Nội dung

MINIREVIEW DYNLL/LC8: a light chain subunit of the dynein motor complex and beyond Péter Rapali, Áron Szenes, László Radnai, Anita Bakos, Gábor Pál and László Nyitray Department of Biochemistry, Eötvös Loránd University, Budapest, Hungary Keywords dynein; hub protein; intracellular transport; linear motif; protein–protein interactions Correspondence L. Nyitray, Department of Biochemistry, Eötvös Loránd University, Pázmány P.s. 1 ⁄ c, H-1117 Budapest, Hungary Fax: +36 381 2172 Tel: +36 1381 2171 E-mail: nyitray@elte.hu (Received 8 February 2011, revised 11 May 2011, accepted 15 June 2011) The LC8 family members of dynein light chains (DYNLL1 and DYNLL2 in vertebrates) are highly conserved ubiquitous eukaryotic homodimer proteins that interact, besides dynein and myosin 5a motor proteins, with a large (and still incomplete) number of proteins involved in diverse biological functions. Despite an earlier suggestion that LC8 light chains function as cargo adapters of the above molecular motors, they are now recognized as regulatory hub proteins that interact with short linear motifs located in intrinsically disordered protein segments. The most prominent LC8 function is to promote dimerization of their binding partners that are often scaffold proteins of various complexes, including the intermediate chains of the dynein motor complex. Structural and functional aspects of this intriguing hub protein will be highlighted in this minireview. doi:10.1111/j.1742-4658.2011.08254.x Introduction Dynein light chains (molecular mass 10–20 kDa) are accessory subunits of the large dynein motor complexes. The LC8 family of light chains (DYNLL1 and DYNLL2 in vertebrates; abbreviated here as LC8), together with the Tctex (DYNLT1 and DYNLT3) and LC7 (Roadblock; DYNLRB1, DYNLRB2) light chains, bind as homodimers to the dimeric cytoplasmic dynein intermediate chains (DYNC1I1, DYNC1I2; abbreviated here as DIC), which are scaffold subunits for cargo binding to the motor complex (for recent reviews see [1–3]; a schematic view of the subunit structure of dynein is shown in this minireview series [3]). LC8 was first described as a subunit of Chlamydomonas axonemal dynein [4], and was subsequently found to bind to all cytoplasmic and most axonemal dyneins [2]. The LC8 genes are present in all sequenced eukaryotic genomes [1] and code for an extremely conserved 10 kDa protein. The Chlamydomonas, Caenorhabditis elegans, Drosophila and mammalian LC8 orthologs share more than 90% identity. The two mammalian paralogs DYNLL1 and DYNLL2 differ only in six out of 89 residues, and they are fully conserved as orthologs. Based on genetic studies, it is clear that at least in metazoans LC8 is an essential protein; knocking out or knocking down of its gene in Drosophila and C. elegans either is embryonically lethal or causes severe pleiotropic phenotypes [5,6]. Since LC8 was identified as tail-binding light chain of not only dyneins but also of myosin 5a [1,7], and was also found to interact with many proteins that Abbreviations DIC, dynein intermediate chain; DYNLL ⁄ LC8, dynein light chain LC8; ERa, estrogen receptor a; HTP, high-throughput; MYO5A, myosin 5a heavy chain; nNOS, neuronal nitric oxide synthase; Pak1, p21-activated kinase; PSD, postsynaptic density. 2980 FEBS Journal 278 (2011) 2980–2996 ª 2011 The Authors Journal compilation ª 2011 FEBS DYNLL ⁄ LC8 dynein light chain P. Rapali et al. were shown to be transported either on microtubules or on the actin filaments, it was widely assumed that LC8 could function as a cargo adapter being simultaneously associated with both the motor and the cargo protein [8–13]. However, recent structural and thermodynamic studies challenged this hypothesis: the two identical binding sites of LC8 and the homodimeric nature of both DIC and myosin 5a heavy chain (MYO5A) make it unlikely that LC8 can bridge the cargo to either motor complex [14–17]. The facts that LC8 interacts with proteins that are not associated with intracellular transport and that it is present in plants that are entirely devoid of dynein motors [2] point to a more general role of LC8. Recently, it has been recognized by analyzing the sequences of LC8 interaction partners that the short linear LC8 binding motifs are located in intrinsically disordered protein segments [18,19]. Moreover, the LC8 binding motif is often located close to coiled-coil or other dimerization domains of the interacting partners. Accordingly, the current view is that LC8 is an essential hub protein that functions as a ‘molecular velcro’: it promotes dimerization and structural stabilization and hence it could allosterically regulate its binding partners in diverse protein complexes