The Mechanism of Dynein Light Chain LC8-mediated Oligomerization of the Ana2 Centriole Duplication Factor*

Background: Ana2 is a conserved centriole duplication factor involved in nascent centriole biogenesis. Results: Two sites in the central domain of Ana2 (Ana2M) bind LC8 and form an Ana2M4-LC88 complex. Conclusion: LC8 potentiates Ana2 tetramerization. Significance: LC8-potentiated Ana2 tetramerization is expected to increase the avidity of Ana2 for centriole factors, including Sas-6, and may drive binding factor oligomerization. Centrioles play a key role in nucleating polarized microtubule networks. In actively dividing cells, centrioles establish the bipolar mitotic spindle and are essential for genomic stability. Drosophila anastral spindle-2 (Ana2) is a conserved centriole duplication factor. Although recent work has demonstrated that an Ana2-dynein light chain (LC8) centriolar complex is critical for proper spindle positioning in neuroblasts, how Ana2 and LC8 interact is yet to be established. Here we examine the Ana2-LC8 interaction and map two LC8-binding sites within the central region of Ana2, Ana2M (residues 156–251). Ana2 LC8-binding site 1 contains a signature TQT motif and robustly binds LC8 (KD of 1.1 μm), whereas site 2 contains a TQC motif and binds LC8 with lower affinity (KD of 13 μm). Both LC8-binding sites flank a predicted ∼34-residue α-helix. We present two independent atomic structures of LC8 dimers in complex with Ana2 LC8-binding site 1 and site 2 peptides. The Ana2 peptides form β-strands that extend a central composite LC8 β-sandwich. LC8 recognizes the signature TQT motif in the first LC8 binding site of Ana2, forming extensive van der Waals contacts and hydrogen bonding with the peptide, whereas the Ana2 site 2 TQC motif forms a uniquely extended β-strand, not observed in other dynein light chain-target complexes. Size exclusion chromatography coupled with multiangle static light scattering demonstrates that LC8 dimers bind Ana2M sites and induce Ana2 tetramerization, yielding an Ana2M4-LC88 complex. LC8-mediated Ana2 oligomerization probably enhances Ana2 avidity for centriole-binding factors and may bridge multiple factors as required during spindle positioning and centriole biogenesis.

Centrioles are cylindrical cellular structures that form the core of centrosomes and basal bodies, organelles responsible for nucleating polarized microtubule networks in the cytoplasm and cilia, respectively. A cell's centriole count largely determines its capabilities, because single centrioles form the base of sensory cilia, whereas multiple centrioles are needed to nucleate motile cilia (1). A centriole pair constitutes the core of the centrosome, needed for bipolar mitotic spindle formation. Centriole structure is largely conserved across metazoans, protists, and some plants, with a characteristic 9-fold radial symmetry established by an inner, 9-spoked cartwheel structure (2). Although different species have different microtubule arrangements in the surrounding blades (singlets, doublets, or triplets, as shown in Fig. 1A) as well as different cartwheel architectures (3), the critical centriole duplication components are conserved. The inner cartwheel recruits centriolar proteins and pericentriolar matrix components to build and elongate the outer centriole wall. During elongation, nine sets of microtubule blades (each a microtubule triplet in the case of Drosophila) form around the centriole perimeter parallel to the longitudinal axis, propagating the organelle's 9-fold radial symmetry.
Canonical centriole duplication is coupled to the cell cycle to limit centriole number (4). A subset of conserved centriole proteins are involved in centriole duplication (5), because their misregulation leads to increased or decreased centriole counts (Fig. 1B). Three key initiation factors include Polo-like kinase 4 (Plk4), spindle assembly abnormal protein 6 (Sas-6), and anastral spindle 2 (Ana2/STIL/Sas-5, found in Drosophila melanogaster, humans, and Caenorhabditis elegans, respectively). Plk4 is a serine/threonine kinase whose catalytic activity is required for centriole duplication. Plk4 phosphorylates a set of both known and unknown components to transmit the centriole duplication signal (6,7). Plk4 is recruited to the centriole through an interaction with Asterless (Asterless/Cep-152 found in D. melanogaster and humans, respectively), but the conserved definitive target of the Plk4 kinase activity remains unknown (7). Downstream of Plk4, nascent centriole construction involves Sas-6 oligomerization to form the inner, 9-fold symmetric cartwheel (8 -10). A third and less studied centriole-initiating factor is the Sas-6-binding protein, Ana2, whose role in centriole duplication is unclear.
Ana2 was identified in a genome-wide screen in which Ana2 depletion caused a decrease in centriole count (11). Ana2 is functionally conserved across metazoan species, with orthologs in humans (STIL), Danio rerio (STIL), and C. elegans (SAS-5) (12). However, the Ana2 sequence has diverged among species, with similarity restricted to an N-terminal Sas-4 binding site (13,14), a central predicted coiled-coil, and a C-terminal STAN (STil/ANa2) domain that binds Sas-6 in vitro (12,15,16) (Fig.  1C; see domain conservation presented in the inset, scored using percentage identity and percentage similarity between species). In Drosophila oocytes, Sas-6 overexpression results in centriole amplification only when Ana2 is dually overexpressed (17). In human systems, expression of Ana2 is essential in maintaining centriole count (18). Furthermore, mutations in Ana2 have been linked to primary microcephaly, leukemia, and cancer (19 -23). How Ana2 and Sas-6 synergistically function remains to be determined. Although the function of Ana2 is poorly understood, recent work has demonstrated that Drosophila Ana2 interacts with the dynein light chain, LC8 (cut up (Ctp)) (24), a ubiquitous protein that binds diverse targets throughout the cell to confer or potentiate target dimerization (reviewed in Ref. 25). The Ana2-LC8 interaction is important for directing spindle orientation during Drosophila larval brain development (Fig. 1D). Loss of either Ana2 or LC8 results in aberrant spindle positioning and defective separation of apicobasal polarity determinants during neuroblast asymmetric cell division.
