INAVA-ARNO complexes bridge mucosal barrier function with inflammatory signaling

Homeostasis at mucosal surfaces requires cross-talk between the environment and barrier epithelial cells. Disruption of barrier function typifies mucosal disease. Here we elucidate a bifunctional role in coordinating this cross-talk for the inflammatory bowel disease risk-gene INAVA. Both activities require INAVA’s DUF3338 domain (renamed CUPID). CUPID stably binds the cytohesin ARF-GEF ARNO to effect lateral membrane F-actin assembly underlying cell-cell junctions and barrier function. Unexpectedly, when bound to CUPID, ARNO affects F-actin dynamics in the absence of its canonical activity as a guanine nucleotide-exchange factor. Upon exposure to IL-1β, INAVA relocates to form cytosolic puncta, where CUPID amplifies TRAF6-dependent polyubiquitination and inflammatory signaling. In this case, ARNO binding to CUPID negatively-regulates polyubiquitination and the inflammatory response. INAVA and ARNO act similarly in primary human macrophages responding to IL-1β and to NOD2 agonists. Thus, INAVA-CUPID exhibits dual functions, coordinated directly by ARNO, that bridge epithelial barrier function with extracellular signals and inflammation.


Introduction
C1ORF106, recently named INAVA (Innate Immune Activator), was identified as a risk factor for the chronic inflammatory bowel diseases (IBD) by genome-wide association studies and targeted exome sequencing (Rivas et al., 2011). Mice lacking the protein altogether show defects in intestinal barrier integrity at steady state and greater susceptibility to mucosal infection (Mohanan et al., 2018). Human macrophages carrying the IBD rs7554511 risk allele have decreased INAVA expression and show multiple defects in myeloid function, including in innate immune NOD2 signaling and cytokine secretion, and in microbial clearance in association with reduced autophagy and ROS production (Yan et al., 2017). Each process is well known to affect gut function in health and disease, but the molecular mechanisms for how they are regulated or interconnected by INAVA are not fully understood.
The cytohesins are guanine nucleotide-exchange factors for the ARF-family of proteins (ARF 1-4), which regulate cell membrane and F-actin dynamics (Donaldson and Jackson, 2011;Stalder and Antonny, 2013). All cytohesins contain a N-terminal coiled-coil (CC) protein-protein interaction region, an enzymatic SEC7 guanine nucleotide-exchange factor (GEF) domain, and a C-terminal PIPbinding PH domain. In their inactive conformation, the cytohesins localize to the cytosol. Full-blown GEF activation, typified by cytohesin 2 (also known as ARNO), requires membrane recruitment via ARNO binding to PIP2 (phosphatidylinositol 4, 5-bisphosphate), and then (activated) ARF-GTP, a product of the ARNO-GEF reaction (Chardin et al., 1996;Cohen et al., 2007;Malaby et al., 2013). This enables an enzymatically-driven positive feedback-loop for rapidly amplifying a localized pool of activated cytohesins and ARF-GTP needed to drive the massive ARF-dependent changes in actin and membrane dynamics that underlie cell spreading and epithelial breakdown (Santy and Casanova, 2001;Stalder et al., 2011).
In this study, we address the mechanism of INAVA action in polarized intestinal epithelial cells and primary human macrophages. We discover dual and mutually-exclusive functions for INAVA and the physical and functional interaction of the INAVA CUPID domain (INAVA-CUPID) with cytohesin 2 ARNO. In epithelial cells, INAVA-CUPID recruits ARNO to lateral membranes where the complex promotes actin assembly that underlies barrier function. This occurs via a novel GEF activity-independent mechanism. In response to the inflammatory cytokine IL-1b, INAVA relocates to cytosolic puncta that function as signalosomes. Here, CUPID acts with the E3-ubiquitin-ligase TRAF6 to enhance inflammatory signaling, and in this case, ARNO binding inhibits CUPID activity. In human macrophages containing the INAVA rs7554511 IBD-risk allele (low-INAVA expressing carriers), wild type INAVA expression enhances, and ARNO expression suppresses IL-1b and NOD2 signaling. Reconstitution with purified proteins in vitro shows biochemically that INAVA-CUPID functions as an enhancer of TRAF6 dependent polyubiquitination, and that this is blocked by ARNO. These results provide a direct mechanistic link between mucosal barrier function and inflammation implicated in human disease. When grown on permeable supports as polarized monolayers, INAVA-deficient cells display increased permeability to 4 kDa FITC dextran and reduced transepithelial electrical resistance (TEER) (Figure 1-figure supplement 1D-F), implicating defects in paracellular permeability and intercellular junctions. Conversely, Caco2BBe monolayers that overexpress INAVA exhibit evidence of increased integrity of intercellular junctions as measured by higher TEER and greater tolerance to depletion of intercellular adherens junctions caused by exposure to low Ca 2+ (Figure 1-figure supplement 1G). These phenotypes are consistent with a recent study using INAVA knock-out mice that supports a role for INAVA in maintenance of intercellular epithelial junctions responsible for mucosal barrier function (Mohanan et al., 2018).

