The Crystal Structure of Dynein Intermediate Chain-Light Chain Roadblock Complex Gives New Insights into Dynein Assembly

Abstract

The roadblock/LC7 dynein light chain is a ubiquitous component of all dyneins and is essential for many diverse processes including proper axonal transport and dendrite growth. In addition, LC7 functions in non-dynein transcriptional activation of the transforming growth factor-β complex. Crystal structures of Drosophila melanogaster LC7 in the apo form and in complex with a segment of the disordered N-terminal domain of dynein intermediate chain (IC) provide the first definitive identification of the IC sequence recognized by LC7. The site, confirmed by isothermal titration calorimetry studies, overlaps the IC sequence considered in the literature to be an IC self-association domain. The IC peptide binds as two amphipathic helices that lie along an extensive hydrophobic cleft on LC7 and ends with a polar side-chain interaction network that includes conserved residues from both proteins. The LC7 recognition sequence on IC and its interface with LC7 are well conserved and are, thus, likely representative of all IC·LC7 structures. Interestingly, the position of bound IC in the IC·LC7 complex mimics a helix that is integrated into the primary structure in distantly related LC7 homologs. The IC·LC7 structure further shows that the naturally occurring robl^Z deletion mutation contains the majority of the IC binding site and suggests that promotion of IC binding by phosphorylation of LC7 is an indirect effect.

Introduction

Cytoplasmic dyneins are large multisubunit protein complexes that are responsible for ATP-driven transport of diverse cargo along microtubules. They play fundamental roles within the cell including mitotic spindle assembly and orientation (1), chromosome segregation (2), and intracellular trafficking of vesicles and mRNA (3). Dyneins are essential for the development and maintenance of neurons (4), and dynein dysfunction is associated with several human diseases, such as lissencephaly (5), neural degeneration (6), and male infertility (7).

Dynein heavy chains are responsible for motor activity, whereas intermediate chain (IC)² and light chain subunits comprise the cargo attachment complex. The N- and C-terminal domains of IC are structurally and functionally independent. The primarily disordered N-terminal domain (N-IC) is central to dynein assembly, regulation, and cargo binding as it contains a self-association domain and the binding sites for the three light chains, the p150^Glued subunit of dynactin, and several putative cargoes (8–11). Dynein light chains Tctex1, LC8, and LC7 are all integral components of both cytoplasmic and axonemal dyneins (12) that bind distinct regions of N-IC (13, 14) (Fig. 1). Tctex1 and LC8 are dimeric structural homologs, and each binds two chains of IC at its dimer interface (15–17). Tctex1 and LC8 show mutually enhanced affinity, as one protein binds two IC chains to form a bivalent IC that has higher affinity for the other light chain (17). This led to the proposal that Tctex1 and LC8 work together to create a poly-bivalent IC duplex that serves as a stable and versatile scaffold providing tighter IC self-association and higher affinity for multiple bivalent binding partners (17).

FIGURE 1.

IC light chain binding regions and constructs used in this study. The upper IC diagram locates the verified binding sites for Tctex1 and for LC8 (hatched) along with a proposed IC self-association domain (black) and the proposed binding site for LC7 (open). The second diagram updates the model to include the LC7 recognition sequence identified in this study (gray). IC recognition sequences for Tctex1 and LC8 are, respectively, residues 110–122 and 126–135 (15–17). The four additional lines define the residue ranges of the IC constructs used in this study. Numbering is relative to the D. melanogaster Cdic2b gene, subscripts T, L, 7, and d7 stand for Tctex1, LC8, LC7, and disrupted LC7 recognition sequences, respectively.

LC7, also called roadblock or km23, is a ubiquitous component of cytoplasmic dyneins. The roadblock name originates from knock-out mutants in Drosophila melanogaster that result in posterior sluggish motility leading to complete paralysis. LC7-null mutants in D. melanogaster have mitotic defects (18) and display phenotypes with defective axonal transport, neuronal blast cell division, and dendrite growth (19). LC7-null mutants result in flagellar assembly and motility defects in Chlamydomonas reinhardtii (20) and disrupted dynein-mediated nuclear distribution in Aspergillus nidulans (21).

NMR and x-ray crystal structures of apoLC7 from Homo sapiens and Rattus norvegicus (22–24) show LC7 is a homodimer structurally unrelated to Tctex1 and LC8. LC7 belongs to an ancient protein superfamily that is widely represented in archaea and bacteria and is implicated in regulation of nucleoside triphosphatase activity (25). Superfamily members share a common structural fold of five β-strands and three α-helices, except the C-terminal α-helix is missing in LC7. LC7 apparently has multiple non-dynein interaction partners including the Rab6 family of GTPase regulators (26), the human reduced folate carrier (27), and the transforming growth factor-β receptor complex (28, 29), but in no case is the molecular-level interaction with a binding partner characterized. Consistent with a role for LC7 in multiple cellular pathways, mutations of LC7 (30) or changes in LC7 isoform expression levels (31) have been observed in ovarian and hepatic cancers.