Although LC8 acts as a processivity factor for the dynein motor by enhancing motor dimerization, it largely plays a dynein motor-independent role throughout the cell to potentiate dimerization of its binding partners (26 -37). It was shown in a yeast two-hybrid screen that LC8 binds two Ana2 fragments: the first fragment spanning residues 1-200 and the second spanning residues 201-274, which includes a predicted ␣-helix highly conserved across fly species (Fig. 1, E and F) (24,12). To date, there is no structural insight into the LC8-Ana2 complex.
Here, we use x-ray crystallography, isothermal microtitration calorimetry (ITC), 3 and size exclusion chromatography with multiangle static light scattering (SEC-MALS) to characterize the interactions between LC8 and Ana2. Our results demonstrate that LC8 dimers bind Ana2 at two distinct sites, the first of which contains a high-affinity, canonical LC8-binding TQT motif (residues 159 -168), whereas the second contains a non-canonical TQC motif (residues 237-246). We present the structures of LC8 bound to peptides encompassing both of the LC8 binding sites of Ana2 as well as the apo-LC8 dimer and highlight the conserved Ana2 features that underlie these different interactions with the peptides. SEC-MALS analysis of WT and mutant Ana2M (residues 156 -251) in complex with LC8 reveals LC8-dependent Ana2M tetramerization in an Ana2M 4 -LC8 8 complex. The Ana2 LC8 binding sites flank a predicted ␣-helix probably involved in Ana2 oligomerization. Our findings suggest that LC8 is responsible for enhancing Ana2 oligomerization and structural stability. LC8-potentiated Ana2 oligomerization has spatial and avidity implications for the Ana2 N-terminal Sas-4 binding motif and its C-terminal Sas-6-binding STAN domain.
Protein Purification for Crystallization-LC8 was purified as described previously for the yeast homologue Dyn2 (29). Briefly, cells expressing GST-LC8 were lysed by sonication and clarified by centrifugation at 23,000 ϫ g for 45 min, and the supernatant was loaded onto a Glutathione Sepharose column (GE Healthcare). The column was washed with buffer A, and the GST-LC8 fusion was eluted in buffer A supplemented with 25 mM glutathione. The GST tag was cleaved with PreScission protease (GE Healthcare). LC8 was subsequently purified on an SP Sepharose Fast Flow column (GE Healthcare) and exchanged into MES storage buffer (25 mM MES, pH 6.0, 50 mM sodium chloride, and 0.1% ␤-mercaptoethanol). LC8 was concentrated to 0.5 mM, snap frozen in liquid nitrogen, and stored at Ϫ80°C. The final LC8 contains an N-terminal five-residue (GPLGS) cloning artifact.
Crystallization of the LC8-Ana2 Peptide Complexes-Final concentrations of 0.5 mM LC8 and 0.6 mM Ana2 pep1 (or 0.75 mM LC8 and 0.9 mM Ana2 pep2) in MES storage buffer were incubated for 30 min on ice. For the LC8-pep1 complex, crystallization followed the hanging drop protocol using 2 l of the LC8/Ana2 pep1 mixture and 2 l of a 1-ml well solution that contained 0.3 M magnesium acetate, 0.1 M sodium cacodylate, pH 6.5, and 26% (w/v) polyethylene glycol 8000. The same method was used for LC8-pep2 in a well solution containing 0.19 M ammonium acetate, 27% (w/v) polyethylene glycol 4000, 0.1% ␤-mercaptoethanol, and 0.1 M sodium acetate, pH 4.6. For both structures, crystals grew at 20°C into rods (pep1) or rounded cubes (pep2) within 3 days and remained at full size for up to 3 weeks. Crystals were transferred into fomblin oil (Sigma) cryoprotectant and flash frozen in liquid nitrogen.
Data Collection, Structure Determination, and Refinement-Diffraction data were collected on LC8-Ana2 crystals (both peptides) at the Advanced Photon Source SER-CAT beamline 22-ID with 1°oscillations over 180°from single crystals. Data were indexed, integrated, and scaled using HKL2000 (38). The LC8-Ana2 peptide structures were determined using the AutoMR molecular replacement program (PHENIX crystallographic suite (39)) and a modified 2PG1 (36) coordinate file in which a monomeric (for LC8-Ana2 pep1) or dimeric (for LC8-Ana2 pep2) apo-Drosophila LC8 search model was used. The models were built using AutoBuild (PHENIX) and refined iteratively through manual builds in Coot (40), followed by refinement runs using phenix.refine against a maximum likelihood target (PHENIX) (39). Refinement statistics were monitored using a free R, calculated using 5.4 or 5.6% of the data for pep1 and pep2, respectively, randomly excluded from refinement (41).
