Cross-regulations of two connected domains form a mechanical circuit for steady force transmission during clathrin-mediated endocytosis

SUMMARY

Mechanical forces are transmitted from the actin cytoskeleton to the membrane during clathrin-mediated endocytosis (CME) in the fission yeast Schizosaccharomyces pombe. End4p directly transmits force in CME by binding to both the membrane (through the AP180 N-terminal homology [ANTH] domain) and F-actin (through the talin-HIP1/R/Sla2p actin-tethering C-terminal homology [THATCH] domain). We show that 7 pN force is required for stable binding between THATCH and F-actin. We also characterized a domain in End4p, Rend (rod domain in End4p), that resembles R12 of talin. Membrane localization of Rend primes the binding of THATCH to F-actin, and force-induced unfolding of Rend at 15 pN terminates the transmission of force. We show that the mechanical properties (mechanical stability, unfolding extension, hysteresis) of Rend and THATCH are tuned to form a circuit for the initiation, transmission, and termination of force between the actin cytoskeleton and membrane. The mechanical circuit by Rend and THATCH may be conserved and coopted evolutionarily in cell adhesion complexes.

In brief

Ren et al. characterize a mechanical circuit formed by two connected and mutually regulated protein domains during clathrin-mediated endocytosis. This circuit ensures robust force transmission from the actin cytoskeleton to the membrane and may represent a universal mechanism to synergize catch bonds and protein condensates at the membrane.

Graphical Abstract

INTRODUCTION

Numerous proteins connect the actin cytoskeleton to membranes and transmit force between them in cell-cell adhesion, cell-matrix attachment, and the internalization of the cell membrane during endocytosis.^1–4 These force-bearing proteins include cadherin/catenin, fibronectin/integrin/vinculin, and epsin/HIP1R, among others.^5–9 Additional proteins regulate the composition and strength of these connections in response to both chemical (e.g., ligand type and density) and mechanical (e.g., substrate rigidity, membrane tension) cues,^10–14 and faulty connections are implicated in a variety of developmental and physiological disorders.^1,15,16

The need to construct stable and adaptive supramolecular mechanical connections poses a challenge for adhesion protein machineries. Proteins must bind tightly to sustain force transmission but may also need to unbind quickly to terminate force transmission to meet dynamic mechanical needs, i.e., produce the right amount of force at the right location and at the right time. Growing evidence suggests that mechanical forces play important roles in protein-protein interactions.^2,17,18 For example, several F-actin-binding proteins in cell adhesion (α-catenin, talin, vinculin) have been found to display catch-bond behavior, where force enhances their binding to F-actin, especially when pulled toward the pointed end of F-actin.^19–22 Moreover, domains in talin and α-catenin can be mechanically unfolded to reveal cryptic binding motifs that selectively recruit proteins, such as vinculin, to strengthen focal adhesions.^23–27 Interestingly, the F-actin-binding domains in cell adhesion proteins are often five-helix bundles,^18,28 and force-induced partial unfolding of the five-helix bundle has been recently shown to underlie a catch-bond mechanism.²⁶ Deletion of the first α helix from αE-catenin eliminated the catch-bond behavior, resulting in stable binding to F-actin that can be described by a single-state slip bond.²⁶ Whether this is a conserved mechanism to form catch bonds in other mechanotransduction pathways is unknown, and in vivo data to compare the force transmission with or without catch-bond formation are missing. In addition, the catch-bond mechanism pertaining to F-actin-binding domains alone is insufficient to explain how the initial binding of these adhesion proteins is achieved in the absence of force when the binding affinity is low and what stops the high-affinity binding after stable catch bonds are formed.

End4p is a homolog of HIP1R in the fission yeast Schizosaccharomyces pombe, and the F-actin-binding domain in End4p is a five-helix bundle named THATCH (talin-HIP1/R/Sla2p actin-tethering C-terminal homology).^28–32 End4p transmits mechanical forces from F-actin to the plasma membrane during clathrin-mediated endocytosis (CME), and we have recently shown that there is a gradient of force along the End4p molecule during CME in vivo.^7,33 The peak force before THATCH reaches ~19 pN, which drops to ~11 pN near the clathrin lattice and ~9 pN near the membrane.⁷ Compared with cell-cell and cell-matrix adhesion that exists in the minutes-to-hours timescale, CME in the fission yeast is transient, where F-actin-driven formation of a vesicle out of an initially flat membrane only lasts ~10 s for each endocytic event.^3,4 Therefore, the binding between F-actin and End4p must be tightly coupled spatially and temporally to ensure robust onset and termination of force transmission. How this is achieved is unclear.

Here, we show that the physical connection between two mechanosensitive domains within End4p, Rend (rod domain in End4p) and THATCH, acts as a mechanical circuit to start force transmission only at the plasma membrane, buffer the transmission of force to a small range, and terminate force transmission when it exceeds a pre-determined magnitude. This circuit may represent a general solution to regulate catch bonds in force transmission.

RESULTS

End4p connects F-actin and the plasma membrane during CME

End4p forms a dimer in fission yeast cells.^33,34 End4p binds to the plasma membrane via its N-terminal ANTH (AP180 N-terminal homology) domain and to F-actin through the C-terminal THATCH domain.^29,35–37 End4p interacts with the clathrin lattice by binding to the clathrin light chain and binds to several SH3-domain-containing proteins through its proline-rich domain (PRD) (Figure 1A).^34,38–40 The extensive interactions between End4p and other endocytic proteins relay forces between F-actin and the plasma membrane.⁷

Figure 1. The unfolding of the USH from THATCH requires at least 7 pN force.

(A) Schematic of an End4p dimer between the lipid membrane and the actin cytoskeleton. The N-terminal ANTH domain binds to PIP2. The proline-rich domain (PRD) binds to SH3-domain-containing protein in the endocytic coat. An unknown motif before the dimerization domain binds to the clathrin light chain. THATCH domain binds to F-actin. The drawing is not to scale.

(B) Structure of THATCH from End4p as predicted by AlphaFold. THATCH comprises a five-helix bundle (core), with the first helix, termed the USH, and a C-terminal tail (latch) that mediates dimerization. An arginine (R1093) on the latch is critical for dimerization. Two cysteine residues, C995 and C1034, indicated by blue dots, form a disulfide bond facing the inside of the last two helices of the THATCH core.

(C) F-actin-binding surface of THATCH core is located at the opposite side of USH.^28,32 The barbed end of F-actin is facing the reader in this configuration. Note that pulling the USH toward the pointed end of F-actin, or away from F-actin, leads to the unzipping of the USH.

(D) Force-extension curves (FECs) of THATCH core pulled in the shearing or unzipping orientation. Different states of the THATCH core are placed next to the corresponding curve. Blue dots indicate C995 and C1034. See the full analysis of FECs in Figure S1.

(E) USH unfolds at 13 ± 1.9 pN (mean ± SD) and refolds at 5.4 ± 0.8 pN when pulled in the shearing orientation.

(F) USH unfolds and refolds at an equilibrium force of 7.2 ± 1.2 pN when pulled in the unzipping orientation, which reflects the F-actin-binding orientation in (C).

