Dispersion from Cα or NH : 4D experiments for backbone resonance assignment of intrinsically disordered proteins

This is a self-archived version of an original article. This version may differ from the original in pagination and typographic details. Author(s): Title: Year: Version: Copyright: Rights: Rights url: Please cite the original version: CC BY 4.0 https://creativecommons.org/licenses/by/4.0/ Dispersion from Cα or NH : 4D experiments for backbone resonance assignment of intrinsically disordered proteins © The Author(s) 2020 Published version Tossavainen, Helena; Salovaara, Santeri; Hellman, Maarit; Ihalin, Riikka; Permi, Perttu Tossavainen, H., Salovaara, S., Hellman, M., Ihalin, R., & Permi, P. (2020). Dispersion from Cα or NH : 4D experiments for backbone resonance assignment of intrinsically disordered proteins. Journal of Biomolecular NMR, 74(2-3), 147-159. https://doi.org/10.1007/s10858-020-00299-w 2020 Vol.:(0123456789) 1 3 Journal of Biomolecular NMR (2020) 74:147–159 https://doi.org/10.1007/s10858-020-00299-w ARTICLE Dispersion from Cα or NH: 4D experiments forbackbone resonance assignment ofintrinsically disordered proteins HelenaTossavainen1 · SanteriSalovaara1· MaaritHellman1· RiikkaIhalin2 · PerttuPermi1,3 Received: 30 August 2019 / Accepted: 2 January 2020 / Published online: 13 January 2020 © The Author(s) 2020 Abstract Resonance assignment of intrinsically disordered proteins is remarkably challenging due to scant chemical shift dispersion arising from conformational heterogeneity. The challenge is even greater if repeating segments are present in the amino acid sequence. To forward unambiguous resonance assignment of intrinsically disordered proteins, we present iHACANCO, HACACON and (HACA)CONCAHA, three Hα-detected 4D experiments with Cα as an additional dimension. In addition, we present (HACA)CON(CA)NH and (HACA)N(CA)CONH, new 4D Hα-start, HN-detect experiments which have two NH dimensions to enhance peak dispersion in a sequential walk through C′, NH and HN, and provide more accurate NH/HN chemical shifts than those that can be obtained from a crowded 1H, 15N-HSQC spectrum. Application of these 4D experiments is demonstrated using BilRI (165 aa), an outer-membrane intrinsically disordered protein from the opportunistic oral pathogen Aggregatibacter actinomycetemcomitans. BilRI amino acid sequence encompasses three very similar repeats with a 13-residue identical stretch in two of them. Keywords Aggregatibacter actinomycetemcomitans· BilRI· Resonance assignment· Intrinsically disordered protein· IDP Introduction Aggregatibacter actinomycetemcomitans is a Gram-negative opportunistic oral pathogen that is linked to periodontitis, infection of tissues supporting the teeth (for reviews, see Fine etal. 2006; Åberg etal. 2015; Fine etal. 2019). Although A. actinomycetemcomitans resides in subgingival multispecies biofilms, it is also able to migrate to underlying vessels and cause systemic diseases such as cardiovascular diseases (Kozarov etal. 2005; Hyvärinen etal. 2012). The host response to biofilms is mediated by inflammatory cytokines. In healthy junctional epithelium of tooth the balance of various cytokines and chemokines ensures that the host defence works appropriately. Some periodontal pathogens are able to impair the balance. Porphyromonas gingivalis can suppress the expression of chemokine interleukin(IL)-8 (Takeuchi etal. 2013) and A. actinomycetemcomitans biofilm is able to sequester and internalize IL-1β, IL-8 and IL-6, which leads to changes in biofilm