INAVA affects the epithelial barrier
To understand how INAVA acts on epithelial junctions, we first examined the localization of INAVA constructs fused to GFP that were stably expressed in Caco2BBe cells. Humans express two   well-differentiated polarized Caco2BBe monolayers, INAVA localized  to tight and adherens junctions, as evidenced by co-localization with E-cadherin and ZO-1 (Figure 1-figure supplement 1I,J). When expressed on its own, the C-terminal region flanking the INAVA-CUPID domain (aa 261-677) also localized to lateral membranes ( Figure 1D, bottom panels). The CUPID domain on its own localized to the cytosol, and the CUPID domain together with the N-terminal flanking region (aa 1-260) localized to cytosolic puncta ( Figure 1D). Thus, the C-terminal region of INAVA is necessary and sufficient for lateral membrane targeting. Lateral membrane localization did not depend upon association with the E-cadherin complex. This was evidenced in cells treated with 6 mM EGTA to disassemble adherens junctions and induce E-cadherin endocytosis. While E-cadherin localized to endocytic vesicles upon EGTA treatment, INAVA remained at lateral membranes ( Figure 1E). Staining for the tight junction protein ZO-1 as a control also remained unchanged. Thus, INAVA localizes to lateral membranes independently of E-cadherin. Overall, these results support a critical role for INAVA in maintaining or regulating epithelial barrier integrity.

CUPID-domain directly interacts with cytohesin family members
To understand the mechanism of action underlying this phenotype, we focused on the single signature feature of INAVA, the CUPID domain. As all four CUPID-containing human proteins are implicated in binding the cytohesins (Ikenouchi and Umeda, 2010;Klarlund et al., 2001;Mohanan et al., 2018;Torii et al., 2014), we tested if cytohesin-binding is involved in INAVA function and explained by CUPID. We first screened the cytohesins 1-3 for binding to INAVA-CUPID (aa 100-261) using a yeast-based protein-protein interaction platform (Schmitz et al., 2009). In all cases, as evidenced by intracellular puncta formation, CUPID assembled with each of the cytohesins: cytohesin 1, cytohesin 2 (ARNO), and cytohesin 3 (GRP1), (Figure 2-figure supplement 1A,B). Coiledcoil domains of the cytohesins were sufficient for direct interaction with CUPID. In a purified system, recombinant INAVA-CUPID was sufficient to pull down the coiled-coil domains of Cytohesin-1, ARNO, and GRP1 ( . To test how INAVA may affect cytohesin function, we generated Caco2BBe intestinal cells stably expressing myc-tagged ARNO or GRP1. These cells, however, rapidly suppressed ARNO and GRP1 expression, as evidenced by the lack of protein detected upon transduction. PMA (phorbol-12-myristate-13-acetate), which strongly activates the CMV promotors used in these transductions, rescued ARNO and GRP1 protein expression, presumably by overcoming the endogenous mechanism(s) of epigenetic gene silencing operating in these cell lines ( Figure 2B, compare lanes 1 and 3; and lanes 5 and 7) (Lö ser et al., 1998). Strikingly, when co-expressed in the same cells, INAVA rescued ARNO expression in the absence of PMA ( Figure 2B, compare lanes 1 and 2), implicating a stabilizing interaction between the two proteins. The effect was specific to ARNO as INAVA did not rescue the cytohesin GRP1 against gene-suppression (compare lanes 5 and 6). Over-expression of INAVA-GFP increased endogenous ARNO protein levels with no apparent changes in TEER or paracellular flux when compared to wildtype cells (  Error bars indicate ± SEM. *p<0.05, †p<0.0001, NS not significant (two-tailed Student's t-test), n = 8-10. (G) Cartoon display domain architecture of INAVA and ARNO and three ARNO activity states. Inactive ARNO is autoinhibited and cytosolically localized. The GEF dependent positive feedback mechanism of ARNO promotes robust levels of trafficking and actin assembly at the plasma membrane. In the context of a polarized epithelial cells this positive feedback mechanism disrupts barrier integrity that leads to epithelial breakdown and cell spreading. In contrast, CUPID domain mediated INAVA-ARNO complex permits non-canonical GEF independent function that allow fine control of lateral actin to maintain barrier homeostasis. DOI: https://doi.org/10.7554/eLife.38539.004 The following figure supplement is available for figure 2:

