Mud is a large structural protein containing long, continuous coiled coil regions at the N-terminus, which are separated from the Pins-binding domain (Mud^PBD) and a putative microtubule-interacting motif at the C-terminus (Figure 1A). Alternative splicing uniquely yields an additional, small coiled-coil domain adjacent to the Mud^PBD (referred to here as “Mud^CC”) within the spindle orienting isoform expressed in NBs (Siller et al., 2006). We previously described a phosphorylation-sensitive intramolecular interaction between the Mud^CC and Mud^PBD domains that regulates its cortical localization and spindle orientation in wing disc epithelial cells (Dewey et al., 2015b), suggesting the tandem Mud^CC-PBD cassette may function as a concerted structural unit. Coiled-coils are regarded as multifaceted protein-protein interaction domains (Truebestein and Leonard, 2016), thus we performed mass spectrometry on samples isolated from pulldowns using recombinant GST:Mud^CC as bait and Drosophila S2 cell lysate as prey to identify new Mud^CC-binding proteins that may contribute to such functions. Analysis in Scaffold4 software of these results yielded 25 unique proteins, among which was Hts [2 peptides at strict 99.9% protein and 95% peptide thresholds; also see (Cutillas and Johnston, 2021)]. Although neither a direct functional link nor physical interaction have been formally established previously between Hts and Mud, both proteins are known to localize to spectrosomes, cytoskeletal structures that play essential roles in the ACD of germline stem cells (Yu et al., 2006), suggesting the association identified here could have functional implications in vivo.

To corroborate the mass spectrometry findings, as well as probe if the interaction is direct, we next conducted equilibrium binding pulldown experiments using recombinant, purified proteins in vitro. Hts was fused to Maltose-binding protein (MBP) and coupled to amylose resin as the solid phase bait; Mud^CC-PBD was purified using a cleavable hexahistidine tag and used as the soluble prey component. As shown in Figure 1, Mud^CC-PBD bound to full-length Hts (Hts^FL), with saturation binding isotherm experiments demonstrating a robust, sub-micromolar affinity. We next performed domain mapping experiments with Hts to identify the Mud binding site. The Hts N-terminal “Head” domain (Hts^HEAD) contains a class II adolase fold (Hts^ALDO) that is known to have lost enzymatic activity but remains otherwise poorly understood functionally (Matsuoka et al., 2000). This is followed by the “Neck” domain (Hts^NECK) that contributes to oligomerization, with tetramers constituting the most likely functional unit. Finally, a C-terminal “Tail” domain (Hts^TAIL), which includes a Myristoylated alanine-rich C-kinase substrate (MARCKS) region, binds F-actin and associates with additional components of the cortical actin cytoskeleton as well as other proteins (Matsuoka et al., 2000). Truncation of neither the Hts^TAIL alone nor in combination with the Hts^NECK impaired binding, implicating the Hts^HEAD as the primary site of Mud interaction (Figure 1B). In support of this inference, Mud binding affinity to the Hts^HEAD domain, as well as the isolated Hts^ALDO subdomain, was statistically indistinguishable from that of Hts^FL (Figure 1C). Finally, we examined Mud binding to an N-terminal truncation that deletes the Hts^HEAD domain, which showed a loss of binding compared to Hts^FL (Figure 1D). We conclude that Mud^CC directly binds the Hts^HEAD domain, specifically within the aldolase fold region.