and networks, the dynein motor complex being only one of them [18]. A Thr1 Gln0 Thr-1 Arg-3 B C Structure of DYNLL ⁄ LC8 alone and in various complexes The solution and crystal structures of LC8 in apo form [8,20–23] and in complex with peptides from six binding partners [DIC, neuronal nitric oxide synthase (nNOS), Bim, Swallow, p21-activated kinase (Pak1), EML3] have been determined [8,14,22,24–26]. Moreover, models with three additional binding peptides have recently been published [27,28]. LC8 has a unique fold (Fig. 1A, B): two-five-stranded, antiparallel b-sheets are responsible for dimerization; each b-sheet contains four strands from one monomer and a fifth strand from the other monomer. These sheets are flanked by two pairs of a-helices at the opposite faces of the dimer. Interestingly, the Tctex ⁄ DYNLT light chain is a structural homolog of LC8 ⁄ DYNLL with no apparent sequence similarity [29,30]. Despite their structural similarity there is no overlap in known targets of the two light chains [1]. The bound ligands of LC8 lie in two identical parallel grooves formed at the two edges of the dimerization interface. The bound peptides form an extra antiparallel b-strand and therefore augment the central b-sheets [8,14,22,24–26] (Fig. 1A, B). Practically all non-identical residues of LC8 paralogs and orthologs in metazoans are located Fig. 1. (A) Structure of DYNLL ⁄ LC8 complexed with a peptide ligand. Crystal structure of human DYNLL2 in complex with the binding motif of EML3 at 1.3 Å resolution (PDB ID: 2XQQ) [26]. DYNLL2 is a homodimer with the monomers related by a twofold axis (cyan and blue). Two peptides (orange) corresponding to the binding motif of EML3 lie in the binding grooves formed at the edge of the dimerization interface of DYNLL2. Side-chains with a key role in the interaction (Arg)3, Thr)1, Gln0, Thr1) are highlighted. (B) Two five-stranded b-sheets are formed at the dimerization interface, each containing four strands from one monomer and the fifth from the other. Residues not fully conserved in LC8 paralogs and orthologs are shown in red. Most of the natural diversity clearly occurs in a-helices and loops far from the ligand, while the binding grooves are highly conserved. The bound peptides augment the fivestranded b-sheets by an antiparallel b-strand. (C) Sequence logo of the binding motif of all hitherto determined and verified LC8 binding motifs. Similar colors indicate similar chemical properties [135]. FEBS Journal 278 (2011) 2980–2996 ª 2011 The Authors Journal compilation ª 2011 FEBS 2981 DYNLL ⁄ LC8 dynein light chain P. Rapali et al. on the outer surface of the homodimer protein and are not involved in binding of target proteins (Fig. 1B). LC8 binding motifs were originally divided into two classes: (K)3X)2T)1Q0T1X2) and [X)3G)2(I ⁄ V))1 Q0V1D2] [8,12,31]. In both classes the central Gln (position 0) caps the N-terminal end of the second a-helix, while the side-chains of residues at positions +1, )1 and )3 interact with the interior of the binding groove. A few LC8 partners contain non-canonical binding motifs that lack the most conserved Gln residue (e.g. MYO5A [32,33], Pak1 [22], GRINL1A [27]), but the overall binding topology of these peptides is similar to that of the canonical ones. In the DYNLL1–Pak1 complex a specific H-bond network compensates for lack of the conserved Gln [22], and similar compensatory interactions are seen in the docking model of DYNLL1–GRINL1A [27] and the very recently determined crystal structure of DYNLL2–MYO5A (L. Radnai, P. Rapali and L. Nyitray, unpublished results). How can a highly conserved binding site promiscuously interact with such a diverse motif set? Available data indicate that conformation of the binding grooves of LC8 is highly dynamic. Binding groove residues of apo-LC8 show conformational exchanges on multiple time scales in NMR spectroscopic experiments, while ligand binding reduces the conformational breathing of these regions [34,35]. Evidence for structural plasticity can also be obtained by comparing crystal structures of apo-LC8 and LC8 complexes [20]. Ligand binding is coupled to a slight opening of the grooves and in this way LC8 can accommodate to different interaction patterns [20]. Finally our detailed kinetic studies on ligand binding of LC8 with different partners also provided evidence of conformational changes required for complex formation. The biphasic binding transients, observed by using a stopped-flow method, can best be described by a conformational selection model [36]. The affinity of LC8 to several monomeric binding peptides was determined and found to be moderately weak (Kd between 0.1 and 40 lm) (Table 1). However, most