composition and metabolic activity (Paino etal. 2011, 2012; Ahlstrand etal. 2017). It has been suggested that A. actinomycetemcomitans bacterial IL receptor I, BilRI is associated with this sequestering activity (Paino etal. 2013; Ahlstrand etal. 2017). BilRI is an outer membrane lipoprotein able to bind IL-1β, IL-8 and IL-10 and tumor necrosis factor (TNF)-α (Ahlstrand etal. 2017). Due to its rather low binding affinity it is assumed that BilRI acts by concentrating cytokines on cell membranes, which are then transferred to other components of the uptake system (Paino etal. 2013; Ahlstrand etal. 2017). We have engaged in the structural characterization of BilRI. BilRI is an intrinsically disordered protein (IDP), as demonstrated by its 1H, 15N HSQC spectrum, which displays very limited signal dispersion (Ahlstrand etal. 2017). In the HN dimension their dispersion is only 0.63ppm. This arises from a feature typical of IDPs, namely a biased amino acid Electronic supplementary material The online version of this article (https ://doi.org/10.1007/s1085 8-020-00299 -w) contains supplementary material, which is available to authorized users. * Perttu Permi perttu.per[email protected] 1 Department ofChemistry, Nanoscience Center, University ofJyväskylä, Jyväskylä, Finland 2 Department ofBiochemistry, University ofTurku, Turku, Finland 3 Department ofBiological andEnvironmental Science, University ofJyväskylä, Jyväskylä, Finland 148 Journal of Biomolecular NMR (2020) 74:147–159 1 3 composition with a pronounced number of polar or charged amino acids, low number of bulky hydrophobic amino acids and lack of aromatic amino acids (Dyson 2016). BilRI amino acid sequence is dominated by alanine (23%), lysine (14%) and aspartic acid (13%) residues. However, unlike the general trend (Dunker etal. 2001), BilRI sequence contains only one proline and one glycine. BilRI is positioned among IDPs in the mean net charge-hydropathy plot (Uversky etal. 2000) (Fig.1a). The paucity in the variability of residue’s nearest neighbors narrows down the chemical shift range of a particular amino acid type, for example there are 12 Lys-AspAla triplets in the sequence, which most likely results in close chemical shifts for the middle aspartic acid. Additionally, the BilRI sequence contains three very similar segments of about forty residues. The longest identical stretch is of 13 residues, present in two of these repeats, covering residues 71–82 and 111–122 (Fig.1b). The 1H, 15N HSQC spectra of IDPs typically being very challenging to scrutinize, the resonance assignment most often relies on spectra other than the classical HN-detected experiments widely used for folded proteins (for reviews see e.g. Sattler etal. 1999; Permi and Annila 2004). Indeed, the conventional approach based on the HNCACB and HN(CO)CACB/CBCA(CO)NH experiments that link the intraand sequential 13Cα and 13Cβ chemical shifts to 1HN and 15N frequencies, are very inefficient for many IDPs due to severe clustering of aliphatic carbon chemical shifts for each residue type. Instead, correlation of 15N and 13C′ frequencies provide much better results for IDPs (Yao etal. 1997; Mäntylahti etal. 2009; Bermel etal. 2012) Another obstacle arises from the increasing chemical exchange rate of amide protons with water at alkali pH and/or elevated measurement temperature (Mäntylahti etal. 2010). Yet another challenge for the assignment originates from the abundancy of proline residues in IDPs. As an N-substituted residue, proline lacks the amide proton, which results in gaps