Novel Biochemical and Structural Insights into the Interaction of Myristoylated Cargo with Unc119 Protein and Their Release by Arl2/3*

Primary cilia are highly specialized small antenna-like cellular protrusions that extend from the cell surface of many eukaryotic cell types. The protein content inside cilia and cytoplasm is very different, but details of the sorting process are not understood for most ciliary proteins. Recently, we have shown that prenylated proteins are sorted according to their affinity to the carrier protein PDE6δ and the ability of Arl3 but not Arl2 to release high affinity cargo inside the cilia (Fansa, E. K., Kösling, S. K., Zent, E., Wittinghofer, A., and Ismail, S. (2016) Nat. Commun. 7, 11366). Here we address the question whether a similar principle governs the transport of myristoylated cargo by the carrier proteins Unc119a and Unc119b. We thus analyzed the binding strength of N-terminal myristoylated cargo peptides (GNAT1, NPHP3, Cystin1, RP2, and Src) to Unc119a and Unc119b proteins. The affinity between myristoylated cargo and carrier protein, Unc119, varies between subnanomolar and micromolar. Peptides derived from ciliary localizing proteins (GNAT1, NPHP3, and Cystin1) bind with high affinity to Unc119 proteins, whereas a peptide derived from a non-ciliary localizing protein (Src) has low affinity. The peptide with intermediate affinity (RP2) is localized at the ciliary transition zone as a gate keeper. We show that the low affinity peptides are released by both Arl2·GppNHp and Arl3·GppNHp, whereas the high affinity peptides are exclusively released by only Arl3·GppNHp. Determination of the x-ray structure of myristoylated NPHP3 peptide in complex with Unc119a reveals the molecular details of high affinity binding and suggests the importance of the residues at the +2 and +3 positions relative to the myristoylated glycine for high and low affinities. The mutational analysis of swapping the residues at the +2 and +3 positions between high and low affinity peptides results in reversing their affinities for Unc119a and leads to a partial mislocalization of a low affinity mutant of NPHP3.

The existence of cilia in higher organisms was discovered a century ago, but research to explore the functional importance of cilia has only recently intensified. Cilia are highly specialized small antenna-like cellular protrusions that extend from the cell surface of almost all eukaryotic cell types. Primary cilia as sensory organelles are important for many cellular functions. They work as control centers of developmental signaling pathways, such as Hedgehog, or the induction of left-right asymmetry (2)(3)(4). Ciliary dysfunction leads to a range of diseases like Meckel-Gruber syndrome, Joubert syndrome, Bardet-Biedl syndrome, nephronophthisis, and polycystic kidney disease, which are commonly identified as ciliopathies (5).
Although the ciliary membrane is the extension of plasma membrane, the lipid composition is different from that of the plasma membrane (6). The composition and concentrations of proteins inside this compartment are also very different from the entire cell (7). A membrane barrier between cilia and the rest of the cell formed by a septin ring has been postulated such that the entry of transmembrane and membrane-associated proteins is tightly regulated (8 -10). This renders the ciliary membrane a very specialized compartment of the cell, which orchestrates proteins to achieve spatially controlled signaling pathways. However, the regulation of entry into and retention inside cilia and signals for such processes are still incompletely understood. It has been proposed that partition of proteins between cilium and cell body is directed by steric hindrance and/or cytoskeletal structures (11) or binding affinities between proteins and cytoplasmic elements (12).
We and others have previously shown that proteins with a C-terminal CaaX box motif, which are post-translationally modified with a farnesyl or geranylgeranyl moiety, are transported via the ␦ subunit of phosphodiesterase 6 (PDE6␦) (1,(13)(14)(15). PDE6␦ is a structural homologue to RhoGDI. It forms a ␤-sandwich fold with a deep hydrophobic pocket, which binds the prenylated cysteine and the adjacent three amino acid residues, two of which determine the affinity of the interaction. We have shown recently that the sorting of cargo between cilia and the rest of the cell depends on the affinity between cargo and PDE6␦, such that cargo with high affinity is destined for cilia, whereas low affinity cargo stays in the rest of the cell or is no longer exclusively retained inside the cilium (1). Unc119a (uncoordinated), also known as HRG4 (human retina gene 4) and Unc119b share 58% sequence identity and are homologous to PDE6␦. The C termini of Unc119a/b share the PDE␦-like ␤-sandwich domain, whereas they are considerably divergent in the N-terminal part. Unc119a has been shown to be localized at the basal body of the cilium, whereas the homologous Unc119b localizes to the transition zone and proximal cilium of RPE cells (16). Unc119a is also found in the eye and highly enriched in photoreceptor cells, in both rods and cones (17), supporting the notion that photoreceptors are considered as a specialized version of cilia (18 -21). Unc119a has been shown to be expressed in eosinophils, T-cells, lung fibroblasts, adrenal glands, cerebellum, and kidney (22)(23)(24)(25). Numerous studies have shown the involvement of Unc119a in the function of Src family kinases, Lyn, Fyn, Lck, and Hck, or as an inhibitor of Abl family tyrosine kinases, although the nature of such interactions had remained obscure (23)(24)(25)(26)(27)(28).
