We generated D. melanogaster
S2 cell lines expressing anillin-GFP. The anillin-GFP fusion rescued loss of endogenous anillin (Fig. S1, available at http://www.jcb.org/cgi/content/full/jcb.200709005/DC1
) and its localization paralleled that of endogenous anillin. In interphase it was nuclear, at metaphase it was uniformly cortical, and in anaphase it accumulated at the equator while being lost from the poles ( and A′; and Video 1, available at http://www.jcb.org/cgi/content/full/jcb.200709005/DC1
; Field and Alberts, 1995
; Echard et al., 2004
). In some highly expressing cells, nuclear anillin-GFP formed filaments not ordinarily seen with anillin immunofluorescence (unpublished data), but these disassembled upon nuclear envelope breakdown and the overexpression had no appreciable effect on the progress or success of cytokinesis.
Figure 1. RhoGEFPbl controls anillin-GFP localization during anaphase via F-actin– and myosin II–dependent and –independent mechanisms. (A–H) Frames from time-lapse sequences of anillin-GFP cells progressing through anaphase/telophase (more ...)
Equatorial localization of anillin-GFP during anaphase occurs via multiple RhoGEFPbl-dependent mechanisms
We tested whether RhoGEFPbl
contributed to anillin localization during cytokinesis. After 3 d of RhoGEFPbl
RNAi or Rho1
RNAi (unpublished data), anillin-GFP localized to the cortex in metaphase but did not relocalize to the equator during anaphase ( and B′; and Video 2, available at http://www.jcb.org/cgi/content/full/jcb.200709005/DC1
), indicating a requirement for RhoGEFPbl
that is consistent with prior analysis of fixed RhoGEFPbl
mutant embryos (Prokopenko et al., 1999
). Because anillin can bind F-actin (Field and Alberts, 1995
) and phosphorylated myosin regulatory light chain (MRLC; Straight et al., 2005
might regulate anillin indirectly through its control of F-actin and myosin II.
We used latrunculin A (LatA) to test whether F-actin was required for anillin-GFP localization. A 30–60-min incubation of 1 μg/ml LatA abolished cortical anillin-GFP localization in metaphase (), indicating an F-actin requirement at this phase. However, when anillin normally relocalizes to the equator (~3–4 min after anaphase onset), anillin-GFP formed punctate structures that became progressively more filamentous over the next few minutes, reaching up to several micrometers in length and having a thickness of ~0.3 μm ( and Video 3, available at http://www.jcb.org/cgi/content/full/jcb.200709005/DC1
). These linear anillin-containing structures contained barely detectable levels of F-actin (Fig. S2) and formed specifically at the plasma membrane and preferentially at the equator, although subsequent lateral movement often led to a more random distribution ( and Video 3). Thus, anillin responds to spatiotemporal cytokinetic cues even after major disruption of the F-actin cytoskeleton. A substantial (albeit incomplete) reacquisition of cortical phalloidin staining was observed in cells fixed after washing out the drug for a few minutes (unpublished data). In live cells, LatA washout immediately after formation of the anillin structures allowed the preformed structures to migrate from a broad to a compact equatorial zone as the cells attempted to complete cytokinesis ( and Video 4). This movement indicates that an F-actin–dependent process can contribute to the equatorial focusing of anillin.
We tested the influence of RhoGEFPbl on anillin behavior in LatA. After RNAi of RhoGEFPbl () or Rho1 (not depicted), anillin-GFP remained cytoplasmic through anaphase. Thus RhoGEFPbl and Rho1 are required for anaphase anillin behavior, whether the cortex is intact or disrupted by LatA treatment.
We tested whether myosin II impacted anillin-GFP localization. Compared with controls ( and A′), RNAi of the gene encoding MRLC spaghetti squash
; Karess et al., 1991
) inhibited cell elongation during anaphase (Hickson et al., 2006
), slowed furrow formation, and delayed and diminished the equatorial localization of anillin-GFP ( and F′; and Video 5, available at http://www.jcb.org/cgi/content/full/jcb.200709005/DC1
). However, unlike after RhoGEFPbl
RNAi, equatorial accumulation of anillin-GFP was not altogether blocked. It was still recruited but in a broad zone (). Furthermore, in the presence of LatA, MRLCSqh
RNAi did not affect the formation of the anillin-GFP structures (). We conclude that myosin II contributes to the equatorial focusing of anillin when the F-actin cortex is unperturbed but that myosin II is dispensable for anillin behavior in LatA.
