A key role for Canoe’s intrinsically disordered region in linking cell junctions to the cytoskeleton

Abstract

Adherens junctions regulate tissue architecture, mediating robust yet dynamic cell-cell adhesion and, via cytoskeletal linkage, allowing cells to change shape and move. Adherens junctions contain thousands of molecules linked by multivalent interactions of folded protein domains and Intrinsically Disordered Regions (IDRs). One key challenge is defining mechanisms conferring robust linkage and mechanosensing. Drosophila Canoe and mammalian Afadin provide superb entry-points to explore how their complex protein structures and shared IDRs enable function. We combined genetic, cell biological and biochemical tools to define how Canoe’s IDR functions during morphogenesis. Unlike many of Canoe’s folded domains, the proximal IDR is critical for junctional localization, mechanosensing and function. In its absence, the mutant protein localizes to nuclei. We took the IDR apart, identifying two conserved stickers that directly bind F-actin, separated by less-conserved spacers. Surprisingly, while mutants lacking the IDR die as embryos with morphogenesis defects, no IDR sub-region is essential for viability. Instead, stickers and spacers act combinatorially to ensure localization, mechanosensing and function.

Etoc summary:

Jensen et al. reveal that the IDR of Canoe is critical for its role in linking cell-cell adherens junctions to the cytoskeleton. Surprisingly, deleting the proximal IDR relocalizes the protein from junctions to nuclei. no IDR sub-region is essential for viability. Instead, stickers and spacers act combinatorially to ensure localization, mechanosensing and function

Introduction

Many key cellular functions are carried out by large multi-protein complexes rather than individual proteins. Some form biomolecular condensates which can contain thousands of molecules linked by multivalent interactions of both folded protein domains and intrinsically disordered regions (IDRs; Banani et al., 2017). Understanding how these assemble is a key challenge for our field. A premier example is the cell-cell adherens junction (AJ), which joins cells to one another, and, via linkage to the actomyosin cytoskeleton, mediates cell shape change and cell migration. At its core is the cadherin-catenin complex. Cadherin extracellular domains mediate homophilic cell adhesion, while their cytoplasmic tails link to actin via β-catenin (βcat) and α-catenin (αcat). While connections between AJs and the cytoskeleton were initially thought to rely on this simple linear linkage, instead it is mediated by a dynamic and multivalent network including many more proteins (Perez-Vale and Peifer, 2020). They provide mechanosensing and mechano-responsiveness, with AJ:actin linkage strengthened when force is exerted on AJs (Yap et al., 2018).

Textbook diagrams often depict individual cadherins linked to their cytoplasmic partners. However, cadherins assemble in much larger complexes, linking in cis to cadherins in the same cell and in trans to cadherins on neighboring cells. Rather than being arrayed continuously along the apicolateral membrane, AJ proteins form discrete puncta (Truong Quang et al., 2013). In Drosophila, the Harris lab counted molecules in spot AJs assembling during cellularization (McGill et al., 2009). Bicellular borders have 3–5 puncta, each containing ~1500 cadherin-catenin complexes and ~200 Bazooka (Baz)/Par3 proteins. However, they are not “standardized”; size and protein number per punctum varies and proteins dynamically enter and leave. As cells change shape during gastrulation, puncta increase in number and decrease in brightness (Schmidt et al., 2023). Similar biomolecular condensates are seen in cultured mammalian cells (e.g. Choi et al., 2016; Erami et al., 2015; Wu et al., 2015), though their architecture varies widely depending on cell type. Tight junctions (TJs) and focal contacts also assemble as biomolecular condensates (Citi, 2020; Sun et al., 2022).

A hallmark feature of many cell junction proteins are long IDRs (Rouaud et al., 2020). IDRs perform many functions in assembling multiprotein complexes. Most have low sequence complexity, with specific subsets of amino acids enriched. They mediate interactions via two broad mechanisms (Choi et al., 2020). First, embedded in IDRs are Short Linear amino acid Motifs (SLiMs), sometimes referred to as “stickers”, serving as binding sites for other proteins. Some SLiMs fold on their own while others fold when binding a partner or bind as extended peptides. SLiMs are embedded in unstructured regions that act as “spacers” and also mediate low affinity interactions. Many IDRs phase separate in vitro, and in some models, phase separation helps drive multiprotein complex assembly in vivo. The IDRs of multiple junctional proteins, including ZO-1 (Beutel et al., 2019), βcat (Zamudio et al., 2019), Par3 (Liu et al., 2020), and Afadin (Kuno et al., 2025) can phase separate. Apart from ZO-1, the roles of this in vivo remains unclear.

Assembling AJs involves multivalent interactions among the many proteins in the AJ network. The Drosophila AJ-cytoskeletal linker Canoe (Cno) and its mammalian ortholog Afadin play key roles in multiple morphogenic movements during development and homeostasis, making them outstanding entry-points for studying AJ assembly and dynamics. Cno regulates diverse processes including apical AJ positioning, mesoderm apical constriction, cell intercalation and axis elongation, and epithelial sheet migration during dorsal closure and head involution (Boettner et al., 2003; Choi et al., 2013; Sawyer et al., 2011; Sawyer et al., 2009). Cno is mechanosensitive, with recruitment to AJs enhanced under tension (Yu and Zallen, 2020), where it stabilizes AJ:cytoskeletal linkage (Perez-Vale et al., 2021). In Cno’s absence, AJ:cytoskeletal connections are disrupted, with AJ gaps and the cytoskeleton broadened under tension. Similar effects are seen in mammalian cells with reduced Afadin (Choi et al., 2016; Sakakibara et al., 2020).

The second reason Cno/Afadin is a superb entry-point is its size and structural complexity. Cno/Afadin is a multidomain protein with five well-conserved folded protein domains followed by a long IDR (Fig. 1A; Gurley et al., 2023). This allows for multivalent interactions with other proteins in the AJ network. The two N-terminal Ras-association (RA) domains each bind the membrane-anchored small GTPase Rap1 when it is in its active state, and this in turn “activates” Cno (McParland et al., 2024b). The PDZ domain binds transmembrane AJ proteins including nectins (Takahashi et al., 1999) and E-cadherin (Ecad; Sawyer et al., 2009). Meanwhile, regions in the IDR bind F-actin (Carminati et al., 2016; Mandai et al., 1997; Sawyer et al., 2009), αcat (Pokutta et al., 2002), and other proteins.

Fig 1. Cno’s IDR is critical for Cno’s role in embryonic morphogenesis and in Cno protein localization.

(A). Drosophila Cno, Human Afadin, and CnoΔProxIDR. Orange boxes=predicted helices. *= helix in mammalian Afadin that binds αcat ‘s M domain. (B-G). Representative cuticles illustrating different strength phenotypes. B-E. Red arrows indicate head skeleton (B-C) or complete head involution failure (D,E). (E, F). Blue arrows=dorsal hole (E) or complete failure of dorsal closure (F). (G). Most extreme defects in epidermal integrity. (H). Distribution of cuticle phenotypes in different mutants. cno^R10/+ data is from Gurley et al., (2023). (I-N, P-U). Embryos, stained to visualize Cno or GFP-tagged CnoΔProxIDR. (I-L, P). Cross-sections, apical up, stages indicated. Yellow arrows=junctional localization. Cyan arrows=nuclei. Red arrow=ectodermal fold. (M, N). Apical view, level of AJs. Yellow arrows=TCJs, cyan arrows=bicellular junctions. (O) Tricellular junction enrichment. n=14 embryos. In this and subsequent Figures TCJ enrichment significance was assessed using Welch’s unpaired T-test followed by Welch’s ANOVA. (Q). Whole embryo. Arrows=enrichment in invaginating midgut. (S-U). Stages 8–9. Yellow arrows=cells rounded up for mitosis. Cyan arrows=nondividing cells. Panels D and G are from Bonello et al., (2018).

To elucidate the molecular mechanisms underlying Cno function, we are systematically taking it apart, deleting specific domains or regions and assessing their roles. We initially hypothesized Rap1 binding to the RA domains activated Cno and Cno then used its PDZ domain and the so-called F-actin binding (FAB) region of the IDR to link Ecad and actin. We tested this, using CRISPR to generate cno mutants cleanly deleting each domain/region (Perez-Vale et al., 2021). Mutants lacking both RA domains die as embryos with severe morphogenesis defects, consistent with our hypothesis. In contrast, neither the PDZ domain nor the FAB region are required for viability or fertility. The RA2 and Dilute domains are also individually dispensable (McParland et al., 2024a; McParland et al., 2024b; Perez-Vale et al., 2021). However, each mutant has reduced function in sensitized assays. This changed our thinking, emphasizing how robustness of AJ-cytoskeletal connections is conferred by protein networks linked by multivalent interactions.

While Cno and Afadin share long IDRs, they differ in length, charge, and sequence composition (Gurley et al., 2023). The C-terminal FAB region is modestly conserved (35% identity; Perez-Vale et al., 2021), but reciprocal BLAST searches with the proximal IDR revealed “no significant similarity”. Binding sites in Afadin’s IDR for ZO-1 (Ooshio et al., 2010) and Lgn (Carminati et al., 2016) do not appear to be conserved in Cno. We used AlphaFold to predict potential structures in the IDRs (Gurley et al., 2023). Afadin’s IDR has multiple predicted α-helices (Fig. 1A). These include a helix known to bind αcat’s M-domain (Maruo et al., 2018; Pokutta et al., 2002; Fig. 1A asterisk), but surprisingly this helix is not conserved in Cno. In contrast, two predicted helices in the middle of both IDRs are conserved (33% identity; Gurley et al., 2023). These two helices in Afadin bind both αcat’s actin-binding domain and actin filaments, stabilizing αcat’s interaction with actin (Gong et al., 2025). There are also predicted helices in the FAB.

The IDRs of both Afadin and Cno bind F-actin (Mandai et al., 1997; Sawyer et al., 2009). Initial studies used fragments covering the C-terminal half of the IDR. Subsequent work narrowed one interacting region in Afadin to the two conserved α-helices in the IDR along with a more C-terminal helix found only in Afadin (Carminati et al., 2016). These conserved helices co-sediment with F-actin in the presence of αcat and enhance αcat:actin binding; cryo-EM revealed the structural basis for this (Gong et al., 2025). In contrast, despite the designation of the C-terminal IDR as the “F-actin binding region” (FAB), no one directly assessed actin binding.

Functional assays of the IDR are limited. Deletion studies suggest a modest role for the FAB in full function of both Afadin (Sakakibara et al., 2018) and Cno (Perez-Vale et al., 2021). Mutants with premature stop codons in Cno’s IDR have strong phenotypes, but we could not rule out effects of nonsense-mediated mRNA decay (Gurley et al., 2023). In mammalian cells, Afadin knockdown (Choi et al., 2016) or knockout (Gong et al., 2025; Sakakibara et al., 2020) broaden actin at bicellular and tricellular junctions (TCJs). A mutant protein lacking both the helix that binds αcat’s M-domain and the two conserved helices did not rescue junctional actin architecture, while adding this region alone provided partial rescue (Sakakibara et al., 2020; Rätze, et al, 2025). More limited analysis of an Afadin mutant retaining the FAB but lacking the proximal IDR revealed that it still localizes to AJs but loses apical enrichment and apical AJs are destabilized (Kuno et al., 2025). Here we extended and expanded this analysis, combining genetic, cell biological and biochemical tools to define the function of the IDR and its stickers and spacers in Cno localization and function in vivo during morphogenesis.

Results

Cno’s proximal IDR is essential for embryonic viability

We previously assessed the role of the FAB, the IDR region with the strongest sequence conservation between flies and mammals. It is dispensable for viability and fertility, but sensitized tests revealed it supports robustness of morphogenesis (Perez-Vale et al., 2021). To test function of the the proximal IDR (ProxIDR), we generated a cno mutant deleting sequences from the PDZ domain to the beginning of the FAB (Fig 1A; Fig S1). This mutant, with an added C-terminal GFP, was engineered into the endogenous cno locus, using the platform we developed (Perez-Vale et al., 2021). We refer to it as cnoΔProxIDR. This and all mutants described below were verified by sequencing genomic DNA, and we used immunoblotting to confirm that appropriately sized, GFP-tagged proteins were produced at roughly wildtype levels (Fig S2A–F).