Phosphorylation is one mechanism that appears to regulate the multiple roles of LC7. LC7 is serine-phosphorylated after transforming growth factor-β receptor activation and binds IC in response to this activation. Two phosphorylation site mutants disrupt the interaction with IC (30), leading to the conclusion that phosphorylation of LC7 is necessary for IC binding. However, at present there is no clear molecular-level picture of processes associated with LC7 phosphorylation/dephosphorylation or of its involvement in IC binding.

As part of our ongoing effort to elucidate structure-function relationships of dynein light chains and build a comprehensive understanding of dynein assembly, we report here the first molecular-level structure of an IC·LC7 complex from D. melanogaster (79% sequence identity to H. sapiens LC7). The IC·LC7 complex structure reveals unexpected insights into dynein assembly and regulation. It also provides interesting evolutionary perspectives. Comparing LC7 to its likely ancestral fold suggests that the multifunctionality of LC7 is the result of its recruitment for new dynein-related function even while ancient functions are maintained.

EXPERIMENTAL PROCEDURES

Protein Preparation

The gene for D. melanogaster LC7 was a generous gift from Dr. T. Hays (University of Minnesota). The LC7 and IC constructs were cloned into pCR2.1 TOPO using the TOPO TA cloning kit (Invitrogen) then subcloned into pET15d (Novagen, Darmstadt, Germany). The IC construct corresponding to residues 212–260 (IC₇) was cloned into pET SUMO (Invitrogen) which contains an N-terminal SUMO fusion protein. All constructs included an N-terminal hexahistidine tag, and their sequences were verified by automated sequencing.

Recombinant proteins were expressed in Escherichia coli BL21(DE3) host cell lines. Cells expressing LC7, IC_TL-d7, IC_TL7, and IC_92–289 were grown in LB media, and cells expressing IC₇ were grown in TB media. All cells were grown at 37 °C to an A₆₀₀ of ∼0.6. LC7 was induced with 0.4 mm isopropyl-1-thio-β-d-galactopyranoside at 22 °C, and IC constructs were induced as previously described (13). Proteins were initially purified using a nickel-nitrilotriacetic acid column (Qiagen); the IC constructs were refolded on the column. The hexahistidine tag was cleaved from LC7 using FactorXa protease (Novagen), and the SUMO tag was cleaved from IC₇ using Ulp1-Sumo protease (32). Final purification of LC7, IC_TL-d7, IC_TL7, and IC_92–289 was done using size-exclusion chromatography (HiLoad Superdex75 26/60, GE Healthcare). Final purification of IC₇ was done using high performance liquid chromatography on a YMC C-18 column in 0.1% trifluoroacetic acid with a linear gradient (40–65%) of acetonitrile. Protein purity was verified by SDS-PAGE and matrix-assisted laser desorption ionization time-of-flight mass spectrometry: LC7 molecular mass = 10,823.6 Da (10,823.4 theoretical), IC_TL7 molecular mass = 20,401.8 Da (20,518.5 theoretical), IC₇ molecular mass = 5,803.6 Da (5,804.5 theoretical); IC_92–289 = molecular mass 24,122.0 Da (23,935.4 theoretical). The deviation from theoretical mass for IC_TL7 may be due to N-terminal demethionation and retention of a Na⁺ ion. The higher mass of IC_92–289 could be due to its low signal to noise and ambiguous centroid mass or a covalent difference as in IC_TL7. Protein concentrations were determined from sequence-based calculation of absorptivity at 280 nm (33).

X-ray Crystallography

ApoLC7 and the IC_92–289·LC7 complex were stored in 15 mm sodium chloride, 5 mm Tris at pH 7.5. The monomeric concentration of LC7 in both apoLC7 and the IC_92–289·LC7 complex was 0.5 mm, with a 2-fold excess of IC_92–289. Crystals were obtained at 4 °C using 2 μl of hanging drops with an equal volume of reservoir solution equilibrated against a 400 μl reservoir.

In attempts to crystallize a supercomplex of IC_92–289·Tctex1· LC8·LC7, IC_92–289 was proteolytically degraded, but some apoLC7 crystals formed. Optimization of these crystals was done using LC7 alone, and diffraction quality apoLC7 crystals were grown using a reservoir of 20% isopropanol, 20% polyethylene glycol 4000, 100 mm sodium citrate at pH 5.6. Rectangular crystals grew to a final size of 0.1 × 0.12 × 0.05 mm³. For IC·LC7, a 1:500 (v:v) of FactorXa protease (Novagen) was added to a stock solution of IC_92–289·LC7 to aid opportunistic crystal lattice formation. Hexagonal rod crystals were obtained using a reservoir of 1 m sodium citrate, 100 mm sodium chloride, 100 mm Tris at pH 7.0. Crystals grew to a final size of 0.25 × 0.06 × 0.06 mm³ within 2 weeks. Mass spectrometry analysis of IC·LC7 crystals showed IC_92–289 was proteolysed with no fragments larger than 9 kDa present. Crystals of the IC₇·LC7 complex having the same morphology and diffraction properties as those of IC·LC7 were reproducibly grown using the same crystallization conditions but with a 1.1-fold excess of IC₇ replacing IC_92–289 and with no FactorXa protease added.

