Filament transport supports contractile steady states of actin networks

23 In all eukaryotic cells, the actin cytoskeleton is maintained in a dynamic steady -state. Actin filaments are continuously 24 displaced from cell periphery, where they assemble, to wards the cell’s center , where they disassemble. Despite this 25 constant flow and turnover, cellular networks maintain their overall architecture constant. How such a flow of material 26 can support dynamic yet steady cellular architectures remains an open question. To investigate the role of myosin-based 27 forces in contractile steady-states of actin networks, we used a reconstituted in vitro system based on a minimal set of 28 purified proteins, namely actin, myosin and actin regulators . We found that, contrary to previous bulk experiments, 29 when confined in microwells , the actin network could self-organize into ordered arrangements of contractile bundles, 30 flowing continuously without collapsing. This was supported by three -dimensional fluxes of actin filaments, spatially 31 separated yet balancing each other. Unexpectedly, maintaining these fluxes did not depend on filament nucleation or 32 elongation, but solely on filament transport . Ablation of the contractile bundles abolished the fl ux balance and led to 33 network collapse. These findings demonstrate that the dynamic steady state of actin networks can be sustained by 34 filament displacement and recirculation, independently of filament assembly and disassembly. 35 36

37 Actin architectures undergo continuous self-renewal, with their individual components constantly 38 dissociating, recycling and reassembling while maintaining an overall stable structure 1–4. This 39 process is maintained by the balance of steady-state active fluxes of incoming and outcoming 40

allowing different subcellular networks of the cell (lamellipodia, filopodia, stress fibers, cell 41 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2025. ; https://doi.org/10.1101/2025.02.21.639071doi: bioRxiv preprint 2 cortex) to be dynamically regulated, providing them with the flexibility to grow, shrink, appear and 42 disappear as needed 5. Hence, most actin architectures are in a “dynamic steady state” (DSS) , 43 characterized by the continuous nucleation and growth of filaments, balanced by their disassembly 44 and recycling 6. Additionally, active contraction and filament transport by myosin motors further 45 contribute to maintaining a stable architecture 2,7–10. All the steady state fluxes, which depend on 46 the rate of assembly and d isassembly, of contraction and of exchange of filaments between 47 subcellular networks, must balance perfectly to establish and maintain the network in a DSS 11. 48 However, due to the multiplicity and interdependencies of these mechanisms, investigating their 49 basic properties in cells is challenging. A recent leap forward has been the use of Xenopus egg 50 extracts encapsulated in droplets to study what role of the balance between assembly and 51 contraction rates plays in establishing a DSS 12–18. However, cytoplasmic extracts, while simplified 52 compared to a whole cell, retain all the biochemical complexity of the cytoplasm. In vitro 53 reconstituted networks, based on the use of a minimal set of purified proteins to build specific 54 network architectures, offer a powerful alternative by allowing precise control over individual 55 components to reveal the basic principles underlying DSS formation and maintenance 11. Recent 56 efforts have indeed proven successful in reconstituting recycling -based coupling in assembly and 57 disassembly mechanisms 19,20. However, reconstituted actomyosin networks typically undergo only 58 transient contraction in response to myosin -induced stress 21–26. More recent studies involving 59 actomyosin on supported lipid membranes 27–29 and/or encapsulated in giant vesicles or droplets 30–60 32, often in the presence of actin nucleators, have shown that DSS -like networks exhibiting energy 61 dissipation without contraction collapse can be reconstituted. However, most reconstituted DSS 62 architectures exhibit no directed flux of actin mass and lack general order, resembling active but 63 disordered networks in which energy is consumed to enhance fluctuations. These DSS likely result 64 from balanced flux throughout space, whereas to achieve steady state currents, like those observed 65 in cells, incoming and outcoming mass fluxes must be spatially separated. Here, we focused on the 66 reconstitution of transport- and contraction-based actin networks assembling into a dynamic steady 67 state. Using lipid-coated micro-engineered devices to confine and guide network self-organization 68 and taking advantage of myosin-based force generation, we identify conditions leading to ordered 69 and contractile dynamic steady-states. 70 The myosin-to-actin ratio modulates the length scale of network coordination 71 Geometrical boundaries are necessary to guide the self -organization of filaments and molecular 72 motors into ordered networks 33. To implement them, we first resorted to micropatterned supported 73 lipid bilayers (SLBs), as their fluidity is key for optimal polymerization of actin filaments 22 (Fig. S1). 74 We thus coated 100 μm-wide circular micropatterns with a lipid membrane containing 0.5 % 75 biotinylated lipids to functionalize the bilayer with a Nucleation Promoting Factor (NPF), (SNAP-76 Streptavidin-WA-His or hereafter referred to as WA, see Methods). The bilayer acts at the same time 77 as a surface passivation and as a tool to localize polymerization only on its surface. The assembly of 78 actin filaments was induced by the addition of actin monomers (0.5 to 1 µM), the Arp2/3 complex 79 (50 nM), and profilin (in a 1:1 molar ratio with actin), (Fig. 1A). This led to the formation of a branched 80 network confined to the micropattern within half an hour, with its assembly rate controlled by the 81 WA and actin concentrations (Fig. 1 A-B-C, Movie S1). We then added myosin motors, fixing the WA 82 concentration at 10 nM, which we found allows modulation of network density solely through actin 83 concentration. Specifically, we used Myosin VI, a processive minus-end directed motor (Fig. 1B-C). 84 Myosin VI (hereafter referred to as myosin) is known to efficiently contract branched network and 85 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2025. ; https://doi.org/10.1101/2025.02.21.639071doi: bioRxiv preprint 3 induce sliding of antiparallel actin filaments without the need to assemble into minifilaments , and 86 has already been used extensively22,26,34. As expected, the addition of 3.3 nM of myosin to 1 μM of 87 actin induced the contraction of the entire network towards the center22. However, the contraction 88 occurred only as a single, transient event . The overall network collapsed in 15-30 minutes, with a 89 peak mean speed of 1.5 μm min-1 (Fig. 1D), after which no notable rearrangement of the network 90 was observed. Interestingly however, halving the concentration of actin monomers to 0.5 μM led to 91 a slower actin polymerization and to a less coordinated contraction. Specifically, the first appearing 92 Arp2/3 complex-based branched clusters were soon connected by elongated bundles that appeared 93 to push and pull on each other, forming multiple, small, independent and unsynchronized contractile 94 regions ( Fig. 1E). As previously shown, the tracking of actin flow revealed local and prolonged 95 disordered motion28, which lasted over more than two hours, i.e. much longer than the time for 96 global and transient network contraction at higher actin densities ( Fig. 1F-G, Movie S2). We 97 interpreted this difference (global collapse vs continuous disordered motion) as a consequence of 98 network entanglement. Dense networks might better propagate the contractile stress and integrate 99 it over the whole sample. However, such global inward contraction is not balanced by peripheral 100 growth as in cells and thus could not reach any steady state and only collapsed. Conversely, the less 101 dense network appeared to be locally balanced by pushing and pulling forces, but did not achieve 102 any global coherence. We then wondered whether enhancing the coherence of less dense networks 103 by limiting the extensile component of the flow through 3D confinement could lead to more ordered 104 and better-synchronized contraction. 105 106 107 108 109 110 Figure 1 Actin and myosin on micropatterns form contractile networks A) Schematic of a functionalized lipid micropattern. A circular 111 pattern coated with a SLB (red) is engraved on a glass slide and functionalized with WA (green circles). WA and the Arp2/3 co mplex 112 trigger the assembly of an actin branched network on the SLB. Myosin is also present to provide contractile stress. Top: tilt ed view. 113 Bottom: side view. B) Actin polymerization curves on micropatterns. By varying the amount of actin and WA on the surface the rate 114 of actin assembly can be tuned, with higher WA (actin) resulting in faster polymerization and a shorter lag phase. C) Microgr aphs of 115 actin polymerization at 1 μM (top) and 0.5 μM (bottom) actin and 10 nM WA, with 50 nM Arp2/3 complex and 1:1 actin:profilin ratio. 