Genetically encoded mechano-sensors with versatile readouts and compact size

Abstract

Mechanical forces are critical for virtually all fundamental biological processes, yet quantification of mechanical forces at the molecular scale in vivo remains challenging. Here, we present a new strategy using calibrated coiled coils as genetically encoded, compact, tunable, and modular mechano-sensors to substantially simplify force measurement in vivo, via diverse readouts (luminescence, fluorescence and analytical biochemistry) and instrumentation readily available in biology labs. We demonstrate the broad applicability and ease-of-use of these coiled coil mechano-sensors by measuring forces during cytokinesis (formin Cdc12) and endocytosis (epsin Ent1) in yeast, force distributions in nematode axons (β-spectrin UNC-70), and forces transmitted to the nucleus (mini-nesprin-2G) and within focal adhesions (vinculin) in mammalian cells. We report discoveries in intracellular force transmission that have been elusive to existing tools.

Currently there are two major categories of sensors to measure mechanical forces at the single molecule level for live biological samples. One uses DNA duplexes (hairpins) that rupture at different levels of force^1–3, and the other one utilizes elastic peptides (nano-springs) that monotonically change their extension in a force-dependent manner^4–6 (Fig. 1a–b). Both are tuned to maximize their conformational change to tensile mechanical forces in the pico-newton (pN) range, and the conformational changes are commonly quantified by Förster Resonance Energy Transfer (FRET) measurements^7–10. Variations of mechano-sensors based on these principles have had great successes and yielded valuable insights into the mechanical workings of cell attachment, cell motility, and embryo development^11–16. Nevertheless, both approaches have inherent limitations in their broad applications. For the DNA duplexes, it is easy to tune their mechanical properties and alter readouts^17–21 (Fig. 1a). However, it is difficult to deliver DNA duplexes into cells and protect them from rapid degradation. Therefore, in practice, DNA duplexes are only applied to measure force on proteins with extracellular domains^10,22. Mechano-sensors based on calibrated elastic peptides, on the other hand, can be genetically encoded within any protein of interest, in theory. However, the inclusion of two fluorescent proteins for FRET measurement results in a large sensor size (>55 kDa) that is rarely tolerated by the host protein (Fig. 1b). Additionally, FRET measurements are inherently challenging to perform, requiring extensive controls and calibration^23,24. These measurements are highly context-dependent, time-consuming, and exhibit a low to moderate dynamic range (5–25%) in vivo^23–26. Signals from FRET-based tension sensors are typically averaged over a population of molecules from unknown distribution and quickly drops to a flat value outside of the sensors’ linear range, resulting in convolution of the forces on single molecules. Although newer versions of FRET-based sensors have improved performance by introducing digital force responses and multiplexed pairs^12,13, smaller genetically encoded mechano-sensors that do not rely on FRET measurements are still desired to democratize molecular force measurements in cell and developmental mechanobiology beyond the few commonly studied molecules.

Figure 1. Rational design and calibration of coiled coils as force-sensing modules.

a, A DNA hairpin (TP9) with defined nucleotide composition unfolds at a characteristic force threshold². Fluorescence readouts (such as a fluorophore-quencher pair) can be added to nucleotides by chemical modifications on the DNA hairpin so that change in fluorescence reports the digital opening of the DNA under force. b, An elastic peptide (HP35) sandwiched by two fluorescent proteins (mTurquoise2 and mNeonGreen) form a genetically encoded force sensor^16,30. The FRET efficiency between the two fluorescent proteins depends on the peptide length, which scales linearly with force magnitude within a small range (e.g. 3–6 pN). c, Calibrated coiled coils can be used as force-sensing modules to correlate mechanical force to their conformational change. Readouts are protein constructs that can only bind to the linkers connecting the two α-helices of the coiled-coil when the coiled-coil is unfolded under force. a-c, Molecules are shown at the same scale, and the structures are predicted by AlphaFold3¹⁰³ and colored according to the residues’ b-factors. Arrows indicate the directions of force pulling. d, Cumulative unfolding probability of the coiled coil force sensor ${力}_{7\mathit{\text{pN}}}$ as a function of force. The jump of the unfolding probability from below 0.1 to above 0.9 within 1 pN indicates a near-digital response to mechanical forces. e-i, Predicted structures of ${力}_{5\mathit{\text{pN}}},{力}_{7\mathit{\text{pN}}},{力}_{10\mathit{\text{pN}}},{力}_{11\mathit{\text{pN}}}$ and ${力}_{13\mathit{\text{pN}}}$ with their GS linkers (GGSSGG) highlighted in cyan are shown at the same scale as a-c. Mechanical stabilities are tuned by mutating the hydrophobic core (amino acids in positions $a$ and $d$) and by changing the total number of heptads. See also Fig. S1 for helical wheel depictions. j, Representative force-extension curves (FECs) obtained by pulling (grey) or relaxing (black) coiled coil sensors using optical tweezers. The conformations of the coiled coils are labeled as follows: 1, folded state; 2, unfolded state; and 2*, partially unfolded state. k, Unfolding force thresholds of coiled coils are calibrated by optical traps. Each dot represents a pulling event. l, The selection of functional linkers and their binding partners offers modularity in force sensor readouts. Readouts with slow unbinding kinetics are force recorders (HiBiT and LgBiT; GFP11 and GFP1–10; TEV cleaving site (TEVcs) and TEVp), while readouts with fast unbinding kinetics are force live reporters (IAAL-K3 and IAAL-E3; SsrA and SspB).

We present a new class of mechano-sensors that retains the programmable digital mechanical response and versatile readouts of DNA sensors, while being fully genetically encoded and substantially smaller than the current peptide-based force sensors^24,26 (Fig. 1c). We achieve this goal through a modular approach by building on force sensors we have previously developed^27–29 (Fig. S11). The force-sensing module is composed of a dimeric anti-parallel coiled coil that unfolds (i.e. fully opens) when the tensile force applied to it is larger than a calibrated force threshold. The readout module is bipartite and made of a) a peptide that links the two α-helices of the coiled coil and b) an interacting partner that binds the linker only when the coiled coil is in an open conformation but not when it is in a closed conformation (Fig. 1c). The binding between different linkers and their binding partners can be engineered to generate binary signals that are gated by the force on the coiled coil (Fig. 1l). Due to their small size and short end-to-end distances (~13 Å, or the typical length of a ~5 amino acid linker), the mechano-sensors can be placed between virtually any domains or within virtually any flexible region of a protein of interest with minimal influence on the protein’s function. Our approach extends and generalizes previous analogous approaches based on larger force-sensing modules, including mutants of the HP35 peptide or Titin I10 domain, which have been limited to proteins in focal adhesions (talin and vinculin)^30–32.

