Closing while flowing: How a shrinking cytokinetic interface avoids membrane self-congestion

Animal cell cytokinesis is classically described as constriction of a RhoA-dependent actomyosin contractile ring, comprising a dense array of circumferentially aligned actin filaments. This view has been extraordinarily productive, organizing more than five decades of work on furrow positioning, RhoA signaling, actomyosin assembly, and ring constriction. The mitotic spindle promotes local RhoA activation at the equatorial cortex (Bement et al., 2005), driving actin assembly, myosin recruitment, and contractile ring organization as the plasma membrane ingresses between the nascent daughter cells. Many core components, regulatory pathways, and mechanical behaviors of the contractile ring are now understood in considerable detail and have been expertly synthesized in numerous reviews (Barr and Gruneberg, 2007; D’Avino et al., 2015; Glotzer, 2017; Green et al., 2012; Pollard and O’Shaughnessy, 2019). Additional organizers and linkers, including anillin, septins, and ERM proteins, associate with this machinery and contribute to robust furrow ingression (D’Avino, 2009; Debadarshini et al., 2025; Fehon et al., 2010; McClatchey, 2014; Mostowy and Cossart, 2012; Piekny and Maddox, 2010; Russo and Krauss, 2021; Spiliotis and Nakos, 2021).

Yet, the contractile ring does not act in isolation. It operates at the apex of a three-dimensional furrow, where the plasma membrane, cortical cytoskeleton, and their associated proteins together form a shrinking membrane–cortex interface. As this interface closes, it must remain mechanically engaged to transmit force, while continually reorganizing as the circumference available to its components decreases. Thus, ring closure concerns not only force generation; it also requires material organization and redistribution at a shrinking boundary.

This problem of material organization and redistribution becomes especially acute when cortical flow is considered. The equatorial RhoA program drives sustained, actomyosin-dependent reeling-in of cortex from the surrounding cortex (Cao and Wang, 1990b; Reymann et al., 2016). Because this inflow carries membrane-associated material sufficiently coupled to the cortex to be advected into the shrinking interface (Alonso-Matilla et al., 2024; Ng et al., 2005; Wang et al., 1994), closure cannot be understood simply as tightening of a preexisting belt. The furrow must shrink while material continues to enter it. This creates a mass-balance paradox: how can the furrow continue closing without becoming congested by the material it gathers? One apparent escape would be to treat the plasma membrane as a freely fluid surface whose lipids simply diffuse away from the narrowing apex. Indeed, the implicit default is often that membrane-associated material remains sufficiently redistributable during constriction. However, direct imaging instead shows membrane and membrane-associated material being advected into the furrow as the boundary shrinks, accompanied by substantial membrane accumulation and folding (Alonso-Matilla et al., 2024). Such folding may accommodate excess membrane area, but it does not by itself resolve the fate of mechanically engaged membrane–cortex material continually reeled into an interface whose effective engagement capacity is decreasing. Treating that material as freely redistributable may be valid in some contexts; the key question is when this approximation holds, when it fails, and what mechanisms permit mechanically engaged material to disengage and redistribute as the interface shrinks. This physical constraint was anticipated most clearly by Thomas Schroeder, whose classic electron microscopy and perturbation experiments showed that the cytokinetic ring is both force-transmitting and rapidly maintained, while its membrane attachments must remain sufficiently dynamic to permit continued remodeling during closure (Box 1).

Physical models have shown how active cortical stresses, actin turnover, cortical flows, and surface tension can account for major features of furrow ingression (Pollard and O’Shaughnessy, 2019; Turlier et al., 2014), yet important mechanistic gaps remain. Wang argued that the standard contractile ring hypothesis needed to be rethought (Wang, 2005), and Pollard subsequently organized the field around major, still-open questions, including how the ring attaches to the plasma membrane, produces force, and disassembles as it constricts (Pollard, 2017). I propose that these questions converge on a further physical constraint: how does a shrinking membrane–cortex interface remain force-transmitting while avoiding congestion by the material that contraction and cortical flow gather into it? In other words, how does it close while flowing?

This Perspective extends the contractile ring framework to accommodate these spatial and material constraints. I develop the argument at two levels. First, a constraint-level view frames cytokinesis as closure of a shrinking membrane–cortex interface under continued material inflow. Second, a mechanistic model proposes how coordinated actomyosin-driven Loading and anillin/septin-facilitated Relief could satisfy this constraint, and how their balance may vary across cytokinetic operating regimes.