N-terminal myristoylation of proteins facilitates their reversible membrane binding activity (29). The mechanism by which myristoylated proteins are recruited to the proper membrane of cellular organelles remains unclear. Unc119a/b have been shown to bind specifically to N-terminal myristoylated proteins (30), and biochemical studies have shown that Unc119a/b proteins are involved in binding and shuttling of N-myristoylated proteins (16, 30 -32), suggesting that the early findings on the involvement of Unc119 in Src kinase function can be related to that observation. The determination of a structure between Unc119a and a lauroylated N-terminal peptide from the ␣ subunit of transducin has shown that Unc119a forms a ␤-sandwich structure very similar to that of PDE6␦ (32,33). Unc119a forms a hydrophobic pocket that accommodates a lipid moiety and a number of N-terminal residues of the peptide. Thus, like PDE6␦, Unc119a/b work as a carrier of these post-transitionally modified membrane-associated proteins. These proteins can also be considered chaperones that shield the lipid from the solvent (16,(32)(33)(34).
Previous studies have revealed a number of myristoylated proteins that interact with Unc119 (16,31,32). Binding of Unc119a/b to their interacting partners has been analyzed either qualitatively via yeast two-hybrid screening and in vitro pull-down assays or quantitatively through isothermal titration calorimetry (32) and polarization measurements (31). Previous studies have measured the affinities of either Unc1119a or Unc119b with the N-terminal myristoylated peptides of GNAT1, NPHP3, and Src (16, 30 -32). It has also been shown for Cystin1 and NPHP3 that the cargo is released from Unc119a only by Arl3⅐GTP and not Arl2⅐GTP (16,31). Recently, we have shown that high affinity binding between farnesylated cargo INPP5E and PDE6␦ leads to INPP5E recruitment to the ciliary membrane, and its release from PDE6␦ by Arl3 exclusively in cilia, whereas low affinity binding between farnesylated Rheb and PDE6␦ results in Rheb localization to the cytosol (1).
Here we have set out to perform a comparative more comprehensive analysis of the interaction between Unc119a/b and different myristoylated proteins, the interaction with Arl2 and Arl3, and finally the release of cargo by Arl2 and Arl3. The myristoylated proteins analyzed in this study are NPHP3, Cys-tin1, GNAT1, RP2, and Src. We find different affinities of Unc119a/b for Arl2 and Arl3 and for cargo peptides and differences in cargo release by Arl2/3. By structure-guided mutational analysis, we show how the difference in affinities can be manipulated between high and low affinity cargo. Our results suggest that similar to the transport of prenylated cargo into cilia, the import of myristoylated cargo might be determined by the cargo-carrier binding affinity and the cargo release specificity mediated by Arl2 and Arl3.

Results
Binding of N-terminal Myristoylated Peptides to Unc119a and Unc119b-Using fluorescence polarization, we measured the binding affinity of both Unc119a and Unc119b to N-terminal myristoylated peptides derived from transducin-␣ (GNAT1), NPHP3, Cystin1, RP2, and Src, which were labeled with fluorescein at the C terminus. Preliminary experiments had suggested that some of the affinities were too high to be reliably measured by equilibrium methods (data not shown; also see below). Thus, we used kinetic measurements via stopped flow instead to determine association and dissociation rate constants and obtain K D (Fig. 1). Fig. 1A shows the association of 1 M fluorescein-labeled N-terminal myristoylated GNAT1 peptide with increasing concentrations of Unc119a. The association of the myristoylated GNAT1 peptide with Unc119a (0.2-20 M) leads to the increase in fluorescence polarization. Plotting of the observed rate constants (k obs ) against the concentration of Unc119a resulted in the determination of the association rate constant (k on ) of the reaction (Fig.  1B). The association rate constant (k on ) for Unc119a is 4.6 M Ϫ1 s Ϫ1 and is very similar to Unc119b (6.3 M Ϫ1 s Ϫ1 ) measured under the same conditions (Fig. 1B). Association rate measurements were also measured for the N-terminal myristoylated peptides from NPHP3, Cystin1, RP2, and Src under similar experimental conditions, and the association rate constants (k on ) obtained are shown as a bar diagram in Fig. 1C and numerically in Table 1. These values are very similar and range from 3.4 to 12 M Ϫ1 s Ϫ1 . Association of myristoylated cargo is generally slightly faster for Unc119b than for Unc119a.