The THATCH domain is composed of a “core” that directly binds to F-actin and a “latch” that mediates dimerization by forming an anti-parallel coiled coil (Figures 1B and S1A).^28,32 The structures of THATCH from human HIP1R and mouse talin-1 reveal an F-actin-binding surface at the opposite side of the first α helix (also named the upstream helix [USH]) (Figure 1C). Dimerization through the latch greatly increases the binding affinity between the THATCH core and F-actin, and the presence of the USH strongly inhibits binding, as shown from bulk biochemical assays.^28,32,41,42 Dimerized THATCH domains from human talin-1 form an asymmetric catch bond with F-actin, where force >5 pN on THATCH toward the pointed end of F-actin results in exceptionally stable binding.²² A mechanistic understanding of such potent catch-bond behavior of THATCH and its regulation and biological relevance in vivo is still missing.

Mechanical force regulates the affinity of THATCH to F-actin by unfolding the USH in vivo

We began by measuring the mechanical stability of the purified THATCH core from End4p using optical tweezers. Unlike previous experiments where the binding between F-actin and THATCH was studied in a reconstituted system, we focused solely on the response of the isolated THATCH domain to tensile forces.⁴³ To characterize the mechanical stability of the THATCH domain, we first pulled the domain from its N and C termini. This corresponds to force transmitted to THATCH by another End4p through latch dimerization. To this end, we directly attached the domain to two beads held in separate optical traps via a biotin tag and a DNA handle. The force applied to the THATCH core was changed by controlling the distance between the two optical traps. The protein transition was detected as the end-to-end extension of the protein-DNA tether between the two beads. As the pulling force increased, the THATCH domain reversibly unfolded and refolded at a force ~11.5 pN between a folded THATCH domain (Figure 1D, state 1) and a partially unfolded intermediate (state 2). The extension change associated with the transition indicates that intermediate state 2 corresponds to the THATCH domain with the USH (Figure 1E). The intermediate state was irreversibly unfolded at 13 ± 1.9 pN (mean ± SD) to another state 3, where the THATCH domain was largely unfolded except for the amino acids between the two crosslinked cysteine residues, which probably remained in a helical hairpin conformation (Figures 1D, state 3, and S1B). When relaxed, the THATCH domain completely refolded at 5.4 ± 0.8 pN force. By introducing a cysteine into the loop between the 4^th and 5^th α helices and crosslinking the cysteine residue to the DNA handle, we pulled the THATCH domain in a different direction. This pulling direction mimics the direction of force directly exerted by F-actin. The THATCH domain reversibly unfolded and refolded with an equilibrium force of 7.2 ± 1.2 pN. The extension change of the transition is consistent with USH unfolding in unfolded state 2 (Figures 1D, 1F, and S1C). The remaining four-helix bundles could sustain large forces up to ~25 pN without unfolding (Figure 1D). The unfolding of the USH is reversible under both shearing and unzipping forces, and the transition is fast, with a lifetime ~50 ms at equilibrium forces. Altogether, our data suggest that a THATCH core where the USH is mechanically dissociated from the remaining four-helix bundle (state 2) can be achieved either by forces applied on the C terminus of the THATCH domain or from the F-actin-binding surface, and at least ~7 pN force is needed to maintain the THATCH core in an open conformation.

To explore the biological function of the USH domain, we went on to study the affinity of THATCH constructs to F-actin with or without the USH in fission yeast cells. In wild-type cells, full-length End4p localizes at endocytic sites, whose location correlates with sites of cell wall deposition, i.e., the cell tips during cell growth and the midline during cell division (Figure 2A).^7,44 Isolated THATCH dimers displayed a diffusive localization throughout the fission yeast cell, indicating a low affinity to F-actin, similarly to THATCH homologs (Figure 2B).^45,46 In contrast, overexpression of THATCH-ΔUSH dimers induced the formation of thick F-actin bundles and large puncta (Figures 2C and S1D).³⁴ Both structures contained endocytic actin crosslinker fimbrin (Fim1p) but not the cytokinetic actin crosslinker α-actinin (Ain1p) (Figures 2D and 2E; Video S1), suggesting that F-actin bundled by THATCH-ΔUSH dimers retains some features of endocytic actin.⁴⁷ The presence of filament bundles depended on THATCH-ΔUSH dimerization, as monomeric constructs that contained a point mutation preventing dimerization of the latch (THATCH-ΔUSH-R1093G) were diffusive (Figure 2F). The F-actin structures induced by THATCH-ΔUSH dimers were remarkably stable, as treatment with Latrunculin A (LatA) at a dose that destroys all F-actin structures in wild-type fission yeast cells⁴⁸ (Figure 2G) had a negligible effect on the F-actin bundles and puncta. To test the activity of THATCH-ΔUSH dimers at the native expression level, we used the 2A peptide to create strains where THATCH-ΔUSH dimers and the rest of End4p (containing or not containing the USH domain) were expressed in the same cell with equal stoichiometry. F-actin bundles and puncta only formed in cells with End4 constructs that did not contain USHs (Figures 2H and 2I). These data demonstrate that the USH is a potent inhibitor of the F-actin-binding ability of the THATCH-ΔUSH dimer and that constitutively active THATCH, even at the native expression level, creates overly stable F-actin structures in fission yeast cells. Combining these results with data from in vitro force spectroscopy (Figure 1), our data suggest that force >7 pN turns THATCH from an “OFF” state, where its affinity to F-actin is negligible, to an “ON” state, where THATCH strongly binds and possibly bundles F-actin (Figure 2J). Therefore, mechanical forces on THATCH directly regulate its affinity to F-actin by controlling the unfolding/refolding of the USH.

Figure 2. USH inhibits the binding of THATCH to F-actin in vivo.

(A) mScarlet-I tagged End4p localizes to endocytic sites in fission yeast cells. Endocytic sites are enriched at cell tips and the division plane.

(B) THATCH is disconnected from End4p by a self-cleaving 2A peptide, and fragments of End4p are visualized by two different fluorescent tags. The N-terminal (N-term) fragment (tagged by mScarlet-I) forms puncta at cell tips and the division plane, and isolated THATCH (tagged by mEGFP) is diffusive in the cytoplasm, indicating a lack of affinity to F-actin.

(C) mEGFP-THATCH-ΔUSH is overexpressed in fission yeast cells where wild-type End4p is tagged with mScarlet-I. The overexpression of mEGFP-THATCH-ΔUSH caused the formation of large puncta and bundles (arrows). Wild-type End4p is recruited to the puncta but not the bundles. See also Video S1.

(D) mEGFP-THATCH-ΔUSH is overexpressed in fission yeast cells where actin crosslinker Fim1p is tagged with mScarlet-I. Fim1p co-localizes with both the puncta and bundle.

(E) mEGFP-THATCH-ΔUSH is overexpressed in fission yeast cells where actin crosslinker Ain1p is tagged with mScarlet-I. Ain1p shows no co-localization with either the puncta or bundle.

(F) mEGFP-THATCH-ΔUSH-R1093G is overexpressed in fission yeast cells where wild-type End4p is tagged with mScarlet-I. The R1093G mutation prevents the dimerization of THATCH. Monomeric THATCH is diffusive in the cytoplasm without forming puncta or bundles.

(G) 50 μM Latrunculin A (LatA) was used to disassemble F-actin in two populations of fission yeast cells. Cells enclosed by a dotted line have overexpression of mEGFP-THATCH-ΔUSH, and cells enclosed by a solid line express Fim1p-mEGFP. Note that puncta and bundles are insensitive to LatA treatment, whereas Fim1p became diffusive in the cytoplasm.

(H) A self-cleaving 2A peptide was inserted after USH to disconnect THATCH-ΔUSH (tagged by mEGFP) from the N-term fragment of End4p (tagged by mScarlet-I). The USH inhibited the formation of puncta and bundles by THATCH-ΔUSH even when it was physically disconnected from THATCH-ΔUSH.

