Journ

alof

Cell

Scie

nce

Sho1p connects the plasma membrane with proteins ofthe cytokinesis network through multiple isomericinteraction states

Karolina Labedzka1, Chen Tian1, Ute Nussbaumer1, Steffi Timmermann1, Paul Walther2, Judith Muller1 andNils Johnsson1,*1Institute of Molecular Genetics and Cell Biology, Department of Biology, and 2Central Facility for Electron Microscopy, Ulm University,James-Franck-Ring N27, 89081 Ulm, Germany

*Author for correspondence (nils.johnsson@uni-ulm.de)

Accepted 19 April 2012Journal of Cell Science 125, 4103–4113� 2012. Published by The Company of Biologists Ltddoi: 10.1242/jcs.105320

SummaryAn understanding of cytokinesis at the molecular level requires a detailed description of the protein complexes that perform central

activities during this process. The proteins Hof1p, Cyk3p, Inn1p and Myo1p each represent one of the four genetically defined andpartially complementary pathways of cytokinesis in the yeast Saccharomyces cerevisiae. Here we show that the osmosensor Sho1p isrequired for correct cell–cell separation. Shortly before cytokinesis Sho1p sequentially assembles with Hof1p, Inn1p and Cyk3p, into acomplex (the HICS complex) that might help to connect the membrane with the actin–myosin ring. The HICS complex is formed

exclusively through interactions between three SH3 domains located in Cyk3p, Hof1p and Sho1p, and five acceptor sites found inCyk3p, Hof1p and Inn1p. Owing to the overlapping binding specificities of its members the HICS complex is best described asensembles of isomeric interaction states that precisely coordinate the different functions of the interactors during cytokinesis.

Key words: Constraint interaction networks, Cytokinesis, Protein interaction, Split–Ubiquitin, Systems biology

IntroductionTo separate mother from daughter, yeast cells undergo a well-

orchestrated set of biochemical reactions: contraction of the

actin–myosin ring (AMR), ingression and fusion of the plasma

membrane, synthesis of a chitin-containing primary and a

secondary septum (PS, SS), and, finally, the partial degradation

of the primary septum to liberate both cells from each other

(Balasubramanian et al., 2004; Bi, 2001; Schmidt et al., 2002).

Although the enzymes responsible for performing these

activities are well known it remains an unresolved issue how

these enzymes are regulated, and how their individual

contributions are integrated in time and space to form a precisely

operating cell–cell separation machinery (Schmidt et al., 2002;

Tolliday et al., 2001). One outstanding question is how the AMR is

connected to the membrane to transform its contraction into

membrane ingression. Candidate proteins for this linkage are the

proteins Hof1p and Inn1p. Both are known to be involved in the

correct progression of cell separation (Kamei et al., 1998;

Lippincott and Li, 1998; Sanchez-Diaz et al., 2008; Vallen et al.,

2000). Their specific roles have, however, not yet been precisely

defined. Hof1p may directly connect the membrane with the AMR

during contraction as the phenotypes of its deletion as well as its

localization during cytokinesis are compatible with such a role

(Meitinger et al., 2011). In addition, Hof1p contains at its

N-terminus a F-BAR domain. Other members of the family of

F-BAR-containing proteins were shown to bind to the actin

cytoskeleton and to simultaneously bind and bend membranes

(Itoh et al., 2005; Rao et al., 2010). However, both features have

not been demonstrated for Hof1p. Hof1p was shown to contact

through its C-terminal SH3 domain Inn1p, and Cyk3p, a further

SH3-domain-containing protein (Jendretzki et al., 2009; Korinek

et al., 2000; Nishihama et al., 2009). Inn1p is essential for

cytokinesis. Its N-terminal C2 domain was postulated to bind to

membranes and the co-precipitation with Iqg1p, a member of the

AMR, was taken as evidence for an important role in linking the

membrane to the AMR (Sanchez-Diaz et al., 2008). An alternative

model suggesting that Inn1p and Cyk3p activate and coordinate

the formation of the SS and PS recently questioned this conclusion

(Nishihama et al., 2009).

One goal of protein–protein interaction analyses is a detailed

interaction map that helps to navigate through the simultaneously

or sequentially occurring events during cellular processes such as

cytokinesis (Vidal et al., 2011). In yeast, many of the relevant

protein interactions have probably already been detected by

genome-wide screens and are easily accessible in public

databases. For example, novel interaction partners of Hof1p

were recently identified through a systematic mapping of the

interaction partners of all yeast genome-encoded SH3 domains by

phage display, peptide spotting, and two-hybrid analysis

(Tonikian et al., 2009). However, a direct incorporation of

large-scale protein interaction data into models of cytokinesis is

not straightforward, as the data sets are still incomplete and

potentially erroneous. Furthermore, most of the applied high

throughput methods are performed under conditions rather distant

from the situation within the living cell and thus lack one level of

information required for realistic and predictive model building.

We recently introduced a Split–Ubiquitin-based approach

(Split–Ub) of collecting and structuring protein–protein

Research Article 4103

Journ

alof

Cell

Scie

nce

interactions to build constraint protein interaction networks (CIN)

(Hruby et al., 2011; Johnsson and Varshavsky, 1994; Muller and

Johnsson, 2008). By experimentally determined constraints these

networks allow to discriminate between different interaction

states. Interaction states are defined as the subset of all detected

interactions of a local network that are simultaneously allowed.

This feature of CINs reduces the number of possible

combinations of protein interactions within a network and thus

provides better frameworks for mechanistically meaningful

models describing their functions and workings (Hruby et al.,

2011).

Here we report on a local protein interaction network around

the integral membrane protein and osmosensor Sho1p. Sho1p,

together with other membrane proteins, senses conditions of high

osmolarity by a still unknown mechanism. It then serves as a

platform for the assembly of the MAP kinase pathway that

transforms the signal into a cellular response (Hohmann, 2009;

Tatebayashi et al., 2007). Our derived CIN suggests an additional

unexpected role of Sho1p in cytokinesis by linking through its

SH3 domain the known members of the cytokinesis machinery

Hof1p, Inn1p and Cyk3p to the plasma membrane.

ResultsNovel binding partners of Sho1p

A Split–Ub-based systematic interaction screen against an array

of 383 Nub fusions (Nub and Cub denote the N- and C-terminal

halves of Ub) revealed novel and known interaction partners of

the Cub–RUra3 fusion of the osmolarity sensor Sho1p

(Sho1CRU; Fig. 1A; supplementary material Table S1).

Although this Nub array is enriched in fusion genes involved in

polarized growth, cytokinesis, and stress responses we noted

among the binding partners a surprising predominance of proteins

involved in cytokinesis (Hruby et al., 2011). To exclude in a first

step those of the detected contacts that merely reflect the co-

Fig. 1. Sho1p interacts with known members of the cytokinesis pathway. (A) Split–Ub interaction assay of yeast strains co-expressing Sho1CRU (left panel)

or Sho11–297CRU (right panel) with the selected Nub fusion proteins. Cells were grown to OD600 of 1 and 4 ml of the cultures spotted as tenfold serial dilutions on

5-FOA plates. Nub without a C-terminally attached ORF (Nub–) serves as a negative control in this and the following Split–Ub assays. Growth of the cells indicates

interactions between the co-expressed Ub fusions. Note that the Sho11–297CRU-expressing cells show some growth even in the presence of non-interacting Nub

fusions. Here, only a significantly elevated growth serves as an indicator of interaction. (B) Protein extracts of yeast cells expressing MYC-tagged Pbs2p