Crystals were pulled through oil before flash-freezing in loops using liquid nitrogen. For apoLC7, 1.95 Å resolution oscillation data (Δϕ = 1.0°) were collected using an in-house Raxis IV system with CuKα radiation. For IC·LC7, oscillation data (Δϕ = 1.0°) were collected using HHMI beam line 5.0.1 at the Berkeley Advanced Light Source. Data were integrated using MOSFLM (35) and scaled using SCALA (36). For IC·LC7, data were input into the diffraction anisotropy server without B-factor sharpening (37), and an elliptical resolution boundary of 2.7 × 3.0 × 3.0 Å (a*, b*, and c*, respectively) was chosen based on the default three F_o/σF_o cutoff. Thus, the IC·LC7 complex has a nominal resolution of 3.0 Å but includes data to 2.7 Å in the a* lattice direction. The apoLC7 space group was P2₁ with a unit cell of β = 101.8°, a = 56.22, b = 58.69, and c = 65.10 Å; that for IC·LC7 was P3₂ with unit cell a = b = 107.13, and c = 65.29 Å. Crystallographic data collection and refinement statistics are summarized in Table 1.

TABLE 1.

Data collection and refinement statistics for apoLC7 and the IC·LC7 complex

	ApoLC7	IC·LC7a
Data
Resolution (Å)	63.72-1.95 (2.00-1.95)b	53.6-3.0 (3.15-3.0)
Completeness (%)	98.9 (98.90)	100 (100)
I/σ	9.15 (2.50)	6.4 (0.80)
Rmeas	0.14 (0.46)	0.09 (1.12)
Rpim	0.07 (0.23)	0.04 (0.47)
Total reflections	271,466	95,560
Unique reflections	27,041	15,919
Refinement
Resolution (Å)	63.72-1.95 (2.00-1.95)	53.6-2.7 (3.15-3.0/2.86-2.7)
No. reflections used	27,041 (1,802)	17,314 (839/393)
Rcryst	0.19 (0.28)	0.20 (0.31/0.42)
Rfree	0.25 (0.37)	0.24 (0.40/0.47)
Average B-factor (all atoms) (Å2)c	27	103
RMSD from ideal
Bond length (Å)	0.022	0.018
Bond angles (°)	1.9	1.9
ϕ,ψ preferred (%)d	95.5	86.1
ϕ,ψ allowed (%)	100	99.2

^a Data statistics are reported for a nominal resolution limit of 3.0 Å for which all collected data were used; as refinement included some additional data to 2.7 Å (see “Experimental Procedures”), the refinement statistics are reported with that cutoff, and information is given for both the 3 and 2.7 Å highest resolution bins. R_meas is the redundancy corrected indicator of the precision of individual measurements (58) and R_pim describes the precision of the averaged measurements (59).

^b Values in parentheses are for the highest resolution bin.

^c Average B-factors are within the acceptable range for their resolution (57).

^d Reported values are based on MOLPROBITY (34). The values using PROCHECK (46) are within 4% of the MOLPROBITY values. The only ϕ,ψ outlier is Lys-85 in loop 6, which has weak density and, therefore, is not reliably determined. Lys-85 has (ϕ,ψ) ≈ (−115°, 55°) in apoLC7, and (ϕ,ψ) ≈ (−80°, 100°) in the IC·LC7 complex.

Phases were determined for both crystals by molecular replacement using PHASER (38) with PDB structure 2HZ5 (24) as the search model. In each case two dimers of LC7 were placed. Electron density maps calculated for the IC·LC7 complex show extra helical density for four IC chains. Model building was performed in Coot (39), and both apoLC7 and the IC·LC7 complex were refined using restrained isotropic B-factor refinement with one translation libration scres domain per chain in REFMAC (40). Non-crystallographic symmetry restraints were applied to the four IC and LC7 chains in the IC·LC7 complex. To confirm the accuracy of the model, unbiased electron density maps were generated starting with a molecular replacement model leaving out side chains for residues Arg-71, Arg-73, Glu-78, and Gln-93. Refinement of this model with poly-Ala helices for α_IC2 and the final model show unbiased density for residues Arg-71, Arg-73, Glu-78, and Gln-93 of LC7 and Glu-252 and Asn-253 of IC.

Solvent-accessible surface area buried upon complex formation was calculated using SURFACE RACER 5.0 (41). The atomic coordinates and structure factors for apoLC7 and the IC·LC7 complex have PDB accession numbers 3L7H and 3L9K, respectively.

Isothermal Titration Calorimetry

Proteins were dialyzed into 50 mm sodium phosphate, 50 mm sodium chloride, 1 mm sodium azide at pH 7.5. Thermodynamics of binding were determined at 25 °C using a VP-ITC isothermal titration calorimeter (MicroCal, Northampton, MA). Experiments were conducted with IC₇ and IC_TL7 in the sample cell and LC7 in the syringe. IC_TL7 was pre-bound by a 4-fold excess of Tctex1 and LC8. Data were processed using Origin 7.0 (OriginLab Corp., Northampton, MA). Heat of dilution, estimated to be equal to the enthalpy of the final injection, was subtracted from the binding data before fitting. LC7 binds IC₇ or IC_TL7·Tctex1·LC8 with a stoichiometry of 1.0 ± 0.1, but whereas IC_TL7·Tctex1·LC8 is well fit by a single-site binding model (A + B → AB, where A and B refer to a single chain of IC and LC7, respectively), IC₇ data show small deviations from the best fit. Nonetheless, a single-site binding model was chosen for IC₇·LC7 in the absence of further information justifying additional fitting parameters. For all experiments the “c value” (c = [protein]_{sample cell} × K_d⁻¹) was within the 5–500 range required for reliable determination of association constants (42) using cell/syringe concentrations of 0.028/0.55 and 0.045/0.55 mm for IC₇ and IC_TL7, respectively. Average values are reported with error estimation from the difference between experimental repeats (Table 2).