The dissociation rate constants were determined by incubating a complex of C-terminal fluorescein labeled N-terminal myristoylated peptides with Unc119a/b in the presence of a 100-fold excess of unlabeled peptides, as shown for the GNAT1 peptide (Fig. 1D). In contrast to k on , the k off values vary considerably among the different peptides over a 900-fold range from 0.0002 to 1.74 s Ϫ1 (Fig. 1E and Table 1). NPHP3 and Cystin1 show the slowest release from Unc119a with 0.002 and 0.006 s Ϫ1 , whereas Src peptide has the fastest off rate close to 1 s Ϫ1 . The rate for RP2 is intermediate between very slow and very fast rates. The dissociation rate constants of Unc119a and -b for NPHP3, Cystin1, RP2, and Src are rather similar, but for GNAT1, there is a Ͼ10-fold difference between Unc119a and Unc119b, with k off values of 0.0023 and 0.035 s Ϫ1 , respectively.
The equilibrium dissociation constants (K D ) for the complexes between Unc119a/b and myristoylated peptides were calculated as the ratio of k off /k on ( Fig. 1F and Table 1). The data revealed three different categories of affinity: (i) the very tight binding with subnanomolar affinities for Cystin1 and NPHP3 and low nanomolar affinity for GNAT1, (ii) the intermediate binding affinity in the two-digit nanomolar range for RP2, and (iii) the low affinity in the submicromolar range for Src (145-252 nM) ( Fig. 1F and Table 1). The affinities of Unc119a and Unc119b for myristoylated peptides are similar except for GNAT1, which shows Ͼ10-fold higher affinity for Unc119a as compared with Unc119b, exclusively due to the different dissociation rates. The affinities determined here by kinetics are different from those determined earlier by an equilibrium method (31). As indicated above, equilibrium binding assays are not quite suitable for such high affinities (picomolar or sub-nanomolar) because such affinity measurements by titration result in straight lines that look like active site titrations and only give upper limits for the K D .
Binding Affinity of Unc119a and Unc119b to Arl Proteins-We have shown previously that Unc119a and PDE6␦ bind to Arl2/3 with high affinity (1,35,36). We showed that the binding affinities of Unc119a and PDE6␦ are 20 -30-fold higher for Arl3 as compared with Arl2 (1, 36) because of the additional contribution by the N-terminal helix of Arl3 to the interaction. To verify whether the same holds true for Unc119b and to compare with Unc119a, we measured the kinetics of interaction using Reactions were carried out at 25°C in buffer A. Observed rate constants (k obs ) of associations were obtained by single exponential fitting of individual curves. B, the observed association rate constants k obs for interaction of Unc119a and Unc119b with N-terminal myristoylated cargo peptide (GNAT1) obtained as in A were plotted against the concentration of Unc119a and Unc119b. The association rate (k on ) is represented by the slope. The k on values for all N-terminal myristoylated cargo peptides (GNAT1, NPHP3, Cystin1, RP2, and Src) are summarized in a bar diagram (C) and in Table 1. D, dissociation of cargo was measured at 25°C in buffer A by monitoring the decrease of fluorescence polarization after incubating a complex of fluorescein labeled peptide with Unc119 (1 M) with a 100-fold excess (100 M) of unlabeled peptide. Dissociation rate constants (k off ) were obtained by single exponential fitting of the data. All k off values are summarized in a bar diagram (E) and in Table 1. F, equilibrium dissociation constants (K D ) were calculated as ratios, k off /k on , and the values are plotted as a bar diagram and summarized in Table 1. G, the sequences of N-terminal myristoylated peptides used in this study. For labeled peptides, fluorescein fluorophore was attached at the C terminus of N-myristoylated peptides. All bar graphs show the average of 5-7 measurements for the experiments performed by stopped flow instruments and an average of three measurements for the experiments performed by the Fluoromax instrument. Error bars, S.D., for (C) n ϭ 5-7 and for (E) n ϭ 3.
full-length Arl2/3 labeled with mant-GppNHp 3 (fluorescent, non-hydrolyzable analog of GTP). The association rates with increasing concentrations of Unc119a and -b were measured with fluorescence polarization using a stopped flow instrument (Fig. 2). Association rate constants for Arl2/3⅐GppNHp were obtained as described above by plotting the observed rate constants of association against the concentration of Unc119a/b ( Fig. 2A). The results are shown as a bar diagram in Fig. 2B and in numerical form in Table 2. The association rate constants (k on ) are on the order of 1-3 M Ϫ1 s Ϫ1 . The association rates are Ͼ3-fold higher for Arl3 than for Arl2 in the case of Unc119b, whereas it is only 2-fold in the case of Unc119a. Dissociation rates were measured by incubating the fluorescent complex of Unc119a/b and Arl2/3⅐mant-GppNHp with a 100-fold excess of unlabeled Arl2/3⅐GppNHp, as shown for Arl2 and Unc119a/b (Fig. 2, C and D). The dissociation rate constants (k off ) are shown as a bar diagram in Fig. 2E and numerically in Table 2. The dissociation rate constants (k off ) are slower for Arl3 than Arl2 for both Unc119a and Unc119b. The equilibrium dissociation constants for Unc119a/b toward fulllength Arl2⅐GppNHp and Arl3⅐GppNHp were calculated as the ratio of k off /k on ( Fig. 2F and Table 2). The kinetic data suggested the binding affinities of Arl3 toward both Unc119a and -b are higher as compared with Arl2. Whereas, the 14-fold difference in affinity between Arl3 and Arl2 toward Unc119a is mediated only by the dissociation rates, the difference for Unc119b is due to a combination of different association and dissociation rates.

