Reversal of contractility as a signature of selforganization in cytoskeletal bundles
Abstract
Bundles of cytoskeletal filaments and molecular motors generate motion in living cells, and have internal structures ranging from very organized to apparently disordered. The mechanisms powering the disordered structures are debated, and existing models predominantly predict that they are contractile. We reexamine this prediction through a theoretical treatment of the interplay between three wellcharacterized internal dynamical processes in cytoskeletal bundles: filament assembly and disassembly, the attachementdetachment dynamics of motors and that of crosslinking proteins. The resulting selforganization is easily understood in terms of motor and crosslink localization, and allows for an extensive control of the active bundle mechanics, including reversals of the filaments’ apparent velocities and the possibility of generating extension instead of contraction. This reversal mirrors some recent experimental observations, and provides a robust criterion to experimentally elucidate the underpinnings of both actomyosin activity and the dynamics of microtubule/motor assemblies in vitro as well as in diverse intracellular structures ranging from contractile bundles to the mitotic spindle.
Introduction
Many cellular functions, from motility to cell division, are driven by molecular motors exerting forces on actin filaments or microtubules held together by crosslinking proteins. This wide variety of processes is powered by an equally wide range of structures, many of which do not display any apparent spatial organization of their components (Verkhovsky et al., 1995; Cramer et al., 1997; Medalia et al., 2002; Kamasaki et al., 2007). While actomyosin structures are overwhelmingly observed to contract (Murrell et al., 2015), the mechanisms underlying this contraction are unclear, as individual motors can in principle elicit extension just as easily as contraction (Figure 1a–b; Hatano, 1994; Sekimoto and Nakazawa, 1998; Lenz et al., 2012a; Mendes Pinto et al., 2013).
Recent investigations into this breaking of symmetry between contraction and extension have focused on three classes of models. The first one is specific to actin filaments, which are very flexible, as it is based on the idea that mechanical nonlinearities, for example, the buckling of individual filaments under compression could suppress the propagation of extensile forces and thus favor contraction (Dasanayake et al., 2011; Lenz et al., 2012b; Ronceray et al., 2016). The second mechanism is relevant for microtubulebased systems, where some (but not all) motors may dwell on the ends of the filaments, which transiently generating a type of organization similar to that found in muscle (Foster et al., 2015; Tan et al., 2018). Finally, in the third type of models the spatial selforganization of the bundle’s motors and crosslinks along undeformable, rodlike actin filaments leads to contraction (Kruse and Sekimoto, 2002; Zumdieck et al., 2007; Zemel and Mogilner, 2009; Oelz et al., 2015; Koenderink and Paluch, 2018). So far, opportunities to discriminate between these models experimentally remain very limited for lack of a clear theoretical prediction setting one apart from the others.
Here we provide such a prediction, namely that the selforganization mechanism implies that filamentmotor bundles robustly extend if taken to certain parameter regimes. This stark qualitative change from contraction to extension is easily detectable experimentally and is not expected in models based on mechanical nonlinearities or motor dwelling. This prediction could also explain the illunderstood extension observed in some microtubule systems (Sanchez et al., 2012; Keber et al., 2014; Roostalu et al., 2018). Our prediction crucially rests on a simultaneous theoretical treatment of the filament polymerization, motor and crosslink dynamics detailed in Sec. ‘Model’. Previous studies of contractility mechanisms only involved partial treatments, whereby the time scale associated with one of these dynamics was effectively assumed to be much larger or smaller than the others. We show in Sec. ‘Selforganization and force distribution’ that the coupled dynamics of these three elements induces a spatial organization of motors and crosslinks along the filaments. Sec. ‘Velocity selection and bundle tension’ then demonstrates that the resulting localization of the motors and crosslinks in the vicinity of the filament ends leads to a switch between contraction and extension, as schematized in Figure 1c. In Sec. ‘Qualitative predictions’ , we discuss the qualitative physics underlying this switch and its experimental relevance, and show that extension arises when the motor runlength and unbinding rate are relatively large compared to the filament length and the crosslink unbinding rate, respectively. We quantitatively compare the resulting tensions to those expected from simple mechanical nonlinearity and motor dwelling models in Sec. ‘Quantitative aspects and alternative models’ . We find that selforganization dominates in tightly connected actomyosin bundles, where filament buckling is hampered, and that it outcompetes dwelling in microtubule systems where the filaments are long enough for enddwelling to be a rare occurrence. Finally, we discuss the conceptual implications of these simple, widely applicable ideas for the understanding of selforganization in active filamentmotor systems.
Model
We consider a bundle of polar filaments of length $L$ aligned in the $x$direction and subjected to periodic boundary conditions. The filaments are rigid, ruling out contraction arising from mechanical nonlinearities (Lenz et al., 2012a). The motor velocity does not depend on its position on the filament, ruling out contraction arising from enddwelling (Liverpool and Marchetti, 2005). A filament may point in the direction of positive or negative $x$, and maintains this polarity throughout the dynamics. At steadystate, filaments constantly grow from their plus ends and shrink from their minus ends at a fixed velocity ${v}_{t}$, a phenomenon known as ‘treadmilling’ in actin throughout which their length remains constant (Alberts et al., 2015; Figure 2a). While the dynamics of microtubules proceeds through somewhat different mechanisms, the insights gained from this simple, analytically tractable dynamics are also applicable there, as further described in the Discussion. Motors and crosslinks constantly bind and unbind from filaments, and we denote by ${\tau}_{m}$ (${\tau}_{c}$) and ${\rho}_{m}^{0}$ (${\rho}_{c}^{0}$) the average motor (crosslink) unbinding time and equilibrium density (Figure 2b). Finally we neglect viscous drag, as further argued in the Discussion.
Once bound to a filament, motors slide towards its plus end with a velocity ${v}_{m}$. The value of ${v}_{m}$ is set by a competition between the propulsive forces of the motors and the restoring forces of the crosslinks, and is to be determined selfconsistently at a later stage of the calculation. In a meanfield description (valid for filaments interacting with many neighbors through many motors and crosslinks), this results in the pattern of motion illustrated in Figure 2c.
Focusing on a single rightpointing filament, the combined effect of motor motion and treadmilling implies that motors move with a velocity ${v}_{m}{v}_{t}$ relative to the growing plus end. Denoting by $\xi $ the distance between the motor and the plus end (Figure 2c), this implies that the number of bound motors per unit filament length ${\rho}_{m}(\xi ,t)$ satisfies the reactionconvection equation:
where ${J}_{m}(\xi ,t)={\rho}_{m}({v}_{t}{v}_{m})$ is the motor current in the reference frame of the plus end, and ${\rho}_{m}^{0}/{\tau}_{m}$ represents the attachment rate of unbound motors from the surrounding solution. Newly polymerized filament sections in $\xi =0$ do not yet have any motors bound to them, implying ${\rho}_{m}(0,t)=0$ if ${v}_{t}>{v}_{m}$; likewise ${\rho}_{m}(L,t)=0$ if ${v}_{t}<{v}_{m}$. Motors bound to two filaments of opposing polarities exert forces on each filament, and we denote by ${f}_{m}(\xi ,t)$ the longitudinal force per unit length exerted by the motors on a rightpointing filament. For independent motors operating close to their stall force (i.e., motors whose velocity is essentially controlled by the external crosslink restoring forces), ${f}_{m}(\xi ,t)$ is the ratio between the stall force of a single motor and the spacing between motors along the filament. It is thus proportional to the local motor density through ${f}_{m}(\xi ,t)={f}_{m}^{0}\times [{\rho}_{m}(\xi ,t)/{\rho}_{m}^{0}]$. In the opposite limit where crosslinks are absent, motors slide filaments along at their unloaded velocity without producing any stresses, which rules out both contraction and extension (Lenz et al., 2012a). Note that motors do not induce internal forces in pairs of filaments with identical polarities, which we thus need not consider here.
The density ${\rho}_{c}(\xi ,\tau ,t)$ of crosslinks of age $\tau $ bound in $\xi $ at time $t$ satisfies the conservation equation