identified LC8 partners are known or predicted to be dimeric (Tables 1 and 2). These interacting proteins, as bivalent ligands, can form dimer-to-dimer complexes with LC8. Bivalent interactions are known to produce significant gains in binding affinity, specificity and functionality due to the avidity effect [37]. Indeed, compared with the monomeric peptides, a two to three order of magnitude increase in the apparent affinity was measured with bivalent DIC, MYO5A and an artificially dimerized target peptide of LC8 [15,26,29]. DIC is a poly-bivalent scaffold for binding of the three classes of dimeric dynein light chains; the implication of poly-bivalency on dynein function will 2982 be discussed later. Interaction partners that contain tandem LC8 binding motifs (p53BP1, Nup159, GKAP, Bassoon, U19, ATMIN; Table 1) are also poly-bivalent ligands. Hitherto only a few 3D structures of dimer-to-dimer LC complexes have been published: a short DIC fragment containing the binding sites for LC8 and Tctex1 in complex with these two light chains [15,29], and LC8 in complex with an artificially dimerized binding motif of EML3 [26]. The dimerization constant of Drosophila LC8 was reported to be moderately weak [38]; however, a more careful measurement showed higher affinity (Kd  200 nm) indicating that under cellular conditions the LC8 pool is mostly dimeric [22]. Nevertheless, at low pH (due to protonation of His55 at the dimerization interface [38–41]) or by phosphorylation of Ser88, the dimers readily dissociate to stable monomers [42,43]. The structure of the monomers, which are unable to bind target proteins, has also been determined [39,41]. As expected, tight binders shift the apparent equilibrium back towards the dimers and hence form complexes even with the Ser88Glu phosphomimetic mutants [24,36]. Such a monomer–dimer transition could have regulatory roles in the interaction network of LC8. Originally, LC8 was identified as a substrate of Pak1 and LC8 phosphorylation on Ser88 was implicated in cancer development, metastasis and triggering macropinocytosis [44,45]; however, recent studies ruled out that Pak1 is able to phosphorylate LC8 [22,36]. Consequently protein kinases involved in the potential dimer–monomer regulatory switch have not yet been identified. Instead of being a substrate, LC8 seems to be involved in the nuclear translocation of Pak1 [46] a function shared with several other LC8 partners (see below). Alternative ways of regulating LC8–target complexes could be the phosphorylation of LC8 binding motif or LC8 binding groove residues. Indeed, binding of the pro-apoptotic Bim is abolished upon JNK-kinase phosphorylation of its LC8 binding motif [47]. Phosphorylation of the LC8 residue Tyr65 by a hitherto unknown kinase [48] could also interfere with target binding. What are the structural consequences of LC8 binding to the target proteins? Characterization of LC8 complexes indicated that LC8 binding facilitates the folding and increases the a-helical content of DIC, Swallow, MYO5A and synthaphilin [32,33,49–51]. The stabilized coiled-coils could provide additional binding platforms in various complexes (Fig. 2A). This structure ⁄ folding promoting ‘chaperon-like’ activity is consistent with the high percentage of potential coiledcoil forming sequences near the LC8 binding motifs (Tables 1 and 2) and could be one of the major FEBS Journal 278 (2011) 2980–2996 ª 2011 The Authors Journal compilation ª 2011 FEBS Organism Adenovirus Human, rat Human ASF virus Rat Human Human Human Fruit fly Mouse Rat Human Fruit fly Human Rat Human, rat Human Papilloma virus Rat Rat Human Rat Caenorhabditis elegans Human Human Human Human Protein name Adenain (ADE41) AIBC1 (BCAS1) ATMINa p54 (E183L) Bassoon (Bsn) BimEL (BCL2L11) Bmf BS69 (ZMYND11) DIC (sw) DIC1 (Dyncli1) DIC2 (Dyncli2) DNMT3A Egalitarian EML3 Gephyrin (Gphn) GKAP (DLGAP1) FEBS Journal 278 (2011) 2980–2996 ª 2011 The Authors Journal compilation ª 2011 FEBS Grinl1A (GCOM1) E4 Hsc73 (Hspa8) KID-1 (Znf354a) MAP4 Mark3 METT-10 Myosin Va (MYO5A) NEK9 nNOS (NOS1) NRF1 Q8TD19 P29475 Q16656 Q9Y4I1 P0CAP1 P06425 P63018 Q02975 P27816 Q8VHF0 Q09357 Q03555 O14490 ⁄ P97836 O88485 Q62871 Q9Y6K1 P92030 Q32P44 Q15326 Q24246 O43521 Q96LC9 P11826 O75363 ⁄ Q3ZB98 O43313 Q4TWM2 O88778 Uniprot DYNLL1 DYNLL1 DYNLL1 DYNLL2 VGMHSKGTQTA AEMKDTGIQVD MEEHGVTQTE QPKDDKNTMTD KQTEDKGVQCE NRCLSIGIQVD SKFQSVGVQVE TEVETREIGVG DHHQDKQTQTP TTIPTKQTQTF SHRTTKSTQTQ SRSGSKSTQTV VVAYPKRSQTS LNAWDNASQAY DYNLL1 ⁄ 2 DYNLL2 DYNLL1 DYNLL1 DYNLL1 DYNLL1 DYNLL1 DYNLL1 dlc-1 VVSYSKETQTP IVTYTKETQTP LVLKDLGIQVD VKLVDAESQTL PSLVSRGTQTE PRMLHRSTQTT TLVYTKQTQTT CITLVKSTQTV KRMLDAQVQTD LESLDIETQTD VTTQNTASQTM PTTANYGSQTE SPMVAQGTQTP RATAEFSTQTP PMSCDKSTQTP TSQEDKATQTL Sequence DYNLL1 DYNLL1 DYNLL1 ddlc1 DYNLL1 ⁄ 2 DYNLL1 ddlc1 DYNLL1 DYNLL2 DYNLL1 DYNLL1 DYNLL1 DYNLL1 DYNLL1 ⁄ 2 Paralog ⁄ ortholog Table 1. DYNLL ⁄ LC8 interaction partners with verified binding motifs. 