Hts localizes to the spectrosome, a cytoskeletal structure that controls spindle anchoring and microtubule polarization in asymmetrically dividing germline stem cells (Lin et al., 1994; Robinson et al., 1994), yet to our knowledge a role in somatic stem cells such as neural stem cells has not been described. Confirmation of direct Mud binding thus prompted us to next investigate Hts expression and localization in central brain NBs. We first used an anti-Hts antibody to examine endogenous protein in third instar larval central brains (Zaccai and Lipshitz, 1996). This analysis revealed extensive Hts expression throughout brain lobes, including in Miranda (Mir)-positive NBs (Figure 2A). This result was similar to studies of Hts expression in the optic lobe at the same developmental stage (Ohler et al., 2011). Hts signal was significantly diminished throughout central brains dissected from the hypomorphic hts ⁰¹¹⁰³ mutant [involving a P-element insertion that significantly reduces Hts protein expression (Ohler et al., 2011)]. Hts staining was also reduced in NBs expressing interfering RNA against hts (hts ^RNAi ) using the NB-specific 1407 ^GAL4 driver (Figures 2A–C). Finally, to determine if Hts potentially plays a functional role in NBs, namely, in their ACD, we counted the number of Mir⁺ NBs in brain lobes among genotypes, as loss of polarity and/or spindle orientation genes often leads to abnormal NB numbers. Indeed, both mutant genotypes showed a reduction in NB population compared to control, which contained the stereotypical ∼100 central brain NBs per lobe (Figure 2D; Supplementary Figure S1). Hts function is explored further in the subsequent sections. Closer examination revealed that Hts was cortically localized in most cells, including NBs and their surrounding progeny, consistent with its association with cortical actin (Matsuoka et al., 2000). Unfortunately, however, this made assessing polarization of endogenous Hts localization in individual NBs challenging and ambiguous.

To circumvent the ubiquitous antibody labeling of Hts, we next expressed a transgenic UAS:mCherry-Hts fusion, again under the control of NB-specific 1407 ^GAL4 driver and examined localization of the native mCherry fluorophore. Notably, mCherry-Hts was apically polarized in mitotic NBs, opposite of the basal Mir crescent (Figure 3). Further inspection of NBs throughout the cell cycle revealed that Hts is nonpolarized with additional cytoplasmic localization during interphase, subsequently becoming polarized early in mitosis (e.g., prophase prior to nuclear envelope breakdown) where it remains through metaphase and anaphase (Figure 3). At telophase, Hts remained apical, segregating primarily into the nascent self-renewing NB progeny cell. At this stage, however, Hts also appeared to begin delocalizing back into the cytoplasm. Interestingly, this cell cycle-associated pattern of Hts localization closely mirrors that of aPKC (Oon and Prehoda, 2019), suggesting a potential functional link with the core apical polarity complex. We conclude that Hts is highly expressed in the larval central brain and a component of the apical polarity domain in mitotic NBs.

Having established Hts expression in the larval central brain and ability to apically polarize in dividing NBs, we next examined whether it contributes to localization of the core apical-basal polarity components. We found that NB-specific expression of hts ^RNAi did not alter apical localization of aPKC nor basal localization of Mir (Figure 4A). Similar results were obtained using the hts ⁰¹¹⁰³ mutant allele. We further found that Pins also formed normal apical crescents following loss of Hts (Figure 4B). Overall, we conclude that Hts does not contribute to the establishment or maintenance of the core Par-mediated apical-basal polarity in NBs.

Having established that Hts directly binds Mud and is capable of apically polarizing in NBs, we next investigated its role in spindle positioning. Using the intact apical aPKC and basal Mir crescents as markers of the polarity axis, we found that expression of hts ^RNAi resulted in a spindle orientation defect similar to the expected impairment seen in response to mud ^RNAi as well as pins ^RNAi (Figure 5). Spindle angles were measured relative to the aPKC crescent center, with 0° representing ideal orientation and 90° indicating complete misalignment (Figure 5A). Plotting of spindle angles measured from numerous individual NBs as a cumulative percentage found a similar distribution between hts, mud, and pins knockdown, with each genotype showing spindle angles broadly distributed relative to control cells that cluster at lower, more precise angles of orientation (Figure 5B). Similar defects were also found in hts ⁰¹¹⁰³ . We conclude that Hts functions as a spindle positioning component likely through its interaction with Mud. However, that loss of Hts did not impact the localization of Pins the apical domain (Figure 4) suggested Hts could potentially impact Mud-dependent spindle orientation through a unique mechanism, which is explored further in the remaining sections.