116 Pictures show the different rate and coverage as actin density is varied. Scale bar is 50 μm. D) At 1 μM actin, the addition of 3.3 nM 117 of myosin leads to global contraction towards the center, as exemplified by the kymographs on the right. The dashed line indicates 118 where the kymographs have been performed. Scale bar is 50 μm. Arrows indicate local actin flow. Red dashed line indicates the 119 position of the pattern. E) At 0.5 μM actin, the addition of 3.3 nM of myosin leads instead to a disordered state with no collapse, as 120 exemplified by the kymographs on the right. The white dashed line indicates where the kymographs have been performed. Arrows 121 indicate local actin flow. Red dashed line indicates the position of the pattern. Scale bar is 50 μm. A zoomed-in part of the sample 122 with arrows indicating the local flow is shown on the bottom. F) Mean speed of actin on patterns as obtained by optical flow. The 123 globally contracting network exhibits a surge in speed leading to global contraction in ~ 30 minutes. The low actin density n etwork 124 instead continues to move in a disordered manner for several hours. G) Direction of the mean actin flow inside the micropatte rn, 125 showing radial contraction (left) for the high actin density and disordered flow for the low one (right). Arrows length indic ates the 126 relative flow magnitude. 127 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2025. ; https://doi.org/10.1101/2025.02.21.639071doi: bioRxiv preprint 4 128 129 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2025. ; https://doi.org/10.1101/2025.02.21.639071doi: bioRxiv preprint 5 130 Actomyosin network self-organizes into contractile DSS in microwells. 131 To confine the system in 3D, we microfabricated cylindrical, 70-µm-wide and 50-µm-high microwells 132 out of NOA photoresist (Fig. 2A, see Methods) . Microwells were coated with a WA -functionalized 133 SLB, filled with the actin mix and closed with an oil layer on top 19,35. We first confirmed that the SLB 134 was also covering the upper oil layer and that lipids were diffusing on all sides (see Fig. S1). Actin 135 filaments inside microwells could polymerize into branched networks (Fig . 2B-C, Movie S3). Network 136 growth was, however, limited by the pool of available components enclosed in the 200 picolitres of 137 volume, resulting in lower surface coverage, i.e. less total actin on the surface of the microwells (Fig. 138 2D). To compensate for the limited available pool of components, we increased the concentrations 139 of the Arp2/3 complex to 100 nM, of actin monomers to 4 μM, and of myosin to 20 nM. In these 140 conditions, we observed again a global but unique contraction of the network towards the center of 141 the microwells (Fig. 2E, top). As done for micropatterns, we then lowered the actin concentration 142 (to 1 µM) to reduce network entanglement. Strikingly, instead of the disordered state observed on 143 2D micropatterns, this led to the assembly of circular bundles, which appeared to continuously form 144 along the well periphery and contract towards the center ( Fig. 2E, bottom, Movie S4). Circular 145 bundles were aligned tangentially to the microwell edges but became more disordered as they 146 contracted toward the center, as quantified by the local nematic order parameter ⟨S⟩, which is 0 for 147 a disordered network and 1 for a perfectly aligned one (Fig. 2F-G, see Methods). Similarly, the mean 148 speed of the continuous contraction was around 0.5 μm min-1 and peaked at the periphery of the 149 well, with the speed profile decreasing towards to the center (Fig. 2F-G). An analysis of the actin 150 flow's divergence revealed that the flow inside the wells was contractile, particularly at the periphery 151 (Fig. S2). However, by contrast with respect to the single and global contraction of dense network 152 that collapsed in 30 minutes, this architecture remained persistent for several hours. Specifically, at 153 the periphery, the radially directed actin flow maintained a roughly constant speed for at least 3 154 hours, gradually decaying without stopping for up to 8 hours ( Fig. 2H, Fig. S3). Over this time, we 155 observed the formation and successive flow of approximately 40 independent bundles at the 156 periphery, with a new bundle appearing every 10 minutes (see kymograph in Fig. 2E). However, this 157 contractile flow did not converge to a single point , as in globally contractile states, but gradually 158 dissolved into an incoherent , fluctuating organization at the center of the well . Myosin was 159 colocalized with the actin bundles at all times, differently than in the globally contracting case in 160 which after network collapse myosin accumulates in a central spot ( Fig. S4). Additionally, the fact 161 that the radial profile of actin intensity and the total amount of actin on the bottom surface remained 162 unchanged over time (Fig. S5) provide evidence of actin mass redistribution, with contracted actin 163 being somehow sent back to the periphery to maintain a constant density on the microwells’ surface. 164 The formation and maintenance of a constant architecture, despite the presence of flow, confirmed 165 that we had identified conditions leading to a contractile DSS. Finally, this DSS appeared resistant to 166 variations in actin density, as it was also observed at 2 µM actin, showing similar characteristics to 167 the 1 µM condition (Fig. 2H, S6A, Movie S5). 168 To further test whether the emergence and maintenance of contractile DSS were favored by high 169 myosin-to-actin ratio, we increased the myosin concentration to 50 nM, compared to the conditions 170 leading to global collapse (4 µM of actin and 20 nM of myosin). This resulted in a similar contractile 171 DSS as previously described, with the constant production of new actin bundles along the periphery, 172 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2025. ; https://doi.org/10.1101/2025.02.21.639071doi: bioRxiv preprint 6 steadily contracting toward the center (Fig. 2.I-J,Movie S5). Thus, higher myosin concentrations favor 173 the transition from global collapse to DSS-like flow. In contrast, decreasing the myosin concentration 174 to 5 nM at 1 μM actin did not produce coherent flow, but instead led to a condensed, fluctuating 175 network (Fig. S6). 176 177 Contractile DSS does not depend on filament assembly and disassembly. 178 Previous descriptions of actin network DSSs stand on the balance between filament assembly and 179 disassembly 6. While no disassembly factor (e.g., ADF/cofilin 6,19) is present in our system, myosin 180 activity could break filaments and cause their disassembly 26,36, making them available again for 181 repolymerization by the Arp2/3 complex. To rule out this hypothesis, we manipulated the nucleation 182 activity through WA variations. First, we favored branched actin filament assembly by increasing the 183 concentration of WA from 20 to 100 nM ( Fig. 3A). We then inhibited branched filament nucleation 184 by removing the Arp2/3 complex. Finally, to completely inhibit disassembly, we stabiliz ed the 185 network with 1 μM of phalloidin (Fig. 3A). Surprisingly, none of these conditions perturbed the self-186 organization of the acto -myosin network, which assembled in all cases into a long -lived DSS with 187 comparable architecture and speed (Fig. 3B-C). This led us to conclude that this peculiar DSS is not 188 based on continuous turnover (i.e., assembly/disassembly) of actin filaments (Movie S6). 189 However, the continuous appearance and contraction of new filament bundles at the periphery 190 indicate that actin filaments must be transferred from the center to the edge to sustain the radial 191 flow. Since we observed individual actin filaments “gliding” on the surface of the SLB in the early 192 stages of network assembly (Fig. 3D, Movie S7), likely due to the binding of myosin on the membrane 193 that could propel the actin 37–40, we reasoned that such transport could randomly redistribute 194 filaments from the center of the network, where they accumulated due to the radial flow, back to 195 the edge of the microwells. We hence tracked individual filaments in these dense networks by 196 adding a small quantity of pre -polymerized, short actin filaments with a distinct fluorescent label 197 (Fig. 3E) to mark fragments of the network . Notice indeed that these filaments are still able to 198 elongate, so they are integrated into the network, allowing to track its motion more easily. We found 199 that most of the filaments were moving tangentially at the periphery, i.e., along the bundles, with a 200 speed of approximately 3 μm min-1, lower than reported walking speeds for myosin VI, as expected 201 if motors are membrane-bound 37,38,40,41 (Fig. 3F-G-H, Movie S8). 202 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2025. ; https://doi.org/10.1101/2025.02.21.639071doi: bioRxiv preprint 7 203 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2025. ; https://doi.org/10.1101/2025.02.21.639071doi: bioRxiv preprint 8 Figure 2: Actin and myosin in microwells assemble into a dynamic steady state A) Schematics of a NOA microwell. 