940 226 1 1281 216 672 647 423 18 418 94 797 429 416 146 153 648 947 78 408 125 104 563 665 139 1423 1527 1499 107 63 First residue 7 8.8 0.04d 2.4 0.1 0.05d 8 0.2d 0.8 0.7 0.004d 1.7 Kd (lM) 1F96, 1CMI 2XQQ, 3P8M 2P2T, 3FM7, 2PG1 1F95 PDB I I I I I I I I I I I I D, I D, Ic D, I D, Ic D, D, D, D, D, – D, D, D, D D, D, – D, D, C E C Cc C – C – – – E [26,61,115] [12,36,58,116] [58,117] [32,33,36] [28] [110] [62] [12,62] [12] [62] [73] [12,28,58,85,87] [12,58,88,89] [31,58] [99] [12,62] [11] [26] Cc C – C C C C [31] [14,29,51] [13,31,58] [12,36,114] [28,113] [70] b [110] [31,111,112] Reference C Cc – – D, Ic D, I D, I D, Ic – C – – C CC ⁄ dimer I I I I – D, D, D, D, Disorder P. Rapali et al. DYNLL ⁄ LC8 dynein light chain 2983 2984 P10193 Q8K3I7 P40688 B5DF41 Q9UHF7 Q23064 Q05127 Q13153 P15198 O56780 P63244 Q8IV61 Q69501 Q12888 P40477 Uniprot DYNLL1 DYNLL1 ddlc1 DYNLL1 DYNLL1 dlc-1 DYNLL1 DYNLL1 ⁄ 2 DYNLL1 DYNLL1 DYNLL1 DYNLL1 DYNLL1 DYNLL1 DYN2 Paralog ⁄ ortholog SASADFDVQTS DNYAESGIQTD VETCNFSVQTF IPVKHNSTQTV KEAVDNGLQTE PSQNNIGIQTM ETVVSAATQTI TPTRDVATSPI RSSEDKSTQTT KSTEDKSTQTP LGVCKYTVQDE RATTSQATQTE TILVSRSTQTG LGHFTRSTQTS GVQMAKSTQTF QDVLRRTVQTR SATSAKATQTD SCMQERAIQTD TEKVDRSTQDE DSISDRHIQTM PKTRNSQTQTD Sequence 1102 1115 1164 1140 1152 1147 1164 212 142 139 135 607 225 237 744 555 286 307 1202 643 66 First residue 1 42 4.5 Kd (lM) 3E2B 3DVT PDB I I I I I I I I I Ic I I D, I D D D D D, I D, I D, D, D, D, D, D, D, D, D, D, – D, D, Disorder C C Cc Cc E C C E – C – – C C C CC ⁄ dimer [110] [62] [12,14,16,67,68] [49] [61,92] [103] [122] [22,36,44] [12,118,119] [12,119] [120] [121] [110] [10,61] [71] Reference ATMIN has at least five verified motifs. b Unpublished (P. Rapali, L. Nyitray and I. Rodriguez-Nacho). c Experimentally verified disordered or coiled-coil region. d Determined with a bivalent ligand. D, I, Binding motif in predicted disorder region determined by DISPROT and IUPRED [132,133]. C, Predicted coiled-coil by COILS [134]. E, Experimentally determined dimer, trimer or oligomer. a HHV-1 Mouse Fruit fly Rat Human C. elegans Ebolavirus UL9 Spice1 Swallow Syntaphilin (Snph) TRPS1 UNC-83 VP35 Human p53BP1 (TP53BP1) Human Rabies virus Mokola virus Human Human HHV-7 Yeast NUP159 PAK1 P protein P protein RACK1 (GNB2L1) RASGRP3 U19 Organism Protein name Table 1. (Continued). DYNLL ⁄ LC8 dynein light chain P. Rapali et al. FEBS Journal 278 (2011) 2980–2996 ª 2011 The Authors Journal compilation ª 2011 FEBS DYNLL ⁄ LC8 dynein light chain P. Rapali et al. Table 2. DYNLL ⁄ LC8 interaction partners with unknown or predicted binding motifs. D, I, Binding motif in predicted disorder region determined by DISPROT and IUPRED [132,133]. C, Predicted coiled-coil by COILS [134]. E, Experimentally determined dimer, trimer or oligomer. Protein name Organism Uniprot Paralog ⁄ ortholog Sequence First residue Disorder CC ⁄ dimer References AMBRA1 Astrin (SPAG5) C20orf117 p26CP CIZ1 Dazl ERa (ESR1) Ewg CHICA (FAM83D) Gag Gag GLCCI1 gurken mRNA IjBa (NFKBIA) KIAA0802 KIBRA MORC3 NudE Phototropin PKIa PTH mRNA RSP3 Human Human Human BIV Human Mouse Human Fruit fly Human HFV Rice Sirevirus Human Fruit fly Human Human Human Human Mouse Vicia Human Human Chlamy.s Q9C0C7 Q96R06 O94964 P19558 Q9ULV3 Q64368 P03372 Q24312 Q9H4H8 P14349 Q109L4 Q86VQ1 TSVTSQGTQTL PETQDSSTQTD VGLASVGTQTI VQLIESKVREKENA KLQKQAQTQTS KKSVDRSIQTV 1097 467 1135 214 340 245 P12759 DYNLL1 DYNLL1 ⁄ 2 DYNLL1 DYNLL DYNLL1 DYNLL1 DYNLL1 ddlc1 DYNLL1 DYNLL1 LC8 DYNLL1 ddlc1 DYNLL1 DYNLL1 DYNLL1 DYNLL1 DYNLL1 LC8 DYNLL1 DYNLL1 LC8 – C C – C E E E – – C C n⁄a – C C C Ca – – n⁄a C [26,61] [61,75] [26,61] [123] [61,72] [9] [93] [117] [26,61] [124] [125] [26,61] [102] [81,82,117] [26,61] [94] [26,61] [104–106] [126] [127] [128] [129] Spn-F TPR UHRF1BP1L ZMYM3 Fruit fly Human Human Human Q9V9Y9 P12270 A0JNW5 Q14202 ddlc1 DYNLL1 ⁄ 2 DYNLL1 DYNLL1 D, I D, I D, I D, I D, I D Da D, I D, I D, I D D, I n⁄a D, Ia D, I D D, I D, I – D, Ia n⁄a D, I D, I D, I – D, I D, I D, I C C – [130,131] [61,77] [26,61] [26,61] a P25963 Q9Y4B5 Q8IX03 Q14149 Q9CZA6 Q8H934 P61925 LSDVDYTTQTV LSVSEVGTQTS 536 402 IDVGISCDLLD SSTRSIDTQTP n⁄a 467 343 NGSRTMGTQTV KQYLDVSSQTD DQGNTAATQTE 1622 275 633 n⁄a KYREDETTQTL LPADATQTQTK VPEADTSTQTD 26 58 112 TATVMPTTQVE QRSVTQATQTS VEMKSKGSQTE 1714 1392 848 Experimentally verified disordered or coiled-coil region. physiological functions of LC8 [18,19]. Because of the mutual avidity effect, LC8 molecules inside the cell are expected to be bound to