NECAPs are negative regulators of the AP2 clathrin adaptor complex

Eukaryotic cells internalize transmembrane receptors via clathrin-mediated endocytosis, but it remains unclear how the machinery underpinning this process is regulated. We recently discovered that membrane-associated muniscin proteins such as FCHo and SGIP initiate endocytosis by converting the AP2 clathrin adaptor complex to an open, active conformation that is then phosphorylated (Hollopeter et al., 2014). Here we report that loss of ncap-1, the sole C. elegans gene encoding an adaptiN Ear-binding Coat-Associated Protein (NECAP), bypasses the requirement for FCHO-1. Biochemical analyses reveal AP2 accumulates in an open, phosphorylated state in ncap-1 mutant worms, suggesting NECAPs promote the closed, inactive conformation of AP2. Consistent with this model, NECAPs preferentially bind open and phosphorylated forms of AP2 in vitro and localize with constitutively open AP2 mutants in vivo. NECAPs do not associate with phosphorylation-defective AP2 mutants, implying that phosphorylation precedes NECAP recruitment. We propose NECAPs function late in endocytosis to inactivate AP2.


Introduction
Eukaryotic cells internalize transmembrane protein cargo, such as laden receptors, by enrobing cargo-containing regions of the plasma membrane with a cytosolic clathrin coat. The GTPase dynamin releases the nascent transport vesicle into the cytosol for subsequent delivery to internal organelles. This fundamental cellular process is called clathrin-mediated endocytosis. The Adaptor Protein 2 (AP2) complex, a heterotetramer of a, b2, m2, and s2 subunits ( Figure 1-figure supplement 1A), actively couples polymerization of the clathrin coat to membrane phospholipids and cargo molecules destined for internalization (Kirchhausen et al., 2014). In this manner, AP2 choreographs turnover of the cell surface and entry into the endolysosomal system, yet it remains unclear how this complex is regulated with spatiotemporal precision to maintain appropriate levels of endocytosis.
AP2 activity is likely regulated at the structural level. Biochemical (Matsui and Kirchhausen, 1990;Rapoport et al., 1997) and structural data suggest that AP2 adopts at least two functionally distinct conformations (Figure 1-figure supplement 1B). In one arrangement, the binding pockets for membrane, cargo, and clathrin are occluded; this orientation is thought to represent a closed, inactive state (Collins et al., 2002). Molecular rearrangement of the AP2 subunits exposes these binding sites and results in an open, active complex that presumably coordinates the formation of endocytic pits (Jackson et al., 2010;Kelly et al., 2014;Kelly et al., 2008).
How then is AP2 inactivated and returned to the cytosol after endocytosis is complete? The heat shock protein Hsc70 (Chappell et al., 1986) and its cofactors (Greener et al., 2000;Umeda et al., 2000;Ungewickell et al., 1995) remove clathrin coats from vesicles. Release of AP2 is also thought to depend on Hsc70 (Hannan et al., 1998) but appears to require additional factors that stimulate the dephosphorylation of m2 (Ghosh and Kornfeld, 2003) and the conversion of vesicular PI(4,5)P 2 into PI(4)P (Cremona et al., 1999;Semerdjieva et al., 2008). Whether there are mechanisms to directly restore the inactive, closed conformation of AP2 has been relatively unexplored.
In this study, we identify adaptiN Ear-binding Coat-Associated Proteins (NECAPs) as AP2 modulators that promote inactivation of the complex. NECAPs were originally identified as endocytic accessory proteins through proteomic analysis of clathrin-coated vesicles (Wasiak et al., 2002) and were shown to bind the AP2 a appendage via a C-terminal WXXF motif (Ritter et al., 2004;Ritter et al., 2003). A conserved N-terminal region of these proteins (the PHear domain) appears to be structurally similar to pleckstrin homology domains and exhibits varying affinity for FxDxF motif-containing endocytic accessory proteins (Ritter et al., 2007). While most invertebrate and fungal genomes encode a single NECAP, vertebrates express two closely-related forms (Dergai et al., 2016;Manna et al., 2015) that are thought to be functionally distinct. Brain-enriched NECAP1 is proposed to modulate the size and number of endocytic structures by regulating the binding of clathrin to the AP2 b2 linker and the coordination of accessory protein recruitment to the a appendage (Ritter et al., 2013). In support of this model, NECAP1 has been localized to clathrin-coated pits at the ultrastructural level (Sochacki et al., 2017). By contrast, ubiquitously-expressed NECAP2 has been proposed to recruit a different clathrin adaptor, AP1, to early endosomes to facilitate fast recycling of receptors back to the cell surface (Chamberland et al., 2016).
Using an unbiased genetic screen in C. elegans, we have unveiled a novel role for NECAPs as proteins that work in opposition of muniscins. We demonstrate that in worms lacking NECAPs (ncap-1 mutants), AP2 accumulates in an open, hyper-phosphorylated state and that heterologous NECAPs restore the closed conformation. NECAPs bind open and phosphorylated forms of the AP2 core but not phosphorylation-defective AP2 mutants, suggesting that NECAPs target the active complex. Together, our genetic and biochemical evidence establish NECAPs as negative regulators of AP2.

