UAS domain of Ubxd8 and FAF1 polymerizes upon interaction with long-chain unsaturated fatty acids.

Ubxd8, a multidomain protein sensor for long-chain unsaturated fatty acids (FAs), plays a crucial role to maintain cellular homeostasis of FAs. Ubxd8 polymerizes upon interaction with long-chain unsaturated FAs, but the molecular mechanism involved in this polymerization remains unclear. Here we report that the UAS domain of Ubxd8 mediates this polymerization. We show that a positively charged surface area in the domain is required for the reaction. Mutations changing the positively charged residues in this area to glutamates prevented long-chain unsaturated FAs from inducing oligomerization of Ubxd8. Consequently, the mutant protein no longer responded to regulation by long-chain unsaturated FAs in cultured cells. Long-chain unsaturated FAs also induced polymerization of Fas-associated factor 1 (FAF1), the only other mammalian protein that contains a UAS domain homologous to that of Ubxd8. These results provide further insights into protein-FA interactions by identifying the UAS domain as a motif interacting with long-chain unsaturated FAs.

Fatty acids (FA) are crucial nutrients for cell survival, yet their overaccumulation is toxic to cells. Thus, cells develop multiple pathways to maintain their homeostasis. One of the pathways is feedback inhibition in FA synthesis. We have previously identifi ed Ubxd8 as a key regulator for the reaction ( 1,2 ). In cells depleted of FAs, Ubxd8 facilitates proteasomal degradation of Insig-1, a membrane protein of the endoplasmic reticulum (ER), through its interaction with the protein ( 1 ). Depletion of Insig-1 triggers proteolytic activation of sterol-regulatory element binding protein (SREBP)-1, a transcription factor synthesized as a membrane-bound precursor ( 1,3 ). This activation allows the into the in vitro assays. The fi nal concentration of ethanol in the reaction mixture was 1% for these assays.

Plasmid constructs
The following plasmids were described in the indicated reference: pCMV-Myc-Ubxd8 encoding human Ubxd8 with fi ve tandem copies of a c-Myc tag at its NH 2 -terminus under control of the CMV promoter ( 1 ); pCMV-Insig1-T7 encoding human Insig-1 followed by three tandem copies of a T7 epitope tag under control of the CMV promoter ( 9 ); and pAcHLT-Ubxd8( ⌬ 90-118) used for producing Ubxd8( ⌬ 90-118) in sf9 cells through recombinant baculovirus ( 2 ). pAcHLT-Ubxd8( ⌬ 90-118, ⌬ 360-445), pAcHLT-Ubxd8( ⌬ 90-118, ⌬ 278-445), pAcHLT-Ubxd8(122-277), pAcHLT-Ubxd8(1-66), pAcHLT-FAF1 and pAcHLT-FAF1(325-491) were generated to produce indicated fragments of Ubxd8 or FAF1 in sf9 cells through recombinant baculovirus as previously described ( 2 ). pCMV-myc-Ubxd8-VenusN encodes human Ubxd8 tagged at the NH 2 -terminus with fi ve tandem copies of a c-Myc tag fused at the COOH-terminus with the NH 2 -terminal fragment of the Venus protein under control of the CMV promoter and a neomycin-resistant gene. pCMV-fl ag-Ubxd8-VenusC encodes NH 2 -terminal fl ag epitope-tagged human Ubxd8 fused at the COOH-terminus with the COOH-terminal fragment of the Venus protein under control of the CMV promoter and a hygromycin-resistant gene. The NH 2 -and COOH-terminal fragments of the Venus gene were amplifi ed through PCR from pBiFC-VN155(I152L) and pBiFC-VC155 (Addgene), respectively. pBiFC bFos-VC155 encoding bFos fused with the COOH-terminal fragment of the Venus protein, pBiFC bJun-VN155(I152L) encoding bJun fused with the NH 2 -terminal fragment of the Venus protein, that long-chain unsaturated FAs also induce polymerization of full-length Ubxd8 in cultured cells. We demonstrate that this polymerization is mediated through the UAS domain present in Ubxd8. Moreover, we show that long-chain unsaturated FAs also specifi cally stimulate polymerization of Fas-associated factor 1 (FAF1), another mammalian protein that contains a UAS domain. These observations suggest that proteins containing UAS domains may subject to regulation by FAs.

Circular dichroism
Purifi ed proteins (1.3 µM) in 0.5 ml buffer B were measured for circular dichroism (CD) spectrum. The spectra were recorded at wavelength ranging 198-250 nm by a J-815 CD spectrometer (JASCO Inc.).

