAP53, SMN-coilin, SMN-SMN, SMN-SmG, and SMN-SART3 (~0.8) (Fig. 3a), suggesting that baseline distances between two proteins were closest for snRNP proteins of the Sm ring SmE-SmG and farthest for SMN-SART3 complexes. Interestingly, CFP-fibrillarin (a methylase that binds to SMN in the CB³¹) and YFP-coilin in the CB exhibited even higher baseline emission ratios than SMN-SmG, reaching almost 1.0 (Fig. 3a), consistent with previous published data that fibrillarin and coilin do not directly interact with each other¹⁴ and that they might be far apart (~>10 nm) (a negative control). After a quick step-function force was applied, coilin-coilin self-interaction exhibited greatest FRET ratio increases from baseline values, followed by coilin-SmE, WRAP53-WRAP53, coilin-Nopp140, SMN-coilin, coilin-SART3, coilin-WARP53, SMN-SMN, SMN-SmG, and SMN-WRAP53 (Fig. 3b, Supplementary Information, Fig. S3, S4). These results suggest that when these protein complexes were exposed to the same magnitudes of stress, their respective FRET ratio changes were in general inversely proportional to their baseline distances between the two proteins, except for SMN-coilin, which had larger FRET ratio changes than SMN-WRAP53, suggesting that distances between the SMN-coilin protein pairs increased more than predicted from their baseline values. Fibrillarin-coilin and SMN-SART3 pairs did not exhibit any FRET change from the baseline (Fig. 3b, Supplementary Information, Fig. S3, S4), consistent with the baseline data that fibrillarin–coilin and SMN-SART3 are already too far apart at baseline so that they are beyond the detection range of FRET changes. In contrast, SmE and SmG proteins strongly interact within the heteroheptameric ring of the spliceosomal snRNP core structure. Thus, it is likely that they could not be dissociated by the magnitude of the applied stress and thus did not exhibit FRET ratio change upon stress (Fig. 3b).

Dissipative behaviors of protein complexes in the Cajal body

In a time-dependent force-deformation response, a phase lag (or a time delay) between the applied cyclic stress at the cell surface and the displacement of a nuclear protein is a measure of rheological properties of the material. Any viscoelastic linkage has a phase lag of between 0 ° and 90°. For multiple linkages of different proteins in a complex protein aggregate that is tethered in a nucleoplasmic matrix, the phase lags would indicate how viscous these interactions are. Under an oscillatory stress of 0.3 Hz, the phase lag of coilin to the magnetic bead displacement (a measure of cytoplasmic strain, which, in turn, was ~5° lagged behind the applied force, see Fig. 1b) was ~33°, 4° more than that of SMN (~29°) inside the same CB (Fig. 3c). The fact that coilin was always 4° lagged behind SMN, irrespective of which fluorescent probe (CFP or YFP) was attached to coilin (Fig. 3d), suggests that these phase lags were not biased by the attached fluorescent probes. The phase lag differences between CFP-coilin and YFP-coilin or CFP-SMN and YFP-SMN were almost 0° (Fig. 3d), suggesting that the 4°-phase difference did not originate from CFPs or YFPs themselves. The 4°-phase difference suggests that the SMN-coilin interactions are viscoelastic and dissipative. The phase lag differences between coilin and WRAP53, between SMN and WRAP53, between coilin and Nopp140, between coilin and SmE were all <2°, suggesting that the dissipative elements (where the input mechanical energy is dissipated into heat) between them were not as significant. SMN-SmG and coilin-SART3 pairs have significant phase lag differences between each protein pair (~2–3°, corresponding to a time delay of 20–30 ms (3° times 3200 ms divided by 360° equals 26.7 ms; one cycle is 3200 ms at a loading frequency of 0.3125 Hz)), indicating that force is dissipated when transmitted from one protein to the other. Interestingly, viscous responses were also evident when a sudden stress (a step function) is applied to the cell surface. The FRET curves exhibited creep responses showing local “material deformation” (two proteins pulled apart): slow responses (creep time, defined as time to reach peak FRET change, ~700–1350 ms) indicate a predominantly viscous behavior (Fig. 4a, b, c, d); in contrast, fast responses (creep time ~300 ms, confirmed by sampling at a time resolution of ~150 ms in Fig. 4e) are likely to be associated with more elastic local nanoenvironment (Fig. 4f, g, h, i, j, k).