Results
Loss of ncap-1 suppresses fcho-1 mutants By mutagenizing worms lacking fcho-1 and selecting for offspring that outcompete their siblings (Hollopeter et al., 2014), we isolated nine independent loss-of-function mutations in ncap-1, which encodes the sole adaptiN Ear-binding Coat-Associated Protein (NECAP) in C. elegans ( Figure 1A). Worms with null mutations in fcho-1 exhibit reduced fitness and require twice as long as wild type worms to populate a culture plate and consume the bacterial food source. Additionally, they display a distinctive 'jowls' phenotype that is indicative of compromised AP2 activity (Gu et al., 2013;Hollopeter et al., 2014). Loss of NCAP-1 suppressed the jowls phenotype ( Figure 1B) and ameliorated the fitness of fcho-1 mutants ( Figure 1C). Expression of fluorescently-tagged NCAP-1 in fcho-1 ncap-1 double mutants restored the fcho-1 phenotype ( Figure 1B and C), confirming that suppression of fcho-1 is due to loss of NCAP-1 function.
Open and phosphorylated AP2 accumulates in ncap-1 mutants We previously discovered that AP2 activity in fcho-1 mutants is also partially restored by amino acid substitutions that specifically destabilize the closed conformation of the adaptor complex, henceforth referred to as 'open AP2 mutations' (Hollopeter et al., 2014). Because loss of NCAP-1 phenotypically mimicked open AP2 mutations, we tested whether AP2 also dwells in the open state in Figure 1. Loss of NCAP-1 suppresses fcho-1 mutants. (A) Gene model of the C. elegans ncap-1 locus. Boxes represent exons. Mutations isolated from the fcho-1 suppressor screen are indicated. The deletion allele, mew39, was used throughout this study as ncap-1. The neighboring gene (Y110A2AR.1) is predicted to encode a receptor expression-enhancing protein (REEP). (B) Animal heads showing jowls phenotype (red arrows). Anterior is up. WT, wild type; RFP:NCAP-1, red fluorescent protein-tagged NCAP-1 single-copy transgene. (C) Starvation assay. Data represent days required for worms to reproduce and consume bacterial food source (top schematic). Bars indicate mean ±SEM for n = 10 biological replicates. DOI: https://doi.org/10.7554/eLife.32242.002 The following figure supplement is available for figure 1:  Representative confocal micrographs of intestinal cells (middle) in worms expressing a GFP-tagged cargo (top schematic). TM, transmembrane domain. The average pixel intensity along a basolateral membrane was measured (bottom). (A-C) Bars indicate mean ±SEM for n ! 8 biological replicates. *p<0.05, **p<0.001, not significant (n.s.), unpaired, two-tailed T-test. (D) m2 protease-sensitivity assay. Western blot analysis of whole worm lysates was used to quantify the amount of full-length m2 (anti-HA, 50 kDa) before (pre TEV, bottom blot) and after protease induction (post TEV, top blot). Band intensities were compared to a tubulin loading control and normalized to the fcho-1(+) ncap-1(+) ratio (values below). (E) m2 phosphorylation assay. Western blot analysis of whole worm lysates to quantify phosphorylated m2 (top blot) relative to total m2 subunit (bottom blot). Values indicate band intensity ratios of phospho m2 compared to total m2, normalized to the fcho-1(+) ncap-1(+) ratio (values below). (D and E) Blots are representative of !3 biological replicates. +, wild type allele; -, deletion allele. DOI: https://doi.org/10.7554/eLife.32242.004 fcho-1 ncap-1 worms. We evaluated AP2 conformation using an in vivo protease-sensitivity assay (Hollopeter et al., 2014). Briefly, a Tobacco Etch Virus (TEV) protease site was inserted into a surface loop of the m2 subunit that becomes more exposed when AP2 is opened (Jackson et al., 2010;Matsui and Kirchhausen, 1990). Following induction of TEV protease expression, the ensemble protease-sensitivity of the AP2 complexes is determined using western blot analysis of m2 in whole worm lysates. In fcho-1 worms, AP2 dwells in a protease-insensitive, closed state (Hollopeter et al., 2014) and AP2 was more protease-sensitive (open) in ncap-1 mutants ( Figure 2D). These results explain the phenotypic rescue of fcho-1 mutants and indicate that AP2 accumulates in an open state in the absence of NCAP-1.
Phosphorylation of m2 correlates with open AP2 in C. elegans (Hollopeter et al., 2014). Because loss of NCAP-1 increases AP2 protease-sensitivity, we wondered whether m2 phosphorylation might also be elevated. We quantified the phosphorylation status of AP2 in vivo by blotting worm lysates with an antibody specific to the phosphorylated threonine in m2 (T160). Loss of NCAP-1 resulted in T160 phosphorylation levels greater than in wild type worms ( Figure 2E). Thus, AP2 adopts a hyperphosphorylated, open state without the action of NCAP-1.