Structural modeling of the UAS domain in Ubxd8
The structure model of the UAS domain in Ubxd8 was generated by using the NMR structure of the FAF1-UAS domain (PDB ID: 2EC4_A) as the template with programs HHPred ( 11 ) and Modeler ( 12 ) incorporated at the MPI Bioinformatics Toolkit web server ( 13 ).

RESULTS
To identify the region in Ubxd8 that is required for long-chain unsaturated FAs to induce polymerization of the protein, we made various deletion mutants in Ubxd8( ⌬ 90-118) fused with a His 6 -tag at the NH 2 terminus, purifi ed the recombinant protein expressed in sf9 cells to homogeneity using Ni 2+ -affi nity followed by size exclusion chromatography, and analyzed the effect of FAs on polymerization of the proteins through blue native PAGE, a technique that allows detection of protein complexes in their native state ( 2,14 ). Oleate (C18:1) and arachidonate (C20:4), two classes of long-chain unsaturated FA, were able to induce polymerization of Ubxd8( ⌬ 90-118) in which the UBX domain (amino acid residues 360-445) ( Fig. 1A ) located at the COOH-terminal end of the protein was deleted ( Fig. 1B , lanes 4 and 5). Similar to Ubxd8( ⌬ 90-118), this polymerization was specifi c to longchain unsaturated FAs as medium-chain unsaturated FA dodecenoate (C12:1) and long-chain saturated FA stearate (C18:0) failed to produce the same effect ( Fig. 1B , lanes  2 and 3). This polymerization remained intact when a stretch of 82 amino acids located between the UAS and UBX domain (amino acid residues 278-359) was also deleted ( Fig. 1C ). These results suggest that the remaining two domains in Ubxd8, namely, the UBA and UAS domains ( Fig. 1A ), may be required for this polymerization. While long-chain unsaturated FAs failed to induce oligomerization of the purifi ed UBA domain (amino acid residues 1-66) ( Fig. 1D ), they were effective in stimulating polymerization of the purifi ed UAS domain (amino acid residues 122-277) ( Fig. 1E ). Consistent with these observations, deletion of the UAS domain from Ubxd8( ⌬ 90-118) signifi cantly inhibited unsaturated FA-induced polymerization of the protein ( Fig. 1F ). and pBiFC bFos( ⌬ zip)-VC155 encoding a mutant bFos, in which the bZIP domain was deleted, fused with the NH 2 -terminal fragment of the Venus protein were all acquired from Addgene. Mutations in Ubxd8 were generated through a site-directed mutagenesis kit (Stratagene) according to the manufacture's protocol.

Transient transfection
On day 0, SRD-13A cells were set up for experiments at 4.5 × 10 5 cells per 60 mm dish. On day 1, cells were transiently transfected with the indicated plasmids with FuGENE6 reagent (Roche Applied Science) according to the manufacturer's protocol. Conditions of incubation after the transfection are described in the fi gure legends.

Immunoblot analysis
Aliquots of the lysate were subjected to SDS-PAGE followed by immunoblot analysis. Antibodies used in the current studies were IgG-9E10 (1 g/ml), a monoclonal anti-T7 (0.4 g/ml), a polyclonal anti-Ubxd8 (1 g/ml), a polyclonal anti-actin (1∶2000 dilution). Horseradish peroxidase-conjugated donkey anti-mouse and anti-rabbit IgGs (0.2 g/ml) were used as the secondary antibody in all immunoblot analysis. Bound antibodies were visualized by chemiluminescence using the SuperSignal substrate system (Pierce) according to the manufacturer's instructions.