The central argument is that ring closure requires not only contractile force generation, but also the continual processing of membrane–cortex material at a shrinking interface. Key spatial and descriptive terms used below are summarized in Box 2.

The cytokinetic apparatus spans a 3D cleavage zone (Fig. 1). At its center is the furrow apex, which harbors the contractile ring proper (CR proper)—a narrow, belt-like strap where F-actin becomes strongly circumferentially aligned and generates hoop tension in that axis. The CR proper lies within a broader RhoA zone and is flanked on both sides by cortical flow fields that deliver actomyosin and other cortical material toward the equator (Khaliullin et al., 2018; Reymann et al., 2016). The equatorial actomyosin network can initially contain relatively disordered filaments (Spira et al., 2017), whereas convergent cortical flow promotes their circumferential alignment, and rapid filament turnover coupled to filament-guided assembly can preserve this organization as the CR proper forms (Li and Munro, 2021; Reymann et al., 2016). As its circumference decreases, continued delivery creates an additional material constraint: material occupying the shrinking furrow apex must be continually reorganized. Thus, the force-bearing apex lies within a larger cleavage zone apparatus through which material continues to move during closure.

Classic surface-marking and cortical flow observations emphasized that furrowing involves organized movement and deformation of the cell surface, not merely contraction of an internal ring (discussed in Chapter 8 of Rappaport, 1996). Direct tracking subsequently showed both equator-directed movement of cortical actin filaments (Cao and Wang, 1990a; Cao and Wang, 1990b) and surface-bound receptor markers (Wang et al., 1994). Membrane-associated material is therefore clearly advected with such flow and concentrated at the cleavage furrow (Alonso-Matilla et al., 2024; Ng et al., 2005; Wang et al., 1994), with substantial membrane folding (Alonso-Matilla et al., 2024). Such accumulation does not necessarily represent mechanically engaged membrane–cortex material: additional area can be accommodated in folded or microvillar reservoirs. In cleaving Xenopus embryos, membrane delivered to the cleavage region is extensively incorporated into dynamic protrusions and folds that can act as a local sink for excess surface area (Danilchik and Brown, 2008), while during Drosophila cellularization, an apical microvillar reservoir supplies the ingressing furrow and is replenished by exocytosis (Figard et al., 2016; Figard et al., 2013). Thus, bulk membrane area can be dynamically stored, supplied, and released without specifying the fate of the mechanically engaged fraction at the furrow apex.

The mechanically relevant “membrane” considered here is therefore not the lipid bilayer in isolation, but membrane–cortex material through which contractile forces gain purchase on the cell surface. Its composition, attachment lifetime, and effective mechanical coupling may vary between cell types, cell regions, and mechanical states. Adhesion, polarity, membrane–cortex coupling, and cortical organization could all influence how readily this material reorganizes under load. These complexities can be coarse-grained by asking when mechanically engaged membrane–cortex material retains sufficient capacity for disengagement and redistribution, and when this capacity becomes a bottleneck.

Several conserved proteins implicated in cytokinesis provide molecular links between this interface and the cortical machinery. ERM-family proteins dynamically link cortical actin to the plasma membrane and are implicated in the regulation of furrow ingression (Carreno et al., 2008; Hiruma et al., 2017; Kunda et al., 2008; Kunda et al., 2012; Roubinet et al., 2011). Anillin interacts with actomyosin, active RhoA, membrane lipids, and septins, contributing to furrow anchorage and contractile ring organization (Field and Alberts, 1995; Jananji et al., 2017; Liu et al., 2012; Piekny and Glotzer, 2008; Straight et al., 2005; Sun et al., 2015; Tian et al., 2015). Septins form curvature-sensitive, membrane-associated higher order assemblies and are recruited to the furrow through anillin in several cell types (Carim and Hickson, 2023; Field et al., 2005a; Hickson and O’Farrell, 2008; Liu et al., 2012; Maddox et al., 2005). Septins can also associate directly with membranes through several mechanisms, including a recently identified transmembrane domain in some animal septins (Perry et al., 2025). Other membrane-remodeling proteins, including BAR-domain proteins such as syndapin (Takeda et al., 2013), may also contribute to the local interface organization and rearrangeability. The membrane itself may also be locally specialized: cholesterol and other lipid species are enriched at cytokinetic membranes (Atilla-Gokcumen et al., 2014; Kunduri et al., 2022; Ng et al., 2005), while perturbation of PI(4,5)P2 can disrupt cortical actin attachment to the plasma membrane during cytokinesis (Field et al., 2005b). Together, these features provide candidate mechanisms for establishing and remodeling membrane–cortex purchase during furrow ingression, although which define the mechanically engaged fraction may differ between cell types.