Release of Myristoylated Cargo by Arl2/3 Proteins-The
Arflike small GTP-binding proteins Arl2 and Arl3 act as displacement factors for lipid-modified proteins bound to the GDI-like solubilizing factor PDE6␦ as well as to Unc119a/b (15,31,33,34). Previously, it has been shown that Arl3, but not Arl2, allosterically regulates the release of ciliary proteins from Unc119a/b (16,31). We thus analyzed the specificity of Arl2and Arl3-mediated release of the high and low affinity peptides analyzed above for both Unc119a and Unc119b (Fig. 3). Arl2⅐GppNHp or Arl3⅐GppNHp was added to the preformed complex of fluorescent peptides from GNAT1, NPHP3, Cys-tin1, RP2, and Src with Unc119a and Unc119b. The release of peptide from the complex is scored as a decrease in fluorescence polarization. Under the conditions used, Arl3⅐GppNHp was able to disrupt the complex of Unc119a with all five peptides (Fig. 3, A and C). In contrast, Arl2⅐GppNHp was only able to disrupt the complex of Unc119a with the low or intermediate affinity RP2 or Src peptides, whereas the high affinity GNAT1, Cystin1, and NPHP3 peptides were fully resistant to release by Arl2 under the conditions used (Fig. 3, B and D). The results were consistent with the previous studies, which used GNAT1 and NPHP3 bound to Unc119a (16,31). When comparing Unc119a and -b, similar results were obtained for the specificity of Arl2 and Arl3 (Fig. 3, A and B versus C and D). These results indicate that the high affinity cargos GNAT1/NPHP3/Cystin1 are only released by Arl3⅐GppNHp, whereas Arl2⅐GppNHp can release only the low affinity cargo RP2 and Src. Myristoylated GNAT1 was partially released by Arl2⅐GppNHp when bound to Unc119b but not Unc119a, possibly due to the 10-fold affinity difference observed above (Fig. 3D).
Because these data were obtained under equilibrium conditions and are somewhat difficult to compare quantitatively, we determined the Arl2/3⅐GppNHp cargo release acceleration from Unc119a/b in the presence of excess unlabeled peptides. The effects of Arl3⅐GppNHp and Arl2⅐GppNHp on the dissociation rates were determined in the presence of a 100-fold excess of unlabeled peptides to silence the back-reaction and were measured as a decrease in polarization in a stopped flow instrument. The data for the release of GNAT1 peptide from either Unc119a or -b are shown in Fig. 4, A and B, respectively, and the dissociation rates for release of GNAT1, NHP3, Cystin1, RP2, and Src are summarized in Table 1. The presence of Arl3 increases the off rates for high affinity peptides GNAT1 (2.8   Table 2. C and D, the dissociations of full-length Arl2⅐GppNHp and Arl3⅐GppNHp from a complex with Unc119a/b (0.5 M) at 25°C in buffer A were measured as decreases of fluorescence polarization after the addition of a 100-fold excess (50 M) of unlabeled Arl2/3⅐GppNHp. E, observed dissociation rate constants (k off ) were obtained by single exponential fitting of the data. All k off values are plotted as a bar diagram and summarized in Table 2. F, equilibrium dissociation constants (K D ) were calculated as ratios of k off /k on , and the values here are plotted as a bar diagram and appear in Table 2    Crystallization and Structural Analysis of Unc119a-NPHP3 Complex-To gain molecular insights and to understand the nature of the binding affinity difference between low and high affinity myristoylated peptides toward Unc119a/b, we aimed to solve the crystal structure of peptides in complex with Unc119a or Unc119b. Using full-length Unc119a/b and myristoylated GNAT1, NPHP3, RP2, and Src peptides of 10-residue length, we did not obtain any suitable crystals. The sequence alignment of Unc119a and Unc119b reveals that they share 58% sequence identity. The N-terminal 50 residues of Unc119b are the most variable region. Because the first 57 residues of Unc119a do not have any effect on peptide binding (16,33), we deleted these residues of Unc119a. However, the complex of the truncated Unc119a with peptides did not crystallize either. Finally, we used myristoylated peptides of 6-residue length and in addition added limited amounts of both trypsin and chymotrypsin to the crystallization setup. With this strategy, we were able to obtain crystals and solve the crystal structure of myristoylated NPHP3 peptide in complex with Unc119a at 2.1 Å resolution (data collection and refinement statistics summarized in Table 3). Trypsin and chymotrypsin protease were used for in situ proteolysis. The flexible loop of Unc119a, which is located at the entry of the myristoylated peptide, is cleaved by trypsin and chymotrypsin and apparently facilitates crystal packing. Superimposition of the NPHP3 peptide-Unc119a complex with the previously solved structure of Unc119a in complex with lauroylated GNAT1 peptide (PDB code 3RBQ) shows that the ␤-sandwich fold of Unc119a in both complexes superimposes well with a root mean square deviation of 0.9 Å (Fig. 5A, left). The myr-istoyl moiety of NPHP3 inserts into the hydrophobic cavity of Unc119a in a very similar pattern as the lauroyl moiety of GNAT1. The two additional carbon atoms of the myristoyl moiety are suited inside the hydrophobic cavity by a higher level of ramification and slightly deeper insertion. Nevertheless, both moieties maintain an identical interaction pattern with the surrounding hydrophobic residues of Unc119a (Fig. 5A,  left).