Negative regulation of AP2 is a conserved function of NECAPs
To determine if the ability to inactivate AP2 is a conserved function of NECAPs in the context of our C. elegans system, we expressed heterologous NECAPs as single-copy transgenes in fcho-1 ncap-1 mutant worms. Mouse NECAP1 and NECAP2, as well as a NECAP from the multicellular fungus, Sphaerobolus stellatus, recapitulated the reduced fitness of fcho-1 mutants ( Figure 3A) and caused AP2 to adopt a more closed, protease-insensitive state ( Figure 3B). It appears that NECAPs have retained the ability to negatively regulate AP2, at least in C. elegans.

NECAPs bind the open and phosphorylated AP2 core in vitro
We investigated the possibility that NECAPs negatively regulate AP2 via a direct interaction. Indeed, affinity-tagged NECAPs co-purified endogenous AP2 complexes from HEK293 cells ( Figure 4A), consistent with previous reports (Ritter et al., 2003). NECAPs are thought to bind the AP2 a adaptin appendage domain and the clathrin binding box in the b2 adaptin hinge region via a C-terminal WXXF motif and an N-terminal PHear domain, respectively (Ritter et al., 2013;Ritter et al., 2003). Because the C-terminal WXXF motif is poorly conserved in C. elegans NCAP-1 (LLDF) and completely absent from the S. stellatus protein (PKRR), it was unclear whether the negative regulation of AP2 by heterologous NECAPs (Figure 3) was mediated via binding to the a appendage. Instead, we were interested to examine whether NECAPs also bind to the AP2 core ( Figure 1-figure supplement 1B) because this would offer a direct route to conformational regulation. We purified recombinant vertebrate AP2 cores lacking ears and linkers and tested binding of these cores to recombinant NECAPs from worms and mice ( Figure 4B). Interestingly, NECAPs did not bind unmodified AP2 cores ( Figure 4B).
Because open and phosphorylated AP2s accumulate in ncap-1 mutants, we reasoned that NECAPs might act upon these modified forms of AP2. To test our hypothesis, we introduced a previously-characterized open AP2 mutation in the m2 subunit (E302K; Figure 1-figure supplement 1C) (Hollopeter et al., 2014) to produce recombinant AP2 cores that dwell in the open state. We also co-expressed AP2 with the kinase domain of AAK1 to purify AP2 cores with phosphorylated m2 (T156) (Hö ning et al., 2005). When these modified complexes were tested in our pulldown assays, we observed that NECAPs bound both open and phosphorylated AP2 ( Figure 4B and C). These data suggest that NECAPs associate with the AP2 complex in a conformation-dependent manner.

NCAP-1 localizes with constitutively open AP2 in vivo
To corroborate our in vitro pulldown results we sought in vivo evidence that NECAPs associate with open and phosphorylated AP2. In C. elegans, fluorescently-tagged AP2 is enriched at the nerve ring, a major neuropil of bundled axons encircling the pharynx (White et al., 1986) ( Figure 5). We used confocal microscopy to observe localization of NCAP-1 tagged with a red fluorescent protein (RFP:NCAP-1) at the nerve ring in live worms. In both wild type and fcho-1 mutant worms, NCAP-1 was not overtly enriched compared to AP2. The low level of NCAP-1 relative to AP2 may indicate that in worms, NCAP-1 is not stably associated with AP2 on membranes. However, introduction of a mutation in m2 known to generate hyper-phosphorylated, constitutively open AP2 (E306K; Figure 1figure supplement 1C) (Hollopeter et al., 2014) enhanced localization of NCAP-1 nearly 2-fold ( Figure 5). In other words, NCAP-1 was abnormally recruited to the nerve ring in the open AP2 mutant suggesting that NECAPs favor association with activated forms of AP2 in vivo. It is possible that the basal level of localization observed in both wild type and fcho-1 mutants may represent NCAP-1 bound to the AP2 appendages (Ritter et al., 2013;Ritter et al., 2003).
The preferential association of NCAP-1 with the hyper-phosphorylated AP2 (E306K) compared to the phosphorylation-defective AP2 (T160A) was not simply due to differences in the extent to which these mutations generate a protease-sensitive, open AP2 complex (Hollopeter et al., 2014). When the E306K and T160A mutations were both present in the same m2 subunit, less RFP:NCAP-1 was localized to the nerve ring compared to when the E306K mutation was combined with another open AP2 mutation (R440S; Figure 1-figure supplement 1C and Figure 5) (Hollopeter et al., 2014). In other words, the effect of the phosphorylation site mutation was dominant, while the two open AP2 mutations were additive, with respect to NCAP-1 association. Thus, phosphorylation site mutations preclude the association of NCAP-1 with open AP2.