We derived our library of force-sensing modules from an artificial heterodimeric antiparallel coiled coil³³, which unfolds at 7.4 ± 0.1 pN (mean ± SEM) (Fig. 1d). Dimeric coiled coils are stabilized by knobs-into-holes packing at the hydrophobic core along with electrostatic pairing on the surface^34,35. We tuned the mechanical stability by 1) varying the amino acids in the hydrophobic core and 2) changing the total number of heptad repeats (Fig. 1e–I, S1a–e). We also used molecular simulations to aid our design (Fig. S1f). The resultant coiled coils have the same end-to-end distance and surface properties when folded, and each sensor reversibly unfolds when the tensile force exerted on it exceeds a predetermined sequence-specific threshold between 5 pN (four heptads) and 13 pN (eight heptads) (Fig. 1j–k). As shown in Fig. 1j, the folded coiled coils start to switch (or flicker at equilibrium) between folded and unfolded conformations when force reaches the threshold (Fig. 1e, state 1 and state 2, respectively), and become completely unfolded when force keeps increasing. The unfolding/refolding transitions are reversible in the relaxation rounds (Fig. 1j, black curves). We took this reversible transition feature as a main criterion to select sensors for further use in live measurements. The lower bound of the unfolding force threshold is set by the minimum number of heptads required to form a coiled coil (3.5 heptads for most known anti-parallel coiled coils)^36,37. Higher force thresholds can be achieved by increasing the number of heptads but with diminishing returns the longer it is extended. The maximum number of heptads is also limited by the increasing likelihood of unfolding hysteresis when the number of heptads exceeds 6 (Fig. 1j, FEC# (6)), suggested by the additional folding intermediate (Fig.1j, state 2*). To detect forces smaller than 5 pN, we engineered the stalk region of mouse cytoplasmic dynein, an anti-parallel coiled coil that unfolds at 3 pN (Fig. 1j, FEC # (1)). Our library of four force-sensing coiled coils (hereafter called ${力}_{\mathit{\text{XpN}}}$, where $力$ (lì) is the Chinese character for “force” and X pN is the opening force threshold) is sufficient to cover the physiological range of most forces on a single molecule with ~2 pN resolution. Note that varying the pulling speed or the replacement of the linkers of similar lengths between the two α-helices has a minimal influence on the unfolding force threshold of the coiled coil, consistent with previous models and measurements^38,39 (Fig. S2).

We developed two categories of readout modules: recorders and live reporters. “Recorders” use bipartite systems that are irreversible (or very slowly reversible), so that they keep emitting signals for a long time after the force-sensing module has unfolded, and therefore “records” (or “integrates”) the maximum force the sensor has experienced in the past (Fig. 1l). Here, we present three recorders: the split-Nanoluc, split-GFP, and TEV systems. The split-Nanoluc system is composed of an 11-aa (amino acid) peptide (HiBit) used as linker between the two α-helices, and the rest of the Nanoluc (LgBit) expressed in the cell^40,41. When force opens the force-sensing element, the exposed HiBit is bound by LgBit, reconstituting a full Nanoluc able to produce luminescence following the addition of a Nanoluc substrate (e.g. furimazine). Similarly, the split-GFP system is composed of GFP11, a 16-aa peptide whose sequence corresponds to the 11^th β-strand of the GFP, used as a linker between the α-helices of the force-sensing module, and GFP1–10, the GFP lacking the 11^th β-strand (not fluorescent by itself), which is expressed in the cell. When force is low, no fluorescence is emitted. Upon opening of the force-sensing element, GFP11 is exposed and binds with GFP1–10, matures, and emits fluorescence⁴². Because the off-rate constant of the split-GFP system is very slow (~hours), fluorescence persists even after the force vanishes⁴³. Since sub-cellular compartments may limit the use of fluorescence or luminescence, we developed an orthogonal detection method based on cleavage by Tobacco Etch Virus (TEV) protease (TEVp). The TEV readout uses as a linker the 7-aa TEV cleavage sequence (TEVcs) which can be cleaved by the TEVp. In a cell expressing the TEVp, the force sensing element is cleaved after it is unfolded by force⁴⁴, and the cleavage can typically be detected using an immunoblot.

“Live reporters” are constituted of a peptide tag used as linker and a complementary peptide or domain fused to a fluorescent protein, which is expressed in the cell (Fig. 1l). Here, we present the reversible binder-tag systems IAAL-E3/IAAL-K3 and SsrA/SspB, an artificial binding pair⁴⁵. Under low force, fluorescence is diffuse, but when high force unfolds the force-sensing element, the fluorescent protein re-localizes to the force sensor. Because the tags used are reversible with a fast off rate constant (see in vivo binding kinetics in Fig. 3j), when force is released, the fluorescent protein detaches, and fluorescence becomes diffuse again⁴⁶. We summarized the strengths and weaknesses of each readout in Table 1, and simulated the kinetics of reporter binding in Note S1. Hereafter, the force-sensing coiled coils are called ${力}_{\mathit{\text{XpN}}}^Y$, where X is the force threshold of the sensor and Y is the linker specific to the readout used to detect its opening (e.g. ${力}_{3\mathit{\text{pN}}}^{\mathit{\text{GFP}}11}$ is the force sensor that opens under forces larger than 3 pN with the split-GFP readout)

Figure 3. Comprehensive force measurements on vinculin.

