Role of Bicaudal C1 in renal gluconeogenesis and its novel interaction with the CTLH complex

Altered glucose and lipid metabolism fuel cystic growth in polycystic kidneys, but the cause of these perturbations is unclear. Renal cysts also associate with mutations in Bicaudal C1 (Bicc1) or in its self-polymerizing sterile alpha motif (SAM). Here, we found that Bicc1 maintains normoglycemia and the expression of the gluconeogenic enzymes FBP1 and PEPCK in kidneys. A proteomic screen revealed that Bicc1 interacts with the C-Terminal to Lis-Homology domain (CTLH) complex. Since the orthologous Gid complex in S. cerevisae targets FBP1 and PEPCK for degradation, we mapped the topology among CTLH subunits and found that SAM-mediated binding controls Bicc1 protein levels, whereas Bicc1 inhibited the accumulation of several CTLH subunits. Under the conditions analyzed, Bicc1 increased FBP1 protein levels independently of the CTLH complex. Besides linking Bicc1 to cell metabolism, our findings reveal new layers of complexity in the regulation of renal gluconeogenesis compared to lower eukaryotes.


Author summary
Polycystic kidney diseases (PKD) are incurable inherited chronic disorders marked by fluid-filled cysts that frequently cause renal failure. A glycolytic metabolism reminiscent of cancerous cells accelerates cystic growth, but the mechanism underlying such metabolic re-wiring is poorly understood. PKD-like cystic kidneys also develop in mice that lack the RNA-binding protein Bicaudal-C (Bicc1), and mutations in a single copy of human BICC1 associate with renal cystic dysplasia. Here, we report that Bicc1 regulates renal gluconeogenesis. A screen for interacting factors revealed that Bicc1 binds the C-Terminal to Lis-Homology domain (CTLH) complex, which in lower eukaryotes mediates degradation of gluconeogenic enzymes. By contrast, Bicc1 and the mammalian CTLH complex regulated each other, and Bicc1 stimulated the accumulation of the rate-limiting gluconeogenic enzyme even in cells depleted of CTLH subunits. Our finding that Bicc1 is required for normoglycemia implies that renal gluconeogenesis may be important to inhibit cyst formation.