(lanes 1–3), Hof1p (lanes 4–6), Cyk3p (lanes 7–9) and Inn1p (lanes 10–12) were treated with bacterially expressed GST (lanes 2, 5, 8, 11) or GST–SH3Sho1 bound

to Sepharose beads (lanes 3, 6, 9, 12). Glutathione eluates (lanes 2, 3, 5, 6, 8, 9, 11, 12) and yeast extracts (lanes 1, 4, 7, 10) were separated by SDS-PAGE

followed by immunoblotting with anti-MYC antibodies. (C) Bacterially expressed AGT (lane 3), AGT fusions of Cyk3160–210 (lanes 1, 4) and its proline-rich

motif-missing mutant (Cyk3160–210DP–AGT; lanes 2, 5) were treated with GST (lanes 1, 2) or GST–SH3Sho1-bound beads (lanes 3–5). Glutathione eluates were

separated by SDS-PAGE followed by immunoblotting with anti-AGT antibodies. (D) Upper panel: cartoon of Hof1p showing the FCH and SH3 domain as well as

the proline-rich motifs at position 333–338 (P1) and 386–392 (P2). Lower panel: as in C except that Hof1269–349–AGT (lanes 1–3) and Hof1366–423–AGT were

precipitated with GST (lanes 2, 5) or GST–SH3Sho1-bound beads (lanes 3, 6). Bacterially expressed AGT fusions of Hof1269–349 and Hof1366–423 are shown in

lanes 1 and 4. (E) As in C except that Inn1131–211–AGT (lanes 1–3), Inn1291–341–AGT (lanes 4–6) and Inn1351–409–AGT (lanes 7–9) were precipitated with GST

(lanes 2, 5, 8) or GST–SH3Sho1-bound beads (lanes 3, 6, 9). Bacterially expressed AGT fusions of Inn1131–211, Inn1291–341 and Inn1351–409 are shown in lanes 1, 4

and 7. (F) As in C except that Msb11–39–AGT (lanes 1–3) was precipitated with GST (lane 2) or GST–SH3Sho1-bound beads (lane 1). Bacterially expressed anti-

AGT-probed Msb11–39–AGT is shown in lane 3. (G) Summary of interaction data: the proline-rich motifs (white nodes) of the proteins Pbs2p, Hof1p, Cyk3p,

Inn1p and Msb1p might compete for binding to the SH3 domain (grey node) of Sho1p. Interactions between proteins are indicated by edges. Proteins with more

than one interaction site are indicated by a thick horizontal line.

Journal of Cell Science 125 (17)4104

Journ

alof

Cell

Scie

nce

localization within the membrane rather than a specific protein–

protein interaction, we repeated the experiment with a truncatedversion of Sho1p. This mutant still harbored its four trans-membrane spanning elements but lacked the C-terminal SH3

domain (Sho11–297). Comparison of the obtained interactionprofiles for Sho1CRU and Sho1p1–297CRU left as SH3-dependentbinding partners the MAP kinase kinase of the osmolaritypathway Pbs2p, the cytokinesis proteins Hof1p and its N-

terminally truncated version Hof198–669, Inn1p, Cyk3p and its N-terminally truncated allele Cyk369–885, the scaffold proteinMsb1p, the chitin synthase Chs3p, Sho1p, and the stimulator of

exocytosis Sro7p (supplementary material Table S1; Fig. 1A)(Hao et al., 2007; Tonikian et al., 2009). To prove a directinvolvement of the SH3 domain of Sho1p (SH3Sho1, residues

298–367) in the potentially cytokinesis-relevant interactions, weprecipitated from yeast extracts MYC-tagged versions of Cyk3p,Inn1p and Hof1p with the bacterially expressed and immobilized

SH3Sho1. As shown in Fig. 1B, all proteins with a well-established role in cytokinesis bound the SH3 domain ofSho1p. These interactions were abolished when the SH3domain was deleted from the full-length Sho1p (Fig. 1A).

Dissecting the SH3Sho1 binding sites

Besides the full-length Cyk3p the Nub array contains two N-terminally truncated versions missing the N-terminal SH3 domain

(Nub-Cyk369–885), or the N-terminal 196 residues (Nub-Cyk3196–885),respectively. Remarkably, Sho1CRU still interacted with Nub-Cyk369–885 but not with Nub-Cyk3196–885, although both Nub

fragments of Cyk3p were expressed at similar levels in the yeast(Fig. 1A; supplementary material Fig. S1). This finding locates thebinding site to SH3Sho1 between residues 69 and 196 of Cyk3p. This

sequence harbors a proline-rich stretch between residues 183 and191 that resembles the binding motif of Pbs2p for Sho1p (Maedaet al., 1995). We expressed residues 160–210 of Cyk3p as a fusion toO6-alkylguanine-DNA alkyltransferase in E. coli (AGT; AGT–

Cyk3160–210). AGT–Cyk3160–210 directly interacted with the purifiedGST fusion of SH3Sho1 (Fig. 1C). Replacement of the prolinesin P185PLPPLP191 of Cyk3p (AGT–Cyk3160–210DP) completely

abolished the binding to GST–SH3Sho1 (Fig. 1C).

Two findings restricted the binding site of SH3Sho1 to theregion between residues 98 and 592 of Hof1p: the Nub fusion ofHof198–669 lacking its FCH domain still bound to Sho1CRU, and

bacterially expressed SH3Sho1 precipitated Hof1p lacking the C-terminal SH3 domain (Nub-Hof11–592) (supplementary materialFig. S1). This central region of Hof1p harbors at positions

333–338 and 386–392 two proline-rich stretches (P1, P2) asprospective docking sites for SH3Sho1 (Fig. 1D). Pull-downassays with bacterially expressed Hof1p fragments showed a

specific and direct interaction between the purified SH3Sho1 andHof1269–349, harboring P1. Hof1366–423, containing P2, did notinteract with SH3Sho1 (Fig. 1D).

Inn1p is a further binding partner of Sho1p with a well-

established role in cytokinesis (Fig. 1A,B) (Nishihama et al.,2009; Sanchez-Diaz et al., 2008; Tonikian et al., 2009). Inn1pcontains multiple proline-rich motifs that are known to bind to

different SH3-domain-containing proteins including Hof1p andCyk3p (Fig. 3C) (Nishihama et al., 2009; Tonikian et al., 2009).To identify the motif in Inn1p responsible for binding to Sho1p,

we expressed in E. coli fragments of Inn1p as AGT fusionproteins, each encompassing different proline-rich stretches.Precipitations with purified GST–SH3Sho1 identified the most

C-terminally located proline-rich motif of Inn1p as the ligand forSH3Sho1 (Fig. 1E).

Msb1p was originally identified as a suppressor of certainbem1 and cdc24 alleles and was later shown to stimulate thesynthesis of b1–3 glucan in the extracellular matrix (Bender andPringle, 1989; Sekiya-Kawasaki et al., 2002). Inspection of the

Msb1p sequence revealed at position 6–12 a region with a closeresemblance to the proline-rich motif in Pbs2p (Maeda et al.,1995). We precipitated the N-terminal 39 residues of Msb1p as

bacterially expressed AGT fusion (Msb11–39–AGT) with thepurified GST–SH3Sho1. The efficient binding supports ourhypothesis that this motif indeed represents the in vivo binding

site for Sho1p (Fig. 1F).

In summary, this detailed protein interaction analysis of Sho1pidentifies as binding sites of its SH3 domain four new proline-

rich motifs in four different proteins. Three of these proteins,namely Hof1p, Cyk3p and Inn1p, are known to play an importantrole in cytokinesis (Fig. 1G).

The architecture of the HICS complex: identification of newbinding partners and dissection of the binding sitesOur interaction analysis suggested that Hof1p, Cyk3p, Inn1p and

Sho1p might act together in a complex. To identify additionalbinding partners of this putative complex we screened Cyk3CRU,Hof1CRU and Inn1CRU against our Nub array (supplementarymaterial Table S1). We could confirm the pairwise interactions

between all three proteins. In addition, Hof1p and Cyk3p but notInn1p were found to interact with Nba1p. Cdc3p was a uniqueinteraction partner of Hof1p. Slt2p, Tpk1p and Bud14p bound

exclusively to Cyk3p. Sla1p, Ubc9p as well as Cdc14p interactedonly with Inn1p (see supplementary material Table S1; Fig. 2A,Fig. 3A,B).

To delineate the individual binding sites for their ligands, wecreated different fragments of Cyk3p, Hof1p and Inn1p as CRU

fusions. The N-terminal fragment of Cyk3 containing both theSH3 domain as well as the SH3Sho1 binding site (Cyk31–267CRU)still interacted with Nub-Hof1p, -Inn1p and -Nba1p but not anylonger with Nub-Slt2p and -Bud14p (Fig. 2A). The isolated

SH3Cyk3 (Cyk31–79CRU) retained the ability to interact withNub-Inn1p and -Nba1p, whereas a Cyk3 fragment harboring theisolated proline-rich stretch (Cyk3119–267CRU) only interacted

with Nub-Hof1p and its N-terminally truncated variant Hof198–669

(Fig. 2A).

