Most proteins in interaction networks have a small number of partners, while a few, called hubs, participate in a large number of interactions and play a central role in cell homeostasis. One highly conserved hub is the eukaryotic protein LC8, which was originally identified as an essential component of the multi-subunit dynein complex, is essential in multiple protein complexes in diverse systems. Indeed, LC8 is broadly distributed troughout the cell (Figure 1A,B), and its more than 100 verified partners can be found in many cellular compartments such as the nucleus, mitochondira, and vesicles (Figure 1C). Intriguingly, LC8 does not passively bind its various partners, but emerging evidence instead suggests that LC8 acts as a dimerization engine that promotes self-association and/or higher order organization of its primarily disordered, monomeric partners. This structural organization process is triggered by long-range allosteric regulation initiated by LC8 binding a pair of disordered chains, forming a bivalent or polybivalent scaffold.

Figure 1. LC8 and its binding partners display broad cellular localization. (A) Live HeLa cells stably expressing LC8-GFP (green) were transiently transfected with the focal adhesion marker TagBFP-paxillin (blue) and the nucleus marker dsRed-tagged histone H2B (red). The top view images shown on the left represent an optical section located next to the coverslip. LC8-GFP is present throughout the cell, but forms puncta at the cell cortex. (B) HeLa cells stably expressing LC8-GFP were fixed with paraformaldehyde (PFA) and stained for the endogenous α-tubulin (red) and with DAPI (blue) to visualize the DNA. LC8 accumulates at the kinetochores (box 1) and at the spindle poles (box 2). (C) Localization information derived from the COMPARTMENTS program demonstrates that LC8 binding partners are localized to all cellular compartments.

We have established novel properties of LC8-partner protein complexes as follows. LC8 binding partners described thus far are invariably intrinsically disordered proteins (IDPs). When bound to LC8, these IDPs always form a complex containing two IDP chains. The active form of LC8 is a homodimer that has, in each monomer, a binding groove specific for an 8-10 residue consensus sequence on an IDP. The LC8/IDP duplex creates an elongated and flexible scaffold that is at the core of large and diverse macromolecular assemblies. Structurally, LC8/IDP scaffolds have three salient properties. i) The duplex attaches additional bivalent partner-proteins, often with enhanced affinity relative to their binding to a single IDP chain. ii) LC8/IDP scaffolds retain a high degree of structural disorder that is important to function. iii) In the duplex, the two IDP chains often self-associate to form new functional domains. One example of these properties can be seen in the eponymous Dynein-LC8 complex (Figure 2).

Figure 2. A model of LC8/IC duplex showing coupling between self-association of dynein intermediate chain IC and LC8 binding.The linker connecting the LC8 binding site (yellow) to the self-association domain (red) remains fully disordered.

An Anchored Flexibility Model explains the ability of LC8 to accommodate a large variety of motif sequences. LC8 recognition is based on fixed interactions between LC8 and TQT that act as an anchor, coupled with inherently flexible interactions between LC8 and non-TQT residues (Figure 3A). The TQT triplets of the 8-10 amino acid recognition motifs are the anchor residues because they are least accessible to solvent and most rigid (Figure 3B). As the LC8/Chica motif complexes illustrate (Figure 4), the ‘anchor’ region of the LC8 binding site engages peptide TQT residues in a network of polar interactions involving buried H-bond donors and acceptors in both LC8 subunits. Such a buried network requires optimal H-bonding geometry, the distortion of which is energetically costly, especially in an apolar milieu. The TQT region of the LC8-motif complex is highly ordered and rigidly packed (anchored) compared to the other more flexible motif positions (Figure 3B). The plasticity of the non-TQT regions of the complex is reflected in the conformational accommodation of numerous sequence permutations with similarly variable affinities (Figure 5). The flexibility resides in non-TQT residues and LC8 residues that bind them, while the anchor is composed of TQT residues and LC8 residues 35’, 36’, and 62-64 (Figure 4). It is likely that these features underlie the evolutionary fitness of LC8-motif interactions.

Figure 3. Properties of the LC8/peptide interface. (A) Sequence logo of LC8 binding motifs derived from sequences of the motifs in the 11 crystal structures reported for LC8/peptide complexes. Height of amino acids indicates their relative frequency at that position. Analysis of all known crystal structures show that the TQT residues are buried and we propose that they function as the buried anchor in an ‘anchored flexibility’ model of TQT motif binding. The TQT “anchor” has clearly complete burial of side chains in the LC8 binding pocket at positions -1 to 1, even for non threonine side chains. Solvent accessibility and B factors for side chains at other positions vary considerably, demonstrating the higher flexibility of the non-TQT regions of the peptide binding groove. (B) Solvent accessible surface area depiction of an example LC8/peptide pair. Color scheme was defined at the atom level using the GetArea program, with magenta representing more solvent exposed, and orange regions more buried atoms.