INAVA-ARNO promotes non-canonical GEF-independent lateral membrane actin assembly
To circumvent the problem of gene silencing in cells stably expressing ARNO and enable further mechanistic studies on INAVA-ARNO interactions, we generated stable doxycycline (dox) inducible myc-tagged ARNO expressing Caco2BBe cells. Upon induction with dox, ARNO localized to the cytosol, consistent with the molecule folding in an auto-inhibited inactive state ( Figure 2C top panels) (Cohen et al., 2007). Co-expression of INAVA-GFP recruited ARNO to lateral membranes, consistent with our findings that ARNO and INAVA form stable protein complexes ( Figure 2C bottom panels). Unexpectedly, given ARNO's canonical actions in regulating F-actin assembly and membrane dynamics (Turner and Brown, 2001), the INAVA-ARNO complex did not cause epithelial cell spreading or disassembly of intercellular contacts. Intercellular junctions remained fully intact as assessed by TEER (Figure 1-figure supplement 1F). We did, however, find enhancement of lateral membrane F-actin assembly ( Figure 2D,E), consistent with the anticipated effects of ARF activation on cortical F-actin (Myers and Casanova, 2008). In these studies, to properly compare the densities of cortical F-actin, we co-cultured cells stably expressing INAVA, or ARNO, or both proteins, alongside non-transduced wild type (WT) Caco2BBe cells ( Figure 2D). Relative densities of lateral membrane F-actin stained with TRITC-phalloidin was measured by imaging both cell types on the same cover slip under the same conditions ( Figure 2E right panels, peak intensities plotted in left panels). Cells expressing INAVA alone showed modest amplified F-actin density at lateral membranes. Expression of ARNO alone had no detectable effect. In contrast, expression of both INAVA and ARNO together strongly amplified lateral membrane cortical F-actin (>2 fold). Remarkably, the effects of INAVA-ARNO on F-actin assembly did not require ARF-GTP binding to ARNO or ARNO's enzymatic GEF activity ( Figure 2F), as we observed the same enhanced densities of lateral membrane F-actin in Caco2BBe cells expressing the ARNO GEF mutant (E156K) or the ARF-GTP binding mutant (K366A). In contrast, ARNO mutants lacking the ability to bind PIP 2 (K268A) failed to enhance F-actin ( Figure 2F, right). Thus, in intact cells, when complexed with INAVA, only PIP binding by ARNO is required for effecting F-actin assembly.
To identify the INAVA region required for ARNO-mediated F-actin effects, we examined cortical F-actin assemblies in Caco2BBe cells expressing ARNO together with just the INAVA-CUPID domain. In this case, because CUPID alone cannot target ARNO to lateral membranes ( Figure 1D , again implicating a unique and non-canonical mechanism of action for the cytohesins in regulating F-actin dynamics when bound to INAVA. In other words, instead of the substantial ARNO GEF-dependent hyperactivation of ARFs leading to massive F-actin rearrangement, cell spreading, and epithelial breakdown (Stalder and Antonny, 2013;White et al., 2010), CUPID binding to ARNO displaces the positive-feedback hyperactivation of ARFs to enable instead a smaller (modulated) effect on cortical F-actin assembly, which we propose is required for barrier integrity ( Figure 2G). Canonically, activated ARNO localizes to the leading edge to drive cell motility, but here we show INAVA localizes ARNO to lateral membranes, where regulation of cell-cell contacts and mucosal barrier function occurs.