204 Proteins are encapsulated inside NOA -based microwell s (NOA indicated in blue ) coated with a SLB (red). As in 205 micropatterns, the SLB is functionalized with WA (green circles). The well is then closed with oil (orange) to seal it. Top: 206 tilted view. Bottom: side view. B) Actin polymerization curves inside microwells at different actin and WA concentrations, 207 confirming that branched actin assembly is maintained in wells and that it can be tuned biochemically. C) Micrographs 208 of polymerization inside microwells at 1 μM actin with 10 and 100 nM WA, 1:1 actin:profilin ratio and 100 nM Arp2/3 209 complex. Surface coverage is reduced with respect to micropatterns. Scale bar is 35 μm. D) Fraction of the area occupied 210 inside microwells or micropatterns by the actin network, showing reduced coverage inside microwells due to the limited 211 monomer pool. Data is taken from the same conditions as Fig 2B and Fig. 1B, i.e. at 1 μM actin, 10 nM WA, 100 nM 212 Arp2/3 complex, 2 μM profilin for microwells and same but with 50 nM Arp2/3 complex for micropatterns . E, top) At 4 213 μM actin, 10 nM WA, 20 nM myosin the network undergoes global contraction. The kymograph on the right shows the 214 behavior over the full recording, contractile radial flow is marked by orange arrows. E, bottom) Reducing the amount of 215 actin to 1 μM leads instead to the formation of an organized architecture at the periphery and a continuous sustained 216 actin flow towards the center. The kymograph on the right shows the behavior over the full recording, peripheral radial 217 flow is marked by orange arrows. Scale bars are 35 μm. F) Mean actin orientation (blue) and mean actin flow (red) inside 218 microwells in the same condition as E, bottom, extracted from averaging several wells (n= 7). G) Radial profiles of the 219 speed (blue continuous line) and of the nematic order parameter ⟨S⟩ (black dashed line). The mean speed is peaked at 220 the well periphery and decreases towards the center. The order parameter does the same, indicating an organized 221 architecture at the periphery that decorrelates towards the center and a correlation between order and flow. H) Mean 222 speed at 1 and 2 μM over time, indicating constant persistent flow for several hours. The shaded area indicates roughly 223 the polymerization time. I) Actin orientation and mean flow at steady state for 4 μM with 20 nM and 50 nM myosin 224 respectively, showing that the contractile flow observed (left) can be turned into DSS-like behavior by increasing myosin. 225 J) Kymograph of a sample at 4 μM and 50 nM myosin with visible radial peripheral flow, indicating that the myosin actin 226 ratio regulates the transition from global contraction to DSS. Scale bar is 35 μm. 227 In the case of DSS -like contraction at the periphery, this could also be attributed to myosin activity 228 sliding antiparallel bundles tangentially against each other. Assuming the slow contraction towards 229 the center is due to this mechanism, with this tangential sliding leading to radial contraction, we 230 obtain an independent estimate for the slower contractile radial flow of 3 μm min-1/π ~ 1 μm min-1, 231 that fits the observed contractile flow. Actin filaments displayed a more random displacement closer 232 to the center. However, averaging over all the observed trajectories, we noticed that the number of 233 actin filaments moving from the center outwards (obtained by the mean radial component of the 234 velocity v, i.e. ⟨r.v⟩, where r is the tangential versor, see Methods) did not balance the inward flux at 235 the periphery, which was found to be directed towards the center. This ruled out the hypothesis that 236 balance was due to transport from center to periphery ( Fig. 3H-I). Conversely, in the tangential 237 direction no net flux (⟨t.v⟩, where t is the tangential versor ) whatsoever is observed. Again, these 238 observations were confirmed across different conditions (Fig. S7). Additional experiments obtained 239 by labelling only a tiny (0.2 %) amount of the actin to obtain individual bright speckles on otherwise 240 unlabelled filaments also revealed motion of actin fibers in both directions, in the center of the wells 241 as well as at the periphery (Fig. S8, Movie S9). 242 Contractile DSS in 2D is associated with 3D recirculation of filaments 243 If the outward flow of moving filaments on the bottom of the microwell could not supply the 244 peripheral network and its contraction, the missing actin mass likely had to come from the volume 245 of the microwell. We hence used confocal microscopy to observe actin filaments in the entire volume 246 of the microwell. First, we noticed that the contractile rings on the lower side of the microwell 247 appeared to be connected to actin cables spanning the volume and forming a “tent -like” structure 248 attached at the tip to the upper side of the enclosed volume (Fig. 4A, Movie S10). We then wondered 249 whether these cables connecting the lower and upper SLBs could support the transport and 250 recirculation of actin filament in 3D. 251 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2025. ; https://doi.org/10.1101/2025.02.21.639071doi: bioRxiv preprint 9 252 Figure 3: The dynamic steady state is not maintained by turnover or 2D transport A) Variation of actin polymerization 253 and stability. Three networks at 1 μM actin and 20 nM myosin are polymerized with 100 nM WA to increase the actin 254 assembly (top), with 1:1 actin:phalloidin ratio to stabilize filaments (middle, labelled as “+ Phalloidin”) and in the absence 255 of the Arp2/3 complex and profilin to only have spontaneous polymerization (bottom, labelled as “no Arp2/3”) . In all 256 cases, the networks assembled into a DSS with comparable architecture and speed over time. Scale bar is 35 μm. B) 257 Mean actin orientation (blue) and mean actin flow (red) as extracted from averaging several wells (n= 7) for the three 258 above samples. C) Mean actin speed over time for the three above samples, compared to the standard conditions with 259 10 nM WA as a reference. D) At the beginning of the experiment, fast frame -rate movies (10 s interval) reveal the 260 presence of individual filaments gliding (one filament is tracked by the yellow arrow an example) and of Arp2/3 clusters 261 expanding due to filaments gliding. Scale bar is 10 μm. E) A network at 1 μM actin, 20 nM myosin, 10 nM WA (left) is 262 doped with 20 nM of pre-polymerized filaments of a different color (middle) so that both the network and its microscopic 263 motion can be observed at the same time (right, composite). Scale bar is 35 μm. F) Tracking of individual filaments 264 reveals both tangential and radial trajectories. G) Histogram of filaments’ speed over n= 7 wells, indicating that the 265 microscopic speed of the network is higher than the resulting mean flow. H) Mean (black) speed of individual filaments 266 as a function of the radius, showing mainly constant speed of individual filaments (not to be confounded with the net 267 actin flow). By analyzing the ratio between radial (red) and tangential (green) speed, we confirm that motion is mostly 268 tangential at the periphery and disordered in the center. I) Net flux of filaments in the radial (circles) and tangential 269 (crosses) direction, indicating an influx of filaments at the periphery but only in the radial direction. 270 271 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2025. ; https://doi.org/10.1101/2025.02.21.639071doi: bioRxiv preprint 10 We took advantage of fast volumetric scanning with confocal spinning disk imaging to track short 272 fluorescent filaments within an unlabeled network in 3D . Strikingly, we observe d numerous 273 individual filaments moving up or down, confirming the circulation of filaments in 3D along the 274 “tent” cables (Fig. 4B-C-D-E-F, Movie S11-S12). The number of filaments moving up and down was 275 similar (Figure 4C, 4E-F) so no net flow was established between the upper and lower layers, which 276 was consistent with the relatively stable actin density observed in the bottom layer. These results 277 were confirmed by both hand-tracked filaments inside the whole well , showing both upwards and 278 downwards trajectories and their transition from moving on the bottom to moving inside the 279 bundles, (Fig. 4E) and by automatically tracked actin filaments in a 10 μm slice near the bottom with 280 higher temporal resolution (Fig. 4F) . Interestingly, upward -moving filaments were more 281 concentrated at the center of the microwells, while downward-moving filaments were more 282 dispersed across the entire microwell ( Figure 4D). Filaments moved at similar speeds in both 283 directions (~3 μm min-1, Figure 4C). At such speeds, they could be transported from the top to the 284 bottom of the well and back to top within an hour, so multiple cycles were expected to occur over 285 the several hours of steady-state behavior. These measurements suggested that the contractile flux 286 of filaments in the bottom layer was balanced by actin transport through the 3D architecture ( Fig. 287 4G, right). 288 To further challenge this hypothesis, we used laser ablation to severe the vertical actin bundles in 289 the middle of the microwell to interrupt the transport along them. Following ablation, the contractile 290 bundles rapidly recoiled and coalesced with both upper and lower layers, which were then isolated 291 (Fig. 4H). This recoil also implies the presence of tension in the network (Fig. 4G., left). In the minutes 292 following the ablation of vertical bundles, the bottom network collapsed and the DSS was 293 interrupted (Fig. 4H-I, Movie S13). The peripheral fibers lost their tangential alignment, and the 294 network globally contracted toward the center of the microwell, thereby stopping the overall flow, 295 as shown by the mean speed analysis (Fig. 4J-K). Conversely, on the same timescale, both control 296 samples without ablation and samples in which the laser pulses were focused next to the cables 297 without severing them (n=2) kept on showing a continuous contraction (Fig. 4J-K, Movie S14). Thus, 298 interrupting transport and releasing tension through ablation results in the interruption of DSS 299 behavior, leading to global contraction. 