a, Vinculin binds to talin and F-actin in focal adhesions and contributes to force transmission between the extracellular matrix and the cytoskeleton. b, Quantification of force on vinculin with the Split-GFP readout. U2OS cells were plated on glass and transfected with plasmids encoding mCherry-vinculin with different force sensors inserted after E883 using the split-GFP readout. GFP fluorescence is normalized to mCherry-vinculin signals. Each dot represents measurement from a segmented focal adhesion. Data are pooled from two independent experiments. Ordinary one-way ANOVA was performed with Tukey’s multiple comparison tests with “No-GFP11” as the control. ****, p<0.0001. c, Force distribution of vinculin, calculated by normalizing the fluorescence signals from ${力}_{5\mathit{\text{pN}}}^{\mathit{\text{GFP}}11},{力}_{7\mathit{\text{pN}}}^{\mathit{\text{GFP}}11}$, and ${力}_{10\mathit{\text{pN}}}^{\mathit{\text{GFP}}11}$ to that from ${力}_{3\mathit{\text{pN}}}^{\mathit{\text{GFP}}11}$. Mean ± SEM. d, Schematic for comparing the forces on vinculin from cells plated on glass or Matrigel for 6 hours. e, Quantification of force on vinculin with the Split-Nanoluc readout. Luminescence is normalized to mCherry-vinculin signals to compare the number of mechanically active vinculin molecules. Multiple comparisons with Kruskal-Wallis test. *, p<0.05. f, Force distribution on vinculin, calculated by normalizing the luminescence signals from different sensors to that from ${力}_{3\mathit{\text{pN}}}^{\mathit{\text{GFP}}11}$. Multiple comparisons with Kruskal-Wallis test. *, p<0.05. g, U2OS cells were transfected with force sensors, collected, and plated onto substrates with different stiffnesses. Luminescence was measured to detect the forces on vinculin after overnight culture. h, Quantification of force on vinculin across substrate stiffness. Data are shown as mean ± SEM with fitted lines of linear regression. The slopes of the fitted lines are not significantly different from zero (F-test). i, Force distribution on vinculin calculated by normalization to ${力}_{3\mathit{\text{pN}}}^{\mathit{\text{GFP}}11}$. Data are shown as mean ± SEM with fitted lines of linear regression. The fraction of forces on vinculin was not changed by the substrate stiffness according to the F-test on the slopes of the fitted lines. j, U2OS cells were transfected by plasmids encoding mCherry-vinculin-E883-${力}_{3\mathit{\text{pN}}}^{\mathit{\text{IAAL}}-K3}$ and mEGFP-IAAL-E3. FRAP was performed to compare the recovery of mEGFP and mCherry signals in focal adhesions. Scale bar: 5 μm. k, Quantification of fluorescence intensities after photobleaching. mEGFP signals recover with ${\mathrm{t}}_{1/2}=1.6\mathrm{s}$, indicating fast binding and unbinding kinetics for the live force reporter. l, U2OS cells were transfected with vinculin-E883-${力}_{3\mathit{\text{pN}}}^{\mathit{\text{IAAL}}-K3}$ with mScarlet-3-IAAL-E3 binder, and vinculin-E883-${力}_{5\mathit{\text{pN}}}^{\mathit{\text{SsrA}}}$ with TagBFP-SspB binder. Signals corresponding to ${力}_{3\mathit{\text{pN}}}^{\mathit{\text{IAAL}}-K3}$ are closer to the cell edge and stronger in focal adhesions than ${力}_{5\mathit{\text{pN}}}^{\mathit{\text{SsrA}}}$. Scale bar, 10 μm. All schematics in this figure are for illustrative purpose only and not drawn to scale. Some schematics were created with BioRender.com.

Table 1.

Comparison of functional linkers for force measurement

	Length (a.a.)	Size of binding partner	Readout	Equipment for detection	Spatial and temporal resolution	Scalability#
GFP11 (SplitGFP)	16	24 kDa (GFP1-10)	Fluorescence reconstitution	Standard fluorescence microscope	Weak spatial resolution and no temporal resolution	On the order of 10 to 100
HiBiT (SplitNanoLuc)	11	18 kDa (LgBiT)	Luminescence reconstitution	Plate reader, Luminescence microscope (needs special setup)	Weak spatial resolution and weak temporal resolution	On the order of 100 or more
TEVcs	7	25 kDa (TEVp)	Protein cleavage	Equipment for Western blot or cellular immunofluorescence	No spatial resolution and no temporal resolution	One the order of 10 or less
IAAL-K3	21	2 kDa (IAAL-E3*)	Fluorescence redistribution	Standard fluorescence microscope, TIRF	Good spatial resolution and temporal resolution	Not scalable
SsrA	11	12 kDa (SspB*)	Fluorescence redistribution	Standard fluorescence microscope, TIRF	Good spatial resolution and temporal resolution	Not scalable

^*Size of IAAL-E3 or SspB alone. Typically, they are tagged to a fluorescent protein (e.g. ~ 27 kDa) for visualization.

^#Number of samples with the same readout that can be measured in parallel at the same time.

We demonstrate the use of our new sensors by measuring the in vivo forces on key force-bearing cytoplasmic proteins. First, we measured the force on Cdc12, the formin that polymerizes actin filaments and connects them to the plasma membrane during cytokinesis in the fission yeast Schizosaccharomyces pombe^47,48. Using the split-Nanoluc (Fig. 2b, Fig. S3) and split-GFP (Fig. 2c, Fig. S4) recorder readouts, we measured 6 pN peak tension on formin Cdc12 (Fig. 2a). Next, we followed the temporal evolution of the force on Cdc12 over the course of cytokinesis using the live reporter readout IAAL-K3/ mEGFP-IAAL-E3 (Fig. 2d). Timelapse imaging indicates that force on Cdc12 starts to exceed 3 pN during ring constriction and drops below 3 pN before the ring fully disassembles, suggesting that the mechanical tension on Cdc12 is down-regulated to match the decreased need for F-actin assembly. These results are the first force measurements on a formin in vivo. Previous in vitro measurements using a microfluidic chamber showed that weak tension on budding yeast formin Bni1 or mammalian formin mDia1 accelerates the formin-mediated polymerization of profilin-actin^49,50. In addition, in vitro single-molecule spectroscopy experiments demonstrated that ~7 pN tension on formin mDia1 accelerates its actin-polymerizing activity by nearly 8-fold, in the presence or absence of profilin^51,52. Our in vivo force measurement of 6 pN on native Cdc12 strongly suggests that mechanical force regulates the activity of formin during cytokinesis⁵³.