Isothermal Microtitration Calorimetry-ITC experiments were carried out at 26°C in MES storage buffer on a MicroCal AutoITC200 (GE Healthcare). Lyophilized peptides were solubilized in MES storage buffer. 19 2-l injections of 1.0 mM Ana2 pep1 were automatically injected into 200 l of 50 M LC8, and 2.0 mM pep2 was automatically injected into 200 l of 100 M LC8. The resulting binding isotherms were analyzed using the Origin version 7.0 software package (OriginLab) and were fit to a single-site, independent binding model. Ana2 peptide control experiments were performed to determine the contribution from each peptide's heat of dilution. These controls involved 19 2-l injections of 1.0 mM Ana2 pep1 or 2.0 mM Ana2 pep2 into a chamber containing 200 l of MES storage buffer. The Ana2 pep1 control isotherm did not reveal significant heat of dilution; therefore, the final five injection values (where binding was saturated in the pep1-LC8 isotherm) were averaged, and this value was subtracted from each injection in the pep1-LC8 experiment. The Ana2 pep2 control isotherm revealed a significant endothermic heat of dilution (data not shown); therefore, these control values were individually subtracted from the corresponding raw experimental values from the pep2-LC8 binding isotherm. Experiments were conducted in triplicate, the internal or external controls were subtracted, and the resulting heats of dilution were averaged to determine respective mean K D and S.D. values.
Ana2M-LC8 Complex Purification for SEC-MALS-The composite pellet of His 6 -LC8 and GST-Ana2M was thawed and lysed by sonication with the addition of phenylmethanesulfonyl fluoride to a final concentration of 200 M. The supernatant was purified over Ni 2ϩ -nitrilotriacetic acid resin (Qiagen) followed by PreScission protease (GE Healthcare) treatment to cleave off the His 6 and GST tags. The LC8-Ana2M complex was subsequently purified over a Superdex 200 size exclusion column (GE Healthcare) and concentrated in SEC-MALS buffer: 25 mM HEPES, pH 7.5, 300 mM sodium chloride, 0.1% ␤-mercaptoethanol. The presence of both components was confirmed by SDS-PAGE. Expression and purification of the Ana2M-SNAP-LC8 complex followed a similar protocol.
Mutagenesis of Ana2M-An Ana2M site 1 LC8-binding mutant (Q165A/T166A) was created using the QuikChange (Agilent Technologies) method on the wild-type GST-Ana2M construct according to the manufacturer's instructions. The mutant GST fusion protein was expressed and co-purified with wild-type LC8 as described above.
SEC-MALS-LC8-Ana2M complexes (100 l) were injected onto a Wyatt WTC-030S5 silicone size exclusion column (for elution of 5-1,250-kDa proteins) in SEC-MALS buffer supplemented with 0.2 g/liter sodium azide and passed in tandem through a Wyatt DAWN HELEOS II light scattering instrument and a Wyatt Optilab rEX refractometer. The light scattering and refractive index data were used to calculate the weight-averaged molar mass and the mass fraction in each peak using the Wyatt Astra V software program (Wyatt Technology Corp.) (42).

Ana2 Contains Two High-affinity LC8 Binding Sites-Drosophila
Ana2 is a 420-residue centriole duplication component that lacks apparent conservation across species barring an N-terminal Sas-4 binding region (13,14), a central predicted coiled-coil domain, and the highly conserved C-terminal STAN (STil/ANa2) motif ( Fig. 1, C, E, and F) (12). Previous studies demonstrated a physical interaction between the N-terminal 274 residues of Ana2 and LC8, a dynein light chain, via yeast two-hybrid screening (24). Structure function analysis indicated that Ana2 contained at least two LC8 binding sites, one within the region spanning residues 1-200 and the second within the region spanning residues 201-274. LC8 binds many subcellular targets across species in a cytoplasmic dynein motor-independent mechanism to promote target dimerization (43), suggesting that LC8 may potentiate Ana2 oligomerization. To map the interactions between Ana2 and LC8, we scanned Ana2(1-274) for potential LC8 binding sites. LC8 target motifs comprise up to 11 contiguous residues, which, although diverse in sequence composition, often contain a K Ϫ3 X Ϫ2 T Ϫ1 Q 0 T 1 or G Ϫ2 I Ϫ1 Q 0 V 1 D 2 motif with the conserved glutamine (Gln 0 ) set as the zero reference point (44). Target peptides with these LC8-binding motifs bind LC8 with K D values in the 0.1-100 M range (44,45). We identified two poten-tial binding sites within Ana2, corresponding to residues 159 -168 (containing a T Ϫ1 Q 0 T 1 sequence) and 237-246 (containing a T Ϫ1 Q 0 C 1 sequence). These two sites flank either end of the conserved predicted coiled-coil (Fig. 1, E and F) and correlate with the two fragments identified via yeast two-hybrid as LC8binding segments.

Structural Basis of LC8-mediated Ana2 Oligomerization
To determine whether these sites were conserved, we aligned several Ana2 sequences from 10 different Drosophila species. Much of the protein is conserved within the genus, with the largest concentration of identity mapping to the STAN motif and an N-terminal region with no predicted secondary structure but involved in Sas-4 binding (Fig. 1E) (13,14). Additional identity maps to the central predicted coiled-coil and the flanking regions that contain the tentative LC8-binding sites that we identified (Fig. 1F). The linkers that bridge the predicted LC8 binding sites with the central, predicted coiled-coil show diversity in both sequence length and composition. When we analyzed the central, predicted coiled-coil domains in Ana2 orthologs (human STIL, zebrafish STIL, and C. elegans Sas-5), only Sas-5 contained a potential QT motif N-terminal to the predicted coiled-coil with the sequence KTVNVSQTVE, sug-gesting that the LC8-Ana2 interaction may be specific to a subset of Ana2 orthologs.