Taking together the FRET changes in an intact cell and in a cell where specific components of the cytoskeleton and of the nuclear skeleton are disrupted or knocked out, we propose a working model of the force response for the protein complex in the CB (Fig 5): the protein complex is stressed via the pathway of integrin-mediated focal adhesion to F-actin, to LINC, to nuclear lamina, to a putative nucleoplasmic element (likely chromatin and/or an actin-dependent filamentous network³²), and then to the CB. Interestingly, at a higher loading frequency of 0.83 Hz, coilin still phase-lagged SMN by ~4°, although the magnitudes of phase lags of both proteins decreased by ~4° (to ~29° for coilin and ~25° for SMN) (Fig. 6a). These frequency-dependent data suggest that a more dynamic force resulted in a more elastic cytoskeleton and/or nucleoskeleton (lower phase lags or less dissipative energy).

Cell culture and transfection

HeLa cells, LMNA-knockout mouse embryonic fibroblast cells, and plectin knockout mouse skin fibroblasts were cultured and maintained in DMEM (Sigma) supplemented with 10% FBS (HyClone), 100U/ml penicillin, 100µg/ml streptomycin, and 2mM L-Glutamine at 37°C in 5% CO₂. For live cells imaging, cells were plated on glass bottom dishes coated with type I collagen (40 µg/ml) overnight. Each fluorescence protein plasmid of interest was transfected at 1 µg per 35mm glass bottom dish using Lipofectamine 2000 twenty hours prior to experiment, following protocols provided by the manufacturer (Invitrogen). The plasmids CFP-SMN^{14, 46}, YFP-SMN^{14, 46}, CFP-coilin⁴⁷, YFP-Coilin⁴⁷, fibrillarin-YFP⁴⁸ constructs are described as referenced. YFP-Nopp140, CFP-WRAP53, YFP-WRAP53, YFP-SART3, CFP-SmE, YFP-SmE, CFP-SmG, YFP-SmG were subcloned into either pECFP-C1 or pEYFP-C1 expression vectors as indicated. mCherry-Lamin A plasmid was a gift from Dr. P. Kalab. The cell culture media was then changed 6 hours after transfection to serum-deprived-media.

Magnetic twisting cytometry, cell stiffness, and tractions

The optical magnetic twisting cytometry method^49,50 was used to exert mechanical stress on the living HeLa cells. Arg-Gly-Asp (RGD)-coated ferromagnetic beads (Fe₃O₄, ~4 µm in diameter) were added to the cells and then incubated for 10 minutes to allow for integrin clustering and formation of focal adhesions surrounding the bead. The beads were magnetized in the horizontal direction by a strong magnetic pulse (1,000 G) for a brief period (<100 µs). During imaging, cells were maintained in CO₂ independent medium without serum (Invitrogen) at 37°C. Cell Young’s modulus in Pa was estimated by quantifying the embedded area of the magnetic bead and using the cell stiffness data (in Pa/nm)⁵¹. Fluorescent beads (0.2-µm in diameter) were embedded in the polyacrylamide gel substrate and their positions were recorded by fluorescence microscopy. The displacement field induced by each individual cell’s tractional forces was determined by comparing the fluorescent bead positions before and after trypsinization (cell-free and thus force-free). The RMS (RMS=root-mean-square) traction was calculated from the displacement field with known substrate stiffness⁵². Cell tractions were quantified after the cells were plated on various stiffnesses (0.6, 2, and 8 kP, corresponding to 0.06% bisacrylamide and 3% acrylamide, 0.05% bisacrylamide and 5% acrylamide, 0.3% bisacrylamide and 5% acrylamide respectively) of collagen-1 coated polyacrylamide gels⁵².

