A MEC-2/stomatin condensate liquid-to-solid phase transition controls neuronal mechanotransduction during touch sensing

A growing body of work suggests that the material properties of biomolecular condensates ensuing from liquid–liquid phase separation change with time. How this aging process is controlled and whether the condensates with distinct material properties can have different biological functions is currently unknown. Using Caenorhabditis elegans as a model, we show that MEC-2/stomatin undergoes a rigidity phase transition from fluid-like to solid-like condensates that facilitate transport and mechanotransduction, respectively. This switch is triggered by the interaction between the SH3 domain of UNC-89 (titin/obscurin) and MEC-2. We suggest that this rigidity phase transition has a physiological role in frequency-dependent force transmission in mechanosensitive neurons during body wall touch. Our data demonstrate a function for the liquid and solid phases of MEC-2/stomatin condensates in facilitating transport or mechanotransduction, and a previously unidentified role for titin homologues in neurons.


Supplementary Material
1 Supplementary Note 1: Optical Trapping 1.1 Single trap stress relaxation of protein condensates in vitro Freshly prepared MEC-2 condensates (t=0 h) showed a rapid and almost complete relaxation, indicating that naive, liquid-like MEC-2 condensates cannot store mechanical stress.This fluid-like signature remained largely unchanged for the first 24 h after condensate formation (Extended Data Fig. 8c).Only after 48 h, MEC-2 showed a significantly slower relaxation time scale despite retaining a constant stiffness (Extended Data Fig. 8d), reminiscent of a glass transition with an age-dependent increase in viscosity (Extended Data Fig. 8e, [1]).Interestingly, we found the same viscoelastic behavior with the condensates composed of mutant MEC-2 (R385H), suggesting that MEC-2 maturation in vitro is a slow process that does not depend on the presence of arginine at position 385 (Extended Data Fig. 8f-h).
Condensates in the single trap configuration were prepared as described in the methods and settled onto the coverglass and strongly attached to it, but were prevented from wetting and thus remained in a droplet shape (Extended Data Fig. 8a).Stress relaxation tests upon step-strain indentations onto the droplets were carried out by addressing a square oscillation protocol to the trap.The bead pushing force against the droplet arose from a change in light momentum that was detected with our direct force sensor.
Upon the pushing event, the force peaked to F p and relaxed exponentially to a resting force, F 0 , which arose from surface tension (Extended Data Fig. 8b).Upon pulling, we observed a nearly symmetric peak followed by a relaxation curve, with opposite sign, due the bead being adhered to the droplet.The force-time curves were fitted with an exponential function that describes the time-dependent decay in force as follows: Parameter τ corresponds to the relaxation time constant of the Maxwell element and equals τ = η/k, where η and k are the viscosity and the stiffness of a dashpot and a spring in series [2].We observed an increase in τ for older condensates (throughout a time frame of 48 h), indicative for Maxwell fluids undergoing hardening upon maturation (aging).This is consistent with a higher ability to store mechanical stress for a longer time (see Main Text, Extended Data Fig. 8c,f), for time scales that are smaller than τ .In turn, the spring stiffness was measured as k = (F p − F 0 )/d, where d = 100 nm is the amplitude of the square oscillation.We did not observe an increase in k in between the different conditions tested.
Thus, droplet ageing did not seemingly affect their stiffness, as measured in this experiment.We noticed that this lack of stiffening is similar to the ageing properties reported by ref. [1] which reported invariant elasticities during the maturation of five different condensates.We note though that our measurement represents the structural stiffness and not necessarily the elastic plateau modulus derived from an active microrheology routine (see below).Importantly, we did not observe a statistically significant change in this assay between condensates composed of pure wild type and mutant MEC-2.Thus, MEC-2 forms rapidly relaxing condensates that undergo a relatively slow ageing process.