Live cell fl uorescent microscopy
On day 0, SRD-13A/pUbxd8-Venus cells were set up at 6 × 10 4 cells per 35 mm glass-bottom dish. On day 1, cells were switched to medium A supplemented with 5% delipidated FCS. On day 2, fl uorescent images were acquired with EX 492/EM 535 nm fi lter of Deltavision RT microscope in a 37°C chamber before and after incubation for 6 h with the indicated FAs in the same medium. ImageJ (http://rsbweb.nih.gov/ij) was used for quantifi cation of fl uorescent intensity. Relative fl uorescent intensity was calculated by dividing the intensity after the FA treatment by that before the treatment. The statistical analysis was performed with one tailed paired t -test.
We then determined the effect of the mutations on longchain unsaturated FA-induced polymerization of full-length If UAS domain is a motif that polymerizes upon interaction with long-chain unsaturated FAs, proteins other than Ubxd8 that contain a UAS domain should also polymerize in response to these FAs. Database search revealed that FAF1, a cytosolic protein ( 15 ), was the only protein expressed in mammalian cells containing a UAS domain homologous to that of Ubxd8. The UAS domain in human FAF1 is 27% identical to that of human Ubxd8 ( Fig. 2A ). We thus expressed the UAS domain of FAF1 and the full-length FAF1 protein in sf9 cells and purifi ed them to homogeneity. Blue native PAGE analysis indicated that long-chain unsaturated FAs also specifi cally induced polymerization of the UAS domain of FAF1 [FAF1 (325-491)] ( Fig. 2B ) as well as the full-length FAF1 protein ( Fig. 2C ).
While the structure of the UAS domain in Ubxd8 remains unsolved, the structure of the same domain in FAF1 has been determined through nuclear magnetic resonance (PDB ID: 2EC4_A). According to the structure, there is a surface area highly enriched in positively charged residues ( Fig. 3A , left  panel). The structural model built on the homology of the UAS domain between Ubxd8 and FAF1 ( Fig. 2A ) predicts that the same positively charged area also exists in Ubxd8 and that this area is made by two adjacent loops, each of which contains three positively charged residues (i.e., K167, R168, and R171 in one loop, and K239, R241, and R242 in the other loop) ( Fig. 3A , right panel). Since the negatively charged carboxyl group is required for long-chain unsaturated FAs to reconstitute the fl uorescent activity of the Venus protein ( Fig. 4A ). To use this approach to determine long-chain unsaturated FA-induced polymerization of Ubxd8, we stably transfected SRD-13A cells with these two plasmids and selected a clone of the cells (SRD-13A/pUbxd8-Venus) in which the Ubxd8 fusion proteins were expressed at an amount no more than that of endogenous Ubxd8 ( Fig. 4B ). We chose SRD-13A cells because they are mutant CHO cells auxotrophic for unsaturated FAs that can be easily depleted of these FAs by incubating them in medium free of FAs ( 2, 10 ). These cells were barely fl uorescent when they were incubated in the absence of FAs ( Fig. 4C, D , panel 1). Supplementation of the incubation medium with long-chain unsaturated Fas, such as oleate (C18:1) or arachidonate Ubxd8 in cultured cells. To perform the experiments, we had to fi rst develop an assay to measure such polymerization. We chose the approach of bimolecular fl uorescence complementation ( 16 ) to investigate long-chain unsaturated FA-induced polymerization of Ubxd8. For this purpose, we generated two plasmids encoding full-length Ubxd8 fused at its COOH-terminus with either NH 2 -terminal or COOH-terminal half of the Venus protein, a variant of yellow fl uorescent protein ( 16 ) ( Fig. 4A ). Cells in which Ubxd8 is