with ${\rho}_{c}(0,\tau ,t)={\rho}_{c}(\xi ,\tau \le 0,t)=0$. Since the crosslink attachment points do not slide on the filament, their advection relative to the plus end is entirely due to treadmilling and the crosslink current reads ${J}_{c}(\xi ,t)={\rho}_{c}{v}_{t}$. The term ${\partial}_{\tau}{\rho}_{c}$ in Equation 2 can be viewed as an advection term along the coordinate $\tau $, which accounts for the fact that the age $\tau $ of a bound crosslink increases linearly with time $t$. While attached crosslinks are thus advected towards increasing $\tau $, newly attached crosslinks all have age $\tau =0$ by definition, which we enforce through the delta function in the source term ${\rho}_{c}^{0}\delta (\tau )/{\tau}_{c}$. As motor forces tend to slide filaments of opposing polarities respective to one another, they are opposed by the restoring forces of the crosslinks, which tend to keep filaments stationary with respect to one another. To describe this competition, we assimilate crosslinks to Hookean springs with elastic constant ${k}_{c}$. The average extension of a crosslink bound to two antiparallel filaments is equal to zero at the time of its binding (denoted as $\tau =0$), but increases as $2{v}_{m}\tau $ as the filaments slide respective to one another (Figure 2d). As each crosslink exerts a Hookean force ${k}_{c}\times (2{v}_{m}\tau )$ on the filament, the crosslink force per unit filament length is obtained by summing this force over all filament ages, yielding ${f}_{c}(\xi ,t)={\int}_{0}^{+\mathrm{\infty}}{k}_{c}\times (2{v}_{m}\tau )\times {\rho}_{c}(\xi ,\tau ,t)\mathrm{d}\tau .$
Results
Selforganization and force distribution
Solving Equations 12, we compute the steadystate force densities exerted by the motors and crosslinks on the filament:
Equation 3 describe a depletion of motors and crosslinks close to the filament ends, with associated depletion lengths ${v}_{t}{v}_{m}{\tau}_{m}$ and ${v}_{t}{\tau}_{c}$, as illustrated in Figure 3. Similar nonuniform motor distributions have previously been studied to explain the lengthdependence of microtubule depolymerization rates (Varga et al., 2006; Reese et al., 2011). The crosslink depletion results from the finite time required to decorate newly polymerized filament sections with crosslinks, while the motor depletion arises from the time required to dress a newly created filament overlap with motors. Provided the filament length is much larger than these depletion lengths, the motor force and crosslink friction asymptotically go to the constant values ${f}_{m}^{0}$ and $2{k}_{c}{\rho}_{c}^{0}{\tau}_{c}$ far from the filament ends as the motor and crosslink densities go to their equilibrium values.
Velocity selection and bundle tension
To understand how the motor velocity ${v}_{m}$ is selected, we define ${v}_{m}^{0}={f}_{m}^{0}/(2{k}_{c}{\rho}_{c}^{0}{\tau}_{c})$ as the speed at which the asymptotic forces ${f}_{m}^{0}$ and $2{k}_{c}{\rho}_{c}^{0}{\tau}_{c}{v}_{m}$ balance each other. The velocity ${v}_{m}^{0}$ thus characterizes the hypothetical motion of infinitelength filaments, where the effects of depletion are negligible. By contrast, shorter filaments undergo both a smaller overall driving force and a smaller friction. Depletion thus affects the velocity ${v}_{m}$, while ${v}_{m}$ itself affects motor depletion as described by Equation 3a. Here we analyze this mutual dependence for finitelength filaments.
Rescaling all lengths by ${v}_{m}^{0}{\tau}_{m}$ and times by ${\tau}_{m}$, we henceforth denote dimensionless variants of previously introduced variables with a tilde and determine ${\stackrel{~}{v}}_{m}$ by demanding that the total force $F={\int}_{0}^{L}\left[{f}_{m}(\xi )+{f}_{c}(\xi )\right]\text{d}\xi $ exerted on a single filament vanishes. Defining $u=({\stackrel{~}{v}}_{t}{\stackrel{~}{v}}_{m})/\stackrel{~}{L}$, we insert Equation 3 into this condition and obtain a transcendental equation for $u$:
where $a={\stackrel{~}{v}}_{t}[1g({\stackrel{~}{v}}_{t}{\stackrel{~}{\tau}}_{c}/\stackrel{~}{L})]$ and $b=\stackrel{~}{L}[1g({\stackrel{~}{v}}_{t}{\stackrel{~}{\tau}}_{c}/\stackrel{~}{L})]$ are two constants and $g(y)=2y(1+2y){e}^{1/y}$ (see Figure 4a). As $a>0$ and $b>0$, Equation 4 gives rise to three regimes illustrated in Figure 4b–c: one where translocation by the motors is faster than treadmilling ($u<0\iff {v}_{m}>{v}_{t}$), one where treadmilling is faster than translocation ($u>0$) and one where one $u<0$ solution coexists with two $u>0$ solutions. We determine the stability of these solutions by perturbing ${\stackrel{~}{v}}_{m}$ by a small quantity $\delta {\stackrel{~}{v}}_{m}$ and assessing whether the overall force $F$ exerted on the filament tends to amplify or suppress this perturbation. We find that all unique solutions are stable (i.e., $\partial F/\partial {\stackrel{~}{v}}_{m}<0$). In the threesolutions regime, the smaller of the two $u>0$ solutions is unstable. The bundle thus chooses one of the other two, resulting in two coexisting stable solutions of opposing signs as illustrated in Figure 4d. As for any firstorder (discontinuous) transition, bundles in this parameter regime will select either value of $u$ depending on their initial condition, and any switching from one to the other involves hysteresis.
We now turn to the contractile/extensile character of a bundle comprised of ${\rho}_{f}$ filaments per unit length. A filament in this bundle is subjected to a total force per unit length $f(\xi )=z[{f}_{m}(\xi )+{f}_{c}(\xi )]$ at location $\xi $, where $z$ denotes the number of interacting antiparallel neighbors of a filament. As the filament tension $T(\xi )$ vanishes at the filament ends [$T(0)=T(L)=0$], its tension in $\xi $ thus reads $T(\xi )={\int}_{0}^{\xi}f({\xi}^{\prime})\text{d}{\xi}^{\prime}$. The contractile or extensile character of our bundle is revealed by its integrated tension across any $x=\text{constant}$ plane. In thick bundles, this plane is intersected by a large number of filaments (namely ${\rho}_{f}L\gg 1$) each intersecting the plane at a random coordinate $\xi $ that is uniformly distributed between $0$ and $L$. As a result, the bundle tension is given by the average $\mathcal{T}={\rho}_{f}{\int}_{0}^{L}T(\xi )\text{d}\xi $. Defining $\stackrel{~}{\mathcal{T}}=\mathcal{T}/(z{\rho}_{f}{L}^{2}{f}_{m}^{0})={\stackrel{~}{\mathcal{T}}}_{m}+{\stackrel{~}{\mathcal{T}}}_{c}$, the respective contributions of the motors and crosslinks to the dimensionless bundle tension are
where the function $h(y)=[4y12{y}^{2}+(2+8y+12{y}^{2}){e}^{1/y}]/[(12y)+(1+2y){e}^{1/y}]$ is illustrated in Figure 4a. As shown in Figure 5a, these expressions can result in either sign for $\mathcal{T}$ depending on the values of $u$ and $h({\stackrel{~}{v}}_{t}{\stackrel{~}{\tau}}_{c}/\stackrel{~}{L})$. As the periodic boundary conditions used here confine the bundle to a fixed length, a bundle with a propensity to extend develops a negative tension $\mathcal{T}<0$ (i.e., is compressed), while $\mathcal{T}>0$ denotes a contractile (tense) bundle. These two behaviors respectively correspond to the situations illustrated in Figure 3b–c and Figure 3d.
Qualitative predictions
To analyze the different regimes accessible to our bundle, we illustrate them in Figure 5b as a function of the original dimensionless parameters ${\stackrel{~}{v}}_{t}$, ${\stackrel{~}{\tau}}_{c}$ and $\stackrel{~}{L}$. As some parameter values yield coexisting metastable solutions for $u$, so can they allow for both contractile and extensile steady states. However, despite this ambiguity at intermediate parameter values, Figure 5b shows that the selforganization mechanism investigated here results in unambiguous extension for broad ranges of parameters.
To understand this reversal of contractility qualitatively, we first consider a contractile situation at low crosslink detachment time ${\tau}_{c}$. From there, increasing ${\tau}_{c}$ results in an enlarged crosslink depletion zone in the vicinity of the filament plus ends (Figure 3), and thus in a relative localization of the crosslinks towards the filament minus ends. This ‘antisarcomere’ organization results in an extensile bundle (Figure 1c, right) at large ${\tau}_{c}$, in contrast with the contractile ‘sarcomere’ structures (Figure 1c, left) found in our highly organized striated muscle.
The extension mechanism reported here differs from behaviors previously modeled in the theoretical literature, both in its cause and in its applicability. Extension has indeed been observed in numerical simulations reported in Belmonte et al. (2017), but only in situations where the motors were assumed to dwell at the filament ends. The discussion section of this reference does however qualitatively predict that extension should result from an ‘antenna’ mechanism similar to the depletion quantitatively modeled here. Another instance of bundle extension is reported in Kruse and Sekimoto (2002). In that model, the extensile behavior stems from the interaction between parallel, not antiparallel filaments, which is at odds with experimental observations that bundlewide force generation in actomyosin requires antiparallel filaments (Reymann et al., 2012). Finally, in twodimensional numerical simulations extensile antiparallel microtubule pairs were reported due to the fact that they typically bind halfway between the two configurations of Figure 1b, implying that their subsequent dynamics predominantly involves extensile configurations (Gao et al., 2015). The freedom required for this midfilament binding is however likely not afforded to tightly bundled filaments, although the nonuniform motor distributions discussed here could also have contributed to the extensile behavior observed in Gao et al. (2015).