The required role for Hts in spindle positioning, despite a lack of effect on Pins polarity, prompted us to next explore more explicitly the mechanism for how Hts impacts function of the Pins/Mud complex. Considering its established role in organization of the actin cortex in other cell types (Matsuoka et al., 2000), we considered the possibility that Hts may influence the formation and behavior of the Pins/Mud complex within the intricate network of interactions comprising the apical polarity domain. Liquid-liquid phase separation (LLPS) has become an increasingly recognized mode of organizing supramolecular protein complexes, including those at the cell cortex, resulting in the formation of dense liquid-like droplets (Case et al., 2019). Furthermore, LLPS can facilitate inclusion of specific complex components at the exclusion of others within droplet phases, ultimately impacting their signaling outputs and cellular functions (Banjade and Rosen, 2014; Ditlev et al., 2018). We began by purifying recombinant Mud^CC-PBD and Pins^TPR−Linker protein domains and examining their ability to phase separate in vitro (Figure 6A). We found these proteins underwent LLPS when combined in a minimal biological buffer, forming numerous individual phase-separated droplets without the need for crowding agents such as PEG or Ficoll (Figure 6D). Isolated droplets underwent extensive fusion events over time consistent with liquid-like behavior (Ganser and Myong, 2020), eventually leading to formation of large wetted droplets (Figures 6E, F). Neither Mud nor Pins phase separated when examined alone, suggesting this phenomenon occurs specifically within the framework of an intact Pins/Mud complex (Figures 6B, C). Supportive of this inference was the fact that an E1939A missense mutation within the Mud^PBD that significantly impairs Pins binding was unable to phase separate with Pins under identical conditions [Figure 6G; (Zhu et al., 2011)]. To further corroborate the co-localization of Pins and Mud within observed LLPS droplets, we chemically labeled purified proteins with rhodamine and fluorescein, respectively, at N-terminal cysteine residues using maleimide-based fluorophores. Neither of these labeled proteins underwent LLPS alone (Figures 6H, I), however; when combined they phase separated under identical conditions to their native counterparts (Figures 6J–L). Importantly, the resulting droplets were uniformly positive for both fluorophore signals, providing strong evidence for this being a complex-specific and -driven process.

We next tested whether Pins can phase separate with its other, mutually-exclusive apical binding partner, the polarity adaptor protein Insc (Schober et al., 1999; Schaefer et al., 2000). In contrast to Mud, combining Pins with Insc did not lead to detectable LLPS droplet formation across a range of concentrations and buffer conditions tested with Mud (Figure 6M), spotlighting a specificity of this process with respect to Pins-binding partners. Failure to phase separate was not due to a lack of a direct interaction, as the purified Insc protein bound Pins with an affinity consistent with previous studies and well below the concentrations tested for LLPS [Supplementary Figure S2; (Zhu et al., 2011)]. We conclude that phase separation is likely a unique property of the Pins/Mud complex, with Pins/Insc instead remaining in a single unmixed phase. We suggest LLPS might therefore serve as a mechanism driving formation of discrete complexes within the apical polarity domain.

Phase separation of heterotypic protein complexes is often driven by multivalent interactions, for example, those generated through repeated modular protein interaction domains in one or more components (Banjade and Rosen, 2014; Banani et al., 2017). Through their intrinsic oligomerization, coiled-coil domains necessarily generate a multivalent protein assembly, thus we next questioned whether an intact Mud^CC is required for phase separation with Pins. In silico primary sequence analysis predicted high coiled-coil confidence in Mud^CC apart from a short segment near the domain center, with a parallel trimer suggested as the most probably assembly. To explore this in further detail, we generated a structural model using CCFold, a threading-based algorithm specifically designed to build coiled-coil models from primary sequence input [Figure 7A; (Guzenko and Strelkov, 2018)]. This model was consistent with sequence analysis, revealing a parallel trimer of idealized coiled-coil structure, save for a short interruption near the domain center. We used this model to generate missense mutations at the N- and C-terminal ends predicted to disrupt coiled-coil formation. Specifically, amino acids were targeted at core-facing, hydrophobic “A” positions and mutated to electrostatic acidic residues (L1788E and I1878D; referred hereafter as “ED”). This Mud^ED mutant was highly expressed in Escherichia coli and stable to recombinant purification similar to wild-type protein; however, it eluted from size-exclusion column at a larger volume consistent with disruption of the coiled-coil trimer (i.e., a smaller molecular weight; Figure 7B).

We then examined the ability of the Mud^ED mutant to undergo phase separation when combined with Pins and found that it was completely devoid of this property compared to wild-type Mud under identical conditions (Figure 7C). Notably, however, Mud^ED fully retained its ability to bind Pins through its intact PBD (Figure 7D). These results demonstrate that Mud coiled-coil assembly is required for phase separation of the Pins/Mud complex, most likely through its creation of a multivalent interaction platform for Pins (e.g., a 3:1 stoichiometry assuming a trimeric Mud^CC), and that the Mud^ED mutant decouples Pins binding from phase separation. Finally, we found that the Mud^ED mutant was significantly impaired in its ability to bind Hts (Figure 7E), consistent with the assembled Mud^CC being the site of this interaction. We conclude that although formation of an intact Mud^CC trimer is dispensable for Pins binding, it is required for phase separation as well as Hts binding.