INAVA-ARNO regulates IL-1b signaling
ARNO has been implicated in IL-1b signaling (Zhu et al., 2012), and given the association of INAVA with inflammatory bowel disease, we examined the effect of IL-1b on INAVA-ARNO function. We found that IL-1b rapidly within 30 min induced the relocalization of INAVA-GFP from lateral membranes to cytosolic puncta in Caco2BBe cells ( Figure 3A). IL-1b-induced cytosolic puncta were also found in human colon HCT8 cells where we found a stronger phenotype ( Figure 3B). In this case, puncta were assembled from cytosolic pools of INAVA-GFP as HCT8 cells do not form intercellular junctions (Vermeulen et al., 1997) where INAVA-GFP normally localizes. INAVA puncta appeared to be associated with IL-1b signaling as IL-1b-induced INAVA-puncta were not efficiently formed in HCT8 cells lacking TRAF6 (CRISPR-Cas9 KO), the ubiquitin E3 ligase required for IL-1b signal transduction ( Figure 3B; Figure 3-figure supplement 1A) (Cao et al., 1996;Lomaga et al., 1999). Notably, the short INAVA-S isoform rarely induced puncta formation in Caco2BBe or HCT8 intestinal cells ( Figure 3A,B). We conclude the N-terminal region of INAVA is required for puncta formation.
To test if INAVA affects IL-1b signaling, we used HEK293T cells expressing INAVA together with a NF-kB-dependent luciferase reporter. The cells were treated with IL-1b or co-transfected with TRAF6, which also induces an inflammatory response. Cells expressing INAVA responded to IL-1b or TRAF6 overexpression by enhanced NF-kB activation, as assessed using luciferase expression ( Figure 3C; numerical values Figure 3-source data 1). Cells expressing the short INAVA-S isoform, however, did not. This result aligns with our finding that the short isoform cannot assemble into cytosolic puncta, further implicating puncta formation in the signaling cascade. We also found that cells expressing INAVA lacking the CUPID domain (CUPID D) or cells expressing only the CUPID domain failed to show amplified IL-1b signal transduction ( Figure 3D; numerical values

INAVA-ARNO amplifies inflammatory signaling in human macrophages
To further address the relevance of these findings, we prepared primary monocyte-derived human macrophages (MDMs) from individuals carrying the INAVA-IBD rs7554511 CC risk allele. These cells have reduced INAVA expression and impaired signaling and cytokine secretion in response to PRR stimulation (Yan et al., 2017). We asked if similar genotype-dependent regulation was observed in response to IL-1b treatment. As predicted from our previous results (Yan et al., 2017), IL-1b-treated MDMs from rs7554511 CC IBD risk carriers showed reduced INAVA protein expression compared to MDMs derived from AA carriers ( Figure 4A). Also consistent with our in vitro results, both IL-1binduced NF-kB and MAPK signaling ( Figure

INAVA enhances protein ubiquitination in cells and in vitro
Because we find a dependence on TRAF6 for INAVA puncta formation and IL-1b inflammatory responses, we hypothesized that INAVA may function in the TRAF6 ubiquitin ligase cascade essential for IL-1b signaling (Xia et al., 2009). To test this, we examined protein polyubiquitination in HEK293T cells expressing HA-tagged ubiquitin. Cells expressing both TRAF6 and INAVA exhibited much higher levels of protein ubiquitination as assessed by anti-HA immunoblot ( Figure 5A Figure 5A, compare lanes 5 and 6 with lane 4), consistent with the evidence that INAVA-S cannot form cytosolic puncta or mediate IL-1b inflammatory responses. We also found that overexpression of ARNO strongly inhibited INAVA-TRAF6 induced ubiquitination ( Figure 5B, compare lane 8 with lane 6), consistent with our discovery that ARNO can act as a negative regulator in this pathway. Thus, INAVA appears to act as an enhancer of the ubiquitin-cascades induced by TRAF6, and this is blocked by ARNO.
To show INAVA is minimally sufficient for this effect, we reconstituted these reactions in vitro using recombinant purified proteins. For these studies, we produced recombinant HA-tagged ubiquitin lacking the terminal two glycines (UbiquitinD2G-HA), which we used as pseudosubstrate to ensure specificity for the ubiquitination-reactions. Because the di-glycine motif at the C-terminus of ubiquitin is required for covalent attachment to target lysines, the UbiquitinD2G-HA pseudosubstrate cannot be attached to other molecules added in the in vitro reactions, and the only detectable product can be ubiquitination of the pseudosubstrate. Control studies showed that the reactions reconstituted with wild type ubiquitin, UBE1 (E1), heterodimer UBC13/UEV1a (E2), and the E3 ligase TRAF6-RING (aa 50-211) produced polyubiquitinated UbiquitinD2G-HA ( Figure 5C, compare lane 14 with all others). Reactions lacking either E1, heterodimer E2, E3, or ubiquitin yielded no product. When the same reactions were reconstituted with the addition of INAVA-CUPID, the polyubiquitination of UbiquitinD2G-HA was amplified ( Figure 5D, compare lane 4 with lane 2). Reactions performed with just UBC13 as the E2 ligase were not as processive, but the amplification by INAVA-CUPID was still clearly apparent ( Figure 5E, compare lanes 4 and 2). Thus, the INAVA-CUPID domain has activity as an enhancer of TRAF6 ubiquitination. This we propose explains the ability of full-length INAVA to promote inflammatory signaling by IL-1b and MDP-induced NOD2 activation in intact cells. Notably, the INAVA-dependent amplification of ubiquitination in vitro was fully inhibited by the addition of recombinant ARNO ( Figure 5D, compare lane 8 with lane 4), explaining how ARNO can act by binding INAVA to negatively regulate the inflammatory response in this pathway.