Membrane–cortex coupling is also remodeled globally during division. At the poles, cortical relaxation and membrane–cortex detachment permit pressure-driven blebbing, which can dissipate mechanical stresses generated during furrow constriction (Rodrigues et al., 2015; Roubinet et al., 2011; Sedzinski et al., 2011). Such responses demonstrate that cytokinetic contraction can drive redistribution and deformation of the membrane–cortex interface well beyond the furrow. They may provide a global route for accommodating material and mechanical load, but do not specify the fate of mechanically engaged material at the furrow apex.

Schroederʼs observations imply that contractile ring shrinkage is accompanied by net loss of ring material (Schroeder, 1972), but cytoskeletal disassembly alone cannot explain shrinkage of the mechanically engaged interface. Actin and myosin can depolymerize or dissociate, but as cortical flow continues to deliver material toward a furrow apex of decreasing circumference, the fate of membrane-associated material must also be specified. Sustained shrinkage therefore requires net exit of material from its mechanically engaged, load-bearing fraction. Importantly, this does not require bulk membrane to leave the furrow. Rather, membrane–cortex material must leave the mechanically engaged fraction of the shrinking apex, requiring changes in its attachment state and/or position relative to the contractile cortex. This is the mass-balance core of “closing while flowing.”

Schroederʼs “firm but brief” constraint provides a physical mechanism for this exit. Membrane must couple strongly enough to contractile elements to transmit force, yet release frequently enough to move or shear relative to the shrinking ring. Transient disengagement therefore permits membrane–cortex material to slip relative to the contractile cortex. Such slip provides a local mode of Relief, allowing material to leave the mechanically engaged fraction rather than remaining trapped at the furrow apex, where it may prevent shrinkage.

If the furrow were a single contractile layer, it would be difficult to understand how the same membrane-coupled structure could simultaneously draw material into the apex and provide an organized route for material to leave it. Instead, studies across diverse experimental systems reveal axial separation between membrane-distal actomyosin and membrane-proximal anillin/septin architectures (Garno et al., 2021; McDonald et al., 2017). This cortical lamination provides a potential spatial solution to the inflow–outflow problem: actomyosin-rich organization can draw and compress membrane–cortex material into the apex, while a distinct membrane-proximal architecture could facilitate its lateral or rearward redistribution toward the flanks (Fig. 1).

Our work in Drosophila S2 cells has provided several indications of such physical separation. Line-profile analysis revealed an axial offset between myosin and the anillin C terminus at the furrow (Kechad et al., 2012), while latrunculin A exposed Rho1-dependent anillo-septin assemblies that remain membrane-associated independently of actomyosin (Hickson and O’Farrell, 2008). Related anillo-septin material can also be displaced outward from the cytokinetic interface and shed during midbody-ring formation, when furrow diameter approaches its minimum (El Amine et al., 2013). Together, these observations show that membrane-proximal cytokinetic material can become physically separated from the force-bearing actomyosin architecture and can leave the interface as a coherent membrane-associated structure.

Nanoscale imaging also shows a layered organization in fission yeast (McDonald et al., 2017), and sea urchin embryos reveal nascent septin-rich filamentous networks emerging beneath myosin clusters (Garno et al., 2021). Lamination therefore establishes the spatial possibility that force generation and material exit need not be assigned to the same cortical layer, providing a structural basis for closing while flowing.

Any satisfactory model of sustained closure must account simultaneously for continued equatorial inflow, turnover of the contractile apparatus, and net exit of membrane material from the mechanically engaged fraction of the shrinking interface. This physical constraint may be more general than the specific mechanochemical model proposed below: wherever cytokinesis involves shrinkage of a mechanically engaged membrane–cortex interface under continued material delivery, that material must somehow leave the shrinking load-bearing fraction. The mass-balance problem may therefore be broadly conserved even where its molecular solution differs; anillo-septin Routing represents one candidate solution, particularly in animal cytokinesis.