300 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2025. ; https://doi.org/10.1101/2025.02.21.639071doi: bioRxiv preprint 11 301 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2025. ; https://doi.org/10.1101/2025.02.21.639071doi: bioRxiv preprint 12 Figure 4: Confocal imaging and laser ablation reveals maintenance of the dynamic steady state through 3D transport 302 A) Confocal side and tilted view of the actin network clearly showing a tent -like 3D organization (1 μM actin, 20 nM 303 myosin, in the absence of profilin and Arp2/3 complex– all panels in this figure are in the same conditions ). Scale bar is 304 35 μm. B) 3D visualization of individual filaments showing trajectories in both directions (top to bottom and vice -versa, 305 indicated by colored arrows). Scale bar is 35 μm. C) Histogram of speed for tracked actin filaments going upwards (blue) 306 and downwards (orange), indicating the absence of net motion. D) Histogram of position of downwards (orange) and 307 upwards (blue) traveling filaments as a distance from the wells’ center. Upwards trajectories are concentrated at the 308 center, whereas downwards one s are more spread out and present at the periphery. Trajectories are tracked in n=6 309 wells, in a 10 μm slice close to the surface and for a total time of 20 minutes. E) Hand-tracked trajectories showing both 310 upwards and downwards motion and the transition from motion on the bottom layer to moving along the 3D cables. 311 Blue points are t=0 and red points are t=60 min. F ) Automatically tracked trajectories, separated by direction: upwards 312 (top, red) and downwards (bottom, blue). G) Schematics of the actin architecture (in green) with tension (magenta) and 313 transport (red) indicated. The network is held together by the 3D structure that also supports transport of actin 314 filaments. Upon ablation the tension is lost, the flow turns to contractile, and transport is interrupted. H) Side (top) and 315 bottom (bottom) view of a microwell right before and after ablation of the cables in 3D (approximate ablation site is 316 indicated by yellow stars). The peripheral architecture is quickly disrupted and the networks contracts. In 3D, cables snap 317 and recoil indicating the presence of tension. Scale bar is 35 μm. I) Kymographs of a control (bottom, with the same dose 318 of UV light but without cutting) and ablated (top) sample. Only the ablated sample contracts, indicating the effect being 319 due to the ablation itself. Yellow dashed line indicates the ablation time, orange arrow the flow direction. Scale bar is 35 320 μm. J) Flow over time in the ablated and control experiments, showing the decay of active flow in the ablated sample. 321 K) Comparison between the flow (red) and the actin architecture (blue) in the control (right) and ablated (left) case. 322 323

324 Altogether, these results demonstrate the self-assembly of a contractile architecture in vitro that can 325 maintain a stable flow of actin filaments for several hours when spatially confined in 3D within closed 326 microwells. In the horizontal planes, the curvature of the spatial boundaries led to the formation of 327 circular bundles, which contracted and induced a radial flow of actin filaments toward the center. As 328 the actin filaments moved inward, they were also pulled out of plane by contractile vertical bundles 329 connecting the two upper and lower horizontal planes. Actin filaments were transported along these 330 vertical bundles, mostly moving toward the center of the well along central bundles or toward the 331 periphery along lateral bundles . Thus, these fluxes supported a recirculation of mass collecting 332 filaments from the center and redistributed them to the periphery of the horizontal layers (Fig. 4G). 333 Hence this represents an exemplary reconstitution of a contractile DSS. The resulting architecture 334 spatially separates different fluxes of actin filaments, producing a net contractile flow on the bottom 335 surface of the well towards the center, compensated by three -dimensional transport of filaments 336 back to the periphery. 337 Interestingly, this DSS did not involve the assembly and disassembly of filaments contrary to the 338 current understanding/appreciation of DSS in living cells 5,6,42 and egg extracts 12–14,16. Instead, the 339 DSS we obtained in vitro was based on filament transport and recirculation along contractile 340 bundles. Our experiments are unable to distinguish between two possible kinds of transport, i.e. 341 gliding due to membrane -bound myosins and myosin -based sliding of neighboring antiparallel 342 filaments. Both mechanisms likely act in unison, as we have evidence for both, resulting in motion 343 of filaments within bundles and a contractile tendency on the bottom. The orientation of filament 344 recirculation was based on the differential transport along vertical bundles, which appeared directed 345 toward or away from the center of the network depending on the position of these bundles ( Fig. 346 4G). These differences may stem from the differences in fiber anchorage in the central or peripheral 347 part of the upper or lower layer since it was shown that the direction of contractile flow is directed 348 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2025. ; https://doi.org/10.1101/2025.02.21.639071doi: bioRxiv preprint 13 toward region of higher friction 22. Importantly, filament transport and exchange has also been 349 shown to be central to the dynamics of cellular networks 10,11. In mesenchymal cells, the network 350 forming the lamella incorporates pre -existing filament from the lamellipodium, displaces them by 351 bundle contraction and translocation, and transmit them to stress fibers 43–45, as we saw in the 352 horizontal networks. Filaments are also transported along static stress fibers by myosin -based 353 contraction, as we saw in the vertical bundles. The mechanism driving the contractile DSS we 354 reconstituted in vitro suggests that DSS of cellular networks could be at least partially supported by 355 filament recirculation in parallel with cycles of filament assembly/disassembly. A clear outlook is 356 then the possibility of combining both mechanisms, transport and recycling. 357 Furthermore, ordered and contractile DSS of actin networks are central to multiple cellular functions 358 such as shape regulation, environment sensing and motility. All in all, this in vitro reconstitution also 359 constitutes a step forward in the implementation of these key functions in minimal artificial cells. 360 361 Data and code availability statement 362 All raw data and code used for this study will be made available in a public repository. 363 Competing interests 364 The authors declare no competing interest. 365 Author contributions 366 AS performed research and analyzed the data. MO performed research. CG, LBo and AC purified 367 proteins and microfabricated structures. JS and YT purified myosin -VI. AS, CG, AC, MT and LBl 368 designed research. AS, MT and LBl wrote the manuscript. All authors reviewed the manuscript. All 369 authors declare no conflict of interest. 370

371 AS acknowledges the support of the European Molecular Biology Organization (EMBO, ALTF 628-372 2022) and of the MSCA Postdoctoral Fellowship program (HORIZON-MSCA-2022-PF-01, proposal 373 ASTER 101108326). This work was supported by the European Research Council (Consolidator 374 Grant 771599 (ICEBERG) to MT and Advanced Grant 741773 (AAA) to LBl). This work benefited 375 from the technical contribution of the joint service unit CNRS UAR 3750. The authors would like to 376 thank the engineers of this unit for their advice during the development of the microwells 377 fabrication technique. JS and YT acknowledge the NIH grant HL004232. 378 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2025. ; https://doi.org/10.1101/2025.02.21.639071doi: bioRxiv preprint 14 379

380 Protein purification 381 Actin was purified from rabbit skeletal muscle acetone powder 46. Monomeric Ca-ATP-actin was purified by gel-filtration chromatography 382 on Sephacryl S-300 at 4°C in G-buffer (2 mM Tris–HCl, pH 8.0, 0.2 mM ATP , 0.1 mM CaCl2, 1 mM NaN3 and 0.5 mM dithiothreitol [DTT]). Actin was 383 labeled on lysines with Alexa-568 or Alexa-647 47. All experiments were carried out with 5% labeled actin, except single pre-polymerized filaments 384 labelled at 20% and networks with speckle labelled at 0.2 %. 385 The Arp2/3 complex was purified from bovine thymus. Calf thymus was cut into approximately 1 cm pieces and mixed with extraction buffer (20 mM 386 Tris pH 7.5, 25 mM KCl, 1 mM MgCl2, 5% glycerol, protease inhibitors) for 1-2 minutes, then shaken in a beaker for 30 minutes. The extract was then 387 centrifuged in a benchtop centrifuge at 1700×g for 5 minutes, and the supernatant clarified at 39,000×g for 25 minutes at 4°C. The supernatant was 388 then filtered over glass wool, the pH was fixed at 7.5 with KOH and centrifugation was carried out at 140,000×g at 4°C for 1 hour. The medium 389 aqueous phase was transferred to a chilled glass beaker, the extract was precipitated with 50% ammonium sulfate and centrifuged at 39,000×g for 390 30 minutes at 4°C. The pellet was resuspended in 10 mL extraction buffer with 0.2 mM ATP , 1 mM DTT and protease inhibitor. It was then dialyzed 391 overnight in Arp2/3 dialysis buffer (20 mM Tris pH 7.5, 25 mM KCl, 1 mM MgCl2, 5% glycerol, 1 mM DTT and 0.2 mM ATP). A GST-WA glutathione 392 sepharose column was prepared and washed with extraction buffer containing 0.2 mM ATP , 1 mM DTT and protease inhibitors. The dialyzed extract 393 was passed over the GST-WA column. Next, the column was washed with 20 mL extraction buffer with 0.2 mM ATP , 1 mM DTT and then with 20 mL 394 extraction buffer with 0.2 mM ATP , 1 mM DTT and 100 mM KCl. The Arp2/3 complex was eluted with 20 mL extraction buffer with 0.2 mM ATP , 1 mM 395 DTT and 200 mM MgCl2, then dialyzed in source buffer A (piperazine-N,N′-bis(2-ethanesulfonic acid) (PIPES) pH 6. 