Having identified this distinct phase separation property of the Pins/Mud complex, we next explored how Hts, through its direct interaction with Mud, might interface with this process. We first tested the ability of Hts to bind an intact Pins/Mud complex, using the MBP:Hts^FL fusion with soluble Pins and Mud proteins. As shown in Figure 8A, Hts bound to Mud^CC-PBD but not Pins^TPR−Linker when each was examined alone. When combined, however, Hts was capable of pulling down both Mud and Pins, demonstrating that Hts, Mud, and Pins can mutually exist in a trimeric complex (Figure 8A). This is consistent with distinct binding sites on Mud for Hts and Pins (the CC and PBD regions, respectively). We conclude that Hts directly binds an intact Pins/Mud complex.

We next examined how addition of purified Hts^FL affected the concentration dependence of Pins/Mud phase separation. We found that addition of Hts significantly lowered the threshold concentrations of Pins and Mud necessary for droplet formation (Figure 8B). This is consistent with other recent findings that recruitment of specific “clients” can impact the phase boundary of otherwise autonomous phase-separating complexes (Banani et al., 2016; Banani et al., 2017). To understand the molecular aspects of this Hts-mediated effect, we then performed domain mapping experiments across identical conditions. Addition of the isolated Hts^HEAD domain, which is sufficient for Mud binding (Figure 1), did not impact the concentration regime under which Pins/Mud could phase separate (Figure 8B), revealing that Mud binding alone is insufficient for the effects of Hts and that additional sequence elements are necessary to influence LLPS. Inclusion of the Neck domain (e.g., Hts^HEAD−NECK, also referred to as Hts^ΔIDR) also did not affect phase separation, implicating the Hts C-terminal region as the region responsible for this property. In silico analysis of Hts sequence using PONDR yielded a high-confidence prediction of an intrinsically disordered region spanning the majority of the C-terminal and MARCKS domains [IDR; Supplementary Figure S3; (Romero et al., 2004; Uversky, 2020); http://www.pondr.com], consistent with previous analyses of human Adducin sequence (Gardner et al., 2013; Kruer et al., 2013). Interestingly, while removal of the entire C-terminal Hts^IDR completely abolished its effects on Pins/Mud LLPS, selective truncation of the Hts^MARCKS domain alone produced an intermediate loss-of-function (Figure 8B), suggesting this short sequence within the IDR imparts an important component of its effects. We conclude that Hts facilitates Pins/Mud phase separation by lowering the concentration-dependent phase boundary and that this effect is mediated through the Hts^IDR.

To further explore how Hts might impact Pins/Mud phase separation dynamics, we examined droplet dynamics over time. Unlike those formed in the absence of Hts, which underwent liquid-like fusions within minutes of forming, phase separated droplets in the presence of Hts^FL were resistant to such events, remaining mostly as small individual droplets even after 1 hour of incubation (Figure 8C). The liquid property of LLPS droplets is known to have functional implications and has been suggested as means of preventing fusion among distinct phase separated complexes to maintain specificity (Boeynaems et al., 2018). The core apical polarity complex (e.g., Par3/Par6/aPKC) in NBs was recently shown to phase separate, and thus Hts-mediated alteration of Pins/Mud droplet dynamics may provide such control. We conclude that phase separated Hts/Mud/Pins complexes have altered biophysical properties compared to Pins/Mud.