If one accepts the constraint-level view, a minimal mechanochemical model follows naturally: equatorial RhoA activity within the cleavage zone powers a radial Loading–Relief cycle that maintains the compression interface while allowing it to shrink circumferentially (Fig.1). In this model, the actomyosin CR proper is a central force-generating module, but not the whole cytokinetic machine. It operates on a membrane–cortex interface that serves as the organizing surface for RhoA activation, actomyosin assembly, and anillo-septin assembly. The mechanochemical model therefore has three coupled elements (Fig. 2): (1) a membrane surface that organizes the machinery; (2) an actomyosin-rich Loading module that generates force and compresses membrane–cortex material; and (3) a membrane-proximal anillo-septin module that promotes Relief and Rerouting. Box 3 summarizes the proposed physical concepts used in this framework.

Actomyosin-rich architectures (Rok/myosin and formin/actin) capture and compress membrane–cortex material at the furrow apex, generating contractile stress and maintaining the force-bearing CR proper. This inward gathering represents “Loading” of the compression zone.

Sustained shrinkage requires exit from the mechanically engaged fraction. Transient disengagement and relative slip provide a direct means for material to move relative to the actomyosin cortex rather than simply being compressed with it. This exit from the mechanically engaged fraction represents “Lateral Relief” of the compression zone.

I propose that membrane-proximal anillo-septin organization exploits the window of opportunity provided by slip to bias disengaged membrane–cortex material laterally away from the furrow apex. RhoA-dependent anillo-septin assemblies could thereby organize the trajectory of slipping material toward the flanking cleavage zone, reducing the likelihood that disengaged material simply reengages actomyosin Loading elements (Carim et al., 2020).

Loading and Relief should not be interpreted as completely separate molecular phases. Membrane-proximal anillo-septin assemblies may form early and be advected into the furrow with other cortical material, so that the same inflow that loads and compresses the interface also concentrates Relief-competent architecture needed for continued shrinkage. The membrane geometry of the cleavage zone could also feed back on this organization. Septin membrane association and higher order assembly are curvature-sensitive (Cannon et al., 2019), raising the possibility that the strongly nonplanar membrane topology generated during furrowing helps localize or stabilize membrane-proximal anillo-septin architectures. Conversely, septin assembly can itself influence membrane organization, so the relationship between membrane curvature and anillo-septin architecture could be reciprocal.

This proposed anillo-septin–dependent Routing mechanism need not be the only route by which the local material-balance constraint is satisfied. Cytoskeletal components can leave by dissociation or depolymerization, whereas membrane folding, microvillar growth, and caveolae can accommodate bulk membrane area and buffer bilayer tension without necessarily reducing occupancy of the mechanically engaged fraction (Andrade et al., 2022; Figard et al., 2013; LaFoya and Prehoda, 2023). Such reservoirs contribute to Relief only to the extent that they reduce the material or mechanical burden on the shrinking interface; specifically for the mechanically engaged fraction considered here, this requires disengagement as material enters them.

Similarly, cytokinesis-associated exocytosis can supply surface area without specifying the fate of material already occupying the force-bearing interface. Targeted secretion can continue when furrow progression is arrested or reversed (Shuster and Burgess, 2002), and supplied membrane can enter folds, protrusions, and microvillar reservoirs (Danilchik and Brown, 2008; Figard et al., 2016; Figard et al., 2013). Exocytosis therefore addresses membrane supply but does not by itself provide Relief.

Endocytosis provides a potentially more direct route. Furrow-specific endocytosis occurs in zebrafish embryos, and endocytic trafficking also operates within the later intercellular bridge (Feng et al., 2002; Frémont and Echard, 2018). Endocytosis could constitute Relief if internalization removes material directly from the mechanically engaged fraction, or act downstream after prior disengagement. Whether endocytosis occurs within the densely occupied force-bearing apex is unclear; Schroeder described this region as a pronounced “zone of exclusion” for recognizable vesicles and organelles (Schroeder, 1972). Resolving endocytic events relative to the force-bearing interface should distinguish these possibilities.