Introduction
Autosomal dominant polycystic kidney disease (ADPKD) is an incurable inherited chronic disorder characterized by progressive kidney enlargement and frequent end-stage renal disease due to numerous fluid-filled cysts that are induced by mutations and loss of heterozygosity in PKD1 or PKD2 genes [1]. Complexes of the corresponding transmembrane proteins polycystin-1 (PC1) and polycystin-2 (PC2) are activated at the cell surface by specific WNT ligands to mediate influx of extracellular Ca 2+ , or by mechanical stimulation of primary cilia that induces Ca 2+ release from intracellular stores in the endoplasmic reticulum [reviewed in 2]. Intracellular Ca 2+ spikes dampen the levels of cAMP by inhibiting adenylate cyclases V (AC5) and VI (AC6) and by stimulating phosphodiesterase (PDE) activity [2][3][4]. PC2 and PC1 also negatively regulate AC6 at the level of its mRNA and protein expression [4,5]. AC6 is important for cystic growth [6] since accumulation of excess cAMP combined with Ca 2+ restriction stimulates proliferation and cyst enlargement through several pathways, including PKA, EGFR, Src, B-Raf/Erk and mTORC1 signaling [2]. In addition, cystic growth in PC1-deficient kidneys is fueled by a Warburg-like metabolic switch characterized by increased anaerobic glycolysis at the expense of oxidative phosphorylation [7]. How PKD1 attenuates glycolysis is unknown [8].
Large cystic kidneys showing hyperactivation of EGFR, Src and mTORC1 also develop in homozygous mutant bpk mutant mice owing to a frame-shifting mutation in one of two alternative transcripts of Bicc1 [9][10][11][12]. Consistent with a role in polycystin signalling, Bicc1 is required to increase PC2 mRNA and protein levels, and its own accumulation is inhibited in Pkd1 mutant kidneys [13,14]. Maintaining normal Bicc1 expression is likely important since mutation or loss of a single copy of BICC1 in human is sufficient to provoke renal cystic dysplasia [15]. Similar to ADPKD kidneys, Bicc1 null mutant kidneys also upregulate AC6 protein levels, accumulate cAMP, and secrete excess Fetuin-A [16,17]. However, a role of Bicc1 in human ADPKD remains to be defined.
Bicc1 consists of three RNA-binding K homology (KH) and two KH-like domains at the Nterminus, a Gly-and Ser-rich intervening sequence (IVS) and a sterile alpha motif (SAM) at the C-terminus. SAM domains are found in over 4000 proteins (SMART, http://smart.emblheidelberg.de), often forming dimers or oligomers by head-to-tail self-association [18]. In bpk mutant mice, the frame-shifted C-terminus of Bicc1 is abnormally elongated by 149 aberrant amino acids that disrupt SAM-SAM interactions, thus underlining the importance of polymerization for Bicc1 function in vivo [19]. Bicc1 and its Xenopus homolog xBic-C are thought to inhibit canonical Wnt signal transduction by sequestering cytoplasmic Dishevelled [19,20] or by directly binding Wnt11 mRNA, respectively [21]. Inhibition of Dishevelled is potentiated by SAM-mediated stabilization of Bicc1 polymers in cytoplasmic foci, whereas KH domains recruit specific mRNAs [16,19,20]. Drosophila Bicaudal-C in "Malpighian" tubules (the structures corresponding to renal tubules) has been shown to bind myc mRNA and negatively regulates d-Myc protein levels [22]. Direct Bicc1 targets in kidneys are elusive, but likely include AC6 and protein kinase inhibitor (PKI) α mRNAs to mediate their loading unto miRNAinduced silencing complexes in a process that strictly depends on SAM domain polymerization [16,19]. However, this is not the only mechanism how Bicc1 regulates mRNA translation. In particular, when bound to, a 3'UTR fragment of Cripto-1 mRNA in early Xenopus embryos, xBic-C directly repressed 5' cap-dependent translation, independently of the SAM domain and of miRNA [23]. Furthermore, binding of Drosophila Bicaudal-C to its own mRNA has been shown to promote poly(A)-tail deadenylation by the CCR4-NOT complex, or to inhibit it, depending on the developmental context [24]. Consistent with potential roles in translation activation, Bicc1 has also been shown to increase the translation of Pkd2 mRNA, in this case by protecting the 3'UTR against miR-17 [14]. Furthermore, a novel translation-activating function of Bicc1 mediated by binding to eIF3 on mRNAs specifically at centrosomes seems to require fine-tuning by the orofacial-digital syndrome protein-1 (OFD1) to suppress renal cyst formation [25]. Thus, besides being epistatic to PC1 and PC2, Bicc1 likely mediates multiple functions that could be relevant in PKD and other cystic kidney diseases.
An unbiased approach to identify Bicc1-dependent processes is to analyse its protein interactome. In Drosophila ovaries, Bic-C interacts with CCR4-NOT deadenylase complex and with cytoplasmic polyadenylation element binding protein (CPEB) [24,26]. In addition, Drosophila Bic-C co-purified a complex of Me31B, Tral, PABP and Cup that is important to correctly secrete and localize the TGF-α homolog Grk during oogenesis [27]. By comparison, little is known about Bicc1 partners in mammals. In HeLa cells, Bicc1 co-immunoprecipitated with several other KH domain-containing proteins such as Sam68, GLD-1, GRP33, and Qk1 [28]. More recently, Bicc1 was identified as a binding partner of the ankyrin repeat-and SAM domain-containing protein ANKS6, which is mutated in cystic kidneys of a subset of nephronophthisis patients [29,30]. ANKS6 also binds the related Bicc1-interacting protein ANKS3, suggesting that all three proteins may act in a common cilia signalling pathway [30,31]. In keeping with a role in cilia signalling, Bicc1 has been detected in primary cilia from proximal tubule epithelial LLC-PK1 cells [32].
Here, we conducted a proteomic screen to identify novel Bicc1-interacting factors. Among 243 proteins that were enriched at least 2-fold by tandem affinity purification (TAP) of Bicc1, we focused on an interaction of Bicc1 with the CTLH complex because the orthologous complex in S. cerevisae triggers glucose-induced degradation of gluconeogenic enzymes. We show that Bicc1 maintains the levels of the gluconeogenic enzymes fructose-1,6-biphosphatase (FBP1) and phosphoenolpyruvate carboxykinase (PEPCK/PCK1), while inhibiting CTLH complex accumulation in mouse kidneys. Epistasis analysis shows that Bicc1 also stimulates FBP1 protein expression in mIMCD3 and LLC-PK1. Even though the CTLH complex did not inhibit FBP1 in these cell-based models, it negatively regulated Bicc1 protein levels, apparently by competing for the same SAM domain surfaces that are also required for self-polymerization and the silencing of Bicc1 target mRNAs. Our results suggest that Bicc1 is a target of CTLH complex and regulates metabolic processes.