To further prove that the proline-rich stretch of Cyk3p isindeed the binding site for SH3Hof1, we mutated this site in the N-terminal Cyk3p fragment (Cyk31–267DPCRU) and tested it against

different Nub fragments of Hof1p (Fig. 2C). This experimentunambiguously mapped the interaction between Hof1p andCyk3p to SH3Hof1 and the proline-rich stretch between residues183 and 191 of Cyk3p (P1). This interaction was direct as the

purified GST–Hof1342–669 precipitated from E. coli extractsAGT–Cyk3160–210 and to a much lower degree AGT–Cyk3160–210DP (Fig. 2D). The small amount of precipitated

AGT–Cyk3160–210DP might have been caused by the highconcentrations of both proteins in the assay. In these in vitro

experiments we had to N-terminally extend the SH3 domain of

Hof1p [GST–Hof1342–669 (GST–SH3Hof1)] as a smaller fragmentincluding SH3Hof1 (residues 490–669) showed no in vitro activitytowards any of its ligands (our unpublished observation).

Do Slt2p and Bud14p bind to an autonomous domain within theC-terminal region of Cyk3p? We assayed a CYK3 fragment

Molecular description of the HICS complex 4105

Journ

alof

Cell

Scie

nce

missing SH3Cyk3 (Cyk3119–885CRU) and the same fragment

harboring a mutated P1 (Cyk3119–885DPCRU; Fig. 2B).

Nub-Bud14p and -Slt2p still interacted with both Cyk3CRU

fusions, whereas Nub-Hof1p was only bound by Cyk3119–885CRU.

Additionally, we detected a weak affinity of both Cyk3CRU

fragments to the N-terminally truncated version Nub-Cyk3196-885

(Fig. 2B). As homodimerization of full-length Cyk3p was never

observed we did not include this interaction in the summary of the

obtained results (Fig. 2E).

Similar to Sho1CRU, Hof1CRU still interacted with

Nub-Cyk369–885 but not with Nub-Cyk3196–885 (Fig. 3A). The

experiment provided additional evidence that the same proline-

rich stretch in Cyk3p served as binding site for Hof1p and Sho1p.

A C-terminal fragment of Hof1p (Hof1343–669CRU) harboring the

SH3 but not its F-BAR domain still interacted with Nub-Cyk3p,

-Cyk369–885, -Inn1p and Nub-Nba1p (Fig. 3A).

Inn1p displays four proline-rich stretches within its sequence

(Fig. 3C) (Nishihama et al., 2009). The interactions of Inn1p with

Cyk3p and Hof1p were already mapped to the proline-rich motifs

P2 and P4 of Inn1p, respectively (Nishihama et al., 2009). To

independently map the binding sites of some of the herein found

interaction partners within the Inn1p sequence we tested different

Inn1CRU fragments against a selection of Nub fusions (Fig. 3B).

The measured interaction between Nub-Cyk3p and Inn11–211

confirmed previous studies (Fig. 3B) (Nishihama et al., 2009).

Hof1p and Hof198–669 interacted with P4 (Inn1367–409) and P3

(Inn11–373). Sla1p and Cdc14p interacted exclusively with

Inn1367–409 (Fig. 3B). This dissection predicted that the SH3

domains of Hof1p, Sla1p and Sho1p might compete for binding the

C-terminal P4 motif of Inn1p (Fig. 3C). The absence of a SH3

domain lets us assume that the binding of Cdc14p is not mediated

by P4 and thus not directly influenced by the presence of the other

ligands (Fig. 3C). A cluster of in vivo phosphorylation sites

in Inn1367–409 might constitute the docking site for Cdc14p

(Bodenmiller et al., 2008).

The identification of non-overlapping binding sites supports

the existence of a so far unknown protein complex consisting of

the four core subunits Hof1p, Inn1p, Cyk3p and Sho1p (HICS

complex; Fig. 3C).

Membrane attachment of SH3Sho1 is a prerequisite forligand binding in vivo

The in vitro precipitations of the proline-rich motifs with the

isolated SH3Sho1 were performed under high concentrations of

both reactants (Fig. 1C–F). In vivo, Hof1p, Cyk3p and Inn1p

locate to the bud neck below the membrane during cytokinesis

while Pbs2p is distributed in the cytosol (Jendretzki et al., 2009;

Nishihama et al., 2009; Sanchez-Diaz et al., 2008). Are the

Fig. 2. Dissection of Cyk3p interaction sites. (A) Upper panels: Split–Ub interaction assays of Cyk3p and truncated versions of Cyk3p as in Fig. 1A. Lower

panel: cartoon of Cyk3p showing the N-terminal SH3 domain, the proline-rich motif between residues 183 and 191, and the trans-glutaminase domain between

residues 500 and 600. (B) Cells co-expressing Cyk3119–885CRU lacking the SH3 domain or the same truncation of Cyk3p with a mutated proline-rich motif

(Cyk3119–885DPCRU), and Nub fusions of the identified Cyk3p ligands were spotted in tenfold serial dilutions on medium containing 5-FOA. Nub–Guk1p-co-

expressing cells served as an additional negative control. (C) Yeast cells containing the C-terminal truncations of Cyk3p still harboring the SH3 domain and the

proline-rich motif (Cyk31–267CRU) or the same truncation of Cyk3p with mutated proline-rich motif (Cyk31–267DPCRU) and co-expressing Nub fusions of Hof1p

or its fragments (see Fig. 1D) were spotted on 5-FOA-containing medium. Expression of Nub fusions were induced by addition of copper to 50 mM, except Nub–

SH3Hof1, which showed high activity even under non-inducing conditions. (D) Pull-down assay as in Fig. 1C, except that AGT (lanes 1, 4), Cyk3160–210–AGT

(lanes 2, 5) and Cyk3160–210DP–AGT (lanes 3, 6) were incubated with immobilized GST–SH3Hof1. Anti-AGT-probed glutathione eluates of the beads are shown in

lanes 1–3 and extracts before incubation are shown in lanes 4–6. (E) Summary of interaction data. Horizontal lines connect SH3 domains (grey nodes) with

proline-rich stretches (white nodes) or other regions within one protein, whereas all other edges indicate interactions between different proteins (black nodes) or

their domains.

Journal of Cell Science 125 (17)4106

Journ

alof

Cell

Scie

nce

interactions between SH3Sho1 and its ligands restricted to the bud

neck? We compared the in vivo interaction profile of the CRU

fusion of SH3Sho1 (SH3Sho1CRU) with that of the full length

Sho1CRU to elucidate the influence of the cellular localization

on its interactions (Fig. 4A). Strikingly, only the Nub fusion of

Pbs2p was found as an interaction partner of the cytosolic

SH3Sho1CRU in this in vivo assay, implicating the membrane

association of SH3Sho1 as a prerequisite for its effective

interaction with Inn1p, Cyk3p, Msb1p and Hof1p (Fig. 4A).

Similar in vivo measurements with the SH3 domains of Cyk3p

and Hof1p suggested that the interactions of the isolated SH3Cyk3

with Inn1p and Nba1p and of SH3Hof1 with Inn1p, Nba1p and

Cyk3p are principally strong enough to recruit the respective

complexes from their pre-mitotic localization to the site of

cytokinesis (Fig. 2A,C, Fig. 3A).

Sho1p co-localizes with its ligands at the site of cytokinesis

Live cell imaging showed that Sho1–GFP appeared at the site of

septum formation shortly after the splitting of the septin ring

(Fig. 4B). Time-lapse studies of cells co-expressing Sho1–

CHERRY together with the GFP fusions of either Hof1p, Inn1p

or Cyk3p established that Sho1p reached the site of cell

separation after Hof1p and Inn1p and shortly after or nearly at

the same time as Cyk3p (Fig. 4B). While the ring characterized

by Inn1p and Hof1p staining contracted during cytokinesis,

Sho1p and Cyk3p appeared more static (Fig. 4B). Both proteins

remained at the opposing membranes of mother and daughter

cells some time after cell separation (Fig. 4B). These data and the

work by others suggest the following temporal sequence of

assembly and disassembly of the HICS complex at the site of cell

division: Hof1p, Inn1p, Cyk3p, Sho1p–Inn1p, Hof1p, Cyk3p,

Sho1p (Lippincott and Li, 1998; Meitinger et al., 2010;

Nishihama et al., 2009).