The myristoyl-glycine ester bond makes hydrophilic interactions with Tyr-131 and Glu-163 (Fig. 5A, right). As in the structure with the GNAT1 peptide, the first residues of NPHP3 also form helical turns. The interaction of the carbon chain with the pocket involves the hydrophobic residues Ile-93, Val-143, Phe-175, Tyr-194, Tyr-134, Phe-91, Phe-137, and Phe-196. Fig. 5B shows that the alanine and serine side chain of the NPHP3 peptide at positions ϩ2 and ϩ3 after the myristoylated glycine (the 0 position) are situated in a tightly structured environment. Comparing the equivalent ϩ2 and ϩ3 residues from the other peptides analyzed here, we observe that the high affinity Unc119 binders (GNAT1, NPHP3, and Cystin1) have small size side chain residues, such as alanine, serine, or glycine (Fig. 5B). In contrast, the low or intermediate affinity Unc119 binders (Src and RP2) possess residues with bigger side chains, such as lysine, phenylalanine, or glutamine. We thus hypothesized that the residues with bigger side chains at the ϩ2 and ϩ3 positions would create steric clashes with Unc119a, resulting in lower binding affinities, as found for Src and RP2.
Mutational Analysis of ϩ2 and ϩ3 Positions-To verify our model, we generated two 10-residue mutant peptides, where the residues in the ϩ2 and ϩ3 positions were swapped between high (NPHP3) and low (Src) affinity binders, creating NPHP3(NK) and Src(AS) peptides. The affinities of these swapped peptides toward Unc119a were measured by titrating increasing amounts of unlabeled NPHP3(NK) (Myr-GTNK-SLVSP) and Src(AS) (Myr-GSASSKPKD) peptides into a preformed complex of a fluorescent RP2 peptide (0.1 M) with Unc119a (0.2 M) and monitoring the displacement of the fluorescent peptide by the unlabeled ones, scored as the decrease of fluorescence polarization. The data were analyzed with a competition model equation derived from the law of mass action as described before (1,37,38), using Origin software. The affinity of NPHP3(NK) to Unc119a decreased to a K D of 940 Ϯ 48 nM, whereas the affinity of Src(AS) increased to a K D value of 6.2 Ϯ 1.8 nM (Fig. 5C). These results clearly show that increasing the size of the amino acid side chains at the positions ϩ2 and ϩ3 decreases the binding affinity with Unc119a, and vice versa. Taken together, the residues at the ϩ2 and ϩ3 positions relative to the myristoylated glycine seem to determine the binding affinity between Unc119a and cargo.
To test whether the reduced affinity of NPHP3(NK) affects its ciliary localization, we used a fragment of NPHP3 (residues 1-203) that is very similar to one described previously for its consistent ciliary localization (16). Constructs containing wild type NPHP3 and mutant NPHP3(NK) were stably transfected into IMCD3 cells, and the ciliary localization of the proteins was measured by quantifying and comparing the fluorescence inside and outside the cilium in ciliated cells. Fig. 6 shows that NPHP3(WT) is highly enriched in cilia compared with the cell body (Fig. 6A), whereas considerable amounts of the NPHP3(NK) mutant are localized to the rest of the cell (Fig. 6B). Quantification of the fluorescence intensity of mCherry-tagged protein was done using CellProfiler. The mean fluorescence intensity ratio between cilia and whole cell shows that the wild type protein has an 8.3-fold enrichment inside the cilia. The NPHP3(NK) mutant loses its almost exclusive ciliary localization and is more distributed over the entire cell, with only a 4.4-fold ciliary enrichment (Fig. 6C).