during the resonance assignment procedure. Although this can be a benefit in the case of globular proteins, it significantly hampers the resonance assignment of disordered systems (Hellman etal. 2014). Several different approaches have been proposed to overcome these obstacles imposed during the assignment procedure. These include increase of dimensionality from conventional 3D to 4–7D spectra (Fiorito etal. 2006; Motáčková etal. 2010; Nováček etal. 2011; Kazimierczuk etal. 2013; Brutscher etal. 2015) as well as detection of non-exchangeable spins 13C′ and 1Hα instead of 1HN (Bermel etal. 2006a, 2009; 2012; Mäntylahti etal. 2010, 2011; Permi and Hellman 2012). Our group has been resorting both to Hα-start, HN-detect pulse schemes (Mäntylahti etal. 2009; Hellman etal. 2014) or complete Hα-detection experiments (Mäntylahti etal. 2010, 2011; Permi and Hellman 2012) for the assignment of IDPs, which overcome hurdles associated with proline assignment and exchange broadening at alkali pH or elevated temperature. In addition to reduced susceptibility towards solvent exchange induced linebroadening, the Hα chemical shift is extremely valuable in structural analysis. The Cα, Hα and C′ shifts are particularly sensitive to the ϕ/ψ angles of the protein backbone and thus the most informative in the estimation of secondary structure content in an IDP (Borcherds and Daughdrill 2018). Comprehensive assignments allow for detailed, residuespecific analysis of structure and dynamics (Konrat 2014). There are therefore grounds for an extra effort towards a more comprehensive backbone resonance assignment. IDPs are often comprised of repetitive amino acid sequences and hence higher dimensionality in combination with high resolution offer superior results. However, the increased dispersion of signals should not be obtained, if possible, at expense of sensitivity. Here we present the resonance assignment of BilRI, whose demanding amino acid sequence necessitated development of a suite of 4D pulse sequences that offer superior signal dispersion with respect to their well-established 3D counterparts without indirect sampling associated sensitivity loss. b 0.0 0.1 0.2 0.3 0.4 0.5 0.20.3 0.40.5 0.6 BilRI Disordered Proteins Ordered Proteins Mean Scaled Hydropathy Absolute Mean Net Charge a DDSKTSPQAEQAKTSVSEAK DAVVNAANDV KDATVEAAKD AQNMAADKMV EVKDAISEKM DAMTTQASEM KDAAVEAAKD AKDAAADKMA EVKDAISEKM DAMATQVNEM KDTAAEAVKD AKDAAADKMT EVKDAVSEKMGATATQTNEM KDAVKSETES K GSHM 100 150 160 170140 180 30 40 50 70 80 9060 110 120 130 Fig. 1 a Mean net charge-hydropathy plot of BilRI. Data for the plot were created with the PONDR predictor (https ://www.pondr .com/). b Amino acid sequence of A. actinomycetemcomitans BilRI. Amino acids are classified by type: yellow, small hydrophilic (A, G, N, P, Q, S, T); green, hydrophobic (I, M, V); red, negatively charged (D, E) and blue, positively charged (K) amino acids. BilRI sequence does not contain C, F, H, L, R, W or Y. The longest 13-residue repetitive segments are underlined in black, and the broken underline indicates a very similar segment. The second longest 11-residue repetitive segments are anderlined in blue, and the triplet K-D-A, which recurs twelwe times in the sequence is underlined in red. The first four residues are a cloning artefact 149Journal of Biomolecular NMR (2020) 74:147–159 1 3 Materials andmethods Protein expression andpurification The gene encoding BilRI (residues 21–181) was cloned to pET15b vector (Novagen) into the NdeI and XhoI sites. This leads to soluble