Dual trap active microrheology of protein condensates in vitro
To further characterize the frequency dependent response of MEC-2 condensates and their derivates, we established a dual-trap active microrheology routine.To prevent condensate adhesion to the substrate which would modify force propagation between the two traps altering the force balance in an unpredicted way, we used the surface passivation reported in ref. [3].Only spherical condensates were found laying on the coverglass and could be detached and manipulated with the optical trap we used for the experiments.A total trapping power of P T = 40 mW at the sample plane was split into two time-shared optical traps with an average power of P trap = 20 mW each.Once two beads were optically trapped, each trapping stiffness, k 1 and k 2 , was obtained by rapidly-scanning the trap across the bead using the Particle Scan routine of the manufacturer's software, as described elsewhere [4].Importantly, the timesharing frequency to generate the two traps at different locations was f T S = 25 kHz, far beyond the relaxation time of our measurements, thus we assumed that the two traps are continuous during the force measurements.
We attached one bead on each opposite side of a protein droplet along the x direction (Fig. 5a-ii) and stretched the droplet.Trap 1, on the left hand side, applied a square or sinusoidal oscillation insults, depending on whether a stress relaxation test or a rheological measurement on the frequency domain was getting acquired, while trap 2 was kept motionless on the right hand side (Extended Data Fig. 10a, Supplementary Video 6).Similar to the single-bead stress relaxation tests (Extended Data Fig. 8b), for square oscillations, force on trap 1 peaked at F (1) p upon every pushing event and relaxed exponentially approaching a resting force, F (Extended Data Fig. 10a-ii).Upon pulling, we observed a similar peak followed by a relaxation curve, with opposite sign, due the bead being adhered to the droplet.An opposite, reaction force on bead 2 was measured with a lower peak force, F p , and was as well followed by a decay curve.We determined the steady-state resting force values after complete relaxation, F (1,2) 0 , through fitting the F-t curves to Eq. 1 and determined the resting droplet spring constant, χ 0 , and thereby obtained a measurement of the surface tension, γ, according to Eq. 3 [5].At the same time, beads 1 and 2 were displaced from the respective trap centers by F 1 /k 1 and F 2 /k 2 .Being d = 100 nm the oscillation amplitude of the square trajectory of trap 1, droplet deformation, ∆x 0 , and thereby χ 0 , read as: Using the expression reported in [5], we obtained the surface tension as follows: The complex spring constant of the protein condensate, χ * (ω), was measured by applying sinusoidal oscillations, with an amplitude of d = 100 nm, at increasing frequencies f = 0.5, 1, 2, ..., 64 Hz (Extended Data Fig. 10a-i).Traps 1 and 2 measured the force acting onto each optically-trapped bead upon the oscillation (Fig. 5a-iii and Extended Data Fig. 10a-i).We calculated the Fourier transforms of the force signals at the driving frequency, using the fast Fourier transform (FFT) algorithm in Matlab.We finally calculated χ * (ω) from the frequency-dependent complex spring constant of the coupled system (trap 1 -droplet -trap 2), χ * sys , as follows: In Eq. 4a, xsys (ω) is the Fourier transform at frequency ω of the length of the coupled system, taken from the optical trap coordinates as x sys (t) = x (2) trap (t).In Eq. 4b, ξ = 3πηD is the friction coefficient of the droplet in a viscous medium with η = 1 mPa•s.In turn, D is the droplet diameter and was measured as trap − ϕ, where ϕ = 1 µm was the diameter of the microbeads.
The complex spring constant captured the main features of a liquid phase separated condensate with Maxwell-like rheology (Extended Data Fig. 10b).At low driving frequencies, χ * (ω) approached the static spring constant, χ 0 , which is a mark for surface tension.At higher frequencies, χ * (ω) augmented considerably, especially in its elastic part, χ ′ (ω).The complex shear modulus of the droplets, G * (ω), was finally obtained as follows [5]: Experimental data of the G * modulus was fitted a non-fractional Maxwell element [2], from which parameters E (Pa) and eta (Pa•s), and the subsequent crossover frequency, ω c = 1/(2π)E/η (Hz), was obtained.The crossover frequency defines the viscous-and elastic-dominant regimes characteristic of Maxwell materials.For ω < ω c the material flows, whereas for ω > ω c the material responds as a solid.