We next wondered if the Hts^IDR was not only required for its effects on LLPS but also sufficient. However, the Hts^IDR is dispensable for Mud binding, which occurs at the Hts^HEAD domain (Figure 1), precluding a direct examination of its effects. To circumvent this dilemma, we designed a rapamycin based chemical induced dimerization (CID) strategy to fabricate an interaction between Mud and the isolated Hts^IDR (DeRose et al., 2013). Specifically, we cloned and purified the human rapamycin-binding FKBP12 and FRB proteins as N-terminal fusions to Mud^CC-PBD and Hts^IDR domains, respectively (Figure 9A). Similar to native protein, FKBP12:Mud (nor Pins) did not phase separate when examined alone (Figures 9B, C). However, when combined with Pins at a high concentration (50 μM), FKBP12:Mud phase separated similarly to that observed with native proteins (Figure 9D), with droplets undergoing similar fusion events over time as well (Figure 9E). This indicates that FKBP12 fusion does not interfere with the normal LLPS dynamics of the Pins/Mud complex. At lower concentration (15 μM), no detectable LLPS was evident (Figure 9F), again consistent with non-tagged Mud protein seen previously. Addition of FRB:Hts^IDR in the absence of rapamycin (using DMSO as a control) did not trigger phase separation within this low concentration regime (Figure 9G). However, when FKBP12:Mud, Pins, and FRB:Hts^IDR were combined in the presence of rapamycin, small droplets were visible (Figure 9H). Notably, these droplets did not undergo fusion events over time (Figure 9I), similar to Pins/Mud droplets in the presence of Hts^FL (Figure 8C). We also tested the “ED” mutant form of Mud fused to FKBP12 as an additional control. This FKBP12:Mud^ED did not phase separate alone (Figure 9J), nor in combination with Pins (Figure 9K). Finally, addition of FRB:Hts^IDR and rapamycin was not capable of eliciting phase separation of FKBP:Mud^ED when combined with Pins (Figures 9L, M). Together, these results suggest that the Hts^IDR, when physically bound with Mud, is sufficient to facilitate phase separation of the Pins/Mud complex.

Drosophila melanogaster stocks were maintained on standard cornmeal medium at 20°C and crosses were raised at 29°C for all experiments unless otherwise noted.

Driver Lines: 1407Inscuteable GAL4 (BDSC, #8751, RRID:BDSC_8751).

RNAi Lines: Hu Li Tai Shao (Hts) RNAi (BDSC, #38283, RRID:BDSC_38283).

Transgenic Lines: UAS-Hts:mCherry (BDSC, #66171, RRID:BDSC_66171).

Mutant alleles: hts⁰¹¹⁰³ (BDSC, #10989, RRID:BDSC_10989).

Whole brains from third instar larvae were dissected in cold (4°C) PBS followed by fixation for 23 min in 4% paraformaldehyde at room temperature. Tissues were washed three times at room temperature for 10 min in PBS-T (1x PBS, 0.3% Triton) and blocked for 1 h at room temperature using block buffer (PBS-T supplemented with goat and donkey serum 1x PBS, 0.3% Triton, 2.5% goat serum, 2.5% donkey serum) then incubated overnight in primary antibody solution diluted in block buffer at 4°C. Following incubation, tissues were washed three times for 20 min in block buffer followed by incubation in secondary antibody diluted in block buffer overnight at 4°C. Larval brains were mounted ventral side up in 80% glycerol and stored at 4°C until imaging.

The following antibodies were used: Rat Anti-Miranda (1:500) (Abcam, #197788, RRID:AB_2936368), Rabbit Anti-PKC (1:1,000) (Santa Cruz Biotechnology, #sc216, RRID:AB_23000359), Mouse Anti-Hts 1B1 (1:100, DSHB, RRID:AB_528289), Rat Anti-Partner of Inscuteable (Pins) (1:500, generous gift from Dr. Chris Doe; University of Oregon, United States, RRID:AB_2569570).

Images were acquired using a Zeiss LSM-780 confocal microscope equipped with a 63x or 40x oil immersion X 1.40 numerical aperture (NA) objective, with pinhole set to 1 Airy Unit (AU) using Zen Software (Carl Zeiss). Ventral Z-stacks (1-μm steps) were acquired, and analysis was performed using FIJI software. Intensity measurements were performed as described [63] with some modifications. Briefly, a region with an abundant neuroblast population (Hts⁰¹¹⁰³), or individual neuroblasts (Hts^RNAi), were selected using the polygon lasso tool in FIJI and Hts intensity was quantified using the Integrated Density (ID) function. Huygens Essential Deconvolution Suite (SVI) was used to deconvolve images and maximum intensity projections were generated. Figures were assembled in Adobe Illustrator.