Beyond the transition from contraction to extension, Figure 5b shows that a second transition can be triggered by a further increase of ${\tau}_{c}$ in the extensile phase, which causes the variable $u\propto {v}_{t}{v}_{m}$ to change sign through a firstorder (for small $\stackrel{~}{L}$) or a secondorder (for large $\stackrel{~}{L}$) transition (Figure 5b). Indeed, the enhanced crosslink depletion associated with a large ${\tau}_{c}$ tends to reduce the friction between filaments, resulting in faster motor motion and thus in a situation where motor sliding outpaces treadmilling ($u<0$).
These two transitions could be observed in vitro, and possibly even in vivo. Indeed, the contractile vs. extensile character of actomyosin bundles is apparent from the direct imaging of reconstituted assays (Thoresen et al., 2011; Reymann et al., 2012) as well as cells (Mendes Pinto et al., 2012). The second, velocityreversal transition, on the other hand, can be monitored in experiments where single filaments are resolved (Murrell and Gardel, 2012). As motor velocity outpaces treadmilling, such filaments will switch from an apparently plusenddirected motion (illustrated in Figure 2a) to a motorinduced, minusenddirected motion. In practice, these transition could be induced in a number of ways, including changes in the monomeric actin concentration or the action of formin (affecting $L$ and ${v}_{t}$), the presence of different types of crosslinks (affecting ${v}_{m}^{0}$ and ${\tau}_{c}$), or modifications of the number or type of motor heads in a thick filament (affecting ${v}_{m}^{0}$ and ${\tau}_{m}$) (Thoresen et al., 2013). Such changes could also be at work in smooth muscle, where the number of myosins in individual thick filaments is regulated dynamically (Seow, 2005). The experimental relevance of these transitions is illustrated by a dashed line in Figure 5b, which shows that both transitions can be probed by varying $L$ between 250 nm and 1 µm while holding ${v}_{m}^{0}=50\text{nm}\cdot {\text{s}}^{1}$, ${\tau}_{m}=5\text{s}$ (Erdmann et al., 2013), ${v}_{t}=100\text{nm}\cdot {\text{s}}^{1}$ (Howard, 2001), and ${\tau}_{c}=1\text{s}$ (Miyata, 1996) fixed.
Quantitative aspects and alternative models
The magnitude of the forces and velocities predicted by our model are on par with those found in vivo, for example in the cytokinetic ring of fission yeast. Indeed, setting $L=1.4\mu \text{m}$, ${\rho}_{f}L=20$, ${f}_{m}^{0}\simeq 7.2\times {10}^{6}\text{N}\cdot {\text{m}}^{1}$ (Wu and Pollard, 2005), ${k}_{c}\simeq 3\times {10}^{4}\text{N}\cdot {\text{m}}^{1}$ (Rief et al., 1999), $z=3$ as in a hexagonal packing of alternating left and rightpointing filaments and $\stackrel{~}{\mathcal{T}}\simeq 0.2$, we find a contractile force $\mathcal{T}\simeq 120\text{pN}$ comparable with the ring tension of 390 pN measured in fission yeast protoplasts (Stachowiak et al., 2014) We also find a characteristic velocity ${v}_{m}^{0}\simeq 5\text{nm}\cdot {\text{s}}^{1}$ similar to that of ring contraction ($\simeq 3\text{}4\text{nm}\cdot {\text{s}}^{1}$). These order of magnitudes retrospectively justify our choice to neglect viscous drag forces in our system, which are of order of $\eta L{v}_{m}^{0}\simeq 7\times {10}^{18}\text{N}$, where $\eta \simeq {10}^{3}\text{Pa}\cdot {\text{s}}^{1}$ is the viscosity of water. Note also that other mechanisms for contractility based on sarcomerelike crosslinking of the filament barbed ends have also been proposed in the specific case of fission yeast (Thiyagarajan et al., 2017).
In this section, we show that in addition to being on par with experimentally observed forces, the tensions generated by our selforganization mechanism exceed those resulting from mechanical nonlinearities (e.g., buckling) or motor dwelling over broad ranges of parameters, suggesting that the selforganization mechanism could be a substantial contributor to bundle tension in vitro and in vivo.
Selforganization vs. bucklinginduced tensions
Bucklinginduced contractility is relevant for bundles comprising flexible (typically actin) filaments. To estimate the associated tension, we use the model of Lenz et al. (2012b), where filaments may locally buckle over a small section bounded by a crosslink and a motor (Figure 6a). Filaments retain their overall linear shape outside of the buckled regions, and it is thus reasonable to assume that selforganization and buckling can operate simultaneously and that they contribute additively to the bundle tension. Here we compare the tension ${\mathcal{T}}_{\text{buckling}}$ induced by the latter mechanism to the tension $\mathcal{T}$ given by Equation 5 as a function of two experimentally adjustable parameters, namely the number $n$ of myosin heads per myosin minifilament (Thoresen et al., 2013), and the average spacing ${\mathrm{\ell}}_{0}$ between two consecutive motors or crosslinks (Lenz et al., 2012b).
To compute the tensions of interest, we extrapolate their value to a bundle with one filament per crosssection, keeping in mind that both $\mathcal{T}$ and ${\mathcal{T}}_{\text{buckling}}$ scale linearly with that number in thicker bundles. To obtain the value of ${\mathcal{T}}_{\text{buckling}}$, we first note that bucklinginduced force generation in an actomyosin bundle requires that the typical compressive forces on a filament exceed the buckling threshold of that filament. As shown in Lenz et al. (2012b), this requires that the typical spacing ${\mathrm{\ell}}_{0}$ between two motors/crosslinks lies between the following two bounds:
where ${k}_{B}T$ is the thermal energy, ${\mathrm{\ell}}_{p}$ the persistence length of a filament, $L$ its actual length, ${F}_{s}$ is the stall force of a single motor, $v$ its unloaded velocity and ${\tau}_{m}$ its detachment time. Qualitatively, the condition ${\mathrm{\ell}}_{0}>{\mathrm{\ell}}_{0}^{}$ accounts for the requirement that the filaments be long enough (and thus floppy enough) for the motor forces to be able to buckle them, while the upper bound ${\mathrm{\ell}}_{0}<{\mathrm{\ell}}_{0}^{+}$ expresses the fact that building up a force sufficient to buckle a filament in bundles where motors and crosslinks are far apart takes so long that spontaneous motor detachment will hinder it. Following Rosenfeld et al. (2003) and Lenz et al., 2012b, we use ${\tau}_{m}={0.96}^{n}\times 3\text{ms}$, as well as ${F}_{s}=n\times 0.1\text{pN}$, where the low value of the stall force per myosin head accounts for their intermittent attachment to the filaments. We also set ${\mathrm{\ell}}_{p}=10\mu \text{m}$, $L=5\mu \text{m}$ and $v=200\text{nm}\cdot {\text{s}}^{1}$ as in Lenz et al. (2012b). In situations where the condition ${\mathrm{\ell}}_{0}^{}<{\mathrm{\ell}}_{0}<{\mathrm{\ell}}_{0}^{+}$ is satisfied, we set the contractile force of a buckled bundled to ${F}_{s}$, consistent with the idea that buckling actin essentially mechanically removes the compressed sections of the filament, allowing individual motors to act as if in a sarcomeric configuration.
Complementing these assumptions with the values of ${v}_{t}$, ${v}_{m}^{0}$, ${\tau}_{c}$ and $z$ used in Sec. ‘Qualitative predictions’, we plot the tension per filament induced by either mechanism in Figure 6b–c, and directly compare them in Figure 6d. Selforganization trivially dominates outside of the region where buckling is allowed, which represents a substantial fraction of the reasonably accessible parameter regimes. Moreover, even in the region where buckling is allowed selforganization dominates for small values of ${\mathrm{\ell}}_{0}$ as long as $n$ does not become very large. These results support the notion that positional selforganization of motors and crosslinks constitutes a viable mechanism for actomyosin force generation despite the possibility of bucklinginduced contraction.
Selforganization vs. dwellinginduced tensions
Motor dwelling at the filament ends takes place in certain types of microtubuleassociated motors (Roostalu et al., 2018; Tan et al., 2018), and could occur in some actinmyosin systems (Wollrab et al., 2018). To estimate the associated tension, we consider a variant of our selforganization model where motors reaching the plus end of a filament dwell there for an average time ${\tau}_{m}$. The motor attachmentdetachment dynamics is infinitely fast everywhere else, and so is that of the crosslinks. This rules out the depletion effects discussed in Sec. ‘Selforganization and force distribution’.
These assumptions imply uniform bulk densities of nondwelling motors ${\rho}_{m}(\xi \ne 0,t)={\rho}_{m}^{0}$ and crosslinks ${\rho}_{c}(\xi ,t)={\rho}_{c}^{0}$. In the case ${v}_{m}>{v}_{t}$, the flow of motors into the filament plus end is equal to ${J}_{m}(\xi =0)={\rho}_{m}({v}_{m}{v}_{t})$, implying that at steady state an average number ${\rho}_{m}({v}_{m}{v}_{t}){\tau}_{m}$ of motors dwells there. Assuming as before that motors operate at their stall force, this implies
where the first term of the righthand side of Equation 7a accounts for the forces exerted on the filament by nondwelling motors, while the second term incorporates the effects both of motors dwelling on the filament of interest (through the delta function) and on other filaments. Our description does not include doublydwelling motors, as the pattern of filament motion illustrated in Figure 2c implies that such filaments are immediately ripped from either one of the filaments they are attached to. In the case ${v}_{m}<{v}_{t}$, the motion of the motors is not fast enough to allow them to reach the plus end. Instead, the depolymerizing minus end catches up to them. This configuration has not been observed or proposed to lead to dwelling to my knowledge, and in the absence of dwelling no tension is generated.