Missense mutations render NECAPs functionally inactive
In the fcho-1 suppressor screen, we isolated two independent missense mutations in the N-terminal PHear domain of NCAP-1: A29D and S84N ( Figure 1A). To investigate the mechanism by which these mutations restore AP2 activity, we engineered them de novo into an RFP:NCAP-1 transgene in an otherwise ncap-1 null background. These modified NECAPs failed to complement ncap-1 mutants ( Figure 6A), were unable to restore the closed conformation of AP2 in vivo ( Figure 6B), and were not recruited to the nerve ring by open, hyper-phosphorylated AP2 ( Figure 6-figure supplement 1A). Importantly, these mutant proteins were not simply unstable; by western blot analysis they appeared to be expressed at levels similar to the wild type version ( Figure 6C). We also introduced the homologous mutations into recombinant M. musculus NECAP2 constructs to evaluate AP2 binding using purified components. Even though these mutant NECAPs appeared to be stable and retained the ability to bind open AP2 (E302K), they specifically failed to bind the phosphorylated AP2 core ( Figure 6-figure supplement 1B). The in vitro pulldown assays suggest that the primary reason the NCAP-1 missense mutants are inefficiently recruited to the nerve ring by the open AP2 mutation in vivo may be that the corresponding adaptor complexes are also phosphorylated. Taken together, our results further suggest that the NCAP-1 mutants fail to inactivate AP2 because they are unable to bind the phosphorylated AP2 core.

Discussion
In C. elegans, loss-of-function mutations in ncap-1 bypass the requirement for FCHO-1 by restoring AP2 activity and promoting the accumulation of the open, phosphorylated form of AP2. These initial results were consistent with two distinct models: (1) NCAP-1 maintains the closed conformation of AP2 until FCHO-1 releases this inhibition, or (2) NCAP-1 generates the closed conformation of AP2, either directly or indirectly. As a consequence, the inactive form of AP2 predominates in the absence of allosteric activation by FCHO-1. Our subsequent biochemical and imaging data are most consistent with the latter model; NECAP homologs from a variety of organisms bind the open, phosphorylated AP2 core in vitro and restore the closed form of the complex in vivo. We also observe increased recruitment of fluorescently-tagged NCAP-1 by open, hyper-phosphorylated forms of AP2 in vivo. We propose that NECAPs act downstream of muniscin function and m2 phosphorylation to inactivate AP2 ( Figure 7).

fcho-1 suppressors disrupt the AP2 inactivation pathway
In the absence of FCHO-1, nematodes exhibit reduced fitness and AP2 dwells in a closed, hypophosphorylated state. Using an unbiased genetic screen, we isolated mutations that independently improve the fitness of fcho-1 genetic nulls. We now recognize that these fcho-1 suppressors occur in three distinct classes: dominant missense mutations that disrupt the closed conformation of AP2 (class 1) or mutate the phosphorylation site on m2 (class 2), and recessive mutations that inactivate NCAP-1 (class 3) (Figure 7). These three classes of mutations appear to rescue fcho-1 by restoring AP2 activity, but we propose that they achieve this through different mechanisms. Previously, the open AP2 mutants (class 1) appeared to behave similarly to the phosphorylation site mutants (class 2) (Hollopeter et al., 2014). However, this study reveals that NECAPs clearly distinguish between these two classes. Open, hyper-phosphorylated AP2 mutants (class 1) appear to recruit NECAPs in excess but must be somewhat resistant to their action in order to sustain AP2 activity in the absence of muniscins. By contrast, phosphorylation site mutants (class 2) appear to bypass the requirement for FCHO-1 by evading NECAP binding altogether. Similar to class 2, mutations that directly disrupt NCAP-1 (class 3) probably enable a basal level of AP2 activity to be sustained by complexes that succeed in attaining the open, phosphorylated state in lieu of allosteric activation by muniscins. Importantly, our data suggest that all three classes of suppressors bypass FCHO-1 by disrupting the Figure 6. Missense mutations render NCAP-1 stable but functionally inactive. Amino acid changes isolated from the fcho-1 suppressor screen ( Figure 1A) were introduced into an RFP-tagged NCAP-1 transgene in fcho-1 ncap-1 worms. +, wild type allele; -, deletion allele. (A) Starvation assay performed as in Figure 1C. Bars represent mean ±SEM for n ! 9 biological replicates. **p<0.001, unpaired, two-tailed T-test. (B) m2 protease-sensitivity assay as in Figure 3B. (C) Western blot analysis to detect HA epitope on NCAP-1 transgenic proteins. (B and C) Band intensities were compared to a beta actin loading control and normalized to the fcho-1(+) ncap-1(+) ratio (B) or to the transgenic wild type form of NCAP-1 (C) (values below). Blots are representative of 2 biological replicates. DOI: https://doi.org/10.7554/eLife.32242.008 The following figure supplement is available for figure 6:  Figure 6 continued on next page process by which AP2 closes. The wealth of mutations in AP2 and NECAP identified in the fcho-1 suppressor screen, combined with our biochemical and imaging data, reveal discrete steps of AP2 conformational and phosphorylation changes during its recycling (Figure 7).
It is curious that the unnaturally active forms of AP2 isolated from our fcho-1 suppressor screen do not exhibit enhanced endocytosis or membrane association when the fcho-1 gene is intact (Hollopeter et al., 2014) (Figure 2). In other words, why is there not a functional consequence associated with hyperactive AP2 in an otherwise wild type background? Perhaps there are compensatory mechanisms that prevent rampant endocytosis or, alternatively, our assays may lack the sensitivity and range necessary to detect the differences. Indeed, endocytic pits become enlarged after knockdown of NECAP1 and these structures could be a consequence of overactive AP2 (Ritter et al., 2013). Although vesicle size was not examined in this paper, regulating the time during which AP2 is phosphorylated and open could directly impact the size of the vesicle that gets internalized.

