Linker Length Drives Heterogeneity of Multivalent Complexes of Hub Protein LC8 and Transcription Factor ASCIZ

Abstract

LC8, a ubiquitous and highly conserved hub protein, binds over 100 proteins involved in numerous cellular functions, including cell death, signaling, tumor suppression, and viral infection. LC8 binds intrinsically disordered proteins (IDPs), and although several of these contain multiple LC8 binding motifs, the effects of multivalency on complex formation are unclear. Drosophila ASCIZ has seven motifs that vary in sequence and inter-motif linker lengths, especially within subdomain QT2-4 containing the second, third, and fourth LC8 motifs. Using isothermal-titration calorimetry, analytical-ultracentrifugation, and native mass-spectrometry of QT2-4 variants, with methodically deactivated motifs, we show that inter-motif spacing and specific motif sequences combine to control binding affinity and compositional heterogeneity of multivalent duplexes. A short linker separating strong and weak motifs results in stable duplexes but forms off-register structures at high LC8 concentrations. Contrastingly, long linkers engender lower cooperativity and heterogeneous complexation at low LC8 concentrations. Accordingly, two-mers, rather than the expected three-mers, dominate negative-stain electron-microscopy images of QT2-4. Comparing variants containing weak-strong and strong-strong motif combinations demonstrates sequence also regulates IDP/LC8 assembly. The observed trends persist for trivalent ASCIZ subdomains: QT2-4, with long and short linkers, forms heterogeneous complexes, whereas QT4-6, with similar mid-length linkers, forms homogeneous complexes. Implications of linker length variations for function are discussed.

Keywords: dimers; duplexes; heterogeneity; hub proteins; intrinsic disorder; linker length; multivalency; transcription factor.

Figure 1

LC8 hub, binding motif, and multivalent partners. (A) Ribbon diagram of the LC8 dimer showing each monomer (orange and cyan) bound to disordered peptides (magenta) that adopt β-strand structure upon binding in LC8’s binding groove (Protein Data Bank code 2P2T, from D. melanogaster). Ribbon diagram is overlayed on a star display of a selection of LC8 partner proteins. Red font denotes monovalent partners, while blue denotes multivalent partners. (B) Amino acid enrichment for each position in the LC8 binding motif. The TQT anchor is boxed in gray. (C) Multivalent LC8 binding partners. Sequence-based predictions of order (red boxes), disorder (black lines), coiled-coil (blue boxes), and LC8 binding motifs (orange bars) are shown. PSIPRED [16] was used to predict order and disorder. Paracoil2 [17] was used to predict coiled-coils. LC8 binding motif locations are shown for Bassoon [18], 53BP1 [19], NUP159 [20], GKAP [21], ASCIZ [22], Chica [23], Panx [24], Pac11 [25], RSP3 [26], and dASCIZ [27]. Panel adapted from Clark et al. [28]. (D) Zoom of QT2–4 from dASCIZ showing the effect of a long linker on flexibility. Panel adapted from Reardon et al. [29].

Figure 2

Alignment of ASCIZ homologs’ domain architecture and dASCIZ constructs. (A) Comparison of 10 ASCIZ homologs, aligned to show the similarity of the LBDs and the linker connecting LBD1 to LBD2. (B) Drosophila ASCIZ LC8 binding region and the constructs utilized in this study, including QT2–4 and QT4–6. (C) Sequences of QT2–4 and QT4–6. (D) Variants that systematically abolish either one (top) or two (bottom) LC8 recognition motifs from QT2–4 through mutation of the anchor triplet to AAA. Hollow boxes indicate sites that have been mutated. Construct nomenclature denotes the binding sites left intact in each construct.

Figure 3

LC8-ASCIZ interactions monitored by ITC. (A–F) Representative thermograms of the titration of LC8 into QT2–4 variants corresponding to the single-site constructs QT2 (A), QT3 (B), and QT4 (C), and the double site constructs QT2,3 (D), QT3,4 (E), and QT2,4 (F). (G–I) Simulated isotherms overlaid on experimental isotherms for each of the double-site variants: QT2,3 (G), QT3,4 (H), and QT2,4 (I). Isotherms were simulated using ΔH and K_d values obtained from single-site isotherms. Fractions of free sites at each point in the simulated isotherms are shown below each, respectively.

Figure 4

Sedimentation velocity analytical ultracentrifugation of double site ASCIZ constructs bound to LC8. SV-AUC titrations of QT3,4 (A), QT2,4 (B), and QT2,3 (C) with LC8 at three separate molar ratios of QT:LC8 (1:1, 1:2, 1:3) and a plot of LC8:QT ratio vs. complex sedimentation coefficient up to a ratio of 1:6. When applicable, populations corresponding to free LC8 are labeled. The dashed lines correspond to the location of the peak seen in QT2–4 at the same ratio.

Figure 5

QT/LC8 complex species and abundance distributions determined by ESI-MS and EM. (A) Native mass spectra of double-site variants at 25 µM are shown with individual and complex species labeled. (B) Abundance distributions of species detected at 25 µM (top) and 5 µM (bottom) for each double-site variant are shown. (C) Monomeric chains of each double site variant with QT2, QT3, and QT4 sites color coded. Upon addition of LC8, 1:2, 1:4, 2:4, and 2:6 complexes form. The most abundant complex species for each QT construct is boxed. (D) Two representative EM images out of 50 captured of QT2–4/LC8 mixture in which bright dots in the images indicate LC8 dimers. Plotted relative populations of complexes seen in EM, and zoomed negatives of all eight 3-mers observed in the 50 images.

Figure 6

AUC, SEC-MALS, and peptide K_d comparisons for the constructs QT2–4 and QT4–6. AUC titration of (A) QT2–4 (replotted data from Reardon et al. [29]) and (B) QT4–6. SEC-MALS chromatogram and mass key for ranges numbered as shown on the chromatogram for (C) QT2–4 and (D) QT4–6 as single proteins and in complex with LC8. LC8 alone trace is shown plotted with both proteins for reference. Highlighted regions of the mass key emphasize the major peaks seen in the SEC-MALS traces of the mixture. (E) Measured K_d values for peptides representing the five binding sites [10] represented across the two analyzed constructs.