A shift of ω c towards lower frequencies indicates that the material responds with a more elastic behavior for a given frequency and has undergone maturation.
With the help of this formalism, we discovered that all condensates behaved as Maxwell materials with a strongly frequency dependent response (Extended Data Fig. 10b,c).Whereas pure MEC-2 condensates and condensates composed of mutant MEC-2 (R385H) and UNC-89 did not show a significant change in their mechanical parameters such as crossover frequency and plateau modulus, the condensates composed of wildype MEC-2 and UNC-89 showed a strong maturation.We found that the crossover frequency shifted to 3 times lower values after 24 h maturation from ≈3Hz to 1Hz, whereas the plateau modulus E slightly increased, even though only marginally in all compositions tested.This subtle increase in E may suggest that MEC-2 is distinct from the glass-like behavior measured previously [1].In support for this observation is the fiber formation, which suggest the appearance of some degree of order during the spontaneous extrusion, which is incompatible with the assumption of a frustrated glass-like state.Indeed, the observation of the stiffening response could be due to increased crosslinks, especially in the binary mixture composed of MEC-2 and UNC-89.These fibrillar structures exhibited a strong increase in stiffness and stress storing capacity, with almost no disspation during the indentation test.The force-time signal upon square actuation perpendicular to it by means of an optically-trapped microbead.
For technical reasons, a rigorous microrheological analysis was impossible, but an ultrastructural analysis of the long range order in these fibers shall provide conclusive answers whether or not MEC-2 UNC-89 mixtures represents a glass or gel-like state.
How does a change in cross-over frequency affect cell signaling and mechanics?It How does the maturation time course compare to the maturation of other condensates?We found that pure MEC-2 showed significant maturation after 48h which is significantly slower than the maturation of PGL-3 droplet measured with active microrheology in ref. [1] and much slower than the aging of SPD-5 [12].We also found that the increase in relaxation time was accelerated in presence of UNC-89, which suggests that binding of the SH3 domain to the MEC-2 proline motif causes structural rearrangements between protein molecules within the condensate.Because our measurements on MEC-2 reveal higher viscosity and elastic modulus at high salt concentrations, the values of the surface tension in the order of 20-60 µN/m from our measurements is in agreement with values derived from FUS droplets [13].
We should note that the viscoelastic time scale measured in the rheology experiment characterizes the relaxation to an external, mechanical perturbation and hence does not necessarily coincide with the diffusion time scale in the FRAP experiment which is derived at equilibrium.As the importance of the rigidity of a material comes only to play under external load (e.g.touch), the optical tweezer and tension sensor experiments provide a more reliable readout for the rigidity transition.The diffusion coefficient thus provides information about the distinct mobility of the labeled MEC-2 in the naive and mature condensates, but cannot resolve differences of the viscoelastic relaxation of unstressed condensates once they are matured.Indeed, the fluctuation-dissipation theorem that relates thermal motion of a particle with the bulk mechanical properties does not necessarily hold in out-of-equilibrium materials such as glassy systems [14] and living cells [15,16].Thus, we shall emphasize here that an aging system is not at thermodynamic equilibrium, thus the relaxation time scale from the active microrheology and the FRAP recovery dynamics may not necessarily coincide [1].
2 Supplementary Note 2: Plasmid design and genome engineering Wild type complementary DNA encoding 61-128 amino acids of UNC-89, which includes the SH3 domain, was subcloned into a pET-24 expression vector (ordered from Twist Bioscience) with an Nterminal polyhistidine affinity tag for posterior UNC-89 purification, giving to pNS75.A previous version including 1-454 amino acids of UNC-89 led to unfolded purified protein.