We first asked whether cnoΔProxIDR was viable over our null allele, cno^R2. No adult progeny were seen (0/174), thus contrasting with cnoΔRA2, cnoΔDIL, cnoΔPDZ, and cnoΔFAB, all of which are adult viable (McParland et al., 2024a; McParland et al., 2024b; Perez-Vale et al., 2021). However, unlike null alleles, cnoΔProxIDR is not zygotically embryonic lethal (7% lethality; n=755; wildtype range=3–8%). To assess CnoΔProxIDR’s function in embryonic morphogenesis, we examined maternal-zygotic mutants, generating females with germlines homozygous for cnoΔProxIDR (Chou and Perrimon, 1996) and crossing them to cnoΔProxIDR/+ males. 50% of the offspring are maternal-zygotic mutants, and we observed 57% embryonic lethality (n=509), consistent with fully penetrant lethality of maternal-zygotic mutants and substantial zygotic rescue. Thus, the proximal IDR is essential for embryonic viability.

The proximal IDR plays important roles in embryonic morphogenesis but some Cno function remains in its absence

As a first assessment of CnoΔProxIDR function in morphogenesis, we examined larval cuticles of dead embryos. This allows us to assess the completion of multiple morphogenic events requiring Cno function, including dorsal closure, head involution and ventral epidermal integrity (Fig 1B–G). Maternal/zygotic cno null mutants have strong defects in all these events (Fig 1H), with fully penetrant failure of dorsal closure and head involution. ~70% of embryos exhibit additional defects in epidermal integrity (as in Fig 1G; Gurley et al., 2023). cnoΔProxIDR maternal-zygotic mutants had substantial morphogenesis defects, but these were not as severe as those of cno null mutants. Only 43% of cnoΔProxIDR mutants had complete failure of dorsal closure and head involution or worse (Fig. 1H; phenotypes in Fig 1F, G). 37% were substantially less severe, with defects in head involution but not in dorsal closure (Fig 1H; as in Fig 1C–E; n=293). Thus, the proximal IDR is important for Cno function in morphogenesis, but some function remains in its absence.

Deleting Cno’s proximal IDR relocalizes the protein from AJs to nuclei

Wildtype Cno localizes to nascent AJs as they assemble during cellularization (Fig 1I, yellow arrow; Bonello et al., 2018; Choi et al., 2013) and remains localized to apical AJs from gastrulation onset through the remainder of development (Fig 1K, yellow arrow; Sawyer et al., 2009). While CnoΔRA and CnoΔRA1 exhibit altered localization during cellularization, all mutant proteins tested thus far localize to AJs after gastrulation onset.

We thus explored CnoΔProxIDR localization. The results were quite surprising. During cellularization, CnoΔProxIDR predominantly localized to nuclei (Fig 1J, cyan arrow). Low levels of CnoΔProxIDR accumulated at spot AJs (Fig 1J, yellow arrows), but unlike wildtype Cno (Fig 1I) this localization was not restricted to apical junctional planes. As gastrulation began, CnoΔProxIDR remained predominantly nuclear, but AJ localization increased (Fig 1L, cyan vs yellow arrows). During germband extension, wildtype Cno localizes to all AJs and is especially enriched at tricellular (TCJs) relative to bicellular junctions (Fig 1M, yellow vs cyan arrows), because of elevated tension at these sites (Yu and Zallen, 2020). At this stage, CnoΔProxIDR retains strong nuclear localization along with some AJ enrichment (Fig 1N), but TCJ enrichment relative to levels at bicellular junctions is lost (Fig 1N, yellow vs cyan arrows; quantified in O). Intriguingly, AJ enrichment of CnoΔProxIDR was elevated in cells that were apically constricting to form folds or the posterior midgut invagination (Fig 1P, Q, red versus yellow arrows), as is true for wildtype Cno.

During stage 8 mitosis begins in the thorax and abdomen. Cells enter mitosis in programmed groups called mitotic domains (Foe, 1989), rounding up and with nuclear contents released into the cytoplasm. Core AJ protein and Cno localization to AJs is somewhat reduced in cells as they round up to divide relative to non-mitotic cells (Fig 1R, yellow vs cyan arrows), perhaps by simple dilution as junctional perimeter increases. However, localization of CnoΔProxIDR was strikingly different. No longer confined to nuclei, CnoΔProxIDR filled the cytoplasm of dividing cells and became substantially enriched at the cell cortex relative to neighboring non-mitotic cells (Fig 1S, yellow vs cyan arrows). Cortical enrichment of CnoΔProxIDR in mitotic cells continued at stage 9 (Fig 1U, cyan vs yellow arrows), in contrast to what is observed with wildtype Cno (Fig 1T). Upon sectioning 1μm deeper, continuing CnoΔProxIDR localization to nuclei of non-mitotic cells is apparent (Fig 1U”, arrow), and nuclear accumulation continued through dorsal closure.

We next asked whether the presence of wildtype Cno could restore normal CnoΔProxIDR localization, as there is evidence from both Drosophila and mammalian cells that Cno/Afadin can oligimerize (Bonello et al., 2018; Mandai et al., 1997). We took advantage of the 50% of embryos receiving a wildtype paternal copy of cno, recognizing them by restoration of staining by our C-terminal Cno antibody. Despite the restoration of wildtype Cno to AJs (Fig. S3A’”), CnoΔProxIDR localization to non-mitotic cells remained reduced (Fig. S3A”, red arrow). Instead, it was predominantly nuclear in non-mitotic cells (Fig. S3B, magenta arrows) and recruited to the cortex of mitotic cells (Fig. S3A, green arrow).

Finally, since CnoΔProxIDR returned to the cell cortex in mitotic cells, we asked if the release of CnoΔProxIDR from nuclei when the nuclear envelope breaks down during mitosis was sufficient to restore other aspects of normal Cno localization. We first used line scans across AJs of non-mitotic and mitotic cells (Fig. S3C–F). As we and others previously observed, junctional levels of Arm, wildtype Cno, and CnoWTGFP are all about 2–3 fold higher in non-mitotic versus mitotic cells (Fig. S3G), though this may simply reflect the expanded cell perimeter due to cell rounding. The opposite is true for CnoΔProxIDR, which is at 2–3 times higher levels at AJs of mitotic versus non-mitotic cells (Fig. S3G). However, return to the cortex does not restore TCJ enrichment of CnoΔProxIDR—it remained substantially less enriched at TCJs of mitotic cells than is wildtype Cno in non-mitotic cells (Fig S3H). Finally cross sections revealed wildtype Cno is apically enriched in AJs of both non-mitotic and mitotic cells (Fig S3I, red arrows vs blue arrows). In contrast, CnoΔProxIDR is present all along the lateral membrane of mitotic cells (Fig. S3J, blue arrows) rather than being apically enriched as it is in non-mitotic cells (Fig. S3J, red arrows). Thus, while we cannot rule out the idea that release from nuclei might restore some aspects of normal localization of CnoΔProxIDR, we think it more likely that the cortical enrichment of CnoΔProxIDR in mitotic cells reflects association with cortical actin. Consistent with this, it also is enriched at the contractile ring of cells undergoing cytokinesis (Fig S3B, blue arrow). Taken together, these data reveal that the proximal IDR is critical for effective recruitment/retention of Cno at AJs, as, when it is missing, the protein substantially re-localizes to nuclei.

Blocking nuclear export leads to accumulation of wildtype Cno in nuclei

While we have never seen wildtype Cno in nuclei, there are multiple reports of nuclear Afadin localization (see Discussion). We thus explored whether wildtype Cno might enter and exit nuclei without accumulating there. To do so, we used leptomycin, a well-studied inhibitor of nuclear export that we previously used to assess nuclear import of the APC tumor suppressor (Roberts et al., 2012). We permeabilized cnoWTGFP embryos using octane (Perez-Vale, et al., 2021), and treated them with either water, as a control, or 10μg/ml leptomycin (Townsley and Bienz, 2000; n=three experiments). In control embryos, GFP-tagged wildtype Cno was strongly junctional, whether detected with antibodies to the GFP-tag or to Cno (Fig 2A, D, red arrows;19 of 19 embryos). In contrast, after leptomycin treatment, AJ accumulation of CnoWTGFP was reduced (Fig 2B, E, red arrows). Instead, CnoWTGFP accumulated in nuclei (Fig. 2C, F, magenta arrows). This was most apparent in earlier embryos (stages 5–8; 24/24 embryos) but became less apparent or not apparent in epidermal cells at later stages (3/8 embryos). However, even during dorsal closure nuclear enrichment was apparent in amnioserosal cells (Fig S4B, C, magenta arrows), although junctional staining was also somewhat restored (Fig S4B, red arrows; in control embryos CnoWTGFP remained junctional; Fig S4A). These differences in apparent nuclear accumulation at different stages will require future analyses. To further verify this we used an alternate permeabilization and fixation procedure and also saw nuclear accumulation of CnoWTGFP in early embryos (Fig. S4D,E vs D-inset). Thus, although wildtype Cno does not accumulate in nuclei, if nuclear export is blocked it does accumulate there, suggesting it is continuously shuttling in and out of nuclei.

Fig 2. Inhibiting nuclear export results in wildtype Cno accumulating in nuclei.

All are stage 5–6 cnoWTGFP embryos, permeabilized with octane and treated either with leptomycin or with water as a control. (A-C). CnoWTGFP visualized with GFP antibodies. (D-F). CnoWTGFP visualized with Cno antibodies. A, D. Control embryos. Arm and Cno are enriched at AJs (red arrows) and there is no sign of nuclear accumulation. B, C, E, F. Embryos treated with leptomycin, visualized at the AJ level (B,E) or more basal (C, F). Accumulation of CnoWTGFP is substantially reduced at AJs (B, E red arrows). Instead, it accumulates in nuclei (C, F, magenta arrows).

cnoΔProxIDR mutants have defects in morphogenesis and AJ integrity but these are less severe than those of cno null mutants

Next, we examined CnoΔProxIDR function in detail. Cno’s first role in morphogenesis is at the onset of gastrulation when ventral cells apically constrict and invaginate to become mesoderm (Fig 3A–B, red arrows). In cno null mutants the contractile actomyosin network detaches from AJs and invagination stalls, leading to fully penetrant defects in invagination and an open ventral furrow (Sawyer et al., 2009). Some embryos zip up at the midline but defects persist. cnoΔRA mutants also have fully penetrant defects in mesoderm invagination, but defect severity is reduced relative to null mutants (Perez-Vale et al., 2021). We examined ventral furrow invagination in cnoΔProxIDR mutants, using our C-terminal anti-Cno antibody that does not recognize CnoΔProxIDR to distinguish maternal-zygotic mutants from zygotically-rescued embryos. Many cnoΔProxIDR mutants had defects in ventral furrow closure (Table S1), but these were substantially less penetrant and less severe than seen in cno null mutants. 6/31 mutants at stages 6–9 had closed ventral furrows, 10/31 had mild defects in closure (short regions where the furrow was open; Fig 3C, red arrow), 8/31 had moderate defects (>30% open), and 7/31 had a wide-open ventral furrow (Fig 3D, red arrows).

Fig 3. cnoΔProxIDR mutants have defects in morphogenesis and junctional integrity, but these are less severe than those of cno null mutants.

Embryos, anterior left, antigens and stages indicated. (A-D). Ventral views, arrows=ventral furrow. (E-F) Lateral views. Red arrows=DV borders, green arrows=AP borders, yellow arrows=Baz restricted to center of DV border. (G-I). Stage 8. Asterisks=all junctional gaps/disruptions. Arrows=selected gaps. (J). Gap quantification. n=14, p value =0.0000113. In this and subsequent Figures significance of gap number was assessed using Welch’s unpaired T-test followed by Welch’s ANOVA. (K-N). Stage 11. Arrows=Failure of cells to resume columnar architecture. (O, P). Wildtype dorsal closure (cyan arrow) and head involution (yellow arrow). Leading edge cells are relatively uniform in width (P). (Q-S). cnoΔProxIDR mutants. (Q). Hyper-constricted and splayed open leading edge cells. (R, S). Defects in head involution (yellow arrows) and dorsal closure (cyan arrow). (T, U). Enhanced resolution imaging of AJs using the Airyscan module. Yellow arrows=junctional puncta enriched for Arm. Cyan arrows=junctional puncta enriched for Cno.

As gastrulation began in cnoΔProxIDR mutants, AJs assembled with Armadillo (Arm) enriched apically but also localized along the lateral interface, thus matching wildtype (Fig 1K vs L). Baz/Par3 also localizes to apical AJs at this stage. In wildtype, Baz localizes all around the cell but is planar-polarized, with two-fold higher accumulation on dorsal-ventral (DV) relative to anterior-posterior (AP) cell borders (Fig 3E, red vs. green arrows; Zallen and Wieschaus, 2004). However, in both cno null and cnoΔRA mutants (Perez-Vale et al., 2021; Sawyer et al., 2011), Baz planar polarity is strongly enhanced. This was also true in cnoΔProxIDR mutants. Baz was strongly enriched on DV cell borders (Fig 3F; red vs. green arrows) and restricted to the center of the DV borders in some cells (Fig 3F, yellow arrows).