not polymerized should not be fl uorescent because the NH 2 -and COOH-terminal halves of the Venus protein are separated from each other ( Fig. 4A ). Polymerization of Ubxd8 should bring the NH 2 -and COOHterminal halves of the Venus protein to close proximity to cells expressing the wild-type bFos fusion protein ( Fig. 5B ). This result suggests that oleate does not enhance fl uorescent intensity of the reconstituted Venus protein. Oleate also had no effect on fl uorescent intensity of the cells expressing the mutant bFos fusion protein ( Fig. 5C ). This observation suggests that oleate does not induce reconstitution of the Venus protein nonspecifi cally.
We then made the same mutations that disrupted longchain unsaturated FA-induced polymerization of Ubxd8( ⌬ 90-118) (K167E, R168E, R171E, K239E, R241E and R242E) in full-length Ubxd8. We stably transfected SRD-13A cells with plasmids encoding the mutant Ubxd8 fused with NH 2 -or COOH-terminal fragment of the Venus protein, and selected a clone of the cells [SRD-13A/pUbxd8(mutant)-Venus] in which expression of the mutant Ubxd8 fusion proteins was similar to that of wild-type Ubxd8 fusion proteins found in SRD-13A/pUbxd8-Venus cells ( Fig. 6A ). We determined the effect of the mutation on long-chain unsaturated FA-induced polymerization of the protein through bimolecular fl uorescence complementation. While arachidonate and oleate markedly enhanced the fl uorescent intensity of the cells transfected with the wild-type Ubxd8-Venus fusion proteins (SRD-13A/pUbxd8-Venus), these FAs failed to increase the (C20:4), dramatically enhanced the fl uorescent intensity of the cells ( Fig. 4C, D , panels 2 and 3). Quantifi cation of the fl uorescent signal in these images showed that these longchain unsaturated FAs increased the fl uorescent intensity by 6-to 7-fold ( Fig. 4E ). This effect was specifi c to long-chain unsaturated FAs, as medium-chain unsaturated FA C12:1 and saturated FA palmitate (C16:0) failed to produce the same result ( Fig. 4C, D , panels 4 and 5, and Fig. 4E ).
As a control, we also examined the effect of long-chain unsaturated FAs on interaction between bFos and bJun analyzed through the same bimolecular fl uorescence complementation approach ( 16 ). SRD-13A cells were transfected with a plasmid encoding NH 2 -terminal half of the Venus protein fused with bJun (bJun-VenusN) and a plasmid encoding COOH-terminal half of the Venus protein fused with either wild-type bFos (bFos-VenusC) or a mutant version of the protein in which the bzip domain required for interaction with bJun was deleted [bFos( ⌬ zip)-VenusC] ( 16 ). The fl uorescent intensity of the cells cotransfected with the plasmid encoding the mutant bFos fusion protein was only 2% of that cotransfected with the plasmid encoding the wild-type bFos fusion protein ( Fig. 5A ). Treatment with oleate did not affect the fl uorescent intensity of the cotransfection with a plasmid encoding either wild-type or the mutant Ubxd8. Arachidonate increased the amount of Insig-1 protein ( Fig. 6D , lanes 3 and 4). This effect was not altered by cotransfection with a plasmid encoding wild-type Ubxd8 ( Fig. 6D , lanes 5 and 6). In contrast to these results, arachidonate was unable to increase the amount of Insig-1 protein in cells cotransfected with a plasmid encoding the mutant Ubxd8 ( Fig. 6D , lanes 7 and 8), an observation suggesting that the mutant Ubxd8 stimulated degradation of Insig-1 even in cells treated with arachidonate. The results shown in Fig. 6C, D suggest that the mutation specifi cally impairs the response of the protein to long-chain unsaturated FAs without affecting the global folding of the protein.