To confirm the ability of the putative LC8-binding sites of Ana2 to bind LC8, we synthesized peptides corresponding to the two predicted Ana2-LC8 binding sites (Fig. 1A) and performed ITC, monitoring the heat released as each peptide was titrated into the calorimeter cell containing purified LC8. Experiments were performed in triplicate, with reported values reflecting the average of all trials. The Ana2 peptide 1 (pep1)-LC8 binding isotherm was exothermic and yielded a K D value of 1.14 Ϯ 0.07 M ( Fig. 2A). Compared with reported LC8-target affinities (100 -0.1 M) (44,45), Ana2 pep1 binds LC8 in the higher-affinity range. The Ana2 pep2 binding isotherm was also exothermic and yielded an experimentally determined LC8 K D value of 12.8 Ϯ 1.5 M (Fig. 2B), a weaker binding affinity than   (50,51), where it phosphorylates both a known and unknown set of substrates in the centriole duplication pathway (7,52). Sas-6 (spindle assembly abnormal-6) oligomerizes to form the first structure observed using electron microscopy; this nine-spoked cartwheel is depicted on the left. In cells, Sas-6 oligomerization is Ana2 (anastral spindle-2)-dependent (9,12,17). Sas-4 (spindle assembly abnormal-4) is thought to recruit triplet microtubule blades and stabilize centriole elongation and maturation (mature centriole shown at the left) (5,13,14). C, comparison of Homo sapiens, D. rerio (D.r.), D. melanogaster, and C. elegans Ana2 orthologs reveals diversity in protein structure. Although the lengths of Ana2 orthologs differ, the presence of a Sas-4 binding domain (red), an Sas-6 binding domain (STAN domain; gray), and a predicted central coiled-coil region remain constant (domains shown as determined previously (12,14)). Inset, alignments of individual Sas-4 binding domains and STAN domains between D. melanogaster and H. sapiens, D. rerio, or C. elegans reveal high percentages of invariant (first value) and similar (second value) residues. D, Ana2 and LC8 form a complex with Mud to orient the mitotic spindle during asymmetric divisions in the developing Drosophila neuroblast (24). Asymmetry is achieved, in part, via differential maturation of the centrosomes. The daughter centrosome forms the LC8-Ana2-Mud complex that coordinates spindle alignment with cortical polarity cues to maintain a stem population (ganglion mother cell (GMC)) (53,24). E, full-length Drosophila Ana2 has an N-terminal Sas-4 binding region (13,14) and a C-terminal STAN motif (12) conserved across functional Ana2 orthologs. The central predicted helical domain is flanked by two LC8 binding sites (site 1, residues 159 -168; site 2, residues 237-246). Residue identity across Drosophila species is noted below in green. F, conservation within the Ana2 central helical domain and LC8 binding sites. Residues with 100% identity are highlighted in green, whereas those with 80% similarity are highlighted in yellow. Note that both the TQT (positions 165-166) and TQC (positions 243-244) sites are conserved within the genus.

Structural Basis of LC8-mediated Ana2 Oligomerization
pep1 but within the commonly reported range of LC8-target affinities.
Crystallization of LC8 Ana2 Pep1 and Pep2 Complexes-To determine the molecular determinants underlying the LC8-Ana2 interaction, we attempted to crystallize LC8 in complex with each synthesized Ana2 peptide. Both LC8-peptide complexes were amenable to crystallization, although diffraction quality crystals formed in different conditions (see "Experimental Procedures"). Ana2 pep1-LC8 crystals diffracted to 1.83 Å resolution and belonged to the space group P2 1 2 1 2 1 (Table 1). Ana2 pep2-LC8 crystals diffracted to 1.9 Å resolution and belonged to the space group P1 (Table 1). To solve both structures, we performed molecular replacement using a search model containing a single Drosophila LC8 chain (for Ana2 pep1) or an LC8 dimer (for Ana2 pep2) without bound peptide, derived from Protein Data Bank entry 2PG1 (36).
The Structure of LC8 Bound to Ana2 Pep1-Four LC8 chains were found in the asymmetric unit. The LC8 chains are paired to form two independent homodimers, each arranged around non-crystallographic 2-fold axes. Clear electron density was evident in the initial F o Ϫ F c map to build four Ana2 pep1 chains (Fig. 3A), two bound to each LC8 homodimer. The structure was built and refined to R and R free values of 17.6 and 20.7%, respectively (see Table 1 for refinement statistics).
The LC8 homodimer forms a composite platform for Ana2 pep1 binding (Fig. 3B). The homodimeric core is characterized by a central 12-stranded ␤-sandwich, each half of which is formed by four ␤-strands from one LC8 chain (␤1 from Val 7 -Asp 12 , ␤4 from His 72 -Leu 78 , ␤5 from Val 81 -Lys 87 , ␤2 from Trp 54 -Gly 59 ), one ␤-strand from the LC8 homodimeric mate (␤3Ј from Gly 63 Ј-Glu 69 Ј), and the Ana2 peptide, which contributes the sixth and final ␤-strand (Fig. 3B). Each ␤-sheet is entirely antiparallel. The ␤-sandwich is flanked on either side by two ␣-helices. Ana2 pep1 binding engages determinants in both LC8 chains, with ␤3 forming a key extended interface with the peptide (Fig. 3B). Peptide binding is stabilized by backbone/ backbone antiparallel ␤-sheet hydrogen bonding (Fig. 4A) as well as several side chain interactions. The Ana2 conserved glutamine Gln 165 (notated as Q 0 in reference to its position in the canonical K Ϫ3 X Ϫ2 T Ϫ1 Q 0 T 1 binding motif) forms key contacts, including van der Waals interactions with both LC8 chains and hydrogen bonds with the Glu 35 Ј side chain carboxylate group and the Lys 36 Ј backbone amide, serving to cap the ␣2Ј helix's N-terminal region (Fig. 4B).