In cno null mutants, cnoΔRA mutants, or in embryos with strong RNAi knockdown of Cno, a subset of AJs become destabilized during germband elongation (Manning et al., 2019; Perez-Vale et al., 2021; Sawyer et al., 2011), particularly TCJs and those along aligned AP borders, which are known or suspected to be under elevated tension (Fig 3G, arrows). We quantified AJ defects in cnoΔProxIDR mutants at stages 7–8, by staining for Arm. We included small or large gaps where three or more cells came together or places where AJ protein accumulation broadened, and scored gaps blinded in wildtype and cnoΔProxIDR mutants. Mutants had a dramatically elevated number of gaps at multicellular junctions—a mean of 5.5 gaps/field in wildtype versus 38 gaps/field in cnoΔProxIDR mutants (Fig 3H vs 3I; quantified in 3J; the one cnoΔProxIDR mutant outlier with near normal gap numbers may be zygotically rescued). However, there was a qualitative difference between cnoΔProxIDR and strong cno mutants—instead of the many broad gaps at AJs seen in strong mutants (Fig 3G, arrows), we primarily saw smaller gaps in cnoΔProxIDR (Fig 3I, I’ arrows).

During stages 9–10 lateral and ventral ectodermal cells experience two challenges to AJs, posed by mitosis and by invagination of ~30% of cells as neural stem cells. These challenges make the ventral epidermis particularly susceptible to reduced cell adhesion. In wildtype, most cells are columnar, as they rapidly return to this shape after division (Fig 3K). In contrast, recovery of columnar cell shape after division is delayed in strong cno mutants, leading to penetrant ventral epidermal integrity defects (Manning et al., 2019; Perez-Vale et al., 2021). At stage 9, most cnoΔProxIDR mutants appeared relatively normal, aside from continued open ventral furrows. However, at stages 10–11 many embryos had ventral cells that failed to return to columnar shape after division. Defects varied in severity: ~38% had mild defects (10/26 embryos), 35% had more widespread failure of columnar cell architecture (9/26 embryos; Fig 3L, arrows), and 27% had more severely disrupted ventral epidermis (7/26 embryos; Fig 3M, N, arrows). During stages 13–14, wildtype embryos complete head involution and dorsal closure (Fig 3O, yellow and cyan arrows, respectively). Consistent with the observed cuticle defects, a subset of cnoΔProxIDR mutants failed to close dorsally (Fig 3R, cyan arrows), head involution was often defective (Fig 3R, S, yellow arrows), and leading-edge cell shapes were hyper-constricted or hyper-extended (Fig 3P vs Q), as is seen in strong cno mutants (Manning et al., 2019). However, few embryos had ventral epidermal holes, consistent with the cuticles, another way in which cnoΔProxIDR mutants are less severe than the null mutant.

Our final analysis of proximal IDR function was to examine its role in AJ substructure. Cadherin/catenin complexes are not continuous along the membrane but assemble into large puncta containing hundreds of proteins (McGill et al., 2009). Enhanced resolution imaging using the Airyscan module revealed that this punctate array of Arm is maintained during germband extension (Fig 3T, yellow arrows; Schmidt et al., 2023). Intriguingly, Cno’s C-terminus is also enriched in puncta, but many puncta have different relative levels of Arm vs Cno (Fig 3T, yellow vs cyan arrows; Schmidt et al., 2023). In cnoΔProxIDR mutants we saw no obvious changes in the punctate Arm localization (Fig 3U, yellow arrows). Polychaetoid/ZO1 loss similarly does not alter Arm’s punctate localization (Schmidt et al., 2023). While CnoΔProxIDR localization to AJs was substantially reduced, if we enhanced the junctional signal, it appeared to localize to a broader zone than is occupied by Arm puncta (Fig 3U, yellow vs cyan arrows). Thus, in strong contrast to our earlier mutants in which folded domains were deleted, the proximal IDR is critical for Cno AJ localization, Cno function in stabilizing AJs and in morphogenesis. However, subtle differences with the cno null mutant phenotype suggest CnoΔProxIDR retains some residual function.

Replacing Cno’s proximal IDR with Afadin’s restores AJ localization and viability but does not provide full function

While the five folded protein domains of Cno and Afadin are well conserved (42–74% sequence identity; Gurley et al., 2023) and the C-terminal FAB has clear blocks of conservation (35% identity), the proximal IDR is highly divergent between flies and mammals (Gurley et al., 2023). We thus began with the hypothesis that Afadin’s proximal IDR would not functionally replace that of Cno. To test this, we created a mutant with Cno’s proximal IDR replaced by Afadin’s (cnoAfadinIDR; Fig 4A). The replacement began 30 amino acids downstream of the PDZ domain, as the preceding region is well conserved among different species and may be a PDZ domain extension. At the C-terminal boundary we ended the deletion 10 amino acids N-terminal to our earlier FAB deletion. For cloning, we engineered restriction sites into these positions encoding short linker peptides (Methods; Fig S1). As a control, we re-inserted the Cno IDR coding sequence into the same location (cnoCanoeIDR; Fig 4A). CnoCanoeIDR accumulated at near-wildtype levels, while CnoAfadinIDR accumulated at somewhat lower levels (~60%; Fig S2C, D).

Fig 4. Afadin’s IDR can restore viability and fertility and largely but not complete restore protein localization.

(A). Drosophila Cno and the IDR replacement mutants. (B-D). Maternal/zygotic mutant adults of the indicated genotype. (E-M). Embryos, anterior left, antigens and stages indicated. (E-G). Stage 9. Both mutant proteins localize to AJs. (H-J). Cross-sections, apical up, stage 6. Yellow arrows=junctional localization. Cyan arrows=nuclei. (K-M). Tricellular junction enrichment. Yellow arrows=TCJs, cyan arrows=bicellular junctions. (N). Quantification of Tricellular junction enrichment. n=14 embryos. (O). Relative accumulation at AJs. n= 18 embryos. Kolmogorov-Smirnov analysis was used to to see if the distributions were different.

We then crossed each mutant to our standard null allele. To our surprise, cnoAfadinIDR/cnoR² flies were adult viable without morphological defects, as were the cnoCanoeIDR/cno^R2 control (Fig 4C, D vs B). However, while cnoCanoeIDR/cno^R2 adults appeared at near-Mendelian ratios (26% versus 33% expected; Balancer homozygotes die; n=827), cnoAfadinIDR/cno^R2 adults were less viable (16% versus 33% expected; n=1167). Both produced viable progeny surviving to adulthood, confirming that maternal-zygotic mutants are viable. Like wildtype Cno (Fig 4E), both CnoCanoeIDR (Fig 4F) and CnoAfadinIDR (Fig 4G) proteins localized to AJs, with enrichment at the apical end of the lateral interface, like wildtype Cno (Fig 4I, J vs H, yellow arrows). However, replacing Cno’s proximal IDR with that of Afadin did not fully eliminate the nuclear localization seen in CnoΔProxIDR (Fig 4J, cyan arrow). We also assessed TCJ enrichment. While CnoCanoeIDR enrichment at TCJs matched that of wildtype Cno (Fig 4L vs K, arrows; quantified in N), CnoAfadinIDR had substantially reduced TCJ enrichment (Fig 4M vs K, arrows; quantified in N). Finally, we examined overall levels of AJ accumulation, comparing progeny of cnoAfadinIDR/cno^R2 parents to CnoWTGFP on the same slide, using the GFP signal. CnoAfadinIDR localization to AJs was reduced about 2-fold (Fig 4O). Thus, replacing Cno’s proximal IDR with that of Afadin restores AJ localization and adult viability, but does not restore enrichment at TCJs under tension.

We developed a sensitized assay to assess function of viable alleles (Perez-Vale et al., 2021). To do so, we reduce the maternal dose of the mutant protein by making mothers transheterozygous for our null allele, cno^R2, and then cross transheterozygotes to one another. When we do this assay with the wildtype allele, crossing +/cno^R2 males and females, 25% of the progeny die: those that are cno^R2/cno^R2. Because of the strong maternal contribution, the dead embryos only have mild defects in morphogenesis, as assessed by examining cuticles, primarily defects in head involution (Fig 5A; Gurley et al., 2023). Dorsal closure and epidermal integrity, which are disrupted in maternal/zygotic null mutants, remain unaffected.

Fig 5. Sensitized assays reveal that cnoAfadinIDR does not have fully wildtype function.

(A-C). Distribution of cuticle phenotypes in different mutants, using cuticle examples in Fig 1. cno^R2/+ data is from Gurley et al., (2023). n values are in the text. (D-F). Junctional gaps, Stage 8. Asterisks=all junctional gaps/disruptions. Arrows=selected gaps. (G). Gap quantification. n=14 fields of view.

We tested CnoCanoeIDR and CnoAfadinIDR function using this assay. When we crossed cnoCanoeIDR/cno^R2 males and females, we saw 42% lethality (n=1317), above the 25% expected if it provided full function. However, the cuticles of dead embryos only had mild defects in head involution, comparable to those seen when crossing +/cno^R2 males and females (Fig 5B vs A). Thus, CnoCanoeIDR provided substantial but not full function, perhaps due to effects of the small disruptions generated in the IDR at the two ends of our replacement allele. CnoAfadinIDR provided substantially less function. When we crossed cnoAfadinIDR/cno^R2 males and females, we saw 64% lethality (n=675), suggesting many cnoAfadinIDR/cno^R2 embryos die. Further, cuticle phenotypes were substantially enhanced, with 45% of embryos exhibiting failure of both head involution and dorsal closure (Fig 5C vs A). These data suggest that the Afadin IDR provides enough function for viability, but its function is not fully wildtype. To further this analysis, we examined the ability of each IDR replacement to stabilize AJs under tension, using as a control the progeny of +/cno^R2 parents. These control embryos had a slightly elevated number of AJ gaps relative to wildtype (Fig 5D,G), as did progeny of cnoCanoeIDR/cno^R2 parents (Fig 5E vs D, green asterisks; quantified in G). In contrast, progeny of cnoAfadinIDR/cno^R2 parents had highly elevated AJ gap number (Fig 5F vs D; quantified in G). There are three genotypes among the progeny (mutant/mutant, mutant/cno^R2, and cno^R2/cno^R2); we think this likely contributes to variability in gap number between embryos. Finally, we scored embryos for ventral furrow invagination defects. 6% of progeny of +/cno^R2 parents have mild defects in this process (McParland et al., 2024a). Progeny of cnoCanoeIDR/cno^R2 parents had a slightly elevated defect frequency (11%), with most defects being minor, whereas progeny of cnoAfadinIDR/cno^R2 parents had a substantially higher incidence of ventral furrow defects (49%; Table S1). Taken together, these data reveal that Afadin’s IDR can function in the Cno context, restoring AJ localization and sufficient function to convert embryonic lethality to adult viability and fertility. However, examination of enrichment at AJs under tension and sensitized functional assays revealed that it does not restore full wildtype function.

ΑlphaFold predicts helical regions in the IDR and some have defined functions in Afadin

Many IDRs contain short motifs serving as binding sites for other proteins. While BLAST searching did not reveal obvious conserved sequences in the proximal IDR, ΑlphaFold provided an alternate approach to look for elements conserved at the structural level. Afadin’s proximal IDR contains four predicted α-helices (Fig 4A, orange boxes; Gurley et al., 2023). The most N-terminal helix precisely overlaps the known binding site for αcat’s M-domain (Maruo et al., 2018). Intriguingly, this helix is clearly absent from Cno’s IDR. Cno only has three predicted α-helices, a nine amino acid glutamine-rich helix and two longer helices. The two longer helices share similarity to two of the Afadin helices, with 33% sequence identity over 72 amino acids (Gurley et al., 2023). Our interest in this conserved region was substantially heightened by recent cryo-EM studies using mammalian proteins. This conserved region of Afadin folds as a forked pair of α-helices and forms a ternary complex with αcat’s actin-binding domain and an actin filament, enhancing their interaction (Gong et al., 2025). We thus hypothesized this region would play an important role in Cno function.

Neither the conserved α-helices nor the IDR’s N- or C-terminal regions are individually essential for viability, but the N- and C-terminal regions are required for full function

We used the two conserved α-helices to subdivide the IDR into three regions: the helices themselves and the regions N-terminal or C-terminal to them. To evaluate how these regions contribute to Cno function, we created a set of mutants (Fig 6A) lacking the conserved helices (cnoIDRΔH), the region N-terminal to them (cnoIDRΔN; ending 14 amino acids before the predicted helices), or the region C-terminal to them (cnoIDRΔC; beginning 9 amino acids after the predicted helices; Methods/Fig S1). All three proteins accumulated at roughly wildtype levels (Fig S2E, F).