Discussion
INAVA-CUPID binds ARNO, targets the molecule to lateral membranes of epithelial monolayers, and enables ARNO to affect F-actin assembly without the need for GEF activity or positive ARF-GTP feedback ( Figure 2G). This is a non-canonical mechanism for ARNO action, and one, we propose, explains how the INAVA-ARNO complex acts as a tunable factor for control of epithelial cell-cell contacts at mucosal surfaces. The impaired barrier integrity observed in INAVA knockout mice and cells can plausibly be explained by loss of INAVA-ARNO cortical actin promoting complexes at lateral membranes. This mechanism of action differs from the model proposed for INAVA mediated degradation of cytohesin 1 delineated in a recent report (Mohanan et al., 2018). It is possible that INAVA acts in different signaling complexes and operates in both ways depending on cell context. In our case, however, using both human epithelial cell lines and primary human macrophages, we find that cytohesins and INAVA cooperate to regulate downstream outcomes.
We also find that INAVA acts an enhancer in the ubiquitin cascades required for IL-1b signaling, and that CUPID is sufficient to reconstitute this activity in vitro. Thus, CUPID can functionally and directly act with the UBC13/UEV1 E2-and TRAF6 E3-ligases in polyubiquitination. The recent proteomic study by Mohanan et al., 2018 of proteins co-isolated by INAVA immunoprecipitation implicated such an interaction between INAVA and another E3-ligase FBXW11 (SCF complex). But whether INAVA acted in protein ubiquitination of cytohesin 1 through FBXW11 was not conclusively tested (Mohanan et al., 2018).
In intact cells, the N-terminal region (aa 1-99) of INAVA is required for puncta formation and CUPID mediated enhancement of the ubiquitin ligases, as only the long INAVA isoform acts in the IL-1b inflammatory responses. INAVA-S is inactive. And it does not form cytosolic puncta, which we suggest represent protein complexes required for signal transduction (signalosomes). In macrophages, the C-terminal region of INAVA is also implicated in inflammatory signaling by mediating scaffolding for NOD2 interacting proteins (Yan et al., 2017).
We also note the signaling puncta formed in response to IL-1b require the relocalization of INAVA from lateral membranes to the cytosol. An analogous relocalization of INAVA, from cytosol to nucleus, was also observed in NOD2 activated macrophages (Yan et al., 2017), though the functional consequences for INAVA in the nucleus have yet to be defined. Thus, CUPID action is likely site-specific, dictated by the N-and C-terminal regions of the protein flanking the CUPID domain.
Strikingly, in the case of inflammatory signaling, INAVA-CUPID is inhibited in the ubiquitination reactions by ARNO, thus, elucidating a previously unappreciated function for ARNO as a negativeregulator of inflammatory responses. The result also shows how ARNO can coordinate CUPID function as it bridges between barrier function and inflammation. Given INAVA's dual activities in IL-1b and MDP signaling cascades, and in F-actin assembly at lateral membranes, we can imagine INAVA (together with ARNO) innately reacting to environmental factors in the gut mucosa as a form of guard receptor (Dangl and Jones, 2001) for the management of intestinal homeostasis.
In summary, our results delineate the molecular activities of CUPID and uncover unexpected and opposing roles of INAVA-ARNO complexes in the maintenance of epithelial barrier function at mucosal surfaces, and in inflammatory signaling for epithelial cells and macrophages. The two mechanisms defined are intrinsically related to the pathogenesis of IBD, and likely will apply as general rules for the CUPID domains of FRMD4A, FRMD4B, and CCDC120 -proteins implicated in neurite outgrowth, and in human cancer, Alzheimer's, celiac, and heart disease (Cappola et al., 2010;Garner et al., 2014;Goldie et al., 2012;Lambert et al., 2013;Torii et al., 2014;Velcheti et al., 2017).