Taken together, our protein interaction maps and the timing of

protein appearances at the bud neck predict that the specific

localization of Cyk3p and Inn1p during cytokinesis depends on

Hof1p but not on Sho1p (Fig. 3C, Fig. 4B). Indeed, the single

deletion of Sho1p had only a minor effect on the localization of

Fig. 3. Dissection of Inn1p and Hof1p interaction sites. (A) Split–Ub

interaction assay as in Fig. 1C with Hof1CRU and a C-terminal fragment of

Hof1p lacking the N-terminal FCH domain and the P1 motif

(Hof1343–669CRU). Split–Ub interaction assay of Inn1CRU or fragments of

Inn1p containing different combinations of the proline-rich motifs P1–P4 (see

C). (C) Upper panel: summary of interaction data as in Fig. 2E. Lower panel:

cartoon of Inn1p showing the N-terminal potential lipid-binding C2 domain

and the four proline-rich motifs. (D) Protein interaction states as in Fig. 3C

but lacking P3 and P4 of Inn1p and either Cyk3p, or Sho1p. /indicates no

genetic interaction (GI), and – indicates a significant negative GI.

Fig. 4. Sho1p localizes to the site of cytokinesis. (A) Split–Ub interaction

assay as in Fig. 1A, between the SH3 domain of Sho1p (SH3Sho1CRU) and its

known interaction partners. (B) Fluorescence microscopy of cells co-

expressing Sho1–GFP and Cdc11–CHERRY, Sho1–CHERRY and Hof1–

GFP, Sho1–CHERRY and Inn1–GFP, and Sho1–CHERRY and Cyk3–GFP.

Except for Cdc11–CHERRY, all fusions were expressed under the control of

the PMET17 promoter from their chromosomal locations. Arrowheads indicate

the first appearance of Sho1p, Cyk3p or Inn1p at the bud neck, or the double

ring of Cdc11p and Hof1p at the bud neck. Scale bars: 5 mm.

Molecular description of the HICS complex 4107

Journ

alof

Cell

Scie

nce

Inn1–GFP and increased the fraction of cells containing Cyk3–

GFP at the bud neck (supplementary material Table S3). The

absence of Hof1p reduced Inn1p and Cyk3p at the bud neck quite

dramatically (supplementary material Table S3). In comparison,

the localization of Sho1p during cytokinesis did not depend on

the presence of Hof1p (supplementary material Table S3).

Sho1p functionally contributes to cytokinesis

The high connectivity of Sho1p within this cytokinesis interaction

network suggests a unique role of Sho1p in this biological process.

Electron microscopy of Dsho1 cells revealed a large fraction of

cells displaying a normally structured primary and secondary

septum. The primary septa of a minority of the cells displayed

additional invaginations of the plasma membrane close to the site

of septum formation (Fig. 5A; supplementary material Fig. S2). In

some cells the ingression of the plasma membrane during

cytokinesis appeared asymmetrically (supplementary material

Fig. S2). More prominently, half of the inspected cells displayed

a secondary septum approximately twice as thick as seen in wild-

type cells. Occasionally, we observed lacunae of cytosol within

this enlarged secondary septum (Fig. 5A,B).

What are the specific functions of Sho1p and its interaction

states within the above-established cytokinesis network? Our

results suggest that Hof1p and Sho1p can compete for binding to

the C-terminal proline-rich motifs in Inn1p and Cyk3p and might

thus perform similar or even redundant functions within this

network (Figs 1–3). To address this question we sporulated a

diploid Dhof1;Dsho1 strain. Spores harboring the double deletion

could be recovered but appeared less frequently than would have

been expected for non-interacting genes (supplementary material

Table S2; Fig. 6F). The diameters of the recovered Dhof1;Dsho1

colonies were roughly half in size compared to the Dhof1

colonies (supplementary material Table S2). We repeated the

sporulation in the presence of a HOF1-expressing centromeric

plasmid and tested the ability of Sho1p and a mutant lacking the

SH3 domain (sho11–240) to replace the plasmid-borne Hof1p in

the Dhof1;Dsho1 strain. Sho1–MYC partially complemented the

growth defect of the strain carrying the double deletion (Fig. 6A).

The complementation disappeared under conditions of low Sho1–

MYC expression (Fig. 6A). Sho11–240 did not complement the

double deletion and instead seemed to exert a negative effect on

the growth of the cells (Fig. 6A,F). This allele-specific

enhancement was confirmed by disrupting the reading frame of

the genomic SHO1 before the start of its C-terminal SH3 domain.

Cells lacking HOF1 and expressing this allele instead of the full

length SHO1 were not viable (Fig. 6B). The loss of the SH3Sho1

was, however, tolerated in cells lacking MYO1 or CHS2

(supplementary material Fig. S3).

To further test the significance of the interactions between

Sho1p and its targets during cytokinesis we overexpressed

Fig. 5. Deletion of SHO1 leads to defects in septum

formation. (A) Electron microscopy of wild-type (a–c)

and Dsho1 cells (d–i). Arrowheads indicate septa

between the cells; double carets indicate the plasma

membrane. LA, cytosol-enclosed lacuna; CW, cell wall.

(B) Distribution of cell wall thickness at the division sites

of separating wild-type and Dsho1 cells.

Journal of Cell Science 125 (17)4108

Journ

alof

Cell

Scie

nce

SH3Sho1 as a cytosolic CHERRY–SH3Sho1 fusion or as a

membrane-anchored SH3Sho1–Sso1p fusion in wild-type cells

and cells lacking HOF1. The expression of SH3Sho1–Sso1p

specifically impaired the growth of Dhof1 cells, implying that the

interactions between Sho1p and its targets become important for

viability under these conditions (Fig. 6C).

Hof1p is the major determinant for Inn1p and Cyk3p

localization at the bud neck (supplementary material Table S3).

To estimate the contribution of Sho1p we compared the

localization of both proteins in cells lacking HOF1 and SHO1

with those lacking only HOF1. Supplementary material Table S3

shows that the deletion of Sho1p further impaired the correct

assembly of Cyk3p and Inn1p at the division site.

Sho1p contacts Inn1p directly at the P4 motif and indirectly at

the P2 motif through Cyk3p (Fig. 3C). To dissect the importance

of either site for the cytokinetic function of Sho1p we created an

allele of INN1 (inn11–211) that lacked the two C-terminal proline-

rich motifs including the direct binding sites to Hof1p and Sho1p

(Fig. 3C). This allele keeps the physical interaction with Sho1p or

Hof1p indirectly through Cyk3p. Sporulation of a heterozygous

inn11–211;Dsho1 strain yielded haploid inn11–211, Dsho1 and

inn11–211;Dsho1 cells. Compared to the wild type a slightly

reduced diameter of the colonies of the inn11–211 spores reflected

the influence of the C-terminal interaction sites on the activity of

Inn1p (supplementary material Table S2). The cells were affected

by a strong defect in cytokinesis (supplementary material Fig. S3).

The further reduced colony diameters of the inn11–211;Dsho1

spores demonstrated a negative genetic interaction between

inn11–211 and SHO1 (supplementary material Table S2; Fig. 3D).

This finding supports the functional importance of Sho1p for

cytokinesis by either recruiting Inn1p to the membrane and/or by

providing a complementary, Inn1p-independent pathway of

cytokinesis. To distinguish between these two alternatives we

sporulated a heterozygous inn11–211;Dcyk3 diploid and found no

obvious genetic interaction between this specific inn1 allele and

CYK3 (supplementary material Table S2). As Cyk3p is necessary

to link Inn11–211 to Sho1p this result argues for an additional

Inn1p-independent function of Sho1p in cytokinesis (Fig. 3D).