Fig 6. Neither the conserved α-helices nor the regions of the IDR N- or C-terminal are individually essential for viability, but the N- and C-terminal regions of the IDR are required for full wildtype function.

(A). Drosophila Cno and the IDR deletion mutants. (B-E). Maternal/zygotic adults of the indicated genotype. (F-I). Distribution of cuticle phenotypes in different mutants, using cuticle examples in Fig 1. cno^R2/+ data is from Gurley et al. (Gurley et al., 2023). (J-M). Junctional gaps, Stage 7. Asterisks=all junctional gaps/disruptions. Arrows=selected gaps. (N). Gap quantification. n=14 fields of view.

Our initial hypothesis was that the conserved helices would be essential. We first examined whether each IDR region was essential for adult viability, crossing each mutant to our null allele. Strikingly, cnoIDRΔH/cno^R2, cnoIDRΔN/cno^R2, and cnoIDRΔC/cno^R2 were all adult viable and produced viable adult progeny. After outcrossing we obtained homozygous mutant lines of each (Fig 6B–E). However, while cnoIDRΔH/cno^R2 progeny were obtained at Mendelian ratios (33% versus 33% expected; Balancer homozygotes die; n= 602), both cnoIDRΔN/cno^R2 and cnoIDRΔC/cno^R2 flies appeared at less than Mendelian ratios (19% for cnoIDRΔN/cno^R2, n= 514; 12% for cnoIDRΔC/cno^R2, n= 523). Thus, none of the three IDR regions is essential for viability or fertility, but reduced viability suggested two regions contribute to full function.

We used our sensitized assay to test this further, crossing males and females transheterozygous for each allele and cno^R2. While the conserved helices were not essential for viability, we still suspected they would be the most important region of the proximal IDR. However, in crosses of cnoIDRΔH/cno^R2 males and females, we saw 29% lethality (n=785), only slightly above the 25% expected if it provided full function. Cuticle phenotypes of the dead embryos were only slightly more severe than those of cno^R2 zygotic mutants, with 81% of embryos exhibiting defects restricted to head involution (Fig 6H vs F). Thus CnoIDRΔH provides nearly wildtype function. In contrast, the other two mutants provided less function. Crossing cnoIDRΔN/cno^R2 males and females led to 34% lethality (n=690), while crossing cnoIDRΔC/cno^R2 males and females led to 53% lethality (n=866). In both cases, cuticle phenotypes were substantially enhanced, with 38% (Fig 6G) or 41% (Fig 6I) of embryos having both dorsal closure and head involution failure.

Deleting the proximal IDR led to frequent gaps at multicellular AJs (Fig 3I). To determine which IDR regions are required to stabilize AJs under tension, we examined AJ gap frequency in each mutant, crossing mutant/cno^R2 males and females, with progeny of +/cno^R2 parents as a control. cnoIDRΔH/cno^R2 mutants had a slightly elevated gap frequency, similar to our +/cno^R2 control (Fig 6L vs J, quantified in N) or to cnoCanoeIDR/cno^R2 mutants. Mean gap frequency trended higher in cnoIDRΔN/cno^R2 mutants but did not reach statistical significance (Fig 6K vs J, quantified in N). Gap frequency was substantially elevated in cnoIDRΔC/cno^R2 mutants (Fig 6M vs J, quantified in N). Finally, we scored ventral furrow invagination (Table S1), with progeny of +/cno^R2 parents as a baseline (6% defects; McParland et al., 2024a). Progeny of cnoIDRΔH/cno^R2 parents exhibited a slightly elevated frequency of defects (11%, all mild; Table S1). Defects were more elevated in progeny of cnoIDRΔN/cno^R2 parents (19%; Table S1). In contrast, both the frequency and severity of defects were substantially elevated in progeny of cnoIDRΔC/cno^R2 parents (52%; Table S1). Together these data reveal that none of the individual IDR regions are essential to restore viability and fertility. While the conserved helices appear largely dispensable for function, the N- and C-terminal IDR regions are required for full wildtype function, with the C-terminal region playing a more important role.

The IDR’s C terminal region is important for nuclear exclusion and together with the N-terminal region contributes to TCJ enrichment

CnoΔProxIDR protein had reduced localization to AJs, loss of TCJ enrichment and surprising mis-localization to nuclei (Fig 1). We thus examined whether any individual proximal IDR region was important for AJ localization. All three mutant proteins clearly localized to AJs at stage 9 (Fig 7A vs B–D). In cross section, both CnoIDRΔN and CnoIDRΔH proteins were strongly enriched at apical AJs (Fig 7F,G vs E, yellow arrows). We next examined TCJs, where wildtype Cno is enriched in response to tension (Fig 7I, yellow vs cyan arrows). CnoIDRΔH enrichment at TCJs relative to bicellular junctions was unchanged from wildtype (Fig 7K, yellow vs cyan arrows, quantified in M). However, CnoIDRΔN enrichment at TCJs was reduced, though the protein retained ~2-fold enrichment (Fig 7J, yellow vs cyan arrows, quantified in M).

Fig 7. The C terminal region of the IDR is important for nuclear exclusion and with the N-terminal region contributes to TCJ enrichment.

(A-D). Stage 9. All mutant proteins localize to AJs, although CnoIDRΔC also localizes to nuclei (D, arrow). (E-H). Cross-sections, apical up, stage 6. Yellow arrows=junctional localization. Cyan arrows=nuclei. (I-L). Tricellular junction enrichment. Yellow arrows=TCJs, cyan arrows=bicellular junctions. (M) Quantification of Tricellular junction enrichment. n=14 embryos. (N). Relative accumulation of CnoIDRΔC at AJs. n=19 embryos. (O, P). Stage 8. Cyan arrows=cells rounded up for mitosis. Yellow arrows=nondividing cells.

In contrast, CnoIDRΔC protein localization was substantially altered. While it accumulated strongly at apical AJs (Fig 7D, H yellow arrow), cross-sections revealed it also accumulated in nuclei (Fig 7H, cyan arrow), resembling CnoΔProxIDR in this regard. TCJ enrichment of CnoIDRΔC protein was essentially eliminated (Fig 7L yellow vs cyan arrows, quantified in M), but AJ levels of CnoIDRΔC protein were not significantly different from wildtype (Fig 7N), in contrast to CnoΔProxIDR. While CnoIDRΔC and CnoΔProxIDR exhibited similar accumulation in nuclei, other aspects of localization differed. CnoΔProxIDR accumulation in AJs of non-dividing cells was substantially reduced relative to wildtype (Fig 1S; 7O yellow arrows), whereas CnoIDRΔC protein accumulated in AJs of non-dividing cells at essentially wildtype levels (Fig 7P, yellow arrows). In addition, as cells round up for mitosis and nuclei disassemble, CnoΔProxIDR protein is released into the cytoplasm and accumulates at high levels at the mitotic cell cortex (Fig 1S, 7O, cyan arrows). In contrast, CnoIDRΔC protein resembles wildtype Cno, with somewhat reduced levels in mitotic cells (Fig 7P, cyan arrows). This may reflect the fact that since less CnoIDRΔC is sequestered in nuclei, there is less released when nuclei breakdown during mitosis, making it more similar to wildtype.

The conserved helices are not sufficient for Cno localization or function

Thus, both the N- and C-terminal proximal IDR spacers are dispensable for viability. Since the two helices are the only long conserved sequence in the proximal IDR and since, in Afadin, they stabilize αcat interaction with actin (Gong et al., 2025), we tested the hypothesis that they are sufficient for IDR function. To do so, we created a mutant retaining only the conserved helices in the proximal IDR (cnoIDRHelixOnly; Fig 8A), by combining the regions deleted in CnoIDRΔN and CnoIDRΔC (Fig 6A). CnoIDRHelixOnly accumulated at near wildtype levels (Fig S2A, B). cnoIDRHelixOnly was not viable over our null allele, cno^R2 (0/558 progeny), ruling out the idea that the helices are sufficient for function. However, unlike cno^R2, but similar to cnoΔProxIDR, cnoIDRHelixOnly was not zygotically embryonic lethal (2% lethality, n=720), suggesting it might retain a small amount of function.

Fig 8. The conserved helices are not sufficient for Cno localization or function.

(A). Drosophila Cno and CnoHelixOnly mutant. (B). Distribution of cuticle phenotypes in different mutants. cno^R10/+ data is from (Gurley et al., 2023). (C-F). cnoHelixOnly embryos, anterior left, antigens indicated. (C, E). Defects in ventral furrow invagination (yellow arrows). (D, E). Defects in resuming columnar cell architecture (cyan arrows). (F). Failure of dorsal closure (yellow arrow) and head involution (cyan arrow). (G). Junctional gaps, Stage 8. Asterisks=all junctional gaps/disruptions. Arrows=selected gaps. (H). Gap quantification. n=14 fields of view. (I). Cross-sections, apical up, stage 6. Yellow arrows=junctional localization. Cyan arrows=nuclei. (J, K). TCJ enrichment. Yellow arrows=TCJs, cyan arrows=bicellular junctions. n=14 embryos. (L). Stage 8. Cyan arrows=cells rounded up for mitosis. Yellow arrows=nondividing cells.

To evaluate CnoIDRHelixOnly function, we generated maternal-zygotic mutants. We observed 55% lethality (n=727), consistent with fully penetrant maternal-zygotic lethality and substantial zygotic rescue. We next assessed function in morphogenesis by examining cuticles. Strikingly, cnoIDRHelixOnly maternal-zygotic mutants had strong defects in morphogenesis that were, on average, more severe than those of cnoΔProxIDR (Fig 8B). 70% of embryos exhibited failure of both head involution and dorsal closure, in contrast to only 31% of cnoΔProxIDR mutants. However, morphogenesis defects were not as severe as maternal-zygotic null mutants, with only 5% having additional defects in epidermal integrity, unlike the 68% of null mutants (Fig 8B). Next, we looked at embryos during development. cnoIDRHelixOnly mutants had moderate to strong defects in ventral furrow invagination (Fig 8C, E, yellow arrows; 78% defective, n=82; Table S1). cnoIDRHelixOnly mutants also had delays in resuming columnar cell architecture after later cell divisions (Fig 8D, E, cyan arrows) and had defects in both head involution and dorsal closure (Fig 8F, cyan vs yellow arrows), like strong cno mutants. To look more closely at underlying AJ defects, we examined AJ gaps. cnoIDRHelixOnly mutants had an exceptional number of AJ gaps, with many if not most multicellular junctions affected. These were more frequent than those seen in cnoΔProxIDR (Fig 8G, quantified in H) and included broader gaps along aligned AP borders (Fig 8G, red asterisks).

CnoIDRHelixOnly protein accumulated in nuclei, with reduced AJ localization (Fig 8I–J). TCJ enrichment of the mutant protein was virtually eliminated (Fig 8J, yellow arrows, quantified in K). During germband extension, AJ localization in non-mitotic cells remained relatively weak (Fig L, yellow arrows), but, as we saw for CnoΔProxIDR protein, in mitotic cells CnoIDRHelixOnly accumulated at high levels at the cortex (Fig 8L, cyan arrows). Thus, the conserved helices are not sufficient for Cno localization or for Cno function, and some phenotypes of cnoIDRHelixOnly mutants are more severe than those of cnoΔProxIDR.

The proximal IDR also plays an important role in Cno function during eye development

Cno also plays important roles in postembryonic development. For example, in the developing eye a field of imaginal disc cells re-organizes through orchestrated cell shape changes into an array of identical ommatidia. Each has a defined number of cone, primary (1°), secondary (2°), and tertiary (3°) cells, and bristles, arranged in precise relationships to one another (Fig 9A; Johnson, 2021). AJ proteins and their regulators play important roles in the eye (Johnson, 2021). Complete loss of Cno dramatically disrupts the epithelium (Walther et al., 2018), while milder mutants have less severe defects (e.g. Matsuo et al., 1997; Matsuo et al., 1999; Miyamoto et al., 1995; Takahashi et al., 1998). Cno’s regulator Rap1 also plays important roles (Yost et al., 2023). Even cno alleles that are viable and fertile, like cnoΔPDZ, exhibit defects in ommatidial cell number and arrangement (McParland et al., 2024a; McParland et al., 2024b; Perez-Vale et al., 2021), providing another tissue in which to assess function.