(I) A self-cleaving 2A peptide was inserted after the USH to disconnect THATCH-ΔUSH (tagged by mEGFP) from the N-term fragment of End4p (tagged by mScarlet-I), and the USH is deleted from End4p. The absence of the USH led to constitutively active THATCH-ΔUSH, which drove the formation of puncta and bundles.

(J) THATCH exists in two states, each with different affinities to F-actin. THATCH_OFF contains a folded USH and has negligible affinity to F-actin, and THATCH_ON has an unfolded USH and strongly binds F-actin. Note that THATCH is drawn as a monomer for simplicity, but dimerization of THATCH is needed for binding to F-actin.

Images are maximum intensity projections of the whole fission yeast cells. The schematic under each image indicates the modifications on End4p. Proteins are expressed at the endogenous level unless labeled OE (overexpression).

The scale bar in (A) applies to all images except (D) and (E), 5 μm. The scale bar in (D) applies to (D) and (E), 4 μm.

A rod-like domain in End4p, Rend, buffers the force on THATCH to terminate force transmission

The peak force on THATCH in vivo is ~19 pN and large enough to unfurl the USH.⁷ To turn THATCH OFF, the tension of the USH must drop below 7 pN. This cannot be achieved by unfolding the remaining four-helix bundle, which is stable up to at least ~25 pN (Figures 1D, S1B, and S1C), or disassembling F-actin because F-actin bound by active THATCH is resistant to disassembly (Figures 2C, 2G, and 2I). THATCH binds to F-actin with a dwell time >40 s under 20 pN tension in vitro (Figure 3C in Owen et al.²²), thus ruling out direct dissociation between THATCH and F-actin during CME, where actin is assembled and disassembled within 10 s.^3,4 A plausible way to reduce tension is to have other domains that unfold under tension so that force on THATCH could be buffered. In talin, 13 rod domains (also called R domains) have been found to unfold with varying stabilities in the 5–25 pN range. The unfolding of R domains buffers tension on talin to 5–10 pN despite significant changes in the total length of talin or fluctuations in talin extension.^49–51 While comparing End4p sequences, we found a structured domain located in front of THATCH (Figures 1A and S1A). A previous study noticed a similarity between this domain and talin THATCH (Figure 5 in Chen and Pollard⁴⁸). This domain is predicted to be a five-helix-bundle by both AlphaFold and RaptorX (Figure 3A).^52,53 We verified the all-helical content of this domain through circular dichroism (Figure S2A). This domain is conserved in all species that contain End4p/Sla2/HIP1R at the structural level but not the sequence level (Figures S2B and S2C), and the predicted structure is most similar to that of R12 from talin, which is also the R domain immediately before THATCH (Figure 3A).^27,49 The five a helices of Rend are not organized in the same way as THATCH’s helices. While consecutive helices of THATCH contact each other, the first helix of Rend does not contact helix 2 but rather helices 3 and 4, making the first α helix of Rend less isolated than THATCH’s first helix (USH) (Figure 3B) and, therefore, potentially less stable when under force. To test this hypothesis, we pulled on purified Rend attached at its N and C termini. We observed an all-or-none unfolding (unfolding extension change = 59 nm) when the pulling force reached 14.8 ± 1.2 pN (mean ± SD) instead of partial unfolding. Rend refolds with a strong hysteresis when the force drops to ~2 pN (Figures 3C, 3D, and S2D). Note that force on Rend could be applied in different configurations than along the N to C termini via putative Rend binding partners. However, without further information on those partners, we did not study Rend unfolding in other configurations. The mechanical properties of Rend indicate that tension before THATCH can only briefly exceed 15 pN before Rend unfolds and that tension on End4p cannot increase again before it first drops below ~2 pN for Rend to refold. Consequently, since the USH refolds when the tension decreases below 7 pN, THATCH is turned OFF and disassociates from F-actin before Rend refolds (Figure 3E). As a result, Rend limits the transmission of tension to 15 pN and effectively turns THATCH OFF once this limit is reached. To test this hypothesis in vivo, we used a recently calibrated coiled-coil force sensor,⁷ cc-14 pN (Figure S3), which opens at 13.9 ± 2.8 pN (mean ± SD), to show that the peak force before Rend is ~16 pN in vivo (Figures 3F–3H), in agreement with our force buffering model.

Figure 3. Rend unfolds at 15 pN to terminate force transmission from THATCH.

(A) Predicted structure of Rend (mustard) docked to the crystal structure of R12 from talin (cyan, PDB: 3DYJ) in two orthogonal views. Rend most resembles R12 among the 13 R domains of talin (root-mean-square deviation [RMSD]: 2.2 Å).

(B) The order of connections of the five α helices are different in Rend than in THATCH. α Helices are represented as numbered circles. Solid lines indicate connections on top of the α helices, and dotted lines indicate connections at the bottom.

(C) Representative FECs of Rend under tension. Two consecutive pulling events from the same molecule are shown to demonstrate the complete refolding of Rend after relaxation. See a full analysis of Rend unfolding in Figure S2.

(D) Rend unfolds at 14.8 ± 1.2 pN (mean ± SD) and refolds at <2 pN, with an unfolding distance of 59 nm. Tensile force on protein domains directly connected to Rend is therefore buffered at 14.8 pN unless the full extension of unfolded Rend is achieved.

(E) Different states of Rend-THATCH domains are listed from top to bottom, and the tension between Rend and THATCH is indicated by “F.” (i) Forces from F-actin are transmitted through THATCH_ON to Rend until they transiently reach a peak magnitude of 18–20 pN.⁷ (ii) Force reaches the unfolding threshold of Rend, and Rend unfolds, and tension between Rend and THATCH_ON starts to drop. (iii) The long unfolding distance of Rend (59 nm) and the fast refolding of the USH prevent its full extension in CME, and tension between Rend and THATCH continues to drop. The USH refolds when tension drops below 7 pN, and THATCH unbinds from F-actin. (iv) Rend refolds after tension drops below 2 pN. (v) End4p with refolded Rend and THATCH could participate in the force transmission cycle again. The mechanical unfolding of Rend inactivates THATCH and buffers force transmission before Rend. Pulling forces are indicated by arrows.

(F) A calibrated coiled coil (cc-14pN) is inserted into End4p before Rend, and End4p is fluorescently tagged. The formation of End4p droplets indicates that force before Rend is large enough to unfold cc-14pN.

(G) A calibrated coiled coil (cc-18pN) is inserted into End4p before Rend, and End4p is fluorescently tagged. The absence of End4p droplets indicates that the peak force before Rend is below 18pN.

(H) cc-14pN is tagged to the C terminus of fluorescently labeled End4p. cc-14pN did not drive the formation of End4p droplets in the absence of force. The schematic under each image indicates the modifications on End4p. Proteins are expressed at the endogenous level.

The scale bar in (F) applies to (F)–(H), 5 μm.