The Nub array identified Nba1p as interaction partner of the

SH3 domains of Hof1p and Cyk3p (Tonikian et al., 2009). An

Inn1p-like scaffold function of Nba1p could however be ruled out

by our observation that the deletion of NBA1 does not impair the

Fig. 6. Genetic interactions. (A) Dhof1;Dsho1 cells carrying a plasmid expressing HOF1 and URA3 were transformed with plasmids expressing the indicated

genes under control of the PMET17 promoter. 4 ml of cultures of 1 OD600 were spotted in tenfold serial dilutions on media containing uracil, 5-FOA and the

indicated concentrations of methionine to repress (1 mM) or modestly induce (70 mM) the PMET17 promoter. (B) Cells expressing the sho11–240 allele and HOF1 in

a URA3-containing plasmid in wild-type or Dhof1 cells were grown on media supplemented with uracil or uracil and 5-FOA. Wild-type and Dhof1 cells served as

controls. (C) Wild-type and Dhof1 cells were transformed with the indicated constructs under control of the PMET17 promoter and spotted in tenfold serial dilutions

on media selecting for the presence of the plasmids and containing no methionine to fully induce the expression of the gene fusions. Cells were grown for 4 days at

34 C. (D) As in A, except that the Dhof1;Dcyk3 cells expressing HOF1 and URA3 from the same plasmid contained different fusion alleles of CYK3 under control

of the PMET17 promoter. Cells were grown on medium containing 0 or 70 mM methionine to fully or modestly induce the expression of the fusion genes. (E) Anti-

MYC immunoblot of extracts originating from cells expressing MYC-labeled CYK3 (lane 2), a mutant of CYK3 lacking the proline-rich stretch between residues

183 and 191 (lane 3) or the empty PRS315 plasmid (lane 1). (F) Visualization of protein networks as in Fig. 3D. – indicates a negative genetic interaction (GI)

between the respective alleles.

Molecular description of the HICS complex 4109

Journ

alof

Cell

Scie

nce

growth of cells expressing the inn11–211 allele (supplementary

material Table S2).

The physical connections between Cyk3p, Hof1p and

Sho1p are relevant for cytokinesis

The simultaneous deletion of HOF1 and CYK3 is lethal for the

cell (Korinek et al., 2000). To evaluate the functional

significance of the interaction of Cyk3p with SH3Sho1 and

SH3Hof1 we tested different alleles of CYK3 for their abilities to

rescue the growth of a Dhof1;Dcyk3 strain (Fig. 6D,F). A CYK3

allele lacking the SH3 domain-encoding sequence (cyk3119–885)

was not functional in this genetic background (Fig. 6D). The N-

terminal fragment of Cyk3p still harboring the SH3 domain and

the proline-rich motif (Cyk31–194–GFP) retained a residual

Cyk3p activity when overexpressed (Fig. 6D). Importantly, a

CYK3 allele (cyk3DP–MYC) where the contact site to Hof1p and

Sho1 was destroyed rescued the Dhof1;Dcyk3 strain poorly and

only upon overexpression (Fig. 6D,F). This very reduced ability

of Cyk3DP–MYC to functionally replace Cyk3p is not due to an

intrinsic instability of the protein as it is expressed at similar

levels as the wild-type protein (Fig. 6E). The experiments assign

SH3Cyk3 an active role during cytokinesis that is supported by the

C-terminal tail domain and the ability of the P1 motif to interact

with Hof1p or Sho1p (Fig. 6F). The binding partner of this C-

terminal domain, Slt2p, was already discussed to activate Chs2p

(Lesage et al., 2004; Lesage et al., 2005; Tong et al., 2004). We

thus propose that independently of Inn1p, Cyk3p might recruit

the MAP kinase Slt2p to modulate the enzymes responsible

for primary and/or secondary septum formation through

phosphorylation. The observation that a SLT2 deletion is not

tolerated in Dmyo1 but in Dchs2 cells further supports this

hypothesis (Rodriguez-Quinones and Rodriguez-Medina, 2009;

Yoshida et al., 2009).

The synthetic lethality between HOF1 and CYK3 additionally

suggests that either of the two proteins acts in separate, partially

complementing branches of cytokinesis (Tolliday et al., 2001).

Through our interaction analysis we were able to define with P1

and the C-terminal domain two further separable regions within

Cyk3p besides the N-terminal SH3 domain (Fig. 2). As none of

both Cyk3p fragments complemented the Dhof1;Dcyk3 strain we

can exclude that either of them alone contributes significantly to

this alternative pathway (Fig. 6D). Hof1p is a multi-domain

protein (Fig. 1D). We tested different fragments for their ability

to complement the function of Hof1p in the absence of Cyk3p.

Fragments missing the FCH, the F-Bar, or the region linking the

F-Bar with the SH3 domain failed to complement, whereas a

hof11–609 allele only lacking the SH3 domain fully rescued the

Dhof1;Dcyk3 double deletion (Fig. 7A). The outcome of these

experiments suggests that the SH3 domains of Cyk3p, Sho1p and

Hof1p collaborate in the same pathway, whereas the more

Fig. 7. Functional dissection of the CYK3-dependent pathway. (A) Dhof1;Dcyk3 cells carrying a plasmid expressing CYK3 and URA3 were transformed with

plasmids expressing the indicated alleles of HOF1 under control of the PMET17 promoter or an empty vector. 4 ml of cultures of 3 OD600 were spotted in tenfold

serial dilutions on media containing uracil, 5-FOA and the indicated concentrations of methionine to modestly (70 mM) or fully (0 M) induce the PMET17

promoter. Growth was monitored after 4 days. (B) Dhof1;Dcyk3;Dsho1 cells carrying a plasmid expressing HOF1 and URA3 were transformed with plasmids

expressing the indicated HOF1 alleles and analyzed as in A, except that cells were grown on medium containing 1 mM Met for 6 days. (C) As in B, except that the

indicated cells expressed low, moderate and high levels of CYK3–MYC (1 mM, 70 mM, 0 M Met). (D) Information flow within the Cyk3p-dependent pathway of

cytokinesis. Protein interactions are indicated with black, and functional activities with red arrows. Hof1p and Sho1p recruit and activate Cyk3p through their SH3

domains. Cyk3p activates Inn1p through its SH3 domain, and at the same time recruits Slt2p and potentially other factors (Z) to the site of septum formation.

Inn1p (either directly or through factor Y) and Slt2p might coordinate primary and secondary septum formation during cytokinesis. Both, Hof1p and Sho1p

perform additional functions during cytokinesis outside the Cyk3p-dependent pathway.

Journal of Cell Science 125 (17)4110

Journ

alof

Cell

Scie

nce

N-terminal domains of Hof1p are additionally involved in at least

one alternative pathway. To reveal whether Sho1p contributes tothis alternative path we repeated the experiments in a strainlacking CYK3, HOF1 and SHO1, but harboring an extra-

chromosomal copy of HOF1. Independently of the introducedalleles, the triple deletion grew less well than the correspondingDhof1;Dcyk3 double deletion (Fig. 7B). hof11–609 complementedthe wild-type allele in the triple knockout less efficiently than in

the double deletion (Fig. 7B). The reintroduction of CYK3 undercontrol of the PMET17 promoter allowed us to compare the growthof the Dhof1 with the Dhof1;Dsho1 cells under different Cyk3p

expression levels. Increasing the expression of Cyk3p partiallycompensated for the single loss of HOF1 or the simultaneousdeletions of HOF1 and SHO1. Taken together, these experiments

demonstrate a negative genetic interaction between SHO1 andCYK3 and indicate that the SH3 domains of Hof1p and Sho1poperate upstream of Cyk3p (Fig. 7C).