To characterize the resulting bundle steadystates, we apply the velocity selection and stability criteria described in Sec. ‘Velocity selection and bundle tension’ and compute the resulting bundle tension. The case ${\stackrel{~}{v}}_{t}\ge 1$ implies that fastdepolymerizing minus ends catch up on the motors as discussed above, ruling out out dwelling and implying ${\stackrel{~}{v}}_{m}=1$ and ${\stackrel{~}{\mathcal{T}}}_{\text{dwell}}=0$. On the other hand, if ${\stackrel{~}{v}}_{t}<1$ motors are faster than treadmilling and localize at the plus ends, inducing extension. In that case, the bundle finds a steady state provided that $\stackrel{~}{L}>2$, with ${\stackrel{~}{v}}_{m}=(\stackrel{~}{L}2{\stackrel{~}{v}}_{t})/(\stackrel{~}{L}2)$ and ${\stackrel{~}{\mathcal{T}}}_{\text{dwell}}=(1{\stackrel{~}{v}}_{t})/2(\stackrel{~}{L}2)$. Finally, if ${\stackrel{~}{v}}_{t}<1$ and $\stackrel{~}{L}<2$, no steady state exists in the system. To understand this, consider a situation where motors start accumulating at the filament plus end, increasing the propulsive force on the filament and thus increasing ${v}_{m}$, leading to a further increase in the number of accumulated motors. In the model, this positive feedback results in an infinite increase in velocity ${v}_{m}$ unless crosslinkinduced friction stops it. If $\stackrel{~}{L}<2$ however, filaments are very short, implying a small number of crosslinks and a comparatively small effective friction, hence the absence of a steadystate. In practice, such a situation is stabilized by effects ignored here, including the onset of motor depletion on the filament or a departure of the motors from their stall conditions.
As the dwelling and selforganization mechanisms both rely on a localization of the motors induced by the filament/motor dynamics, they result in quantitatively similar tensions as long as crosslink localization remains limited, that is for small values of ${\stackrel{~}{\tau}}_{c}$ (Figure 7). However, selforganization dominates over dwelling in situations where extended depletion profiles are allowed to develop, that is when crosslinks are longlived or filaments are long (large ${\stackrel{~}{\tau}}_{c}$ or large $\stackrel{~}{L}$), as well as when fast treadmilling prevents motor dwelling (${\stackrel{~}{v}}_{t}>1$). We thus expect that the selforganization mechanism will be a substantial contributor to force generation in microtubulemotor systems not only in the obvious cases where motors do not dwell on the filament, but also in many situations where they do.
Discussion
In contrast with the static organization of striated muscle, many nonmuscle actomyosin structures as well as motormicrotubules assemblies are very dynamic, and their components continuously assemble and disassemble even as they exert forces on their surroundings. While numerical simulations are useful to investigate these systems (Oelz et al., 2015; Kim, 2015; Ennomani et al., 2016), the many parameters involved have until now hampered systematic explorations of all possible dynamical regimes. By contrast, our analytical approach allows us to derive a complete phase diagram for selforganized motorfilament bundles. Lenz et al. (2012b), Lenz (2014) and Belmonte et al. (2017) derive similarly useful analytical results applicable to situations dominated by filament nonlinearity or motor dwelling. We thus uncover two previously unreported, experimentally observable transitions between bundle contraction and extension, and between plusenddirected and minusenddirection apparent filament motion. These transitions could serve as experimental signatures of selforganizationdriven cytoskeletal force generation.
While some of our simplifying assumptions may affect the accuracy of our quantitative predictions, the simplicity of the underlying mechanisms make our qualitative statements very robust. Indeed, we predict extension whenever the filament plus end polymerizes quickly enough to induce significant crosslink depletion in its vicinity, resulting in a sarcomerelike organization. This depletion is insensitive to whether filament disassembly occurs through depolymerization at the minus end or cofilininduced severing (Theriot, 1997). It is additionally relevant for dynamical microtubules, whose minus ends are mostly static while the plus end grows slowly before quickly retracting in a socalled ‘catastrophe’. Our model will thus accurately predict the tension resulting from plusend depletion during the growth phase, while fast catastrophes can be seen as more or less instantaneous filament deletion events without significant effect on bundle tension. Depletion is also present whether crosslinks detach at a constant rate as assumed here, or unbind increasingly quickly under increasing force (Miyata, 1996). Turning to motors, we note that although small motor numbers may add significant density and velocity fluctuations to our meanfield model, motors are depleted on average in newer filament overlaps even in the presence of these fluctuations. While we describe this effect as a consequence of delayed motor binding from the surrounding solution, a similarly reduced force could also arise in bundles densely covered by motors, for instance due to a delay in fully aligning the myosin minifilament with the two antiparallel actin filaments to allow all myosin heads to fully participate in filament sliding. Finally, while we assume that motors always exert their stall force and thus acquire a velocity inversely proportional to the effective crosslink friction, introducing a more complicated motor forcevelocity relationship would slightly complicate this dependence quantitatively, but not qualitatively.
Our prediction of a robust extensile regime provides a stringent test to validate or invalidate the selforganized force generation model in specific experiments. As such, it constitutes an important statement even for systems in which extension is not observed, as it implies that the absence of extension in certain parameter regimes argues against selforganization mechanisms in favor of mechanical nonlinearity (in actin) or motor dwelling (in microtubules) models. As an illustration, Thoresen et al. (2011) and Thoresen et al. (2013) report a setup where actomyosin bundles contract in the absence of treadmilling, in contrast with the prediction of Figure 5b that ${v}_{t}=0$ implies extension. This discrepancy tends to disqualify selforganized contraction in this setup, and retrospectively validates the proposal made in Lenz et al. (2012b) that mechanical nonlinearities and specifically filament buckling dominate this assay. Conversely, selforganized force generation is likely to play a role in a number of in vivo actomyosin contractile structures where buckling is not observed (Cramer et al., 1997; Kamasaki et al., 2007) and where actin treadmilling dynamics plays a crucial role (Mendes Pinto et al., 2012). In microtubules, extension is observed consistent with our prediction when the filament polymerization/depolymerization dynamics is blocked in vitro (Sanchez et al., 2012; Keber et al., 2014), while both extension and contraction can arise in more complex in vivo situations (Patel et al., 2005; Foster et al., 2015). Both extension and contraction have also been reported in actin bundles, which cannot buckle due to their large stiffness (Stam et al., 2017).
Beyond steadystate contraction or extension, transitions between these two states could help understand several in vivo behaviors involving alternating contractions and expansions of the actomyosin cortex, including cell area oscillations observed during Drosophila, C. elegans, and Xenopus development (Martin et al., 2009; Solon et al., 2009; RohJohnson et al., 2012; He et al., 2010; Kim and Davidson, 2011; Levayer and Lecuit, 2012) or propagating actomyosin contractility waves (Allard and Mogilner, 2013). We speculate that such oscillations could arise through a Hopf bifurcation involving the rapid switching between a contractile and an extensile metastable state in the multiplesolution regime of Figure 4.
It would be interesting to see how the mechanisms described here apply to two or threedimensional actomyosin assemblies, whose richer geometry allows for additional actomyosin force generation mechanisms (Lenz, 2014). More refined approaches could also include discussions of the onset of positional ordering of the filaments themselves within the bundle (Kruse et al., 2001; Kruse et al., 2003; Friedrich et al., 2012). While such ordering is suppressed by filament diffusion (Zemel and Mogilner, 2009) and is not observed in many disordered actomyosin bundles (Cramer et al., 1997; Kamasaki et al., 2007), its onset during the formation of stress fibers is quite dependent on actin filament dynamics, suggesting a role for the mechanisms considered here (Hu et al., 2017). Finally, the fundamental principles for the dynamical depletion of motors and crosslinks described here could serve as guiding principles in our developing understanding of selforganized contractility in the cytoskeleton (Nakazawa and Sekimoto, 1996; Kruse and Jülicher, 2000; Kruse and Sekimoto, 2002; Oelz et al., 2015; Belmonte et al., 2017).
Appendix 1
Data availability
This study does not involve the generation or analysis of data.
References