Rend forms protein condensates at the plasma membrane to initiate force transmission

We also discovered that Rend has an additional function in organizing End4p molecules in vivo. End4p constructs that are disconnected from THATCH formed large membrane-associated puncta in fission yeast cells, whereas End4p constructs disconnected from Rend-THATCH did not (Figures 4A and 4B). End4p lacking Rend (end4-ΔRend) was more cytoplasmic than wild-type End4p (Figures 4C, 4D, and 4F) and had impaired localization to endocytic sites (Figure 4G), suggesting that Rend increases the local concentration of End4p. Rend deletion also slowed cell growth (Figure S4A). We saw similar defects after replacing Rend with R3 from talin (Figures 4E–4G and S4A). To test if Rend could cause protein puncta formation independent of other domains from End4p, we compared the localization of isolated Rend constructs in fission yeast cells. The overexpressed Rend dimer was diffusive in the cytoplasm (Figure 4H). A membrane localization signal (single copy of the PH domain from fission yeast Plc1p) was sufficient to induce the formation of Rend puncta from dimerized, but not monomeric, Rend (Figures 4I and 4J). In the same vein, the attachment of dimerized, but not monomeric, Rend to the C terminus of a BAR-domain-containing membrane protein, Pil1p, reshaped the linear furrows (eisosomes) into spherical puncta (Figures S4B–S4E). These data demonstrate that Rend can mediate protein puncta formation at the plasma membrane in an autonomous fashion. In vivo protein condensation is often under phospho-regulation.^54,55 Since Rend contains a threonine (T841) that loosely conforms to the recognition motif of Ark1/Prk1 family kinase^56–58 (Figures S4F–S4H; Video S2), we wondered if the phosphorylation state of T841 could control the formation of puncta in vivo. We tested this hypothesis by constructing a phospho-mimetic and a non-phosphorylatable Rend (T841E and T841A, respectively). The phospho-mimetic Rend did not form puncta, while the non-phosphorylatable Rend did (Figures 4K and 4L), demonstrating that the phosphorylation state of T841 can control the assembly and disassembly of Rend and End4p. The ability to form protein puncta is not a general property of all R domains, as only R12, but not R3, from talin formed protein puncta when dimerized and attached to a PH domain (Figures 4M and 4N). Taken together, our data demonstrate that membrane localization of Rend leads to strong self-association of Rend that increases its local concentration, possibly through condensation (Figure 4O). Consequently, in End4p, Rend promotes the activation of THATCH in at least two ways: (1) increased concentration of THATCH increased the number of binding events to F-actin even in the low-affinity OFF state and (2) because the motion of Rend is restricted after self-association, force from an activated THATCH of an End4p molecule will be channeled through its dimerization tail to unfold and activate the THATCH of another End4p molecule (Figures 1E and 4O). By linking THATCH to Rend, THATCH is only activated at the endocytic site when a sufficient amount of End4p molecules are recruited, and Rend spatially coordinates the activation of THATCH.

Figure 4. Membrane proximity drives the formation of Rend puncta.

(A and B) Rend-THATCH (A) or THATCH (B) is disconnected from End4p by a self-cleaving 2A peptide, and fragments of End4p are visualized by two different fluorescent tags. The N-term fragment (tagged by mScarlet-I) forms puncta at cell tips and the division plane only when it contains Rend, and isolated Rend-THATCH (tagged by mEGFP) is diffusive in the cytoplasm.

(C) Wild-type End4p with a fluorescent tag.

(D) Rend is deleted from fluorescently tagged End4p. End4p-ΔRend has an increased partition in the cytoplasm.

(E) Rend is replaced by R3 from talin, and End4p is fluorescently tagged. Replacement of Rend by talin R3 increased the partition of End4p in the cytoplasm.

(F) Quantification of the cytoplasmic ratio of End4p for (C)–(E). The deletion of Rend increased the End4p cytoplasmic ratio, and talin R3 could not rescue this phenotype. Mean with SD. n = 5, 12, and 13. Each dot represents a field of view with >10 cells. Kruskal-Wallis test and Dunn’s test for pairwise comparison. *p < 0.05 and **p < 0.005.

(G) Temporal evolution of the number of End4p molecules at endocytic patches as determined by patch tracking. The deletion of Rend, or the replacement of Rend by talin R3, inhibited the accumulation of End4p molecules at endocytic sites.

(H) Overexpressed Rend dimer is diffusive in fission yeast cytoplasm. Rend dimerization is mediated by the latch from the End4p THATCH domain.

(I) Overexpressed Rend dimer with a membrane localizing signal (PH domain) led to the formation of droplets in the cytoplasm.

(J) Overexpressed monomeric PH-Rend is diffusive in the cytoplasm and did not form droplets. The R1093G mutation in the latch from End4p THATCH prevents dimerization.

(K) T841A phospho-dead mutation on overexpressed PH-Rend had no influence on the formation of droplets.

(L) T841E phospho-mimetic mutation on overexpressed PH-Rend inhibited the formation of droplets.

(M) Overexpressed PH-R3 dimer did not form droplets in fission yeast cells.

(N) Overexpressed PH-R12 dimer formed droplets in fission yeast cells.

(O) Rend_OFF is converted to Rend_ON by localization to the membrane. Membrane-induced formation of Rend puncta primes the binding between THATCH and F-actin in two ways: (1) locally increasing the concentration of THATCH and (2) restricting the movement of THATCH so that force from stochastically activated THATCH can be used to activate nearby THATCH.

The scale bar in (A) applies to all yeast cell images, 5 μm.

Direct connection and mutual regulation of Rend and THATCH form a mechanical circuit to coordinate force transmission during CME

We propose that two domains of End4p, Rend and THATCH, form a mechanical circuit that regulates the transmission of force from F-actin to the plasma membrane during endocytosis (Figure 5A). Both Rend and THATCH have an ON and an OFF state (Rend_ON and Rend_OFF and THATCH_ON and THATCH_OFF, respectively). Rend is turned ON when it is folded and makes puncta at the membrane and turned OFF when it is unfolded by >15 pN force. THATCH is turned ON when force >7 pN unfolds the USH, increasing its affinity to actin, and turned OFF when force drops <7 pN and the USH refolds with the rest of THATCH into a five-helix bundle. A cycle of force transmission involves the sequential changes in the ON and OFF states of Rend and THATCH, where the magnitude of force triggers the transition between states. To test this model, we made constructs that mimicked Rend_ON, Rend_OFF, THATCH_ON, and THATCH_OFF. Our model predicts that THATCH deletion results in an End4p with a constitutive THATCH_OFF, unable to apply force on Rend and, therefore, unable to remove Rend from the membrane. This was confirmed by experiments showing large aggregates in end4-ΔTHATCH mutants (Figure 5B). A similar phenotype can also be observed when force on Rend is removed by disconnecting Rend and THATCH (Figure 2B), preventing the dimerization of THATCH (Figure S6 in Ren et al.⁷), or preventing actin assembly with LatA (Figure S5 in Chen and Pollard⁴⁸). By replacing Rend with a domain able to multimerize but without mechanosensitive properties (Figure S5), force through THATCH was able to remove Rend puncta from the membrane but could not dissolve the puncta, resulting in Rend droplets in the cytoplasm (Figure 5C). Note that membrane localization of Rend is still required for the formation of End4p puncta (compare Figures 5C and S5D). Deletion of the USH mimicking a constitutive THATCH_ON resulted in the formation of large End4p droplets in the cytoplasm, consistent with our model prediction (Figure 5D). Note that the inclusion of the Rend domain to THATCH_ON transformed the linear F-actin structures into spherical ones (compare Figures 5D and 2C). The partial unfolding of Rend (Figures 5E and S6) and the deletion of Rend (Figure 5F) both reduced the recruitment of End4p to endocytic sites, thereby impeding End4p from entering the force transmission cycle. In summary, a faulty circuit by Rend and THATCH results in a permanent association with the membrane (Figure 5B) or F-actin (Figures 5C and 5D) or failure to associate with either the membrane or F-actin (Figures 5E and 5F).

