Borg5 restricts contractility and motility in epithelial MDCK cells

Septins have emerged as versatile modulators of actin filament (F-actin) and microtubule (MT) dynamics (Lam and Calvo, 2019; Spiliotis, 2010, 2018; Spiliotis and Nakos, 2021). They exist in mammalian cells as palindromic heteromeric complexes composed of septin proteins from four sequence-homology groups (Field et al., 1996; Iv et al., 2021; Kinoshita, 2003; Kinoshita et al., 2002; Martins et al., 2023; Sellin et al., 2011, 2014; Sirajuddin et al., 2007); these building blocks can further anneal into filaments (Bridges et al., 2014; Field et al., 1996; Kinoshita et al., 2002; Woods and Gladfelter, 2021). The ability of septins to act as a modulator is thought to stem from three features: (1) septins can directly interact with F-actin and MTs (Dolat et al., 2014; Kuzmić et al., 2022; Mavrakis et al., 2014; Smith et al., 2015; Targa et al., 2019; Verdier-Pinard et al., 2017), which causes F-actin bundling and MT stabilization or growth, respectively (Bai et al., 2013; Bowen et al., 2011; Farrugia et al., 2020; Iv et al., 2021; Kuzmić et al., 2022; Nakos et al., 2022; Schmidt and Nichols, 2004); (2) septins bind membrane lipids and might provide a template for actin polymerization and possibly MT association with cellular membranes (Benoit et al., 2023; Dolat and Spiliotis, 2016; Hagiwara et al., 2011; Kinoshita et al., 1997; Martins et al., 2023); and (3) septins scaffold or compete with F-actin- and MT-modifying proteins resulting in a multitude of regulatory effects (reviewed in Spiliotis and Nakos, 2021). This diversity is enabled by the various septin paralogs interacting with modifying proteins. For example, septin 7 mediates recruitment of HDAC6 to MTs, which then become subjected to tubulin deacetylation (Ageta-Ishihara et al., 2013), whereas septin 2 competes with MAP4 for MT binding, which the authors linked, in HeLa cells, to increased MT-stability upon septin depletion (Kremer et al., 2005). Conversely, a splice variant of septin 9 (9_i1) functions as an MT-associated protein (MAP) and increases MT stability instead (Kuzmić et al., 2022). Given that septins 2 and 9_i1 coexist in the same septin complexes, their concomitant action could yield opposing effects on MT stability. In a similar apparent antagonism, septin 2 recruits myosin II and myosin-activating enzymes to F-actin (Joo et al., 2007), whereas septin 9 competes with myosin for actin binding (Smith et al., 2015). Such counteracting activities suggests that higher-order regulation defines the outcome, but such regulatory mechanisms have yet to be characterized.

Interaction of septins with the actin or MT cytoskeleton itself is regulated in vertebrates by Borg (for ‘binder of Rho GTPases’) proteins (Burbelo et al., 1999; Joberty et al., 1999, 2001), a family of five largely disordered proteins (Borg1–Borg5), which share a conserved α-helical domain that binds to the interface of septin 6 and 7 (Castro et al., 2023; Sheffield et al., 2003) as well as a Cdc42-binding Crib domain, which links Borg5-mediated septin regulation to Cdc42 activity (Farrugia and Calvo, 2017) and prompted the alternative denomination of Borgs as Cdc42 effector proteins (Cdc42EP1– Cdc42EP5). Available data suggest that Borg proteins enhance septin-mediated F-actin bundling, and promote septin-associated stress fibers and acto-myosin contractility (Calvo et al., 2015; Farrugia and Calvo, 2017; Farrugia et al., 2020; Hirsch et al., 2001). Borg2/3 depletion can even shift septin-association from F-actin to MTs (Salameh et al., 2021), supporting a view in which Borg proteins enable or facilitate F-actin associated septin functions (Farrugia and Calvo, 2016; Tomasso and Padrick, 2023). The molecular mechanisms underlying these regulations are still largely unknown. Only Borg2 has a putative actin-binding domain (Calvo et al., 2015), which could mediate enhanced septin–actin interaction; other possible mechanisms include Borg-induced changes in septin conformation, a scenario supported by observations of altered septin polymerization in the presence of Borg3 in cell-free assays (Soroor et al., 2021), and a hypothetical Borg-mediated recruitment of third-party proteins to septins.

Much of septin- and Borg-mediated F-actin regulation has been characterized in directionally migrating cells (Cohen et al., 2018; Farrugia and Calvo, 2017; Liu et al., 2014), in which the actin cytoskeleton underpins front–rear polarity which depends on three types of stress fibers: (1) transverse arcs intersecting with (2) radial stress fibers, which support the lamellae at the migratory front, and (3) ventral stress fibers providing contractility at the rear (Vallenius, 2013). Like directionally migrating cells, epithelial cells exhibit front-rear polarity when in an undifferentiated, mesenchymal state, that is during development, during mesenchymal transition in disease and during wound healing. Epithelial differentiation is associated with a dramatic switch from a motile or migratory phenotype to a stationary phenotype in which cells are isotropic, and largely lack stress fibers and protrusive lamella, but feature a circumferential actin belt linked to E-cadherin at zonula adherens (ZA) as well as actin filaments parallel to the lateral cortex (Lamouille et al., 2014; Morris and Machesky, 2015; Nelson, 2009; Scarpa et al., 2015; Thiery, 2002). The latter govern ‘columnarization’, an increase in cell height (Cavey et al., 2008; Kalaji et al., 2012). The role of septins in actin reorganization from a mesenchymal to epithelial pattern is not well understood. During epithelial differentiation, septins have been primarily linked to having a role in the MT-based machinery for the polarized delivery of exocytic membrane cargo, which is crucial for the establishment of distinct surface domain (Bowen et al., 2011; Spiliotis et al., 2008), and in supporting the actin association of cell–cell junctional complexes (Sidhaye et al., 2011; Wang et al., 2021). Likewise, despite evidence that Borg proteins are frequently mis-expressed in epithelial-derived cancers (reviewed in Tomasso and Padrick, 2023), and that Borg5 mutations are cancer drivers in oral squamous carcinoma (Campbell et al., 2021), Borg proteins have not yet been attributed any role in the establishment of the epithelial cytoskeletal networks. Indeed, unlike in other cultured cells, overexpression of Borg proteins in the kidney-derived epithelial model cell line MDCK fails to cause any obvious morphological changes or polarity defects (Hirsch et al., 2001). This prompted us to investigate how loss of function of endogenous Borg protein(s) in MDCK cells affects septin-dependent cytoskeleton organization and epithelial morphology. Given that available protein expression data suggested that MDCK cells express predominantly Borg5 (Cdc42EP1) (Harwood et al., 2023), which was also detected in a separate proximity ligation assay in the vicinity of apical junctional complexes in MDCKs (Tan et al., 2020), we focused our study on Borg5.

We found that Borg5 depletion prevented the acquisition of mature epithelial morphology and kept MDCK cells highly motile. In contrast to previously reported Borg functions, Borg5 in MDCK cells did not promote but instead limited stress fiber formation and acto-myosin contractility, and curbed MT-dependent lamella associated with mesenchymal migration. Identification of a large number of putative Borg5 interaction partners indicates that Borg5 has the ability to expand and/or modify the repertoire of septin-associated proteins.