Traveling waves in actin dynamics and cell motilityCurrent Opinion in Cell Biology 25:107–115.https://doi.org/10.1016/j.ceb.2012.08.012

A theory that predicts behaviors of disordered cytoskeletal networksMolecular Systems Biology 13:941.https://doi.org/10.15252/msb.20177796

General mechanism of actomyosin contractilityPhysical Review Letters 107:118101.https://doi.org/10.1103/PhysRevLett.107.118101

Stochastic dynamics of small ensembles of nonprocessive molecular motors: the parallel cluster modelThe Journal of Chemical Physics 139:175104.https://doi.org/10.1063/1.4827497

Sarcomeric pattern formation by actin cluster coalescencePLOS Computational Biology 8:e1002544.https://doi.org/10.1371/journal.pcbi.1002544

Multiscale polar theory of microtubule and motorprotein assembliesPhysical Review Letters 114:048101.https://doi.org/10.1103/PhysRevLett.114.048101

Actinbinding proteins in cell motilityInt Rev. Cytology 156:199–273.https://doi.org/10.1016/s00747696(08)62255x

Tissue elongation requires oscillating contractions of a basal actomyosin networkNature Cell Biology 12:1133–1142.https://doi.org/10.1038/ncb2124

BookMechanics of Motor Proteins and the CytoskeletonSunderland, MA: Sinauer Associates.