Figure 5. Rend and THATCH form a mechanical circuit for force transmission.

(A) Force transmission cycle by Rend and THATCH in CME. Five states are depicted to indicate the sequential transition of the ON and OFF states of Rend and THATCH. The cycle starts from the transition of state #5 to state #1 and proceeds irreversibly. State #1 (Rend_ON, THATCH_OFF): End4p is localized to the membrane, and Rend forms puncta. State #2 (Rend_ON, THATCH_ON): THATCH is activated and stably binds to F-actin. Forces are transmitted through THATCH to Rend and further into the endocytic coat to deform the membrane. State #3 (Rend_OFF, THATCH_ON): force on Rend exceeds 15 pN, and Rend unfolds. The tension between Rend and THATCH drops. State #4 (Rend_OFF, THATCH_OFF): force on THATCH drops below 7 pN. THATCH refolds and unbinds F-actin. The tension between Rend and THATCH continues to drop. State #5 (Rend_ON, THATCH_OFF): Rend refolds, and End4p is ready to enter another force transmission cycle. Note that states #2, #3, and #4 are transient states. End4p domains are simplified and not drawn to scale.

(B) Deletion of THATCH led to the formation of large End4p puncta at the membrane because THATCH activation is needed for the transition from state #1 to state #2.

(C) Replacing Rend with an open coiled coil mimics constitutive Rend_ON, where the intermolecular interaction between End4p cannot be terminated by force. This led to the formation of End4p droplets in the cytoplasm.

(D) Deletion of USH led to constitutive THATCH_ON, which led to the formation of End4p droplets in the cytoplasm. (C) and (D) had similar phenotypes because constitutive Rend_ON leads to THATCH_ON, and states #2 and #3 are transient states.

(E) Replacing Rend with a partial open version of Rend prevents the refolding of Rend and leads to a higher cytoplasmic ratio of End4p and less End4p at endocytic sites (see also Figure S6).

(F) Deletion of Rend prevents the formation of Rend puncta at the membrane and results in a higher cytoplasmic ratio of End4p and less End4p at endocytic sites (see also Figures 4D and 4E). (E) and (F) had similar phenotypes because state #4 is a transient state.

(G) Mutual regulation of Rend and THATCH forms a circuit. Membrane localization of Rend primes the activation of THATCH, and force through THATCH in-activates Rend at 15 pN.

(H and I) Mathematical modeling of THATCH and Rend shows that THATCH catch bond and Rend mechanosensitive unfolding buffer the forces transmitted by actin during endocytosis.

(H) Force response of the Rend-THATCH connection when the input force is proportional to the amount of F-actin present at endocytic sites. The input force is transmitted from THATCH to Rend, and the output force is recorded at the membrane-proximal end of Rend. The transmitted force is defined as the product of the input force and the ratio between actin-bound THATCH to total actin. Gray line: input force; blue line: catch-bond THATCH; orange line: slip-bond THATCH; green box: regime where the transmitted force is buffered.

(I) Force transmitted by the Rend-THATCH connection as a function of a constant input force produced by actin. Blue line: catch-bond THATCH; orange line: slip-bond THATCH; blue dotted line: catch-bond THATCH in RendΔ mutant; vertical dashed line: Rend unfolding force; vertical dotted-dashed line: THATCH USH unfolding force (or catch-bond threshold force); green box: regime where the transmitted force is buffered to vary less than 27%, as the input force varies over ~3-fold.

The scale bar in (B) applies to (B)–(F), 5 μm.

To quantitatively probe the functionality of the Rend-THATCH circuit for force transmission, we developed a mathematical model for the response of End4p force transmission to varying input forces (Figure 5G; Tables S2–S4). We restricted our study to the life cycle of a fixed pool of membrane-bound End4p and actin filaments at an endocytic site (states 1–4 in Figure 5A). In this model, we considered the Rend domain as folded (Rend_ON) or unfolded (Rend_OFF) and the THATCH domain as bound in a catch bond to an actin filament (THATCH_ON) or not (THATCH_OFF). Before binding to any actin filament, an End4p molecule is in the Rend_ON&THATCH_OFF state (state 1 in Figure 5). It captures an actin filament according to mass action kinetics, and THATCH immediately becomes a catch bond (Rend_ON&THATCH_ON) (state 2 in Figure 5), as we assume actin filaments continuously produce force (in other words, End4p acts as a clutch). End4p can detach from actin filaments as a function of force, following catch-bond and slip-bond behavior (we used realistic parameters inspired by the literature^22,59,60). The Rend domain unfolds (Rend_OFF&THATCH_ON) according to a rate following a step function that transitions around 15 pN as measured in this paper (state 3 in Figure 5). At this point, End4p does not bear force anymore since it is buffered by the unfolded Rend, and the actin filament is released (Rend_OFF&THATCH_OFF) (state 4 in Figure 5). Rend refolds rapidly (Rend_ON&THATCH_OFF) (state 1 in Figure 5) and is ready to restart the cycle and bind a new actin filament. The force transmitted to the endocytic membrane is calculated as the force produced by actin times the ratio of End4p attached to the actin filament in a force-bearing configuration (i.e., Rend_ON&THATCH_ON).

We simulated the model assuming the input force produced by the actin filaments is proportional to the amount of actin at endocytic sites, i.e., the input force from F-actin followed a smooth ramp of force from 0 to 20 pN before it decreased (Figure 5H). The transmitted force rapidly plateaued for most of the time course, showing that the force is buffered. Changing the type of bond (slip bond or catch bond) between THATCH and F-actin did not qualitatively influence the buffering effect, demonstrating that it is the Rend-THATCH connection that buffers the force transmission from F-actin toward the membrane. To better understand the buffering response, we performed simulations using a step function of different magnitudes as input force (Figure 5I). In all cases, the system reached a steady state in less than 1 s, and the force transmitted at steady state varied non-monotonically with the input force (Figure 5I). Slip-bond THATCH-actin binding led to very inefficient force transmission that became worse with increasing forces. In contrast, catch-bond THATCH-actin binding allowed for force buffering at intermediate forces and sustained linear force transmission for higher forces, with an efficiency plateauing around 30% for forces above Rend’s unfolding force threshold. Our simulations also showed that Rend unfolding was critical for force buffering since without Rend unfolding (e.g., in RendΔ mutants), the transmitted force increased with increasing input force in the case of a catch-bond THATCH-actin connection. Our data show that combining a catch-bond THATCH and Rend unfolding allows End4p to exquisitely buffer the transmitted force over a large range of input forces (Figure S9). For input forces between 9 and 30 pN (i.e., a ~3-fold range), the transmitted force is buffered between 4.8 and 8.4 pN (i.e., within ~25% of an average value of 6.6 pN). Therefore, our simulations demonstrate that the Rend-THATCH connection, by combining the catch-bond mechanism and mechanical unfolding, filters the stochastic (i.e., noisy) input force from endocytic actin assembly to a sustained and virtually constant force transmission to the membrane.