Immunofluorescence (IF) analysis (Fig. 1), the specificity of which we validated in Borg5-depleted MDCK cells stably expressing a dox-inducible Borg5 shRNAmir (immunoblot Fig. 1F; IF, compare −dox in Fig. 1A,C,E to +dox in Fig. S1A–C), revealed that Borg5 levels and distribution changed with density. In single cells, Borg5 localized in patches with septin 2 and stress fibers under the nucleus (Fig. 1A, basal, blue arrowhead) and where F-actin demarcated the lamellipodia-cell body interface (Fig. 1A, basal, pink arrowhead) or at F-actin bundles at the cell cortex (Fig. 1A′, pink arrowhead). Borg5 was also present at the free cortex above the attachment plane (Fig. 1A, mid). The Borg5 population under the nucleus decreased with confluency (Fig. 1B,C, compare basal plane images). In contacting subconfluent cells Borg5 also prominently localized along the lateral domain (Fig. 1B, mid, arrowhead) but was excluded from ZO1-labeled tight junctions (TJs; ZO1 is also known as TJP1) (Fig. 1D). With increasing polarization in mature monolayers, overall Borg5 levels decreased (Fig. 1F, compare −dox, M and H) and Borg5 accumulated at the apical domain (Fig. 1C, apical), and, when cells were grown on permeable filter substrates, at the ZA (Fig. 1E).

To analyze the contribution of Borg5 to MDCK polarization, we generated knockdown (KD)-rescue pools of MDCK cells, in which the inducible shRNAmir was co-expressed with Myc-tagged RNAi-resistant mouse Borg5 (Borg5KD+WT). We established by immunoblotting, that the recombinant protein replaced endogenous Borg5 and that it was expressed above endogenous levels (Fig. 2A). IF analysis revealed that, even when overexpressed, Borg5–Myc localized like endogenous Borg5 to subnuclear stress fibers (Fig. 2B,D) and to lateral and apical domains, although the apical population was increased at expense of the lateral one (Fig. 2C, z-views). Overexpressed Borg5 caused recruitment of septin 2 to the apex (Fig. 2C, z-views). The higher Borg5 intensity under the nucleus also increased septin levels there (Fig. 2B), although this increase was not observed in mature monolayers [Fig. 2D, compare Sept 2 at the basal plane in cells expressing and not expressing (marked by asterisks) Borg5–Myc] and did not reflect an increase in total septin 2 levels (Fig. 2A, compare −dox and +dox). Importantly, increased Borg5 levels at the basal cortex did not translate into increased stress fiber formation (Fig. 2B,D, compare F-actin in cells with high and low Borg5 levels) and plotting Borg5 against stress fiber intensity at the basal cortex yielded no correlation between both (Fig. 2E). This suggests that recombinant Borg5 does not, as reported for other Borg proteins, promote septin-dependent stress fiber formation.

The enrichment of Borg5 at several different subcellular domains hinted at multiple functions. To dissect them, we first investigated Borg5KD phenotypes in single cells 3 h after seeding on collagen-coated coverslips.

MDCK morphology upon attachment results, as in other cell types (Even-Ram et al., 2007; Sato et al., 2020), from the outcome of two antagonistic processes: (1) Rho-dependent acto-myosin II-activity, which promotes circumferential actin bundles and restricts spreading; it dominates when MTs are disrupted by nocodazole treatment [Fig. 3B, Ctrl (−dox), nocodazole], and (2) MTs targeting the cortex and activating Rac1, which causes formation of large lamellae as well as MT-dependent neurite-like extensions; it dominates when myosin II is inhibited by blebbistatin [Fig. 3B, Ctrl (−dox), blebbistatin] (Even-Ram et al., 2007; Rafiq et al., 2019; Sato et al., 2020). In untreated Borg5KD −dox controls, most cells featured small lamellae and there were few cell extensions (Fig. 3A, −dox, DMSO; quantification in Fig. 3F,G, KD −dox). Borg5-depleted cells, by contrast, were more likely to form large lamellae and extensions like blebbistatin-treated cells (Fig. 3A, +dox DMSO; quantification in Fig. 3F,G, KD +dox), suggesting that MT-activities prevailed. Because septins can support either F-actin or MT organization, we hypothesized that Borg5 depletion altered the allocation of septins to the two cytoskeletal systems. Indeed, whereas septin 2 was loosely aligned with cortical and internal F-actin in control cells (Fig. 3D, −dox, Fig. 1A,A′; Fig. S2C), septin populations distinct from F-actin organization were apparent in Borg5KD cells (Fig. 3D, +dox; Fig. S2D,E); they either radially emanated from the cell center (Fig. 3D, KD+dox; Fig. S2D, blue arrowheads) or concentrated in neurite-like cell extensions (Fig. 3D, KD+dox, yellow arrowhead). We quantified this phenomenon in Fig. S2D′ by scoring the septin organization in cells as being either aligned with or distinct from the trajectories of main F-actin bundles. A quantitative pixel-by-pixel colocalization analysis proved less meaningful in this case, because septins only approximate rather than precisely colocalize with a subset of F-actin fibers (Martins et al., 2023). Colabeling for septin 2 with tubulin and F-actin, or of septin 9 with acetylated tubulin and F-actin, furthermore indicated that in Borg5KD cells, septin filaments aligned with MTs rather than F-actin when both cytoskeletal systems were present in the same plane (Fig. S2E,D, blue arrowheads), whereas in control cells septins appeared largely distinct from MT trajectories (Fig. S2B,C).

To determine whether Borg5-dependent differences in MT–septin alignment account for the observed differences in morphology and F-actin organization, we compared control and Borg5KD cells under conditions of nocodazole-mediated MT disruption. Nocodazole treatment during spreading resulted in robust septin alignment with F-actin in both control and Borg5KD cells (Fig. 3E, ctrl −dox, KD+dox), supporting the notion that MTs compete with F-actin for septin binding. Borg5-depleted cells nevertheless failed to organize the cortical F-actin bundles characteristic of control cells; they instead frequently accumulated parallel or stellate stress fibers in the cell center, linked to robust vinculin-positive focal adhesions (Fig. 3E, KD+dox; quantified in Fig. 3H,I). This indicates that Borg5-mediated suppression of MT activities alone does not account for the Borg5KD phenotype and suggests that Borg5 has additional MT-independent roles in F-actin organization.

Co-expression of the Borg5 shRNAmir with WT Borg5 prevented increased lamella formation and outgrowth of neurite-like extensions (Fig. 3C,F,G, KD+WT), the aberrant stellate stress fibers after nocodazole treatment (Fig. 3E, KD+WT) and increased septin–MT alignment (Fig. S2A,A′), indicating that the phenotypes were indeed due to Borg5 depletion and furthermore revealing that MDCK morphology in single cells is not altered by elevated Borg5 levels.

Taken together, in single cells, Borg5 restricts MT-dependent lamellae and promotes the establishment of cortical actin bundles as opposed to a cell center-based actin network.