Expression of wild type and R385H mec-2::mCherry in vivo
Wild type complementary DNA encoding full-length MEC-2 (A isoform) was subcloned into pBCN27 [17] to replace puromycin resistance gene, and fused to mCherry, generating pMK8.Site-directed mutagenesis was used to introduce the R385H single point mutation (u26 allele) in MEC-2, giving pMK9.A truncated version of MEC-2 (1-370 aa) was generated to delete the C-term from the pMK8 template giving pNS79.The three plasmids were integrated by the MosSCI method [18] in the EG6699 strain, that contains compatible MosSCI landing sites in Chr.II, leading to MSB87, MSB88 and MSB1151, respectively.MSB1151 was crossed to MSB99 to incorporate the mec-2(u37) allele (W119Stop) in the endogenous locus (Chr.X) leading to MSB1157.Overexpression of a region of the C-terminus (365-396 amino acids) of MEC-2 containing PRM was done by cloning MEC-2 wild type or R385H mutant from a gBlock under TRN-specific mec-17p, thus generating pNS10 and pNS12, respectively, and injecting them as extrachromosomal array.

Generation of the u37 allele in mec-2
The mec-2(u37) allele (W119Stop), which introduces a premature stop codon, was reproduced by CRISPR/Cas9 genome editing as described in [19,20].Two crRNAs were designed to cut few base pairs before the target site together with a donor consisting of a ssODN with 35 basepair (bp) homology arms flanking the polyspacer adjacent motif (PAM) sequence, the desired single point mutation and 5 other silent mutations to facilitate the posterior screening of the edit by PCR.Briefly, the Cas9-crRNA-tracrRNA RNP complex and the homology repair template (HDR) were assembled in Mili-Q water, together with the Cas9 complexes and HDR for the marker gene dpy-10, to introduce the semi-dominant cn64 allele.20-30 young adult hermaphrodites were injected with the CRISPR mix and recovered onto individual plates.After 3 days cultured at 25ºC, the progeny was screened based on the dpy or roller phenotype and singled onto individual plates.Mothers were lysed, screened by PCR for the corresponding edit and verified by sequencing.Sequences of crRNAs and ssODN donors are provided in Supplementary Data Table 6.

Expression of mec-2 in hypodermis
For mec-2 expression in hypodermis, the wrt-2 hypodermal promotor was amplified from genomic DNA (1376 bp) and mec-2::mCherry::coLOVpep was amplified from pNS13, both were cloned into pNMSB35 backbone giving pNS70.It was injected at 30 ng/ul leading to MSB991.pNS70 was used as a template to incorporate the R385H single point mutation in MEC-2 by site-directed mutagenesis, giving to pNS78.
It was injected at 30 ng/ul leading to MSB1144.

Generation of the FRET constructs
The TSMod cassette containing mTFP, 40-amino acid-long flexible linker, mVenus and a TEV protease site, was amplified from pMG319 [21] and inserted between 370-371 amino acids of MEC-2, leading to pNS2.This plasmid was used as a template to introduce the R385H mutation by site-directed mutagenesis to yield pNS24.The TSMod cassette was also inserted at the C-term of MEC-2 in the plasmid pMK13.These plasmids were integrated by the MosSCI method [18] in Chr II in strains with mec-2(u37) background in the endogenous copy (Chr X), generating MSB341, MSB357 and MSB74, respectively.
The low FRET construct was made by replacing the TSMod cassette in pNS2 with a mTFP-TRAF-mVenus cassette derived from pMG352 [21], which constitutively separates the donor and acceptor fluorophores, and it was injected as extrachromosomal array giving to MSB907.A TEV protease site was fused to mCherry through a spliced leader SL2 (gpd-2-gpd-3) under TRN-specific mec-17 promotor, giving to pMK97.It was injected into MSB341 leading to MSB403.Expression of the FRET construct in hypodermal cells was made by replacing the promotor of pNS2 by wrt-2p from pNS70, giving to pNS82, which was injected to N2, generating MSB1173.