The Structure of LC8 Bound to Ana2 Pep2-The Ana2 pep2-LC8 crystal contains three LC8 dimers in the P1 unit cell. One LC8 dimer is bound to two Ana2 pep2 chains (Fig. 3, A and C), whereas the other two LC8 dimers are in the apo form with crystal packing sterically occluding the peptide binding sites. The structure was built and refined to R and R free values of 18.5 and 23.7%, respectively (see Table 1 for refinement statistics).
Ana2 pep2 binds in a manner similar to Ana2 pep1, extending either side of the LC8 core ␤-sandwich and making several backbone interactions with LC8 ␤3 (Figs. 3C and 4A). Gln 0 of Ana2 pep2 participates in similar interactions as observed in the LC8-Ana2 pep1 structure; however, pep2 contains a noncanonical cysteine residue at the ϩ1-position, Cys 244 . To our knowledge, this is the first example of an LC8 target with a cysteine in the ϩ1-position. In contrast to the canonical threonine at the ϩ1-position, Ana2 pep2 Cys 244 is angled into the

is the ith measurement and ͗I(h)͘ is the mean of all measurements of I(h) for Miller indices h.
LC8 peptide binding groove, with its side chain engaging LC8 Glu 35 Ј, Arg 60 , Asn 61 , Phe 62 , Tyr 77 , and Ala 82 . Specifically, the cysteine's terminal sulfhydryl group forms a 3.6-Å electrostatic interaction with the backbone carbonyl of LC8 Arg 60 (Fig. 4C). This shift allows for extended backbone-backbone contacts, including interactions between Ana2 Cys 244 and LC8 Phe 62 , as well as between Ana2 Ile 246 and LC8 Arg 60 (Fig. 4D). As a result, the Ana2 pep2 C-terminal region differentially engages the LC8 dimer as compared with Ana2 pep1, whose respective determinants are positioned 3-5 Å away (Fig. 4D).

The LC8 Binding Pocket Undergoes Structural Shifts to Accommodate Ana2
Peptides-In addition to observing an LC8-Ana2 pep2 complex in the P1 unit cell, two sets of apo-LC8 homodimers were also present. As observed previously (45), the apo-LC8 binding pocket is narrower than the peptide-bound cleft observed in both Ana2-bound LC8 structures (Fig. 5A). Several LC8 residues that directly engage the Ana2 peptides are swung toward the peptide binding pocket in the apo state, including Asn 10 , Lys 36 Ј, Tyr 65 , Thr 67 , Phe 73 , Tyr 75 , and Tyr 77 , highlighting the mobility of LC8 side chains upon target binding. . Structures of LC8-Ana2 complexes reveal LC8 homodimers bound to two parallel Ana2 peptides. A, LC8-Ana2 complex structures were determined using peptide-free LC8 search models. Initial F o Ϫ F c electron density for the Ana2 peptides is shown in green and contoured at 2.0 (pep1) and 1.65 (pep2). Final 2F o Ϫ F c electron density is shown below in gray with the final Ana2 pep1 and pep2 model included; electron density is contoured at 2.0 (pep1) and 1.0 (pep2). Final models of the respective LC8-Ana2 peptide complexes are presented in the top left (LC8-Ana2 pep1) and top right (LC8-Ana2 pep2) with peptides in the same orientation for reference. B and C, the final structures of LC8 bound to Ana2 pep1 (orange; B) and Ana2 pep2 (cyan; C) are shown looking down the complex's 2-fold axis (left) and after a 90°rotation about the y axis (right). The center schematic in B and C summarizes the secondary structure elements that comprise a single ␤-sheet in the LC8-peptide complexes. Each ␤-sheet is extended by the third ␤-strand contributed by the LC8 homodimeric mate (purple) as well as the bound Ana2 peptide (pep1 shown in orange (B); pep2 shown in cyan (D)). The final ␤-sheet comprises a total of six strands and is flanked by two ␣-helices (shown in mint, behind the sheet). JULY 25, 2014 • VOLUME 289 • NUMBER 30

JOURNAL OF BIOLOGICAL CHEMISTRY 20733
Ana2 Employs a Unique Tandem Set of LC8 Binding Motifs-LC8 targets vary widely in their binding affinity and motif composition, both within and beyond the canonical K Ϫ3 X Ϫ2 T Ϫ1 Q 0 T 1 or G Ϫ2 I Ϫ1 Q 0 V 1 D 2 sequence motifs. Interestingly, both Ana2 LC8 binding motifs combine features from each canonical motif (pep1, A Ϫ3 G Ϫ2 T Ϫ1 Q 0 T 1 D 2 ; pep2, T Ϫ3 G Ϫ2 T Ϫ1 Q 0 C 1 D 2 ). Both Ana2 LC8 sites have threonine residues at the Ϫ1-position, as found in the K Ϫ3 X Ϫ2 T Ϫ1 Q 0 T 1 motif, and both have glycine and aspartate residues at the Ϫ2and ϩ2-positions, as found in the G Ϫ2 I Ϫ1 Q 0 V 1 D 2 motif. Neither Ana2 site employs a basic residue at the Ϫ3-position, which is often