AP2 phosphorylation modulates NECAP recruitment
NECAPs form stable complexes in vitro with open and phosphorylated AP2 cores, suggesting that activated AP2 may be the endogenous substrate of NECAPs. Indeed, imaging data suggest NECAP dynamics during endocytic pit formation mimic those of clathrin (Taylor et al., 2011), consistent with our model that AP2 activation precedes NECAP recruitment. The accumulation of phosphorylated AP2 in the absence of NCAP-1 indicates that phosphorylation may also precede NECAP recruitment but is it unclear whether the phospho-threonine is an essential feature of the NECAP-AP2 interface. Although changing the threonine to a phosphorylation-defective alanine does appear to disrupt the association of NECAPs with AP2 ( Figures 4C and 5), we have not excluded the possibility that this mutation simply precludes NECAP binding by altering the conformation of AP2. Indeed, we observe that NECAPs bind open AP2 cores that have not been phosphorylated (the E302K cores, Figure 4B) supporting the model that NECAPs bind to a face of active AP2 that does The core function of NECAPs While NECAPs are widely conserved across eukaryotic organisms (Dergai et al., 2016), the precise function of this protein family has remained enigmatic. Vertebrate NECAPs exhibit different tissue distributions and are proposed to function in mutually exclusive pathways. For example, NECAP2 was recently shown to regulate AP1 instead of AP2 (Chamberland et al., 2016). Despite this proposed divergence, both vertebrate NECAPs rescue loss of NCAP-1 in the nematode, as does a fungal NECAP that lacks the C-terminal a adaptin-ear binding WXXF motif. These results indicate that the capacity to negatively regulate AP2 has been conserved among NECAPs. Based on the missense mutations that disrupt NECAP activity in vivo and AP2 binding in vitro ( Figure 6 and Figure 6-figure supplement 1), we believe this inhibitory function may involve a direct interaction of the N-terminal PHear domain (Ritter et al., 2007) with the phosphorylated AP2 core. How do NECAPs sustain the AP2 cycle? We propose the following model ( Figure 7): muniscins such as FCHO-1 promote the open state of AP2, which is phosphorylated on the m2 subunit by the AP2-associated kinase. NECAPs then bind the AP2 core to counteract the open state, presumably by facilitating a conformational change or dephosphorylation event later in endocytosis. This recycles AP2 back to its inactive state in the cytosol and renders the complex available for another round of endocytosis. sequence-based reagent rEP360 this paper Gene-specific target of crRNA: TGAAGTGTCTCGTAACAAGA sequence-based reagent rGB156 this paper Gene-specific target of crRNA: CAAATCACGTCTCAAGTGAC sequence-based reagent rGB155 this paper Gene-specific target of crRNA: TTGGGTGAAGTTCTAGCATC sequence-based reagent rEP254 this paper Gene-specific target of crRNA: CGGGCTGTCGAGGTTCCAGT sequence-based reagent rEP676 this paper Gene-specific target of crRNA: CACAAAATATCGAGAACTAT Continued on next page 2013). For differential interference contrast microscopy ( Figure 1B), worms were mounted on 5% agarose pads in PBS pH 7.4 with 20 mM sodium azide.