8, 25 mM KCl, 0.2 mM ethylene 396 glycol-bis(β-aminoethyl ether)-N,N,N′,N′-tetraacetic acid (EGTA), 0.2 mM MgCl2 and 1 mM DTT) overnight. The Arp2/3 complex was then loaded 397 onto a MonoS column and eluted with source buffer B (piperazine-N,N′-bis(2-ethanesulfonic acid) (PIPES) pH 6.8, 1 M KCl, 0.2 mM ethylene glycol-398 bis(β-aminoethyl ether)-N,N,N′,N′-tetraacetic acid (EGTA), 0.2 mM MgCl2 and 1 mM DTT). The Arp2/3 complex was dialyzed in storage buffer (10 399 mM Imidazole pH 7.0, 50 mM KCl, 1 mM MgCl2, 0.2 mM ATP , 1 mM DTT and 5% glycerol), flash-frozen in liquid nitrogen and stored at -80°C. 400 Human profilin was expressed in BL21 DE3 pLys Escherichia coli cells and purified according to 48. Snap-Streptavidin-WA-His (pETplasmid) was expressed 401 in Rosetta 2 (DE3) pLysS (Merck, 71403). Culture was grown in TB medium supplemented with 30  μg/ml kanamycin and 34 μg/ml chloramphenicol, 402 then 0.5 mM isopropyl β-D-1- thiogalactopyranoside (IPTG) was added, and protein was expressed overnight at 16°C. Pelleted cells were resuspended 403 in Lysis buffer (20 mM Tris pH8, 500 mM NaCl, 1 mM EDTA, 15 mM Imidazole, 0.1% TritonX100, 5% Glycerol, 1 mM DTT). Following sonication and 404 centrifugation, the clarified extract was loaded on a Ni Sepharose high-performance column (GE Healthcare Life Sciences, ref 17526802). Resin was 405 washed with Wash buffer (20 mM Tris pH8, 500 mM NaCl, 1 mM EDTA, 30 mM Imidazole, 1 mM DTT). Protein was eluted with Elution buffer (20 mM 406 Tris pH8, 500 mM NaCl, 1 mM EDTA, 300 mM Imidazole, 1 mM DTT). Purified protein was dialyzed overnight 4°C with storage buffer (20 mM Tris pH8, 407 150 mM NaCl, 1 mM EDTA, 1 mM DTT), concentrated with Amicon 3KD (Merck, ref UFC900324) to obtain concentration around 10  μM then 408 centrifuged at 160,000 g for 30 min. Aliquots were flash-frozen in liquid nitrogen and stored at −80°C. 409 Human Myosin VI construct (1021 amino-acids; Met1 – Ala1021) used in this study was similar to a previously published construct design49, but contained 410 a C-terminal GCN4 leucine zipper, for a dimerization sequence, and an eGFP added for fluorescence observation. A C -terminal FLAG sequence was 411 included for affinity purification. The Myosin VI construct was expressed in the presence of calmodulin, using the Sf9/baculovirus expression system 412 and purified using standard methods as described previously 50. Myosin VI samples were either stored as drops (20 μl per drop), or aliquoted into thin-413 walled PCR tubes (5 – 20 μl per tube), then frozen in liquid nitrogen for storage 50. 414 Slide silanization 415 Silane-PEG30K (Creative PegWorks) is dissolved at 1 mg/ml in 96% ethanol with 0.1% HCl, stirred several hours in the dark at 70 °C to completely 416 dissolve the silane and then stored in the dark. 417 Slides (and coverslips for patterns) are cleaned with 96% ethanol and rinsed with ddH2O, then sonicated 30 minutes at 60 °C in 2% Hellmanex. Slides 418 are then rinsed and stored in water overnight, then they are plasmatized for 5 minutes before dipping and storing them in the silane solution. Right 419 before use, they are rinsed in ethanol and abundantly washed in water and finally dried. 420 Microwells and micropatterns 421 Micropatterns are engraved on a silane-coated coverslip using a quartz mask under exposure to UV light. Silanized coverslips were put in tight contact 422 with a quartz-chrome printed photomask (Toppan Photomask). Tight contact was maintained using a vacuum holder. The PEG-Silane layer was burned 423 with deep UV (190nm) through the non-chromed windows of the photomask, using a UVO cleaner (Model No. 342A-220, Jelight), at a distance of 1cm 424 from the UV lamp with a power of 6mW/cm2, for 30 s. After UV exposure, slides are quickly detached from the mask with a suction pump and then 425 the sample is assembled. 426 For microwells, a SU8 mold with pillars was prepared using standard protocols and silanized with Trichloro(1H,1H,2H,2H-perfluoro-octyl)silane for 1 h 427 and heated for 1 h at 120°C. From the SU8 mold, a PDMS primary mold was prepared (Dow, SYLGARD 184 silicone elastomer kit) with a 1:10 w/w ratio 428 of curing agent. PDMS was cured at 70°C for at least 2 h. PDMS primary mold was then silanized with Trichloro(1H,1H,2H,2H-perfluoro-octyl)silane for 429 1 h and heated for 2 h at 100°C. PDMS was then poured on top of the PDMS primary mold to prepare the PDMS stamps. Coverslips were cleaned with 430 the following protocol: They were first wiped with ethanol (96%), then washed with water. They were then sonicated for 30 min in Hellmanex 2% at 431 60°C. After this second sonication, coverslips were rinsed in several volumes of mqH 2O and kept in water until use. Just before use, coverslips were 432 dried with compressed air. Finally, PDMS stamps were cut into pieces and placed on the coverslips with the pillars facing the coverslip. A droplet of 433 NOA 81 (Norland Products) was then placed on the side of the PDMS stamp, and NOA was allowed to go through the PDMS stamp by capillarity. When 434 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2025. ; https://doi.org/10.1101/2025.02.21.639071doi: bioRxiv preprint 15 the NOA filled stamp completely, it was polymerized with UV light for 10 min (UV KUB2/KLOE; 100% power). After polymerization of the NOA, PDMS 435 stamp was removed and the excess of NOA was cut. Then, an additional UV exposure of 2 min was done and the microwells were pl aced on a hot 436 plate at 60°C overnight to tightly bind the NOA to the glass. Wells have a diameter of 70 μm and a height of 50 μm. 437 Lipids/SUV preparation 438 l-α-phosphatidylcholine (EggPC; Avanti, 840051C) PEG-Biotin (DSPE-PEG(2000) Biotin, Sigma Aldrich) and Atto 390- labeled DOPE (Atto-390 DOPE, Atto 439 tec) were used. Lipids were mixed in glass tubes as follows: 98.5 % EggPC (10 mg/ml), 0.5 % DSPE-PEG(2000) Biotin and 1% Atto390-DOPE (1 mg/ml). 440 The mixture was dried with nitrogen gas. The dried lipids were incubated in a vacuum overnight. After that, the lipids were hydrated in the SUV buffer 441 (10 mM Tris (pH 7.4), 150 mM NaCl, 2 mM CaCl2) to the desired concentration. The mixture was tip-sonicated on ice for 10 min. The mixture was then 442 centrifuged for 10 min at 20,238 g to remove large structures. The supernatants were collected and stored at 4°C for no more than two weeks. 443 Sample preparation 444 For micropatterns, a flow chamber was assembled using double-edge tape (height 100 μm) placed on the coverslip containing the micropatterns. SUVs 445 were incubated for 10 minutes, washed with 0.5 mL of SUV buffer, followed by a 60 μl rinse with HKEM, a 5 minutes passivation with 1% BSA and by 446 again rinsing with 60 μl HKEM. The desired concentration of WA (60 μl) was incubated for 10 minutes and then rinsed abundantly with 0.5 mL HKEM, 447 before inserting 60 μl of the protein mix. 448 For microwells, a flow chamber was assembled using double -edge tape (height 1 80 μm) placed on the coverslip containing the microwells to stick 449 them on a silanized glass slide. The sample was rinsed once with SUV buffer, then 60 μl of solution containing the SUVs was incubated for 10 minutes, 450 washed abundantly with 1 mL SUV buffer, followed by 100 μl of HKEM buffer (50 mM KCl, 15 mM HEPES pH = 7.5, 5 mM MgCl 2, 1 mM EGTA). The 451 sample was passivated with 1% BSA (Sigma) for 5 minutes, then rinsed again with 100 μl HKEM. If necessary, the desired amount of WA was incubated, 452 diluted in HKEM for 10 minutes and rinsed again with 1 mL HKEM. Finally, 60 μl of the protein solution is flown in the chamber, which (after 20 s of 453 incubation to allow proteins to enter the wells and diffuse) is closed with paragon oil (Paragon scientific Viscosity Reference Standard RTM13). 454 All samples are immediately imaged on a TIRF microscope (Controlled with MetaMorph software) with a 100x objective (Olympus UApo N, 100x 1.49 455 Oil), with a 2 minutes time interval. The protein solution contains the desired concentration (diluted in HKEM buffer) of actin and profilin (1:1 ratio), 456 Myosin VI, 100 nM Arp2/3, 0.25 % methylcellulose, 2.7 mM ATP , 5 mM DTT, 0.2 mM DABCO. Actin polymerization curves are obtained by the total 457 integrated fluorescence of TIRF movies containing no myosin. Area coverage is obtained by thresholding the same movies to isolate the actin network 458 and then computing the percentage of pixels above threshold with respect to the number of pixels of the micropattern or microwell surface. 459 Individual filaments 460 To dope the sample with pre -polymerized filaments, 5 μM of actin (20% labelled with 647 -Actin) are polymerized with HKEM and left at room 461 temperature in the dark. Right before the experiment is started, 20 nM of actin is added. As the actin is not stabilized, care is taken of always adding 462 the monomeric actin first to avoid filaments’ depolymerization when diluted in the absence of monomers. Individual filaments are imaged with TIRF 463 microscopy with a 1 minute time interval while the rest of the network is imaged with a 2 minutes time interval. 