seen in high-affinity LC8 interactors (46), including Nek9 (a kinase that regulates mitotic spindle formation and chromosome separation) (47) and the dynein intermediate chain (DIC; a dynein motor complex component used in cargo recognition) (36) (Fig. 5, A-C). Both Nek9 and DIC LC8 target sites contain a lysine at the Ϫ3-position (K Ϫ3 ) that interacts with the LC8 D12Ј side chain carboxyl group and promotes relatively strong LC8-binding interactions (K D values on the order of 0.1-0.2 M) (36, 47) (Fig. 5, B and C). Ana2 sites 1 and 2 contain an alanine and threonine, respectively, at position Ϫ3 that do not Interactions are presented where atoms are less than or equal to 3.5 Å apart (hydrogen bonds and electrostatic interactions; shown in red for pep1 and pink for pep2) and 4.5 Å apart (van der Waals contacts; shown in dark gray for pep1 and light gray for pep2). Boxes completely filled in reflect similar LC8 interaction modes with each peptide, whereas those boxes that are half-filled indicate unique, peptide-specific interactions. B, conserved Gln 165 of Ana2 pep1 forms hydrogen bonds to LC8Ј residues Glu 35 Ј and Lys 36 Ј. C, Cys 244 of Ana2 pep2 forms an electrostatic interaction with LC8 residue Arg 60 . D, the Ana2 pep2 (cyan) C-terminal region forms extensive backbone hydrogen bonds with LC8 and is positioned differently than Ana2 pep1 (orange), which has been overlaid on the LC8-Ana2 pep2 structure for comparative purposes. In contrast to the Ana2 pep2 Cys 244 backbone carbonyl and the Ile 246 backbone amide that interact with LC8 Phe 62 and Arg 60 , respectively, the comparable Ana2 pep1 determinants (indicated with magenta arrows) are splayed and rotated away from LC8. JULY 25, 2014 • VOLUME 289 • NUMBER 30 engage the D12Ј side chain carboxyl (Fig. 5A). At the Ϫ3-position, Pak1 (a kinase that regulates cell motility and, together with LC8, plays a role in cancer transformation (48)) is an interesting point of comparison. Pak1 contains the non-canonical LC8 binding sequence V Ϫ3 A Ϫ2 T Ϫ1 S 0 P 1 I 2 and has the weakest affinity for LC8 (K D of 42 M) of the peptides we used for comparison. Like both Ana2 peptides, Pak1 employs a non-charged residue at the Ϫ3-position, but similar to the Ϫ3 lysine in Nek9 and DIC, positionally equivalent aliphatic side chain determinants are used to engage LC8, highlighting the ability of LC8 to accept side chain variability at the Ϫ3-position.

Structural Basis of LC8-mediated Ana2 Oligomerization
We next examined how the conformations of Ana2 pep1 and pep2 compared with other LC8 binding peptides by aligning LC8-peptide complex structures (Fig. 5, D-F). Ana2 pep1 aligns well with other LC8 binding peptides, including Nek9, DIC, and Pak1 (Fig. 5E). However, the C-terminal region of Ana2 pep2 departs from this common LC8-bound architecture. The aforementioned Ana2 pep2 cysteine, Cys 244 , at position ϩ1 is angled into the LC8 peptide-binding groove, effectively positioning the peptide's C-terminal region closer to the LC8 homodimer. In contrast, Ana2 pep1, Nek9, DIC, and Pak1, each of which has a threonine at position ϩ1, splay away from the LC8 dimer, with their C-terminal regions positioned ϳ3-5 Å from the comparative location of Ana2 pep2 (Fig. 5E). Underlying the differential position of Ana2 pep2 is comparative placement of the Cys 244 backbone carbonyl in the same loca-tion of the threonine (Thr 1 ) side chain hydroxyl as found in Ana2 pep1, Nek9, and DIC (Fig. 5F).
LC8 Mediates the Solubility and Oligomerization State of Ana2M-Multiple attempts to express various Ana2 constructs containing either or both of the LC8 binding sites yielded insoluble protein, making it difficult to study the oligomeric state of Ana2 in the absence of LC8. However, co-purification of Ana2M (residues 156 -251, encompassing both LC8 binding sites and a central predicted helical domain; Fig. 1B) and LC8 yielded a stable, soluble complex that could be purified via the His 6 affinity tag of LC8 followed by size exclusion chromatography. The canonical role of LC8 as a dimerization "hub" led us to predict that the purified LC8-Ana2M complex would form a heterohexamer, with two Ana2M chains forming a central coiled-coil flanked at either end by LC8 homodimers. To experimentally determine the Ana2M-LC8 complex's mass and stoichiometry, we analyzed the complex using SEC-MALS.