Aberrant stellate stress fibers also characterized Borg5-depleted cells when they were plated at confluency (Fig. 4A). Like the situation in single cells, MT disruption did not abolish the stress fiber phenotype (Fig. 4A, nocodazole); the heightened cell isotropy in nocodazole-treated cells (see Fig. 6B,C) clarified, however, that the stellate stress fiber hub localized to the cell center under the nucleus, with the emanating stress fibers ending in focal adhesions, which aligned between neighboring cells at the periphery. Active myosin II, as measured with an antibody to phosphorylated myosin light chain (ppMLC; herein referring to MLC2 phosphorylated at Thr18 and Ser19) was enriched at the stress fiber hub (Fig. S3A,B), and overall ppMLC intensity at the basal domain was greatly increased compared to that in controls (Fig. S3C,D). Acto-myosin asters resembling this stellate organization can form in vitro through myosin II-mediated F-actin polarity sorting whereby unidirectional bipolar myosin filaments gather F-actin filaments at their plus ends (Köster et al., 2016; Miller et al., 2018). In MDCK (Nakano et al., 1999) and HeLa cells (Ishizaki et al., 1997), stellate stress fiber organization with strong FAs, as seen in Borg5KD cells, was triggered by expression of an activated form of Rho kinase 1 (ROCK1). We found that Rho kinase (ROCK; herein referring to both ROCK1 and ROCK2) but not myosin light chain kinase (MLCK) inhibitors abolished the stress fibers in Borg5KD cells (Fig. S3C,D). Therefore, Borg5 depletion likely exaggerates cortical stress fiber formation under the nucleus, which is known to involve myosin II (Lehtimäki et al., 2021) by causing excessive ROCK-mediated myosin II activation at this location.

In addition to the stress fibers, Borg5 depletion caused large peripheral ruffles, as revealed by labeling for α-actinin 4 (ACTN4), an actin-crosslinking protein (Fig. 4F). Fluorescent timelapse imaging of actin dynamics with GFP–tagged α-ACTN4 (Movie 1) showed that together this actin organization yielded highly contractile cells in which cell-spanning stress fibers were mechanically linked to the large, dynamic ruffles [Movie 1, KD (+dox)]. In control cells, parallel stress fibers aligning with the long cell axis appeared mechanically unconnected to much smaller ruffles [Movie 1, ctrl (−dox)].

Altered F-actin organization at the level of the basal cortex in Borg5-depleted cells was accompanied by two morphological changes at cell–cell contacting domains. First, by a lack of circumferential cortical F-actin above the plane of attachment (Fig. 4A, compare F-actin in −dox and +dox at 0.5–1.5 µm above the basal domain). Cortical F-actin bundles, which are re-organized from peripheral stress fibers upon cell-cell adhesion, give raise to the MDCK lateral cortex (Rajakylä et al., 2020), which in turn triggers compaction (Adams et al., 1998), an increase in cell height with concomitant decrease in cell area. We determined that when grown as cell islands, Borg5KD cells had indeed a shorter lateral domain (Fig. 4C, height, KD±dox) but a larger footprint (Fig. 4B,C, area/nuclei, KD±dox) than the corresponding control cells. If and how the reduced cortical actin density contributes to the compaction phenotype remains to be established.

Second, Borg5KD cells exhibited higher E-cadherin tension, which we determined using an established FRET-based tension sensor (Borghi et al., 2012) (Fig. 5A,B). This result is consistent with the notion that stress fiber contractility in epithelial cells is counterbalanced by cell–cell adhesion tension (Maruthamuthu et al., 2011). The tension increase upon Borg5 depletion resulted in a modest but significant decrease in cell–cell adhesion (Fig. 5C,D). We deduced the cell–cell adhesion strength from the size of cell aggregates that formed in the absence of cell–matrix adhesion after cells were subjected to defined trituration (Elbert et al., 2006; Foty, 2011). Occasional disruption of cell–cell adhesion combined with migration in confluent Borg5KD monolayers yielded a streaming phenotype (Loza et al., 2016; Sadati et al., 2013) in which groups of highly stretched cells moved in unison, resulting in ‘swirls’ (Fig. 5E,F, KD+dox; Movie 2); in control monolayers of similar density streaming activity was muted by comparison (Fig. 5E,F, KD−dox; Movie 2). The level of cell stretching, an indicator of anisotropic forces in directionally migrating cells, was determined as the ratio of major to minor axis length in fixed cells (Fig. 6B–D); this analysis confirmed reduced cell isotropy (Fig. 6C, compare −dox and +dox in DMSO) and a trend to higher variance in cell shapes (Fig. 6D, compare −dox and +dox in DMSO) in dox-induced Borg5KD cells than in their −dox controls.

Because Borg5 depletion in single cells promoted MT-dependent cell shape changes, we interrogated the role of MTs in the observed streaming phenotype by utilizing nocodazole. Nocodazole treatment reduced streaming (Fig. 6A; Movie 3, compare DMSO and NZ) as well as cell stretching (Fig. 6B,C compare +dox DMSO versus NZ) and the variation in cell shape (Fig. 6D, compare +dox DMSO versus NZ), indicating that these features depend on MT-supported protrusive activities. Nocodazole as well as Rac1 depletion also decreased the ACTN4-positive protrusions (Fig. 4F,H, compare +dox DMSO versus NZ; Fig. S4), indicating that ruffle formation is contingent on MT-dependent Rac1 activation.

MT analysis showed that there were stabilized (i.e. acetylated MTs) at the basal domain in Borg5KD cells that were not seen in controls (Fig. 4D), which was reflected in higher basal acetylated tubulin fluorescence intensity (Fig. 4E, KD±dox). Segments of these MTs aligned with septin 9-positive filaments (Fig. 4D, zoom1, 2), which in Borg5-depleted cells showed little co-occurrence with F-actin. Given that septins can stabilize MTs (Bai et al., 2013; Kuzmić et al., 2022), Borg5 depletion might contribute to MT stability by allowing more extensive septin–MT interactions.

Taken together, Borg5 depletion prior to MDCK monolayer establishment induces changes in F-actin and MT organization that yield highly contractile cells that exhibit migratory behavior in confined spaces.

Co-expression of WT Borg5 with the shRNAmir-construct rescued all Borg5KD phenotypes described above. Thus, WT Borg5 prevented increased acetylated tubulin IF intensity at the basal domain (Fig. 4E, KD+WT ±dox; Fig. S5D), the stellate stress fiber phenotype (Fig. S5A), compromised cell compaction (Fig. 4C, KD+WT±dox; Fig. S5B) and the extensive ACTN4-positive protrusions (Fig. 4G, KD+WT±dox; Fig. S5C). WT Borg5 expression also prevented the tension increase on E-cadherin adhesions (Fig. 5B, KD+WT) and the reduced cell–cell coherence (Fig. 5C,D, KD+WT) observed in Borg5KD monolayers. This indicates not only specificity of the shRNA but also that, unlike what occurs on reduction of Borg5 levels, overexpression of Borg5 had no adverse effect on these aspects of MDCK morphology.

To assess whether Borg5 plays similar roles in other cell types, we analyzed the effects of Borg5 depletion in HeLa cells, where ROCK1 activation, like in MDCK cells, induces stellate stress fibers (Ishizaki et al., 1997). We found that as in MDCK cells, HeLa cell Borg5 colocalized with septin 2 on stress fibers under the nucleus (Fig. S6A) and that the F-actin population there increased upon Borg5 depletion (Fig. S6A,D, yellow arrowheads; nuclear area fraction covered by F-actin quantified in Fig. S6B). Also, similar to what occurred in nonadherent MDCK cells (compare to Fig. 3E), Borg5-depleted HeLa cells frequently lacked compressed circumferential F-actin bundles but instead presented with cortex-parallel F-actin stacks (Fig. S6A,D, blue arrowheads, phenotype frequency quantified in Fig. S6C). By contrast, Borg5 depletion did not cause the septin9–MT alignment (Fig. S6D) or the formation of lamellae or neurite extensions, as seen in MDCK cells, suggesting that the actin, but not the MT-dependent phenotypes are conserved in HeLa cells.