464 Flow, tracking and fiber alignment 465 Actin flow is recorded with a custom made Python3 script using OpenCV’s optical flow library. Briefly, images of the SLB are thresholded to identify 466 the inside of the microwells or micropattern and its center, and thresholded a second time in the actin channel to isolate th e actin. The script then 467 computes for each pixel the flow of the fluorescence intensity. Each well (pattern) from the same condition is analyzed to obtain the mean speed over 468 time. The flow is averaged in a running window of 3 frames (6 minutes) and then all the tracks from different microwells are averaged. This also allows 469 to compute the mean flow at each position inside the wells. To compute the mean flow and orientation across all wells, wells are aligned and, in each 470 position of the sample, the flow from different wells in the same condition is averaged together using only data at steady state . 471 Individual filaments’ tracking and fiber alignment are obtained using scripts adapted from 51. 472 Filaments are tracked by identifying elongated contours, and trajectories are reconstructed based on their position in consecutive frames. Briefly, if 473 two contours in consecutive frames are the closest ones, they are assumed to be part of the same trajectory provided their sh ape is similar enough 474 and their distance is below a threshold of 5 μm. From these trajectories we can extract the position and speed of filaments across different wells in 475 the same conditions, which are used to obtain the speed distribution, the radial and the tangential flow based on the alignment of the speed vector v 476 with the radial versor r having as origin the center of the circular well (i.e. by the dot product r.v), and its perpendicular versor t (tangential flow t.v). 477 The alignment of fibers (unit vector n at each position) is instead obtained by the local intensity gradient which is assumed to be perpendicular to the 478 mean alignment direction. The gradient is computed, at each position, as the average a box of side 2 μm centered at the given point, from which the 479 mean direction of the gradient is extracted and hence the mean alignment of fibers perpendicular to it. This data can also be used to compute locally 480 the nematic tensor, whose eigenvalue is the order parameter at each position. The procedure is explained in 51. The mean alignment across different 481 wells in the same conditions is obtained by aligning wells, using the local orientation n of each well to build an effective nematic tensor at each position 482 on the surface of the well, whose biggest eigenvector is the local alignment averages over all wells, while the corresponding eigenvalue indicates how 483 coherent such alignment is across different wells and is used to compute the magnitude of the vectors in the figures. 484 All the mean quantities (flow, orientation, order parameter) are obtained at steady state and computed in the time between 1 hour after 485 polymerization and 3 hours after polymerization. All values are given as mean and standard deviation, with the number of replicates indicated. 486 Ablation and confocal 487 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2025. ; https://doi.org/10.1101/2025.02.21.639071doi: bioRxiv preprint 16 Confocal recordings are obtained by spinning disk confocal (Nikon Ti Eclipse, equipped with a spinning scanning unit CSU-X1 Yokogawa and a R3 retiga 488 camera from QImaging). Photoablation was performed using the iLas2 device (Gataca Systems) equipped with a passively Q -switched laser (STV-489 E,ReamPhotonics) at 355 nm producing 500 picosecond pulses. Laser displacement, exposure time and repetition rate were controlled via ILas software 490 interfaced with MetaMorph (Universal Imaging Corporation). Laser photoablation and subsequent imaging were performed with a CFI Super Fluor 491 100X/1.3 NA oil objective. Stacks for each well are taken at low laser intensity and wide z -step (~ 1 μm) to avoid photobleaching. Lines to cut the 492 bottom surface of the wall or small (1 μm2) circular spots in 3D to cut cables are illuminated, while the bottom or the whole 3D volume is imaged. 493 3D tracking of fibers 494 To track filaments in 3D starting from confocal data, first bright spots are identified by thresholding at different heights in the same time frame. 495 Filaments are reconstructed by binding together two spots i and j if they are, at consecutive slides in z, closer than 3 μm, provided that no other bright 496 spot is closer to either i or j. From this, filaments are obtained and their center of mass is computed. Tracks in time are reconstructed by connecting 497 together centers of mass in consecutive time frames as in 2D, but computing their relative distances in 3D and connecting trajectories closer than 3 498 μm in the XY plane and of 5 μm in the Z direction, again provided no other center of mass is closer. 499 500 501 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2025. ; https://doi.org/10.1101/2025.02.21.639071doi: bioRxiv preprint 17

502 503 (1) Lappalainen, P .; Kotila, T.; Jégou, A.; Romet-Lemonne, G. Biochemical and Mechanical Regulation of 504 Actin Dynamics. Nature Reviews Molecular Cell Biology 2022, 23 (12), 836–852. 505 (2) Lomakin, A. J.; Lee, K.-C.; Han, S. J.; Bui, D. A.; Davidson, M.; Mogilner, A.; Danuser, G. Competition for 506 Actin between Two Distinct F-Actin Networks Defines a Bistable Switch for Cell Polarization. Nature 507 cell biology 2015, 17 (11), 1435–1445. 508 (3) Rafelski, S. M.; Theriot, J. A. Crawling toward a Unified Model of Cell Motility: Spatial and Temporal 509 Regulation of Actin Dynamics. Annual review of biochemistry 2004, 73 (1), 209–239. 510 (4) Theriot, J. A.; Mitchison, T. J. Actin Microfilament Dynamics in Locomoting Cells. Nature 1991, 352 511 (6331), 126–131. 512 (5) Vitriol, E. A.; McMillen, L. M.; Kapustina, M.; Gomez, S. M.; Vavylonis, D.; Zheng, J. Q. Two Functionally 513 Distinct Sources of Actin Monomers Supply the Leading Edge of Lamellipodia. Cell reports 2015, 11 (3), 514 433–445. 515 (6) Goode, B. L.; Eskin, J.; Shekhar, S. Mechanisms of Actin Disassembly and Turnover. Journal of Cell 516 Biology 2023, 222 (12), e202309021. 517 (7) Banerjee, S.; Gardel, M. L.; Schwarz, U. S. The Actin Cytoskeleton as an Active Adaptive Material. Annu. 518 Rev. Condens. Matter Phys. 2020, 11 (1), 421–439. https://doi.org/10.1146/annurev-conmatphys-519 031218-013231. 520 (8) Vignaud, T.; Copos, C.; Leterrier, C.; Toro-Nahuelpan, M.; Tseng, Q.; Mahamid, J.; Blanchoin, L.; 521 Mogilner, A.; Théry, M.; Kurzawa, L. Stress Fibres Are Embedded in a Contractile Cortical Network. 522 Nature materials 2021, 20 (3), 410–420. 523 (9) Hu, S.; Dasbiswas, K.; Guo, Z.; Tee, Y .-H.; Thiagarajan, V .; Hersen, P .; Chew, T.-L.; Safran, S. A.; Zaidel-Bar, 524 R.; Bershadsky, A. D. Long-Range Self-Organization of Cytoskeletal Myosin II Filament Stacks. Nature 525 cell biology 2017, 19 (2), 133–141. 526 (10) Letort, G.; Ennomani, H.; Gressin, L.; Théry, M.; Blanchoin, L. Dynamic Reorganization of the Actin 527 Cytoskeleton. F1000Research 2015, 4. 528 (11) Théry, M.; Blanchoin, L. Reconstituting the Dynamic Steady States of Actin Networks in Vitro. Nature 529 Cell Biology 2024, 26 (4), 494–497. 530 (12) Ierushalmi, N.; Malik-Garbi, M.; Manhart, A.; Abu Shah, E.; Goode, B. L.; Mogilner, A.; Keren, K. 531 Centering and Symmetry Breaking in Confined Contracting Actomyosin Networks. Elife 2020, 9, 532 e55368. 533 (13) Malik-Garbi, M.; Ierushalmi, N.; Jansen, S.; Abu-Shah, E.; Goode, B. L.; Mogilner, A.; Keren, K. Scaling 534 Behaviour in Steady-State Contracting Actomyosin Networks. Nature physics 2019, 15 (5), 509–516. 535 (14) Krishna, A.; Savinov, M.; Ierushalmi, N.; Mogilner, A.; Keren, K. Size-Dependent Transition from Steady 536 Contraction to Waves in Actomyosin Networks with Turnover. Nature Physics 2024, 20 (1), 123–134. 537 (15) Sakamoto, R.; Izri, Z.; Shimamoto, Y .; Miyazaki, M.; Maeda, Y . T. Geometric Trade-off between 538 Contractile Force and Viscous Drag Determines the Actomyosin-Based Motility of a Cell-Sized Droplet. 539 Proceedings of the National Academy of Sciences 2022, 119 (30), e2121147119. 540 https://doi.org/10.1073/pnas.2121147119. 541 (16) Sakamoto, R.; Miyazaki, M.; Maeda, Y . T. State Transitions of a Confined Actomyosin System Controlled 542 through Contractility and Polymerization Rate. Phys. Rev. Research 2023, 5 (1), 013208. 543 https://doi.org/10.1103/PhysRevResearch.5.013208. 544 (17) Tan, T. H.; Malik-Garbi, M.; Abu-Shah, E.; Li, J.; Sharma, A.; MacKintosh, F. C.; Keren, K.; Schmidt, C. F.; 545 Fakhri, N. Self-Organized Stress Patterns Drive State Transitions in Actin Cortices. Sci. Adv. 2018, 4 (6), 546 eaar2847. https://doi.org/10.1126/sciadv.aar2847. 547 (18) Pinot, M.; Steiner, V.; Dehapiot, B.; Yoo, B.-K.; Chesnel, F.; Blanchoin, L.; Kervrann, C.; Gueroui, Z. 548 Confinement Induces Actin Flow in a Meiotic Cytoplasm. Proc. Natl. Acad. Sci. U.S.A. 2012, 109 (29), 549 11705–11710. https://doi.org/10.1073/pnas.1121583109. 550 (19) Colin, A.; Kotila, T.; Guérin, C.; Orhant-Prioux, M.; Vianay, B.; Mogilner, A.; Lappalainen, P .; Théry, M.; 551 Blanchoin, L. Recycling of the Actin Monomer Pool Limits the Lifetime of Network Turnover. The EMBO 552 Journal 2023, 42 (9), e112717. https://doi.org/10.15252/embj.2022112717. 553 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2025. ; https://doi.org/10.1101/2025.02.21.639071doi: bioRxiv preprint 18 (20) Bleicher, P .; Sciortino, A.; Bausch, A. R. The Dynamics of Actin Network Turnover Is Self-Organized by a 554 Growth-Depletion Feedback. Scientific Reports 2020, 10 (1), 6215. https://doi.org/10.1038/s41598-555 020-62942-8. 556 (21) Bendix, P . M.; Koenderink, G. H.; Cuvelier, D.; Dogic, Z.; Koeleman, B. N.; Brieher, W. M.; Field, C. M.; 557 Mahadevan, L.; Weitz, D. A. A Quantitative Analysis of Contractility in Active Cytoskeletal Protein 558 Networks. Biophysical Journal 2008, 94 (8), 3126–3136. https://doi.org/10.1529/biophysj.107.117960. 