Collectively, these data suggest that multiple RhoGEFPbl
-dependent inputs control anillin localization. The slowed equatorial accumulation of anillin when myosin II function was impaired indicates a myosin II–dependent input. That reassembly of the cortical F-actin network (after washout of LatA) allowed preformed anillin structures to move toward the cell equator indicates an F-actin–dependent input. This is consistent with the concerted actions of myosin II and F-actin driving cortical flow, as observed in other cells (Koppel et al., 1982
; Cao and Wang, 1990
; Wang et al., 1994
; DeBiasio et al., 1996
; Fishkind et al., 1996
), and is reminiscent of the coalescence of cortical nodes during contractile ring assembly in Schizosaccharomyces pombe
(Wu et al., 2006
). However, the F-actin– and myosin II–independent behavior of anillin in LatA indicates an additional RhoGEFPbl
-dependent input. We tested possible involvement of Dia or Rok, the Rho effectors responsible for cytokinetic F-actin assembly and myosin II activation, respectively. Neither Dia
RNAi, alone or combined ( and Fig. S3, available at http://www.jcb.org/cgi/content/full/jcb.200709005/DC1
), prevented formation of the anillin structures in LatA, indicating that they too were dispensable. Thus, RhoGEFPbl
can control anillin behavior in anaphase via a previously unrecognized route. Immunofluorescence analysis revealed extensive colocalization between endogenous Rho1 and anillin-GFP in LatA, indicating that Rho1 was itself a component of these structures (). Although other Rho effectors could be involved, our findings are consistent with prior indications that Rho1 and anillin may directly interact (Suzuki et al., 2005
Equatorial accumulation of myosin II during anaphase can occur independently of F-actin and anillin, but stable furrow positioning requires anillin
We examined myosin II localization (), as it can bind anillin and is controlled by RhoGEFPbl
. As previously reported (Dean et al., 2005
-GFP was able to localize to the equatorial membrane independently of F-actin, and in doing so formed filamentous structures resembling those observed with anillin-GFP ( and Video 6, available at http://www.jcb.org/cgi/content/full/jcb.200709005/DC1
). Indeed, anillin and MRLCSqh
(detected as either MRLCSqh
-GFP or with an antibody to serine 21–phosphorylated pMRLCSqh
) colocalized (), although they were often offset as if labeling different regions of the same structures ().
Figure 2. Myosin II localization can occur independently of F-actin and anillin but stable furrow positioning requires anillin. (A and B) Frames from time-lapse sequences of cells expressing MRLCSqh-GFP progressing through anaphase. (A) Control cell expressing (more ...)
We tested the effects of anillin RNAi on MRLCSqh
-GFP localization. MRLCSqh
-GFP recruitment and furrow initiation appeared normal, but within a few minutes of initiation, furrows became laterally unstable and oscillated back and forth across the cell cortex, parallel to the spindle axis, in repeated cycles, each lasting ~1–2 min and eventually subsiding to yield binucleate cells after ~20 min ( and Video 7, available at http://www.jcb.org/cgi/content/full/jcb.200709005/DC1
). The phenotype was very similar to that reported for anillin RNAi in HeLa cells (Straight et al., 2005
; Zhao and Fang, 2005
) and represents a requirement for anillin at an earlier stage than previously noted in D. melanogaster
(Somma et al., 2002
; Echard et al., 2004
). Thus, a conserved function of anillin is to maintain furrow positioning during ingression. A potential mechanism is proposed in the following paragraph.
In LatA, anillin RNAi did not prevent equatorial MRLCSqh
-GFP recruitment, but instead of appearing as persistent linear structures distorting the cell surface, a more reticular and dynamic structure lacking cell surface protrusions was observed (compare and Video 8 [available at http://www.jcb.org/cgi/content/full/jcb.200709005/DC1
] with and Video 6; and compare with ). Thus myosin II can localize independently of both anillin and F-actin but the filamentous appearance of myosin II in the presence of LatA requires anillin, indicating that anillin can influence myosin II behavior in the absence of F-actin, whereas myosin II appeared capable of influencing anillin behavior only in the presence of F-actin ().