Cell culture
HEK293T and MCF7 were grown on DMEM 10% FBS with pen/strep. Caco2BBe were grown with DMEM 15% FBS with pen/strep. HEK293T, MCF7 and Caco2BBe were authenticated using Short Tandem Repeat (STR) analysis and tested negative for mycoplasma contamination. Human peripheral blood mononuclear cells (PBMCs) were prepared from peripheral blood using Ficoll-Paque (GE Dharmacon, Lafayette, CO), and monocytes then isolated and cultured for 1 week for monocyte derived macrophages (MDMs).

Confocal and cell imaging
Brightfield images were taken using an inverted microscope coupled to a Nikon camera. Confocal images were taken using a spinning disk confocal head (CSU-X1, PerkinElmer, Boston, MA) coupled to an inverted Zeiss Axiovert 200M microscope (Carl Zeiss, Jena, Germany). Imaging system was operated using Slidebook (Intelligent Imaging Innovations, Denver, CO).

Confocal imaging line scans
Caco2BBe WT cells were cocultured with stably expressed INAVA or INAVA-GFP and myc-ARNO constructs. Cells were grown on coverslips for 2 days, fix and stain as above. F-actin was stained with TRITC-phalloidin. ImageJ plot histogram function was used for line scans analysis. Line width was set at 10 pixels and the fluorescence F-actin intensity across cell-cell junctions. Peak positions were aligned based on the maximum peak value along the line.

Immunocytochemistry
Cells plated either on glass coverslips or transwell filters were washed with PBS and fixed with 4% paraformaldehyde for 30 min. Cells were washed with PBS, permeabilize with PBS + 0.2% saponin and blocked with PBS + 0.2% saponin+5% normal goat for 1 hr. Primary antibodies were incubated for at least 1 hr at room temperature to overnight at 4˚C. Cells were washed several times with PBS + 0.2% saponin and then incubated for 30 min with Alexa Fluor secondary antibodies, TRITCphalloidin and DRAQ5 where indicated. Cells were washed and mounted on to slides.

Protein purification
Plasmids were transformed into Rosetta2DE3pLys (EMD Millipore) and grown on 2XYT (BD) plates with appropriate antibiotics. Bacterial colonies were scrapped and inoculated onto starter liquid culture of 2XYT + antibiotics and grown at 37˚C for 3-5 hr. The 5-10 ml starter culture were inoculated into large scale flasks containing 2XYT + antibiotics flasks and grown to OD600 = 0.8-1.0. Cultures were induced with 0.4 mM IPTG overnight at room temperature. Cells were harvested resuspended in 10 mM Tris pH8.0+150 mM NaCl 0.5 mM Triton-X100 +1 mM PMSF and lysed with an Emulsiflex-5 (Avestin). Cells were spun 16 kg for 10 min and supernatant transfer to Cobalt beads (Goldbio. com) or Glutathione resin (Goldbio.com) depending if the protein were His-tagged or GST-tagged, respectively. HiTrap QHP (GE Life Sciences) buffer A = 20 mM Tris pH 8.5 and buffer B = 20 mM Tris pH 8.5 + 1M NaCl. Superdex 200 10/300 GL buffer = 20 mM Tris pH 7.5, 100 mM NaCl. UBC13-UEV1a was purified as a complex where the plasmids were cotransformed and coexpressed. UBC13-UEV1a complex was purified by cobalt purification and then onto HiTrap QHP. Cytohesin constructs were purified by cobalt and onwards to HiTrap QHP where bound fractions were collected. GST-INAVA-CUPID (pGEX-TEV) and His-MBP-INAVA-CUPID (pET24a) constructs were purified by glutathione or cobalt respectively and onwards to HiTRAP QHP where bound fractions were collected. GST-INAVA-CUPID+His-Cytohesin-CC-SEC7 (pET28a) complex were coexpressed in Rosetta2-DE3pLlys and purified by tandem cobalt to glutathione resin and then onto Hi-Trap QHP. Complex fractions were collected. Traf6-His (aa 50-211) in pET24a was transformed into SHuffle T7 Express (NEB). Starter culture, inoculating of large culture and induction with IPTG was performed as above with addition of 50 mM zinc chloride added to the expression media. Purification of Traf6-His6 (aa 50-211) was purified with cobalt purification and samples were further purified by size exclusion Superdex 200. UbiquitinD2G-HA-His6 (pET24a) was expressed in Rosetta2DE3pLys, expressed and purified with cobalt and Superdex200. His-DN17ARF1 was purified by cobalt resin and chelated with 2 mM EDTA. GDP nucleotide was added at 20 molar excess of His-DN17ARF1 protein concentration along with 60 mM MgCl 2. Protein was then dialyzed to 10 mM Tris pH 8, 150 mM NaCl +5 mM MgCl 2 . All proteins were flash frozen with liquid nitrogen and stored in aliquots at À80˚C.