Fig 9. Multiple regions of the IDR are important for Cno function in the developing eye.

(A). Diagram of the complex cell arrangement in a wildtype ommatidium with a key to defects observed in mutants. (B-J). Representative pupal eyes from each genotype, highlighted to illustrate defects observed. (K). Quantification of mean number of defects per ommatidium with statistical analysis of difference from wildtype CnoGFP. Box and whisker plot. Dots are individual eye discs. n= 10 eye discs with 77–115 total ommatidia per genotype. Ordinary one-way ANOVA using Šídák’s multiple comparisons test.

We thus generated transheterozygous animals with each allele over cno^R2, using GFP-tagged wildtype Cno as a control. +/cno^R2 animals do not have significant numbers of defects relative to wildtype (McParland et al., 2024b). Eye discs were dissected at 40 hours after puparium formation, and fixed, stained and imaged for Ecad and N-cadherin, and also imaged in the GFP-channel to visualize mutant Cno proteins. We scored defects in the stereotyped cell arrangement using an established scoring system (Johnson and Cagan, 2009), enumerating defects in the number and arrangement of cone, 1°, and 3° cells, and bristles per ommatidium (Table S2). We scored 10 eye discs with 77–115 total ommatidia per genotype.

Wildtype eyes have occasional defects: ~0.5 defects per ommatidium. Wildtype GFP-tagged cno has a defect frequency similar to wildtype (Perez-Vale et al., 2021). We replicated that here (Fig 9B, K; 0.14 defects/ommatidium), and it was our baseline control. For cnoCanoeIDR and cnoAfadinIDR we scored animals heterozygous with cno^R2, as we never obtained a homozygous cnoAfadinIDR line. cnoCanoeIDR/cno^R2 mutants had defect numbers that were not significantly elevated (Fig 9C, K; 0.58 defects/ommatidium). In contrast, cnoAfadinIDR/cno^R2 mutants had a substantially elevated number of defects (Fig 9D, K; 3.8 defects/ommatidium). We next examined mutants with smaller deletions. cnoIDRΔH/cno^R2 mutants had defect frequency similar to wildtype (Fig 9F, K; 0.48 defects/ommatidium). In contrast, both cnoIDRΔN/cno^R2 mutants (Fig 9E, K) and cnoIDRΔC/cno^R2 mutants (Fig 9G, K) had substantially elevated numbers of defects (3.0 and 3.8 defects/ommatidium, respectively). With these mutants we also could look at homozygotes and saw similar, though weaker, effects (Fig 9H–K; cnoIDRΔN=1.5 defects/ommatidium; cnoIDRΔC=1.9 defects/ommatidium; vs cnoIDRΔH=0.32 defects/ommatidium). These results precisely paralleled each of our sensitized assays in embryos, with CnoCanoeIDR and CnoIDRΔH retaining near wildtype function, CnoIDRΔN modestly reduced in function and CnoAfadinIDR and CnoIDRΔC providing significantly less than wildtype function.

Both the conserved helices and the FAB can bind F-actin

The IDR’s surprising robustness raised significant questions. No individual region of the proximal IDR or FAB was essential for viability and all mutants retained substantial function. We were particularly surprised at the dispensability of the conserved helices, as they are the only region of Afadin known to bind F-actin directly (Carminati et al., 2016; Gong et al., 2025). In contrast, while the C-terminal IDR is traditionally referred to as the FAB (most recently defined as the C-terminal 77 amino acids; Sakakibara et al., 2020), no one ever directly assessed whether it can bind F-actin.

We used an F-actin co-sedimentation assay to test if recombinant proteins of specific regions in the Cno IDR bind F-actin. Following centrifugation, actin filaments pellet (Fig 10A), and one can assess F-actin association by evaluating whether proteins co-sediment. αcat’s actin-binding domain was a positive control. When centrifuged alone, it remained in the supernatant, but 70% pelleted when mixed with F-actin (Fig 10A, E). The only previous assessment of Cno’s ability to bind F-actin used a long construct extending from just before the conserved helices to the end of Cno (Sawyer et al., 2009), similar to the construct originally used to define Afadin’s F-actin binding (Mandai et al., 1997). We replicated our earlier work, using a fragment of Cno extending from 6 amino acids N-terminal to the conserved helices to Cno’s C-terminus (H+C+FAB; Fig S5). This was N-terminally tagged with GST for purification and mixed with F-actin. 32% co-sedimented with F-actin (Fig 10B, F). We next examined whether adding αcat’s actin binding domain enhanced interaction of H+C+FAB with F-actin. The increase did not reach statistical significance (Fig 10B, F). When we purified H+C+FAB, we obtained both the full-length fusion protein (Fig 10B, blue arrowhead) and several C-terminal degradation products (Fig 10B, red arrowhead). Intriguingly, the truncated products co-sedimented less effectively, as we saw previously (Sawyer et al., 2009), suggesting some C-terminal sequences in this fragment are important for F-actin association.

Fig 10. Both the conserved helices and the FAB can bind F-actin.

(A-D) SDS-PAGE of the soluble (S) and pellet (P) fractions of different αcat or Cno constructs with or without F-actin. (A). The αcat actin-binding domain co-sediments with F-actin. (B). A Cno IDR fragment from the conserved helices to the C-terminus (H+C+FAB) co-sediments with F-actin, and this is not enhanced by addition of αcat. (C). The conserved helices (H) modestly co-sediment with F-actin, and this is not enhanced by addition of αcat. (D). The FAB peptide co-sediments with F-actin, and this interaction is enhanced by addition of αcat. (E). Quantification revealing that αcat actin-binding domain co-sedimentation is not substantially enhanced by any Cno peptide. n= 3 gels. One way ANOVA followed by Dunnett’s multiple comparison test. (F). Quantification of co-sedimentation of different Cno peptides with or without α-catenin. n= 4 gels. Two way ANOVA followed by Tukey’s multiple comparisons test. (G). Structural alignment of AlphaFold 3 models of the Cno (red) and Afadin (orange) FAB on actin. One F-actin protofilament is colored in grey, the other in green. (H). Zoomed in view of Cno FAB from yellow boxed region in G, showing conserved residues. (I). Sequence alignment of the Drosophila Cno and rat Afadin H regions, with identity shown in green font, similarity in yellow font. AlphaFold 3-predicted helices are indicated above or below the respective sequence.

Next, we examined two IDR regions which, in Afadin, were either previously demonstrated to bind F-actin (the conserved helices (H)) or have a name suggesting binding experiments were done (the FAB). We generated constructs allowing us to express and purify these two peptides (Fig S5, S6A), and first used circular dichroism to examine whether they are helical in solution. As a positive control, we examined Cno’s Dilute domain which produced minima at 208 and 222 nm, well aligned with its predicted secondary structure (Fig S6B; McParland et al., 2024a). Afadin’s H region has a forked helical structure in the F-actin:αcat:Afadin structure (Gong et al., 2025). While many individual helices adopt structure upon binding a partner, the Cno H peptide showed helical character on its own (Fig S6C). To determine if this was due to oligomerization, we analyzed the oligomeric state of the H peptide using SEC-MALS, which revealed a single monomeric species (Fig S6E). In contrast, the FAB peptide had a single minimum at 200 nm, indicative of a disordered peptide (Fig S6D). As the FAB has predicted helices, it may adopt helical character upon binding a partner.

To directly assess F-actin binding, we tested each fragment for F-actin co-sedimentation. After purification, both the H and FAB peptides ran as two bands (Fig S7A, C), and we used mass-spectrometry to identify the precise nature of each. The H region peptides were a full-length version and one lacking the C-terminal 9 amino acids (Fig S7A, B), while the FAB peptides were full-length and one lacking the C-terminal 28 amino acids (Fig S7C, D). We first quantified the longer of the two peptides in each case. The H peptide exhibited modest co-sedimentation with F-actin, which was above the background seen in F-actin’s absence, and adding αcat did not enhance H peptide co-sedimentation (Fig 10C, F). In contrast, 55% of the FAB peptide co-sedimented with F-actin and adding αcat significantly increased co-sedimentation (to 76%; Fig 10D, F). While both the full-length and truncated H peptides bound actin similarly (Fig S7E), the longer of the two FAB peptides found F-actin more effectively (Fig S7F), suggesting the 28 C-terminal amino acids removed in the shorter fragment enhance actin association.

The Alushin and Takai labs found that Afadin’s coiled-coil (CC) region enhances association of αcat’s actin-binding domain with F-actin (Gong et al., 2025; Sakakibara et al., 2020). However, the mammalian CC region contains two sub-regions, an alpha-helix that binds the M-domain of αcat (CC-N) and the recently revealed forked helices that binds both F-actin and αcat’s F-actin binding domain (CC-C; Gong et al., 2025). Cno’s IDR lacks the former region. Only Gong et al. assessed CC-C separately and found that it significantly enhanced αcat binding to F-actin. In contrast, we found no enhancement by Cno’s H-peptide (Fig 10E). We also assessed whether the FAB enhanced αcat binding to F-actin. There was a trend toward enhancement, but it did not reach statistical significance (Fig 10E). These data suggest possible differences in αcat or actin interaction of the fly and mammalian proteins.

These results, combined with what is known about Afadin, suggest at least two distinct regions of Cno’s IDR bind F-actin: the conserved helices (H) and the FAB. ΑlphaFold 3 allows us to model protein interactions. We used this tool to model potential interactions of H or the FAB with F-actin. Cryo-electron microscopy revealed that the H region of Afadin binds both αcat’s actin-binding domain and F-actin, stabilizing this interaction (Gong et al., 2025). In this structure, it forms a forked structure of two α-helices (Fig S8B). The H peptide is reasonably well conserved between mammals and Drosophila and all four residues important for interactions of the mammalian H peptide with αcat and F-actin (Gong et al., 2025) are perfectly conserved (Fig S8A). We used ΑlphaFold 3 to model the Cno H peptide bound to a Drosophila actin filament along with the actin-binding domain of Drosophila αcat. This produced a high-confidence model (as assessed by pLDDT and PAE; Fig S8D–E) very similar to that reported for the Afadin H peptide for the region of overlap (Fig S8B vs C). A conserved PW-R motif on the N-terminal helix engaged residues on actin and one αcat molecule, while a conserved LE—F—R motif in the second helix engaged a second αcat (Fig S8C, F, G). Thus, we think it is likely that the H peptide interacts with αcat and F-actin in similar modes in both animals.

Next, we used ΑlphaFold 3 to model potential interactions of the Cno FAB peptide with actin filaments and compared this with a model of the Afadin FAB and F-actin. While the Cno and Afadin predictions varied in placement of the N- and C-terminal regions of each FAB, both predictions converged with confidence in the placement of a highly conserved FAB region on one actin subunit in the filament (Fig 10G–I; Fig S9A, B). This FAB helix-turn-helix motif is predicted to use conserved hydrophobic and charged residues to engage features of actin, primarily on subdomain 1 (Fig 10G, H). These data are consistent with our biochemical assays and suggest the FAB indeed is a second actin binding site.

Discussion

Intrinsically disordered regions (IDRs) play important roles in assembling multiprotein complexes with diverse functions from transcriptional regulation to cell signaling to cell-cell and cell-matrix adhesion. Many proteins in the AJ and TJ contain IDRs. However, the roles of the IDRs and how their complex sequence structure contributes to their function in vivo remains a key question in the field. Here we address this using the AJ-cytoskeletal linker Cno, defining the IDR’s roles in protein function and localization, and then taking it apart, defining sequence features important for its biochemical and cell biological functions.

Cno’s IDR plays important roles in protein function

Many junctional proteins include IDRs, but in only a few cases have their functions been explored. The most extensive characterization was of the TJ scaffold ZO-1. It phase separates in vitro and recruits other TJ proteins into condensates. ZO-1’s C-terminal IDR is important for TJ localization and assembly of functional TJs (Beutel et al., 2019). TJ puncta become more continuous via actin interaction (Beutel et al., 2019) and spread along the apical junction complex by interactions with PATJ (Pombo-Garcia et al., 2024). Afadin’s C-terminal IDR can also phase separate in vitro and in vivo and plays a role in localizing Afadin to apical AJs (Kuno et al., 2025). Functional studies of the Afadin (Kuno et al., 2025) or Cno IDR (Gurley et al., 2023; Perez-Vale et al., 2021) are more limited, with studies in cultured MDCK cells suggesting a role in maintaining columnar cell architecture (Kuno et al., 2025).