their numerous and (mostly) homodimeric partners rather than being in free, uncomplexed state. The molecular glue function of LC8 does not necessarily induce structural stabilization of the target protein; if LC binding causes spatial constraints in the target, it could lead to dissociation of pre-existing dimeric domains or destabilization of binding platforms preventing additional interactions, and ⁄ or could inhibit enzymatic or other activities of the partner proteins (Fig. 2B). LC8-induced dissociation of a binding platform has not been described yet. Besides stabilization or destabilization of homodimeric partners we envisage one situation in which an LC8 dimer can bind two different ligands and cause heterodimerization: if the two targets have an independent (even weak) interaction domain ⁄ motif near the LC8 binding motifs (Fig. 2C). Hitherto no such LC8 interaction or complex has been described. The DYNLL/LC8 binding peptide is a short linear motif LC8 binds to a loose consensus sequence, a short linear motif. Such motifs are usually localized in disordered segments and the implied plasticity and versatility in their molecular recognition is a clear advantage in eukaryotic interactomes [52–54]. These properties must also contribute to the promiscuous interaction network around LC8. The segment containing the binding motif in DIC was experimentally shown to be disordered by biochemical and NMR studies [34,51,55], and the same results was obtained with a MYO5A fragment [32]. Bim and Bmf, two LC8 binding pro-apoptotic proteins, are intrinsically disordered along almost their entire length [56]. Among the LC8 interactors, disprot database [57] also lists estrogen receptor a (ERa), IjBa and PKIa as proteins that have experimentally verified intrinsically disordered domains. FEBS Journal 278 (2011) 2980–2996 ª 2011 The Authors Journal compilation ª 2011 FEBS 2985 DYNLL ⁄ LC8 dynein light chain P. Rapali et al. A B C D E Fig. 2. (A) Possible interaction modes of DYNLL ⁄ LC8 with its targets. Interaction of LC8 with a partner that contains a potential or preformed coiled-coil domain near the LC8 motif could lead to homodimerization or coiled-coil stabilization. This is the only experimentally proved model of LC8 complex formation. The newly formed structure could act as a platform for further interactions. (B) If the LC8 binding motif is localized near interacting globular domains, LC8 binding could pry apart the domains by steric constraints and might destroy further interaction sites or inhibit other activities. (C) The same destabilizing effect may occur if the LC8 binding site is located within a coiled-coil domain (not shown). Heterodimerization of two targets could occur if two LC8 binding motifs are located near two weakly interacting domains. (D) Heterodimeric coiled-coils could also form by this mechanism (not shown). LC8 could function as a direct cargo adapter on dynein if one assumes that two homodimeric LC8–target complexes interact via their ligands. (E) Such an interaction between two ligandbound LC8 complexes via antiparallel b-strands of the ligands has been observed hitherto only as a crystal contact in the crystal structure of the LC8–EML3 complex. Subunits of LC8 are colored as in Fig. 1; the interacting EML3 peptide ligands are brown and green (PDB 3P8M). There are additional although less likely theoretical interaction modes that are not depicted here. Close inspection of the most recent list of validated binding peptides (Table 1) indicates that several motifs represent a mix of residues from the previously described two motif classes (see above). It suggests that even though the two putative motif types show some distinctive thermodynamic differences [14,36] their physiological significance is rather questionable. A 2986 sequence logo visualization of all known binding motifs (50 in 41 proteins) shows that the most frequently occurring and therefore likely key binding determinants of the motif are the most conserved Gln0, the flanking Thr+1 and Thr)1 residues and Asp)4 (Fig. 1C). We depict 11 residues both on the logo and in Tables 1 and 2 since the binding site on FEBS Journal 278 (2011) 2980–2996 ª 2011 The Authors Journal compilation ª 2011 FEBS DYNLL ⁄ LC8 dynein light chain P. Rapali et al. Fig. 3. (A) Partial interaction networks of selected DYNLL ⁄ LC8 binding partners. Functional categories that involve at least five LC8 ⁄ DYNLL interactors are shown with different colors. More details of the interactions, including a full reference list, is found in Tables 1 and 2. (B) Egalitarian (Egl, in Drosophila), NudE (in mammals) ⁄ NUD-2 (in Caenorhabditis elegans), UNC-83 (in C. elegans) were originally suggested