Loss of Bicc1 increases WDR26 protein in newborn kidneys
To evaluate potential interactions with Bicc1 in vivo, we compared the expression levels of CTLH subunits in wild-type (WT) and Bicc1 -/newborn mice. Western blot analysis revealed 2-fold higher WDR26 protein levels in the knockout compared to WT (n = 5, P 0.01) and a similar trend for GID4 and RanBP9, but not for TWA1 (Fig 2A-2D). By contrast, no changes were observed in WDR26 or GID4 mRNAs (Fig 2E). These results indicate that Bicc1 attenuates the levels of some but not all CTLH subunits in vivo.

Bicc1 maintains normoglycemia and normal renal expression of the gluconeogenic enzymes FBP1 and PEPCK
In S. cerevisiae, the Gid complex polyubiquitinates FBP1 and PEPCK and induces their degradation to switch from gluconeogenesis to glycolysis in high glucose [37,39]. To assess glucose metabolism, we first measured blood glucose in WT and Bicc1 -/neonates. Average glucose levels decreased below 60 mg/dl compared to 80 mg/ml in wild-type (Fig 3A). Furthermore, RT-qPCR and Western blot analysis revealed 3.4-fold less PEPCK mRNA and 10-fold less protein, whereas FBP1 decreased 2-fold specifically at the protein level and only in Bicc1 -/kidneys (Fig 3B and  3C). By contrast in livers, where Bicc1 localizes to cholangiocytes and not to hepatocytes [19], FBP1 and PEPCK expression were unchanged (Fig 3D). These results establish that Bicc1 increases PEPCK and FBP1 expression in newborn kidneys and maintains normoglycemia.

Bicc1 upregulates FBP1 while attenuating the accumulation of several CTLH subunits
Since Bicc1 did not affect FBP1 transcription, we chose FBP1 to further assess how Bicc1 influences its protein level. Ectopic expression of HA-Bicc1 in HEK293T cells increased the accumulation of FBP1 (Fig 4A), whereas RNAi depletion of Bicc1 in mIMCD3 cells [16,40] decreased it (Fig 4B). Alternatively, we inactivated Bicc1 by CRISPR/Cas9 editing using single guide RNA. While Western blot confirmed the loss of Bicc1 protein, FBP1 levels varied among independent sgBicc clones (Fig 4C, S1 Fig). To evaluate whether Bicc1 regulates FBP1 through the CTLH complex, we depleted WDR26 or Twa1 or both by RNAi. FBP1 levels did not increase but rather decreased, and only in sgBicc cells, pointing to potentially complex layers of inhibitory mechanisms (Fig 4D). Depletion of WDR26 similarly failed to enhance FBP1 accumulation in LLC-PK1 proximal tubule cells, even though RNAi of Bicc1 inhibited it ( Fig  4E). FBP1 also was not significantly stabilized in cells treated with the proteasome inhibitor MG132 or with H+-ATPase inhibitor Bafilomycin A1, or both (S1 Fig). Thus, Bicc1 does not upregulate FBP1 simply by inhibiting the CTLH complex.
To assess the contribution of Bicc1 polymerization, we also tested the regulation of FBP1 by the non-polymerizing D913;K915;E916/AAA mutant Bicc1 MutD [19]. Bicc1 MutD failed to alter FBP1 expression (Fig 4F), indicating that the effect of wild-type Bicc1 described above is specific and likely depends on SAM domain-mediated polymerization. The CTLH complex and proteasomal degradation attenuate Bicc1 levels Rather than degrading FBP1, CTLH complex may target Bicc1. In keeping with this idea, depletion of CTLH subunits in mIMCD3 cells significantly increased endogenous Bicc1 protein, but not its mRNA (Fig 5A). To assess Bicc1 degradation, we treated mIMCD3 cells for 4 hrs with MG132 or Bafilomycin A1 or empty vehicle. Treatment with MG132 increased Bicc1 protein levels without inducing slow migrating intermediates characteristic of polyubiquitination (Fig 5B). By contrast, an apparent increase in Bicc1 levels after treatment with the H + -ATPase inhibitor Bafilomycin A1 was not statistically significant, and both drugs combined gave variable results, reflecting increased toxicity.
Previously, we have shown that Bicc1 is stabilized in large cytoplasmic foci by self-polymerization of its SAM domain [18,19]. To distinguish whether the CTLH complex interacts with Bicc1 in cytoplasmic foci or with a non-polymerized pool, or both, we performed co-immunostainings of HA-Bicc1 and endogenous RanBP9 in HEK293T cells. We only observed colocalization with RanBP9 peripheral to and outside bright Bicc1 foci and only at increased laser intensities (Fig 5C), indicating that access of CTLH complexes to macroscopically visible Bicc1 self-polymers may be limited.