Cytoskeletal turnover could also contribute to Relief when it produces net loss of mechanically engaged material, although rapid exchange alone need not do so. Actin within the contractile ring can exchange rapidly, but turnover produces net loss from the CR proper only when vacated positions are not equivalently refilled. As discussed previously using a “musical chairs” analogy (Carim et al., 2020), rapid exchange can maintain stable occupancy if departing components are continuously replaced. Relief and Routing of membrane–cortex material could instead reduce the local availability of sites for renewed actin assembly, biasing turnover toward net disassembly. Conversely, cytoskeletal turnover could create opportunities for the transient disengagement and slip required for Relief. Together, these processes could couple progressive consumption of the CR proper to redistribution of the membrane–cortex interface, extending Schroederʼs original insight.

The resulting passage of mechanically engaged membrane–cortex material through the interface can be described as throughput: Loading and Relief describe entry into and exit from the engaged fraction, whereas throughput describes the resulting material flux. At the level of a local furrow sector, interface assembly and interface shrinkage may therefore represent successive operating states of the same material-processing system: initial Loading builds and maintains a force-bearing interface, whereas continued operation during geometrical closure increasingly requires sufficient exit from the mechanically engaged fraction to maintain throughput.

According to this mechanochemical model, closing while flowing is implemented by a single equatorial RhoA program with two coupled outputs. Rok/myosin and formin/actin drive Loading by building the membrane-distal contractile layer and compressing membrane–cortex material at the apex, whereas membrane-proximal anillo-septin assembly promotes Relief by biasing material laterally toward the flanks. The shrinking apex therefore behaves as a material-processing bottleneck: insufficient Relief relative to Loading should favor accumulation and congestion, whereas sufficient Relief should permit sustained closure. Local variation in this balance could contribute to differences in furrow progression around the circumference, including asymmetric or unilateral ingression which can itself be associated with spatially asymmetric cortical flows (Menon et al., 2017).

Anillin is ideally positioned within this system because it contributes to actomyosin-rich Loading architecture (Field and Alberts, 1995; Jananji et al., 2017; Kučera et al., 2021; Straight et al., 2005), while its membrane-proximal association with septins places it to support Lateral Relief (Carim and Hickson, 2023; El Amine et al., 2013; Garno et al., 2021; Hickson and O’Farrell, 2008; Kechad et al., 2012). Consistent with anillin influencing the distribution rather than simply the magnitude of RhoA-dependent outputs, disrupting RhoA binding to anillin altered effector organization and shifted Caenorhabditis elegans furrowing from asymmetric toward more symmetric ingression (Lebedev et al., 2025). Relief need not be equally limiting throughout furrowing: as compression increases, closure may become progressively more sensitive to the ability to clear the bottleneck, so that defective Relief may first alter symmetry, robustness, or closure regime before strongly reducing bulk ingression rate.

A further consequence is that Relief could redistribute RhoA signaling activity and membrane–cortex material. The RhoGEF ECT2 is usually concentrated near the furrow apex (Husser et al., 2022), whereas RhoA itself often occupies a broader, anillin-dependent zone (Bement et al., 2005; Piekny and Glotzer, 2008). Because membrane-proximal anillo-septin assemblies are themselves RhoA-dependent and enriched in RhoA (Carim and Hickson, 2023; Hickson and O’Farrell, 2008), their lateral redistribution could carry active RhoA away from its site of generation, contributing to the broader RhoA zone. This is consistent with anillin-dependent enhancement of RhoA membrane residence and furrow localization (Budnar et al., 2019; Piekny and Glotzer, 2008).

Several recognizable phenotypes become more coherent when anillin and septins are viewed as regulators of membrane–cortex material coupling and throughput rather than only as actomyosin scaffolds. In Drosophila and mammalian cells, anillin depletion can allow furrow assembly and ingression to initiate, but furrows often become laterally unstable, oscillatory, and unable to close to completion (Goldbach et al., 2010; Hickson and O’Farrell, 2008; Piekny and Glotzer, 2008; Straight et al., 2005; Zhao and Fang, 2005). These phenotypes likely arise from several contributing defects including microtubule-dependent anchoring via Citron kinase and/or centralspindlin (Bassi et al., 2013; D’Avino et al., 2008; El Amine et al., 2013; El-Amine et al., 2019; Gai et al., 2011; Gregory et al., 2008), but they are also consistent with continued actomyosin-driven Loading becoming poorly matched by Relief. In this regime, additional contractility may increase compression and promote congestion rather than productive closure.

Importantly, reduced Relief capacity need not produce outright failure. In the C. elegans zygote, anillin/septin-depleted furrows can still ingress, but become more symmetric and more sensitive to mild reduction of myosin function (Maddox et al., 2007). Such phenotypes are consistent with a system operating closer to its material-processing bottleneck: sufficient capacity remains for ingression, but the margin for accommodating changes in Loading or geometry is reduced.