When Borg5 depletion was induced after the establishment of polarized mature monolayers, the most conspicuous phenotype were distortions of the apical surface and TJs. These included a high variation in the shape and surface area of apical domains (Fig. 7A–C, KD −dox versus +dox) and frequent fingerlike undulations of ZO1-labeled TJs [Fig. 7A,B, ctrl versus KD; Fig. 7B′, quantified as ratio of perimeters at TJs and adherens junctions (AJs)]. These were accompanied by ‘sunken’ apices in which the top of the apical domain was at level with tight junctions rather than dome-like (Fig. 7A, ctrl versus KD; Fig. 7A′, quantified as distance from apex to TJs). TJ tortuosity results from a difference in forces that neighboring cells exert on shared TJs (Lynn et al., 2020). Myosin II inhibition with blebbistatin eliminated the exaggerated TJ tortuosity (Fig. 7B,B′, compare KD with and without blebbistatin) and restored a dome-like appearance to the apical domain (Fig. 7A,A′, compare KD with and without blebbistatin), indicating that high myosin contractility of the mechanically coupled TJs and apical domains caused both phenomena. The apparent apical force imbalances are unrelated to stress fiber formation as we observed a similar phenomenon when cells were plated on a soft matrix (collagen-functionalized polyacrylamide gels of 2 kPa) where stress fiber formation did not occur (Fig. 7C, 2 kPa PAA). It manifested as high variability in apical cell shape as measured by the circularity index (Fig. 7C′).

Rescue experiments with recombinant Borg5 revealed that elevated Borg5 likewise distorted the cell apex (Fig. 7D). High apical Borg5 levels were associated with small apical surfaces in comparison to neighbors expressing lower levels of Borg5 (Fig. 7E, blue versus pink arrowheads) without distorting cell shape along the basolateral aspect of the cells (Fig. 7E, blue versus pink asterisks), resulting in different apical/basal surface ratios for high and low/no Borg5–Myc-expressing cells in the same monolayer (Fig. 7E′). Borg5 and co-recruited Septin2 could be seen localized around F-actin fibers spanning across the apical surface, often emanating from the apical median, and apparently pulling on the apical cortex (Fig. 7F). These unusual apical F-actin fibers are likely the direct result of the exaggerated Borg5 and septin levels.

Thus, F-actin organization at the basal domain and restriction of cell–cell adhesion tension requires a minimal level of Borg5 and is insensitive to Borg5 overexpression, whereas apical tension homeostasis is sensitive to both reduced and elevated Borg5 levels.

Borg proteins are thought to promote septin interactions with F-actin, which in turn facilitates actin polymerization and/or bundling (Lam and Calvo, 2019; Tomasso and Padrick, 2023). Our findings suggest a different function for Borg5 in MDCK cells as Borg5 overexpression did not increase basal stress fiber levels; instead, stress fiber formation at sites of Borg–septin colocalization was induced upon Borg5 depletion. This prompted us to investigate whether Borg5 has functional partners that might inhibit or modulate the effects of septins on F-actin. To that end, we performed mass spectrometry analysis of proteins that specifically co-isolated with HALO-tagged Borg5 from MDCK Tx100 lysates (Fig. 8A). The results from three independent experiments revealed several highly abundant proteins and a larger number of less abundant putative binding partners, isolated with fewer peptide spectral matches (#PSM) (Table S1). Septins (septin 9>7>2 in order of #PSM) were among the low abundant interactors (Fig. 8B); thus, the robustness of Borg5 co-isolation might not correlate with the extent of the interactions in vivo. We speculate that because Borg5 is predicted to be largely disordered, pH and ionic strength of the lysis buffer and exposure to proteins that Borg5 does not see in intact cells, might all influence the in vitro Borg5 interactome.

Among the putative Borg5 interactors we selected myosin IIA (MyoIIA, herein referring to the MYH9 heavy chain) for further investigation because myosin activity was associated with each of our identified Borg5-regulated processes and because septins have been reported to bind and regulate myosin II function.

We established, using recombinant proteins in cell-free assays, that His–Borg5 interacted with the GST-tagged rod domain of MyoIIA (MyoIIA-Rod) directly (Fig. 8C,D, compare lanes 1 and 2, in Coomassie and His-IB; red asterisks indicate migration of Borg5). Given that the MyoIIA rod domain is also a reported septin-binding partner (Joo et al., 2007), we further investigated whether septins modified the MyoIIA-Rod interaction with Borg5. To that end, we first confirmed that recombinant His-tagged Borg5 bound to immobilized septin 2–septin 6–septin 7 (hereafter septin 2-6-7) or septin 9–septin 6–septin 7–septin 2 (hereafter septin 9-6-7-2) complexes (Fig. 8F). These complexes were prepared according to Castro-Linares et al., (2022) by bacterial co-expression of 6×His-tagged septin 2 with either untagged septin 6 and Strep-tagged septin 7, or with septin 6, septin 7 and Strep-tagged Septin 9, and their sequential purification first on nickel followed by Streptag resin (Fig. 8E). We then included septin 9-6-7-2 complexes in three-way binding assays. Pre-incubation of Borg5 with septin complexes reduced the recruitment of recombinant His–Borg5 to immobilized GST–MyoIIA-Rod (Fig. 8C,D, asterisk, compare lanes 2 and 4 in Coomassie gels and His blots, respectively). Conversely, pre-incubation of His-tagged MyoIIA-Rod with His–Borg5 diminished Borg5 binding to immobilized septin complexes (Fig. 8G, compare Borg5 in lanes 1 and 3 in eluted fractions). Collectively these findings indicate that septins and MyoIIA-Rod hinder the interactions of each other with Borg5. Septin or MyoIIA-Rod interaction with His–Borg5 also reduced the tendency of Borg5 to appear in low-speed (14,000 g) pellets (Fig. 8H, compare Borg5 in lane 1 to that in lanes 4–6), likely by stabilizing a conformation that prevents Borg5 aggregation or multimerization. Although we were unable to determine whether Borg5 in turn interfered with the reported septin–MyoII interaction, because under the salt conditions (300 mM) required to minimize His–Borg5 aggregation, MyoIIA-Rod binding to septins was weak (Fig. 5D, note lack of visible His-Sept2 (black arrow) in lanes 3,4 of His-IB) and the effect of Borg5 on MyoIIA-Rod–septin interaction is too inconsistent to allow a conclusion, we speculate that Borg5 finetunes the reported septin-mediated myosin II activation. As Borg5 depletion in contacting cells increased stress fiber-associated ROCK-dependent myosin activity (Fig. S2B), Borg5 might competitively interfere with myosin–septin binding directly or with the reported recruitment of MyoIIA-activating proteins, including ROCK, to septin scaffolds (Joo et al., 2007).

Our findings reveal that, in polarized MDCK cell monolayers, Borg5 prevents excessive contractility, surface anisotropy and migration. We attribute this role to the combined effects of three separable functions, as below.