DiscussionThe aim of this study is to better understand the complexinterplay among the proteins of the cytokinesis machinery. Wecomplemented our protein interaction analysis by measuring

genetic interactions between alleles of members of theestablished network that lacked defined protein interaction sitesfor their binding partners. Following this experimental strategy

supported by microscopic studies we could define theosmosensor Sho1p as a novel member of the cytokinesisnetwork. Its contribution to cytokinesis depends on its C-

terminal SH3 domain that is responsible for interactions withthe proteins Hof1p, Inn1p, and Cyk3p. Sho1p is the only memberof the herein defined HICS complex that is an integral membrane

protein and thus permanently embedded in the cell membrane.Through interaction with Hof1p and Inn1p, Sho1p may present alink between the AMR and the plasma membrane duringcytokinesis (Fig. 8A). As the phenotype of a SHO1 deletion is

mild compared to the phenotype of a MYO1 deletion it is highlylikely that Sho1p acts with other proteins to fulfill this function.Potential candidates are again Hof1p and Inn1p, as both genes

display negative genetic interactions with SHO1 (supplementarymaterial Table S2).

Our data collectively suggest that the role of Sho1p incytokinesis is not linked to its function as an osmosensor. The

functional consequences of its interactions with the cytokineticproteins thus differ from its reported interaction with Fus1pduring mating. Here, the interaction interrupts the MAP-kinase

pathway of the high-osmolarity response to increase the toleranceof the cells to the conditions that occur during cell–cell fusion(Nelson et al., 2004).

By including Sho1p we can upgrade our knowledge on the

choreography of events during cytokinesis (Jendretzki et al.,2009; Lippincott and Li, 1998; Meitinger et al., 2011; Nishihamaet al., 2009). Hof1p assembles as a double ring at the bud neck of

large budded cells. In late anaphase Hof1p relocalizes to a singlering between the split septins. This event is accompanied by thephosphorylation of the central region of Hof1p by Dbf2p

(Meitinger et al., 2011). Inn1p and later Cyk3p assemble withHof1p at that central ring to from a AMR–Hof1p–Inn1p–Cyk3pcomplex (Meitinger et al., 2011; Nishihama et al., 2009). The

delivery of Sho1p into the region between the split septin ringsduring anaphase increases its effective concentration, therebyfacilitating the binding of SH3Sho1 to the proline-rich region

within the RLS of Hof1p and the proline-rich motifs of Inn1p and

Cyk3p to finally form the HICS complex (Fig. 4B, Fig. 8).

Reflecting their different roles during membrane contraction and

septum synthesis, the subunits of the HICS complex display

extensive negative genetic interactions among each other. One

may speculate that mediating different yet related activities

through one multifunctional complex might allow their extremely

precise and efficient coordination in time and space (Fig. 7D,

Fig. 8A) (Nishihama et al., 2009; Sanchez-Diaz et al., 2008).

The HICS complex contains with Cyk3p, Hof1, and Sho1p three

SH3-domain-containing proteins. Two, namely Cyk3p and Hof1p,

serve also as SH3 acceptors and all three bind with partially

overlapping specificities to the scaffold protein Inn1p (Fig. 3C).

These features increase the potential combinatorial diversity of this

interaction network through multiple protein interaction states and

might equip the HICS complex with unique properties normally

not encountered in protein complexes. For example, a 1:1:1:1

stoichiometry of its members can assemble into seven isomeric

interaction states that only differ in the positions of the edges

(Fig. 8B). This rich molecular heterogeneity might find its

counterpart in the fluctuations of the relevant parameters during

cytokinesis such as membrane–AMR distance, curvature of the

membrane, chemical composition of the membrane and of the

AMR. The HICS complex might adjust to these changes by

selecting the best fitting from its spectrum of interaction states and

accommodating varying combinations of further interacting

proteins (Fig. 3C, Fig. 8B).

Fig. 8. Models for the organization of the HICS complex. (A) The HICS

complex connects the membrane with the AMR. Interactions of Hof1p and

Inn1p with the membrane are still hypothetical and the interactions of both

with the AMR are not yet understood. Shown is a HICS complex with a

1:1:1:2 stoichiometry. Other configurations are equally likely but not shown.

(B) The seven isomeric interaction states of the HICS complex with a 1:1:1:1

stoichiometry. (C) A high concentration of Sho1p leads to Sho1p

oligomerization that condenses the isolated interaction states into a net of

proteins with a highly flexible arrangement of interactions.

Molecular description of the HICS complex 4111

Journ

alof

Cell

Scie

nce

Sho1p has the potential to form homo-oligomers(supplementary material Table S1) (Hao et al., 2007). Onceexceeding a critical concentration, these oligomers might

condense the isolated interaction states into a single net(Fig. 8C). Owing to the highly flexible arrangements ofinteractions this net might display liquid-like properties that

could define a new compartment of cell–cell separation below themembrane (Brangwynne et al., 2009; Hyman and Brangwynne,2011). The structure might then serve as attachment site for the

AMR to couple AMR contraction to the ingression of the plasmamembrane. It is even conceivable that during the final steps ofcytokinesis this conformational flexibility might allow this

structure to tether the two opposing membranes for theirfusion. The interaction of Sho1p with Sro7p, a stimulator ofexocytosis, as well as their negative genetic interaction supportsthis hypothesis (supplementary material Table S1) (Aguilar et al.,

2010).

Materials and MethodsGrowth conditions, yeast strains and genetic methods

Standard culture media and standard yeast genetic methods were used as described(Guthrie and Fink, 1991; Dunkler et al., 2012). Media for the Split–Ubiquitininteraction assay and media to select for the loss of URA3-containing plasmidscontained 1 mg/ml 5-fluoroorotic acid (5-FOA). Yeast strains JD47, JD53 andJD51 are as described previously (Dunkler et al., 2012).

Construction of Nub and Cub fusion genes and other molecularmanipulations

Construction of Nub and Cub fusion genes were essentially as described previously(Dunkler et al., 2012; Hruby et al., 2011). Insertion of the PMET17 promoter, genedeletion and epitope tagging were performed by PCR-based methods as described(Janke et al., 2004; Knop et al., 1999). The GST fusions were obtained by placingthe ORF of the respective gene or gene fragment behind the E. coli GST sequenceon the pGEX-6P-1 vector (GE Healthcare, Freiburg, Germany). The AGT fusionswere expressed from the vector pAGT-express, a pet15b derivative, where theORF of the respective gene was inserted into a multi-cloning site located betweenthe upstream 6xHIS-tag-coding sequence and the downstream AGT-codingsequence.

Mutations in CYK3 were introduced by using primers annealing at the respectivesite but carrying a mismatch to change the codon by the ensuing PCR. The mutatedsequence of Cyk3P between position 185 and 192 reads ASLASLA instead ofPPLPPLP. Descriptions of the constructs and yeast strains used in this study can befound in the supplementary material Tables S4, and S5.

Split–Ub interaction analysis

Large scale Split–Ubiquitin assays were performed as described (Dunkler et al.,2012; Hruby et al., 2011). Measuring interactions between individual Nub and Cub

fusion proteins by spotting yeast cells expressing both fusions onto 5-FOA and SDUra-containing media was essentially as described previously (Eckert andJohnsson, 2003).

In vitro binding assays

GST-fusion–Sepharose slurries were incubated for 1 h at 4 C under rotation witheither 1 ml of yeast cell extract containing the MYC-tagged proteins or with E. coliextracts containing AGT fusion proteins. Beads with bound proteins wereseparated by 2 min centrifugations at 6000 rpm and washed four times withPBS. The supernatant was discarded and 100 ml elution buffer (50 mM Tris pH 7.0with 20 mM reduced glutathione) was added, and gently mixed for 10 min at 4 C.Finally, beads were again centrifuged for 2 min at 6000 rpm and the supernatantwas transferred to a fresh tube, mixed with loading buffer [120 mM Tris-HClpH 6.8, 4% (w/v) SDS, 20% (v/v) glycerine, 0.015% (w/v) Bromophenol Blue,100 mM dithiothreitol (DTT)], boiled for 5 min, and analyzed by SDS-PAGE andimmunoblotting. Immunoblotting was done as described previously (Hruby et al.,2011).

Preparation of yeast cell extracts

For pull-down experiments, yeast cell cultures were grown to an OD600 of 1.5,pelleted, washed once in ice-cold water, and then transferred into liquid nitrogen.The cell pellets were ground in liquid nitrogen using a mortar. The cell powder wascollected in protein extraction buffer (50 mM Hepes pH 7.5, 150 mM NaCl, 1 mMEDTA) containing 16 protease inhibitor cocktail Complete (Roche Diagnostics,Mannheim, Germany), 1 mM DTT and 1 mM phenylmethylsulfonyl fluoride

(PMSF). 0.1% Triton X-100 was added, incubated for 10–20 min on ice, andextracts were clarified by centrifugation.