This Perspective may also change how experimentally nonclosing furrows are interpreted. HeLa cells treated with the myosin II inhibitor blebbistatin form nascent furrows in which anillin accumulates in prominent filamentous structures punctuated by myosin (Straight et al., 2005). Such structures could represent membrane-proximal anillo-septin organization that continues to assemble when actomyosin-dependent compression and geometrical advance are strongly reduced. Blebbistatin also slows F-actin turnover (Guha et al., 2005; Murthy and Wadsworth, 2005), raising the possibility that actomyosin-dependent compression normally promotes cytoskeletal turnover as material is gathered and processed through the furrow. In this regime, Routing architecture may assemble without the mechanical activity needed to couple cytoskeletal turnover, transient disengagement, and net material exit to interface shrinkage.

Reciprocally, depletion of anillin or septins can increase rather than decrease furrow F-actin (Jordan et al., 2016), while anillin depletion increases myosin II accumulation and disrupts its normal asymmetry (Hsu et al., 2023). Thus, loss of anillin/septin does not simply weaken assembly of the contractile machinery. Instead, impaired Relief could allow membrane–cortex material to remain available for renewed actomyosin engagement, biasing ongoing turnover toward equivalent replacement rather than net disassembly. Continued or increased actomyosin accumulation after anillin/septin loss would then reflect a shift in material processing and engagement, which could increase productive contractility in some regimes but promote excessive compression, instability, or congestion in others.

The same logic may help rationalize dosage and genetic-interaction effects. Intermediate anillin/ANI-1 levels in C. elegans produce maximal constriction speed, whereas both higher and lower levels reduce performance (Descovich et al., 2018), consistent with anillin tuning a working range in which membrane–cortex coupling is strong enough for force transmission but permissive enough for remodeling and Relief. Similarly, the otherwise surprising suppression of formin-deficient cytokinesis by reducing septin and/or anillin function, either experimentally or through cell type–specific regulation (Connors et al., 2024; Jordan et al., 2016; Lebedev et al., 2023), is consistent with simultaneous reduction of Loading and Relief partially restoring their balance.

Cytokinesis spans symmetric and unilateral furrows, asymmetric divisions, and persistent intercellular bridges (Gerhold et al., 2022; Sugioka, 2022), while closely related cells can differ in their genetic requirements (Connors et al., 2024; Davies et al., 2018; Husser et al., 2022; Ozugergin et al., 2022; Ozugergin and Piekny, 2022). Even actomyosin rings within a common cytoplasm can exhibit distinct closure kinetics and component dynamics (Linehan et al., 2025). The Loading–Relief framework treats these as different operating regimes of a shared mechanochemical system. Accordingly, geometry need not impose a universal constriction rate: ring size predicts initial constriction rate in early C. elegans embryos (Carvalho et al., 2009), whereas this relationship is developmentally restricted and polarity-sensitive in mouse embryos (Paim and FitzHarris, 2022).

Asymmetric divisions producing unequally sized daughter cells provide another important example. In Drosophila neuroblasts, cortical polarity can position an asymmetric furrow even in the absence of the mitotic spindle (Cabernard et al., 2010), while cortical myosin and Rho1-dependent organization are themselves spatially patterned during division (Connell et al., 2011; LaFoya and Prehoda, 2023; Montembault et al., 2017; Montembault et al., 2023). The two sides of the same furrow can therefore present different membrane–cortex landscapes to the cytokinetic machinery. Polarity-dependent differences in local coupling and rearrangeability could alter the effective balance of Loading and Relief, thereby changing material throughput and the mechanical conditions for geometrical advance on the two sides of the furrow.

Open and incomplete contractile structures extend this idea around the circumference. Local sectors of cytokinetic rings can behave semiautonomously, consistent with descriptions of locally organized contractile units and sectoral furrow behavior (Carvalho et al., 2009; Khaliullin et al., 2018; Silva et al., 2016; Werner et al., 2024). The open-ended cytokinetic bands described in zebrafish provide an especially striking example: locally anchored portions can ingress even after the band is experimentally interrupted, demonstrating that productive advance does not require constriction of a complete ring; moreover, the rate of advance is gated by cell cycle–dependent changes in microtubule-mediated cytoplasmic rheology (Kickuth et al., 2026). In Loading–Relief terms, each sufficiently competent sector could process material locally, while whether and how rapidly it advances also depends on the surrounding mechanical environment.