DISCUSSION
The current study identifi es UAS domain as a motif polymerizing upon interaction with long-chain unsaturated FAs. UAS domain is a divergent member of the large thioredoxinlike protein superfamily ( 17 ). However, no other members of this superfamily are known to interact with FAs. We show that a predicted positively charged surface patch composed by six lysine and arginine residues located at two adjacent loops in UAS domain are important for long-chain unsaturated FAs to induce polymerization of the UAS domain. Thus, it is possible that long-chain unsaturated FAs may bind to the surface of the UAS domain. Such binding may render the protein surface more hydrophobic, a condition that may facilitate self-association of the UAS-FA complexes. Alternatively, these positively charged residues may not be required for FA binding but critical for polymerization of the domain upon interaction with the FAs. Obviously, structural analysis of the UAS domain in complex with long-chain unsaturated FAs is required to differentiate these possibilities. However, the heterogeneous sizes of Ubxd8 polymers pose a considerable challenge for such analysis.
A major discovery in the current study is that proteins containing a UAS domain may subject to regulation by long-chain unsaturated FAs. Indeed, Ubxd8 is such a protein. We show that long-chain unsaturated FAs induce polymerization of Ubxd8 to prevent the protein from stimulating degradation of Insig-1. FAF1 is the only other mammalian protein that contains a UAS domain homologous to that of Ubxd8. We show that long-chain unsaturated FAs stimulated polymerization of purifi ed recombinant FAF1, an observation suggesting that the function of FAF1 may also be regulated by these FAs. FAF1 has diverse functions ( 18 ), but none of the functions so far has been demonstrated to be regulated by FAs. FAF1 was identifi ed as a regulator for nuclear factor kappa-light-chain-enhancer of activated B cells (NF-B) (19)(20)(21), a transcription factor that activates proinfl ammatory cytokines ( 22 ). It was reported previously that saturated but not unsaturated FAs activated production of proinfl ammatory cytokines through activation of NF-B (23)(24)(25). These observations suggest that the FA sensor involved in this regulation should tell the difference between saturated and unsaturated FAs. Our in vitro analysis indicates that FAF1 meets this requirement. Thus, fl uorescent intensity of the cells transfected with the mutant Ubxd8-Venus fusion proteins [SRD-13A/pUbxd8(mutant)-Venus] ( Fig. 6B ). This result suggests that long-chain unsaturated FAs are unable to induce polymerization of the mutant Ubxd8 in cultured cells.
The lack of the response of the mutant Ubxd8 to longchain unsaturated FAs may be caused by a specifi c defect in response to these FAs or a global defect in protein folding. If the mutation specifi cally disrupts the interaction with longchain unsaturated FAs, the mutant protein should still be able to bind Insig-1 and to stimulate its degradation in FAdepleted cells, but the effect can no longer be inhibited by these FAs. On the other hand, if the mutation affects protein folding, the mutant protein should not be functional so that the protein is not expected to bind or degrade Insig-1. To differentiate these possibilities, we fi rst examined the interaction between Insig-1 and the mutant Ubxd8. For this purpose, we transfected SRD-13A cells with a plasmid encoding Insig-1 together with a plasmid encoding either wild-type or the mutant Ubxd8, treated the cells with the proteasome inhibitor MG132 to block degradation of Insig-1, and determined their interaction through a coimmunoprecipitation experiment. Insig-1 was coimmunoprecipitated with wildtype Ubxd8 in cells depleted of FAs but not those treated with arachidonate ( Fig. 6C , lanes 3 and 4). Insig-1 was also coimmunoprecipitated with the mutant Ubxd8, but the amount of Insig-1 coimmunoprecipitated was no longer reduced by treatment with arachidonate ( Fig. 6C , lanes 5 and 6).
We then determined the effect of the mutant Ubxd8 on degradation of Insig-1. To this end, we transfected SRD-13A cells with a plasmid encoding Insig-1 with or without densed with CoA immediately following their entry to cells. Consequently, free FAs are believed to be kept in very low concentration so that they may not reach high enough concentration to bind UAS domain-containing proteins in cytosol. This traditional view has been challenged by a recent study in which the amount of free FAs and their CoA conjugates in cells was measured directly. The result shows that the amount of free FAs is 100 times more than that of their CoA conjugates in cultured cells ( 29 ). Moreover, free FAs have been reported to be concentrated in membranes in their protonated form, and they are readily released into aqueous environment adjacent to the membranes as FA anions ( 30 ). Thus, the local concentration of free FAs at the interface between cytosol and ER membranes may future work is required to determine whether FAF1 is involved in FA-mediated activation of NF-B.
In addition to Ubxd8 and FAF1, Ubxd7 is the only other mammalian protein that contains a domain remotely resembling that of the UAS domain found in Ubxd8 and FAF1 ( 26 ). Ubxd7 facilitates degradation of hypoxia-inducible factor 1 alpha (HIF1 ␣) ( 26 ), a transcription factor activating genes in response to hypoxia ( 27 ). Ubxd7 also functions as an activator for cullin-RING ubiquitin ligases ( 28 ). It is currently unknown whether functions of Ubxd7 are regulated by long-chain unsaturated FAs.
Our fi nding that UAS domain is a sensor for long-chain unsaturated FAs suggests that these free FAs may serve as signaling molecules. It is assumed that free FAs are con- On day 1, they were transfected with a plasmid encoding T7 epitope-tagged Insig-1 (0.3 µg/dish) and a plasmid encoding wild-type or the mutant Ubxd8 fused with the Myc epitope tag (0.3 µg/dish), followed by incubation in medium A supplemented with 5% delipidated FCS. On day 2, cells were treated with 100 M arachidonate as indicated in medium A supplemented with 5% delipidated FCS and 30 M MG132 for 6 h. Detergent lysates of these cells were subjected to immunoprecipitation with anti-Myc to precipitate transfected Ubxd8. Pellets (representing 0.5 dish of cells) and supernatants (representing 0.1 dish of cells) of the immunoprecipitation were subjected to immunoblot analysis. (D) SRD-13A cells transfected with indicated plasmids were treated as described in (C), except that the cells were incubated in the absence of MG132. Cell lysates were subject to immunoblot analysis. be high enough to trigger polymerization of membraneanchored Ubxd8. Future studies are required to further delineate the roles of free FAs in cell signaling.
Ubxd8 is known to localize in the ER in FA-depleted cells but transport to the lipid droplets in cells treated with excess FAs as a result of increased synthesis of triglycerides (TG) (31)(32)(33). While TG synthesis is not required for Ubxd8 to stabilize Insig-1 in cells treated with unsaturated FAs ( 1 ), translocation of Ubxd8 to lipid droplets may be required to execute another function of the protein in maintaining cellular FA homeostasis, namely, regulating TG metabolism ( 2,33 ). We previously reported that Ubxd8 inhibited TG synthesis in FA-depleted cells by inhibiting the activity of diacylglycerol acyltransferase ( 2 ), a committed step in TG synthesis ( 34 ). A recent report showed that translocation of Ubxd8 to lipid droplets resulted in inhibition of TG lipolysis ( 33 ). It will be interesting to determine whether UAS domain-mediated polymerization of Ubxd8 is required for the protein to transport to lipid droplets in cells loaded with FAs.