Cargo assay
The cargo assay was performed essentially as previously described (Hollopeter et al., 2014). Worms expressing an artificial cargo (oxSi484) were imaged on a Ziess LSM 880 confocal microscope (Biotechnology Resource Center, Cornell University, Ithaca, NY) with a 40x water immersion objective. All strains were imaged in one session with the same laser settings. Images were analyzed in Fiji (Schindelin et al., 2012). A user defined line was drawn along the membrane between intestinal segments 2 and 3, or segments 3 and 4, and the average pixel intensity was measured along the line. Strains are listed in Supplementary file 1A.
Primary antibody incubations occurred for 1 hr at room temperature or overnight at 4˚C. Secondary antibody incubations occurred at room temperature for 30 min -1 hr in the dark. TBST was used for all washes. Blots were imaged using Bio-Rad ChemiDoc MP systems and band intensities were quantified using the associated ImageLab software.
We used SYPRO Ruby Protein Gel stain (Lonza, Rockland, ME) to visualize the in vitro pulldowns in Figure 4B and in Figure 6-figure supplement 1B. Gels were fixed in 50% methanol/7% acetic acid for 30 min and stained with SYPRO for 4-16 hr at room temperature in the dark with gentle agitation. Gels were then washed 2 Â 15 min in 10% methanol/7% acetic acid and then 2 Â 5 min in water before imaging on a Bio-Rad ChemiDoc MP system.

C. elegans NECAP transgenes
We generated C. elegans targeting vectors for expression of NECAPs as single-copy transgenes (Frøkjaer-Jensen et al., 2008). For Figure 1B and C, Figure 5, and Figure 6-figure supplement 1A, a mini-gene encoding NCAP-1 was constructed using the Multisite Gateway System (Invitrogen). The first 4 exons of the ncap-1 gene were amplified with primer pair oGH731 + 3, while the other half of the coding sequence was amplified from cDNA using primer pair oGH734 + 6. The two amplicons were recombined using the Gibson assembly reaction (Gibson et al., 2009) and then recombined with the [1-2] donor vector using BP clonase (Invitrogen). The entry clone was amplified with oGH698 + 738 and a worm codon-optimized TagRFP-T (Ed Boyden, MIT, Cambridge, MA) was amplified with oGH526 + 8 and inserted upstream of the ncap-1 coding sequences using Gibson assembly. The resulting [1-2] entry clone (pGH505) was recombined with a [4-1] entry containing the ubiquitous dpy-30 promoter, the unc-54 3'UTR in a [2-3] entry and the [4-3] destination vector pCFJ150 (Christian Frøkjaer-Jensen, University of Utah) using LR clonase (Invitrogen) to generate the MosSCI targeting vector (pGH495) that was injected into worm strain EG6699 (Frøkjaer-Jensen et al., 2012).

Tissue culture pulldowns
The heterologous expression of affinity-tagged proteins in HEK293 cells followed by HaloTag isolation from cell lysates and western blots analysis of the purified proteins was performed as described (Hollopeter et al., 2014). Mammalian vectors for the expression of HaloTag fusions with the worm and mouse NECAPs were generated by amplifying the coding sequences from cDNA and inserting them by Gibson assembly into a custom-built HaloTag expression vector (Banks et al., 2014) that was amplified using oGH953-4. The C. elegans NCAP-1 (NM_061997.5; amplicon oGH955-6) expression vector is pGH500, while the M. musculus NECAP1 (BK000656.1; amplicon oGH957-8) is pGH501, and M. musculus NECAP2 (BK000657.1; amplicon oGH959-60) is pGH502. Plasmids and oligonucleotides are listed in Supplementary files 1B and 1C, respectively. The identity of the HEK293 cells was 100% matched by STR Profiling at ATCC. Cells were negative for mycoplasma using ATCC PCR kit #30-1012K on January 10, 2017.