Figure 2. Modular assembly of mechano-sensors with versatile readouts in multiple biological systems.

a, Schematic of the role of Cdc12 during cytokinesis in the fission yeast S. pombe. Cdc12 nucleates and polymerizes actin filaments, and bridges F-actin to the cytokinetic nodes on the plasma membrane. Force sensors are inserted after A216 of cdc12 genomic location. b, Force measurement on Cdc12 using the split-NanoLuc readout. Ordinary one-way ANOVA was performed with Tukey’s multiple comparison tests with “No-HiBiT” as the control and only displayed for pairs where p values is less than 0.05. ****, p<0.0001. Each dot represents one measurement with >1000 cells. Data are pooled from three independent repeats. c, Force measurement on Cdc12 using the split-GFP readout. Ordinary one-way ANOVA was performed with Tukey’s multiple comparison tests with “No-GFP11” as the control and only displayed for pairs where p values is smaller than 0.05. ****, p<0.0001. Each dot represents measurement from a single cell. Data are pooled from more than three independent experiments. d, Timelapse of force on Cdc12 (${力}_{3\mathit{\text{pN}}}^{\mathit{\text{IAAL}}-K3}$ and mEGFP-IAAL-E3 binder) and the cytokinetic ring (Rlc1-sfCherry) during fission yeast cytokinesis. Sad1-mCherry is used to locate the dividing spindle pole body and to time cytokinesis. Force on Cdc12 starts to build up above 3 pN at the beginning of ring maturation and drops below 3 pN before the ring fully disassembles. Arrowheads indicate the recruitment of mEGFP to Cdc12 when ${力}_{3\mathit{\text{pN}}}^{\mathit{\text{IAAL}}-K3}$ is unfolded by force. Scale bar, 5 μm. e, Schematic of Ent1 and its interacting partners during clathrin-mediated endocytosis (CME) in the fission yeast. Force sensors are inserted after P571 at ent1 genomic location. ENTH, N-terminal lipid-binding domain. ACB, actin cytoskeleton-binding domain. CBM, clathrin-binding motif. f, Force measurement on Ent1 using the split-NanoLuc readout. ~6 pN force is detected on Ent1 and the deletion of CMB decreased the luminescence signal detected, indicating a smaller fraction of Ent1 molecules under force. Ordinary one-way ANOVA was performed with “No-HiBiT” as the control and only displayed for pairs where p values is smaller than 0.05. *, p<0.05; **, p<0.01; ***, p<0.001; ****, p<0.0001. Each dot represents one independent experiment with >1000 cells. See also Fig. S5 for controls and force measurement on different sites of Ent1. g, Dual-color TIRF is used to simultaneously track the force on Ent1 (${力}_{3\mathit{\text{pN}}}^{\mathit{\text{IAAL}}-K3}$ and mEGFP-IAAL-E3 binder) and actin dynamics (Fim1-mScarlet-I) in fission yeast cells. Scale bar, 5 μm. h, Montage of a representative CME event. Note that force on Ent1 increases to above 3 pN ~30 s before actin assembly begins. Arrowheads denote the first and last frames where force on Ent1 is above 3 pN. i, Position of the sensory PLM neuron in the nematode C. elegans. Spectrins (UNC-70 and SPC-1), shown as spirals in the zoomed region, form the central building block of the membrane associated periodic skeleton (MPS), which consists of actin rings that are interspaced by spectrin tetramers below the plasma membrane throughout the entire length of the axon. Force sensors were inserted into the genomic unc-70 locus between spectrin repeats 8 and 9 (after R1167). Only one coiled coil is shown for clarity. See Fig. S6 for details of force sensor insertion. j, Representative maximum projections of C. elegans strains expressing UNC-70 with force sensors together with GFP1–10 under a PLM specific promoter (mec-17p) in wildtype or in a loss of function mutant of unc-115(ky275). Signal intensity is color coded according to the displayed color bar. Scale bar, 50 μm. The position of the cell body (*) and the rectum (#) are indicated. Background fluorescence outside of the axon region was not included in quantification. Example images are rescaled in X and Y dimension (1:2) to enlarge the neurite diameter for better visualization. k, Quantification of fluorescence in j. Multiple comparisons with the Kruskal-Wallis test. *, p<0.05. ****, p<0.0001. Each dot represents measurement from a single axon. l, Mini-Nesprin-2G transmits forces from cytoplasmic actin filaments to SUN in the nucleus. ${力}_{5\mathit{\text{pN}}}^{\mathit{\text{TEVcs}}}$ and ${力}_{10\mathit{\text{pN}}}^{\mathit{\text{TEVcs}}}$ were inserted between the N-terminal filamentous actin-binding CH domain (1–485) and the C-terminal KASH SUN-binding domain (6525–6874). m, Immunoblot of cell lysates from U2OS cells transfected with ${力}_{5\mathit{\text{pN}}}^{\mathit{\text{TEVcs}}}$ or ${力}_{10\mathit{\text{pN}}}^{\mathit{\text{TEVcs}}}$ in mini-Nesprin-2G. Note that cleavage of ${力}_{5\mathit{\text{pN}}}^{\mathit{\text{TEVcs}}}$ is observed upon co-transfection with TEVp, whereas ${力}_{10\mathit{\text{pN}}}^{\mathit{\text{TEVcs}}}$ is resistant to TEVp activity (TEVp-HA +). Representative Western blot of two independent repeats. All schematics in this figure are for illustrative purpose only and not drawn to scale. Some schematics were created with BioRender.com.

Second, we measured the forces on epsin Ent1, an adaptor protein involved in clathrin-mediated endocytosis. Ent1 contains an actin cytoskeleton-binding (ACB) domain and a clathrin-binding motif (CBM)^54,55 (Fig. 2e). Although the ACB domain has been shown to be critical for connecting the plasma membrane to the actin cytoskeleton^55,56, we suspected that the CBM may also contribute to the force relay due to the positioning of the clathrin coat within the F-actin layer (Fig. 2e). Indeed, using the split-Nanoluc readout, we detected a peak force of 6 pN before the ACB domain (at P571) (Fig. 2f). Importantly, elimination of the C-terminal CBM decreased the force by 75%, indicating that substantial tension on Ent1 is mediated by its interaction directly with the clathrin lattice, and not just from the actin cytoskeleton, as previously thought^27,55 (Fig. 2f, Fig. S5). We corroborated this conclusion by detecting forces above 3 pN both before and after the ACB domain (at A637), using the split-GFP readout as well as the TEVp readout (Fig. S5d–k). In agreement with this result, dual-color TIRF imaging of the force on Ent1 and the F-actin crosslinking protein fimbrin, which is used as a fiduciary marker of filamentous actin at endocytic sites (Fig. 2g), indicates that force on Ent1 starts to accumulate ~30 seconds before F-actin starts assembling^57,58 (Fig. 2h, Fig. S5l–n). This surprising result directly demonstrates for the first time that forces during endocytosis are not exclusively produced by actin assembly^55,59,60. Actin may interact with Ent1 at the later stages of endocytosis to pull the Ent1 tail from the clathrin layer to the actin layer⁶¹.