First, Borg5 depletion induces formations of a septin population that is distinct from F-actin trajectories and which overlapped with MTs. Although in polarized cells, few MTs run parallel to the basal domain, we observed that Borg5KD cells had an increased basal population of acetylated tubulin, segments of which colocalized with septins; this population of stable MTs could arise by MT plus end capture at the cortex, as plus-end capture stabilizes MTs (Kaverina et al., 1998). Based on these findings, we propose a model in which Borg5 restricts the reported septin-dependent MT plus end targeting into the cortex (Bowen et al., 2011) where MTs, through their known role in sequestering the myosin II-activating GEF-H1, antagonize the formation of circumferential ventral stress fibers and promote Rac1-mediated lamellae (Even-Ram et al., 2007; Rafiq et al., 2019; Seetharaman and Etienne-Manneville, 2019). Borg5 depletion in MDCK cells might thus have a similar effect – promoting septin association with MTs rather than with F-actin – as has been associated with the downregulation of Borg2 and Borg3 upon Taxol-mediated MT stabilization (Salameh et al., 2021).

Second, independent of its effect on MTs, Borg5 depletion causes the formation of ROCK-dependent stellate stress fibers with a myosin II-enriched hub under the nucleus and focal adhesion anchorages at the cell periphery. Acto-myosin asters resembling this organization form transiently during developmental processes in vivo (Blanchard et al., 2010; Munro et al., 2004; Vasquez et al., 2014) where they have been associated with cortical contractile myosin pulses (Martin, 2020). Given that Borg5 depletion yielded stellate stress fibers that were stable, it might be worthwhile investigating whether septins and Borg proteins contribute to their oscillatory behavior in vertebrate morphogenesis. Myosin accumulation in the center of asters in Borg5-depleted MDCK cells might deplete myosin elsewhere and prevent myosin II-dependent F-actin remodeling into lateral cortical F-actin bundles, thereby exacerbating the MT-dependent myosin II inhibition at the periphery discussed above. Lack of peripheral F-actin bundles in turn might favor lamellae and inhibit the development of the lateral domain and cell height.

Finally, Borg 5 also regulated myosin-dependent contractility at the apical domain where Borg5 was enriched in compacted MDCK monolayers. Here, Borg5 was crucial for the equalized apical force balance that ensures isotropic shape and size of the TJ-delineated apical domains of cells within the monolayer. Apical forces are transmitted between cells via contractile F-actin associated with their apical junctional complexes (AJCs) (Citi, 2019). These complexes are in turn mechanically coupled to the cytoskeleton forming the apical cortex, which in MDCK cells consists of an isotropic interconnected network of actin and myosin filaments (Klingner et al., 2014). Apical shape changes can be due to altered forces at TJs, as has been shown experimentally by altering their composition (Odenwald et al., 2018; Tokuda et al., 2014; Brückner and Janshoff, 2018; Hatte et al., 2018; Martínez-Ara et al., 2022; Rouaud et al., 2023), or it can be triggered by changes in contractility of the apical cortex itself, as during apical constriction in gastrulation (Martin, 2020; Vasquez et al., 2014). Given that Borg5 does not localize to TJs but to the apical cortex, it likely regulates contractility there. We expect the underlying mechanisms of action to be different from those operating at the basal domain, because increased apical Borg5 levels upon overexpression caused de novo formation of apical radial F-actin fibers, whereas increased Borg5 levels at ventral stress fibers did not alter stress fibers there.

Beyond differences in F-actin regulation at apical and basal domains of MDCK cells, the net effect of Borg5 loss of function also differs between cell types. Thus, we found that in HeLa cells, Borg5 depletion promoted subnuclear stress fibers but not the MT-dependent lamellae; furthermore, an inhibition of directional vascular endothelial cell migration reported in Borg5 knockout mice was associated with an apparent decrease, rather than increase, in cellular myosin II activity (Liu et al., 2014). Among cell type-specific differences expected to modify phenotypic outcomes are: (1) the expression profile of other Borg family proteins, which, in the absence of Borg5, might execute alternative septin-mediated cytoskeletal organization; (ii) the abundance of the Septin 9_i1 splice variant, which mediates septin alignment with MTs in the absence of Borg proteins (Kuzmić et al., 2022); and (iii) the constellation of available Borg5-binding partners in each cellular context. Borg5 is predicted to be highly disordered, which likely explains the large number of interaction partners we co-isolated with Borg5 from cell lysates and which poises Borg5 to a engage in a wide range of low-affinity interactions. This in turn is a feature well suited to either ‘toning’ or implementing several alternative scaffolding functions of septins.

Given the central place myosin II occupies in Borg5-dependent F-actin organization, we believe that the direct interaction of Borg5 with the septin 2-binding rod-domain of myosin IIA and the competition of septins for this interaction, which we established in cell-free assays, is a crucial part of the cellular role of Borg5. Septin 2 has been shown to recruit myosin IIA to actin filaments and might scaffold ROCK proteins to mediate myosin II activation (Joo et al., 2007). Competition for Borg5 binding could release myosin IIA from this scaffold, thereby limiting septin-mediated myosin II activation, a scenario consistent with increased myosin II activity in Borg5-depleted MDCK cells. However, our Borg5 interactome also included several RhoGEFs (including ARHGEF2) as well as the actin-crosslinking proteins α-actinin 1, α-actinin 4 and filamin A, which if recruited to septin-aligned F-actin filaments, could stimulate Rho-ROCK-mediated myosin activation or F-actin bundling, respectively, a scenario more consistent with previously reported Borg functions in other contexts. The challenge ahead is thus to characterize unique Borg5-containing protein complexes that are associated with specific phenotypes.

MDCK and HeLa cells were obtained from Enrique Rodriguez-Boulan (Weill Cornell Medical College, New York, USA) and maintained below confluence in DMEM (Corning #10-013-CV) with 10% fetal calf serum at 37°C in a 5% CO₂ humidified atmosphere. Cells were yearly tested for mycoplasm but were not independently authenticated. For experimental analyses, cells were cultured on plastic (for biochemical experiments), MatTek glass bottom dishes (P35G-1.5-14-C; MatTek Corporation) (for time lapse imaging), or on glass coverslips (CLS-1760-012; Chemglass), 0.4 µm pore Clear Polyester Transwells (#3460, Corning) or on polyacrylamide hydrogels (for IF analysis). Induction of the dox-inducible expression cassettes in MDCK cells occurred by pre-treatment with 1 µg/ml dox (doxycycline hyclate; Sigma Millipore #D-09891) for 48 h prior to re-plating at a density of either 10⁵ or 3×10⁵ cells/cm² and continuously cultured in the presence of 1 µg/ml dox for 24 h. For single-cell analysis, cells were re-plated after 72 h of induction at 2.5×10³ cells/cm² for 3 h.

For canine Borg5 KD, two target sequences were used [5′-CGGTGAGCAAGTTCACCTTTGA-3′ (#1) and 5′-CCGTGACCGAGACCATGATAGT-3′ (#2)]; each yielded a 70–80% reduction in Borg5 protein levels when expressed as shRNAmirs in the dox-inducible pSLIK lentivirus expression system (Addgene #25737; Shin et al., 2006). We utilized #1 for all experiments presented but observed similar morphological changes with #2. All recombinant Borg5 constructs were based on mouse Borg5–Myc cDNA as used previously (Vong et al., 2010) and obtained from the authors. We co-expressed Borg5–Myc together with the shRNAmir #1 from a single dox-dependent transcript in the pEN_TTmiRc2 vector of the pSLIK lentivirus expression system. We transduced MDCK cells with lentiviruses generated with the corresponding pSLIK-Hygro plasmids and selected pools and clones of Hygromycin-resistant cells. His–Borg and HALO-tagged WT Borg5 were generated by PCR cloning into vectors pet28a+ (Novagen) and pFC14K (Promega), respectively. In the pFC14K vector, the C-terminal HALO tag is separated from Borg5 by a TEV protease cleavage site.