Here we define the mechanistic roles of Cno’s IDR in vivo during morphogenesis. We and others divided the C-terminal IDR into two regions: the relatively conserved C-terminal FAB and the longer, less well conserved region referred to here as the proximal IDR. Deleting the FAB had no effect on Cno AJ localization, including its enrichment in AJs under tension, suggesting regions outside the FAB confer F-actin association. Further, the FAB is dispensable for viability and fertility, though sensitized assays revealed CnoΔFAB does not retain fully wildtype function (Perez-Vale et al., 2021). Here we deleted the proximal IDR. The effect was much more dramatic. cnoΔProxIDR mutants were embryonic lethal with strong defects in multiple events of morphogenesis. AJs under tension were destabilized. However, CnoΔProxIDR retained some residual function, with reduced defects in epidermal integrity relative to the null allele. Thus, the IDR ranks with the RA1 domain as the most critical region of Cno protein assessed so far.

Cno’s IDR is important for AJ localization and its deletion leads to nuclear relocalization

Localization of CnoΔProxIDR was strikingly altered: AJ localization was substantially reduced, and instead the protein re-localized to nuclei. We initially wondered if this was a biological artifact caused by creation of a novel NES. However, recent analysis of Afadin revealed nuclear localization of a construct similar to CnoΔProxIDR (“AfadinΔC”; Kuno et al., 2025). Prominent nuclear localization is apparent in their Figs 3, 4, and 6. Our data together with these suggest there is a nuclear export signal (NES) in the C-terminal spacer of the proximal IDR, as CnoIDRΔC also accumulates in nuclei. Two contrasting but not mutually exclusive hypotheses are consistent with the observed phenotype. First, deleting the IDR may lead to sequestration in nuclei, preventing an otherwise functional protein from localizing to AJs. Second, AJs may be the preferred site of Cno localization because multivalent interactions with other AJ proteins retain it there and the IDR may play a direct role in these interactions. Deleting the proximal IDR may reduce affinity for AJs, releasing Cno and allowing its nuclear import.

Our data using the nuclear export inhibitor leptomycin revealed that acutely blocking all nuclear export reduces the localization of wildtype Cno to AJs and leads to its accumulation in nuclei. This supports the first hypothesis. However, our full data set suggest a more complex situation. While CnoIDRΔC localizes prominently to nuclei, it also retains strong localization to AJs and rescues viability and fertility. This suggests nuclear localization per se does not lead to sequestration, preventing a protein lacking the NES from localizing to AJs. CnoAfadinIDR also has weak nuclear localization, suggesting its NES is not fully functional in Drosophila. However, it also rescues viability. Thus, we suspect sequestration is only part of the reason CnoΔProxIDR fails to localize to AJs and has substantially reduced function. It may have a reduced ability to interact with AJ partners, possibly as part of AJ biomolecular condensates, or it may retain full function when not sequestered. In the future, it will be important to test this by blocking nuclear import of CnoΔProxIDR, or tethering it to the plasma membrane, as we did with Armadillo/beta-catenin (Cox et al., 1999).

Multiple junctional proteins have known or suspected roles in transcriptional regulation (Arm/βcat (Peifer, 1997); p120ctn (Daniel and Reynolds, 1999); ZO-1 (Balda et al., 2003)). Our data raise the question of whether Cno has any normal functions in nuclei. There are predicted nuclear localization signals between the RA domains (Rothenberg et al., 2023) and when we expressed an RA-PDZ protein in Drosophila it localized to nuclei (Bonello et al., 2018). However, we have never observed wildtype Cno in nuclei during any stage of embryogenesis, though we cannot rule out low level accumulation there. There is a long history of reports of Afadin localizing to both AJs and nuclei, beginning with its initial cloning by Takai’s lab (Mandai et al., 1997) and continuing in papers from that lab and others (e.g., Buchert et al., 2007; Miyahara et al., 2000; Nishioka et al., 2000; Ooshio et al., 2010), but in most cases it was either described as an observation whose significance was not clear or dismissed as an artifact. The short isoform of Afadin lacking the FAB and part of the conserved IDR α-helices is more abundant in nuclei than full length Afadin (Buchert et al., 2007; VanLeeuwen et al., 2014). This aligns with our Cno data revealing that the region C-terminal to the IDR helices is important for nuclear exclusion. There are a few examples in the literature suggesting Afadin has a nuclear role. Activating the neurotransmitter receptor NMDA (VanLeeuwen et al., 2014) or the Estrogen receptor (Sellers et al., 2018) can trigger nuclear translocation of Afadin, indirectly promoting Histone H3 phosphorylation. Akt-mediated phosphorylation of Serine 1718 in Afadin’s FAB can also trigger nuclear accumulation of Afadin with effects on transcription and cell migration (Elloul et al., 2014; Xu et al., 2015). However, while the Akt recognition site is in a conserved region, Cno has a non-phosphorylatable Alanine in place of S1718. In the future, it will be important to define whether Cno has nuclear roles. Tethering Arm/βcat to the plasma membrane largely rescued AJ function but failed to rescue its nuclear role in Wnt signaling (Cox et al., 1999). Similar experiments with Cno could help resolve potential nuclear roles.

Different regions of Cno’s IDR contribute combinatorially to localization and function

Previous work revealed the surprising robustness of Cno. Proteins individually lacking folded domains or conserved regions, including RA2, the DIL domain, the PDZ, and the FAB, restored viability and fertility, though all were required for full function in sensitized assays (McParland et al., 2024a; McParland et al., 2024b; Perez-Vale et al., 2021). Further, all these mutant proteins localized to AJs, and only the RA domains were essential for mechanosensitive recruitment to AJs under tension. These data supported a model in which Cno is recruited to AJs by diverse multivalent interactions, with many individual interactions not essential for localization or function.

In contrast, CnoΔProxIDR had substantially reduced AJ localization, making the proximal IDR the first Cno region with a clear role in this. To define the underlying mechanism, we took the IDR apart. Our series of smaller IDR deletion mutants reveal that different regions of the IDR must act combinatorially to mediate AJ location and Cno function. cnoIDRΔN, cnoIDRΔH, cnoIDRΔC, and cnoΔFAB are all viable and fertile mutants and all proteins localize effectively to AJs. However, sensitized assays suggest all except cnoIDRΔH fail to provide fully wildtype function. Two mutant proteins, CnoIDRΔN and CnoIDRΔC, have partial or strong reduction in enrichment in AJs under tension, consistent with their modest deficits in function. These data further emphasize the resilient and multivalent nature of Cno localization to AJs. In this way, Cno resembles the apical polarity protein Baz/Par3, as no Baz protein domain is essential for cortical localization, though each contributes to cortical anchorage (McKinley et al., 2012).

Given this, what are the precise roles of each region of the IDR—N, Helices, C, and FAB— and how do they contribute to function? The conserved helices have well defined biochemical roles in Afadin—cryo-EM reveals they bind F-actin and αcat’s actin-binding domain to stabilize this interaction. Sequence conservation, F-actin binding assays, and modeling are consistent with these helices playing a similar role in Drosophila Cno. Thus, we were quite surprised at the modest deficits in function seen in cnoIDRΔH, which was the least affected of our mutants. Further, the cnoIDRHelixOnly mutant reveals the conserved helices are not sufficient for function, while its slight “dominant-negative” phenotype suggests the helices alone can disrupt function of AJs. Perhaps CnoIDRHelixOnly can bind some, but not all, of its necessary protein partners, thus sequestering them and further disrupting AJ-cytoskeletal linkage. The FAB, despite its name, was not previously demonstrated to bind F-actin. Our data reveal it does so, and that its interaction with F-actin is further enhanced by αcat. However, like cnoIDRΔH, cnoΔFAB is also viable and fertile. Superimposing the predicted Cno FAB structure on the Cno H-αcat-F-actin model revealed steric clash, primarily between Cno FAB and αcat (Fig S9C), suggesting mutually exclusive binding to any one site on the filament. Thus, the ability of αcat to enhance FAB F-actin binding may be due to interactions across neighboring binding sites on the lattice, or allosteric changes to the F-actin lattice that enhance binding to proximal sites on the filament. We note that the >260 residues between the Cno H region and the FAB would enable both regions to engage non-overlapping sites on the actin filament lattice, or even a second, neighboring actin filament. One possibility is the conserved helices and the FAB act in a redundant fashion, with each interacting with F-actin and perhaps αcat. In the future it will be important to test a mutant removing both regions.

The functional roles of the IDR N- and C-terminal “spacers” are also intriguing. Each spacer differs in length, sequence composition and charge between Afadin and Cno. The N-terminal spacer has no sequence identity between Afadin and Cno, while the C-terminal spacer only shares a single short, conserved motif. One role they could play is simply to provide space/volume between more conserved “stickers” like the conserved helices and FAB. Consistent with this, Afadin’s IDR largely replaced Cno function, at least to the extent of restoring viability and fertility. Neither spacer is individually required for viability or fertility, and thus their individual “spacer” functions are not essential. However, deleting both strongly reduced Cno function, consistent with roles as spacers and/or functional redundancy in interactions with binding partners. In the future, it will be interesting to scramble the sequence of one or both spacers and see if function is retained—this would help distinguish between roles as spacers or embedded sequence motifs.

Our work also revealed two clear functions of the “spacers”. First, the C-terminal spacer appears to contain an NES, as deleting it led to nuclear accumulation. We used the LocNES software (Xu et al., 2015) to predict NES sequences in Cno and the highest ranked (HPISNLAKELNQLTM; score =0.67) is in this region and the hydrophobic resides are conserved in Dipera. Despite limited sequence identity in this region, Afadin’s IDR largely but not completely restored nuclear exclusion. Second, the C-terminal region is critical for Cno’s mechanosensing. Deleting it or replacing it with that region from Afadin, virtually eliminated enrichment of mutant Cno proteins at TCJs. This loss correlated with reduced function in our sensitized assays, but, to our surprise, proteins lacking this mechanosensitive recruitment still restored viability. The only other mutants with altered TCJ enrichment are cnoIDRΔN and mutations in the RA-domains (McParland et al., 2024b; Perez-Vale et al., 2021)—future work to define mechanosensing mechanisms will be key.

Ultimately, the question of the robustness of AJ:cytoskeletal linkages is a broader one. While we interpreted the phenotypes of our cno mutants by suggesting combinatorial effects of binding interactions (compensation) mediated by individual domains, things are likely more complex. We need to be careful about defining molecular mechanisms solely from organismal phenotypes that operate at very different scales. Future experiments will need to combine this work with super-resolution imaging and other approaches to fully define how the network of proteins that mediate adhesion-cytoskeletal interactions at AJs act at the supra-molecular level.

Methods

Cloning of cnoΔIDR constructs

Most constructs were generated using Azenta Life Science (Waltham, MA, USA) cloning and mutagenesis service. Details and sequences are in Fig S1. cnoΔIProxIDR was generated by removing all nucleotides corresponding to amino acids G1130-Q1926 in the Canoe protein using cnoWT-GFP as the starting construct (Perez-Vale et al, 2021). Then using Azenta Life Science’s gene synthesis service, we generated a small sequence containing a 5’ SbfI RE site and a 3’ SpeI RE site with a small Gly-Gly linker to maintain the proper reading frame (5’-GCCTGCAGGGGCACTAGT-3’) and subcloned that into the deletion generated above. This new cnoΔProxIDR-SbfI-SpeI construct was then used to generate all the subsequent mutants via RE cloning using those SbfI and SpeI overhangs. This construct was verified by PCR (ΔProxIDR forward primer 5’ – GCGTGGTAATCGCCGTATGTC – 3’ and ΔProxIDR reverse primer 5’ – CAACACAATCTCGACGCCT – 3’) to confirm sequence and further verified by RE digest and western blot analysis. cnoIDRΔN was generated in a similar manner. deleting all nucleotides corresponding to amino acids G1130-V1566, cnoIDRΔC removed amino acids A1671-Q1926, cnoIDRΔH removed amino acids A1567-I1670, and cnoIDRHelixOnly removed all amino acids except those between A1567-I1670. All of these constructs were generated by Azenta Life Sciences. The cnoCanoeIDR construct was generated by adding back amino acids G1130-Q1926 into the SbfI-SpeI site resulting in a final product that contained linker amino acids PAGG upstream and GTS downstream of the endogenous IDR sequence. cnoAfadinIDR was generated similarly by adding nucleotides corresponding to amino acids A1074-A1727 of the rat Afadin protein into the SbfI-SpeI sites. All constructs were verified by DNA sequencing. Each construct also contains the w⁺ selectable marker and is flanked by attR and attL sites and was integrated into the attP site at the cnoΔΔ locus (Bloomington Drosophila Stock Center stock 94023; Perez-Vale et al., 2021). These constructs were sent to BestGene (Chino Hills, CA, USA) for embryo injections. DNA was injected into PhiC31/int^{DM. Vas}; cnoΔΔ embryos (BDSC stock 94023). F1 offspring were screened for the presence of the w⁺ marker and outcrossed to w; TM6B, Tb/TM3, Sb (BDSC stock 2537) to generate a balanced stock over TM3. Each stock was verified by PCR, sequencing and western blot analysis. We outcrossed these stocks to a yw stock with a wild type 3^rd chromosome to remove potential passenger mutations from the mutant chromosomes. This occurred for multiple generations selecting for the linked w⁺ marker in each generation. We were thus able to generate homozygous stocks for some of our mutant alleles.