The SAM domain mediates Bicc1 binding to the CTLH complex through its polymerization interfaces
To estimate the size of Bicc1-CTLH complexes, we fractionated HA-tagged Bicc1 on sucrose density gradients, using CNOT1 as a control (Fig 6A). HA-Bicc1 shifted a peak of CNOT1 from fractions 7-9 to fractions 9-11, all of which also contain ribosomal protein S6 (RPS6) [19] (S2 Fig). By contrast, RanBP9, TWA1, ARMC8, and MKLN1 peaked with low molecular weight (LMW) HA-Bicc1 in fraction 5, and GID4 and WDR26 peaked in fractions 1-3 or 7, respectively. These results indicate that the average size of Bicc1-CTLH complexes is smaller than the ribosome-sized complexes of Bicc1 with CNOT1. Since CTLH complexes did not accumulate in cytoplasmic Bicc1 foci (Fig 5C) or in high molecular weight (HMW) fractions (Fig 6A), we tested whether they bind Bicc1 independently of its SAM domain. Coimmunoprecipitation experiments in transfected HEK293T cells showed that deletion of the SAM domain inhibited Bicc1 binding to CTLH subunits, whereas deletion of the KH domains did not (Fig 6B). To validate that binding involves the SAM domain, we used GST-SAM fusion protein in pull-down assays in HEK293T cell extracts. Irrespective of the presence or absence of exogenous HA-Bicc1, GST-SAM pulled down endogenous TWA1, whereas GST alone did not (Fig 6C). As an additional control, we analyzed another SAM domain protein using Flagtagged ANKS3. Coimmunoprecipitation analysis revealed no Flag-ANKS3 binding to RanBP9, WDR26 or TWA1 above background levels (Fig 6D). However, co-transfection of full length ANKS3 or of the truncated mutant ANKS3ΔC lacking the C-terminal region distal of its SAM domain inhibited CTLH complex binding to HA-Bicc1 (Fig 6E), indicating that ANKS3 competes with CTLH complex for Bicc1. Previous structure analysis suggests that the SAM domain of ANKS3 interfaces with the ML and EH surfaces of the Bicc1 SAM domain, thereby allowing the bulky C-terminal region of ANKS3 to block Bicc1 polymer elongation [41]. To distinguish whether ML or EH surfaces of the Bicc1 SAM domain or their oligomerization also mediate CTLH complex binding, we mutated them individually [19] (S3 Fig). Control mutations MutA and MutF outside the ML and EH surfaces did not inhibit Bicc1 binding to any of the CTLH subunits analyzed, including WDR26, TWA1 and ARMC8. Binding of ARMC8 was also unaffected by mutations MutC or MutE that disrupt the ML or EH surface, respectively. In sharp contrast, disruption of either of these surfaces impaired WDR26 and TWA1 binding, and this defect was not rescued even if Bicc1 MutC and MutE were coexpressed to permit the formation of MutC/MutE dimers (S3 Fig). These data suggest that SAM domain oligomerization increases CTLH complex binding.