Third, we inserted the force sensors using the split-GFP recorder readout into C. elegans β-spectrin (UNC-70), which forms rungs between periodic F-actin rings in the submembrane region of axons (termed membrane-associated periodic skeleton or MPS)^62,63 (Fig. 2i, Fig. S6). We measured a peak force of ~8 pN on β-spectrin, which was dependent on the integrity of the MPS lattice, as impairment of MPS lattice formation in unc-115(ky275) mutants was required to maintain tension^64–66 (Fig. 2j–k). This result suggests that the maximal force experienced by a single β-spectrin in freely-moving worms is higher than that measured in paralyzed worms, where FRET-based force sensors reported an average force of 1.5 pN at the same location on β-spectrin⁶⁷.

Fourth, we measured the force in the linker of nucleoskeleton and cytoskeleton (LINC) complex responsible for force transmission across the nuclear envelope and nuclear positioning^68–70. To this end, we inserted the mechano-sensor into mini-Nesprin-2G, a smaller functional version of the Nesprin-2G protein that associates with actin on the cytosolic side and the SUN protein within the lumen of the nuclear envelope^71,72 (Fig 2l). Upon co-transfection with the TEV protease, mini-Nesprin-2G with cc-5 pN was cleaved, while the mini-Nesprin-2G with cc-10 pN remained resistant to proteolytic cleavage (Fig. 2m, S7). We confirmed these measurements using the split-GFP and the split-Nanoluc readouts (Fig. S8). Our results are consistent with previous studies showing that mini-Nesprin-2G is under actin-dependent tension in the single pN range^73,74.

To showcase the transformative power of our new sensor design, we measured the forces on one of the well-studied mechano-sensitive protein, vinculin, using multiple readouts to investigate its behavior in different settings. Vinculin is a key focal adhesion protein, which is believed to be under ~2 pN average tension, as determined from previous ensemble measurements with FRET-based force sensors in adherent cells^4,11,75 (Fig. 3a). However, vinculin is also known to be a mechanosensitive protein that changes conformation under 5–15pN force to modulate its activity (e.g. vinculin becomes an actin filament catch bond when under ~8pN force)^76,77. In addition, the average force reported by FRET-based sensors is likely an underestimate because any vinculin under forces far outside of the linear range of the FRET-sensor emits the same signal as vinculin under force at the sensors’ bounds^10,16,24,27. Last, ensemble in vivo measurements using FRET sensors are unable to represent the distribution of forces on the vinculin (i.e. are all vinculin molecules under the same force? Or are there a few vinculin molecules under very large force while most are under very low force? Etc.). Using our sensors with the split-GFP recorder readout, we showed that the majority of vinculin molecules (~90%) under tension are experiencing force above 5 pN, and that ~10% bear forces above 10 pN (Fig. 3b–c, S9). Similar ratios were measured using the split-Nanoluc readout using newly adherent cells on different substrates (Fig. 3d–f). These results demonstrate that a small percentage of vinculin molecules mediate most of the tension between talin and F-actin^12,13. Our data also demonstrate that fewer vinculin molecules were under tension on Matrigel than on glass, with a reduced fraction of vinculin molecules under tension above 5 pN (Fig. 3e–f). To test if substrate stiffness alone changes the force on vinculin, we plated cells with the split-Nanoluc sensors on a 96-well plate with varying stiffness in each well (0.1 kPa to 100 kPa) (Fig. 3g), which, to our knowledge, is the first high-throughput single molecule force measurement on vinculin. We did not detect a strong dependence of force on vinculin on substrate stiffness (Fig. 3h–i), in agreement with a previous force measurement with a smaller sample size⁷⁸. From our measurements, the large fraction of vinculin molecules with force above 5 pN (~70%) across stiffness is consistent with the in vitro activation force threshold of single vinculin molecules, a force level needed for the conformational change of vinculin to relieve self-inhibition and to bind F-actin^11,76,79,80. The much smaller fraction of vinculin molecules with force above 7 pN (~10%) implies that activated vinculin may buffer force, possibly mediated by the flexible linker between its head and tail domains or by disengaging from other force transmitting proteins^81,82. To demonstrate that the binding and unbinding kinetics of our IAAL-E3/IAAL-K3 reporter system is fast (sub-second to second time scales), we used Fluorescence Recovery After Photobleaching (FRAP) to measure a recovery rate with a half-time of 1.6 s (Fig. 3j–k). We also demonstrated that the size of the fluorescent tags is not limiting the detection ability of our sensors in the crowded environment of focal adhesions (Fig. S10).

To gain a spatial understanding of the forces on vinculin, we used two pairs of orthogonal binder/tag (IAAL-E3/IAAL-K3 and SsrA/SspB) to label mechanically active vinculin molecules and detect two levels of force (respectively 3 pN and 5 pN) with two different colors (respectively mScarlet-I and TagBFP) (Fig. 3l). Consistent with the split-GFP and split-Nanoluc readouts, we measured a smaller fraction of focal adhesions with TagBFP signals (i.e. forces larger than 5 pN) than with mScarlet-I signals (i.e. forces larger than 3 pN). In addition, our results demonstrate that TagBFP signals are further away from the cell edge, likely coinciding with more mature focal adhesions (Fig. 3l)^4,11,75,79. Overall, our design of mechano-sensors enabled us to collect new data on vinculin in a short time and with a high throughput, to produce a multi-view study of the forces on vinculin with unprecedented precision

In summary, we present here a new modular approach for in vivo force measurement using mechano-sensors that are genetically encoded, small, compatible with a versatile readout selection, and accessible to virtually any lab⁸³. We expect the calibrated coiled coils are compatible with a broader range of readouts not directly tested in this paper, including FRET-based readouts⁸⁴. The validity of our approach is independently demonstrated on different proteins and in multiple biological systems. We envision this modular architecture of mechano-sensors to herald an explosion of in vivo force measurements and to truly open the gate to quantitative mechanobiology.

