Crystal structure of dynein light chain TcTex-1.

TcTex-1, one of three dynein light chains of the dynein motor complex, has been implicated in targeting and binding cargoes to cytoplasmic dynein for retrograde or apical transport. Interactions between TcTex-1 and a diverse set of proteins such as the dynein intermediate chain, Fyn, DOC2, FIP1, the poliovirus receptor, CD155, and the rhodopsin cytoplasmic tail have been reported; yet, despite the broad range of targets, a consensus binding sequence remains uncertain. Consequently, we have solved the crystal structure of the full-length Drosophila homolog of TcTex-1 to 1.7 A resolution using MAD phasing to gain insight into its function and target specificity. The structure is homodimeric with a domain swapping of beta-strand 2 and has a fold similar to the dynein light chain, LC8. Based on structural alignment, the TcTex-1 and LC8 sequences show no identity, although the root mean square deviation between secondary structural elements is less than 1.6 A. Moreover, the N terminus, which is equivalent to beta-strand 1 in LC8, is splayed out and binds to a crystallographic dimer as an anti-parallel beta-strand at the same position as the neuronal nitric-oxide synthase peptide in the LC8 complex. Similarity to LC8 and comparison to the LC8-neuronal nitricoxide synthase complex suggest that TcTex-1 binds its targets in a similar manner as LC8 and provides insight to the lack of strict sequence identity among the targets for TcTex-1.

Cytoplasmic dynein is a protein motor complex that uses ATP hydrolysis to transport cargoes toward the minus end of microtubules. It is involved in a number of important processes, including nuclear migration, mitotic spindle orientation, Golgi and vesicular transport, and mRNA transport (for reviews see Ref. 1). Human immunodeficiency virus, herpes simplex virus, poliovirus, adenovirus, and rabies virus also exploit the dynein motor complex for intracellular transportation (2). Although dynein and the other motor complexes, ki-nesin and myosin, are distinct in function and architecture, they each encode a motor domain on one polypeptide and use accessory proteins to bind distinct cargo (1). Although there are over 45 members of the kinesin family and 16 members of myosin family in the human genome, there are only 4 members of dynein and only 1 cytoplasmic dynein (1). Therefore, the means by which cytoplasmic dynein is capable of participating in so many processes and targeting distinct cargoes remains an open and intriguing question.
Cytoplasmic dynein is a large, multimeric complex with a molecular mass of 1. The light chains fall into three classes, typified by TcTex-1, LC8, and LC7/Robl. The TcTex-1 class includes variants RP3 (54% identity to TcTex-1 in humans) and a more distant relative, TcTex-2 (20% identity to TcTex-1 in humans) that has a 70-residue N-terminal extension and is associated with axonemal dynein. (Two additional human TcTex-1 variants with N-terminal extensions are also found in BLAST searches but have not been described experimentally.) The LC8 class includes variants DNCL1 (93% identity to LC8 in humans) and the axonemal DNCL4 (35% identity), which also has an extended N-terminal tail (13 residues). The LC7/Robl class has at least two members (73% identical in humans).
One of the best examples implicating light chains in targeting cargo to dynein is the interaction between TcTex-1 and rhodopsin (13). TcTex-1 knockouts failed to transport rhodop-sin to the apical surface in vivo, and using RP3, a TcTex-1 homolog (55% identity), to rescue the knock-out also failed (30). Similarly, studies in polarized (Madin-Darby canine kidney cells showed that mutations linked to retinitis pigmentosa in the cytoplasmic tail of the rhodopsin receptor abolished Tc-Tex-1 binding (13). Most interestingly, TcTex-1 and RP3 are temporally and spatially regulated (31). Other functions associated with TcTex-1 include non-Mendelian meiotic drive (32), the production of functional sperm in Drosophila (33), and regular oscillatory nuclear movement in the meiotic prophase in fission yeast (34).
