De novo structure determination of a 27.5 kDa protein- RNA complex: - PowerPoint PPT Presentation

De novo structure determination of a 27.5 kDa protein- RNA complex: a Dead End for classical NMR approaches? 15 July 2017 | CCPN conference 2017, Stirling University, UK Malgosia Duszczyk, Departement D-BIOL, Institute of Molecular Biology and Biophysics Frédéric Allain lab

Dead End (Dnd1) in zebrafish development & mice • Vertebrate-specific germ cell viability mediator • Expressed in primordial germ cells (PGCs) and is essential for their correct migration in zebrafish • Dnd1 deletions lead to PGC death • Ter mutant mice display germ-cell loss and testicular tumors Weidinger et al. Curr. Biol. 13 (2003), 1429-1434

Dnd1 is linked to miRNA regulation & cancer • Two miRNA targets of miR-430 (nanos; tdrd7) persist in zebrafish germ cells while miR-430 is present • The human homologue miR-373 family acts as an oncogene by repressing tumour supressors LATS2 and p27 • Genetic screens found that Dnd1 is associated with this phenomenon by protecting mRNAs from miRNA-mediated repression in human cell culture and zebrafish PGCs • HOW?? Through interaction with conserved U-rich regions (URRs) in 3’UTRs of these targets Dnd1 blocks miRNA accessibility Kedde et al. (Agami group NKI) Cell 131, 1273–1286, December 28, 2007

Canonical RRM fold & RNA recognition RNP1:[R/K]-G-[F/Y]-[G/A]-[F/Y]-[I/L/V]-X-[F/Y] • The RNA Recognition Motif is RNP2:[I/L/V]-[F/Y]-[I/L/V]-X-N-L the most abundant RBD in higher vertebrates N • Binds primarily ssRNA • babbab fold: 4-strand beta C 3’ sheet packed on two alpha- 3’ helices • Two conserved consensus RNA sequences (RNP1 & RNP2) with exposed aromatic 5’ residues accommodate two RNA nucleotides by stacking • RNA is usually bound 5’-3’ in the direction from b 4 to b 2 Adapted from 2UP1 (Ding et al. 1999)

To exert their different functions RBPs often harbor multiple not identical copies of RBDs U1A,U2B’’ Pre-mRNA splicing Alternative-splicing Sex-lethal, HuD mRNA stability PABP Alternative-splicing PTB/hnRNP L RRM2 RRM1 RRM3 RRM4 mRNA stability/ CPEB1/4 Alternative-splicing RRM1 RRM2 Alternative-splicing hnRNP A1 RRM1 RRM2 TDP43 Alternative-splicing RRM1 RRM2 microRNA-mediated Dnd1 repression inhibition dsRBD RRM2 RRM1 Canonical RNPs Non-canonical RNPs

Both RRMs of Dnd1 necessary for tight binding to p27 3’UTR URR1 5’-AAGCGUUGGA UGUAGCA UUAUGCAAUUAGGUUUUUCCUUAUUUGCUUCAUUGUACUACCUGUGUAUAUAGUUUUUACCUUUUA UGUAGCA CAUAAACUUU-3’ miRNA seed 1a 1b 2a 2b miRNA seed URR1 URR2 2) 1) + CUUAUUUG Dnd1 RRM1 + p27 URR1b RNA Dnd1 RRM2 + p27 URR 1/2 Kd = 41uM No Binding ITC 3) + CUUAUUUG NMR 15N HSQC Dnd1 RRM12 + p27 URR1b RNA ‘fingerprint’ Kd = 1.2uM RRM12 RRM12+ CUUAUUUG 27.5 kDa complex ITC

Specificity? Dnd1 RRM-p27URR interaction data

CUUAUUUG binding site mapping by NMR RNP2 RNP1 Kedde: Y94>C mutant inactive VFIGRL RGFAYA • Note that some putative RNA binding residues cannot be traced due to large shift or exchange in complex (e.g. RNP1 Y94 ) • As expected from mutational analysis RNA binding by canonical RRM1 beta- sheet surface but unexpectedly non-canonical elements on RRM1 ( a0b0 ) and 2 ( a2b4)

RRM12-URR1-CUUAUUUG complex 15N relaxation 18ns ~30kDa

NMR spectroscopy is an important method to solve small to medium sized RNAs and protein-RNA complexes PDB statistics (09.07.2017) Method: Xray solNMR EM ssNMR All structures: Molecule All 117968 11827 102 1594 RNA only 783 522 29 1 Protein-RNA 1557 116 348 0 under 40kDa: Molecule All 44736 11564 75 76 RNA only 624 516 7 1 Protein-RNA 185 109 10 0

Solving protein-RNA structures using NMR spectroscopy: pipeline • Production of Dnd1 RRM12-CUUAUUUG complex: Recombinant protein expression Purchase of short unlabeled RNA oligos or synthesis using isotope labeled phosphoramidites In vitro transcription, used for longer RNAs using 13C/15N labeled and/or deuterated nucleotides not possible From F. Nelissen RUNijmegen