Recombinant AP2 cores
Bicistronic vectors expressing the hexahistidine-tagged mouse AP2 b2 trunk along with mouse m2 were based on pGH424 (Hollopeter et al., 2014) and generated as follows: Mutations were introduced in m2 using Gibson assembly, and primers oGH847-8 for E302K and oGB24 + 33 for T156A. A thrombin protease site (not utilized in this study) was inserted after amino acid 236 of m2 using oGB34-35. The construct encoding wild type b2 trunk/m2 is pGB21, while b2 trunk/m2(E302K) is pGB19 and b2 trunk/m2(T156A) is pGB31. The bicistronic vector expressing the GST-tagged mouse AP2 a trunk along with rat s2 (pGH504) was generated as follows: The a trunk/s2 hemicomplex was amplified from the original expression vector (Collins et al., 2002) using primers oGH1249-50 and recombined with the pACYCDuet vector backbone (amplicon oGH1246-47) using Gibson assembly. The thrombin cleavage site between the a trunk and GST was replaced with the human rhinovirus (HRV) C3 protease site (not utilized in this study) with oGH1227-8 and Gibson assembly. Simultaneous expression of all four subunits of the AP2 core was achieved by transforming E. coli BL21 (DE3) cells (New England Biolabs, Ipswich, MA) with a pair of the bicistronic vectors and co-selecting for ampicillin resistance (pGB21, pGB19 or pGB31) and chloramphenicol resistance (pGH504). To generate phosphorylated AP2 cores, the kinase domain of mouse AAK1 (amino acids 1-325, BC141176.1) was amplified from plasmid template using oEP17-18 and inserted into the pRSFDuet backbone amplified with oEP13 + oGH338 using Gibson assembly. The AP2 subunits were coexpressed with this vector (pEP82) by including selection for kanamycin resistance. Plasmids and oligonucleotides are listed in Supplementary files 1B and 1C, respectively.
2xYT culture media containing the appropriate antibiotics was inoculated with an overnight culture of bacterial cells expressing AP2 cores and incubated at 37˚C with shaking at 180-200 RPM until OD 600 =~1.0. The temperature was decreased to 18˚C for 1 hr before expression was induced with 100 mM isopropyl b-D-1-thiogalactopyranoside (IPTG) for 20-24 hr. Cells were harvested by centrifugation, washed with lysis buffer (see below), and snap frozen in liquid N2 prior to storage at À80˚C. Cell pellets were resuspended in 50 mL lysis buffer per liter of initial culture volume. GST lysis buffer consists of PBS pH 7.4, 1 mM DTT, 60 mg/mL DNase I (grade II from bovine pancreas, Roche), 2.5 mM MgCl 2 , 0.3 mg/mL lysozyme (Sigma), and 1 mM phenylmethylsulfonyl fluoride (PMSF) (Millipore, Billerica, MA) with the addition of 1 tablet of cOmplete EDTA-free Protease Inhibitor Cocktail (Roche) for every 100 mL. The cell slurry was sonicated (20 s pulses at 20% amplitude for a total of 8 min; 50 mL at a time) using a Q700 sonicator (Qsonica, Newtown, CT). Lysates were cleared by centrifugation (~20000 x g) and filtration (0.2 mm). Cleared filtrate was rotated for 1 hr at 4˚C with equilibrated GST resin (GE Healthcare Life Sciences, Uppsala, Sweden), 1 ml resin per liter of the initial culture volume. The filtrate and resin were poured over a gravity column, washed with PBS pH 7.4 + 1 mM DTT, and AP2 was eluted with 50 mM Tris, 150 mM NaCl, 10 mM reduced glutathione, and 1 mM DTT, pH 8.0 (Fischer BioReagents, Fairlawn, NJ). The eluate buffer was exchanged with TBS pH 7.6 + 1 mM DTT and AP2 cores were concentrated to~0.5 mg/mL using an Amicon Centrifugal Filter Unit with a 100 kDa cutoff (Merck Millipore). Aliquots of AP2 cores were snap frozen in liquid N2 prior to storage at À80˚C.
Because we sometimes noted heterogeneity in our recombinant NECAPs, we built a set of vectors to purify NECAPs with cleavable C-terminal affinity tags. The NECAPs used for pulldowns in Figure 4B and Figure 6-figure supplement 1B were purified using the Intein Mediated Purification with an Affinity Chitin-binding Tag (IMPACT, NEB) system, with slight modifications. The Mxe GyrA intein tag and chitin binding domain (CBD) from pTXB1 (NEB) were amplified using oGB52-3 (which adds a hexahistidine tag onto the C-terminus of the CBD). This amplicon was inserted into the pET21b expression vector (PCR product of oGB47 + oGH1231) in a three-piece Gibson reaction along with one of the following NECAPs: pGB29 included C. elegans NCAP-1 (NM_061997.5, amplified with oGB48-9 from pGH500), pGB27 included M. musculus NECAP1 (BK000656.1, amplified with oGB48 + 50 from pGH501) and pGB28 included M. musculus NECAP2 (BK000657.1, amplified with oGB48 + 51 from pGH502). The NECAPs were amplified such that they include the N-terminal HaloTag and TEV protease site from the mammalian expression vectors (see Tissue culture pulldowns). To generate the NECAPs used in Figure 6-figure supplement 1B, mutations were introduced in NECAP2 (pGB28) using Gibson assembly with oEP642-3 (A32D; pGB91) and oEP644-5 (S87N; pGB94). To generate a HaloTag control protein for these pulldowns (pGB81), pGB28 was amplified with oGB52 + 174 and recombined in a one-piece Gibson reaction to remove the NECAP sequence. All plasmids and oligonucleotides are listed in Supplementary file 1B and 1C, respectively.
Cultures were subsequently grown as described for AP2 cores (above) except expression was induced with 300 mM IPTG and cells were harvested after 16-20 hr incubation at 18˚C.
For 6xHis-HaloTag-NECAP2 (pGH503) and 6xHis-HaloTag (pGH494) cell pellets collected from 500 mL of culture were re-suspended in 50 mL nickel lysis buffer (20 mM HEPES pH 7.5, 500 mM NaCl, 30 mM imidazole, 5% glycerol, 5 mM BME) with the addition of 60-200 mg/mL DNase I (grade II from bovine pancreas, Roche), 2.5 mM MgCl 2 , 1 tablet of cOmplete EDTA-free Protease Inhibitor Cocktail (Roche), 0.3 mg/mL lysozyme (Sigma), and 1 mM PMSF. Cells were lysed by sonication using a Q700 sonicator (Qsonica) at 20% amplitude with 20 s pulses for 4 min total, centrifuged at~20000 x g for 30 min and filtered to 0.2 mm. The hexahistidine-tagged proteins were purified using nickel-charged 5 mL HiTrap Chelating HP columns (GE Healthcare Life Sciences) on a BioLogic LP system (BioRad). After sample loading, columns were washed with nickel lysis buffer and bound proteins were eluted with nickel elution buffer (nickel lysis buffer with 1 M imidazole). Fractions (1 mL each) containing the protein of interest were combined, buffer exchanged with 20 mM Tris pH 8, 5% glycerol, 1 mM EDTA, 1 mM DTT, and 50 mM NaCl and purified by ion exchange chromatography using a HiTrap Q HP column (GE Healthcare Life Sciences) on the BioLogic LP system (BioRad). After sample loading, proteins were eluted using a salt gradient (50 mM to 1000 mM NaCl). Elution fractions containing the protein of interest were combined and aliquots were snap frozen in liquid N2 prior to storage at À80˚C.
To purify recombinant NECAPs (HaloTag-NECAPs-intein-6xHis) or the HaloTag-intein-6xHis control we used a modified version of the IMPACT system. Cell lysis was performed as described for other hexahistidine-tagged proteins (above) except BME was omitted from the lysis buffer to inhibit intein self-cleavage. Cleared lysates were loaded onto gravity columns containing nickel resin (Thermo Scientific, 0.5 mL per 500 mL initial culture volume) and washed extensively. Columns were then flushed with nickel lysis buffer containing 40 mM BME, plugged, and incubated at 4˚C for 30 to 40 hr in order to cleave proteins from the intein tag. The released proteins were collected, buffer exchanged with TBS pH 7.6 + 1 mM DTT, and concentrated using an Amicon Centrifugal Filter Unit with 50 kDa MW cutoff (Merck Millipore) or with 30 kD MW cutoff for the HaloTag control. Aliquots were snap frozen in liquid N2 prior to storage at À80˚C.