Recently, NMR experiments measured the chemical shifts and assigned the secondary structure of elements of TcTex-1; however, sample difficulties prevented the determination of the three-dimensional structure (9,35). Nonetheless, in their work, Mok et al. (9) identified a 19-residue stretch of the dynein intermediate chain that bound to TcTex-1 and showed that the N-terminal fraction of this peptide produced substantial chemical shifts in a set of residues in TcTex-1. Comparison of this sequence with sequences from a set of reported targets of TcTex-1 identified a region of basic residues at the N terminus, (R/K)(R/K)XX(R/K), suggesting a potential TcTex-1-targeting sequence. However, this finding remains ambiguous particularly considering that this sequence is not found in the binding regions identified for a different set of TcTex-1 ligands including rhodopsin, Fyn, and Trk receptors. In fact, another consensus sequence, (VS(K/H))T/S)X(V/T)(T/S)(N/Q)V, was reported recently (14) but again is only shared by a fraction of the reported TcTex-1 ligands. To better understand how TcTex-1 recognizes its ligands that share little sequence identity, we have solved the crystal structure of TcTex-1 as an initial step in defining the structural determinants of its target recognition.

MATERIALS AND METHODS
Cloning, Protein Purification, and Characterization-Initially, we cloned the mouse TcTex-1 homolog, purified it, and crystallized it. X-ray diffraction experiments were carried out, but the diffraction was very weak, and the initial cell constants and space group suggested that there were 18 -21 molecules per asymmetric unit. Consequently, we switched to the Drosophila homolog.
Full-length Drosophila TcTex-1 cDNA clones were obtained by PCR using EST-clones (LP04056 and LD36706) purchased from Research Genetics and placed into pET24d between the NcoI and XhoI sites (36). PCR primers were synthesized by the biopolymer core facility (Columbia University). The sequence of this construct was confirmed by DNA sequencing, and the expression plasmid was transformed into BL21-DE3 expression cells (Novagen). Overexpression was carried out at 37°C for 4 h.
The cells containing Drosophila TcTex-1 were sonicated and clarified by ultracentrifugation. The clarified lysate was passed over a HiLoad 26/10 Q FastFlow column (Amersham Biosciences), and protein was eluted with a NaCl gradient (100 mM to 1 M). TcTex-1 elutes at ϳ200 mM NaCl in 50 mM Tris-HCl, pH 8.0. The protein was concentrated to a volume of 1-3 ml and applied to HiLoad 26/60 Superdex 75 (Amersham Biosciences) at rate of 2 ml/min. Coomassie staining indicated that the protein was more than 95% pure, and mass spectrometry (M r ϭ 12481.2 atomic mass units) showed that the 111-residue protein was not modified. Selenomethionyl (SeMet)-labeled TcTex-1 was prepared and purified in a similar fashion except the cells were grown in a selenomethionine minimum media (37).
Characterization of Ternary Structure-Size exclusion chromatography was carried out to determine the oligomerization state of TcTex-1. Moreover, because there was a question of a pH-dependent monomerdimer equilibrium in LC8, the Superose 12 column (Amersham Biosciences) was equilibrated with two different buffers (38). The low pH buffer contained 50 mM MES, 100 mM NaCl, 1 mM EDTA, and 1 mM DTT, pH 5. The high pH buffer contained 50 mM CHES, 100 mM NaCl, 1 mM EDTA, and 1 mM DTT, pH 9. The protein applied to the column was in 10 mM Tris, 100 mM NaCl, 1 mM EDTA, and 1 mM DTT, pH 8.0. The column was calibrated using bovine serum albumin (66 kDa), ovalbumin (45 kDa), chymotrypsin (25 kDa), and lysozyme (14 kDa) for both buffers. TcTex-1 (25 l of 2 mg/ml solution) was applied to the column. The elution volume was measured and compared with the calibration graph as well as to each other.