Promotor trapping of unc-89
The unc-89 promotor expression vector, was generated by amplifying 4 kb upstream to unc-89 gene and the first three exons, including the SH3 domain, from C. elegans genomic DNA (Supplementary Data Table 7).It was transcriptionally fused to GFP through a spliced leader SL2 (gpd-2/gpd-3), generating pNS49, which was injected in MSB87 animals, leading to MSB656.

Tagging of unc-89 SH3 domain
The plasmid to tag UNC-89 SH3 domain (63-127 amino acids) was generated by amplifiying the SH3 motif from N2 genomic DNA and cloning it under the TRN-specific mec-18 promotor from pMK105.It was fused to GFP with a 5 amino acids linker from IR83 giving to pNS41.For spc-1 SH3 tagging, the mec-17 promotor and spc-1 SH3 motif were taken from the pMK32 backbone and fused to GFP with a 3 amino acids linker from pDD282 giving to pMK101.They were injected at 20 ng/ul giving to MSB493 and MSB544, respectively.

Insertion of TEV site in mec-2
Two TEV cleavage sites preceded by a flexilinker were inserted between amino acids 370-371 of MEC-2 by CRISPR/Cas9 genome editing.Two crRNAs were used to cut exon 9 of mec-2 A isoform and a 134 bp ssODN was used to repair it and incorporate flexilinker::x2TEV, see Supplementary Data Table 6.It was done on N2, generating MSB1127.Animals were verified by sequencing.The resulting animals were touch sensitive (Fig. 1h) and were injected with TEV protease (pMK97) leading to MSB1141.

Mutation of PRM in mec-2
The mutation of the prolines in the PRM motif of MEC-2 to alanines ( 377 PPxxP 381 → 377 AAxxA 381 ; MEC-2∆PRM) in the endogenous locus of C. elegans was done by CRISPR/Cas9 genome editing, as descrived before, giving to MSB1236.Sequences of crRNA and ssODN are provided in Supplementary Data Table 6.

Generation of (mEGFP(loxP)::AID knock-in)unc-89
The Nested CRISPR/Cas9 genome editing [22] was used to knock-in mEGFP at the unc-89 gene.Two crRNAs were used to cut the N-term of unc-89 and it was repaired by 200 bp ssODN containing parts 1 and 3 of mEGFP including a loxP within a synthetic intron of the mEGFP, along with a flexible linker and a degron site (AID).A new PAM site and a protospacer sequence was inserted in the first fragment to allow the in-frame insertion of the remaining sequence mEGFP2, designed as an IDT gBlock.For the second step, the same universal crRNA mentioned in [22] was used to make the double stranded break (see Supplementary Data Table 6).The correct in-frame insertion of the full length mEGFP was sequence verified and correct UNC-89 expression was checked by green fluorescence expression in muscles.The knock-in was done on top of MSB87, generating MSB523.

Generation of split wrmScarlet unc-89
When we tagged the endogenous unc-89 locus at the N-terminus with GFP (Extended Data Fig. 5c), we visualized that its expression was largely restricted to the muscles (Extended Data Fig. 5k).With the previous functional results in mind (Fig. 3a-c and Extended Data Fig. 5e,g), we hypothesized that UNC-89 is expressed in TRNs (Supplementary Data Table 1 and Extended Data Fig. 5b) in quantities that cannot be detected above background level if it is only tagged with a single fluorophore at the endogenous locus.Thus, we sought to visualize UNC-89 distribution using multiplexed split fluorescent protein complementation [23] by tagging the endogenous locus with 5x wrmScarlet (11) and co-expressing wrmScarlet(1-10) selectively in TRNs.To visualize an enhanced endogenous level expression of unc-89, it was used the multiplexed split fluorescence protein complementation [23].5 motifs of wrmScarlet barrel 11 separated by flexilinkers were inserted at the N-term of unc-89 by CRISPR/Cas9 genome editing, see Supplementary Data Table 6.The resulting strain was injected with mec-4p::wrmScarlet 1-10 (pNS50) as extrachromosomal array in order to restrict the expression selectively in TRNs and led to MSB791.
2.12 Generation of the conditional and constitutive unc-89 knock-out unc-89 knock-out was generated by CRISPR/Cas9 genome editing using two crRNAs to cut exon 3 of unc-89 and a 108 bp ssODN that led to a frame-shift and absence of the largest isoforms (a,b,e,f,k,l,m,n,o), see Supplementary Data Table 6.It was done on MSB523 mEGFP(loxP)::degron::unc-89, generating MSB590.Animals were verified by sequencing and by absence of green fluorescence in muscles.The resulting animals move normally but show a slight delay in development and body size (Extended Data Fig. 5d).These animals were crossed to MSB341 in order to assess FRET in a unc-89 mutant background.