559 (22) Colin, A.; Orhant-Prioux, M.; Guérin, C.; Savinov, M.; Cao, W.; Vianay, B.; Scarfone, I.; Roux, A.; De La 560 Cruz, E. M.; Mogilner, A.; Théry, M.; Blanchoin, L. Friction Patterns Guide Actin Network Contraction. 561 Proc. Natl. Acad. Sci. U.S.A. 2023, 120 (39), e2300416120. https://doi.org/10.1073/pnas.2300416120. 562 (23) Claessens, M. M. A. E.; Tharmann, R.; Kroy, K.; Bausch, A. R. Microstructure and Viscoelasticity of 563 Confined Semiflexible Polymer Networks. Nature Physics 2006, 2 (3), 186–189. 564 https://doi.org/10.1038/nphys241. 565 (24) Köhler, S.; Bausch, A. R. Contraction Mechanisms in Composite Active Actin Networks. PLoS ONE 2012, 566 7 (7), e39869. https://doi.org/10.1371/journal.pone.0039869. 567 (25) Belmonte, J. M.; Leptin, M.; Nédélec, F. A Theory That Predicts Behaviors of Disordered Cytoskeletal 568 Networks. Molecular Systems Biology 2017, 13 (9), 941. https://doi.org/10.15252/msb.20177796. 569 (26) Reymann, A. C.; Boujemaa-Paterski, R.; Martiel, J. L.; Guérin, C.; Cao, W.; Chin, H. F.; De La Cruz, E. M.; 570 Théry, M.; Blanchoin, L. Actin Network Architecture Can Determine Myosin Motor Activity. Science 571 2012, 336 (6086), 1310–1314. https://doi.org/10.1126/science.1221708. 572 (27) Linsmeier, I.; Banerjee, S.; Oakes, P . W.; Jung, W.; Kim, T.; Murrell, M. P . Disordered Actomyosin 573 Networks Are Sufficient to Produce Cooperative and Telescopic Contractility. Nature Communications 574 2016, 7, 12615. https://doi.org/10.1038/ncomms12615. 575 (28) Sonal; Ganzinger, K. A.; Vogel, S. K.; Mücksch, J.; Blumhardt, P .; Schwille, P . Myosin-II Activity 576 Generates a Dynamic Steady State with Continuous Actin Turnover in a Minimal Actin Cortex. Journal 577 of Cell Science 2019, 132 (4), jcs219899. 578 (29) Seara, D. S.; Yadav, V.; Linsmeier, I.; Tabatabai, A. P .; Oakes, P . W.; Tabei, S. M. A.; Banerjee, S.; Murrell, 579 M. P . Entropy Production Rate Is Maximized in Non-Contractile Actomyosin. Nature Communications 580 2018, 9 (1), 1–10. https://doi.org/10.1038/s41467-018-07413-5. 581 (30) Dürre, K.; Keber, F. C.; Bleicher, P .; Brauns, F.; Cyron, C. J.; Faix, J.; Bausch, A. R. Capping Protein-582 Controlled Actin Polymerization Shapes Lipid Membranes. Nature Communications 2018, 9 (1), 1–11. 583 https://doi.org/10.1038/s41467-018-03918-1. 584 (31) Loiseau, E.; Schneider, J. A. M.; Keber, F. C.; Pelzl, C.; Massiera, G.; Salbreux, G.; Bausch, A. R. Shape 585 Remodeling and Blebbing of Active Cytoskeletal Vesicles. Science Advances 2016, 2 (4), e1500465–586 e1500465. https://doi.org/10.1126/sciadv.1500465. 587 (32) Sakamoto, R.; Murrell, M. P . F-Actin Architecture Determines the Conversion of Chemical Energy into 588 Mechanical Work. Nature Communications 2024, 15 (1), 3444. 589 (33) Vignaud, T.; Blanchoin, L.; Théry, M. Directed Cytoskeleton Self-Organization. Trends in cell biology 590 2012, 22 (12), 671–682. 591 (34) Ennomani, H.; Letort, G.; Guérin, C.; Martiel, J. L.; Cao, W.; Nédélec, F.; De La Cruz, E. M.; Théry, M.; 592 Blanchoin, L. Architecture and Connectivity Govern Actin Network Contractility. Current Biology 2016, 593 26 (5), 616–626. https://doi.org/10.1016/j.cub.2015.12.069. 594 (35) Yamamoto, S.; Gaillard, J.; Vianay, B.; Guerin, C.; Orhant-Prioux, M.; Blanchoin, L.; Théry, M. Actin 595 Network Architecture Can Ensure Robust Centering or Sensitive Decentering of the Centrosome. The 596 EMBO Journal 2022, e111631. https://doi.org/10.15252/embj.2022111631. 597 (36) Pollard, T. D.; Blanchoin, L.; Mullins, R. D. Molecular Mechanisms Controlling Actin Filament Dynamics 598 in Nonmuscle Cells. Annu. Rev. Biophys. Biomol. Struct. 2000, 29 (1), 545–576. 599 https://doi.org/10.1146/annurev.biophys.29.1.545. 600 (37) Sciortino, A.; Bausch, A. R. Pattern Formation and Polarity Sorting of Driven Actin Filaments on Lipid 601 Membranes. Proceedings of the National Academy of Sciences 2021, 118 (6), e2017047118. 602 https://doi.org/10.1073/pnas.2017047118. 603 (38) Grover, R.; Fischer, J.; Schwarz, F.; Walter, W.; Schwille, P .; Diez, S. Transport Efficiency of Membrane-604 Anchored Kinesin-1 Motors Depends on Motor Density and Diffusivity. bioRxiv 2016, 064246. 605 https://doi.org/10.1101/064246. 606 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2025. ; https://doi.org/10.1101/2025.02.21.639071doi: bioRxiv preprint 19 (39) Pernier, J.; Morchain, A.; Caorsi, V.; Bertin, A.; Bousquet, H.; Bassereau, P .; Coudrier, E. Myosin 1b 607 Flattens and Prunes Branched Actin Filaments. Journal of Cell Science 2020, 133 (18). 608 https://doi.org/10.1242/jcs.247403. 609 (40) Rogez, B.; Würthner, L.; Petrova, A. B.; Zierhut, F. B.; Saczko-Brack, D.; Huergo, M.-A.; Batters, C.; Frey, 610 E.; Veigel, C. Reconstitution Reveals How Myosin-VI Self-Organises to Generate a Dynamic Mechanism 611 of Membrane Sculpting. Nature Communications 2019, 10 (1), 3305. https://doi.org/10.1038/s41467-612 019-11268-9. 613 (41) Rock, R. S.; Rice, S. E.; Wells, A. L.; Purcell, T. J.; Spudich, J. A.; Sweeney, H. L. Myosin VI Is a Processive 614 Motor with a Large Step Size. Proc. Natl. Acad. Sci. U.S.A. 2001, 98 (24), 13655–13659. 615 https://doi.org/10.1073/pnas.191512398. 616 (42) Aroush, D. R.-B.; Ofer, N.; Abu-Shah, E.; Allard, J.; Krichevsky, O.; Mogilner, A.; Keren, K. Actin Turnover 617 in Lamellipodial Fragments. Current Biology 2017, 27 (19), 2963–2973. 618 (43) Burnette, D. T.; Shao, L.; Ott, C.; Pasapera, A. M.; Fischer, R. S.; Baird, M. A.; Der Loughian, C.; Delanoe-619 Ayari, H.; Paszek, M. J.; Davidson, M. W. A Contractile and Counterbalancing Adhesion System Controls 620 the 3D Shape of Crawling Cells. Journal of Cell Biology 2014, 205 (1), 83–96. 621 (44) Burnette, D. T.; Manley, S.; Sengupta, P .; Sougrat, R.; Davidson, M. W.; Kachar, B.; Lippincott-Schwartz, 622 J. A Role for Actin Arcs in the Leading-Edge Advance of Migrating Cells. Nature cell biology 2011, 13 623 (4), 371–382. 624 (45) Tojkander, S.; Gateva, G.; Husain, A.; Krishnan, R.; Lappalainen, P . Generation of Contractile 625 Actomyosin Bundles Depends on Mechanosensitive Actin Filament Assembly and Disassembly. Elife 626 2015, 4, e06126. 627 (46) Spudich, J. A.; Watt, S. The Regulation of Rabbit Skeletal Muscle Contraction: I. Biochemical Studies of 628 the Interaction of the Tropomyosin-Troponin Complex with Actin and the Proteolytic Fragments of 629 Myosin. Journal of biological chemistry 1971, 246 (15), 4866–4871. 630 (47) Isambert, H.; Venier, P .; Maggs, A. C.; Fattoum, A.; Kassab, R.; Pantaloni, D.; Carlier, M.-F. Flexibility of 631 Actin Filaments Derived from Thermal Fluctuations: Effect of Bound Nucleotide, Phalloidin, and 632 Muscle Regulatory Proteins. Journal of Biological Chemistry 1995, 270 (19), 11437–11444. 633 (48) Almo, S. C.; Smith, D. L.; Danishefsky, A. T.; Ringe, D. The Structural Basis for the Altered Substrate 634 Specificity of the R292D Active Site Mutant of Aspartate Aminotransferase from E. Coli. Protein 635 Engineering, Design and Selection 1994, 7 (3), 405–412. 636 (49) Nishikawa, S.; Arimoto, I.; Ikezaki, K.; Sugawa, M.; Ueno, H.; Komori, T.; Iwane, A. H.; Yanagida, T. 637 Switch between Large Hand-over-Hand and Small Inchworm-like Steps in Myosin VI. Cell 2010, 142 (6), 638 879–888. 639 (50) Wang, F.; Chen, L.; Arcucci, O.; Harvey, E. V.; Bowers, B.; Xu, Y .; Hammer, J. A.; Sellers, J. R. Effect of ADP 640 and Ionic Strength on the Kinetic and Motile Properties of Recombinant Mouse Myosin V. Journal of 641 Biological Chemistry 2000, 275 (6), 4329–4335. 642 (51) Sciortino, A.; Neumann, L. J.; Krüger, T.; Maryshev, I.; Teshima, T. F.; Wolfrum, B.; Frey, E.; Bausch, A. R. 643 Polarity and Chirality Control of an Active Fluid by Passive Nematic Defects. Nature Materials 2023, 22 644 (2), 260–268. 645 (52) Goehring, N. W.; Chowdhury, D.; Hyman, A. A.; Grill, S. W. FRAP Analysis of Membrane-Associated 646 Proteins: Lateral Diffusion and Membrane-Cytoplasmic Exchange. Biophysical Journal 2010, 99 (8), 647 2443–2452. https://doi.org/10.1016/j.bpj.2010.08.033. 648 649 650 651 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2025. ; https://doi.org/10.1101/2025.02.21.639071doi: bioRxiv preprint 20 List of Movies 652 Movie S1: Actin polymerization on micropatterns. TIRF imaging of branched actin network assembly on lipid 653 micropattern, related to Fig 1 B-C. Conditions are 1 μM actin, 20 nM WA, 50 nM Arp2/3 complex, 2 μM profilin. White 654 circle indicates the position of the micropattern. 1 image was taken every 1 minute over 1 hour. The movie was 655 compressed in JPEG at 10 frames per second. 656 Movie S2: Actomyosin dynamics on micropatterns. TIRF imaging of the formation and dynamics of a disordered (left) 657 and globally contractile (right) actomyosin network on lipid micropatterns. Conditions are 0.5 μM actin, 20 nM WA, 50 658 nM Arp2/3 complex, 2 μM profilin and 3.3 nM myosin (left, disordered) and the same but with 1 μM actin on the right 659 (global contraction). White circle indicates the position of the micropattern. 1 image was taken every 2 minute over 2 660 hours. The movie was compressed in JPEG at 10 frames per second. 661 Movie S3: Actin polymerization in microwells. TIRF imaging of branched actin network assembly on supported lipid 662 bilayer inside microwells, related to Fig. 2 B-D. Conditions are 1 μM actin, 10 nM WA, 100 nM Arp2/3 complex, 2 μM 663 profilin. White circle indicates the position of the microwell. 1 image was taken every 2 minutes over 48 minutes. The 664 movie was compressed in JPEG at 10 frames per second. 