HeLa cells were reverse transfected with a combination of two Borg5 siRNAs (Qiagen) using RNAiMax (Invitrogen) according to the manufacturer's instructions; after 48 h cells, were re-plated onto coverslips coated with 10 µg/cm² rat tail collagen I (Corning, # 354236) at a density of 20,000 cells/cm² and cultured overnight. The Borg5 target sequences were: 5′-CACGGACGGCCACTCCAGCTA-3′ and 5′-TTCTCTGCGCTTGAACATCTA-3′.

cDNAs of EcadTSMod TFP/YFP (Borghi et al., 2012), provided by Nicolas Borghi (Institute Monod, Paris, France), MyoIIA-pEGFPC3, provided by Anne Bresnick, Albert-Einstein College of Medicine, Bronx, USA, ACTN4-pEGFPN1 (Shao et al., 2010), provided by Alan Wells, University of Pittsburgh, USA were introduced by transient transfection with AMAXA nucleofector II (Lonza) and analyzed 24 h upon transfection. Rac1 siRNA (sequence, 5′-UUUACCUACAGCUCCGUCUCCCACC-3′, custom synthesized by Sigma Millipore) is based on a previously published target sequence (Nakahara et al., 2015) and was introduced by reverse transfection with RNAiMax (Invitrogen) according to the manufacturer's instructions. Allstars negative control (Qiagen #1027281) was used as RNAi control. Transfected cells were dox-induced and cultured for 48 h before being re-plated at the desired cell density for an additional 24 h prior to analysis.

ML-7, Y27632, blebbistatin and nocodazole (Cayman Biologicals) were added from freshly made 1000× stocks in DMSO.

Fabrication of 2 kPa collagen-functionalized polyacrylamide (PAA) gels was undertaken according to the detailed protocol previously provided (Tse and Engler, 2010). Briefly, gels were polymerized on amino-silanated 12 mm circular coverslip(s) by lowering them onto 25 µl drops of a PAA mix placed on chloro-silanated glass slide(s). For an estimated elasticity modulus of 2 kPa, gels of 4% acrylamide and 0.1% Bis-acrylamide were polymerized with a 1/1000 volume of each tetramethylethylenediamine (TEMED) and 10% (w/v) ammonium persulfate (APS). Amino-silanation of the coverslips was by sequentially covering them with 0.1 M NaOH, which was heat evaporated, and reacting with 3-aminopropyltriethoxysilane (APES) and, after extensive rinses, crosslinking with 0.5% glutaraldehyde. Chloro-silanated glass slide(s) were prepared by spreading them with dichlorodimethylsilane (DCDMS), wiping off the excess coating and rinsing them with double-distilled (dd)H₂O. Collagen I-functionalization of the PAA gels occurred by UV crosslinking 0.2 mg/ml sufosuccinimidyl-6-(4′-azido-2′-nitrophenylamino)-hexanoate (sulfo-SANPAH; Pierce Biotechnology) with a UV 360 nm lamp at a distance of 3 inches for 10 min onto the PAA and subsequently reacting the crosslinker with 20 µg/cm² rat tail collagen I (Corning, #354236) in 50 mM HEPES, pH 8.5 overnight at 37°C. Coverslips were sterilized by UV irradiation at 254 nm in a UV-Stratalinker 2400 prior to use.

For the cell dissociation assay, 10⁴ cells (Borg5KD in −dox or precultured in +dox) were cultured overnight in DMEM with 0.1% serum (to prevent proliferation) in polyHEMA coated 96-U-well plates with gentle rocking. For coating, 50 µl of a 2% polyHEMA (poly 2-hydroxyethyl methacrylate; Sigma cat. #P3932) solution in ethanol was dried overnight in the well under rocking at 37°C. Cell spheroids were triturated by five passages up and down a P200 tip set at 180 µl and transferred in 25 µl aliquots onto glass slides, covered with a coverslip. Random pictures of the cell aggregates were taken by phase contrast with a 5× objective lens with an Axiovert 200 M microscope (Carl Zeiss, Oberkochen, Germany).

Cells were fixed in either ice-cold methanol (ACTN4, β-actin) or with 4% paraformaldehyde (PFA; all other antigens) at room temperature for 15 min and quenched with 50 mM NH₄Cl in PBS. For colabeling of tubulin and F-actin with septin 2, cells were fixed in 0.25% glutaraldehyde, 4% PFA, 0.1% Triton X-100, 0.6 U/ml CF 647–phalloidin (Biotum, #00041) in PHEM buffer (60 mM PIPES, 25 mM HEPES, 10 mM EGTA, 2 mM MgCl₂ pH ∼7.0) and quenched in 10 mg/ml borohydride in PBS. PFA-fixed samples were permeabilized with 0.2% Triton X-100. Blocking and antibody incubation was in 10% FCS and 1% BSA. The following primary antibodies were used for immunofluorescence (IF) and/or immunoblotting (IB): Cdc42EP1/Borg5 (Proteintech, 27904-1-AP, IB 1:500, IF 1:100); Myc tag (clone 9E10, home-made, IB and IF 1:500) or c-Myc tag (Proteintech #16286-1-AP, IF 1:250, IB 1:1000); Vinculin (Novus Biologicals, #NB120-11193, IF 1:100 or Proteintech 26520-1-AP, IF 1:600), ZO1 (clone R26.4C; Stevenson et al., 1986, 1:50 IF); septin 2 (Proteintech #11397-1-AP, IF 1:300 and #60075-1-Ig, IB 1:1000; BiCell #0022, IF 1:100) Septin 9 (Proteintech # 10769-1-AP, IF 1:300), pThr18/pSer19-MLC2 (Cell Signaling #3674, IF 1:200), E-cadherin (clone RR1, deposited by Barry Gumbiner, University Virginia, with DSHB, Antibody Registry # AB_528114, IF 1:10), tubulin (clone YL1/2, Novus Biologicals, #NB600, IF 1:500), acetylated tubulin clone [6-11B-1] (Millipore Sigma #T6793 IF 1:200), HALO-tag (Promega #g928a, IB 1:500), β-actin [2D4H5] (Proteintech #66009-1-IG, IF: 1:500, IB 1:5000), Actinin α4 (ACTN4; Proteintech 19096-1-AP, IF 1:250), GAPDH [clone1E6D9] (Proteintech # 60004-1-Ig, IB 1:2000), 6xHis-tag [clone 1B7G5] (Proteintech # 66005-1-IG, IB 1:5000), Rac1 (BD Bioscience #610650, IB 1:500), and phospho-tyrosine ([clone PY20], Santa Cruz Biotechnology, sc508, IF 1:200).

Primary antibody labeling was at 4°C overnight (for all phospho-antigens) and for 1 h at room temperature for all other antigens. DAPI (Sigma #D8417), Atto 647N–Phalloidin (Sigma #65906) and the following F(ab′)₂ fragment secondary antibodies from Jackson ImmunoReseach Laboratory were incubated at a dilution of 1:500 for 1 h at room temperature: Rhodamine Red™-X (RRX) AffiniPure™ F(ab′)₂ fragment donkey anti-mouse-IgG (H+L) # 715-296-151; Alexa Fluor^® 488 AffiniPure™ F(ab′)₂ fragment donkey anti-rabbit-IgG (H+L) # 711-546-152; Alexa Fluor^® 647 AffiniPure™ F(ab′)₂ fragment donkey anti-rat-IgG (H+L) # 712-606-153.