DISCUSSION

The mechanical stabilities of R domains from talin have been characterized in great detail,^{23,49–51,61} and the formation of catch bonds between talin’s THATCH or other structurally similar five-helix bundles and F-actin have been investigated in comparable depth.^{18,19,22,25,26} However, the importance of the direct connection between R domains and THATCH in the initiation, transmission, and termination of mechanical forces was not well understood (Figure 5A). In this study, we identified an R domain in End4p, Rend, and found that it is critical for promoting the binding between THATCH and F-actin to initiate force transmission, spatially coordinating force transmission, and terminating force transmission after reaching a pre-determined magnitude. Thus, the mechanical circuit formed by Rend and THATCH ensures robust force transmission during CME.

Key parameters in building a mechanical circuit

Several factors are crucial for building a functioning circuit using Rend and THATCH. Because forces are transmitted from THATCH to Rend, the mechanical stability of Rend needs to be higher than that of THATCH to allow stable force transmission toward the membrane (Figures 1D and 3C). A low mechanical stability of Rend turns THATCH OFF prematurely. Replacing Rend with R3 from talin, a domain that unfolds at 5 pN,^49,62 led to slow cell growth (Figure S4). Moreover, R3 failed to enrich THATCH near the membrane (Figure 4E), resulting in 40% less End4p molecules at endocytic sites (Figure 4G). We unsuccessfully tried to replace Rend with R12 from talin, a domain that unfolds at an unknown value between 10 and 20 pN⁴⁹ and forms puncta near membranes (Figure 4N). We suspect that an overly stable R domain is detrimental to CME because both R and THATCH would be locked in the ON state (Figures 5C and 5D), blocking endocytosis and possibly resulting in lethality.

The hysteresis of Rend and the unfolding extensions of Rend and THATCH are critical to the irreversibility of the force transmission cycle (Figures 3C and 3D). The unfolding of the USH (14 nm) could be realized during CME, and this change in distance well matches the difference in averaged HIP1R length (10 nm) as revealed by FerriTag in electron microscopy (EM), where extended HIP1R molecules are found close to the tip of the endocytic pit, consistent with the idea that force through THATCH is mainly transmitted to the tip region for membrane invagination.⁶³ In contrast, EM data of endocytic sites in yeast show that Rend may not be fully extended because its unfolding extension (59 nm) would stretch End4p beyond the actin meshwork around the endocytic structure.^64,65 The fast refolding of the USH prevents the full stretching of Rend (Figure 1D). Because Rend unfolds in an all-or-none fashion and refolds when the force on Rend vanishes (<2 pN), the drop in tension in the Rend and THATCH connection cannot be reversed before THATCH detaches from F-actin (Figure 3E). Therefore, a higher refolding force for the USH (7 pN) than for Rend (<2 pN) prevents the re-engagement of THATCH to F-actin once sufficient force transmission has been achieved (Figure 5A).

Our in vivo force measurement shows that the peak force before THATCH is ~19 pN, which is large enough to be transmitted by a catch-bond THATCH.⁷ The peak force measured in vivo on the N terminus of Rend is ~16 pN, in strong agreement with the Rend’s unfolding force measured with optical tweezers (15 pN) (Figure 3F) and higher than the peak force measured in vivo on End4p close to the membrane and the clathrin lattice (~9 and ~11 pN, respectively).⁷ Our results strongly suggest that the mechanical stabilities of Rend and THATCH are likely tuned by evolution to match their roles in CME. If this is true, then we predict that the mechanical stabilities of Rend and THATCH in homologs of End4p in budding yeast (Sla2) and mammals (HIP1R) are lower, as less mechanical force is needed for CME in these species compared to fission yeast.^4,33,66,67 In addition, the unfolding force of Rend in these homologs is expected to be higher than the unfolding force of THATCH, which is expected to be higher than the refolding force of Rend for the mechanical buffering circuit to function. Recent work by Owen et al. hints that the unfolding of the USH from talin’s THATCH might be around 5 pN,²² which makes us speculate that for talin’s R12-THATCH to operate as a force buffering circuit similar to End4p’s Rend-THATCH, the in vivo force before talin’s THATCH is in the 5–10 pN range, considering R12 unfolds between 10 and 20 pN.⁴⁹

Protein condensate formation at the membrane initiates force transmission, and large mechanical forces dissolve condensates

The ability of Rend to form puncta close to the membrane offers a simple mechanism to initiate the binding between THATCH and F-actin by locally increasing its concentration and rerouting force transmission (Figure 4O). The direct and indirect membrane localization domains of End4p likely regulate the formation of Rend puncta during CME.^34,35,39 The spherical droplets formed by dimerized Rend close to the membrane are suggestive of a liquid-like material property (Figures 4I and S4C). Cell adhesion proteins, including talin and vinculin, have been recently shown to undergo membrane-induced two-dimensional (2D) phase separation,^68,69 possibly promoted by non-coding mRNAs,⁷⁰ and several endocytic proteins indeed form condensates to either initiate CME or cause membrane bending and scission.^70–73 We have found that several endocytic proteins, including early coat proteins Ede1p, Yap18p, and Syp1p and late coat protein End3p, are recruited to the droplets formed by End4p, while late coat proteins Ent1p, Shd1p, and Pan1p are not (Figure S7). Determining how Rend puncta contribute to membrane deformation will require further studies. Whether or not they form bona fide phase-separated condensates, the self-association of Rend close to the membrane increases its local concentration and primes the binding of THATCH to F-actin (Figure 4O). This feature is reminiscent of the transmembrane signaling by other membrane receptors, such as integrin, immune receptors, and receptor tyrosine kinases (RTKs), whose clustering is a prerequisite to their activation for downstream signaling.^11,55,74,75 The neighboring THATCH and F-actin binding may be enhanced in a cooperative manner, as has been shown for the αE-catenin actin-binding domain.¹⁹ We speculate that the cooperativities of the Rend and THATCH domains help accelerate the binding between THATCH and F-actin to achieve a switch-like engagement between the endocytic coat and the F-actin meshwork. Moreover, a layer of self-associating Rend spatially coordinates the transmission of forces toward the endocytic coat and filters the F-actin fluctuations in the outmost layer (Figures 5H and 5I). Additional layers made by protein-protein interactions likely exist in the endocytic coat, of which the clathrin lattice is an obvious candidate. Forces transmitted from F-actin to the endocytic coat are thus collected and redistributed through several layers before reaching the membrane.⁷

The unfolding of Rend by mechanical forces retrieves Rend from the membrane, and the putative phosphorylation of Rend by Ark/Prk kinases at T841 may be involved in dissolving Rend puncta during CME, although mechanical forces seem to play a major role^56,76 (Figure S4H). The regulation of Rend puncta during CME by both mechanical and chemical signals is an interesting direction for future exploration, and there may be regulation crosstalk between mechanical forces, protein crowding, and kinase specificity/activity.^77,78 The unfolding of Rend causes a localized mechanical disconnection between the endocytic coat and the F-actin right at the tip of the endocytic pit, which could be used to release the energy stored in crosslinked F-actin to pinch off the endocytic vesicle after sufficient energy accumulation.^64,79 Quantitative modeling of this process that adds spatial dimensions to the model presented in this study (Figure 5) will be informative in explaining the later stages of CME.