Yeast two-hybrid mapping of the mammalian CTLH complex reveals direct binding of ARMC8 to Bicc1 if SAM domain polymerization is inhibited
To further map the topology of CTLH and Bicc1 complexes, we conducted yeast two-hybrid assays. The strongest interactions among CTLH subunits were those of TWA1 with ARMC8, RMND5A, and RANBP9, whereas RanBP9 and RMND5A in turn strongly bound WDR26 and MAEA, respectively, (Fig 6F, S4 Fig). GID4 interacted with ARMC8 (GID5), albeit more weakly than its yeast homolog [42]. In contrast, no CTLH subunit alone directly associated with Bicc1, or vice versa. To test whether docking sites are buried in Bicc1 polymers, we coexpressed individual CTLH subunits with polymerization mutant Bicc1 MutD. Compared to wild-type, Bicc1 MutD showed increased binding of ARMC8 in two-hybrid assays, but not of other individual CTLH subunits (Fig 6F, S5 Fig). To independently validate the influence of SAM domain polymerization, we repeated the TAP-tag purification using Bicc1 MutD, and at lower salt concentrations that allowed Bicc1/CTLH complexes to retain MKLN1 ( Table 2). We found that compared to WT, Bicc1 MutD co-purified significantly less CTLH complex, even though binding to other proteins such as CAD, FASN and ACACA was not inhibited ( Table 2).
Taken together, these data suggest that the polymerization interfaces of the Bicc1 SAM domain act together with SAM-independent ARMC8 binding to recruit CTLH complexes (Fig 6G).

Discussion
Cystic growth in ADPKD reportedly involves anaerobic glycolysis accompanied by a decrease in the expression of gluconeogenic enzymes [7]. Here, a proteomic screen and validation by coimmunoprecipitation, yeast two-hybrid and density gradient fractionation assays revealed that Bicc1 interacts with the CTLH complex, the mammalian ortholog of the Gid complex that mediates degradation of gluconeogenic enzymes in S. cerevisiae. CTLH complexes recruited Bicc1 both independently of SAM domain polymerization via ARMC8 and through SAM domain surfaces that can be blocked by ANKS3 or buried in large Bicc1 self-polymers, suggesting CTLH complexes target Bicc1 oligomers devoid of an ANKS3 cap. Our analysis in cultured mIMCD3 cells indicates that CTLH complex curbs the levels of Bicc1 but not of the gluconeogenic enzyme FBP1. Nevertheless, deletion of Bicc1 in mice led to hypoglycemia and  diminished the expression of FBP1 and PEPCK specifically in kidneys but not in liver, correlating with increased accumulation of CTLH complex. These results reveal multi-layered regulation of FBP1 expression by Bicc1 and CTLH complexes: While Bicc1 overall mediates a net increase in FBP1, inhibition of self-polymerizing Bicc1 by the CTLH complex and vice versa provides a mechanism to adjust the levels of Bicc1.

The protein interactome of Bicc1 suggests a role in metabolism
Bicc1 and the ankyrin and SAM domain proteins ANKS3 and ANKS6 all suppress renal cysts and can coprecipitate each other [29][30][31]. In our interactome screen, stringent co-purification with TAP-tagged Bicc1 significantly enriched ANKS3 but not ANKS6, confirming that Bicc1 more tightly binds ANKS3 and independently of ANKS6 [43]. Since recent metabolomic analysis links ANKS3 to amino acid metabolism [44], future studies should investigate whether this process also depends on Bicc1. TAP-tag purification of Bicc1 also enriched several CCR4-NOT deadenylase subunits including CNOT1, NOT-5, -6/6L and -9. In addition, we found that Bicc1 can bind to the multifunctional enzyme carbamoyl-phosphate synthetase 2, aspartate transcarbamylase, dihydroorotase (CAD), which is rate-limiting for pyrimidine synthesis [45], and acetyl-CoA carboxylase alpha (ACACA) and fatty acid synthase (FASN) which function sequentially in fatty acid synthesis [46]. While regulation of these enzymes by Bicc1 remains to be tested, these interactions point to potential roles in several metabolic pathways.

Known functions of CTLH subunits
Here, we focused our attention on the CTLH complex. The core of the homologous Gid complex in S. cerevisiae is formed by Gid1, Gid5 and Gid9, held together by Gid8. In turn, Gid5 and Gid9 recruit Gid2 [47,48]. Gid2 and Gid9 are RING domain E3 ubiquitin ligases. Each contains a CTLH motif that is also found in Gid1 and Gid8 and in mammalian RanBP9 (GID1), RMND5a (GID2), TWA1 (GID8), MAEA (GID9), WDR26 (GID7) and in MKLN1 [34,47]. The CTLH subunit most enriched by TAP tag-purified Bicc1 was WDR26, a cytoplasmic WD40 domain protein that can recruit substrates to the E3 Ub ligase CUL4 and inhibit MAPK-induced activation of serum response transcription factors [49,50]. Bicc1 also enriched the CUL4-binding protein DDB1, but no CUL4. Possibly, WDR26 and/or DDB1 are substrate-specific adaptors for more than one E3 Ub ligase. RanBP9 was discovered as a partner of Ran, which regulates mitotic spindle assembly and nuclear and ciliary protein import [51][52][53]. RanBP9 also binds the KH domain protein Fragile Mental Retardation (FMRP) [54]. However, two-hybrid assay detected no RanBP9 interaction with Bicc1 KH domains.