For FRET imaging, the magnetic twisting field (0, 1.8, 3.5, 8.8, 14, 17.5, or 24.5 Pa) was applied as a step function (a constant magnetic field) to the magnetic beads that are bound to the surface of the cells. A Leica inverted microscope was integrated with a magnetic twisting device and a Dual-View system (Optical Insights) to simultaneously acquire both CFP and YFP emission images in response to stress. CFP/YFP Dual EX/EM (FRET) (OI-04-SEX2) has the following filter sets: CFP: excitation, S430/25, emission S470/30; YFP: excitation, S500/20, emission S535/30. The emission filter set uses a 515 nm dichroic mirror to split the two emission images. For FRET imaging, we excited at CFP wavelength and measured the emission of both CFP and YFP simultaneously on the same screen using a CCD camera (Hamamatsu C4742-95-12ERG) and a 63x 1.32 N.A. oil-immersion objective. Exposure time was 290 ms or 150 ms.

For Cajal Body (CB) dynamics imaging, an oscillatory magnetic field (24.5 Pa at 0.3 or 0.83Hz) was applied. CFP-SMN and YFP-Coilin images before, during, and after mechanical loading were captured at every 0.32 or 0.09 seconds interval over a period of ~15 s respectively. Bright-field images of the cell and the magnetic bead during oscillatory load were also obtained for bead displacement analysis.

To minimize photobleaching, the cells were treated with 0.5mM Trolox (6-Hydroxy-2,5,7,8-tetramethylchroman-2-carboxylic acid; Sigma) 18 hrs⁵³ before imaging and illuminated with 12% light intensity by using a 120 W Hg lamp (Leica).

Statistical Analysis

A two-tailed Student’s t-test was used for all statistics.

Technical issues related to FRET analyses

Are there potential spontaneous movement artifacts caused by moving structures inside the nucleus? It is known that the largest speed of Cajal body movement can reach ~0.9 µm/min (15 nm/s)⁴¹, which may account for the observed spontaneous movements of the CBs in our study. However, we have observed that in spontaneous movements of CBs before force application there is no separation of coilin from SMN, i.e., both proteins move at the same rate and thus there is no FRET change of the protein complex. More importantly, the anti-correlation behavior between CFP-SMN (increase in emission) and YFP-coilin (decrease in emission) and differential displacements between SMN and coilin in response to force provide a strong evidence that the ratio changes are force-induced FRET changes and not movement artifacts. In addition, we excited at CFP wavelength and simultaneously measured emissions from both CFP and YFP without any time delay. The CFP-SMN transfection (without transfection of YFP-coilin) and excitation yielded only a minimal leak through the YFP channel emission. Therefore the CFP/YFP ratio change is a measure of FRET change which is due to association or dissociation of one protein from the other by force. Since we compared this FRET change in the absence and the presence of applied stress, any bleaching and/or spontaneous FRET changes are accounted for or normalized.

Image Analysis

A customized Matlab (Mathworks) program was used to obtain CFP/YFP emission ratio and CB displacement map. For CFP/YFP emission ratio analysis, CFP and YFP images at each time point were first background-substracted, and aligned pixel-by-pixel by maximizing the normalized cross-correlation coefficient of CFP and YFP images. YFP image was then thresholded to generate a binary mask so that the pixel value inside the fluorescent particle was set to 1, and that outside the fluorescent particle was set to 0. Aligned CFP/YFP emission ratios were displayed as a linear pseudocolor. To obtain the displacement field of Coilin and SMN, CFP and YFP images were divided into small arrays of 11 × 11 pixels that overlap by five pixels. The corresponding arrays at the same location between two adjacent images taken at different phases during the twisting cycle were compared in the Fourier domain. This sensitive method can detect displacements of the fluorescent particles to the resolution of 4–5 nm⁵⁴.

For Cajal body dynamics analysis, a custom matlab program was developed by converting the florescent image into black and white. The centroid coordinates of each fluorescence particle obtained from the binary image were used to calculate the mean square displacement (MSD) of coilin and SMN. The MSD before, during, and after mechanical loading were calculated based on equation (1)

where r(t) is the fluorescence particle position and Δt is the time step (0.32s in our case). The same procedure is performed on bright-field images to obtain the magnetic bead MSD.