recombinant BilRI protein with N-terminal His-Tag with a thrombin cleavage site. Production of 13C, 15N labeled BilRI was carried out by transforming plasmids into the BL21(DE3) cells. Cells were grown in M9 minimal media, supplemented with 1g/l of 15NH4Cl and 2g/l 13C-d-glucose as the sole nitrogen or nitrogen and carbon source, respectively. Cell culture was incubated at 37°C and temperature was decreased to 16°C when OD of the cell culture reached 0.4 and protein production was induced with 1mM IPTG when OD of the cell culture reached 0.6. Cells were further incubated at 16°C for Table 1 Data acquisition parameters Experiment Points in F1 (ms) Points in F2 (ms) Points in F3 (ms) Points in F4 (ms) Sampling % Number of scans 4D iHACANCO 80 (16.6) 13C′84 (19.1) 15N 48 (3.9) 13Cα1024 (80.1) 1Hα10 4 4D HACACON 64 (13.2) 15N 64 (14.5) 13C′48 (6.9) 13Cα1024 (80.1) 1Hα10 4 4D (HACA)CONCAHA 64 (13.2) 13C′64 (14.5) 15N 48 (6.9) 13Cα1024 (80.1) 1Hα20 4 4D (HACA)CON(CA)NH 74 (26.3) 13C′80 (24.7) 15N 82 (25.3) 15N 1024 (71.2) 1HN7 4 4D (HACA)N(CA)CONH 74 (26.3) 13C′80 (24.7) 15N 82 (25.3) 15N 1024 (71.2) 1HN7 4 3D iHA(CA)NCO 116 (24.0) 13C′200 (45.5) 15N – 1024 (80.1) 1Hα25 8 3D HA(CA)CON 220 (50.0) 15N 230 (46.8) 13C′– 1024 (80.1) 1Hα25 8 3D (HACA)CON(CA)HA 156 (32.3) 13C′220 (50.0) 15N – 1024 (80.1) 1Hα25 8 15 N (ppm) 1 H (ppm) 13 C (ppm) 181K 164A134A 175K 87A39A 53A 71E 57A 104A 137A 144A,, 56E 82A 173A 61A 42A 96E 151E, 162A 179E 38E 45N 55V 106A 146A, 89E 24K 47A 94A 131K 152V 171K,,,,, 105A 112V 145A,, 128N 132D 129E 79K 119K 159K,, 36V 127V 70V 69M 90M, 169E 108K 160M 148K 168N 149M 68K 41D 49D 60D 64M,,, 99K 50V 59K 102K 28Q 111E 30E 43V 174V 136E 138V 95V 37S 156V 67D 83M 62Q 35S 123M 180S 176S 48N 85T 23S 167T 150T 88S 25T 178T 54T 165T 163T 125T 161G 177E 34T 84T 133T,, 107D 147D, 63N 157S, 26S 77S 117S,, 76I 116I, 142K 109M 33K 40K 100D 172D,, , 103D 143D, 21D 80M 92D 130M,,,20M 81D 120M 121D 140D,, ,, 22D 52D 74D 114D 154D,,,, 158E 78E 86Q 118E 166Q,, , 98A 135A, 58A 66A, 72V 91K 126Q,, 44V 75A 115A 122A 155A,, ,, 65A 93A 97A 141A,, 29A 32A 46A 139K,,, 51K 73K 153K, 31Q 170M, 113K 101A 110A 124A,, x x 120 8.28.3 7.0 7.5 8.08.5 a 121 181K 164A 134A 175K 87A 39A 139K 71E 57A 97A 104A 124A 122A56E 75A 82A 173A 61A 42A 96E 162A 179E 38E 66A 166Q 45N 94A 55V 106A 89E 24K 105A 128N 132D 129E 79K 120M 36V 114D 127V 70V 169E 108K 143D 160M, 148K 168N 149M 100D 68K 41D 99K 50V 59K 102K 28Q 111E 30E 43V 116I 174V 136E 138V 95V 77S 37S 156V 157S 147D 67D 83M 62Q 35S 123M 180S 176S 48N 85T 23S 167T 150T 88S 133T 25T 178T, 54T 165T 163T 125T 161G 177E 34T 84T 107D 63N 117S 26S 76I 142K 60D 49D 109M 64M 40K 172D 33K 103D 119K 159K 21D 80M 121D 74D 154D, 81D 20M 22D 158E 86Q 126Q, 78E 118E 146A 98A 135A, 58A 171K 47A 131K 91K 151E 44V 115A 155A 141A 110A 137A 144A 65A 93A 29A 53A, 32A 46A 51K 153K 73K 152V 145A 31Q 113K 112V 101A 1:90M 170M, 18S 172174176178180 ** * 19H * * 123 45 2:69M 3: ,92D 130M 4:52D 5:140D 176.0 123 122 72V 176.2 b 110 115 120 125 8.4 110 115 120 125 Fig. 2 a 2D 1H, 15N-HSQC spectrum of BilRI. The peaks are labeled with residue numbers and one-letter amino acid codes. Crosses indicate peaks found at lower contour levels. Residue numbering corresponds to that of whole BilRI protein (1–181) although the construct used was shorter (21–181). b 2D CON spectrum of BilRI. The peaks are labeled with residue number and amino acid code of the amide nitrogen in the C′-NH pair. The peak of the only proline of the BilRI amino acid sequence, which resonates at 172.8 (26Ser C′), 138.1 (27Pro NH) ppm