cDNA cloning for bacterial protein expression was performed using PCR amplified fragments obtained from an S2 cell cDNA library template. Generation of 6×His fusion proteins (e.g., Mud^CC-PBD, Pins^TPRs-Linker, Insc^Asymm, and various Hts constructs) utilized the pBH plasmid with 5′-BamHI or -BglII and 3′-XhoI or SalI restriction sites. Generation of GST:Pins fusion utilized the pGEX plasmid with 5′-BglII and 3′-SalI restriction sites. Generation of MBP:Hts fusions utilized the pMAL plasmid with 5′-NdeI and 3′-SalI restriction sites. Site-directed mutagenesis was carried out with a standard PCR protocol using KOD-XL DNA polymerase (EMD Millipore, catalog #71087).

All proteins were expressed in BL21 (DE3) E. coli under induction of isopropyl β-D-1-thiogalactopyranoside (IPTG) and grown in standard Luria–Bertani broth supplemented with 100 μg/mL ampicillin. Transformed cells were grown at 37°C to an OD₆₀₀ ∼0.6 and induced with 0.2 mM IPTG overnight at 18°C. Cells were harvested by centrifugation (5,000 × g for 10 min), and bacterial pellets were resuspended in lysis buffer and flash-frozen in liquid nitrogen. Cells were lysed using a Branson digital sonifier and clarified by centrifugation (12,000 × g for 30 min).

For 6×His-tagged proteins, cells were lysed in N1 buffer (50 mM Tris pH8, 300 mM NaCl, 10 mM imidazole) and coupled to Ni-NTA resin (Thermo Fisher Scientific, catalog #88222) for 3 h at 4°C. Following extensive washing, proteins were eluted with N2 buffer (50 mM Tris pH8, 300 mM NaCl, 300 mM imidazole). The 6×His tag was removed using TEV protease during overnight dialysis into N1 buffer. Cleaved products were reverse affinity purified by a second incubation with Ni-NTA resin and collection of the unbound fraction. Final purification was carried out using an S200-sephadex size exclusion column equilibrated in storage buffer (20 mM Tris pH8, 200 mM NaCl, 2 mM DTT).

For GST- and MBP-tagged proteins used as bait in pulldown assays, cells were lysed in lysis buffer (50 mM Tris pH 8, 300 mM NaCl, 2 mM DTT), and lysate was then clarified by centrifugation (12,000 × g for 30 min) and immediately flash frozen in liquid nitrogen for storage at −80°C.

Equivalent amounts of GST- or MBP-fused bait construct lysates (Pins or Hts) were absorbed to glutathione or amylose agarose, respectively, for 30 min at 4°C and washed three times to remove unbound protein. Subsequently, soluble prey proteins were added at varying concentrations for 2 h at 4°C with constant rocking in wash buffer (20 mM Tris, pH 8, 120 mM NaCl, 1 mM DTT, and 0.2% Triton-X100). Incubation for different times (e.g., 1 or 3 h at 4°C, or 1 h at room temperature) produced similar results, indicating that this experimental framework had established equilibrium binding conditions. Reactions were then washed four times in wash buffer, and resolved samples were analyzed by coomassie blue staining of SDS-PAGE gels. All gels shown in figures are representative of at least 3 independent experiments.

All interactions were quantified using ImageJ software. Briefly, gel images were converted to greyscale and individual band intensities were measured using the boxed “Measure” analysis tool. The size of measurement box was kept the same across all concentrations and was initially determined by the size of the largest bound band, typically at the highest concentration tested. To ensure accurate measurements of bound proteins, the intensities of bands for bound prey were normalized to that of the corresponding band for bait protein under each respective condition. Binding curves shown in figures plot these normalized intensities (expressed as arbitrary units, “AU”) as a function of prey protein concentration. Dissociation binding constants were calculated in GraphPad Prism using a one-site binding isotherm regression analysis. All plots and statistics were also performed in Prism.

Proteins were combined at various concentrations in minimal buffer without addition of molecular crowding agents (20 mM Tris, pH 7.5; 120 mM NaCl) and equilibrated at room temperature for 5 min. Solutions were subsequently added to glass microscope slides in chambers assembled from affixed coverslip spacers. Coverslips were applied and samples were imaged at indicated time points for formation of LLPS droplets using a Nikon IX83 brightfield and fluorescence microscope.