Limitations of the study

In this study, we have not systematically characterized in details all the different combinations of coiled coils and readouts in vitro. However, the modularity of our designs and the cross-validation of measurement in vivo using different readouts we have performed suggest each module is likely independent from each other. It is possible that due to differences in the dissociation rates of the binding pairs, the disappearance of signal after force is released vary slightly depending on the live reporter readout used, which may potentially create a discrepancy in the measurement of force thresholds between different readouts (systematic error). Another limitation of the live reporters presented in this paper is that they rely on local enrichment of fluorescence from the cytoplasm to the subcellular structures containing the proteins under force. Therefore, it may require optimization of the expression of the fluorescent binder probes, and the parameters for image acquisition and image processing to increase the signal-to-noise ratio. Supplementary note S1 presents a mathematical model and simulations for the sensors and their readouts, and provides a detailed discussion of the biochemical and biophysical parameters to consider for best results.

Material and Methods

Protein purification and biotinylation

DNA sequences coding coiled coils were synthesized from IDT and subcloned into BL21 (DE3) E. coli (New England BioLabs) for protein expression. Coiled coils were expressed with GST fusion at the N-terminus and Avi-tag at the C-terminus. Cleared bacteria lysates were purified by binding to Glutathione Sepharose 4B beads (GE Healthcare). The GST tag was cleaved by PreScission Protease (Sigma-Aldrich) afterwards. Avi-tag in purified coiled coil proteins was exchanged into biotinylation buffer (25 mM HEPES, 200 mM potassium glutamate, pH 7.7) and biotinylated with 50 mM bicine buffer, 50 μg/mL BirA, pH 8.3, 10 mM ATP, 10 mM magnesium acetate, and 50 μM d-biotin (Avidity) at 4°C overnight.

DNA handle preparation and crosslinking with protein

The DNA handle used for protein attachment in single-molecule experiments was produced via PCR, yielding a 2,260 base pair fragment. This DNA fragment was modified to include a thiol group (-SH) at one end and two digoxigenin moieties at the other. For the crosslinking process, as described previously^85,86, DNA handle was conjugated first with DTDP at pH 5.5 and excess DTDP was removed by spin column. Then purified proteins were mixed with the DTDP-treated DNA handle at a molar ratio of 50:1 (protein to DNA). This mixture was prepared in a 100 mM phosphate buffer containing 500 mM NaCl at pH 8.5 and incubated overnight at room temperature.

Single-molecule manipulation experiments

All single-molecule experiments were performed using dual-trap high-resolution optical tweezers^85,87,88. In brief, the two optical traps were created by focusing two orthogonally polarized beams through a water-immersed 60× objective with a 1.2 numerical aperture (Olympus). These beams originated from a single 1,064-nm laser generated by a solid-state laser (Spectra-Physics) and were then split. One of the beams was deflected by a mirror mounted on a piezoelectrically controlled stage (Mad City Labs), which could tilt along two orthogonal axes, allowing for relative movement between the traps. The outgoing laser beams were collimated by a second water immersion objective, split, and projected onto two position-sensitive detectors (Pacific Silicon Sensor). Bead displacements were detected using back-focal plane interferometry. To prepare the DNA handle crosslinked proteins, they were incubated with anti-digoxigenin antibody-coated polystyrene beads (2.17 μm in diameter, Spherotech) for 15 minutes. This mixture was then diluted with 1 mL of phosphate-buffered saline (PBS) and introduced into the top channel of a microfluidic chamber. Meanwhile, streptavidin-coated beads (1.86 μm in diameter) were introduced into the bottom channel of the chamber. Both channels were connected to a central channel via capillary tubes, where the beads were trapped. Once a bead from each type was captured, a single protein was tethered between them by bringing the two beads close together. The tethered molecule underwent pulling and relaxation by adjusting the trap separation at a rate of 10 nm/s. The optical tweezers experiments were conducted in PBS at 23 (± 1) °C. The buffer in the central channel was supplemented with an oxygen scavenging system as described elsewhere. Data were processed using MATLAB codes, also described elsewhere^85,89, and unfolding forces were determined from the force-extension curves.

Steered molecular dynamics (SMD) simulation

Structures of coiled coils were predicted from AlphaFold2⁹⁰. We rotated the coiled coil so that the vector from the N terminus to the C terminus aligned with the x-axis, ensuring a consistent initial condition for all coiled coils. We used Qiwik to generate NAMD’s configuration files. The total displacement was set to 200 Å with a margin of 15 Å for water molecules, and the salt concentration was set to 0.15mol/L. We anchored the first residue of the coiled coils and pulled the last one. The speed at which we pulled the SMD’s dummy atom was 2.5e-5 A/fs (2500 m/s), and the spring constant was 1e-2kcal/mol/Ų (6.949^e−1 pN/ Å). This unrealistically high pulling speed was set up to obtain results within a reasonable time (~60 ns of simulated time, or ~1 week simulation, per coiled coil).

Before running the SMD simulation, we performed a minimization for 5,000 steps, followed by annealing by increasing the temperature from 60 K to 300 K. We increased the temperature by 1-degree increment and ran 600 steps between each increment. In total, it took 14,400 steps to increase the temperature to 300 K. Finally, we equilibrated the system by simulating it for 500,000 steps. We used the result of the equilibrated system as the initial condition of the SMD simulation. We used a step size of 2 fs for all the simulations mentioned.

To measure the force threshold, we measured the minimum force required to open the coiled coil in each iteration and then took the average over all forces. We used the Potential of Mean Force as described by Park and Schulten⁹¹.

Selection of insertion sites for force sensors

To choose insertion sites on a protein of interest, we typically avoid known folded domains, and preferentially choose a disordered region that has no known post-translational modification or binding partner using protein databases that are species specific (e.g. Pombase for S. pombe or Wormbase for C. elegans) or general (e.g. Uniprot). When possible, we recommend using protein structure prediction tools (e.g. AlphaFold) to check the protein structure with and without the insertion of the force sensors.

Yeast strains and media

The S. pombe strains in this paper are listed in Supplementary Table S1. The strains were created through CRISPR-mediated genome editing as outlined previously^92–94 and confirmed through sequencing of colony PCR products. Yeast cells were cultured in YE5S medium (Yeast Extract supplemented with 0.225 g/L each of uracil, adenine, lysine, histidine, and leucine) and imaged on gelatin pads made with EMM5S (Edinburgh Minimum media supplemented with 0.225 g/L each of uracil, adenine, lysine, histidine, and leucine). Before imaging, yeast cells were grown at 32 °C with continuous shaking at 200 rpm overnight, allowing them to reach the exponential growth phase with OD_595nm values between 0.3 and 0.5.