Bicc1 attenuates renal expression of CTLH complex subunits
Bicc1 significantly diminished the accumulation of the CTLH subunit WDR26, and we noted a similar trend for RanBP9 and GID4. In S. cerevisiae, the Gid complex is activated by the accumulation of Gid4 within minutes after addition of glucose to degrade gluconeogenic enzymes [37,38]. A function for human GID4 is elusive. We detected GID4 in HEK293T cells and found that it co-fractionated with other CTLH subunits in sucrose density gradients. GID4 also interacted with RanBP9 and RMND5a in two-hybrid assays. Although this topology differs from S. cerevisiae where Gid4 is recruited by Gid5 [42], these data show that human GID4 can bind the mammalian CTLH complex. TAP-tagged Bicc1-SH copurified CTLH complexes that contained no detectable GID4. Possibly, Bicc1 reduces GID4 expression or inhibits GID4 recruitment to mammalian CTLH complex. Although, since GID4 expression was low and since binding to CTLH complexes is below detection even in the absence of Bicc1 [34], we could not conclusively test this hypothesis.

The CTLH complex binds Bicc1 in part via its self-polymerizing SAM domain and limits Bicc1 accumulation
We found that depletion of CTLH complex subunits by RNAi increased endogenous Bicc1 levels in mIMCD3 cells. In the simplest model, the CTLH complex accelerates Bicc1 turnover by direct binding. In keeping with this idea, Bicc1 interacted with the CTLH subunit ARMC8 in two-hybrid assays. ARMC8 has been shown to reduce the accumulation of E-cadherin and of α-and β-catenin to promote cell invasiveness in various cancers [55][56][57]. ARMC8 targets αcatenin for proteasomal degradation independently of polyubiquitination, despite associating with E3 ligases of the CTLH complex [47,58]. In two-hybrid assays, ARMC8 bound polymerization mutant but not wild-type Bicc1 or its isolated KH or SAM domains. Furthermore, in mammalian cells, ARMC8 and other CTLH subunits clearly segregated from HMW Bicc1 polymers during density gradient fractionation. SAM domain sequestration in HMW polymers may reduce direct access of ARMC8 to the intervening sequence between KH and SAM domains. In keeping with this model, only a diffusely distributed Bicc1 pool peripheral to microscopically visible cytoplasmic polymers overlapped with RanBP9 foci. Nevertheless, the Bicc1 SAM domain and its polymerization interfaces reinforced binding to assembled CTLH complexes, pointing to cooperativity among its subunits. In keeping with this idea, truncated Bicc1ΔSAM or polymerization mutants only inefficiently coimmunoprecipitated the subunits WDR26 and TWA1, and ARMC8 binding to polymerization mutant Bicc1 MutD during 2-step tandem affinity purifications was weakened compared to wild-type Bicc1. CTLH complex recruitment was also inhibited by overexpressed ANKS3, which competes for free Bicc1 SAM domain interfaces to cap their self-polymerization [43]. Taken together, these observations suggest that ARMC8 primes Bicc1 for recruiting additional CTLH subunits to exposed SAM domains. While the CTLH complex is not the only regulator of Bicc1 degradation, our finding that SAM domain polymerization interfaces mediate this interaction suggests that CTLH complexes preferentially target LMW Bicc1 oligomers at equilibrium with larger polymers when they are uncapped of ANKS3 (Fig 6G). It will be interesting to determine whether LMW forms of Bicc1 outside cytoplasmic foci may associate with cilia or centrosomes, and whether their availability is regulated by cell metabolism or vice versa.