TABLE 2.

Thermodynamic parameters for LC7 binding to IC constructs at 25 °C

Values are the average and the standard deviation of three (IC₇) and two (IC_TL7) experiments.

Cell	Syringe	Kd	ΔG°	ΔH°	−TΔS°
		μm	kcal/mol	kcal/mol	kcal/mol
IC7	LC7	5.7 ± 0.5	−7.2 ± 0.1	−15.0 ± 0.3	7.8 ± 0.3
ICTL7·Tctex1·LC8	LC7	2.0 ± 0.1	−7.8 ± 0.1	−14.8 ± 0.5	7.1 ± 0.7

Size Exclusion Chromatography and Multiangle Light Scattering

To determine association states, samples from isothermal titration calorimetry (ITC) were run on an analytical size-exclusion column (Superdex200 10/300, GE Healthcare) at 0.5 ml/min in 200 mm sodium sulfate, 50 mm sodium phosphate, 1 mm sodium azide at pH 7.3. The monomeric concentrations of Tctex1, LC8, LC7, and the IC_TL7·Tctex1·LC8·LC7 complex were 0.135, 0.135, 0.068, and 0.045 mm at loading, respectively. Multiangle light scattering (mini-Dawn, Wyatt Technology) and refractive index (ProStar 350, Varian) data were collected and processed using ASTRA v5.1.9.1 (Wyatt Technology). The peak corresponding to the IC_TL7·Tctex1·LC8·LC7 complex was collected during elution and concentrated 10-fold for SDS-PAGE analysis.

RESULTS

Crystal Structure of ApoLC7

The apoLC7 crystal structure was solved to 1.95 Å with two LC7 homodimers in the asymmetric unit, each dimer with the expected LC7-fold seen for H. sapiens and R. norvegicus apoLC7 (22–24); that is, a pair of α helices (α₁, α₁′) flanking a continuous antiparallel 10-stranded β-sheet (strand order β₂-β₁-β₅-β₄-β₃·β_3′-β_4′-β_5′-β_1′-β_2′) on one face and a 2-helix bundle (α₂·α_2′) on the opposite face (Fig. 2a). The four LC7 chains in the asymmetric unit have somewhat different N-terminal electron density, with chain B being the best defined (starting residues are Leu-9 for chain A, Met-1 for chain B, and Val-5 for chains C and D). In contrast, the C-terminal density of each chain is equivalent, ending at Asn-94 with residues 95–97 too disordered to model. Conformational differences among the four chains are observed for helix α₁ and the loops connecting strands β₁ to β₂ (L₂) and β₄ to β₅ (L₆) (Fig. 2c). The differences in α₁ among the four chains are due to variations in the angle of the loop connecting α₁ to β₁ (L₁), with α₁ acting as a rigid rod to amplify differences N-terminal to L₁. These differences appear related to crystal packing near L₁.

FIGURE 2.

Structures of apoLC7 and the IC·LC7 complex. a, two orthogonal views are shown of the LC7 homodimer. Chains B (steel blue) and D (gray) are shown with secondary structural elements labeled for chain B. b, the same two views are shown of the IC·LC7 complex with chains of IC in orange and light orange and IC α-helix 1 (α_IC1, residues 223–230) and 2 (α_IC2, residues 233–254) labeled. c, shown is a stereoview of an overlay of the four independent apoLC7 chains (semitransparent bright blue) onto one LC7 chain from the IC·LC7 complex as colored in panel b. Regions moving upon IC binding occur in α-helix 1 (α₁), Loop 2 (L₂), and Loop 6 (L₆). d, shown is a stereoview of the IC·LC7 interface. Side chains for hydrophobic residues burying greater than 5 Ų and the residues of the fingerprint region are shown. Gly-243 Cα position is indicated.

Interestingly, conformational variations among the four chains in the asymmetric unit recapitulate differences between previously reported structures of apoLC7, with chain B similar to the average conformation in NMR structures (22, 23) and the orientation of α₁ in chain A similar to α₁ in the 2.1 Å H. sapiens LC7 crystal structure (24). The latter also shows the first five residues and loop L₆ are too disordered to fully model.