to function as dynein regulatory proteins associated to the motor complex via LC8 bound to the DIC (DYNC1I). An alternative scenario is shown here. The LC8 orthologs do not directly link the above proteins to the motor complex; instead the dimerizing ⁄ stabilizing effect of LC8 allows the scaffold proteins to bind to different adaptors ⁄ regulators (e.g. NudE to LIS1), to dynein subunits (NudE to DIC) or to the dynactin complex (Egl and UNC-83 to the dynamitin subunit of dynactin) via separate domains ⁄ linear motifs. UNC-84 is a membrane protein attaching the dynein motor to the nucleus for transport in C. elegans in complex with DLC-1 ⁄ UNC-83 ⁄ Bicaudal D and the C. elegans ortholog of Egl. The Dlc ⁄ Egl ⁄ BicD complex is responsible for effective mRNA transport in Drosophila; Egl is a direct mRNA binding protein [101]; Dlc may also bind specifically gurken mRNAs [102]. p53BP1 was shown to associate with the dynein motor complex in dyneinmediated p53 translocation. Based on experimental results the cargo adapter for p53 is not LC8 but the Hsp90–immunophilin complex that is associated to dynactin. Only experimentally verified physical interactions are shown. LC8 could accommodate up to 11 residues. Nevertheless, only seven residues form the core binding motif as noted earlier [58]. The above mentioned four residues were found critical for binding in mutational studies, structure comparisons and a semi-quantitative pepscan analysis [8,22,58]. In Table 1 we also list all affinity values (measured with monovalent and ⁄ or bivalent peptide ligands) that have been determined hitherto. The only clear correlation between binding strength and sequence is that lack of Gln0 decreases the overall affinity of the peptide. A recent highthroughput (HTP) in silico analysis of the contribution of key residues and their context to the global binding energy of all known eukaryotic linear motifs including the LC8 recognition motif also highlighted the importance of the central three residues plus Lys)3 (instead of Asp)4) as key binding determinants for LC8 binding. The so-called ‘contextual residues’ of short linear motifs, in general and in LC8 complexes as well, seem to contribute more to the binding specificity than to the binding affinity of the interaction [59]. The same tendency was observed in a pepscan analysis of LC8 binding motifs [58]. Recently, we have conducted the first comprehensive and quantitative experimental analysis of the binding motif preference of LC8 using a directed evolution approach, phage display. The in vitro evolved and thermodynamically selected pattern (from 109 sequences) resembles the natural one except at positions )4 and )5. Directed evolution identified position )5 as a significant contributor to the binding energy. A monovalent form of the consensus peptide (based on 25 selected individual sequences) binds to LC8 with a Kd of 0.08 lm (which is an affinity an order of magnitude higher than the previously known tightest binding Bmf peptide) and in a bivalent format with sub-nanomolar dissociation constant. Interestingly, the selected consensus is present in EML3, a human microtubule binding protein involved in mitosis. The crystal structure of the LC8–EML3 peptide revealed how the affinity-enhancing Val)5 is accommodated in a shallow binding pocket on LC8 [26]. DYNLL/LC8 as a hub protein: an ever-increasing interaction network A thorough literature mining revealed 66 proteins and two mRNAs that are reported to interact with LC8 with relatively high confidence (Tables 1 and 2). In 42 of these proteins 55 LC8 binding motifs were identified and verified by small-scale experimental methods (Table 1). The very recently identified LC8 interacting protein ATMIN has at least five binding motifs. However, they have not been unequivocally assigned to the FEBS Journal 278 (2011) 2980–2996 ª 2011 The Authors Journal compilation ª 2011 FEBS 2987 DYNLL ⁄ LC8 dynein light chain P. Rapali et al. potential 11 consensus-like motifs (P. Rapali, L. Nyitray and I. Rodriguez-Nacho, unpublished results), and therefore only one representative site is shown in Table 1. In the rest of the LC8 partners the binding motif is either unknown (eight proteins, e.g. ERa, IjBa, NudE) or was predicted by us based on the consensus motif. We did not include the more than 400 additional proteins that were deposited in protein–protein interaction databases (available through the intact database [60]). Either these interactions were collected by low confidence HTP methods or the experiments did not prove that the interaction is direct, i.e. binary. We made exceptions only with eight proteins that were identified as LC8 binding partners by an innovative