Expression and immobilization of AGT and GST fusion proteins

E. coli cells (BL21, Amersham, Freiburg, Germany) expressing AGT or GSTfusions were grown at 37 C to an OD600 of 0.6, and the expression of the AGT andGST fusion proteins were induced by 0.5 mM IPTG. The cells expressing the GSTfusions were shaken for additional 4 h at 37 C, harvested by centrifugation, andresuspended in PBS containing 1 protease inhibitor cocktail, 0.4 mM PMSF and 1mg/ml lysozyme (Sigma-Aldrich Chemie, Steinheim, Germany). After incubationfor 10 min at 4 C, cells were sonicated, and the lysates were clarified bycentrifugation at 13,000 rpm for 15 min at 4 C, and either stored at 280 C orimmediately incubated with PBS-equilibrated glutathione–Sepharose 4B beads(GE Healthcare, Freiburg, Germany) for 1 h at 4 C under rotation. Finally, boundmaterial was washed four times in 16 PBS.

The cells expressing AGT fusions were incubated under inducing conditions for6 h at 18 C. Extract preparation was as above but containing additional 0.5%Triton X-100 in the lysis buffer.

Electron microscopy

Exponentially growing yeast cells were arrested with a-factor to a finalconcentration of 5 mg/ml over 2–3 h. After washing out the a-factor, the cellswere released into the fresh medium and allowed to proceed through one cellcycle. Samples for electron microscopy were prepared according to Walther andZiegler (Walther and Ziegler, 2002) with some modifications. Cell pellets wereresuspended in fresh medium at 30 C and rapidly frozen on the Saphire glasssurface in a Compact 01 high-pressure freezer (M. Wohlwend GmbH, Sennwald,Switzerland). Samples were subjected for freeze substitution (with 0.2% osmiumtetroxide, 0.1% uranyl acetate, 5% water in acetone), washed with water-freeacetone, stepwise embedded in Epon, polymerized and cut with a Leica UltracutUCT ultra-microtome. Ultrathin sections (about 70 nm) were post-stained withuranyl acetate and observed with a Philips 400 transmission electron microscope atan accelerating voltage of 80 kV. Images were collected by a CCD camera andanalyzed with the EMMENU software and Photoshop. Measurements of septumthickness were performed with ImageJ software.

Microscopy, live cell imaging

Differential interference contrast (DIC) and fluorescence microscopy wereperformed using the Delta Vision System (Applied Precision, Issaquah, USA).The System is equipped with the Olympus IX71 microscope, with a 1006Plan-Apo/1.4 objective lens and the charge-coupled-device (CCD) camera(Photometrics, Munchen, Germany). Images were acquired and analyzed withthe softWoRx software of the Delta Vision System, and processed using AdobePhotoshop. Fluorescent proteins were visualized using a live cell filter set (ChromaTechnology Corp., Olching, Germany) using the EGFP (lex470–lem525) andmCherry filter (lex572–lem632), respectively.

Yeast strains were grown overnight in liquid selective media, diluted the nextday, and grown with or without methionine to mid-log phase at the indicatedtemperatures. Cells were fixed after washing once in selective media by incubationin 4% paraformaldehyde for 10 min at room temperature. Cells were washed twicein 16 PBS before analysis.

For time-lapse microscopy exponentially grown cells were diluted to a densityof 36106 cells/ml and mounted on an agarose pad. Temperature was held constantat 30 C by a Delta Vision System-supplied temperature chamber.

AcknowledgementsWe thank Rong Li for the plasmid pLP8.

FundingThe work was supported by a Deutsche Forschungsgemeinschaftresearch grant [grant number JO 187/5-1 to N.J.]; and the MarieCurie Research Training Network of the European Union [grantnumber PROSA 019475 to K.L.].

Supplementary material available online at

http://jcs.biologists.org/lookup/suppl/doi:10.1242/jcs.105320/-/DC1

ReferencesAguilar, P. S., Frohlich, F., Rehman, M., Shales, M., Ulitsky, I., Olivera-Couto, A.,

Braberg, H., Shamir, R., Walter, P., Mann, M. et al. (2010). A plasma-membraneE-MAP reveals links of the eisosome with sphingolipid metabolism and endosomaltrafficking. Nat. Struct. Mol. Biol. 17, 901-908.

Balasubramanian, M. K., Bi, E. and Glotzer, M. (2004). Comparative analysis ofcytokinesis in budding yeast, fission yeast and animal cells. Curr. Biol. 14, R806-R818.

Journal of Cell Science 125 (17)4112

Journ

alof

Cell

Scie

nce

Bender, A. and Pringle, J. R. (1989). Multicopy suppression of the cdc24 budding

defect in yeast by CDC42 and three newly identified genes including the ras-related

gene RSR1. Proc. Natl. Acad. Sci. USA 86, 9976-9980.

Bi, E. (2001). Cytokinesis in budding yeast: the relationship between actomyosin ring

function and septum formation. Cell Struct. Funct. 26, 529-537.

Bodenmiller, B., Campbell, D., Gerrits, B., Lam, H., Jovanovic, M., Picotti, P.,

Schlapbach, R. and Aebersold, R. (2008). PhosphoPep–a database of protein

phosphorylation sites in model organisms. Nat. Biotechnol. 26, 1339-1340.

Brangwynne, C. P., Eckmann, C. R., Courson, D. S., Rybarska, A., Hoege, C.,

Gharakhani, J., Julicher, F. and Hyman, A. A. (2009). Germline P granules are

liquid droplets that localize by controlled dissolution/condensation. Science 324,

1729-1732.

Dohmen, R. J., Stappen, R., McGrath, J. P., Forrova, H., Kolarov, J., Goffeau, A.

and Varshavsky, A. (1995). An essential gene encoding a homolog of ubiquitin-

activating enzyme. J. Biol. Chem. 270, 18099-18109.

Dunkler, A., Muller, J. and Johnsson, N. (2012). Detecting protein-protein interactions

with the Split-Ubiquitin sensor. Methods Mol. Biol. 786, 115-130.

Eckert, J. H. and Johnsson, N. (2003). Pex10p links the ubiquitin conjugating enzyme

Pex4p to the protein import machinery of the peroxisome. J. Cell Sci. 116, 3623-3634.

Guthrie, C. and Fink, G. R. (1991) Guide to Yeast Genetics and Molecular Biology.

San Diego, CA: Academic Press Inc.

Hao, N., Behar, M., Elston, T. C. and Dohlman, H. G. (2007). Systems biology

analysis of G protein and MAP kinase signaling in yeast. Oncogene 26, 3254-3266.

Hohmann, S. (2009). Control of high osmolarity signalling in the yeast Saccharomyces

cerevisiae. FEBS Lett. 583, 4025-4029.

Hruby, A., Zapatka, M., Heucke, S., Rieger, L., Wu, Y., Nussbaumer, U.,

Timmermann, S., Dunkler, A. and Johnsson, N. (2011). A constraint network of

interactions: protein-protein interaction analysis of the yeast type II phosphatase

Ptc1p and its adaptor protein Nbp2p. J. Cell Sci. 124, 35-46.

Hyman, A. A. and Brangwynne, C. P. (2011). Beyond stereospecificity: liquids and

mesoscale organization of cytoplasm. Dev. Cell 21, 14-16.

Itoh, T., Erdmann, K. S., Roux, A., Habermann, B., Werner, H. and De Camilli, P.

(2005). Dynamin and the actin cytoskeleton cooperatively regulate plasma membrane

invagination by BAR and F-BAR proteins. Dev. Cell 9, 791-804.

Janke, C., Magiera, M. M., Rathfelder, N., Taxis, C., Reber, S., Maekawa, H.,

Moreno-Borchart, A., Doenges, G., Schwob, E., Schiebel, E. et al. (2004). A

versatile toolbox for PCR-based tagging of yeast genes: new fluorescent proteins,

more markers and promoter substitution cassettes. Yeast 21, 947-962.