Confocal microscopy for yeast cells

Live imaging of S. pombe cells was performed on a 25% gelatin pad at room temperature, using a Nikon TiE inverted microscope with a CSU-W1 Confocal Scanning Unit from Yokogawa Electric Corporation (Tokyo, Japan). The microscope was equipped with a CFI Plan Apo 100X/1.45NA Phase objective from Nikon. Image acquisition was performed using an iXon Ultra888 EMCCD camera from Andor (Belfast, UK). For imaging GFP, an excitation wavelength of 488nm from an argon-ion laser was used, and the fluorescence emission was filtered via a single band pass filter 510/25 in the Spectra X system. mCherry and sfCherry2 were excited from a 561nm argon-ion laser, and the fluorescence emission was filtered through a single band pass filter 575/25 in the Spectra X system. To image the entire yeast cell, 21 optical slices were collected with a thickness of 0.5 μm, followed by maximum intensity projection to generate 2D images. For image display and analysis, the Fiji distribution of ImageJ provided by the National Institutes of Health (NIH, USA) was used⁹⁵. S. pombe cells were also imaged on gelatin pads on BC43 CF tabletop confocal (Andor) with 100x Plan Apochromat oil immersion objective and preset excitation/emission combinations for GFP, mCherry and sfCherry.

TIRF microscopy

Live TIRF imaging of yeast cells was performed at room temperature using gelatin pads prepared as previously described^96,97. Imaging was conducted using a Nikon Ti2E inverted microscope equipped with an Abbelight module (Cachan, France), equipped with a Teledyne Kinetix Scientific CMOS camera (01-KINETIX-M-C; Teledyne Photometrics, AZ, USA), 488-nm and 561-nm solid-state diode lasers, 525/50 and 600/40 single bandpass filters (for mEGFP and mScarlet-I imaging, respectively). Both channels were imaged simultaneously using a beam splitter, at 1-second intervals. Image display and analysis were performed using the Fiji distribution of ImageJ (National Institutes of Health, USA)^95,98.

Measurement of bioluminescence (yeast)

Yeast cells were grown to exponential phase in YE5S medium, pelleted, and resuspended in EMM5S to a final density of 1 OD_595nm/μL. 150 μL cells were mixed with 15 μL diluted substrate (Nano-Glo Live, Promega, 1:50 dilution.) and loaded into a black 96 well plate (#675086, Greiner Bio-One) before measuring luminescence on a plate reader (BioTek synergy H1, Agilent). Luminescence signals were recorded 5 minutes after mixing cells with the substrate and normalized to the OD_595nm of each well. Measurements were kept in a linear range predetermined as in Fig. S3.

Measurement of bioluminescence (mammalian cells)

U2OS cells were cultured in a 4-compartment dish (#627870, Greiner Bio-One) till ~90% confluency. Cells were transfected with plasmids (pJB425, pJB413, pJB435, pJB427) using Lipofectamine 3000. 24 Hours after transfection, cells were removed from each compartment by trypsin digestion, pelleted and separated into three equal volumes (100 μL each). One volume was mixed with 20 μL Nano-Glo Lytic buffer containing 1:50 diluted substrate and 1:100 diluted LgBiT for luminescence measurement in a black 96-well plate <5 mins after digestion. The other volumes were added either directly into a well of a black 96-well plate, or into a well precoated with 10 μL Matrigel (Corning). Luminescence was measured 6 hours after plating by adding Nano-Glo Lytic buffer containing 1:50 diluted substrate and 1:100 diluted LgBiT. For the Matrigel coated well, 10 μL culture media was taken out before measurement to keep the volumes consistent. Raw luminescence values were normalized by cell density in each well. For luminescence measurements on mCherry-vinculin, mCherry signals were used for normalization.

To measure the forces on vinculin across stiffness gradients, U2OS cells were transfected with plasmids (pJB425, pJB413, pJB435, pJB427) using Lipofectamine 3000. 24 Hours after transfection, cells were collected by trypsin digestion, pelleted and separated into 11 equal volumes (100 μL each) before being replated onto a 96 well plate with gradient stiffness (SW96G-HTS-COL-EA, Matrigen) overnight. Luminescence was measured by adding 20 μL Nano-Glo Lytic buffer containing 1:50 diluted substrate and 1:100 diluted LgBiT. Raw luminescence values were normalized by mCherry-vinculin signals in each well.

FRAP of vinculin live force reporter

U2OS cells were co-transfected with pJB355 and pJB357 following standard protocols with Lipofectamine 3000. Cells were imaged 24 hours after transfection. FRAP was performed on FV3000 using a 60X objective (NA 1.42). After correcting for photobleaching (ImageJ, NIH), the recovery of fluorescence was fitted with a regression curve f(t)=MF*(1-exp(−k*t)), where MF is a mobile fraction and k is a turnover rate.

Nematode Microscopy and Data Analysis

Nematodes were cultured at 20°C on nematode growth medium plates that were seeded with OP50 bacteria. Detailed animal preparation for microscopy was published previously⁹⁹. In brief, animals at their larval stage L4 were mounted on 10% agarose patch and paralyzed in a droplet of 10 mM Levamisole diluted in M9 medium. Images were acquired with a DMi8 inverted microscope (Leica) that is equipped with a VT-iSIM system (Biovision) and an ORCA-Flash 4.0 camera (Hamamatsu). The microscope was controlled by the MetaMorph Advanced Confocal Acquisition Software. Images were acquired with an HC PL APO 40x/1.30NA OIL CS2 objective at a 488nm laser line. Raw images were processed and analyzed in Fiji/ImageJ v2.3.0/1.53f51^95,98. Images were acquired in single layers and then stacked into maximum projections. To capture the intensity signal along the entire length of PLM, which could not be acquired in a single field of view, multiple images along the length of the neurite were taken and stitched into a single image using the pairwise-stitching plugin with a linear blending fusion method¹⁰⁰. To determine the mean fluorescence along the PLM neurite, a 5-pixel thick line was drawn along the center of the neurite and a signal intensity profile was generated by using the plot profile function and the signal intensity was averaged. To subtract background fluorescence, the same signal intensity profile was acquired by shifting the drawn line from the center of the neurite by a few pixels into the non-neuronal tissue directly contacting the neurite. The signal intensity was calculated in arbitrary units as ${\mathrm{I}}_{\text{mean}}={\mathrm{I}}_{\text{mean}}\left(\text{neurite}\right)-{\mathrm{I}}_{\text{mean}}\left(\text{background}\right)$.