Fly work

All experiments were performed at 25°C unless noted otherwise. Flies of the yellow white genotype [Bloomington Drosophila Stock Center (BDSC), stock 1495] were used as controls and are referred to in the text as wildtype (WT). Maternal/zygotic mutants of cnoΔProxIDR and cnoIDRHelixOnly were made using the FRT/ovoD approach. Third-instar larvae generated by crossing cnoΔProxIDR/TM3 or cnoIDRHelixOnly/TM3 virgin females to P{ry[+t7.2]=hsFLP}1, y[1] w[1118]; P{neoFRT}82B P{ovo−D1−18}3R/TM3, ry[*], Sb[1] males were heat-shocked in a 37°C water bath for 2 hours each on two consecutive days. Then, virgin female progeny with the germline genotype hsFLP1; P{neoFRT}82B P{ovo−D1−18}3R/P {neoFRT}82B cnoΔProxIDR were collected and subsequently crossed with cnoΔProxIDR/TM3 Sb males. The embryos generated from this cross were analyzed. Maternal/zygotic mutants identified by absence of staining with our Cno antibody which recognizes an epitope in the proximal IDR, and referred to in the text as simply cnoΔProxIDR. Sensitized assays were done using mutant flies that are heterozygous for each cno-mutant allele with the null allele cno^R2 as described in Perez-Vale et al, 2021. cno RNAi used the progeny of females carrying two copies of the matGAL4 driver and the TRiP line GL00633 in the v22 vector targeting cno (Bloomington stock 38194), as in Manning et al., 2019.

Immunoblotting and quantification

Table S3 lists the antibodies and dilutions used. We determined accumulation levels wildtype Cno, GFP-tagged mutants, and α-tubulin by immunoblotting embryonic lysates that were collected at 4–8 hours after egg laying. Embryonic lysates were generated as in (McParland et al., 2024b). Briefly, embryos were collected into 0.1% Triton X-100, dechorionated for 5 minutes in 50% bleach, and washed three times with 0.1% Triton X-100. Next, lysis buffer (1% NP-40, 0.5% Na deoxycholate, 0.1% SDS, 50 mM Tris pH 8.0, 300 mM NaCl, 1.0 mM DTT, HaltTM Protease and Phosphatase Inhibitor Cocktail (Thermo Fisher Scientific, #78442) (100×), and 1 mM EDTA) was added, embryos were ground with a pestle for ~20 seconds and then placed on ice. After 10 minutes on ice, embryos were once again ground with a pestle for ~20 seconds and subsequently centrifugated at 16,361×g for 15 minutes at 4°C. Protein concentration was assessed using the Bio-Rad Protein Assay Dye, recording absorbance at 595 nm with a spectrophotometer. Protein lysates were run on 7% SDS-PAGE and proteins transferred onto nitrocellulose membranes. Membranes were blocked using 10% bovine serum albumin (BSA) diluted in Tris-buffered saline with 0.1% Tween-20 (TBST) for 1 hour at room temperature. Primary and secondary antibodies were diluted in 5% BSA diluted in TBST. Primary antibody incubations were performed overnight at 4°C, and secondary antibody incubation was performed for 45 minutes at room temperature. We used the Odyssey CLx infrared system (LI-COR Biosciences) to image the membranes, and band densitometry was carried out using Empiria Studio^® Software (LI-COR Biosciences).

Embryo fixation and immunofluorescence

Embryo fixation and immunofluorescence were performed as follows. Flies were placed into cups containing apple juice agar plates with added yeast paste. Embryos were then collected using a paint brush and 0.1% Triton-X-100. They were then bleached to remove the chorion layer using 50% bleach and washed three times with 0.03% Triton X-100 with 68 mM NaCl. They were then fixed in 95°C in a 0. 03% Triton X-100 with 68 mM NaCl and 8 mM EGTA salt solution for 10 seconds and brought back down to temp on ice for 30 minutes. The vitelline membrane was removed by vigorous shaking in a 1:1 solution of n-heptane and 95% methanol:5% EGTA and then washed three times with a 95% methanol and 5% EGTA solution. To prepare for staining, embryos were washed three times with 5% normal goat serum (NGS; Thermo Fisher Scientific) and 0.1% saponin in phosphate-buffered saline (PBS) (PBSS-NGS) and then blocked for an hour using 1% NGS in PBSS-NGS. Antibodies were diluted (see antibody dilution table below) in a PBS solution containing 1% bovine serum albumin and 0.1% saponin and incubated overnight at 4°C. Embryos then received three 15-minute washes with PBSS-NGS and incubated in secondary antibodies for 2 hours at room temperature. Embryos were then mounted on glass slides using a homemade Gelvatol solution (recipe from the University of Pittsburg’s Center for Biological Imaging). Embryos were imaged using a 40X/NA 1.3 Plan-Apochromat oil objective on a LSM 880 confocal laser-scanning microscope (Carl Zeiss, Jena Germany) running on ZEN Black 2009 software. Using Photoshop (Adobe, San Jose, CA) all images were adjusted to have uniform input levels, brightness, and contrast.

Embryo permeabilization and drug treatment

For embryos in Fig. 2 and Fig S4A–C, embryo permeabilization for drug treatment was performed according to Perez-Vale et al., 2021, with slight deviations. Flies were allowed to lay eggs on apple juice agar plates with yeast paste for 4.5h-overnight at 25°C. Embryos were collected and dechorionated in 50% bleach and washed two times with 0.9% NaCl. We then removed the solution and incubated for 2 hours in 1:1 octane/0.9% NaCl while rotating on a nutator with either control (i.e., water) treatment, or 10μg/ml Leptomycin B treatment (Target Mol, T15735); as in Townsley and Bienz, 2000). Embryos were then washed two times with heptane followed by a 30-min incubation in 1:1 heptane:8% formaldehyde in PBS+8 mM EGTA. We then devitellinized the embryos by vigorously shaking in 1:1 heptane:95% methanol with 5% EGTA. The embryos where then washed three times with 95% methanol/5% EGTA and stored in this solution at −20°C until ready to be immunostained as described above. For Fig. S4D, E, embryos were permeabilized with D-limonene (National Diagnostics; Histoclear; Perez-Vale et al., 2021). Embryos were collected after 4 hours at 25°C on yeasted agar plates. Embryos were then dechorionated using 50% bleach and washed once with water. They were then incubated from 30 min to an hour at room temperature vigorously shaking at 250 rpm in 1:1 Histoclear/heptane with 10 μg/ml of Leptomycin (Target Mol, T15735) drug treatment or control (i.e., water) treatment. After incubation the embryos were immediately heat shocked and heat fixed and immunostained as described above.

Automated TCJ/multicellular junction enrichment analysis

We performed membrane segmentation using the EpySeg graphical user interface (GUI) (Aigouy et al., 2020), which employs a pre-trained deep learning model for image segmentation. Specifically, we utilized the “v2” model from the segmentation_models library (https://github.com/qubvel/segmentation_models), with the following architectural specifications: Model Architecture: Linknet, Backbone Network: VGG16, Activation Function: Sigmoid, and Input Image Dimensions: Dynamically adjusted (0 width and height). The segmentation was performed on representative images, with predictions generated based on the Armadillo staining channel of de-identified images to achieve a single-blinded analysis. After membrane segmentation, we used FIJI (ImageJ) and the Analyze Skeleton plugin (Arganda-Carreras et al., 2010) (https://github.com/fiji/AnalyzeSkeleton) to identify the intersections of tricellular and multicellular borders. The intersection points were then isolated and used as reference points of interest for subsequent analysis. We next used the ModularImageAnalysis (MIA) software developed in the Wolfson Bioimaging Facility at the University of Bristol (Cross et al., 2024). This software allows for fully unique and customizable workflows and analysis. Using this, we loaded the segmented Armadillo image from EpySeg as the skeleton marking all cell borders. We also inputted the location of tri- and multi-cell borders from FIJI. We then preformed two iterations of dilations of these points to make a larger area of the border intersections. This larger area was then subtracted from the skeleton to give us disconnected bi-cells with the ends of the bi-cell edges no longer overlapping the cell-intersection area determined previously and with sufficient separation of signal for larger TCJs. Lastly, we used the GFP channel as the input. We preformed manual thresholding of the signal, elimination of background signal/noise, and manually selected a 512 pixel by 512 pixel area to perform the analysis on (representing one-quarter of the full 1024 image). Using the GFP signal we measured the signal intensity along each bi-cell border as well as its associated tri- and multi-cell junction and compared the ratio of these two intensity values to determine the enrichment at TCJs.

Analysis of junctional gaps

For each genotype, we selected closeups of similar magnifications of seven stage 7 and seven stage 8 embryos stained to visualize Arm, which outlines AJs, and selecting lateral views from the ventral midline toward the amnioserosa. These were then coded, randomized, and scored blind for gaps in tricellular or multicellular junctions, or places where the junctional protein signal was broadened. Samples were then unblinded, and total gap number was then recorded per field of view.

Statistics and quantification

Graphs were made using GraphPad Prism version 10.4.1 (GraphPad Software, Boston, MA, USA). For box and whisker plots, boxes represent 25th–75th percentile ratios and whiskers represent the 5th–95th percentiles. For bar charts, Bar = mean, error bars represent 95% Confidence Intervals. Statistical analyses were performed in Prism using Welch’s unpaired t-test or Brown–Forsythe and Welch ANOVA test (TCJ and Gap analysis) and unless otherwise noted, error bars represent 95% Confidence Intervals. The Welch’s test is parametric and assumes a normal distribution. Data distribution was assumed to be normal but this was not formally tested. Kolmogorov-Smirnov analysis was used for the comparison of GFP levels between CnoIDRΔC and CnoIDRAfadin compared to CnoWT-GFP. Significance of eye disc defects was calculated using ordinary one-way ANOVA using Šídák’s multiple comparisons test. For biochemical experiments the following statistical tests were used. Fig 10E, S7E, and S7F: One way ANOVA followed by Dunnett’s multiple comparison test. Fig. 10F: Two way ANOVA followed by Tukey’s multiple comparisons test.

Cuticle preparation and analysis

Embryonic cuticles were prepared according to (Wieschaus and Nüsslein-Volhard, 1986). Females were placed in cups and eggs were collected on apple juice agar plates with yeast for less than 24 hours at 25°C. Eggs were then aligned on a fresh apple juice agar plate without yeast and incubated at 25°C for 48 hours to allow embryos to develop fully and viable embryos to hatch. Unhatched embryos were collected in 50% bleach and dechorionated for 5 minutes. The bleach was removed and they were transferred to glass slides into 100μl of 1:1 Hoyer’s medium:lactic acid. The slides were incubated at 60°C for 24–48 hours. They were then stored at room temperature. Images were taken using a Nikon Labophot with a 10x Phase 2 lens, and captured on an iPhone, and placed into categories based on morphological criteria, including counting unfertilized eggs and removing them from the total. At least two separate crosses were used for each genotype.

Scoring eye phenotypes

Flies were maintained at 25°C and prepupae selected and stored in humidified chambers until dissection at 40h after puparium formation (APF). At 40h pupae were submerged in a droplet of 1.5x PBS, removed from the pupal casing, and decapitated. The head casing was then removed to expose the brain complex, and it was rinsed in 1.5x PBS twice to remove any excess fat and tissue. The brain complex was then fixed in 3.7% formaldehyde for 20 min, washed in 1.5x PBS twice, and blocked in 1.5x PBST for 30 min. The tissue was then placed in rat anti E-Cadherin (1:50; DSHB) and rat anti N-Cadherin (1:50; DSHB) overnight at 4°C before being rinsed in 1.5x PBS twice and blocked in 1.5x PBST for 30 min. The tissue was then stained with Goat anti-Rat 568 secondary (1:100; Life Tech) to detect AJs and retinas and imaged with a Zeiss LSM 980 with Airyscan 2 using a 63x/1.4 oil Plan Apochromat objective. Patterning errors were scored in 10 eye discs per genotype spanning 77–115 ommatidia per genotype. Data were analyzed for statistical significance using PRISM, normality was determined using the Kologorov-Smirnov test, and significance was calculated using an ordinary one-way parametric ANOVA. Images were processed for publication using FIJI and Adobe Photoshop.