Septins are required for the filamentous structures in LatA and septin recruitment depends on anillin
Septins are multimeric filament-forming proteins that can bind anillin in vitro and function with anillin in vivo (Kinoshita et al., 2002
; Field et al., 2005
; Maddox et al., 2005
). Using an antibody to the septin Peanut (Neufeld and Rubin, 1994
), we found that in nontransfected S2 cells, septinPnut
localized to the cleavage furrow and midbody where it colocalized with anillin (). Unexpectedly, the septinPnut
antibody also strongly labeled bundles of cytoplasmic ordered cylindrical structures, each ~0.6 μm in diameter and of variable length (up to several micrometers; and Fig. S3). These staining patterns could be greatly reduced by septinPnut
RNAi and were thus specific (unpublished data). The cylindrical structures did not appear to be cell cycle regulated, as they were apparent in interphase, mitotic, and postmitotic cells. They also did not colocalize with anillin (), nor did their stability rely on anillin (unpublished data). Incubation with 1 μg/ml LatA before fixation inevitably led to disassembly of most of these large structures; however, the resulting distribution of septinPnut
depended on the cell cycle phase. In LatA-treated interphase cells, when anillin is nuclear, septinPnut
formed cytoplasmic rings, ~0.6 μm in diameter, which are similar to the Septin2 rings seen in interphase mammalian cells treated with F-actin drugs or in the cell body of unperturbed ruffling cells (Kinoshita et al., 2002
; Schmidt and Nichols, 2004
). In LatA-treated mitotic cells, septinPnut
was diffusely cytoplasmic (or barely detectable) in early mitosis, whereas in anaphase/telophase, it localized to the same plasma membrane–associated anillin-containing filamentous structures described in , whether marked by anillin-GFP () or antibody staining of endogenous anillin (Fig. S3).
Figure 3. Anillin recruits septin to the cleavage furrow and to filamentous structures in LatA. (A–C) Nontransfected (A and B) and anillin-GFP–expressing (C) S2 cells fixed during anaphase/telophase and labeled with antibodies to septinPnut (middle; (more ...)
We analyzed anillin behavior after septinPnut
RNAi. Although unable to fully deplete septinPnut
, we found that anillin could localize to the equatorial cortex in regions devoid of detectable septinPnut
(Fig. S3), which is consistent with findings in Caenorhabditis elegans
(Maddox et al., 2005
). Importantly, in septinPnut
-depleted cells, anillin-GFP still localized to the plasma membrane in LatA but no longer appeared filamentous ( and Video 9, available at http://www.jcb.org/cgi/content/full/jcb.200709005/DC1
), indicating that septinPnut
is essential for the filamentous nature of the structures and that Rho1 can promote the association of anillin with the plasma membrane independently of septinPnut
. However, in this case the plasma membrane to which anillin-GFP localized subsequently exhibited unusual behavior. It was internalized in large vesicular structures, apparently in association with midzone MTs (Fig. S3 and Video 9). Although we do not understand this phenomenon, it may be related to events induced by point mutations in the septin-interacting region of anillin that give rise to abnormal vesicularized plasma membranes during D. melanogaster
cellularization (Field et al., 2005
We tested the effects of anillin RNAi on the localization of septinPnut. Using Dia as a furrow marker (), 3 d of anillin RNAi prevented the furrow recruitment of septinPnut (). In LatA-treated cells, anillin RNAi did not affect the formation of septinPnut rings in interphase cells (not depicted), but it greatly reduced the formation of septinPnut-containing structures during anaphase/telophase (). Thus, anillin is required for the furrow recruitment of septinPnut and for the formation of septinPnut-containing structures in 1 μg/ml LatA. In contrast, Dia could still localize to the equatorial plasma membrane after combined anillin RNAi and LatA treatment (), indicating that it can localize independently of both anillin and F-actin. Thus, although Dia partially colocalized with anillin in LatA (Fig. S3), this likely reflected independent targeting to the same location rather than an association between anillin and Dia.
Our data argue that Rho1, anillin, septins, and the plasma membrane participate independently of F-actin in the formation of a complex. However, anillin, septins and F-actin can also form a different complex in vitro,
independently of Rho (Kinoshita et al., 2002
). Perhaps two such complexes dynamically interchange in vivo.
A potential role for anillin in anchoring the cleavage furrow to midzone MTs
We tested the involvement of MTs in anillin behavior in LatA. Overnight incubation with 25 μM colchicine effectively depolymerized all MTs in mitotic cells and promoted mitotic arrest, as expected (unpublished data). Using Mad2 RNAi to bypass the arrest (Logarinho et al., 2004
), we observed anillin-GFP during mitotic exit in the absence of MTs and in the presence of LatA (). Under such conditions, anillin-GFP formed filamentous structures very similar to those formed when MTs were present, indicating that MTs were dispensable for their formation. However, the structures appeared uniformly around the plasma membrane rather than restricted to the equatorial region (), which is consistent with the role MTs play in the spatial control of Rho activation (Somers and Saint, 2003
; Bement et al., 2005
; Yuce et al., 2005
; Piekny et al., 2005
; Kamijo et al., 2006
Figure 4. Anillin-containing structures in LatA form independently of MTs but associate with MT ends at the plasma membrane. (A) Time-lapse sequence of a cell expressing anillin-GFP and mcherry-tubulin exiting mitosis in the presence of colchicine (16 h) and LatA (more ...)