Our experiments reveal that ciliary cargo proteins NPHP3, Cystin1, and GNAT1 bind to Unc119a and Unc119b with picomolar to low nanomolar affinity, whereas the non-ciliary cargo Src has a submicromolar affinity. RP2, which has been shown to localize at the ciliary base, binds with an affinity in the 2-digit nanomolar range. As is typical for such binary interactions, the difference in affinity is mostly dictated by the dissociation rates, which differ between Unc119a and -b by a similar factor, whereas the association rates are very similar. There is no significant difference between Unc119a and -b for the binding affinities of myristoylated peptides except for GNAT1. Its interaction with Unc119a is 10-fold tighter and is determined by a 10-fold difference in the off-rate. Unc119a was originally found as a retina-specific gene named HRG4. It is highly expressed in photoreceptor cells, a specialized form of cilia, and the protein is reported to be localized in the inner segments and photore- ceptor synapses (51). Our quantitative and comparative study for Unc119a/b binding to myristoylated cargo suggests that the binding affinity between cargo and Unc119a/b seems to be very important for the Unc119a/b-mediated sorting into the ciliary compartment. This is a scenario similar to what we have recently demonstrated for the sorting of prenylated cargo, such as INPP5E, into cilia (1). The cell biology experiments show the mislocalization of the mutant NPHP3 construct, which is no longer highly enriched in cilia but is now visible over the whole cell. We propose that this mislocalization results from its reduced affinity to Unc119 protein. The fact that the localization cannot be completely reversed by reducing the affinity argues for an additional retention signal operating inside cilia, something that has also been proposed for the localization of prenylated proteins (1).
Although Arl2 and Arl3 share several GTP-dependent interacting partners like the GDI-like solubilizing factors Unc119a, Unc119b, and PDE6␦, in addition to BART and BARTL1 (52), their cellular functions are distinct. Our data show that high affinity myristoylated cargoes (GNAT1, NPHP3, and Cystin1) bound to Unc119a/b are specifically released only by Arl3, whereas lower affinity cargo is released by both Arl2 and Arl3. One exception is GNAT1, which can be partially released from its complex with Unc119b but not with Unc119a by Arl2, possibly due to the 10-fold lower affinity of GNAT1 to Unc119b as compared with Unc119a. We have previously shown that the Arl3-specific GEF Arl13B (53) is exclusively localized inside cilia, whereas RP2, the Arl3-specific GAP (44), localizes to the base of cilia and the periciliary region (47,48,52), thus making cilia an Arl3⅐GTP-enriched compartment. This suggests that Arl3 can release NPHP3 and Cystin1 cargo only inside the cilia and GNAT1 in the outer segment of photo-receptors, whereas Arl2 is capable of releasing intermediate affinity (RP2) and low affinity cargo (Src) outside cilia. Our data support the notion that (i) Unc119a/b proteins regulate the trafficking of myristoylated transducin ␣ subunit (GNAT1), NPHP3, and Cystin1 into the cilia, and (ii) Arl2/3 proteins regulate the sorting of myristoylated cargo into the cilia. The fact that the affinities of Unc119 for its ciliary myristoylated cargo are higher than for Arl3 suggests a very high concentration of activated Arl3 inside the cilium and/or that, in addition to the release mechanism by Arl3, a retention signal for cargo proteins must be in operation to drive the equilibrium toward full release. Such retention could be achieved by the interaction with the ciliary membrane or other ciliary interacting partners.
Unc119a has been shown to localize to the centrosome at the ciliary base, whereas Unc119b localizes to the transition zone and proximal cilium of RPE cells (16). Unc119a and Unc119b proteins share a 58% identity. The sequence comparison of Unc119a and -b shows that the N terminus (residues 1-55) is the most variable region and is not involved in the binding to myristoylated proteins. However, the N terminus might be important and responsible for the different localization of the orthologues. The crystal structure revealed that Unc119a forms a hydrophobic pocket to accommodate the myristoyl moiety that is formed by the residues Phe-91, Ile-93, Tyr-131, Phe-137, Val-143, Glu-163, His-165, Phe-175, Tyr-194, Phe-196, and Tyr-234 of Unc119a (Fig. 5A). The sequence comparison between Unc119a and -b shows that these residues are conserved in Unc119b (Fig. 7), suggesting a similar mode of interaction between Unc119a/b and myristoylated cargo.