Backbone assignment RRM12 – CUUAUUUG complex 227 aa’s + 8nt RNA = 27.4 kDa Fractional random deuteration (auto-induction in D2O medium) & TROSY Bruker experiments: trHNCO trHN(CA)CO trHNCA trHN(CO)CA trHNCACB trHN(CO)CACB trHNCACB2H3D CBCACONH3D 15N-NOEsy on 2H sample Both 0.4mM and comparable 93.4% assigned (22 prolines) measurement time

Structure determination strategy • Free protein precipitates too rapidly for 3D experiments and in complex above 298K If possible solve free protein structure first and work at as high temperature as possible • Backbone assignment using TROSY-based triple resonance experiments on randomly fractionally deuterated 27.4 kDa complex (93% assigned) • Few signals in HCCCONH type sidechain experiments and many missing and overlapped peaks in HcCH- and hCCH-TOCSY so sidechain assignment must be done mainly through NOESYs: Look for HN-HA/HB correlations in HN plane of n & n+1 at expected CA & CB shifts, compare to HN NOE strips, then move further out the sidechain at expected C-shift >> it helps a lot if you have homology model Use HcCH/hCCH-TOCSY or COSY (less overlap) where possible to confirm intra-spin system correlation

Structure determination strategy • 3D-13C-HMQC-NOESY for higher sensitivity – NOE in acquisition dimension for higher resolution • Some RNA-interacting residues exchange broadened: some, but not full improvement by addition of excess RNA • Lots of overlap (e.g. 124 methyl groups) If possible, try segmental labeling (protein ligation) • Sidechain assignment of free RRM2 • Transfer of assignments to RRM12 • 1H assignment completeness 80% (75% RRM1, 88% RRM2) • High number of prolines (22) > still many not assigned in RRM1

Protein-RNA complex structure calculation protocol Peak picking & automated NOE assignment using ATNOS/CANDID T.Herrmann et al. J. Mol. Biol. 319 (2002) 209 Manual adjustment of peak lists, automatic NOE assignment using CYANA noeassign (more flexible approach, manual assignments kept) List of protein-protein distance constraints Manually assigned intra-RNA and intermolecular distance constraints H-bonds, dihedral angles, RDCs Structure calculation using CYANA: 100 structures Güntert et al. J. Mol. Biol. 273 (1997), p. 283 Structures Final refinement with Amber 12.0

ATNOS/CANDID protocol Essential to obtain correct fold for multidomain protein: Calculate single domains first using: • 3D 15N NOESY (H2O) • 3D 13C ali-NOESY (D2O & H2O) • 3D 13C aro-NOESY (H2O) • 2D NOESy (D2O) • H-bonds (based on non-exchanged HNs in D2O HSQC) • TALOS restraints • 140 upls from manually assigned NOEs Then combine doing run with all chemical shifts and upls obtained from single domain runs

Strategy for structure calculations/improving structural statistics of Dnd1 RRM12 • ATNOS/CANDID fails to pick full peak lists – manual adjustments: Spectrum # peaks AC # peaks man 3D 15N-NOESY 1947 3569 3D 13C-NOESYali(D2O) 2568 3956 3D 13C-NOESYali(H2O) 3627 5369 3D 13C-NOESYaro 382 420 • Peak lists, partly manually assigned (intensities) fed into CYANA for noeassign procedure • Using ‘KEEP’ for keeping manual assignments Systematic underestimation of distances caused by spin diffusion Adjusting calibration parameters to improve statistics dref for median peak volume (4.0 > 4.6A) upl_values (2.4..5.5A > 2.7..6.0A) elasticity (changevol) = small increase violated upls

Cycle-dependent parameters of automated NOE assignment: may be changed in noeassign.cya 1.0 > 1.25 1.25 > 1.5 Changevol=true J Biomol NMR (2015) 62:453–471

Structure of Dnd1 RRMs within the RRM12-CUUAUUUG complex Mean global bb Non-canonical C (130..208) RMSD: 0.64 extensions N a0b0 RRM1 b4b1b3b2 a1 a2 a3 Mean global bb (6..124) RMSD: 0.92 C Canonical RRM Novel RRM - RRM2 babbab fold abb_babbab - fold b4b1b3b2 C-term a helix N 6-strand beta-sheet a1 a2 independent packed on triple helix

Structural ensemble RRM12-CUUAUUUG complex C RRM2 Input MD: Intra protein: 3602 restraints Intra RNA: 171 restraints Intermolecular: 93 restraints RDCs: 112 restraints (NH) H-bonds 158 restraints 3’ 5’ Mean global bb (6..208) RMSD: 1.48 RNA (U3-U7): 1.50 RRM1 N

RDCs for improving local structures & domain-domain orientation • Why RDCs? T1/T2/hetNOE measurements show one rigid complex domain:domain contacts almost 100% via few RNA residues better orientation should be achieved with long distance restraints Tested alignment media: • Pf1 at 12.5mg/ml > 15Hz D2O splitting > too strong alignment at low salt (100mM KPO4 pH 6.5) > severe broadening even after dilution • C12E5/hexanol at 4.2% OK > careful of temperature dependency of aligned phase!!! • N-H in aligned phase quite broadened > improvement with deuterated sample?

Initial fit RDCs to complex structure RRM1 RRM2