Role of Bicc1 in glucose metabolism
A switch of glucose metabolism from glycolysis to gluconeogenesis is important during fasting and at birth when newborns feed on milk, which is low in glucose. During prolonged fasting, up to 25% of gluconeogenesis derives from extrahepatic sources such as kidney and intestine [59]. Prompted by its interaction with the CTLH complex, we tested whether Bicc1 regulates glucose metabolism. Despite hypoinsulinemia, blood glucose levels in Bicc1 -/neonates decreased more than 1.3-fold below wild-type levels. Among possible causes, we considered a role for Bicc1 in gluconeogenesis. Whereas PEPCK converts oxaloacetate into phosphoenolpyruvate, FBP1 catalyzes the rate-limiting step of fructose 1,6-bisphosphate hydrolysis to fructose 6-phosphate. RT-qPCR analysis revealed a 3.4-fold decrease in PEPCK mRNA levels specifically in Bicc1 -/kidneys but not in liver, and both PEPCK and FBP1 were further downregulated at the protein level. In the absence of glucose, ATP production largely depends on fatty acid oxidation. A resulting increase in the levels of ketone bodies can lead to systemic acidosis if kidneys cannot keep up with acid secretion. Acid secretion into urine requires ammonium production by glutaminolysis, combined with increased gluconeogenesis that drives bicarbonate synthesis and transport into the blood to buffer plasma pH [60]. Normally, hypoglycemia stimulates glucagon-induced upregulation of PEPCK expression and gluconeogenesis through cAMP/PKA signalling. Failure to induce renal PEPCK expression and gluconeogenesis despite elevated cAMP levels as observed in Bicc1 mutants predicts a defect in acid-base balance, which in turn likely accelerates disease progression. Indeed, accumulation of ammonium in kidney cortex combined with impaired urinary secretion stimulates interstitial nephritis, and it promotes cyst formation in the Han:SPRD rat model of ADPKD [61][62][63].

Regulation of FBP1
In S. cerevisiae, the Gid complex inhibits gluconeogenesis by targeting PEPCK and FBP1 for proteasomal degradation by the proline N-end rule [37,64]. FBP1 levels decreased almost 2-fold in Bicc1 mutant kidneys without a corresponding decrease at the mRNA level. Depletion of Bicc1 by RNAi also decreased FBP1 protein in cultured mIMCD3 and LLC-PK1 cells, whereas ectopic expression of Bicc1 in HEK293T cells increased it. To test whether endogenous Bicc1 stabilizes FBP1 by inhibiting the CTLH complex, we depleted the subunits TWA1 or WDR26. Knockdown of the CTLH complex by RNAi did not stabilize FBP1. We also observed no accumulation of FBP1 intermediates characteristic of polyubiquitination in cells treated with MG132. In basal-like breast cancer, FBP1 expression can be transcriptionally repressed by Snail1 to stimulate glycolysis and confer a metabolic advantage to cancer stem cells [65], and nearly all renal cell carcinomas deplete FBP1 to promote their aggressiveness [66]. Future studies should investigate whether this process involves Bicc1 or its interaction with CTLH complexes.

Analysis of kidneys and plasma glucose levels in Bicc1 mutant mice
Bicc1 +/heterozygous mice on a C57Bl/6 mouse genetic background were maintained in ventilated cages at the animal facility of the Ecole Polytechnique Fédérale de Lausanne (EPFL) [20].
All animal experiments were approved by the Veterinary Service of the Swiss canton of Vaud or by the Institutional Animal Care and Use Committee and adhered to the guidelines in the Guide for the Care and Use of Laboratory Animals (National Research Council. 2011. Guide for the care and use of laboratory animals, 8th ed. National Academies Press, Washington, DC). Newborn mice were decapitated at postnatal day P2, and total blood glucose was measured using the Accu-Check Aviva Nano glucometer (Roche). Kidney and livers were immediately snap frozen in liquid nitrogen after dissection. For use, the tissues were grinded with a pestle in lysis buffer and sonicated using a Bioruptor device (Diagenode). After centrifugation, protein concentration was quantified using Bradford assay (Bio-Rad).

Plasmids
To generate TAP-tagged MmBicc1 and MmBicc1-D913K915E916/AAA (Bicc1 MutD), a previously described HA-tagged mouse Bicc1 cDNA [20] and HA-tagged mouse Bicc1 MutD [19] were amplified by PCR and fused in-frame as a Kpn I restriction fragment to the SH tag in the inducible expression vector pcDNA5/FRT/TO (courtesy of Aebersold Lab). Alanine substitutions of residues D913, K915 and E916 in HA-Bicc1 have been described previously [19]. Briefly, the mutated DNA fragment was amplified by overlap extension PCR and then subcloned between Bgl II and Xba I sites of pCMV-SPORT6::HA-Bicc1. DNA fragments for Y2H assays were cloned by PCR in plasmid pACT2 distal to the GAL4 transactivation domain coding sequence, or in plasmid pGBKT7 distal to the GAL4 DNA-binding domain coding sequence using primers with unique restriction sites. The luciferase reporter plasmids Luc-AC6 3'UTRprox and Luc-PKIα 3'UTRprox in the vector pCS2+ have been described [16].