Experimental Procedures
Plasmids and Proteins-Constructs of full-length human Unc119a and Unc119b were cloned into the pGEX-4T3 and pET28a vectors containing an N-terminal GST and His 6 tag, respectively. The construct of Unc119a(58 -240) in pGEX-2T we received as a gift from the laboratory of Prof. Baehr and was later recloned in pET28 vector. The C-terminal His 6 tag fulllength constructs of Arl2 and Arl3 in pET20b vector were already available in the laboratory. All proteins were expressed as GST or His 6 fusion proteins from Escherichia coli BL21(DE3) CodonPlusRIL, isolated in a first step by affinity chromatography on a glutathione-Sepharose and Talon column respectively, and purified after proteolytic cleavage of GST in a second step by size exclusion chromatography (Superdex S75), as described before (35,44,54). After purification, both Arl2 and Arl3 full-length proteins were bound to GDP, as detected via HPLC. The exchange for the GTP analog GppNHp and to mant fluorophore-labeled mant-GppNHp was done as described before (35,54). The amount of protein-bound nucleotide was analyzed and quantified by C18 reversed-phase column with HPLC.
Peptides Synthesis of the Peptides-Solid phase peptide synthesis was carried out on a 0.1-mmol scale using a CEM-Liberty peptide synthesizer equipped with a CEM-Discover microwave. Washing steps between coupling and deprotection were carried out in DMF and DCM using 1 ml of solvent per 100 mg of resin. The Fmoc protecting group was removed with a solution of piperidine in DMF (20%, v/v), 1 min at 30°C (intensity ϭ 40 watts), and 5 min at 70°C (intensity ϭ 40 watts). All amino acid couplings were performed in DMF and repeated twice. Typically, Fmoc-Xaa-OH (4 eq, 0.2 M), HCTU (4 eq), and DIPEA (8 eq) were reacted for 10 min at 80°C (intensity ϭ 20 watts). Upon the completion of the automated synthetic cycles, the resin was washed with DCM (5 ml ϫ 5), DMF (5 ml ϫ 5), and DCM (5 ml ϫ 5). The C-terminal allyl group was removed with a mixture of Pd(PPh 3 ) 4 (20 mol %), PhSiH 3 (14 eq) in dry THF (3 ml for 0.1 mmol of peptide on the resin) under an argon atmosphere for 12 h to deprotect them. Upon the completion of the reaction, the resin was filtered under vacuum and washed with dry THF (5 ml ϫ 5), DCM (5 ml ϫ 5), DMF (5 ml ϫ 5), and DCM (5 ml ϫ 5). Fmoc-(PEG3)-NH 2 (4 eq with regard to the resin loading) was dissolved in dry DMF (0.2 M). [O-(6-chloro-1H-benzotriazol-1-yl)-N,N,NЈ,NЈ-tetramethyluronium hexafluorophosphate] (HCTU) (4 eq) and diisopropylethylamine (DIPEA) (8 eq) were subsequently added, and the resulting mixture was shaken for 5 min. The peptide-containing resin was shaken with the reaction mixture for 4 h at the ambient temperature. The resin was filtered under vacuum and washed with DCM (5 ml ϫ 5), DMF (5 ml ϫ 5), and DCM (5 ml ϫ 5). The coupling and the washing step were repeated twice. The Fmoc group was removed by shaking the resin with piperidine (20% in DMF, 5 ml) for 40 min at the ambient temperature twice. The resin was washed with DMF (5 ml ϫ 5). Fluorescein isothiocyanate (5 eq) and DIPEA (5 eq) were dissolved in dry DMF (0.2 M), and the resin was shaken with the mixture for 4 h at the ambient temperature. The resin was washed with DCM (5 ml ϫ 5), DMF (5 ml ϫ 5) and DCM (5 ml ϫ 5). The coupling and the washing steps were repeated twice. The peptides were fully deprotected and cleaved from the resin with 5 ml of the mixture of DCM/ TFA/TES (50:25:25). The resin was shaken with the deprotection mixture for 2 h and filtered into a round bottom flask. The resin was washed with DCM (5 ml ϫ 3), DCM/MeOH (1:1, 5 ml ϫ 3), and MeOH (5 ml ϫ 3). The combined liquids were diluted with toluene (10 ml) and concentrated under reduced pressure. The resulting slurry was diluted again in toluene (10 ml) and co-evaporated again. The co-evaporation was repeated three times. The resulting crude products were purified by preparative HPLC using a C4 column and characterized by HRMS. The purity of the peptides exceeded 95%.
Fluorescence Measurements-All fluorescence polarization measurements were performed in buffer containing 30 mM HEPES, pH 7.5, 5 mM MgCl 2 , 100 mM NaCl, and 3 mM DTE (buffer A) at 20°C. The kinetic measurements were performed with a stopped flow instrument (Applied Photophysics, Leatherhead, UK) in the polarization mode, and fluorescence polarization experiments were performed with a Fluoromax-2 spectrophotometer instrument (Horiba Jobin Yvon, Munich, Germany). The excitation wavelengths were 366 nm for mant and 490 nm for fluorescein fluorophore, whereas the emission wavelengths used for mant and fluorescein fluorophore were 450 and 520 nm, respectively. Emission in stopped flow was detected through a cut-off filter (Schott glass) of 420 and 500 nm for mant and fluorescein, respectively. Data were analyzed using GraFit version 5.0 (Erithracus Software).