Crystallographic Analysis-Purified protein was concentrated to ϳ40 mg/ml in 10 mM Tris-HCl, pH 8.0, 100 mM NaCl, 5 mM EDTA, and 5 mM DTT. Crystals were grown from 100 mM CHES over a pH range of 9 -10 and 2.    1. TcTex-1 structure. A, the overall structure of TcTex-1 is a dimer consisting of two ␣-helices followed by four ␤-strands, where the second ␤-strand is swapped with the crystallographic protomer (one protomer in blue and the other in yellow). In green are residues 8 -13 from an adjacent crystallographic symmetry mate that form an anti-parallel ␤-strand. B, 90 o rotation of TcTex-1 as shown in A about an axis that runs horizontal in the page. This view illustrates how the N terminus is splayed away from the "compact" structure. The ordered N-terminal amino acids, 8 -13, from crystallographic symmetry mates have been added and are shown in green. C, stereoview of C ␣ trace of a TcTex1 protomer as shown in A rotated 90 o about an axis that runs vertical in the page (every 10th residue is labeled). The electron density covering residues 8 -13 of the crystallographic symmetry shown in stereo indicates that the N-terminal tail is well defined. D, stereoview approximately along the 3-fold screw axis. Each dimer is individually colored. Note that the N terminus from each is splayed away and bound to a crystallographic partner.
Drosophila TcTex-1 crystals diffract to Bragg spacings beyond 1.5 Å. MAD data were collected at four wavelengths: 1 (0.99187 Å) at the low energy side remote from the absorption edge; 2 (0.97950 Å) at the inflection point; 3 (0.97897 Å) at the peak of selenium absorption curve; and 4 (0.96672 Å) at the high energy side remote from the absorption edge. One MAD data set was collected to a limit of 2.0 Å spacing, and another set was collected to 1.5 Å using a different SeMet crystal. For the 2.0 Å data set, 100 images (15 s exposure deg Ϫ1 image Ϫ1 ) were collected per wavelength for each of the direct beam and inverse beams, whereas 65 images (60 s exposure deg Ϫ1 image Ϫ1 ) were collected for the 1.5 Å data set.
Structure Determination and Refinement-The data were scaled and reduced using HKL (39). The crystal belongs to the space group P3 2 21 with cell dimensions of a ϭ b ϭ 60.60 Å and c ϭ 48.49 Å Tables I and  II). The selenium sites were determined using Shake-n-Bake (40), and the protein phases were calculated using SHARP (41). The low resolution phases (2.0 Å) were extended to 1.7 Å using DM of CCP4 suite (42). The initial model was generated using ARP-wARP (43) and completed using ONO (44); the model was then refined using CNS (45) and ARP-wARP. The final refined model contains residues 8 -111 of Drosophila TcTex-1, which correspond to residues 10 -113 of human Tc-Tex-1. The N-terminal 7 residues have no electron density in the maps, and thus they are disordered in the crystals. A summary of data collection and refinement statistics is given in Table III.

RESULTS AND DISCUSSION
Overall Structure of TcTex-1-The crystal structure has one molecule per asymmetric unit; however, each is paired by crystallographic dyad axes indicating that TcTex-1 is homodimeric. Size exclusion chromatography further confirms its dimeric state. Also, we could not produce a monomer from the dimeric form by changing the pH (data not shown) as found for LC8 studies (38). Each protomer core contains two ␣-helices followed by four ␤-strands (Fig. 1A). The N terminus of TcTex-1 is splayed away from the folded core and bound to an adjacent crystallographic "dimer" as an additional ␤-strand (␤0) (Fig.  1B). The second ␤-strand (␤2) is domain swapped and makes extensive main chain hydrogen bonds and side chain interactions with the other protomer. Sequence alignment shows that TcTex-1 is highly conserved across a number of species, particularly in ␤-strands (Fig. 2).
TcTex-1 Shares the Same Fold as PIN/LC8 -Despite the lack of sequence identity, TcTex-1 has the same fold as PIN/LC8. As shown in Fig. 3A, the ␣-helices, ␤-strands, and loops are generally longer in TcTex-1 compared with PIN/LC8, but mostly to the N-terminal half of the protein (based on the orientation of the bound N-terminal tail). The r.m.s.d. 1 between common elements is less than 1.6 Å. A structure-based sequence alignment of TcTex-1 and LC8 has zero sequence identity (Fig. 2). An alignment based solely on sequence, on the other hand, produced 11% identity, but this was achieved by overlapping the C terminus of LC8 with the N terminus of TcTex-1.