Cross-regulation of mechanosensitive domains may be a conserved mechanism for robust force transmission

The discovery of Rend on End4p suggests a possible evolutionary connection between End4p/Sla2/HIP1R and talin. Both are multidomain linear molecules that bind the membrane at the N terminus, contain one or several R domains in the middle, and bind to F-actin at the C terminus through THATCH (Figure S8A). In addition, we showed in this paper that unfolded Rend, but not folded Rend, binds to vinculin head domain 1 (Vd1) in fission yeast cells (Figures S8B–S8G). Unknown binding partners of Rend before and after unfolding may help regulate CME, analogous to force-dependent protein binding/unbinding to talin R domains.^27,50 Because endocytosis likely predates multicellularity, cells may have coopted the endocytic machinery to build cell-matrix or cell-cell connections during evolution. Consistent with this idea, in a recent lab evolution experiment, several genes related to endocytosis and cytoskeletal organization were shown to be implicated in the transition into a multicellular life cycle.⁸⁰ One fungus species, Allomyces macrogynus, contains both HIP1R and talin, and structural predictions suggest they respectively contain 1 and 12 R domains besides THATCH.⁸¹ It will be informative to explore if R domain duplication or deletion led to the apparent functional homology between End4p/Sla2/HIP1R and talin. It will also be meaningful to check if similar mechanical circuits represent a general solution for robust mechanotransduction in vivo.

Limitations of the study

In vivo data supporting the existence of different states of the mechanical circuit are snapshots of genetically edited yeast cells and lack kinetic information. Rend domains close to the membrane may display emergent behaviors that are not captured by in vitro single-molecule assays, and we do not have direct evidence to show how Rend condensates promote the activation of THATCH. Live-cell single-molecule tracking of mutant End4p molecules may help reveal the dynamics of this mechanical circuit.

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Julien Berro (Julien.berro@yale.edu).

Materials availability

Further information and requests for yeast strains and plasmids should be directed to and will be fulfilled by the lead contact.

Data and code availability

• All of the data reported in this paper will be shared by the lead contact upon request.

• Reactions, parameters values, and differential equations used to generate the model are included in Tables S2–S4.

• Any additional information required to reanalyze the data reported in this work paper is available from the lead contact upon request.

STAR★METHODS

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Yeast strains and media

The S. pombe strains used in this investigation are detailed in Table S1. These strains were constructed through CRISPR-mediated genome editing as outlined in our prior publication⁸² and confirmed through sequencing of the colony PCR products.⁸³ Fission yeast cells were cultivated in YE5S medium (Yeast Extract supplemented with 0.225 g/L each of lysine, uracil, adenine, histidine, and leucine) and imaged on gelatin pad made with EMM5S medium (Edinburgh Minimum media supplemented with 0.225 g/L each of lysine, uracil, adenine, histidine, and leucine). Yeast cells were incubated at 32°C with continuous shaking at 200 rpm overnight, allowing them to reach the exponential phase with OD_595nm values ranging from 0.3 to 0.5.

Growth assay

10 μL of yeast cells from an overnight culture were diluted to OD_595nm = 0.1. Subsequently, serial dilutions of cells were prepared (10⁰, 10¹, 10², and 10³) and spotted onto YE5S plates. The plates were then placed in either a 32-degree or 37-degree incubator and incubated for 48 h before imaged.

METHOD DETAILS

Protein expression, purification, and sequences

Coding sequences of Rend and THATCH were amplified from fission yeast (S. pombe) genome through colony PCR and cloned into pGEX-6P-1 vectors (Sigma-Aldrich) by Gibson cloning (New England BioLabs). The coding sequence for cc-14pN was synthesized through IDT and codon optimized for S. pombe. Sequences were confirmed and introduced into BL21 (DE3) E. coli (New England BioLabs) for protein expression. Proteins were expressed with GFP fusion at the N terminus, purified by attachment to Glutathione Sepharose 4B beads (Cytiva), and the GST tag was cleaved by PreScission Protease (Sigma-Aldrich). Avi-tag in purified 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 magnesium acetate, 10 mM ATP, and 50 μM d-biotin (Avidity) at 4°C overnight. The amino acid sequences of three protein constructs used for the single-molecule pulling experiments are shown below.

THATCH_Shear

GCSGSNSGLLNAPGENIEELVDNQLAETAQAIQQAILRLQNIAAKPKDDSLSPSELQVHDSLLSASIAITEAIARLIKAATASQAEIVAQGRGSSSRGAFYKKHNRWTEGLISAAKAVARATTTLIETADGVVNGTSSFEHLIVACNGVSAATAQLVAASRVKANFASKVQDHLEDAAKAVTEACKALVRQVESVALKAKEVQHEDFSSGGSGNGGSGSGLNDIFEAQKIEWHE

THATCH_Unzip

GSGLNDIFEAQKIEWHEGGSGNGGSGLLNAPGENIEELVDNQLAETAQAIQQAILRLQNIAAKPKDDSLSPSELQVHDSLLSASIAITEAIARLIKAATASQAEIVAQGRGSSSRGAFYKKHNRWTEGLISAAKAVARATTTLIETADGVVNGTSSFEHLIVACNGVSAATAQLVAASRVKANFGCGASKVQDHLEDAAKAVTEACKALVRQVESVALKAKEVQHEDFSS

Rend

GCSGSNSEKLDDIVDSVLATGIQRLDTSLYELDSPMHAGNQYATPEFILSTIENASNNATDFSTAFNNYFADGPNADHSEVINGVNLFSTAIYEVANNAKGLSRTTGDDQGSDRFVGLSRDLVNMAKRFLSSLFSVNTRKMDVNVKTDLVIGENIELQRYLQQLTQYSEKFLNKESENTVG GGSGNGGSGSGLNDIFEAQKIEWHE

The Avi-tag sequences and the cysteine residues crosslinking to the DNA handle are underlined.

Protein-DNA handle crosslinking

The DNA handle for protein attachment in single-molecule experiments was generated by PCR (2,260 bp) and contained a thiol group (-SH) at one end and two digoxigenin moieties at the other. Crosslinking of the DNA handle to proteins was performed as previously described.⁸⁴ In short, purified proteins were mixed with the DTDP-treated DNA handle at 50:1 M ratio in 100 mM phosphate buffer with 500 mM NaCl, pH 8.5 and incubated overnight at room temperature.

Single-molecule manipulation experiments and data analysis

All single-molecule experiments were conducted using dual-trap high-resolution optical tweezers.⁴³ DNA handle crosslinked proteins were prepared by incubating them with anti-digoxigenin antibody-coated polystyrene beads of 2.17-μm diameter (Spherotech) for 15 min. The mixture was diluted with 1 mL phosphate-buffered saline (PBS) and injected into the top channel of a microfluidic chamber. Streptavidin-coated beads of 1.86-μm diameter were injected into the bottom channel of the chamber. Both the bottom and top channels were connected to a central channel by capillary tubes, and the beads were trapped in the central channel. Once one bead from each type was trapped, a single protein was tethered between them by bringing the two beads close. The tethered molecule was subjected to pulling and relaxation by increasing or decreasing the trap separation at 10 nm/s, respectively. The optical tweezers experiment was conducted in PBS at 23 (±1) °C. The buffer in the central channel contained the PBS buffer supplemented with an oxygen scavenging system as described elsewhere. The data were processed by MATLAB codes as described elsewhere, and the unfolding forces were determined from the force-extension curves. We modeled the unfolded polypeptide and the DNA handle with a worm-like chain model for a semi-flexible chain.⁸⁵ The stretching force $F$ and the entropic energy $E$ of the polymer chain are related to its extension $x$, contour length $L$, and persistence length $P$ by the follow formula

$$F=\frac{k_{B}T}{P}\left[\frac{1}{4{\left(1-\frac{x}{L}\right)}^2}+\frac{x}{L}-\frac{1}{4}\right].$$ (Equation 1)

The persistence length was chosen as 40 nm for DNA and 0.6 nm for the unfolded polypeptide. The contour length of a polypeptide was calculated from its number of amino acids with 0.365 nm per amino acid. The extension of a protein $x_P$ consists of the contributions from its folded or structured portion and the unfolded portion. The folded portion of the protein was modeled as a rigid body with an intrinsic length $h$, which was determined from the protein structure. The extension of the unfolded portion $x_{uf}$ was modeled by Equation 1. Therefore, the extension of a single protein as a function of its tension $F$ can be expressed as

$$x_P=h+x_{uf}.$$ (Equation 2)

The extension of a protein-DNA tether $x_{PD}$ is the sum of the extensions of the protein and the DNA handle $x_{DNA}$, or,

$$x_{PD}=h+x_{uf}+x_{DNA}.$$ (Equation 3)

We used this equation to nonlinearly fit the measured force-extension curves to derive the contour lengths of unfolded polypeptides in the protein, which confirmed the structural model for the unfolded intermediate state.