Crystallization and Structure Determination-The myristoylated N-terminal peptide of NPHP3 (Myr-GTASSL) was dissolved in 100% DMSO to make 50 mM stock solution. 500 M solution of Unc119a(58 -240) was mixed with N-terminal myristoylated NPHP3 peptide at a 1:1 molar ratio in a buffer containing 25 mM Tris-HCl (pH 7.5), 150 mM NaCl and 3 mM DTE.
In situ proteolysis was applied prior to the screening at 20°C by the addition of both proteases, trypsin and chymotrypsin (at a 1:1000 (w/w) ratio each), as described before (56). The crystals appeared in several conditions containing ammonium sulfate. The final crystallization condition that was optimized was 1.75 M (NH 4 ) 2 SO 4 , 0.1 M CAPS (pH 10.0), and 0.2 M Li 2 SO 4 . Cryoprotectant solution containing the mother liquor in addition to 20% (v/v) glycerol was used for flash freezing the crystals. X-ray diffraction data were collected at the X10SA beamline of the Suisse Light Source, Villigen. Data were processed by the XDS program. Molecular replacement was carried out using Molrep from the CCP4 suite and Unc119a from the Unc119a-N-terminal lauroylated transducin-␣-mimicking peptide complex (PDB code 3RBQ) used as a search model. The model was further built by WinCoot, and the refinement was done with REF-MAC5. Refinement and data collection statistics are summarized in Table 3.
Cell Culture and Generation of Stable Cell Lines-Mouse renal epithelial cells from the inner medullary collecting duct (IMCD3) were cultured at 37°C and 5% CO 2 in DMEM/F-12, HEPES complemented with 10% fetal bovine serum and 2 mM L-Glutamine (Life Technologies). The genome of the IMCD3 cells contained a stably integrated FRT cassette (IMCD3 Flp-In, a kind gift from M. V. Nachury; Flp-In cell line technology by Life Technologies), which enabled the generation of stable cell lines as described previously (55,57). Briefly, IMCD3 Flp-In cells were seeded in 6-well plates at a density of 100,000 cells/ well. On the next day, the cells reached a confluence of 40 -60% and were cotransfected with the following two vectors: (i) the modified pG-LAP5-mCherry vector containing the gene of interest, a flippase recognition target site, and a hygromycin resistance gene and, (ii) the pOG44 vector (Life Technologies) encoding the FLP recombinase. Transfection was accomplished using Lipofectamine 2000 (Life Technologies). Beginning 2 days after transfection, cells were selected with 100 -200 mg/ml hygromycin in the complemented culture medium. Expression of the respective mCherry-tagged proteins was verified by immunoblotting with an antibody against mCherry (1:2000; MPI-CBG antibody facility).
Immunostaining and Microscopy-IMCD3 Flp-In cells stably expressing the respective mCherry-tagged protein were plated on poly-L-lysine-coated coverslips in 6-well plates. Each well contained 100,000 cells. On the following day, cilia were induced by serum starvation for 48 h. After washing in PBS, cells were fixed with 4% formaldehyde in cytoskeletal buffer (2.75 M NaCl, 100 mM KCl, 25 mM Na 2 HPO 4 , 8 mM KH 2 PO 4 , 40 mM MgCl 2 , 40 mM EGTA, 100 mM PIPES, 100 mM glucose, pH 6.0) for 20 min. Cells were washed twice in PBS and permeabilized with 0.3% Triton X-100 in cytoskeletal buffer for 10 min. After rinsing in 0.1% Tween 20 in PBS, cells were incubated in 10% FBS in PBS for 30 min for blocking. To immunostain primary cilia, mouse 6-11B-1 anti-acetylated tubulin antibody (1:5000; Sigma T6793) in 10% FBS in PBS was incubated overnight at 4°C. After washing four times with 0.1% Tween 20 in PBS, Alexa Fluor 488 anti-mouse secondary antibody (1:800; Life Technologies A-11001) was added for 45 min at room temperature. Coverslips were rinsed three times in 0.1% Tween 20 in PBS and once in PBS. Nuclei were stained with DAPI (Serva) for 1 min, diluted 1:10,000 in PBS. Cells were washed three times in PBS, and the coverslips were fixed on glass slides using Mowiol (Merck). Images were taken using an Olympus IX81 microscope with a CCD camera and a 60ϫ oil immersion objective with a numerical aperture of 1.35 using identical settings for each image to ensure comparability.