Guanine exchange assays
Recombinant HisARNO-CC-SEC7 alone or complexed with GST-INAVA-CUPID (1 mM) were incubated with MANT-GTP (25 mM) and DN17ARF1-GDP (40 mM) for 15 min at 25˚C. Guanine exchange reactions were measured by fluorescence at ex/em 355/445 nm using a TECAN fluorescent plate reader. Background from control samples with MANT-GTP and DN17ARF1-GDP was subtracted.

Knockdown and CRISPR knockout generation
shRNAs were cloned into pTY-shRNA-EF1a-Puro2a-GFP. Guides were cloned into plentiCRISPRv2 and lentivirus was generated in HEK293T cells by cotransfection with psPAX2 and pVSVG by polyethyleneimine (PEI) transfection. For each INAVA knockout line, two unique pair of guides were designed to target the introns that flank exon 2 of INAVA which knockout both isoforms. Lentivirus that target the 5-prime and 3-prime ends were pooled and co-transduced into Caco2BBe cells. Cells

Low calcium assay
Caco2BBe lines were grown on 24-well 0.4 mM polycarbonate transwell filters were grown at least 1 week. Cells were washed and grown in DMEM calcium free media, supplemented with 50 mM calcium +10% FBS. Transepithelial electrical resistance was measured before and 24 hr after low calcium treatment.

Permeability assays
Caco2BBe lines were grown on 24-well 0.4 mM polycarbonate transwell filters for three days. Cells were washed and changed to DMEM + 10% FBS no phenol red media. 1 mg/ml FITC-dextran 4 Da was added apically. Cells were incubated for 2 hr at 37˚C + 5% CO2. Basolateral media was collected and read in a 6-well plate reader ex/em 490/525 nm.

qRT-PCR
RNA was isolated by QIAshredder and RNeasy Mini kit (Qiagen), reverse transcribed with Super-Script II (ThermoFisher) and qPCR by SYBR Green (BioRad) on a CFX96 (BioRad). Samples were normalized to actin.

Luciferase assays
Firefly luciferase was assayed using Bright-Glo (Promega) or with homemade luciferase reagent as stated previously (Baker and Boyce, 2014). Renilla luciferase reagent was made and assayed as stated previously (Baker and Boyce, 2014).

Cells-based polyubiquitination assays
HEK293T wild type or stably expressed INAVA-GFP constructs were transfected with HA-Ubiquitin and where indicated myc-INAVA, myc-TRAF6 or FLAG-TRAF6 and myc-ARNO constructs by PEI transfection method between 16-20 hr. Stable INAVA-GFP lines had better comparable INAVA expression to each other than transient transfection. Cells were washed two times with PBS and lysed in RIPA +2 mM EDTA +protease inhibitors. Lysate were spun 20 kg for 5 min and supernatant was collected. Sample buffer was added and heated for 10 min at 95˚C. Samples were load loaded onto SDS-PAGE gels and immunoblotted with anti-HA and detected by HRP using SuperSignal West Femto (Fisher Scientific). Images were taken with the Azure c300 system.

Statistical analysis
Significance was assessed using two-tailed t-test or two-way ANOVA where indicated p<0.05 was considered significant. Lines over multiple bars indicate same significance values for these bars. Graphs were generated using GraphPad Prism.