TcTex-1 Dimerization Interface-The interface between each TcTex-1 protomer is extensive and predominantly hydrophobic (supplemental Fig. 1). In our calculations, each protomer buries 1532 Å 2 of surface area, whereas each protomer in PIN/LC8 buries 760 Å 2 . It is interesting to note that there is a solvent channel running through the center of the TcTex-1 dimer (Fig. 3B), whereas the PIN/LC8 structure does not have such a feature. The function, if any, of this solvent channel is unclear. It does, however, produce a dimer interface with two hydrophobic patches. These patches include residues Met-66, Met-68, Leu-103, Leu-75Ј, and the methyl of Thr-77Ј for the N-terminal half of the molecule and residues Leu-49, Thr-53, Ile-62, Cys-81Ј, Tyr-82, Trp-83, Phe-107Ј, and Leu-109 for the C-terminal half (nonprimed and prime residues denote individual protomers).
Several polar interactions also participate in and possibly stabilize the dimeric state. One that stands out is between the FIG. 2. Sequence alignment. The TcTex-1 family is well conserved across all species. The sequence identity between the Drosophila and human is 59%. The largest difference is between Drosophila and Chlamydomonas reinhardtii, with a sequence identity of 55%. Human RP3 is 55% identical to human TcTex-1. Despite the similarity in folds, LC8 and TcTex-1 show no sequence identity based on structural alignment. The secondary structure elements of TcTex-1 are shown above the Drosophila sequence, and the secondary structure elements for LC8 are shown below its sequence. Residues that line the putative binding cleft are highlighted with an asterisk above, and those that are conserved or similar between TcTex-1 and RP3 are shaded violet and green, respectively. highly conserved His-76Ј (Asn in Caenorhabditis elegans), part of the domain swapped ␤-strand, and completely conserved Tyr-34 in the ␣1-␣2 loop and the highly conserved Asn-38 in the ␣-helix 2 (Ser in Chlamydomonas sp.). Because the molecule is symmetric, each interaction listed occurs twice.
Potential TcTex-1 Target-binding Site-TxTex-1 is similar in tertiary fold to LC8. Based on how it binds the N-terminal tail from a neighboring molecule in the crystal lattice, it is likely to bind its targets in the same fashion as LC8.
Peptides derived from putative LC8 cargo bind as an extended anti-parallel strand to swapped ␤-strand 2 of LC8, making eight backbone hydrogen bonds. This strand is surrounded by ␣-helix 2 on one side and ␤-strands 3 and 4 on the other.
All LC8 targets encode an invariant glutamine that caps the ␣-helix 2. The residues that precede and succeed this glutamine bind to hydrophobic sites derived from residues in the ␤-strands 3 and 4. Similarly, in TcTex-1, the residues directly before (Gln-9) and after (Ile-11) Phe-10, which corresponds to Gln-10 of LC8, are both buried in hydrophobic cavities.
A major difference between the binding groove of LC8 and the putative TcTex-1 groove is at the site on LC8 that interacts with the invariant glutamine in LC8 targets (Fig. 3, B and C). In TcTex-1, ␣-helix 2 is extended by an additional N-terminal turn that presents conserved residues His-34 and Asn-38 to the site.
The mechanism used by LC8 and TcTex-1, as we presume, to bind their respective targets provides a possible explanation of the low sequence identity observed among the LC8 target peptides and among the TcTex-1 target peptides. Arguments like those that explain sequence promiscuity in subtilisin substrate (46) and major histocompatibility complex-antigen complexes (47) also apply here; backbone hydrogen bonds and hydrophobic interactions do not place absolute restrictions on the sequence identity.