Plasmids and transient transfections for TEVp expressing cells

The TEV module was cloned using standard cloning techniques into the nesprin TSmod 25 in pcDNA3.1 (https://www.addgene.org/browse/sequence/244038/)¹⁰¹. The mCerulean and mVenus sequences were removed and replaced with either the 5 pN or 10 pN sensor leading to an insertion between the N-terminal actin-binding CH domain (1–485) and the C-terminal KASH SUN-binding domain (6525–6874) (Fig. 2l)

U2OS cells were cultured on coverslips and transfected using Lipofectamine 3000 reagent (ThermoFisher) according to the manufacturer’s instructions. Immunoblot protocols were executed as previously described¹⁰². Briefly, 10 μg of protein were resolved by 10% SDS–PAGE gels and transferred onto polyvinylidene fluoride membranes (Bio-Rad). Membranes were blocked in 4% (wt/vol) milk in PBS + 0.1% (vol/vol) Tween-20 (Sigma-Aldrich). Primary and horseradish peroxidase-conjugated secondary antibodies were diluted in blocking buffer. A ChemiDoc gel imaging system (Bio-Rad) was used to visualize chemiluminescence.

For immunofluorescence, cells were fixed with 4% PFA/PBS for 15 min, permeabilized with 0.3% Triton X-100/PBS for 3 min, and blocked with 4% BSA/PBS (Sigma-Aldrich) for 1 hour. Coverslips were incubated for one hour with primary antibodies: FLAG (Millipore Sigma F3165, 1:500), HA (Roche 11867423001, 1:500), or LaminA (Invitrogen MA1–06101, 1:500). Samples were washed 2x with PBS for 5 min, incubated with Alexa Fluor (Life Technologies) secondary antibodies for one hour, and then washed 3x with PBS and mounted with ProLong Gold Antifade reagent + DAPI (Thermo Fisher P36935). Images were acquired on an LSM 880 laser scanning confocal microscope (Zeiss) with Airyscan using a C Plan-Apochromat 63×/1.40 oil DIC M27 objective using ZEN 2.1 software (Zeiss).

Statistical evaluation

Statistical evaluation was performed with GraphPad Prism (version 7). The dataset was first tested for normality distribution by using the DÁgostino and Pearson test to judge the use of non- vs parametric statistical tests. For formin Cdc12 and Vinculin force measurements, one-way ANOVA with Tukey’s multiple comparison tests was used with MATLAB (MathWorks, version 2021).

Supplementary Material

Figures S1–S11 and Table S1–S4

Note S1

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Monoclonal ANTI-FLAG M2 antibody	Millipore	F3165
Anti-HA High Affinity	Roche	3F10 (11867423001)
Lamin A Monoclonal Antibody (133A2)	Invitrogen	MA1–06101
Chemicals, peptides, and recombinant proteins
PreScission Protease	Sigma-Aldrich	GE27-0843-01
d-biotin	Avidity	Bio-500
Glutathione Sepharose 4B beads	Cytiva	17075601
polystyrene beads	Spherotech	PP10-20-10
Lipofectamin3000	ThermoFisher	L3000001
See Table S4 for sequences of calibrated coiled-coils	This study	N/A
Critical commercial assays
Nano-Glo live cell assay system	Promega	N2011
Nano-Glo HiBiT Lytic detection system	Promega	N3030
Softwell 96 HTS	Matrigen	SW96G-HTS-COL-EA
Experimental models: Cell lines
Human: U2OS	Laboratory of Christian Schlieker	N/A
Human: U2OS with stable integration of H2B-tagBFP-P2A-GFP1–10	Made in the Berro lab	N/A
Experimental models: Organisms/strains
See Table S1 for the list of S. pombe strains	This study	N/A
See Table S2 for the list of C. elegans strains	This study	N/A
See Table S3 for the sequences of DNA templates to generate C. elegans strains	This study	N/A
Recombinant DNA
pGEX6PI-Oakley-p1	This study	pJB297
pGEX6PI-Oakley-p1-OnePlus	This study	pJB298
pGEX6PI-Oakley-p1-LV-L3S	This study	pJB337
pGEX6PI-Oakley-p1-TwoPlus	This study	pJB339
pGEX6PI-Oakley-p1-FourPlus	This study	pJB340
pmCherry-Vinculin-E883–4G-DyneinStalk-GFP11	This study	pJB348
pmCherry-Vinculin-E883–4G-Oakley-p1-LV-L3S-GFP11	This study	pJB349
pmEGFP-IAAL-E3	This study	pJB355
pmCherry-Vinculin-E883–4G-DyneinStalk-IAAL-K3	This study	pJB357
pmCherry-Vinculin-E883–4G-Oakley-p1-GFP11	This study	pJB363
pmCherry-Vinculin-E883–4G-Oakley-p1-OnePlus-GFP11	This study	pJB364
pmCherry-Vinculin-E883–4G-Oakley-p1-LV-L3S-HiBiT	This study	pJB413
pmScarlet3-IAAL-E3-tPT2A-Vinculin-E883–4G-DyneinStalk-IAAL-K3	This study	pJB417
pmCherry-Vinculin-E883–4G-DyneinStalk-HiBiT	This study	pJB425
pmCherry-Vinculin-E883–4G-Oakley-p1-OnePlus-HiBiT	This study	pJB427
pmCherry-Vinculin-E883–4G-Oakley-p1-HiBiT	This study	pJB435
pmEGFP-mScarlet3-IAAL-E3-tPT2A-Vinculin-E883–4G-DyneinStalk-IAAL-K3	This study	pJB439
pTagBFP-SspB_nano-tPT2A-Vinculin-E883–4G-Oakley-p1-LV-L3S-SsrA	This study	pJB449
Software and algorithms
GraphPad Prism 7	GraphPad	https://www.graphpad.com
ImageJ	NIH	https://imagej.net
MATLAB	MathWorks	https://www.mathworks.com/products/matlab.html