In vitro pulldown assays
Pulldown assays were performed essentially as described in Hollopeter et al., 2014 except the protease cleavage step was 3 hr and the input gel sample represented 20% of the 'prey' protein mixture.

Nerve ring microscopy
Worms expressing APA-2:GFP (mewSi1) and RFP:NCAP-1 (mewSi2) in an ncap-1 background were imaged on an Zeiss LSM 880 confocal microscope (Biotechnology Resource Center) with a 40x water immersion objective. APA-2:GFP(mewSi1) is molecularly similar to APA-2:GFP(oxSi254) (Gu et al., 2013) except the entry clones ([4-1]Pdpy-30, [1-2]apa-2 cDNA, and [2-3]GFP:unc-54 3'UTR) were recombined with the [4-3] MosSCI vector (pCFJ210) to target the ttTi4348 site in EG6701 worms (Frøkjaer-Jensen et al., 2012). Fluorophores were excited with 488 nm (GFP) and 561 nm (RFP) lasers. All strains were imaged in one session with the same laser settings. For each worm, a single confocal slice through the approximate sagittal section of the nerve ring was analyzed in Fiji (Schindelin et al., 2012). Two regions of interest (ROI) corresponding to both dorsal and ventral sections of the nerve ring along with an ROI outside of the worm (to control for background signal) were user defined. The average pixel intensities in both the GFP and RFP channels were determined and the background values were subtracted from the nerve ring values.
CRISPR-Cas9 generation of m2 and NCAP-1 mutations CRISPR/Cas9 edits were generated using the dpy-10 co-conversion strategy (Arribere et al., 2014) with ribonucleoprotein (RNP) complexes (Paix et al., 2015). For generating m2 mutations, gonads of young adult hermaphrodites were injected with RNP mixes containing~3.7 mg/mL Cas9 (purified inhouse), 1 mg/mL tracrRNA, 0.08 mg/mL dpy-10 crRNA, 16.7 mM dpy-10(cn64) roller repair, 0.4 mg/mL gene-specific crRNA, and 16.7 mM gene-specific repair. The gene-specific crRNAs and oligonucleotide repairs were as follows: to generate m2(E306K) -rEP360 and oGB154 (introduces an NruI site for genotyping), to generate m2(T160A) -rGB156 and oGB130 (introduces an NgoMIV site), and to generate m2(R440S) -rGB155 and oGB159 (introduces a PvuI site). F1 progeny exhibiting the roller phenotype were placed on individual culture plates and allowed to produce offspring for 1-2 days before being lysed in 50 mL 1X Phusion GC buffer (NEB) with 0.4 U Proteinase K (NEB) by freezing at À80˚C and then heating at 65˚C for 1 hr followed by 95˚C for 15 min. The targeted region of the genome was amplified using PCR and the amplicon was digested to identify correctly edited worms. Mutations were confirmed by sequencing the PCR product.
All materials and resources described in this article are available upon request with no restrictions.