Crystal Structure of the IC·LC7 Complex

IC_92–289 bound to LC7 fortuitously crystallized in the presence of protease. Using these crystals, the IC·LC7 complex was solved at a nominal resolution of 3.0 Å with the asymmetric unit containing two LC7 homodimers binding two IC chains per dimer. Each chain of IC and LC7 in the asymmetric unit has similar electron density, with IC chains initially observed as helical density not accounted for by LC7. Data interpretation at 3 Å is challenging, so the N-to-C directionality of IC segments were determined by modeling a poly-Ala chain in both directions, with one orientation clearly fitting better to the unbiased electron density. The unique sequence and registry of IC was determined from the pattern of side chain densities, including buried Leu, Ile, and Phe residues at the IC·LC7 interface. This ultimately guided the modeling of residues Leu-221 through Thr-258 for each IC chain and Gln-3 through Asp-97 for each LC7 chain. A set of unbiased electron density maps provide strong support for the final interpretation (Fig. 3 and supplemental Fig. 1). Based on this interpretation, the IC₇ construct, encompassing IC residues 212–260 (Fig. 1), was designed for use in subsequent studies. As a further verification that the IC segment bound in the IC·LC7 crystals was correctly identified, IC·LC7 crystallization conditions (without in situ proteolysis) produce isomorphous crystals of the IC₇·LC7 complex. The IC₇·LC7 crystals grew within 4 days, and a low resolution data set shows helical density for IC₇ bound to LC7. These crystals diffract similarly to the original IC·LC7 crystals produced from in situ proteolysis and, thus, were not analyzed further.

FIGURE 3.

Unbiased electron density evidence for the IC·LC7 complex. a, F_o − F_c electron density shows the IC present in the IC·LC7 complex after molecular replacement with an LC7 model (steel blue). A Cα trace of the final IC model (with 4 Cα atoms labeled) and some side chains of LC7 are included for reference. Side chains for residues Arg-71, Arg-73, Glu-78, and Gln-93 (cyan) are omitted. Density is contoured at 3σ (green) and −3σ (magenta). b, 2F_o − F_c and F_o − F_c maps calculated after inclusion in the refinement of a 19-residue poly-Ala helix (carbon, oxygen, and nitrogen atoms are colored orange, red, and blue, respectively) are modeled into the strong helical density seen in panel a with an N-to-C directionality parallel to α₁ of LC7. 2F_o − F_c density is contoured at 2σ (blue), and F_o − F_c density is contoured as in panel a. Labels are based on the final interpreted model. c, 2F_o − F_c and F_o − F_c density for LC7 side chains Arg-71, Arg-73, Glu-78, and Gln-93 (gray/steel blue) and IC side chains Glu-252 and Asn-253 (light orange/orange) after refinement of a model with the complete IC·LC7 complex except these side chains. Potential hydrogen-bonding side chains are connected by dashed lines with the hydrogen bond from Glu-252 drawn halfway between ηN1 and ηN2 of Arg-71 to reflect uncertainty in side chain position in this analysis. 2F_o − F_c density is contoured at 1σ (blue), and F_o − F_c density is contoured as in panel a.

The final IC·LC7 model shows IC binds LC7 as two amphipathic helices separated by a turn, with IC chains entering and exiting parallel to the LC7 2-fold symmetry axis (Fig. 2b). The first IC α-helix (α_IC1, residues 223–230) packs against β₄ and L₂ of LC7, the turn (residues 231–232) caps α₁ of LC7, and the second α-helix (α_IC2, residues 233–254) packs against α₁, β₄, β₅, L₁, and L₆ of LC7. α_IC1 has weaker electron density (and higher B-factors) than α_IC2. α_IC1 and α_IC2 have two and five complete helical turns, respectively, with α_IC2 having a bend after the third helical turn. IC residues 255–258 have a non-regular extended conformation and interact with residues from L₁ and the C terminus of LC7 (Fig. 2b).

The IC·LC7 binding surface is large, comprising 22 IC residues and 27 LC7 residues and accounting for close to 2400 Ų of surface, of which nearly 75% (≈1800 Ų) is nonpolar (Fig. 2d, and supplemental Fig. 2 and Table 1). Multiple packing interactions are formed by each of the highly conserved residues Phe-235, Phe-238, Ile-246, and Leu-250 of IC and Val-5, Val-31, Phe-69, Phe-87, and Leu-89 of LC7. The C-terminal end of α_IC2 contains a network of buried inter- and intrachain hydrogen bonding and salt bridge interactions that may constitute a specificity-determining fingerprint for IC·LC7 recognition.

The fingerprint region contains well conserved polar residues including Glu-252 and Asn-253 of IC and Lys-16, Arg-71, Arg-73, Glu-78, and Gln-93 of LC7. Although side chains in a 3 Å structure are sometimes poorly defined, the residues of the fingerprint region that are largely buried at the IC·LC7 interface have well defined side chain densities (Fig. 3c). Glu-252 and Arg-73 are partially solvent-exposed and have the weakest side chain densities, whereas Arg-71 is almost completely buried and has a strong side chain density (supplemental Table 1). Arg-71 stacks against Arg-71′, as observed in other systems (43, 44) while interacting with Glu-252 and Glu-252′ at the LC7 2-fold axis (Fig. 3c). A polar interaction network bridges the Asn-253 side chain, Arg-71 guanidino group, Gln-93 side chain, and the Lys-16 backbone (Fig. 3c).