Indirect immunofluorescence
HEK293T cells were transfected with 1 μg of HA-Bicc1 in 6 well plates. After 24 hrs, the cells were plated on coverslips. At 48 hrs post-transfection, cells were washed with phosphate-buffered saline (PBS) and methanol-fixed during 10 mins at -20˚C. Cells were washed again with PBS and incubated with blocking solution containing 1% BSA for 1 hr at room temperature. Mouse anti-HA and rabbit anti-RanBP9 antibodies were diluted 1:500 and 1:100, respectively, in blocking solution and added to the cells during 2 hrs at room temperature. After washing with PBS, cells were incubated with Alexa 647-conjugated secondary donkey anti-mouse and Alexa 488-conjugated donkey anti-rabbit IgG during 1 hr at room temperature in the presence of 4',6-diamidino-2-phenylindole (DAPI). Pictures were acquired by confocal microscopy using a Zeiss LSM700 microscope.

Luciferase reporter assays
To measure luciferase expression, HEK293T cells were plated in 24-well plates and transfected on the following day with the indicated plasmids (100 ng/well) and with a lacZ expression vector (50 ng/well) in triplicate samples using jetPEI (Polyplus Transfection). 24 hrs after transfection, cells were lysed in buffer containing 25 mM Tris-phosphate, pH 7.8, 2 mM DTT, 2 mM 1,2-diaminocyclohexanetetraacetic acid (CDTA), 10% glycerol, and 0.5% Triton X-100. Cell extracts were diluted 20-fold, and luminescent counts detected in a Centro XS3 LB960 microplate luminometer (Berthold Technologies). Values were normalized to β-galactosidase activity measured by spectrophotometry in a Safire2 microplate reader. Results represent mean values from at least three independent experiments. Error bars show standard errors of the mean (SEM). Student's t-test was used to calculate P values.

Co-immunoprecipitation
HEK293T cells were transfected with the indicated expression plasmids in 10 cm dishes (we used 2 μg/dish). 24 hrs after transfection, cells were washed with PBS and proteins were extracted with lysis buffer containing 1X TBS, 1mM DTT, 0.05% NP-40, 1x protease inhibitor cocktail (Roche) and phosphatase inhibitor cocktail 3 (Sigma). Endogenous Bicc1 was immunoprecipitated from mIMCD3 cell extracts. In brief, confluent cells in 10 cm dishes were washed with PBS and resuspended in extraction buffer as described above. After sonication and centrifugation at 10'000 × g for 10 min, supernatants were incubated on a rotating wheel for 2 hrs at 4˚C with anti-HA agarose antibody or anti-FLAG M2 affinity gel (Sigma), or with protein G Sepharose beads (GE Healthcare) precoated with Bicc1 custom antibody [19] or pre-immune IgG (R&D Systems). Beads loaded on Mobicol columns were washed with 10 ml of Tris . HCl pH 7.4, 100 mM NaCl, 2.5 mM MgCl 2 , 1 mM DTT, and 0.02% NP-40 washing buffer and resuspended in Laemmli buffer. Elutions were fractionated on SDS-PAGE gels and analyzed by Western blotting.

GST pull-down
Recombinant GST-Bicc1 SAM domain was produced in the E. coli BL21 strain (Novagen) as previously described using pGEX-1λT expression vector [16]. GST

Sucrose density gradient fractionation
Subconfluent HEK293T cells were transfected with HA-Bicc1 or empty vector in 10 cm dishes (2 μg DNA/dish) using 3 plates per condition. Cell extracts were prepared after 24 hrs as described above for the GST pull-down assay. Continuous 15 to 60% sucrose gradients were prepared as described previously [19]. Fractions were recovered manually starting from the top, fractionated on SDS-PAGE gels, and analyzed by Western blotting. γ-tubulin was used as a control.

Statistical analysis
Error bars represent the standard error of the mean (SEM). Two-tail student-t test was used to compare the differences between 2 conditions to calculate p-values. 1-way ANOVA and Turkey's multiple comparison test was used to compare groups of unpaired values and determine the significance (p-value) of every mean compared to every other mean.