As reported previously (29), purified LC8 eluted primarily as a dimer (Fig. 6A, light green trace indicating the Rayleigh ratio) with a mass of 21.6 kDa (Fig. 6A, dark green trace indicating the molecular weight). Surprisingly, Ana2M-LC8 formed a stable complex with a mass of 117.1 Ϯ 5.9 kDa (average of four experiments from two independent protein purifications; Fig. 6A, red traces). This is approximately twice the mass an Ana2M 2 -LC8 4 heterohexamer would form (68 kDa). Adding excess purified LC8 to the Ana2M-LC8 complex did not shift or increase the  Purified Ana2M (residues 156 -251, 11 kDa) remained soluble only when co-purified with excess LC8 and behaved as a single species throughout the purification, which included affinity tag chromatography followed by two sizing columns. A, detection of the LC8-Ana2M complex on a sizing column coupled with multiangle static light scattering shows a single peak (pink trace; Rayleigh ratio) at 117 Ϯ 5.9 kDa (red; molecular weight measurement). The same experiments with a Q 0 T 1 to A 0 A 1 mutation show a single peak (light blue trace; Rayleigh ratio) at 84.8 Ϯ 2.5 kDa (dark blue trace; molecular weight measurement). LC8 alone elutes as a dimer with a mass of 21.6 kDa (dark green trace; molecular weight measurement). B, SNAP-LC8 elutes as a single species (light green trace; Rayleigh ratio) at 59.3 kDa (dark green traces), corresponding to a dimer. Co-purification of SNAP-LC8 with Ana2M yielded a complex that eluted from the size exclusion column in a broad peak, with a shoulder characteristic of complex dissociation (light purple trace, Rayleigh ration). Experimentally determined molecular weight across the broad peak indicated complexes of varying size, ranging from 290 to 150 kDa (dark purple traces; different parts of the peak were integrated to determine the contributing sizes). All experiments are consistent with the formation of a stable LC8 8 -Ana2M 4 complex. mass of the eluted complex but yielded a second peak that eluted later with an experimentally determined mass of 21 kDa, correlating with excess LC8 homodimers (data not shown). These results led us to postulate the existence of an LC8-mediated Ana2M tetramer comprising four Ana2M molecules and four LC8 homodimers (Ana2M 4 -LC8 8 ; Fig. 7).
To test our predicted stoichiometry, we reasoned that mutating the first LC8 binding site of Ana2 to compromise LC8 binding would result in an Ana2M 4 -LC8 4 complex with a corresponding mass of 90.8 kDa. We mutated the first LC8 binding site of Ana2M (Q165A and T166A or "QT/AA") and co-purified the Ana2M QT/AA -LC8 complex, noting that the second LC8-binding site was sufficient for LC8-mediated Ana2M solubilization. The Ana2M QT/AA -LC8 mutant retained its solubility and showed an excess of unbound LC8 during size exclusion chromatography. In accordance with forming an Ana2M 4 -LC8 4 heterooctomer (predicted mass of 90.8 kDa), the complex had an apparent mass of 84.8 Ϯ 2.5 kDa (Fig. 6A, blue traces).
In our experiments, Ana2M and LC8 were approximately the same mass (11.4 and 11.3 kDa, respectively), making it difficult to determine the relative contribution of LC8 and Ana2M to the complex. To independently confirm the composition of the proposed Ana2M 4 -LC8 8 complex, we purified an N-terminally SNAP-tagged LC8 (monomer mass, 30 kDa) alone and in complex with Ana2M. The SNAP tag served the sole purpose of increasing the mass of the LC8 construct to see how this change in turn altered the mass of the Ana2M-LC8 complex. An Ana2M 4 -SNAP-LC8 8 complex would have a mass of 286 kDa. SNAP-LC8 was expressed with an N-terminal His tag and behaved similar to wild-type LC8 throughout purification. Purified SNAP-LC8 eluted from the SEC-MALS column as a homodimer with an apparent mass of 59.3 kDa, indicative that the SNAP tag does not interfere with LC8 dimerization (Fig. 6B, dark green trace). Ana2M co-purified with SNAP-LC8, suggesting that SNAP-LC8 retained target-binding capabilities. The Ana2M-SNAP-LC8 complex eluted broadly from the SEC-MALS column with experimentally determined masses ranging from 290 kDa (early portion of the elution peak) to 150 kDa (later portion of the elution peak) (Fig. 6B, purple trace). The early portion of the elution peak mass correlates with an Ana2M 4 -SNAP-LC8 8 complex, whereas the 150 kDa shoulder suggested that the SNAP tag may sterically hinder Ana2 tetramerization, yielding a Ana2M 2 -SNAP-LC8 4 subspecies.

DISCUSSION
Ana2 is an integral component of the centriole duplication pathway, but how it works with Sas-6 and Sas-4 and whether LC8 plays a role in this pathway remain to be determined. The Sas-6 dimer interactions that facilitate cartwheel formation are very weak (K D of Ͼ100 M), making it unlikely that Sas-6 could spontaneously build cartwheels in a cellular context at endogenous levels. Additionally, Sas-6 overexpression promotes centriole amplification only when Ana2 is co-overexpressed, suggesting that Ana2 plays a supporting role in enhancing Sas-6 oligomerization and cartwheel formation (17). One mechanism by which Ana2 could promote Sas-6 oligomerization would be if Ana2 itself were oligomeric. This idea is supported by recent evidence that Ana2 binds LC8, a dynein light chain that plays a ubiquitous role as a dimerization machine (24). Our work provides insight into the Ana2-LC8 quaternary structure and establishes a foundation upon which the Ana2 tetramer's avidity effects on Sas-6 oligomerization can be investigated.