665 Movie S4: Dynamic steady state (DSS) inside microwells. TIRF imaging of the formation of the dynamic steady state 666 inside microwells, related to Fig. 2E, bottom. Conditions are 1 μM actin, 10 nM WA, 100 nM Arp2/3 complex, 2 μM 667 profilin and 20 nM myosin VI. 6 different wells are shown. 1 image was taken every 2 minutes over 8 hours. The movie 668 was compressed in JPEG at 10 frames per second. 669 Movie S5: Dynamic steady state (DSS) inside microwells in different conditions. TIRF imaging of the formation of the 670 dynamic steady state inside microwells, related to Fig. 2G-K, bottom. Conditions are 2 μM actin and 20 nM myosin VI 671 (top left), 1 μM actin and 50 nM myosin VI (top right), 4 μM actin and 20 nM myosin VI (bottom left), 4 μM actin and 672 50 nM myosin VI (bottom right). All wells are at 10 nM WA, 100 nM Arp2/3 complex, 2 μM profilin. 1 image was taken 673 every 2 minutes. The movie was compressed in JPEG at 10 frames per second. 674 Movie S6: DSS in microwells does not depend on actin turnover. TIRF imaging of the formation of the dynamic steady 675 state inside microwells, even if actin assembly rate and turnover is varied. Related to Fig. 3A-C. Turnover is modulated 676 through variation of WA (100 nM WA and 100 nM Arp2/3 complex, left), addition of phalloidin (1 μM phalloidin, 677 center) and removal of the Arp2/3 complex (no profilin and no Arp2/3 complex, right). Otherwise, conditions are 1 μM 678 actin, 2 μM profilin and 20 nM myosin. 1 image was taken every 2 minutes over 1 hour and 45 minutes. The movie was 679 compressed in JPEG at 10 frames per second. 680 Movie S7: Presence of gliding inside microwells. TIRF imaging of the initial states of actin polymerization inside 681 microwells in the presence of myosin, revealing the presence of gliding that extend actin clusters. Related to Fig. 3D. 682 Conditions are 1 μM actin, 10 nM WA, 100 nM Arp2/3 complex, 2 μM profilin. 1 image was taken every 10 seconds 683 over 2 minutes. The movie was compressed in JPEG at 5 frames per second. 684 Movie S8: Observation of individual filaments inside the actin network. TIRF imaging of the dynamic steady state 685 inside microwells in the presence of individual filaments labelled with a different fluorophore (cyan). The rest of the 686 network is shown in red. Related to Fig. 3E. Conditions are 1 μM actin, 10 nM WA, 100 nM Arp2/3 complex, 2 μM 687 profilin, 20 nM of individual filaments. 1 image was taken every 2 minutes for two hours. The movie was compressed 688 in JPEG at 5 frames per second. 689 Movie S9: Speckled actin reveals the microscopic motion of the network. TIRF imaging of the contractile network 690 with low content of labelled actin (0.2 %) to obtain actin “speckles”, allowing visualization of the microscopic motion of 691 filaments. Related to Fig. S8. Conditions are 1 μM actin, 10 nM WA, 20 nM myosin, 100 nM Arp2/3 complex, 2 μM 692 profilin. 1 image was taken every 10 seconds over 420 seconds. The movie was compressed in JPEG at 10 frames per 693 second. 694 Movie S10: 3D reconstruction of the actin network inside microwells. Confocal imaging of the network in three 695 dimensions, rotated for illustration. Related to Fig. 4A. Conditions are 1 μM actin, 50 nM myosin VI. The movie was 696 compressed in JPEG at 10 frames per second. 697 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2025. ; https://doi.org/10.1101/2025.02.21.639071doi: bioRxiv preprint 21 Movie S11: Visualization of individual filaments inside the microwell in 3D. Confocal imaging of individual filaments, 698 side view, related to Fig. 4B. Individual filaments hand tracked are shown, related to Fig. 4F. Conditions are 1 μM 699 unlabelled actin, 50 nM myosin VI, 20 nM labelled filaments. One image is taken every 45 seconds for 14 minutes and 700 15 seconds. The movie was compressed in JPEG at 5 frames per second. 701 Movie S12: Visualization of individual filaments inside the microwell in 3D (slice). Confocal imaging of individual 702 filaments in a small slice of 5 μm close to the well’s bottom, side view. Related to Fig. 4 C-E. Conditions are 1 μM 703 unlabelled actin, 50 nM myosin VI, 20 nM labelled filaments. One image is taken every 45 seconds for 14 minutes and 704 15 seconds. The movie was compressed in JPEG at 5 frames per second. 705 Movie S13: Ablation in microwells. Confocal imaging of the well’s bottom before, during and after ablation. Related to 706 Fig. 4 G-I. Conditions are 1 μM actin, 50 nM myosin VI. Ablation time is marked with three asterisks and the time is 707 with respect to the ablation time. One image is taken every 2 minutes for 4 hours and 26 minutes. The movie was 708 compressed in JPEG at 20 frames per second. 709 Movie S14: Control experiment of ablation in microwells (without severing). Confocal imaging of the well’s bottom 710 before, during and after ablation (without severing). Related to Fig. 4 G-I. Conditions are 1 μM actin, 50 nM myosin VI. 711 Ablation time is marked with three asterisks and the time is with respect to the ablation time. One image is taken 712 every 2 minutes for 4 hours and 26 minutes. The movie was compressed in JPEG at 20 frames per second. 713 714 715 716 717 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2025. ; https://doi.org/10.1101/2025.02.21.639071doi: bioRxiv preprint 22 Supporting Figures 718 719 Figure S1: a) Time lapse of photobleaching and recovery of the supported lipid bilayer on the top and bottom layer of a microwell. Scale bar is 35 μm, 720 time interval between frames is 1.5 s. b-d) FRAP (Fluorescence Recovery After Photobleaching) curves for different experimental conditions: a 721 micropattern (a), a microwell (b) and comparison between top and bottom of a microwell (c). The bleached area is a rectangular region and the 722 resulting recovery curve is fit as in Reference 52 (dashed line) to obtain a diffusion coeffient D=(2.8±0.3) μm2/s (fit result, MEAN+STD). 723 724 Figure S2: Divergence of the actin optical flow velocity for different conditions as a function of the distance from the center, showing overall 725 contraction (negative divergence). 726 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2025. ; https://doi.org/10.1101/2025.02.21.639071doi: bioRxiv preprint 23 727 728 Figure S3: Complete time trace of the mean flow of an actin network at 1 μM Actin, 10 nM WA, 100 nM Arp2/3 complex, 1 μM profilin and 20 nM 729 myosin VI, showing roughly constant, slowly decaying flow for up to 8 hours (n=7 wells). 730 731 Figure S4: a-b) Comparison of myosin VI behavior between DSS (a) and global contraction (b) inside microwells. In the case of DSS (a), actin (red) and 732 fluorescent myosin VI (green) are superimposable, both at steady-state (left) and over time. In the case of global contraction (b), myosin collapses 733 actin clusters and accumulates at its center. Conditions in (a) are 1 μM Actin, 10 nM WA and 20 nM myosin VI; in (b) actin is 4 μM, other conditions 734 are the same. Scale bar is 35 μm, time interval is 10 minutes. 735 736 Figure S5: a-b) Time lapse of the variation of the actin fluorescence intensity at the bottom of a microwells in two different conditions, over the first 737 two hours, showing an almost constant actin organization. Scale bar is 20 μm, time interval is 10 minutes. Conditions for (a) are 1 μM, 10 nM WA, 100 738 nM Arp2/3 complex and 20 nM myosin, conditions for (b) are the same but actin is at 2 μM. c-d) Analysis of the actin density on the bottom layer over 739 time for two different samples: (c) 1 μM actin (n=7), (d) 2 μM actin (n=9). Both samples show constant density on the bottom of the microwells (c-d, 740 left plots) and a constant radial profile of the actin intensity averaged between 1 and 3 hours of experiment (c-d, right plots). Left plots are normalized 741 by the final actin intensity, right plots show the integrated intensity of actin inside the microwells on the surface. 742 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2025. ; https://doi.org/10.1101/2025.02.21.639071doi: bioRxiv preprint 24 743 Figure S6: a) Phase diagram at different actin, WA and myosin VI concentrations. Unless specified, parameters are: 1:1 actin:profilin molar ratio, 100 744 nM Arp2/3 complex, 20 nM myosin VI. Data in the absence of the Arp2/3 complex (labelled as “no Arp2/3”) is acquired in the absence of profilin. 745 Actin and WA are indicated on the x and y axis, myosin VI variations are explicitly mentioned, and arrows connect the same conditions (except for 746 myosin VI concentration). b) Behavior of the system at 1 μM actin, 10 nM WA and 5 nM myosin VI, showing no DSS-like behavior. Total time is 1 hour, 747 time interval is 15 m. Scale bar is 35 μm. 748 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2025. ; https://doi.org/10.1101/2025.02.21.639071doi: bioRxiv preprint 25 749 Figure S7: a-b) Data obtained from tracking of individual filaments inside the network. Histograms of speeds (a) and radial and tangential flux (b) for 750 different conditions, all showing similar behavior. 751 752 Figure S8: a) Image of the network at 1 μM Actin, 10 nM WA, 20 nM myosin VI, 100 nM Arp2/3 complex, 1 μM profilin with only 0.2% of the network 753 labelled to obtain speckles. b) Enlargements of a central and a peripheral zone. c) Kymographs of the motion of bright spots along the red dashed 754 lines in (b) showing transport in both directions both at the peripheral bundles and in the center. Top, periphery. Bottom, center. All scale bars are 35 755 μm. Related to Movie S9. 756 757 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2025. ; https://doi.org/10.1101/2025.02.21.639071doi: bioRxiv preprint