Longrange selforganization of cytoskeletal myosin II filament stacksNature Cell Biology 19:133–141.https://doi.org/10.1038/ncb3466

Threedimensional arrangement of Factin in the contractile ring of fission yeastThe Journal of Cell Biology 178:765–771.https://doi.org/10.1083/jcb.200612018

Determinants of contractile forces generated in disorganized actomyosin bundlesBiomechanics and Modeling in Mechanobiology 14:345–355.https://doi.org/10.1007/s1023701406082

Architecture shapes contractility in actomyosin networksCurrent Opinion in Cell Biology 50:79–85.https://doi.org/10.1016/j.ceb.2018.01.015

Selfpropagating patterns in active filament bundlesPhysical Review Letters 87:138101.https://doi.org/10.1103/PhysRevLett.87.138101

Continuum theory of contractile fibresEurophysics Letters 64:716–722.https://doi.org/10.1209/epl/i2003006182

Actively contracting bundles of polar filamentsPhysical Review Letters 85:1778–1781.https://doi.org/10.1103/PhysRevLett.85.1778

Growth of fingerlike protrusions driven by molecular motorsPhysical Review E 66:031904.https://doi.org/10.1103/PhysRevE.66.031904

Requirements for contractility in disordered cytoskeletal bundlesNew Journal of Physics 14:033037.https://doi.org/10.1088/13672630/14/3/033037

Contractile units in disordered actomyosin bundles arise from Factin bucklingPhysical Review Letters 108:238107.https://doi.org/10.1103/PhysRevLett.108.238107

Biomechanical regulation of contractility: spatial control and dynamicsTrends in Cell Biology 22:61–81.https://doi.org/10.1016/j.tcb.2011.10.001

Bridging the microscopic and the hydrodynamic in active filament solutionsEurophysics Letters 69:846–852.https://doi.org/10.1209/epl/i2004104140

Strength and lifetime of the bond between actin and skeletal muscle αactinin studied with an optical trapping techniqueBiochimica Et Biophysica Acta (BBA)  General Subjects 1290:83–88.https://doi.org/10.1016/03044165(96)000037

Forcing cells into shape: the mechanics of actomyosin contractilityNature Reviews Molecular Cell Biology 16:486–498.https://doi.org/10.1038/nrm4012

Polarity sorting in a bundle of actin filaments by TwoHeaded myosinsJournal of the Physical Society of Japan 65:2404–2407.https://doi.org/10.1143/JPSJ.65.2404

Crowding of molecular motors determines microtubule depolymerizationBiophysical Journal 101:2190–2200.https://doi.org/10.1016/j.bpj.2011.09.009

Single molecule force spectroscopy of spectrin repeats: low unfolding forces in Helix bundlesJournal of Molecular Biology 286:553–561.https://doi.org/10.1006/jmbi.1998.2466

Myosin IIb is unconventionally conventionalJournal of Biological Chemistry 278:27449–27455.https://doi.org/10.1074/jbc.M302555200

Myosin filament assembly in an everchanging myofilament lattice of smooth muscleAmerican Journal of PhysiologyCell Physiology 289:C1363–C1368.https://doi.org/10.1152/ajpcell.00329.2005

Mechanism of cytokinetic contractile ring constriction in fission yeastDevelopmental Cell 29:547–561.https://doi.org/10.1016/j.devcel.2014.04.021

Accelerating on a treadmill: adf/cofilin promotes rapid actin filament turnover in the dynamic cytoskeletonThe Journal of Cell Biology 136:1165–1168.https://doi.org/10.1083/jcb.136.6.1165

A node organization in the actomyosin contractile ring generates tension and aids stabilityMolecular Biology of the Cell 28:3286–3297.https://doi.org/10.1091/mbc.e17060386

Reconstitution of contractile actomyosin bundlesBiophysical Journal 100:2698–2705.https://doi.org/10.1016/j.bpj.2011.04.031

Thick filament regulation of contractility in selforganized actomyosin bundlesBiophysical Journal 104:655–665.https://doi.org/10.1016/j.bpj.2012.12.042

Yeast kinesin8 depolymerizes microtubules in a lengthdependent mannerNature Cell Biology 8:957–962.https://doi.org/10.1038/ncb1462

Polarity sorting drives remodeling of actinmyosin networksJournal of Cell Science 132:jcs219717.https://doi.org/10.1242/jcs.219717

Motorinduced sliding of microtubule and actin bundlesPhysical Chemistry Chemical Physics 11:4821.https://doi.org/10.1039/b818482h
Decision letter