Persistent intercellular bridges provide another operating regime. In the C. elegans germline, incompletely constricted cytokinetic rings become incorporated into the rachis and persist as open bridges while retaining cytokinetic machinery and undergoing continued remodeling (Amini et al., 2014; Bauer et al., 2021; Gerhold et al., 2022; Goupil et al., 2017). While a conventional furrow may be thought of as closing while flowing, a persistent bridge may have the related problem of staying open while flowing. Importantly, geometrical stationarity does not necessarily imply low material throughput. A nonclosing interface could reflect weak Loading and Relief, as is likely in a blebbistatin-inhibited furrow, or substantial rates of both, with entry into and exit from the mechanically engaged fraction approximately balanced, as may occur in a persistent rachis bridge. Furrow advance may therefore report the net geometrical outcome of material processing, not the magnitude of the underlying material throughput.

The Loading–Relief framework treats furrows as material-processing interfaces whose geometries can advance, remain stationary, or change locally while material continues to flow through them. Material engagement, material trajectory, and interface geometry are related, but not equivalent, readouts. Tests should therefore measure them together, relating exit from the mechanically engaged fraction and material trajectories to actomyosin Loading, anillo-septin organization, and geometrical progression across different operating regimes. Closing furrows directly test the mass-balance requirement during interface shrinkage, while stationary or transiently nonclosing interfaces may be especially informative because material processing can be resolved from geometrical advance.

The most direct test is to track material through the furrow interface while simultaneously measuring its geometrical behavior. Direct membrane readouts—such as lipidated or GPI-anchored reporters, photoactivatable/photoswitchable membrane probes, fluorescent membrane dyes, or surface-bound lectin markers—should reveal whether material advected into the furrow apex subsequently remains there, disengages from the load-bearing cortex, or redistributes toward the flanking cleavage zone. In a closing regime, material exit need not appear as absolute movement away from the equator: it could continue moving inward while moving rearward relative to the advancing furrow apex. Material trajectories should therefore be analyzed relative to the apex where the CR proper operates. Local photoconversion or pulse-labeling approaches would be particularly informative because they could distinguish material already occupying the apex from material subsequently delivered by cortical flow.

The critical measurement is the fate of the mechanically engaged fraction in relation to interface geometry, not simply total membrane accumulation at the furrow, since that can be accommodated in folds or microvillar projections. During closure, persistent retention of mechanically engaged material at the apex despite decreasing circumference would argue against the proposed local mass-balance mechanism, whereas measurable exit from the engaged fraction would support it. In stationary regimes, the key question is whether continued Loading and exit continue despite little geometrical change, or whether both have slowed.

If a septin-rich, membrane-proximal Routing architecture is structurally distinct from the contractile Loading layer, septins should show a reproducible axial offset from actomyosin and distinct kinematics, such as differential advection or turnover, or progressive deposition behind the apex during closure. Because anillin is proposed to contribute to both Loading and Relief, septins provide a cleaner readout of the membrane-proximal Routing architecture. Super-resolution and 3D imaging in sea urchin embryos already provide a precedent for scoring septin-rich filament networks beneath myosin clusters (Garno et al., 2021). Coupling these measurements to direct membrane trajectories would test whether lamination is associated with material redistribution rather than merely spatial separation. Persistent septin-rich structures deposited behind the advancing furrow apex should covary with redistribution of membrane–cortex material and may retain spatial information about previous material trajectories through the interface.

If anillo-septin organization promotes net exit from the mechanically engaged fraction, perturbing it should shift the relationship between contractile assembly, material retention, and geometrical progression. Such perturbations may preserve or even enhance actomyosin intensity, width, or residence time while increasing material retention at the apex and reducing lateral redistribution. Septin depletion provides one experimental avenue, but a particularly sharp test would selectively disrupt anillo-septin organization while preserving anillin’s actomyosin-associated functions, e.g., using a RhoA binding–defective mutant (Carim and Hickson, 2023; Sun et al., 2015). In closing regimes, the predicted signature is a mismatch between contractile assembly and material exit, with increased occupancy associated with intermittent, oscillatory, stalled, and/or more symmetric closure. A complementary test would ask whether restoring Relief without proportionally increasing contractility improves productive advance.