Cloning and purification of Canoe fragments and the Alpha Catenin actin binding domain

DNA encoding the Drosophila melanogaster Cno H region (residues 1567–1670) and Cno FAB (residues 1967–2051) were generated using PCR (H region forward primer, 5’-GGCAGGACCCATATGGCGTCCAATCAAGGGAACAATCGTC; H region reverse primer, 5’-GCCGAGCCTGAATTCTTA ACTGACCTGACCGCCGCCAGCCAATC--3’; FAB forward primer, 5’-GACTATCGTCATATGGCTGCCCTCTGGAACAC-3’; FAB reverse primer, 5’-GACTATCGTCTCGAGTTAGTGCACCGCCGCGTCTATATC-3’;) and sub-cloned into pET28 (Millipore Sigma, Burlington, MA) using Nde1 and EcoR1 restriction enzymes and T4 DNA ligase (New England Biolabs, Ipswich, MA). DNA encoding Drosophila melanogaster Cno H+C+FAB (residues 1560–2051) was generated using PCR (H+C+FAB forward primer, 5’-CGCTGTTCGGGTACCATGGGATCCCCGCCAAAGGGCAGCTATGTGGCGTCC-3’; H+C+FAB primary reverse primer, 5’-GCCGAGCCTGAATTCTTAGTGCACCGCGTCTATATCTCGTTG-3’; secondary reverse primer to append a C-terminal His₆ tag, 5’-GCCGAGCCTGAATTCTTAGTGGTGATGATGGTGATGGTGCACCGCGTCTATATCTCG-3’) and sub-cloned into pGEX-6p2 using BamH1 and EcoR1 restriction enzymes and T4 DNA ligase (New England Biolabs). A full length Drosophila alpha-catenin cDNA was obtained from the DGRC on a Whatman FTA disc, stock number FI19832 (DGRC Stock 1649238; https://dgrc.bio.indiana.edu//stock/1649238; RRID:DGRC_1649238) and was used as the template for PCRs that amplified the Drosophila alpha-catenin long ABD (Dm long α-cat: residues 667–917) (Dm long α-cat forward primer, 5’-CGCTGTTCGGCTAGCGATCAAACCGTTGACGAATATCCCGATATAAG-3’; Reverse primer 5’-CACTCGTTCAAGCTTTTAAACAGCGTCAGCAGGACTCTGG-3’) coding region, which was then sub-cloned into pET28 (Millipore Sigma, Burlington, MA) using NheI and HindIII restriction enzymes and T4 DNA ligase (New England Biolabs, Ipswich, MA). All plasmids were sequence verified.

Plasmids were transformed into Escherichia coli BL21 DE3 pLysS cells and grown to an optical density at 600 nm of 0.8 in medium containing antibiotic (50 μg/l kanamycin or 100 μg/l ampicillin for respective vectors) at 37°C. The temperature was lowered to 20°C (excluding the GST-Cno H+C+FAB-H₆ construct, which was kept at 37°C), and protein expression induced with 100 μM IPTG for 16 h. Cells were harvested by centrifugation, resuspended in buffer A [25 mM Tris-HCl pH 8.5, 300 mM NaCl, 10 mM imidazole and 0.1% β-mercaptoethanol (β-ME)] at 4°C and lysed by sonication (the GST-Cno H+C+FAB buffer A was supplemented with 1% Triton X-100 and 10% glycerol). Phenylmethylsulfonyl fluoride was added to 1 mM final concentration. Cells debris was pelleted by centrifugation at 17,000 × g for 1 h. Supernatants for constructs cloned in the pET28 vector were loaded onto a 10 ml Ni²⁺-NTA column (Qiagen, Hilden, Germany). The column was washed with 1 L buffer A, and the protein batch eluted with 100 mL buffer B (buffer A supplemented with 290 mM imidazole). Supernatant for the GST-Cno H+C+FAB-H₆ construct was loaded onto a 10 ml Glutathione Sepharose 4 Fast Flow column (Cytiva, Marlborough, MA). The column was washed with 1 L of buffer N [25 mM Tris-HCl pH 8.5, 300 mM NaCl, and 0.1% β-ME] and the protein batch eluted with 100 mL buffer N supplemented with 50 mM Glutathione. Eluate was then loaded onto a 10 ml Ni²⁺-NTA column, washed, and eluted as described above. N-terminally His-tagged proteins received 5 μl (1.2 mg/ml) of bovine α-thrombin (Prolytix, Essex Junction, VT) and 1 mM CaCl₂, and were incubated at room temperature for 3 hours and then 4°C for 48 h to proteolytically cleave off the N-terminal His₆ tag. Protein was then filtered consecutively over 0.5 ml benzamadine sepharose (Cytiva, Marlborough, MA) and 10 ml Ni²⁺-NTA resin (Qiagen, Hilden, Germany), and the flowthrough was collected. All constructs were dialyzed against 4L buffer N for 20 hours using 3.5K molecular weight cutoff (MWCO) SnakeSkin dialysis tubing (Thermo Fisher Scientific, Waltham, MA). Protein was concentrated in a Millipore Sigma 3K MWCO centrifugal concentrator, aliquoted and stored at −80°C. Canoe Dilute Domain (aa 613–1006) cloning and purification was described in McParland, et al. 2024

F-actin co-sedimentation assays

F-actin co-sedimentation assays were adapted from Gong et al., 2025 as follows. Lyophilized rabbit skeletal muscle actin (Cytoskeleton, Inc., Denver, CO) was reconstituted to 10 mg/ml (233 μM) with 100 μL deionized water and then diluted to 20 μM in Actin Buffer (2 mM Tris pH 8.0, 0.1 mM CaCl₂, 0.2 mM ATP, 0.5 mM DTT) and stored at −80°C in 112 μL aliquots. Actin was polymerized by mixing 112 μL of thawed monomeric actin with 12.6 μL 10x KMEI buffer (10X = 500 mM KCl, 10 mM MgCl₂, 10 mM EGTA, 100 mM imidazole pH 7.0) supplemented with 5 mM ATP and incubated at room temperature for 1 hour (final Actin concentration of 18 μM). Cno fragment constructs were diluted in 1X KMEI buffer to a final concentration of 13.5 μM, either alone or mixed with 9 μM αcat ABD construct, and pre-cleared by ultracentrifugation at 80,000 rpm (278,000 × g) in a TLA-100 rotor (Beckman Coulter, Brea, CA) for 15 minutes at 4°C. 31.1 μL of the supernatant was mixed with either 38.9 μL of 18 μM polymerized actin or 38.9 μL of 1X KMEI buffer to give final concentrations of Cno at 6 μM, αcat at 4 μM, and actin at 10 μM. Mixtures were incubated for 30 minutes at room temperature, then ultracentrifuged at 80,000 rpm (278,000 × g) for 30 minutes at 4°C. ¾ of the supernatant (52.5 μL) was mixed with 17.5 μL of 4x SDS-PAGE loading buffer. The remaining ¼ of the supernatant was removed, the pellet was washed twice with 1X KMEI buffer, and then dissolved in 4/3 volume of 1x SDS-PAGE loading buffer (93.3 μL). Samples were separated by SDS-PAGE, stained with Coomassie Brilliant Blue, scanned on a LI-COR Odyssey scanner (LI-COR Biosciences, Lincoln, NE), and quantified with FIJI.

Size exclusion chromatography - multi-angle light scattering

A 100-μl sample of the D.m. Cno H region construct (aa 1567–1670; 231 μM) was injected onto a Superdex 200 10/300 GL gel filtration column (Cytiva, Marlborough, MA) in 25 mM Tris pH 8.5, 100 mM NaCl, 0.1% β-ME, 0.2 g/L sodium azide, and run in-line with a Wyatt DAWN HELIOS II light scattering instrument and a Wyatt Optilab t-rEX refractometer (Waters | Wyatt Technology, Santa Barbara, CA). Molecular weight was calculated with light scattering and refractive index data using the Wyatt Astra V software package (Waters | Wyatt Technology Corp.). SECMALS data presented are representative of experiments conducted in duplicate.

Circular dichroism

D.m. Cno constructs (Dilute domain (aa 613–1006); IDR H region (aa 1567–1670); FAB (aa 1967–2051)) were exchanged into 10 mM sodium phosphate, pH 7.4, 50 mM sodium fluoride and diluted to a final concentration of 0.2 mg/ml in 300 μl volume. Initial circular dichroism spectra of constructs were collected at 20°C using a Chirascan-V100 spectrometer (Applied Photophysics, Leatherhead, UK). Spectra were recorded from 260 to 185 nm with a step size of 0.5 nm using a 1 mm-path-length cuvette. The time per point was maintained at 1.25 sec. CD melt spectra were obtained, recording values at 208 and 220 nm (as well as 200 nm for the FAB construct) in 1°C steps from 20–94°C with the time per point maintained at 1.25 sec. After each melt, a final spectrum was recorded from 260 to 185 nm at 94°C for each construct. A base-line CD spectrum of the buffer alone was taken and subtracted from each spectrum. CD data presented are representative of experiments conducted in duplicate.

AlphaFold 3 structural modeling

Structural modeling was performed using AlphaFold 3 via alphafoldserver.com (Cross et al., 2024). To generate a model of the Cno H region bound to F-actin in the presence of the high-affinity alpha-catenin actin binding domain region, the following were input to AlphaFold 3: D.m. actin 5C (6 copies), 6 ATP molecules, 6 Mg²⁺ ions, D.m. αcat (aa 719–917, 3 copies), D.m. Cno (aa 1567–1670, 2 copies). The best model produced an F-actin structure composed of two three-subunit protofilaments with three alpha-catenin actin binding domains and two Cno H regions bound, similar to the 9dva cryo-EM structure of the mammalian homologs (Gong et al., 2025). The composition of the model presented in figures was trimmed to show five actin subunits (bound to ATP and Mg²⁺), two alpha-catenin actin binding domains, and one Cno H region, akin to the chains presented in the 9dva structure. To generate a model of the Cno FAB region bound to F-actin, the following were input to AlphaFold 3: D.m. actin 5C (6 copies), 6 ATP molecules, 6 Mg²⁺ ions, D.m. Canoe (aa 1934–2051, 1 copy). The best model produced an F-actin structure composed of two three-subunit protofilaments with one Canoe FAB region bound. To generate a model of the Afadin FAB region bound to F-actin, the following were input to AlphaFold 3: Rattus norvegicus (R.n.) beta-actin (6 copies), 6 ATP molecules, 6 Mg²⁺ ions, R.n. Afadin (aa 1712–1829, 1 copy). The best model produced an F-actin structure composed of two three-subunit protofilaments with one Afadin FAB region bound. PyMOL (Schrödinger, New York, NY) was used for structural alignment and image generation. Predicted Aligned Error (PAE) matrices were generated using PAE Viewer (Elfmann & Stülke, 2023).

Detailed author contributions

Corbin Jensen led analysis of protein localization and embryonic phenotypes, with contributions from Noah Gurley, Avery Mathias and Yufei Xiao. Noah Gurley carried out immunoblotting analysis. Avery Mathias analyzed pupal eye phenotypes. Leah Wolfsberg and Kevin Slep carried out the biochemical analyses, with help from Zixi Zhou. Leptomycin experiments were done by Corbin Jensen, Maik Bischoff and Sarah Clark. The manuscript was written by Mark Peifer and Kevin Slep, with editorial input from all of the authors.

Supplementary Material

Online Supplemental Material

Fig S1 describes the design of our cno mutant constructs. Fig S2 uses immunoblotting to assess accumulation levels of mutant proteins. Fig S3 reveals that CnoΔProxIDR localization is not restored by association with wildtype Cno and release from nuclei in mitotic cells does not fully restore normal localization of CnoΔProxIDR. Fig S4 reveals that Leptomycin treatment results in the accumulation of wildtype Cno in nuclei. Fig S5 describes our Cno bacterial expression constructs. Fig S6 reveals that the Cno H region is monomeric and helical while the FAB is disordered, Fig 7 includes Mass spectrometry revealing the nature of C-terminal truncations of the purified Cno H region and the FAB. Fig S8 includes modeling predicting that the Cno H region engages F-actin and α-catenin. Fig S9 includes modeling predicting that the Cno FAB engages F-actin. Table S1 quantifies ventral furrow defects in our different mutants. Table S2 includes detailed data on ommatidial defects in our different mutants. Table S3 includes primary and secondary antibodies used.