HTP approach providing low false-negative and false-positive detection rates [61], because these were also identified by our very recent high-confidence in vitro evolution based prediction method [26]. The latter prediction revealed at least 100 additional novel LC8 binding partners in the human proteome (not included here). Each contains a binding sequence in the intrinsically disordered region of the putative interactor [26]. Further studies are needed to verify that these are indeed genuine components of the LC8 interaction hub. The known binding motifs are almost exclusively (94%) located within disordered regions of the LC8 partner protein, as judged by two independent disorder predictors. In Table 2 only those predicted motifs are shown that are located in intrinsically disordered regions of the partner protein. The majority of the LC8 partner proteins contain coiled-coil predicted sequences in close proximity to the known or putative binding motif (78%) or it was shown experimentally that they form dimers or higher oligomers (in nine proteins). The above two characteristics are intimately associated with the LC8 interaction network as previously proposed based on a much smaller set of LC8 binding proteins [18,19]. There are only two reported exceptions in which the LC8 binding linear motif is more probably located in an ordered domain of the interactor. Hsc73 was identified in rat brain lysate by a low-throughput proteomic method [62]. The KXTQT-type binding motif is located in a well-defined 3D structure; the surface exposed b-strand might fit into the binding pocket of LC8 [63]. RACK1 was identified as an LC8 target by the yeast two-hybrid method. It forms a ternary complex with LC8 and BimEL upon apoptotic stimulation [64] and the putative LC8 binding motif is located in a modeled b-propeller WD domain of RACK1 [65]. DNMT3A, a DNA methyltransferase, is the only LC8 partner in our list that we now consider as false-positive. It was identified as an LC8 partner from rat brain 2988 by affinity chromatography and by pepscan analysis [12,62]; however, the crystal structure of the enzyme [66] shows that the binding motif is located within a globular domain and it is highly unlikely that this sequence would be accessible for LC8 interaction. What are the functions of the LC8 binding partners and what could they tell us about the cellular role of LC8? Apparently, LC8 is involved in a wide variety of functions. Nevertheless, its already identified binding partners appear to represent only a few, often overlapping, essential cell functions and ⁄ or protein complexes (Fig. 3A). In the following paragraphs we select a few examples from these clusters and describe their functions and the possible roles of LC8 in the complexes. Intracellular transport Dynein and LC8 were proposed to be involved in targeting the Swallow protein and the bicoid mRNA in Drosophila oocytes [67]. Intensive studies on the structural aspects of the LC8–Swallow interaction revealed that it is unlikely that LC8 directly links the Swallow– mRNA complex to the dynein motor complex [14,68]. Very recent results ruled out the role of Swallow in bicoid mRNA transport; instead, it was found to be localized to the plasma membrane, where it functions indirectly in bicoid mRNA anchoring [69]. Syntaphilin is targeted to axonal mitochondria and to microtubules as well. A model was proposed in which LC8 serves as the ‘stabilizer’ of a coiled-coil structure in syntaphilin for facilitating its docking ⁄ anchoring to a mitochondrial receptor. Such a physical coupling between LC8 and syntaphilin may control mitochondrial mobility and density in axons and at synapses [49]. Drosophila Dazl is an RNA-binding protein essential for gametogenesis. It was proposed that Dazl travels along the microtubule network in association with the dynein complex and controls the subcellular distribution of a specific set of mRNAs [9]. Bassoon is an LC8 interactor linking the complex to retrograde transport of Golgi-derived vesicles in neurons. It was convincingly shown that Bassoon and LC8 are co-transported by the dynein complex [70]; however, it is still not clear how Bassoon is associated with the motor complex. Nuclear transport The yeast LC8 ortholog Dyn2 dimerizes and stabilizes the Nup82–Nsp1–Nup159 complex, a module of the nuclear pore filaments. Dyn2 binds to five tandem motifs located between a disordered Phe-Gly repeat and a coiled-coil domain of Nup159 forming a rigid FEBS Journal 278 (2011) 2980–2996 ª 2011 The Authors Journal compilation ª 2011 FEBS DYNLL ⁄ LC8 dynein light chain P. Rapali et al. ‘beads-on-a-string’ structure. Dyn2 could play a role in organizing the disordered Phe-Gly