Whereas α₁, L₂, and L₆ are the most structurally variable regions of apoLC7 (Fig. 2c), the four LC7 chains in the asymmetric unit of the IC·LC7 complex have consistent positions and strong electron density for these segments despite the lower resolution of the IC·LC7 complex. Interestingly, the α₁ position in chain A of apoLC7 is close to that of α₁ in the IC·LC7 complex and also contains a hydrogen bond between Nϵ of Gln-93 and the carbonyl oxygen of Lys-16; conversely, the α₁ position in chain B (which also matches the NMR solution structure) undergoes a 7 Å shift. All four chains have consistent shifts of 2.5 Å for L₂ and 4 Å for L₆ (Fig. 2c).

Thermodynamics of IC·LC7 Complex Formation

ITC was used to further confirm the newly determined LC7 recognition sequence and to establish the thermodynamic parameters governing IC·LC7 association (Fig. 4, a and b, Table 2). IC constructs used are given in Fig. 1. The designed IC₇ peptide, containing the crystallographically determined recognition residues 221–258, binds LC7 with moderate affinity of K_d = 5.7 μm. For comparison, LC7 binds to the more physiologically relevant IC_TL7·Tctex1·LC8 complex with a K_d = 2.0 μm. For both complexes, ΔG°, ΔH°, and −TΔS° values are quite similar, consistent with the IC₇ segment containing all of the LC7 interactions important in the larger IC_TL7 construct (Table 2). Data for IC₇·LC7 binding show slight deviations from a theoretical best fit for a single-site binding model; however, this is not true for LC7 binding to the IC_TL7·Tctex1·LC8 complex. For further confirmation of the LC7 recognition sequence, we determined if LC7 binds IC with a disrupted recognition sequence (IC_TL-d7). The absence of detectable binding (data not shown) implies that residues 238–260 are important for binding.

FIGURE 4.

Representative ITC and gel filtration data for LC7 binding to IC at 25 °C. ITC data with thermograms (top panels) and isotherms (bottom panels) for LC7 binding to IC₇ (a) and IC_TL7 (b) (pre-bound by a 4-fold excess of Tctex1 and LC8). Solid lines correspond to the non-linear least squares fit. c, a sizing gel (Superdex200, 10/300) elution profile for samples of the IC_TL7·Tctex1·LC8·LC7 complex produced in ITC experiments (from panel b) gave a single peak for the complex along with peaks for excess Tctex1, LC8, and LC7, and dotted lines indicate multiangle light scattering determined mass for each peak. LC8 and LC7 are not resolved, and separate injections of apoIC_TL7, LC8, and LC7 were used to determine individual masses and verify co-elution of LC8 and LC7 (data not shown). The void volume for this column is 10 ml. d, is an SDS-PAGE analysis of the eluted IC_TL7·Tctex1·LC8·LC7 complex (from panel c). The quaternary complex (second lane from the left) has bands of approximately equal staining density and mobility matching free IC_TL7 (lane 3), Tctex1 (lane 4), LC7 (lane 5), and LC8 (lane 6).

Analytical gel filtration chromatography coupled to multiangle light scattering gave molecular masses of 20.9 and 112.2 kDa for a dimeric apoIC_TL7 and a 2:2:2:2 complex of IC_TL7·Tctex1·LC8·LC7 (Fig. 4c), in good agreement with the 20.5 and 110.6 kDa theoretical masses. SDS-PAGE analysis of the peak corresponding to the quaternary complex shows IC_TL7, Tctex1, LC8, and LC7 are present at comparable concentrations (Fig. 4d). The association state of IC₇ was confirmed to be monomeric as it migrates slower than a similar sized IC construct that is monomeric and disordered (17, 45) (supplemental Fig. 3).

DISCUSSION

Essential Residues for IC·LC7 Binding

Residues in the IC·LC7 interface are well conserved among species (Fig. 5a, supplemental Fig. 4), suggesting universality of this interaction mode among D. melanogaster and vertebrate dyneins. Two conserved patterns are notable; the first is helix α_IC2, which is amphipathic in all IC sequences and packs into the hydrophobic cleft created by α₁, β₄, and β₅ of LC7. The second is electrostatic interactions between Asn-253 of IC with Arg-73 and Gln-93 of LC7, which replaces an intrachain salt bridge between Arg-73 and Glu-78 in apo LC7, and between Glu-252 and Glu-252′ from each IC chain with Arg-71 and Arg-71′ of LC7 (Fig. 3c). The electrostatic interaction network formed in the complex may explain why IC constructs ending at residue 250 (23, 47, 48) and 237 (this work) do not bind LC7 even though they contain most of the hydrophobic packing interactions. Consistent with a substantial electrostatic interaction, IC·LC7 binding is abolished in 1 m NaCl (47). These data support the conclusion that the specific interactions between Glu-252 and Asn-253 of IC with conserved polar residues of LC7 form a fingerprint region necessary for strong IC·LC7 binding.

FIGURE 5.