Raymond E GoldsteinReviewing Editor; University of Cambridge, United Kingdom

Anna AkhmanovaSenior Editor; Utrecht University, Netherlands

Francois NedelecReviewer
In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.
Acceptance summary:
This is a very nice theoretical paper on the collective dynamics of cytoskeletal bundles. The author critically examines several competing models for the competition between contraction and expansion of such bundles, and develops a systematic and very clear analysis that focuses on just one of the important issues, putting aside enddwelling and filament buckling. He finds the possibility of reversal of contractility and argues that this supports the hypothesis of collective, selforganized behaviour.
Decision letter after peer review:
Thank you for submitting your article "Reversal of contractility as a signature of selforganization in cytoskeletal bundles" for consideration by eLife. Your article has been reviewed by two peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by Anna Akhmanova as the Senior Editor. The following individual involved in review of your submission has agreed to reveal their identity: Francois Nedelec (Reviewer #2).
The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.
Summary:
The author examines the dynamics of a circular bundle of filaments predicting regimes in which the bundle can contract or expand, in the presence of passive crosslinkers and crosslinking motors. The theory is done in 1D, following a meanfield approach. While all the assumptions are fairly standard in the field, their combination is interesting. Particularly filament treadmilling (assembly at the barbed end that is exactly compensated by disassembly at the pointed end) effectively provides a 'motor' quality to the crosslinkers, as they appear to move towards the pointed end, while the true motors move in the opposite direction. As a consequence, the system can be tuned to be contractile or extensile by changing the treadmilling velocity, which affects the speed of motors and crosslinkers move toward the ends. Treadmilling also leads to filament movements, but within the assumed of periodic boundaries, this has little impact on the overall length of the bundle. The buckling of filaments is not considered. The results are not entirely surprising, but the possibility for a network to expand is insufficiently discussed in the literature, although it is certainly an important behavior of cytoskeletal systems. The theory itself is clearly explained and convincing. It is of high quality and in line with that the author and others have done before on contraction, but now including extension, and this is new.
Minor points:
"Our prediction crucially rests on a simultaneous theoretical treatment of the filament, motor and crosslink dynamics detailed in the subsection “Motor and crosslink dynamics”. Previous studies of contractility mechanisms only involved partial treatments, whereby the time scale associated with one of these dynamics was effectively assumed to be much larger or smaller than the others."
We would agree that for most previous work this is true, but in Belmonte, Leptin and Nédélec, 2017, crosslinkers and motors were treated symmetrically, without making an assumption relating to different time scales. Please, note the supplementary material of this work, is not a simulation, but an analytical result.
Figure 2 legend: We were confused at first that ξ and thus ρ_{m}(ξ) would be measured with L=0 at the barbed end, and not the other way around, which seems more natural (0 = minusend).
"In the opposite limit where crosslinks are absent, motors slide filaments along at their unloaded velocity without producing any stresses, which rules out both contraction and extension [Lenz et al., 2012a]"
This is only true if one neglects the viscous drag, and this assumption should be repeated here ('without any stress' is an exaggeration).
"Indeed, while extension has been observed in numerical simulations reported in Belmonte, Leptin and Nédélec, 2017, this behavior was directly tied to a choice of somewhat exotic motor enddwelling properties that favor antisarcomeric structures."
Belmonte, Leptin and Nédélec, 2017, discussed the consequence of a 'depletion zone' (called with a different name: antenna effect) on which is a cornerstone of the current work, and have predicted that they would lead to the extension of the network. We fail to see how this can be reduced to 'antisarcomeric structures' unless this would apply similarly to the current work.
"This system additionally does not manifest a transition from contraction to extension as the parameters of the model are continuously varied.". Please, refer to Figure 3D [Belmonte, Leptin and Nédélec, 2017] and correct (or delete) the statement.
"Another instance of bundle extension is reported in Kruse and Sekimoto, 2002. In that model, the extensile behavior stems from the interaction between parallel, not antiparallel filaments, which is at odds with experimental observations that tension generation in actomyosin requires antiparallel filaments [Reymann et al., 2012]." This statement appears inconsistent because it compares extension to contraction, and it seems possible to generate opposite forces, on different configurations (parallel vs. antiparallel filaments). At least that would be the expectation of many readers, and the statement, therefore, does not work. Perhaps the author could find something more positive to say about [Kruse and Sekimoto, 2002] which lead the way to the submitted work. We see it as a clear demonstration that the depletion zone can lead to movements between two parallel filaments, and thus break the symmetry that impairs extension/contraction in general.
Subsection “Quantitative Aspects and Alternative Models” seems flawed, as we expect the balance of forces during cytokinesis for echinoderms and S. pombe to be very different, one having a cell wall that dominates the mechanical behavior. Thus it is invalid to extrapolate from one condition to the other.
A number of formulations: "has been suggested to occur in some actinmyosin systems [Wollrab et al., 2018]", "somewhat exotic", "Finally, the motor properties leading to extension in this model may be relevant for microtubule systems, but are not for actin" suggest that the author does not believe that enddwelling motors could exist for actin.
These statements are surprising especially in a theory paper. Can we really consider our knowledge of the cytoskeleton complete enough today to exclude a possibility that is already accepted for microtubules?
Finally, this comment is also relevant to the Discussion, specifically the fact that the absence of certain features can be taken as a proof of some other mechanism: I would argue that we do not know all the mechanisms operating in a cytoskeletal system, and with this in mind, can the author be confident enough to use the absence of a predicted behavior to make a strong conclusion about some other mechanism?
https://doi.org/10.7554/eLife.51751.sa1Author response
Minor points:
"Our prediction crucially rests on a simultaneous theoretical treatment of the filament, motor and crosslink dynamics detailed in subsection “Motor and crosslink dynamics”. Previous studies of contractility mechanisms only involved partial treatments, whereby the time scale associated with one of these dynamics was effectively assumed to be much larger or smaller than the others."
We would agree that for most previous work this is true, but in Belmonte, Leptin and Nédélec, 2017, crosslinkers and motors were treated symmetrically, without making an assumption relating to different time scales. Please, note the supplementary material of this work, is not a simulation, but an analytical result.
I thank the reviewers for critiquing this sentence. What I meant was that this work was to my knowledge the first to consider motor, crosslink as well filament treadmilling dynamics. Belmonte, Leptin and Nédélec, 2017 only describes the first two, and thus effectively assumes that treadmilling is much slower (it does consider filament turnover in one instance, but through a schematic filament removal/addition simulations protocol that is quite different from a detailed polymerization mechanism; additionally this protocol is not included by its analytical approach). I have clarified that the reference to filaments in this sentence regards their treadmilling/polymeration through the following modification of the sentence:
“Our prediction crucially rests on a simultaneous theoretical treatment of the filament polymerization, motor and crosslink dynamics detailed in subsection “Motor and crosslink dynamics”. […] We show in subsection “Selforganization and force distribution” that the coupled dynamics of these three elements induces a spatial organization of motors and crosslinks along the filaments”.
On the issue of numerical vs. analytical results, I have made the following addition to the conclusion highlighting the analytical contents of Belmonte, Leptin and Nédélec, 2017:
“[…] our analytical approach allows us to derive a complete phase diagram for selforganized motorfilament bundles. References Lenz et al., 2012b; Lenz,
2014; Belmonte, Leptin and Nédélec, 2017 derive similarly useful analytical results applicable to situations dominated by filament nonlinearity or motor dwelling.”
Figure 2 legend: We were confused at first that ξ and thus ρ_{m}(ξ) would be measured with L=0 at the barbed end, and not the other way around, which seems more natural (0 = minusend).
I added a comment stressing the convention used to minimize the risk of a reader being confused by this:
“Note that the coordinate ξ is measured from the filament’s plus end” (legend of Figure 2).
"In the opposite limit where crosslinks are absent, motors slide filaments along at their unloaded velocity without producing any stresses, which rules out both contraction and extension [Lenz et al., 2012a]"
This is only true if one neglects the viscous drag, and this assumption should be repeated here ('without any stress' is an exaggeration).