If Relief depends on material being able to disengage and redistribute, physical perturbations that restrict relative membrane–cortex movement should increase apex retention and alter the relationship between contractile input and geometrical progression without directly altering the core RhoA machinery. Lectin-based membrane pinning provides one possible approach to assess this. Combining such perturbations with direct material tracking would test whether reduced mobility shifts the operating regime: in closing interfaces, it may promote accumulation, oscillation, or stall, whereas in a dynamically maintained nonclosing interface, it could instead transiently favor constriction by reducing Relief relative to Loading.

Polarity-patterned asymmetric divisions that give rise to different-sized daughter cells provide a natural within-cell test because opposing sides of the same furrow can experience different membrane–cortex landscapes while using the same core cytokinetic machinery. In Drosophila neuroblasts, where polarity can position and bias furrow ingression (Cabernard et al., 2010; Connell et al., 2011; Montembault et al., 2017; Montembault et al., 2023), simultaneous measurement of material trajectories, contractile Loading, and anillo-septin organization on the two sides of the furrow could test whether local differences in geometrical advance or furrow position correlate with differences in Loading, Relief, or their balance. Such divisions therefore provide a stringent within-cell test of whether local membrane–cortex environment alters cytokinetic operating regime despite shared core machinery.

Elements of these signatures have already been observed across systems, but rarely as coordinated readouts of material processing at the cytokinetic interface. The discriminating advance will be to relate material engagement and trajectories, cortical lamination/Routing architecture, and geometrical progression within the same experiments.

A key next step is predictive formalization. Minimal models could represent material Loading into a cytokinetic interface together with finite capacity for disengagement and exit, and ask how their balance generates different geometrical regimes. Interface shrinkage provides the critical mass-balance case, but the same framework should accommodate productive, intermittent, and stationary states with different underlying rates of material processing. Imaging-derived measurements of material trajectories, apex occupancy, local Loading–Relief balance, and interface geometry could then constrain these models and generate falsifiable predictions for transitions between regimes. The aim is to convert the qualitative mass-balance framework developed here into a predictive one.

More broadly, the framework proposed here is intended as a practical way of seeing and scoring the cleavage apparatus differently. Cytokinesis cartoons tend to depict the plasma membrane as a homogeneous, compliant bilayer and the furrow as a largely one-layer contractile belt. In contrast, the present view treats the relevant object as a dynamic, laminated membrane–cortex interface in which mechanically engaged material must remain sufficiently dynamic for force transmission, material redistribution, and geometrical change to coexist. Closure places a particular demand on this system because continued Loading into an interface of decreasing circumference requires sufficient exit from the mechanically engaged fraction. Different cell types and cytokinetic geometries may satisfy these requirements in different ways; identifying those solutions should reveal which features of cytokinesis are universal physical requirements and which are context-dependent molecular implementations.

I thank all past and present members of the Hickson Lab for their contributions to the development of these ideas, especially Amel Kechad, Sabrya Carim, Nour El-Amine, Silvana Jananji, Guillaume Chambaud, and Vlad Martin. I thank Janet Iwasa (The Animation Lab, University of Utah) for help in developing the figures. I thank Jean-Claude Labbé, Alisa Piekny, Jeremy Carlton, and Pat OʼFarrell for constructive comments on various drafts of the manuscript. I thank the anonymous reviewers for their insightful comments that greatly helped sharpen the argument.

Work in my laboratory is supported by the Canadian Institutes of Health Research (PJT-183759) and the Natural Sciences and Engineering Research Council of Canada (RGPIN-2021-03324). I apologize to colleagues whose work could not be cited because of space limitations. Generative artificial intelligence (ChatGPT, OpenAI; multiple model versions up to and including 5.6 Sol) was used extensively in developing this Perspective. The tool was used recursively and iteratively, under detailed direction by the author, for scientific dialogue, conceptual development and organization, and drafting and revision of text. Its outputs were repeatedly interrogated, corrected, rejected, or refined by the author, and substantial portions of the text, including the Abstract, originated in AI-generated drafts. The author determined the scientific framework, interpretation and selection of evidence, arguments, and conclusions, reviewed and verified the final manuscript, and takes full responsibility for its accuracy, originality, and content.

Author contributions: Gilles R.X. Hickson: conceptualization, funding acquisition, investigation, project administration, resources, and writing—original draft, review, and editing.