Jendretzki, A., Ciklic, I., Rodicio, R., Schmitz, H. P. and Heinisch, J. J. (2009). Cyk3

acts in actomyosin ring independent cytokinesis by recruiting Inn1 to the yeast bud

neck. Mol. Genet. Genomics 282, 437-451.

Johnsson, N. and Varshavsky, A. (1994). Split ubiquitin as a sensor of protein

interactions in vivo. Proc. Natl. Acad. Sci. USA 91, 10340-10344.

Kamei, T., Tanaka, K., Hihara, T., Umikawa, M., Imamura, H., Kikyo, M., Ozaki,

K. and Takai, Y. (1998). Interaction of Bnr1p with a novel Src homology 3 domain-

containing Hof1p. Implication in cytokinesis in Saccharomyces cerevisiae. J. Biol.

Chem. 273, 28341-28345.

Knop, M., Siegers, K., Pereira, G., Zachariae, W., Winsor, B., Nasmyth, K. and

Schiebel, E. (1999). Epitope tagging of yeast genes using a PCR-based strategy: more

tags and improved practical routines. Yeast 15, 963-972.

Korinek, W. S., Bi, E., Epp, J. A., Wang, L., Ho, J. and Chant, J. (2000). Cyk3, a

novel SH3-domain protein, affects cytokinesis in yeast. Curr. Biol. 10, 947-950.

Lesage, G., Sdicu, A. M., Menard, P., Shapiro, J., Hussein, S. and Bussey, H. (2004).

Analysis of beta-1,3-glucan assembly in Saccharomyces cerevisiae using a synthetic

interaction network and altered sensitivity to caspofungin. Genetics 167, 35-49.

Lesage, G., Shapiro, J., Specht, C. A., Sdicu, A. M., Menard, P., Hussein, S., Tong,

A. H., Boone, C. and Bussey, H. (2005). An interactional network of genes involved

in chitin synthesis in Saccharomyces cerevisiae. BMC Genet. 6, 8.

Lippincott, J. and Li, R. (1998). Dual function of Cyk2, a cdc15/PSTPIP family

protein, in regulating actomyosin ring dynamics and septin distribution. J. Cell Biol.

143, 1947-1960.

Maeda, T., Takekawa, M. and Saito, H. (1995). Activation of yeast PBS2 MAPKK by

MAPKKKs or by binding of an SH3-containing osmosensor. Science 269, 554-558.

Meitinger, F., Petrova, B., Lombardi, I. M., Bertazzi, D. T., Hub, B., Zentgraf, H.and Pereira, G. (2010). Targeted localization of Inn1, Cyk3 and Chs2 by the mitotic-exit network regulates cytokinesis in budding yeast. J. Cell Sci. 123, 1851-1861.

Meitinger, F., Boehm, M. E., Hofmann, A., Hub, B., Zentgraf, H., Lehmann, W. D.

and Pereira, G. (2011). Phosphorylation-dependent regulation of the F-BAR proteinHof1 during cytokinesis. Genes Dev. 25, 875-888.

Muller, J. and Johnsson, N. (2008) Split-ubiquitin and the split-protein sensors:chessman for the endgame. Chembiochem 9, 2029-2038.

Nelson, B., Parsons, A. B., Evangelista, M., Schaefer, K., Kennedy, K., Ritchie, S.,

Petryshen, T. L. and Boone, C. (2004). Fus1p interacts with components of theHog1p mitogen-activated protein kinase and Cdc42p morphogenesis signalingpathways to control cell fusion during yeast mating. Genetics 166, 67-77.

Nishihama, R., Schreiter, J. H., Onishi, M., Vallen, E. A., Hanna, J., Moravcevic,K., Lippincott, M. F., Han, H., Lemmon, M. A., Pringle, J. R. et al. (2009). Role ofInn1 and its interactions with Hof1 and Cyk3 in promoting cleavage furrow andseptum formation in S. cerevisiae. J. Cell Biol. 185, 995-1012.

Rao, Y., Ma, Q., Vahedi-Faridi, A., Sundborger, A., Pechstein, A., Puchkov, D.,

Luo, L., Shupliakov, O., Saenger, W. and Haucke, V. (2010). Molecular basis forSH3 domain regulation of F-BAR-mediated membrane deformation. Proc. Natl.

Acad. Sci. USA 107, 8213-8218.Rodrıguez-Quinones, J. F. and Rodrıguez-Medina, J. R. (2009). Differential gene

expression signatures for cell wall integrity found in chitin synthase II (chs2Delta)and myosin II (myo1Delta) deficient cytokinesis mutants of Saccharomycescerevisiae. BMC Res. Notes 2, 87.

Sanchez-Diaz, A., Marchesi, V., Murray, S., Jones, R., Pereira, G., Edmondson, R.,Allen, T. and Labib, K. (2008). Inn1 couples contraction of the actomyosin ring tomembrane ingression during cytokinesis in budding yeast. Nat. Cell Biol. 10, 395-406.

Schmidt, M., Bowers, B., Varma, A., Roh, D. H. and Cabib, E. (2002). In buddingyeast, contraction of the actomyosin ring and formation of the primary septum atcytokinesis depend on each other. J. Cell Sci. 115, 293-302.

Sekiya-Kawasaki, M., Abe, M., Saka, A., Watanabe, D., Kono, K., Minemura-Asakawa, M., Ishihara, S., Watanabe, T. and Ohya, Y. (2002). Dissection ofupstream regulatory components of the Rho1p effector, 1,3-beta-glucan synthase, inSaccharomyces cerevisiae. Genetics 162, 663-676.

Tatebayashi, K., Tanaka, K., Yang, H. Y., Yamamoto, K., Matsushita, Y., Tomida,

T., Imai, M. and Saito, H. (2007). Transmembrane mucins Hkr1 and Msb2 areputative osmosensors in the SHO1 branch of yeast HOG pathway. EMBO J. 26, 3521-3533.

Tolliday, N., Bouquin, N. and Li, R. (2001). Assembly and regulation of thecytokinetic apparatus in budding yeast. Curr. Opin. Microbiol. 4, 690-695.

Tong, A. H., Lesage, G., Bader, G. D., Ding, H., Xu, H., Xin, X., Young, J., Berriz,G. F., Brost, R. L., Chang, M. et al. (2004). Global mapping of the yeast geneticinteraction network. Science 303, 808-813.

Tonikian, R., Xin, X., Toret, C. P., Gfeller, D., Landgraf, C., Panni, S., Paoluzi, S.,

Castagnoli, L., Currell, B., Seshagiri, S. et al. (2009). Bayesian modeling of theyeast SH3 domain interactome predicts spatiotemporal dynamics of endocytosisproteins. PLoS Biol. 7, e1000218.

Vallen, E. A., Caviston, J. and Bi, E. (2000). Roles of Hof1p, Bni1p, Bnr1p, andmyo1p in cytokinesis in Saccharomyces cerevisiae. Mol. Biol. Cell 11, 593-611.

Vidal, M., Cusick, M. E. and Barabasi, A. L. (2011). Interactome networks and humandisease. Cell 144, 986-998.

Wach, A., Brachat, A., Pohlmann, R. and Philippsen, P. (1994). New heterologousmodules for classical or PCR-based gene disruptions in Saccharomyces cerevisiae.Yeast 10, 1793-1808.

Wach, A. (1996). PCR-synthesis of marker cassettes with long flanking homologyregions for gene disruptions in S. cerevisiae. Yeast 12, 259-265.

Wach, A., Brachat, A., Alberti-Segui, C., Rebischung, C. and Philippsen, P. (1997).Heterologous HIS3 marker and GFP reporter modules for PCR-targeting inSaccharomyces cerevisiae. Yeast 13, 1065-1075.

Walther, P. and Ziegler, A. (2002). Freeze substitution of high-pressure frozensamples: the visibility of biological membranes is improved when the substitutionmedium contains water. J. Microsc. 208, 3-10.

Yoshida, S., Bartolini, S. and Pellman, D. (2009). Mechanisms for concentrating Rho1during cytokinesis. Genes Dev. 23, 810-823.

Molecular description of the HICS complex 4113