We have identified two LC8 binding sites in Ana2, conserved within the Drosophila genus, that flank a central domain with predicted helical structure (Fig. 1E). Although the exact binding sites are not apparent in other metazoan species, the presence of a central predicted coiled-coil is conserved across Ana2 orthologs from C. elegans Sas-5 to human STIL and suggests a role in oligomerization. This is supported by a report that the C. elegans Sas-5 N-terminal region (containing the central predicted coiled-coil) forms a tetramer in solution (49). Although Sas-5 tetramerization in vitro is not LC8-dependent, its oligomeric state parallels the LC8-dependent tetramerization that we observe with Ana2.
Dynein light chains often bind targets proximal to an endogenous oligomerization domain, potentiating target dimerization. Both of the Ana2 LC8-binding sites are an amalgam of the canonical K Ϫ3 X Ϫ2 T Ϫ1 Q 0 T 1 and G Ϫ2 I Ϫ1 Q 0 V 1 D 2 LC8-binding motifs. Using ITC, we have shown that Ana2 pep1 binds LC8 with micromolar affinity (K D ϭ 1.1 M). Our crystal structure of LC8 bound to Ana2 pep1 shows an LC8 2 -Ana2 pep1 2 binding mode, with the Ana2 pep1 canonical TQT sequence contributing key binding determinants.
Our second identified LC8-binding site (Ana2 site 2, pep2) flanks the central helical domain's C-terminal region and is composed of the sequence T Ϫ3 G Ϫ2 T Ϫ1 Q 0 C 1 D 2 . Ana2 pep2 binds LC8 with lower affinity (K D ϭ 13 M) than Ana2 pep1. Our data indicate the formation of an LC8 8 -Ana2 4 complex, which may have implications for the role of Ana2 in centriole duplication by clustering multiple Sas-4-binding (red ellipses) and Sas-6-binding (gray ellipses) domains. Each LC8 homodimer locally mediates parallel dimerization of Ana2. The model, as presented, portrays the central, predicted ␣-helix as a tetramerization domain. Whether this domain forms a tetrameric four-helix bundle remains to be determined, but it is presented as a parallel four-helix bundle (above) and an antiparallel four-helix bundle (below).

Structural Basis of LC8-mediated Ana2 Oligomerization
Our crystal structure of LC8 bound to Ana2 pep2 also has an LC8 2 -Ana2 pep2 2 binding mode. Interestingly, Ana2 pep2 adopts a unique architecture when bound to LC8 that contrasts with other LC8-peptide structures. The Ana2 pep2 Cys 244 at position ϩ1 is positioned deeper into the LC8 binding groove. The affinities we report for the LC8-Ana2 peptide interactions probably underestimate the stability of the biological complex involving full-length Ana2 and LC8. Because our solution studies support interactions between LC8 homodimers and a tetrameric Ana2M region, we anticipate that avidity effects will increase the complex's stability beyond the affinities we report for LC8 and Ana2 pep1 and pep2. This is consistent with the finding that a stable Ana2-LC8 complex can be extracted from Drosophila cell lysate (24). We note that within the genus Drosophila, the two segments that bridge the predicted central coiled-coil with the two flanking LC8 binding sites are not conserved in sequence or length. We predict that these segments serve as general spacers that link the LC8 binding sites to the Ana2 coiled-coil oligomerization domain and maintain a general length that enables LC8 homodimers to bind and potentiate Ana2 oligomerization without sterically compromising coiled-coil formation.
Our data support a model in which LC8 stabilizes an Ana2 tetramer (Fig. 7). An Ana2 tetramer may spatially arrange its conserved C-terminal STAN motifs to interact with Sas-6 and promote the Sas-6 oligomerization that underlies centriole cartwheel formation. Our SEC-MALS analysis of the Ana2M-LC8 complex reveals a stable, single-species complex consisting of four Ana2M molecules and eight LC8 molecules (Ana2M 4 -LC8 8 ). This stable complex was purified over two successive sizing columns, demonstrating its ready formation, and yielded a similar experimental mass in two independent purifications and SEC-MALS assays. Mutating the first Ana2M LC8 binding site as well as adding a SNAP tag to LC8 supported the Ana2M 4 -LC8 8 stoichiometry (Fig. 7).
The Ana2-LC8 interactions that we characterized raise important questions about the role of Ana2 in centriole duplication. Previous work has shown that the C-terminal half of Ana2 binds the N terminus of Sas-6 in Drosophila (12), implicating a possible role for the conserved STAN domain of Ana2 in Sas-6 binding. In our model, LC8 binds and stabilizes an Ana2 tetramer that may structurally organize four trans STAN domains at one end of a parallel tetramerization domain or two trans STAN domains at either end of an antiparallel tetramerization domain (Fig. 7). In either configuration, the oligomeric state of Ana2, coupled with its ability to bind Sas-6, is predicted to enhance Sas-6 oligomerization and cartwheel formation. This correlates with cellular studies in which Sas-6 and Ana2 dual overexpression was required for cartwheel formation, suggesting that Ana2 potentiates Sas-6 cartwheel formation, potentially through oligomerization (17). Recent cryotomographic studies of nascent centriole architecture reveal auxiliary protein density connecting the Sas-6-based cartwheel to Sas-4 and the distal microtubule triplets (3). Given the integral role of Ana2 in Sas-6 cartwheel formation as well as evidence that it binds both Sas-6 and Sas-4, Ana2 is a likely candidate for this density. More work is needed to determine if Ana2 can bridge Sas-6 and Sas-4 and whether the LC8-Ana2 interaction plays a role in this Ana2 function, as it does in neuroblast asymmetric cell division. Our work outlines the structural basis of the LC8-Ana2 interaction, with implications for its role in Ana2 structure and function at the centriole.