This is correct. One has to keep in mind that viscous drag is negligible compared to the other forces discussed here, as currently discussed in the Discussion section. Following the reviewers’ recommendation I now explicitly state this assumption earlier in the section:
“Finally we neglect viscous drag, as further argued in the Discussion”.
"Indeed, while extension has been observed in numerical simulations reported in Belmonte, Leptin and Nédélec, 2017, this behavior was directly tied to a choice of somewhat exotic motor enddwelling properties that favor antisarcomeric structures."
Belmonte, Leptin and Nédélec, 2017 discussed the consequence of a 'depletion zone' (called with a different name: antenna effect) on which is a cornerstone of the current work, and have predicted that they would lead to the extension of the network. We fail to see how this can be reduced to 'antisarcomeric structures' unless this would apply similarly to the current work.
I take the reviewers’ overall point that my previous discussion of Belmonte on was needlessly critical, and have modified the language to offer a more nuanced discussion.
On the specific issue of the antenna effect, I now specifically mention its presence in the work of Belmonte, keeping in mind that it is only qualitatively discussed in the Discussion section rather than quantitatively modeled as I do here. I removed the “antisarcomeric” comment:
“Extension has indeed been observed in numerical simulations reported in Belmonte, Leptin and Nédélec, 2017, but only in situations where the motors were assumed to dwell at the filament ends. The Discussion section of this reference does however qualitatively predict that extension should result from an “antenna” mechanism similar to the depletion quantitatively modeled here.”
"This system additionally does not manifest a transition from contraction to extension as the parameters of the model are continuously varied.". Please, refer to Figure 3D [Belmonte, Leptin and Nédélec, 2017] and correct (or delete) the statement.
The original idea behind this comment was to contrast a situation where the transition is induced by a change in the composition of the system vs. its intrinsic parameters. This is however a very inessential point, and I deleted it according to the reviewer’s suggestion.
"Another instance of bundle extension is reported in Kruse and Sekimoto, 2002. In that model, the extensile behavior stems from the interaction between parallel, not antiparallel filaments, which is at odds with experimental observations that tension generation in actomyosin requires antiparallel filaments [Reymann et al., 2012]." This statement appears inconsistent because it compares extension to contraction, and it seems possible to generate opposite forces, on different configurations (parallel vs. antiparallel filaments). At least that would be the expectation of many readers, and the statement, therefore, does not work. Perhaps the author could find something more positive to say about Kruse and Sekimoto, 2002, which lead the way to the submitted work. We see it as a clear demonstration that the depletion zone can lead to movements between two parallel filaments, and thus break the symmetry that impairs extension/contraction in general.
The reviewers are correct in pointing out that my statement is confusing due to its clumsy use of the word “tension”, which may be taken as a synonym for “contraction”. I modified it slightly to more accurately portray the underlying reasoning:
“[…] which is at odds with experimental observations that bundlewide force generation in actomyosin requires antiparallel filaments [Reymann et al., 2012].”
I do not agree with the characterization of Kruse and Sekimoto, 2002, as a proof that depletion zones (in the sense of the present manuscript, which is close to the “antenna effect” of Belmonte, Leptin and Nédélec, 2017) can lead to movements between parallel filaments. Indeed Kruse and Sekimoto, 2002, is not concerned with the depletion/antenna effect, but rather with motor traffic jams, which essentially give rise to an emergent motor enddwelling.
Subsection “Quantitative Aspects and Alternative Models” seems flawed, as we expect the balance of forces during cytokinesis for echinoderms and S. pombe to be very different, one having a cell wall that dominates the mechanical behavior. Thus it is invalid to extrapolate from one condition to the other.
I agree with the reviewers that the extrapolation discussed in the first paragraph of the subsection “Quantitative aspects and alternative models” (and only there) is a bit of a stretch. I have thus replaced the numerical force extrapolated from echinoderms by a direct fission yeast measurement, which gives a result consistent with the one produced by the model using fission yeast parameters:
“The magnitude of the forces and velocities predicted by our model are on par with those found in vivo, e.g., in the cytokinetic ring of fission yeast. Indeed, setting L = 1.4µm, […] we find a contractile force T ' 120pN comparable with the ring tension of 390pN measured in fission yeast protoplasts (Stachowiak et al., 2014).”
A number of formulations: "has been suggested to occur in some actinmyosin systems [Wollrab et al., 2018]", "somewhat exotic", "Finally, the motor properties leading to extension in this model may be relevant for microtubule systems, but are not for actin" suggest that the author does not believe that enddwelling motors could exist for actin.
These statements are surprising especially in a theory paper. Can we really consider our knowledge of the cytoskeleton complete enough today to exclude a possibility that is already accepted for microtubules?
My position is not to exclude this possibility, but to point out that end dwelling is certainly more speculative in actomyosin than in microtubules, and should thus not be our default assumption. It seems to me that this position is indistinguishable from that expressed in the reviewers’ own work [Belmonte, Leptin and Nédélec, 2017], where the Figure 2 presents nondwelling motors as “actinlike” whereas Figure 3 presents a mixture of dwelling and nondwelling as “microtubulelike”. To better reflect this in the text I have modified the formulations that the reviewers objected to.
In the order of the citations above: “and could occur in some actinmyosin systems [Wollrab et al., 2018]”
“in situations where the motors were assumed to dwell at the filament ends”
[sentence deleted] (subsection “Qualitative predictions”).
Finally, this comment is also relevant to the Discussion, specifically the fact that the absence of certain features can be taken as a proof of some other mechanism: I would argue that we do not know all the mechanisms operating in a cytoskeletal system, and with this in mind, can the author be confident enough to use the absence of a predicted behavior to make a strong conclusion about some other mechanism?
I agree with the reviewer’s careful take on the issue, and have modified some of the language of the Discussion accordingly, replacing my previous uses of “ruling out” selfassembly mechanisms by “arguing against” and “tending to disqualify”.
https://doi.org/10.7554/eLife.51751.sa2Article and author information
Author details
Funding
FP7 People: MarieCurie Actions (PCIG12GA2012334053)
 Martin Lenz
H2020 European Research Council (Stg677532)
 Martin Lenz
LabEx PALM (ANR10LABX0039 PALM)
 Martin Lenz
Agence Nationale de la Recherche (ANR15CE13000403)
 Martin Lenz
The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.
Acknowledgements
I thank Alex Mogilner for sharing Oelz et al. (2015) before publication, Pierre Ronceray for enlightening discussions and Samuel CazayusClaverie, Michael Murrell, Guglielmo Saggiorato and Danny Seara for comments on the manuscript. This work was supported by Marie Curie Integration Grant PCIG12GA2012–334053, ‘Investissements d’Avenir’ LabEx PALM (ANR10LABX0039PALM), ANR grant ANR15CE13000403 and ERC Starting Grant 677532. My group belongs to the CNRS consortium CellTiss.
Senior Editor
 Anna Akhmanova, Utrecht University, Netherlands
Reviewing Editor
 Raymond E Goldstein, University of Cambridge, United Kingdom
Reviewer
 Francois Nedelec
Publication history
 Received: September 9, 2019
 Accepted: March 5, 2020
 Accepted Manuscript published: March 9, 2020 (version 1)
 Version of Record published: March 19, 2020 (version 2)
Copyright
© 2020, Lenz
This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.
Metrics

 846
 Page views

 172
 Downloads

 8
 Citations
Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.
Download links
Downloads (link to download the article as PDF)
Open citations (links to open the citations from this article in various online reference manager services)
Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)
Further reading

 Physics of Living Systems
Foraging mammals exhibit a familiar yet poorly characterized phenomenon, ‘alternation’, a pause to sniff in the air preceded by the animal rearing on its hind legs or raising its head. Rodents spontaneously alternate in the presence of airflow, suggesting that alternation serves an important role during plumetracking. To test this hypothesis, we combine fully resolved simulations of turbulent odor transport and Bellman optimization methods for decisionmaking under partial observability. We show that an agent trained to minimize search time in a realistic odor plume exhibits extensive alternation together with the characteristic castandsurge behavior observed in insects. Alternation is linked with casting and occurs more frequently far downwind of the source, where the likelihood of detecting airborne cues is higher relative to ground cues. Casting and alternation emerge as complementary tools for effective exploration with sparse cues. A model based on marginal value theory captures the interplay between casting, surging, and alternation.

 Physics of Living Systems
Computational model reveals why pausing to sniff the air helps animals track a scent when they are far away from the source.