Fixed cells were imaged by confocal microscopy on a TCS SP5 confocal microscope (Leica Microsystems, Wetzlar, Germany) using a HCX PL APO 40×/1.25–0.75 oil CS objective or an HCX PL APO 63×/1.4-0.60 oil λ_BL CS objective on glass coverslips mounted in nonhardening, glycerol-based aqueous mounting medium (DABCO). Confocal (pinhole, 1 Airy Unit; pixel size, 160.5 nm) x-y-z and x-z-y stacks were taken at a stepsize of 0.3–1 µm. Images were processed with LAS AF v.2.6.0.7266 (Leica Microsystems), ImageJ (Fiji, version 2.1.0/1.53c) or v.1.52i (National Institutes of Health) and Adobe CS6 (Adobe Inc.) software. Where indicated in the figure legends, a subset of planes was merged using the Z-projection function with maximal intensity; 3D projections were generated with the ‘3D projection’ function, using the brightest point method, with 1 µm slice spacing and 10° angle rotation with interpolation.

Confocal live-cell imaging was conducted with the HCX PL APO 63×/1.4-0.60 oil λ_BL CS objective on MatTek chambers at 37°C in a CO₂-enriched environmental chamber in growth medium without Phenol Red. Images were collected every 30 s. Image stacks were processed in ImageJ by applying a 1-pixel Gaussian blur, a background subtraction and contrast adjustment.

Phase-contrast imaging was performed on an Axiovert 200 M microscope (Carl Zeiss, Oberkochen, Germany). Cells were imaged using an EC Plan-Neofluar 10×/0.30 Ph 1 objective on MatTek chambers at 37°C in a CO₂-enriched atmosphere in growth medium without Phenol Red. Cells in Fig. 2C were imaged with an EC Plan-Neofluar 5×/0.15 Ph 1 objective. Images were acquired with a Hamamatsu ORCA-R² cooled-CCD camera controlled with AxioVision v4.8.1.0 (Carl Zeiss) software. For time-lapse experiments, images were collected every 5 min, using an exposure time of 100 ms and 1×1 camera binning.

Förster resonance energy transfer (FRET) upon transfection of the EcadTSMod TFP/YFP was measured in live cells. FRET efficiency was calculated with the acceptor-bleaching FRET module (FRET-AB) of the LAS AF software. For the setup, an HCX PL APO 40×/1.25-0.75 oil CS objective was used to obtain pre- and post-bleaching confocal (pinhole, 1 AU; pixel size, 63.1 nm; line average, 4) x-y sections. For the bleaching, a region of interest at the cell–cell contact was exposed to the 488 argon laser (main power; 80% with 514 nm laser line; 100%, for 10 iterations). The FRET efficiency was evaluated only in regions of interest with at least 75% of reduced fluorescence intensity in the acceptor species.

Integrated density fluorescence was determined in ImageJ from individual x-y image planes with a set threshold after subtracting background fluorescence deduced from unlabeled image planes. For comparison between experiments, values were normalized to the lowest measured value in a dataset.

For each data point, the average Borg5 fluorescence intensity along a line crossing F-actin filaments under the nucleus was plotted against the average F-actin peak intensity along that line. Peak intensities were recorded from Image J Profile Plots. For normalization, the lowest intensity from each data set was set to 1.

Individual adjacent cells were tracked with the manual option of TrackMate v6.0.0 by following a nucleolar mark in each of the 30 frames. x-y coordinates and pixel distance were exported to PrismGraph v.7.0 software for documentation. Pixel size values (0.645 µm for 10× and 1.29 µm for 5× lens) were calibrated to actual distance.

BL21(DE3) cells (New England Biolabs #C2527I) were transformed with either His–Borg5 in vector pet28a, His–MyoIIA-Rod domain (aminos acids 1339–1960) in pet28a and provided by Anne Bresnick, or GST-MyoIIA-Rod, which is the same MyoII fragment cloned into in pGEX4T-3 vector (Cytiva GE28-9545-521). For septin complex isolation, pnEA-vH_His-TEV-Sept2 (Addgene #174491) was co-tranformed with pnCS_SEPT6_SEPT7-TEV-Strep (Addgene #174499) or pnEA-vH_His-TEV-SEP2_SEP6 (Addgene #174497) was transformed together with pnCS_SEPT7_SEPT9_i1-TEV-Strep (Addgene #174500).

Cultures were grown in Terrific Broth (12 g/l tryptone, 24 g/l yeast extract, 4 ml/l glycerol, 17 mM KH₂PO₄, 72 mM K₂HPO₄) until OD₂₆₀ 2-3, and recombinant protein expression induced with 1 mM IPTG for 3 h at 30°C; cell pellets were stored frozen at −20°C until processing as below.

For His–Borg5, cells were thawed and resuspended on ice in PBS, boiled for 10 min and then plunged into an ice-water bath for 10 min. The 20,000 g pellet fraction was resuspended in fresh 8 M urea buffer containing 10 mM Tris-HCl, 500 mM KCl and10 mM imidazole (pH 7.5) and bound to Ni-NTA agarose (Invitrogen, #R90115) in batch, washed in binding buffer, followed by binding buffer in the absence of urea and eluted with 1.5 bed volume of 500 mM imidazole,10 mM Tris-HCl pH 7.5, 2 mM MgCl₂ and 300 mM KCl. Protein containing fractions were pooled and dialyzed against 10 mM Tris-HCl pH 7.5, 2 mM MgCl₂ and 300 mM KCl and snap-frozen in aliquots. We also purified His–Borg5 under native conditions as described for His–MyoII-Rod and septins below; proteins from both preparation methods bound septin complexes equally, but when prepared under native conditions, more degradation occurred, and even with an Mg-ATP incubation step, heat-shock proteins co-purified.

Septin complexes were isolated as described previously (Castro-Linares et al., 2022). Briefly, cells were resuspended in septin buffer (10 mM Tris-HCl, 2 mM MgCl₂, 300 mM KCl) supplemented with 10 mM imidazole, 10 mM Mg₂SO₄, 1 mM DTT, benzonase (Sigma# E8263-5KU) 0.1 µl/ml buffer, EDTA-free protease inhibitor cocktail (MedChem Express #HY-K0011) and 1 mg/ml lysozyme (Sigma Millipore #L-6876). Lysis was by sonication and subsequent addition of 0.5% Triton X-100. Cleared lysates were incubated with nickel agarose, washed with 10 column volumes of septin buffer with 20 mM imidazole and eluted with 500 mM imidazole in septin buffer. Fractions were pooled according to protein content, dialyzed against septin buffer with 1 mM DTT and snap-frozen in single-use aliquots. For Borg5-Septin binding assays, septin aliquots were column-loaded onto Strep-Tactin^® Sepharose^® (IBA, #2-1201-002), washed with septin buffer and eluted with 5 mM D-biotin in septin buffer.