Location of residues impacted in the robl^Z mutant; differences between human LC7 isoforms and phosphorylation sites. a, sequences of LC7 from D. melanogaster (Dm), the robl^Z mutant of D. melanogaster (robl^Z), and H. sapiens LC7a and LC7b isoforms (Hsa and Hsb, respectively) are shown. Indicated also are the LC7 secondary structure (above), residues with more than 5 Ų of surface buried by IC (orange bars), Dm LC7 residues differing from Hsa and Hsb LC7 (cyan background), potential phosphorylation sites (30) (red background), and conserved sequence differences between Hsa and Hsb (yellow background) (see supplemental Fig. 4a for a complete sequence comparison of LC7 isoforms in vertebrates). Ser-33 (30) and Ser-38 (55) are both potential phosphorylation sites and conserved sequence difference. The 12-residue insertion of robl^Z is shown in lowercase letters. b, surface representation of LC7 (white) with IC bound (light orange/orange) show residues with more than 5 Ų surface buried by IC divided into those that are retained in robl^Z (dark blue), those that are not retained in robl^Z (bright blue), and those that may be functionally substituted by the 12-residue insertion of robl^Z (violet). The robl^Z mutant of LC7 results in the loss of residues essential for the LC7 homodimer but leaves at least 70% of the IC·LC7 interface (dark blue). c, potential phosphorylation sites of LC7 (red) and conserved sequence differences between LC7 isoforms (yellow) are shown. Solvent-accessible surface area calculations were determined using SURFACE RACER 5.0 (41). Figures were generated using Pymol (56).

The Newly Identified LC7 Recognition Sequence on IC Is Different from That Expected in the Literature

The crystallographic LC7 recognition sequence corresponding to IC residues 221–258 is upstream from the sequence implied by truncation mutations (47) and commonly presumed to be the LC7 recognition site (23, 47, 48). The proposal that the LC7 recognition sequence on IC corresponds to residues 250–289, immediately preceding a WD40 repeat domain, derives from two sets of observations. First, Susalka et al. (47) showed that LC7 binds to IC between residues 1–322 but does not appreciably bind to a construct ending at residue 250, suggesting that the binding sequence begins after IC residue 250 (numbering relative to D. melanogaster Cdic2b gene form). Second, the region preceding IC residue 250 maps to the IC self-association domain (48). Similarly, others proposed that apoIC dimerizes through an interchain coiled-coil based on sequence prediction involving residues 207–237 (Fig. 1) (Fig. 6 of Nurminsky et al. (49)). Experimental data consistent with formation of an IC-IC interchain coiled-coil includes enhanced helix content arising from light chain binding (Fig. 3 of Nyarko et al. (50) and Fig. 10 of Benison et al. (45)) as well as enhancement of IC·light chain binding with an IC construct that contains the proposed self-association domain (Fig. 4 of Williams et al. (16)).

FIGURE 6.

Bound IC mimics a helix contained in LC7 homologs. The IC·LC7 complex (light orange/orange and gray/steel blue, respectively) (a), the homodimeric Mgl complex (gray/steel blue) (b), and the heterodimeric MP1·p14 complex (gray and steel blue, respectively) (c) are shown. Arrows indicate the N-to-C directionality of helices α_IC2 and its cognate helix in the other complexes. The figures were generated using Pymol with PDB accession codes 1J3W and 1SKO for Mgl and MP1·p14, respectively (56).

NMR spectra of LC7 titrated with an IC construct of residues 250–289 show only six LC7 residues with chemical shift changes >0.08 ppm (23). Furthermore, the K_d of 100 μm reported in (23) is 20–50-fold weaker than the binding affinity of IC constructs reported here. The weak interaction and minimal chemical shift differences support our conclusion that IC residues 250–289 are not the true recognition sequence.

LC7 Binding Prevents IC Self-association at the Proposed Self-association Domain

Because the LC7 recognition sequence identified here, residues 221–258, overlaps with the predicted IC self-association domain, residues 207–237 (Fig. 1), any IC self-association in the presence of bound LC7 must occur N-terminal to Leu-221, which is the most N-terminal IC residue in the IC·LC7 crystal structure and 50 Å from Leu-221′ of the other IC chain (Fig. 2b). Far from positioning the two IC chains for mutual interactions in the self-association domain, it is not possible for LC7 to hold Leu-221 and Leu-221′ farther apart. If IC residues N-terminal to residue Leu-221 adopt the extended polypeptide-II conformation observed for unstructured peptides (51) and head directly toward each other, more than 12 residues from both chains would be required to close the 50 Å gap before any interaction occurs. Thus, with LC7 bound, a coiled-coil cannot form between IC residues 207–237.

LC7-induced Fit and IC Disorder-to-order Transition in Dynein Assembly

Conformational variations among the four apoLC7 chains is likely associated with collective motion of the N-terminal helix and disorder in loops 2 and 6 (Fig. 2c). In IC-bound LC7 these regions accommodate IC in a well defined average conformation in the IC·LC7 complex consistent with induced fit and mass action selection of LC7 forms receptive to IC binding.

To bind LC7, IC undergoes a disorder-to-order transition. In solution, free IC is primarily disordered but samples residual helix-like structure (45). Bound IC adopts a helical structure in the complex that matches its predicted secondary structure remarkably well, including the turn between the two helices (supplemental Fig. 5). An analogous disorder-to-order transition in IC also accompanies binding of Tctex1 or LC8, where the recognition sequence acquires a β-strand structure that adds to the β-sheet of the respective light chain (15–17). When fully assembled the IC·Tctex1·LC8·LC7 complex forms an IC duplex with two parallel IC chains bound to three homodimeric light chains; non-interacting regions of IC apparently retain disorder and associated flexibility (17, 45, 50).