Shown by immunofluorescence, the LatA-induced anillin structures localized to the ends of nonoverlapping astral MTs directed toward the equator (, endogenous anillin).Live imaging of cells coexpressing cherry-tubulin and anillin-GFP revealed bundles of MTs associating with the filamentous anillin-GFP structures as they formed (; and Video 10, available at http://www.jcb.org/cgi/content/full/jcb.200709005/DC1
). Colocalization between anillin-GFP structures and MT ends persisted over many minutes, even after considerable lateral movement at the membrane. Thus, although the anillin structures formed independently of MTs, they stably associated with MTs. These findings support prior biochemical evidence for interactions of MTs with both anillin and septins (Sisson et al., 2000
) and reveal a potential positive-feedback loop in which MTs directed where Rho1–anillin–septin formed linear structures at the plasma membrane, whereas the structures in turn associated with the MT ends. An MT plus end–binding ability of anillin–septin could explain the furrow instability phenotype elicited by anillin RNAi (). Accordingly, anillin may physically link Rho1 to MT plus ends during furrow ingression, thereby promoting the focusing and retention of active Rho1, thus stabilizing the furrow at the equator ().
Figure 5. A model for anillin regulation and function during cytokinesis. Pathways controlled by RhoGEFPbl, including well-documented inputs into F-actin and myosin II (black) and a novel, separate input (pink) via anillin and septins (A). Through equatorial stimulation (more ...)
Rho-dependent association of anillin with the plasma membrane
These live-cell analyses highlight an unusual behavior of the Rho-dependent anillin-containing structures at the plasma membrane. Initially forming beneath and parallel to the plasma membrane (), the structures then often lifted on one side to appear perpendicular to the cell surface while remaining anchored at their base by MTs (; and Video 10). We interpret this reorientation as reflecting avid binding to and subsequent envelopment by the plasma membrane. Although intrinsically stable, the structures exhibited dynamic movement within the plane of the plasma membrane and were capable of sticking to one another, via their ends, giving rise to branched structures (, Y shape) that were also capable of breaking apart. Anillin has a pleckstrin homology domain within its septin-interacting region and a membrane-anchoring role of anillin has long been postulated (Field and Alberts, 1995
). Our data support such a role and suggest that it is controlled by Rho.
The relationship between events in LatA and the normal events of cytokinesis
Our data highlight the complexity of RhoGEFPbl signaling and lead to a model in which multiple Rho-dependent inputs synergize to control anillin behavior during cytokinesis ().
At the appropriate time and location of normal cytokinesis, several proteins, including Rho1, MRLCSqh
, Dia, anillin, and SeptinPnut
, localized to the equatorial membrane in the presence of LatA. Of these (and apart from Rho1 itself), anillin and SeptinPnut
were uniquely and specifically required for the formation of the linear filamentous structures that we describe. The behaviors of these structures are consistent with prior studies of ordered assemblies of septins and anillin (Oegema et al., 2000
; Kinoshita et al., 2002
) and of interaction between anillin and MTs (Sisson et al., 2000
). Although such structures are not normally seen in furrowing cells, anillin localizes to remarkably similar filamentous structures in the cleavage furrows of HeLa cells arrested with the myosin II inhibitor blebbistatin (see Fig. 5 in Straight et al., 2005
It seems unlikely that LatA induced the structures we describe through nonspecific aggregation of proteins. Rather, we propose that LatA blocked a normally dynamic disassembly of Rho1–anillin–septin complexes (by blocking an F-actin–dependent process required for the event) and that continued assembly promoted formation of the linear structures (). Because blebbistatin slows F-actin turnover (Guha et al., 2005
; Murthy and Wadsworth, 2005
), blebbistatin (Straight et al., 2005
) and LatA (this study) may have induced filamentous anillin-containing structures via a common mechanism. A dynamic assembly/disassembly cycle involving anillin could promote transient associations between the plasma membrane and elements of the contractile ring and MTs, properties that could contribute to furrow stability and plasticity. Finally, because local loss of F-actin accompanies and may indeed trigger midbody formation (Schroeder, 1972
), LatA treatment may have stabilized events in a manner analogous to midbody biogenesis and could therefore be useful in understanding this enigmatic process.