His–MyoIIRod was purified according to the septin protocol except that, to remove associated heat-shock proteins, the lysate was incubated with 2 mM Mg-ATP for 10 min at 37°C prior to nickel agarose binding; GST–MyoIIRod was prepared similarly, except for omission of imidazole in the buffers and binding to glutathione–Sepharose 4B-CL (Bioworld #20181088-1) and storing the isolated protein on the resin in septin buffer with sodium azide for up to 14 days.

Protein concentrations were estimated from Coomassie-stained gels and proteins were combined in equimolar ratios. Incubations were in septin buffer, supplemented with 1 mM DTT and 0.05% Tween 20 at 4°C for 90 min, including batch-binding to glutathione–Sepharose- and Streptag–Sepharose-immobilized proteins. After binding, Sepharose resins were washed four times in septin buffer and elution was with 5 mM D-biotin (Acros Organic, AC230090010) for Streptag-Sepharose binding and with SDS-PAGE buffer for glutathione–Sepharose binding.

PAGE gels were transferred onto Immobilon-FL membrane (Millipore) and blocked with 5% milk; primary antibody incubation was in 1% BSA and 1% fish serum or SEA BLOCK (ND-R0999 Novatein Biosciences), DyLight 680- or 800-coupled secondary antibodies were incubated in 5% milk. Blots were imaged with a Laser scanner Fuji Typhoon NIR Plus, (Amersham) or LiCor Odyssee (LiCor) and analyzed/quantified with ImageQuant (Amersham) or LiCor Image Studio Lite software. Loading was controlled by GAPDH. The raw scanned blots as well as full sized Coomassie gels are shown in Fig. S7.

2×10 cm dishes of confluent MDCK cells, transiently transfected with Borg5-TEV-HALO or control mock transfected cells, were lysed in 10 mM Tris-HCl pH 7.5, 150 mM NaCl, 0.5 mM EDTA, 1 mM DTT, 0.5% Triton C-100 and Protease Inhibitor Cocktail (Promega G6521). DNA was sheared by passing lysate through a 30 g needle and insoluble material removed by a 20 min 20,000 g centrifugation. Cleared lysate was incubated for 2 h with end-over-end rotation at 4°C with HALO-Trap agarose (Chromotech/Proteintech, #ota), the HALO resin was washed fours time with lysis buffer and incubated with 1 µl His-TEV protease (Genscript# C744N34) in 50 mM Tris-HCl and 100 mM NaCl pH 8.0 plus 5 mM DTT at 4°C overnight to release Borg5 from the HALO moiety. The supernatant and two washes (without DTT) were twice passed over a nickel-agarose spin column to remove TEV protease, precipitated and solubilized in SDS sample buffer.

Proteins were reduced with 2 µl of 0.2 M dithiothreitol (Sigma) for 1 h at 57°C. Samples were cooled to room temperature and then alkylated with 2 µl of 0.5 M iodoacetamide for 45 min at room temperature in the dark. NuPAGE LDS Sample Buffer (1×) (Invitrogen) was added to the samples and the samples loaded onto a NuPAGE^® 4-12% Bis-Tris Gel 1.0 mm (Life Technologies). The samples were run until just passed the stacking region to remove any LCMS incompatible reagents. The gel was then run for 20 min at 200 V, stained with GelCode Blue Stain Reagent (Thermo Scientific) and the entire gel lane excised and destained with 1:1 (v/v) methanol and 100 mM ammonium bicarbonate. Gel pieces were partially dehydrated with acetonitrile then further dehydrated using a SpeedVac concentrator. For proteolytic digestion 300 ng of trypsin (modified, Promega) was added, followed by 200 µl of 100 mM ammonium bicarbonate. The digestion was allowed to proceed overnight at room temperature with light agitation. To stop the digestion and assist with peptide extraction a solution of 5% formic acid in acetonitrile 1:2 (v:v) was added and incubated for 15 min with agitation. The solution was transferred into a new tube and the procedure repeated two more times. The combined solutions were dried down in a SpeedVac concentrator to remove the acetonitrile. The samples resuspended in 0.1% acetic acid and loaded onto equilibrated Ultra-Micro SpinColumns™ (Harvard Apparatus) using a microcentrifuge. The spin columns were washed three times with 0.1% trifluoroacetic acid and the last wash with 0.5% acetic acid. Peptides were eluted with 40% acetonitrile in 0.5% acetic acid, followed by 80% acetonitrile in 0.5% acetic acid. The solutions combined and dried down using a SpeedVac concentrator. The samples were reconstituted in 0.5% acetic acid and stored at −80°C until analysis.

An aliquot of each sample was LC separated on an Easy-nLC 1200 HPLC (Thermo Fisher Scientific) by loading it onto an Acclaim PepMap trap column (2 cm×75 µm) in-line with an EASY-Spray analytical column (50 cm×75 µm ID PepMap C18, 2 μm bead size). The sample was gradient eluted into a Thermo Fisher Scientific Orbitrap Eclipse Tribrid Mass Spectrometer using the following gradient: in 5 min to 5%, 60 min to 35%, 10 min to 45% and another 10 min to 100% solvent B and hold at 100% B for 10 min (solvent A, 2% acetonitrile in 0.5% acetic acid; solvent B, 80% acetonitrile in 0.5% acetic acid). The flow rate was set to 200 nl/min. High resolution full mass spectrometry (MS) spectra were acquired with a resolution of 240,000, an automatic gain control (AGC) target of 10⁶, a maximum ion time of 50 ms, and a scan range of 400–1500 m/z. All MS/MS spectra were collected using the ion trap in rapid scan mode with an AGC target of 2×10⁴, maximum ion time of 18 ms, one microscan, 0.7 m/z isolation window, and a normalized collision energy (NCE) of 27.f

The MS/MS spectra were searched against the UniProt (www.uniprot.org) Canis lupus familiaris database with common lab contaminants using Sequest within Proteome Discoverer 1.4. Searches were performed using the digestion enzyme trypsin permitting two missed cleavages, peptide length of 6 to 144, a precursor mass tolerance of ±10 ppm, a fragment mass tolerance of ±0.4 Da, variable modification of oxidation on methionine, deamidation on glutamine and asparagine, and a fixed modification of carbamidomethyl on cysteine. The results were filtered to better than ≤1% peptide; the protein false discovery rate (FDR) was determined by searching against a decoy database, and only proteins with at least two unique peptides were reported.

Unless otherwise indicated data points represent independent experiments. Explanation of their acquisition and number of independent experiments are noted in the figure legends. All analysis was carried out using GraphPad Prism. Error bars are represented as mean±s.d.. For comparison of two sample populations, we used a parametric two-tailed t-test (paired or unpaired as indicated in the figure legend) when sample values in individual experiments were normally distributed. Where normality could not be established, samples were compared by a nonparametric Mann–Whitney test. We used the convention for representation of P-values as follows: *P<0.05; **P<0.01, ***P<0.001.

We thank Dr Yixian Zheng (Carnegie Institution for Science, Baltimore, Maryland) for the mouse Borg5 cDNA, Dr N. Borghi (Institute Monod, Paris, France) for the E-Cadherin biosensor cDNA, Dr A. Bresnick (Einstein Medical College) for providing the MyoIIA–GFP and His–MyoIIRod cDNAs, A. Wells (U. Pittsburgh) for the ACTN4GFP cDNA, and colleagues who deposited cDNAs we obtained from Addgene (listed in the Materials and Methods section).

The peer review history is available online at https://journals.biologists.com/jcs/lookup/doi/10.1242/jcs.261705.reviewer-comments.pdf