repeats within the NPC scaffold to facilitate nucleocytoplasmic transport [71]. LC8 interacts with the replication factor Ciz1 [61,72]. In a proposed model LC8 brings Ciz1 to the nucleus, where it binds Cdk2 and p21. These complexes may play a regulatory role in cell cycle progression of cancer cells [72]. METT-10 is a C. elegans nuclear protein, a putative methyltransferase that acts to inhibit germ cell proliferation. Interaction of METT-10 with LC8 promotes its nuclear accumulation [73]. ATMIN is an ATM-interacting protein the association of which with LC8 might prevent nuclear accumulation of ATMIN or regulate its association with other nuclear proteins involved in detecting DNA damage (P. Rapali, L. Nyitray and I. Rodriguez-Nacho, unpublished results). Mitosis NEK9 is a pleiotropic regulator of mitotic progression, participating in the control of spindle dynamics and chromosome separation [31,74]. LC8 binding probably stabilizes the dimeric Astrin which in complex with SKAP is targeted to bioriented kinetochores [75]. EML3 could have a role in correct metaphase chromosome alignment [76]. The above three LC8 partners were found associated with LC8 in a HTP screen to characterize chromosome segregation protein complexes and were also predicted to have LC8 binding motifs based on our in vitro evolution assay [26]. The nucleoporin Tpr functions during mitosis as a spatiotemporal regulator of spindle checkpoints and it is involved in recruitment of checkpoint proteins to dynein [77]. Apoptosis/autophagy BimL and Bmf are BH3-only pro-apoptotic proteins thought to be normally sequestered to dynein and MYO5A motor complexes via DYNLL1 and DYNLL2, respectively [13,78]. Specific apoptotic stimuli liberate them from the cytoskeleton, in complex with the respective LC8 isoforms, allowing them to translocate to Bcl-2 proteins thereby activating apoptosis. Surprisingly, it was found that the in vivo target specificity of the two highly similar LC8 isoforms is determined by a single surface residue (Tyr41 in DYNLL1 and His41 in DYNLL2) [21]. In vitro the two LC8 isoforms bind the targets with the same affin- ity. The molecular surface around the ‘specificity residue’ might make contacts with other components of their respective motor or cytoskeletal complexes. The contribution of additional binding motifs of the intrinsically disordered Bim and Bmf [56] in their specific localization cannot be ruled out either. AMBRA1 is a component of a multiprotein complex that regulates autophagy and development of the nervous system in mammals [79]. The Drosophila ortholog of LC8 is required for the regulation of autophagy and cell death; however, its interaction partner(s) has not been identified yet [80]. Binding of DYNLL1 inhibits TNFa-induced NFjB activation by interacting with IjBa, thereby preventing its phosphorylation by IjBa kinase, its nuclear translocation and its regulatory role in apoptosis [81,82]. Very interestingly this interaction is redox regulated: TNFa induces the production of reactive oxygen species, which in turn oxidize DYNLL1 (on Cys2 which is an isoform-specific residue) resulting in the dissociation of the complex and NFjB activation. A novel disulfide reductase, TRP14, contributes to the NFjB inhibitory activity by maintaining LC8 in its reduced state [82,83]. Postsynaptic density (PSD) PSD is a dynamic complex crucial for receptor immobilization at both excitatory and inhibitory synapses. Scaffold proteins constitute one major group of proteins present at the PSD and two of them are LC8 binding partners. Gephyrin is critical for glycine- and GABA-receptor clustering and also interacts with many other proteins, including several cytoskeletal components [84]. LC8 binds to a disordered linker domain between two globular dimerization ⁄ oligomerization domains [85,86]. It was shown that the gephyrin–LC8 complex together with the Gly-receptor is involved in transport processes by the dynein complex [87]. GKAP is an important scaffold molecule involved in the assembly of a multiprotein complex at excitatory synapses. Only the DYNLL2 paralog was identified as an interactor of GKAP in vivo, and it was suggested that this interaction is involved in recruiting nNOS to the PSD [88] and in the trafficking of PSD-95 ⁄ GKAP complex by the MYO5A motor [89]. An alternative model could be that these three LC8 interactors bind independently to PSD components and to the motor protein; nNOS is indeed able to interact with PSD-95 [90]. However, the interaction domain or motif responsible for GKAP ⁄ PSD-95 binding to the myosin motor still needs to be identified. FEBS Journal 278 (2011) 2980–2996 ª 2011 The Authors Journal compilation ª 2011 FEBS 2989