CRE/loxP and AID degradation
Tissue specific unc-89 knock-out was generated by inserting an in-frame loxP site at the C-term of the unc-89 largest isoforms by CRISPR/Cas9 genome editing.Two crRNAs were used to cut the C-term of unc-89 and it was repaired by a 129 bp ssODN that carried a loxP and 5 silent mutations for posterior screening by PCR (see Supplementary Data Table 6).It was injected in MSB523 (mEGFP(loxP)::AID::unc-89) giving to MSB930.The correct in-frame insertion was sequence verified and correct unc-89 expression was checked by green fluorescence expression in muscles.MSB930 was crossed to MSB926, which carries the TRN-specific mec-17p::CRE and the mec-4p::TIR [24] leading to MSB953.Alternatively, MSB930 was crossed to MSB933 which carries the panneuronal (rgef-1)p::CRE leading to MSB941.
likely not accessible, and have therefore been filtered out.Note, MEC-2 has another highly conserved PRM, predicted to bind to canonical class 2 and non-canonical class 1 SH3 domains, highlighted in boldface.
offers a potentially unifying framework to understand how the ion channel kinetics are governed by the mechanical response of the viscoelastic environment.We suspect that for stimulation frequencies slower than the relaxation time constant, forces are effectively dissipated, as the material relaxes faster than the stimulation is propagated.If the material is part of the force transmission pathway, e.g.links the ion channel to the ECM or the cytoskeleton, these slow stimulations are dampened, which leads to an interruption of the mechanical transmission.Such mechanical frequency filters have been proposed to account for the frequency-selection in C. elegans touch[6,7,8].Our values of the cross over frequency (≈1Hz) in the mature binary droplets represents the lowest bound at which the elastic and viscous properties are equal.From this, we reason that effective force transmission shall only happen for faster stimulation, which is indeed seen in C. elegans touch.Our observation that the crossover frequency shifts to lower values in presence of UNC-89 suggest that the mature condensates behave as a solid, and thus conservative, over a larger frequency range.Thus, it offers a unifying principle by which the material environment of mechanical strain sensors become sensitive to external mechanical stimulations in a frequency dependent manner[6,9,10,11].It also may suggest a principle by which a regulated change in the viscoelastic properties enables a transient switch between rapidly adapting or slowly adapting mechanosensors.In this view, coacervation of different binding partners in the condensed phase may trigger viscoelastic changes with consequences on the force transmission and ion channel open/closing.

2. 1
Expression of wild type and R385H mec-2 and unc-89 for in vitro purification Wild type complementary DNA ranging from position 371 to 481 of MEC-2, which includes the Cterminal domain, was subcloned into a pCoofy expression vector (donated by Carlo Carolis lab) containing an N-terminal polyhistidine affinity tag and a NusA solubility tag for posterior MEC-2 purification, giving to pNS66.This plasmid was used as a template to incorporate the R385H single point mutation (u26 allele) in MEC-2 by site-directed mutagenesis, giving to pNS73.The plasmid pNS66 was used as a template to mutate the prolines of the PRM motif into alanines ( 377 PPSLP 381 → 377 AASLA 381 ; MEC-2∆PRM) by site-directed mutagenesis, giving to pNS80.This plasmid was used as a template to incorporate the R385H single point mutation (u26 allele) in MEC-2 by site-directed mutagenesis, giving to pNS81.