Our prediction for the importance of hydrophobic peptide residues on the binding targets to TcTex-1 differs from the basic motif, (R/K)(R/K)XX(R/K), proposed by Mok et al. (9). Glutathione S-transferase pull-down studies showed that a peptide derived from the dynein intermediate chain, LGRRLNKLGVSKVTQVDF, could bind TcTex-1. Titration of a truncated peptide (underlined) produced chemical shifts that map to the N-terminal half of the TcTex-1 structure (supplemental Fig. 2C) (9); however, this truncated positively charged peptide could not pull-down TcTex-1 (9). The chemical shift differences coincide with the negative charged surface of TcTex-1 (Fig. 3B) and likely reflect adventitious electrostatic binding.
A Potential Second Binding Site-It was shown recently that exchanging a single residue that differs between the two LC8 FIG. 3. A, superposition of TcTex-1 and LC8. Superposition of common secondary structure elements of TcTex-1 and LC8 produced a 1.6-Å r.m.s.d. fit. The stereo view of the superposition shows that TcTex-1 (blue) in general has longer ␣-helices, ␤-strands, and loops, particularly in the N-terminal half of the molecule than LC8 (red) (50). B, the electrostatic surface of TcTex-1. The N-terminal tail of the crystallographic related molecule has been included in a stick representation. It binds to the hydrophobic cleft in the C-terminal half of the molecule. Gln-9 and Ile-11 are pointed to ␤-strands 3 and 4 of the cleft, whereas Phe-10 and Val-12 point toward the ␣-helix 2. The C-terminal half of TcTex-1 is negatively charged and may interact with lysines and arginines. The N terminus (residues 8 -13) that points out of the plane of the page has been removed for visual clarity. The electrostatic surface was calculated with adaptive Poisson-Boltzmann solvers (51) and visualized with pmol (50). C, the electrostatic surface of LC8.LC8 was superimposed on to TCTex-1 and has been simply translated to the right of TcTex-1. The surface representation indicates a similar hydrophobic groove but much less surface area to the N-terminal half of the molecule (based on the direction of the bound peptide). The nNOS peptide is shown as a stick representation. Note that the N-terminal tail of TcTex-1 and the nNOS peptide both bind anti-parallel to the swapped ␤-strand and superimpose well in the hydrophobic region (r.m.s.d. C ␣ ϭ 0.96).
isoforms (98% identity) altered partitioning of the isoform from myosin V to the dynein intermediate chain (48). This position maps to ␣-helix 2, away from the nNOS-and Bim-binding site (48). Most interestingly, a point mutation in TcTex-1 that is associated with T-specific overexpression of TcTex-1, a phenotype associated with transmission ratio distortion or meiotic drive, maps to nearly an identical region on the surface (supplemental Fig. 2, A and B) (49). Additional differences between TcTex-1 and RP3 also map to solvent-exposed areas away from the hydrophobic pocket (data not shown). As in the case of DLC2, the differences found in the solvent-exposed helix suggest the possibility of an additional binding site or regulatory role. Consequently, it is possible that the flanking helices form a secondary binding site for additional targets.
Post-translational Modification of TcTex-1 Should Not Affect Ligand Binding-Campbell et al. (18) and Mou et al. (20) found that TcTex-1 is tyrosine-phosphorylated. There are four tyrosines in human TcTex-1 and three are conserved throughout the family (Fig. 2). Moreover, the conserved three are either buried or involved in extensive side chain interaction. However, there is an incompletely conserved tyrosine, Tyr-4, in the human, bovine, and the torpedo sequences, whereas it is a phenylalanine in mouse and rat and valine in Chlamydomonas. This position corresponds to the flexible region that is splayed away from the dimer core and, thus, potentially available for post-translational modification. However, Campbell et al. (18) showed that phosphorylation of TcTex-1 did not affect the binding of Fyn.
Conclusions-We have shown that TcTex-1 is similar to LC8 in tertiary and quaternary structure despite no sequence identity. Based on this similarity and common modes of peptide binding, we suggest that TcTex-1 binds its targets in a similar manner.