Figure 4.Comparisons of crystal structures of the LC8-QT1p and LC8-QT4p complexes.(A) Chica peptides QT1p (orange) and QT4p (green) overlaid after alignment of the mainchain residues of their respective LC8 homodimers. The side chains of LC8 residues that form hydrogen bonds with the motif peptides are shown as sticks and labeled. For the sake of clarity, only one peptide bound to LC8 is shown. A schematic of (B) QT1p (orange) bound to LC8 compared to (C) QT4p (green) bound to LC8 highlights similarities and differences in LC8 residues involved in the hydrogen bonding or stabilizing interactions based upon the peptide sequence. The sequence of each peptide is shown above each diagram with capital letters corresponding to residues that are observed in the crystal structure, while those in small letters are part of the peptide but are not observed and underlined residues are nonnative.

While it is not clear at this stage why some LC8 partner proteins are multivalent, it is clear that they are numerous and biologically diverse, and that their LC8-recognition sites are present exclusively in intrinsically disordered segments of the partner. For some LC8 partners with multiple binding sites, various functions of the multiplicity have been proposed (Figure 6). In Nup159, the multiple sites act cooperatively so that binding of the first LC8 enhances binding of the next, and apparently contributes to the stability of the complex.

Figure 5. Analysis of LC8 binding and nonbinding motifs reveals distinct positional preferences. (A) Motif preferences for LC8 binding partners. “Ф” denotes hydrophobic residues; “X” signifies any residue (unless certain residues are disfavored); underlined “X” signifies any residue but with strong preferences for particular residues; “+” denotes positively charged amino acids. Physiochemical properties beneficial for binding are colored dark blue or light blue, based on magnitude, and deleterious properties are colored in red. (B) All known tightly binding sequences (Kd < 10 μM) are cropped to eight amino acid motifs, and built using the Chimera molecular modeling software. This includes LC8 sequences found on the LC8Hub database, as well as those determined in this paper.

Electron microscopic studies of nuclear pore assembly suggest that incorporation of Nup159 into the nuclear pore does not occur until all LC8 sites are filled. In ASCIZ, the multiple motifs could act as a sensor for LC8 cellular concentration. ASCIZ is an LC8 transcription factor, and it is possible that the level of bound LC8 could modulate its LC8-dependent transcription activity. In Ana2, binding of two LC8 dimers promotes self-association of Ana2 into a tetramer, possibly by aiding formation of a coiled-coil region located between the two binding sites. In Pac11, two LC8 sites flank a nascent helix which forms a self-associated helix/helix interface upon LC8 binding.

Figure 6. Structural models of LC8/Swallow, LC8/Ana2, LC8/DIC and LC8/Nup159. LC8 dimer is shown in green and disordered partners in grey and red. LC8 binds intrinsically disordered partners (IDPs) and forms duplex scaffolds with unique features, namely: parallel alignment of two IDP chains cross-linked by reversible, non- covalent protein-protein interactions; formation of newly ordered, self-associated domains resulting from interactions of identical sequences within each chain; and enhancement, relative to the same IDP in monomeric form, of binding affinity for additional bivalent ligands and of self-association tendency. With Swallow, binding of LC8 stabilizes a B-strand at the LC8 binding interface (yellow) and promotes self- association of a dimeric coiled-coil domain about 20 amino acids distant from binding. With Ana2, binding of two LC8 dimers promotes self-aggregation of Ana2 into a tetramer, possibly by aiding formation of a coiled-coil region located between the two binding sites. With DIC, binding of LC8 promotes formation of self-association domain (red) 100 amino acids distant from binding. With Nup159, binding of LC8 bind at five recognition motifs in tandem forms a rigid duplex. The multiple sites act cooperatively so that binding of the first LC8 enhances binding of the next, with three sites as sufficient for tight binding. Addition of the last two is proposed to contribute to formation of a rigid rod that exposes the Nup FG-repeats into the cytoplasmic side of the complex. Electron microscopic studies of nuclear pore assembly suggest that incorporation of Nup159 into the nuclear pore does not occur until all LC8 sites are filled.