Microscopy

Live cell imaging was performed on a 25% gelatin pad at room temperature, using a Nikon TiE inverted microscope equipped with a CSU-W1 Confocal Scanning Unit manufactured by Yokogawa Electric Corporation (Tokyo, Japan). The microscope was fitted with a CFI Plan Apo 100X/1.45NA Phase objective from Nikon. Image acquisition was carried out using an iXon Ultra888 EMCCD camera provided by Andor (Belfast, UK). For strains tagged with mEGFP, an excitation wavelength of 488nm from an argon-ion laser was used, and the emitted fluorescence was filtered using a single band-pass filter 510/25 in the Spectra X system. Strains tagged with mScarlet-I were excited using a 561nm argon-ion laser, and the emitted fluorescence was filtered through a single band-pass filter 575/25 in the Spectra X system. To capture images of the entire cell, 21 optical sections were collected with a section thickness of 0.5 μm, and these sections were then projected by using maximum intensity to generate 2D images. For image display and analysis, the Fiji distribution of ImageJ provided by the National Institutes of Health (NIH, USA) was used.⁸⁶

Patch tracking

Cells were loaded into six-well iBidi μslides (manufactured by iBidi in Munich, Germany) pretreated with 0.1% poly-L-lysine sourced from Peptide Institute in Osaka, Japan after EMM5S washes. The mEGFP signal was captured and employed for patch tracking. The fluorescence signal was acquired from 5 optical sections, each spaced 0.5 μm apart, and centered around the cells’ mid-plane. Imaging occurred at intervals of 1 s per frame over a span of 1 min.

Following corrections for irregular field illumination and camera-generated noise, the progress of fluorescence intensity over time was traced and quantified within the z-sum projected movies. This was accomplished using an updated version of the PatchTrackingTools toolkit designed for the Fiji distribution of ImageJ, a software package developed by the National Institutes of Health (NIH) in the USA. The resulting data were processed as follows: for each specific strain, measurements of the signal derived from individual tracks, acquired from multiple movies, were aligned, and then averaged. Custom post-processing scripts programmed in MATLAB R2019a (created by Mathworks, located in Natick, USA) were utilized for this purpose, following the methods described.¹³ The conversion of fluorescence intensity into the count of molecules was achieved using a previously established calibration curve.⁵⁹

The graphs illustrating the temporal changes in protein molecule count were produced using MATLAB, and the statistical comparison between different strains was carried out using Welch’s t test applied to the average peak molecule counts, as indicated in the accompanying table.

Modeling of the Rend and THATCH force buffering

We implemented the circuit of Figure 5B representing the different conformational states of the Rend and THATCH domains of End4p at endocytic sites. We did not include state #5 (End4p in the cytoplasm). Transitions between states followed mass action kinetics and were influenced by input forces produced by actin on End4p (Table S2). Actin input forces followed a sine over half a period to mimic forces proportional to the amount of actin present at endocytic patches over the course of endocytosis⁵⁹ (Figure 5H) or were kept constant to determine the steady state buffered force response as a function of input force (Figure 5I). The buffered force was calculated as the product between the input force and the ratio of End4p in a catch bond state with folded Rend (Rend_ON|THATCH_ON). Rate constants values were chosen according to literature or within typical ranges when no value was available. Changing parameter values did not qualitatively change the outputs of the model. Simulations were started with a fixed amount of actin and End4p unattached to actin in the Rend_ON|THATCH_OFF state, so that the total amount of End4p and actin remained constant over the course of the simulation. The model was implemented using the Simbiology plugin (6.4.1) of MATLAB (R2023a).

QUANTIFICATION AND STATISTICAL ANALYSIS

Statistical details can be found in the figure legends of Figures S4C and S4E. Each dot represents averaged value as quantified from a field of view with >10 cells. Microsoft Word was used to tabulate data. Plots and statistical analyses for cellular data were carried our using GraphPad Prism 7. Plots for single-molecule data and for modeling were performed in MathWorks MATLAB. Kruskal-Wallis test was used together with Dunn’s multiple comparison test for pairwise comparison. *: p < 0.05; **: p < 0.005; ****: p < 0.0001.

Protein cytoplasmic ratio and eisosome circularity

For fluorescently tagged End4p constructs, images of the entire cell were acquired through sectioning and projected by using the average intensity to form a 2D image. The average fluorescence intensity of the cytoplasm was calculated and divided by the average fluorescence intensity of the entire cell to give the cytoplasmic ratio. Quantification was performed using ImageJ Fiji.

Eisosome circularity was quantified by first segmenting images through fluorescence intensity to identify structures with Pi1l-mEGFP signal, and then using “Circularity” function within “Analysis” to calculate the average circularity of eisosomes. Quantification was performed using ImageJ Fiji.

In vivo force measurement with calibrated coiled-coils

Calibrated coiled-coils are inserted into different positions of End4p through genome editing at the endogenous loci. Force-dependent unfolding of coiled -coils in End4p leads to intermolecular entanglement and the formation large End4p condensates, which differ from normal End4p endocytic patches in circularity, size, and fluorescent intensity. End4p condensates are more circular, larger, and have increased fluorescent intensity than End4p patches. An ImageJ plugin was created to automatically segment images of fission yeast cells with fluorescently tagged End4p, and the occurrence of End4p condensates are used as a readout for in vivo force measurement.⁷

KEY RESOURCES TABLE.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Chemicals, peptides, and recombinant proteins
PreScission Protease	Sigma-Aldrich	GE27–0843–01
Glutathione Sepharose 4B beads	Cytiva	17075601
d-biotin	Avidity	Bio-500
polystyrene beads	Spherotech	PP10–20–10
Experimental models: Organisms/strains
See Table S1 for the list of S. pombe strains	This study	N/A
Recombinant DNA
pGEX6PI-THATCH_857–1058	This study	pJB331
pGEX6PI-R1_684–857	This study	pJB332
pGEX6PI-THATCH_857–1058-Unzip	This study	pJB341
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
Other
See Table S2 for reaction rates and forces used in this model	This study	N/A
See Table S3 for kinetic parameters used in this model	This study	N/A
See Table S4 for differential equations used in this model	This study	N/A

Highlights.

• 7 pN force is required to unfurl THATCH for stable binding to F-actin

• Rend terminates force transmission by unfolding when force exceeds 15 pN

• Dimerized Rend forms protein condensates at the cell membrane to activate THATCH

• The coupling between Rend and THATCH forms a self-regulated mechanical circuit