Prebinding of Tctex1 or LC8, whose recognition sequences are three residues apart, results in a 50-fold mutual IC affinity enhancement for the other light chain due to multivalent interactions (17). Prebinding of Tctex1 and LC8 is expected to display a similar multivalent enhancement for IC binding to LC7 albeit of lower magnitude due to the longer (85 residue) disordered linker separating the LC8 and LC7 recognition sequences (supplemental Fig. 5). However, prebinding of Tctex1 and LC8 increases binding affinity of IC_TL7 to LC7 by less than 3-fold relative to IC₇ binding to LC7. It is interesting to note that LC7 binding to prebound IC_TL7 is better fit to a single-site model, suggesting that Tctex1 and LC8 binding restrict non-productive LC7 binding (Fig. 4, a and b).

Retained Interface Residues in robl^Z Deletion Mutation

A naturally occurring deletion mutation of LC7 (robl^Z), with the entire dimer interface missing, has a phenotype more severely affected than the completely null allele (18). One possible reason for the severity of the robl^Z phenotype is retained binding activity of robl^Z sufficient to compete with LC7 dynein interactions (18). Although this suggestion seems unlikely considering that robl^Z mutation in D. melanogaster results in a 54-residue deletion and a 12-residue insertion between β₂ and β₅ (Fig. 5a) and that a similar mutation in H. sapiens LC7 (30) that lacks the 12-residue insertion shows no interaction with IC, it is interesting to note that the retained sequence of robl^Z accounts for structural elements with more than 70% of the IC binding surface (Fig. 5b). This supports the possibility that a small population of D. melanogaster robl^Z adopts native-like conformations of α₁, β₁, and β₅ that could be stabilized by IC binding.

Regulation by Phosphorylation Is an Indirect Effect

There is an apparent contradiction between in vivo and in vitro effects of phosphorylation on IC·LC7 binding. For in vivo mammalian systems, IC binds only phosphorylated LC7 (28) and the S33A or S74A (numbering relative to D. melanogaster) single site mutants abolish IC·LC7 binding (30). However, in this work and in studies of mammalian IC there is ample evidence that IC binds unphosphorylated LC7 in vitro (47, 48). In the IC·LC7 structure, the two potential LC7 phosphorylation sites are distant from the IC binding site (Fig. 5c), suggesting that phosphorylation does not directly affect LC7 binding to IC but may modulate LC7 binding to other proteins. It is possible that interactions between unphosphorylated LC7 and an as yet unidentified protein(s) can either directly block the IC binding site or indirectly prevent IC binding by limiting the structural transitions of α₁, L₂, and L₆ seen between apo- and IC-bound LC7.

Vertebrates have two LC7 isoforms (LC7a and LC7b) that are expressed simultaneously in most tissues (31) but at varying levels (18, 31, 52). Interestingly, the few residues that have conserved sequence differences between paralogs (supplemental Fig. 4) are primarily located on LC7 surfaces distant from the IC binding interface (Fig. 5c). The different phosphorylation sites between isoforms and their distance from the IC binding interface suggests that functional differences between isoforms could be modulated by phosphorylation.

The Helix Segment of Bound IC Is Analogous to an Integral Part of LC7-related Proteins

LC7, homodimeric Mgl, and the heterodimeric MP1·p14 are structural homologs belonging to an ancient superfamily of small subcellular adaptor proteins (22, 23, 25) (Fig. 6). A major structural difference between these homologs and LC7 is that Mgl and MP1·p14 contain an amphipathic C-terminal helix (α₃) not present in apoLC7. Interestingly, α₃ packs in the spot occupied by α_IC2 in the IC·LC7 complex, although α₃ and α_IC2 lie in opposite directions. Binding of IC to LC7 adds an element of secondary structure to the apoLC7 fold to make it more like its homologs. The addition of an element of secondary structure to the fold of a light chain also occurs in IC binding to apoTctex1 and apoLC8 (53). In these cases a disordered segment of IC is incorporated as a single β-strand into each of the folds of apoTctex1 or apoLC8 and, in a manner analogous to IC·LC7, IC completes their fold topology. Integration of secondary structural elements into the fold of its binding partners appears to be a common theme for assembly of disordered IC with dimeric light chains and may exemplify a general mechanism for assembly of an elongated flexible scaffold in multisubunit complexes.

Assuming the common ancestor of this superfamily has the C-terminal helix, then the IC binding function of LC7 is a relatively recent adaptation for this superfamily. The remarkable placement of α_IC2 at the position of α₃ in other superfamily members suggests that LC7 or even the IC·LC7 complex might bind Rab6, the human reduced folate carrier, or the transforming growth factor-β receptor complex at sites different from the IC binding site and equivalent to where other LC7 superfamily proteins bind their partners. The conformational diversity of apoLC7 suggests that structural flexibility plays a role in its development of new specialized function(s) while retaining ancestral function(s) (54). In this view, the loss of the C-terminal helix was a key step in the evolution of conformational and functional diversity of LC7.