is not shown. Asterisks indicate impurities. Both 2D spectra were acquired at 800MHz 1H frequency, 25°C from a 1mM BilRI sample at pH 6.5 150 Journal of Biomolecular NMR (2020) 74:147–159 1 3 16h and collected by centrifugation. Cells were disrupted with sonication and resulting supernatant was clarified by centrifugation with 30,000×g. Clarified supernatant of BilRI was applied to the 1-mL His GraviTrap column (GE Healthcare) and the His-Tag was removed by thrombin protease (GE Healthcare) digestion according to the manufacturer’s instructions. Protease digested mixture was applied to His GraviTrap column. BilRI, without His-Tag, eluted with flow-through. Flowthrough was concentrated to volume of 1ml with Vivaspin 2 concentrator. Concentrated BilRI sample was applied into the Superdex 75 16/60 gel filtration column (GE Healthcare). Buffer used in gel filtration contained 20mM sodium phosphate (pH 6.5) and 50mM NaCl (NMR buffer). Fractions with pure BilRI were pooled and concentrated for NMR studies. The gel filtration was performed by using the ÄKTA Purifier FLPC purification system (GE Healthcare). abc d e f 151Journal of Biomolecular NMR (2020) 74:147–159 1 3 NMR spectroscopy BilRI NMR experiments were acquired using 0.5–1.0mM 15N, 13C labeled protein samples in 5/95% D2O/H2O at pH 6.5. Chemical shifts were referenced to external 2,2,-dimethyl-2-silapentane-5-sulfonic acid (DSS). All data were acquired at 25°C on a Bruker AVANCE III HD 800MHz spectrometer, equipped with a TCI 1H/13C/15N cryoprobe. In addition to the new 4D experiments described here, the following experiments were used in the resonance assignment: 2D 1H, 15N-HSQC, constant time 1H, 13C-HSQC, 13C-detected 2D CON (Bermel etal. 2006b), 3D HN-detected HNCACB, CBCA(CO)NH, HNCO (reviewed in (Sattler etal. 1999; Permi and Annila 2004) and i(HACA)CO(CA) NH (Mäntylahti etal. 2009), 3D Hα-detected HA(CA)CON, iHA(CA)NCO and (HACA)CON(CA)HA (Mäntylahti etal. 2010, 2011). All 3D/4D experiments were collected using non-uniform sampling (Table1). Sampling densities were 25% for the 3D experiments and 7–20% for the 4D experiments. NMR data were processed with TopSpin 3.5 (Bruker Inc) and analyzed with CcpNmr Analysis v. 2.4.2 (Vranken etal. 2005). BilRI chemical shifts have been deposited to the BMRB database (www.bmrb.wisc.edu) with accession code 27824. Results anddiscussion Peaks in the BilRI 2D 1H, 15N-HSQC spectrum display overwhelming overlap (Fig.2a) and a very narrow distribution in the HN dimension. As it was later revealed, all but three of the 21 aspartic acid residues’ peaks are located in the middle region pile of peaks, 8.27–8.38, 121.2–121.9ppm, together with 11 of the 13 methionine residues’ amide peaks. The Asp and Met α signals heavily overlap also in the 1H, 13C CT-HSQC (Suppl. Fig. S1). The CON spectrum, on the other hand, shows remarkably well dispersed signals (Fig.2b) with peaks from all C′–NH pairs present. We thus first attempted the assignment with spectra having the CON spectrum as the root spectrum, namely Hα-detected experiments HA(CA)CON, iHA(CA)NCO and (HACA)CON(CA) HA (Mäntylahti etal. 2010, 2011). 3D HN-detected experiments HNCACB, CBCA(CO)NH, HNCO and i(HACA) CO(CA)NH (Mäntylahti etal. 2009) were acquired to collect HN, C′, Cα and Cβ chemical shifts. With this set of seven 3D spectra, we assigned the majority of backbone resonances. However, there were several ambiguous assignments, in particular within the aforementioned repeating segments. Moreover, while precise Hα and C′ shifts were obtained from the 3D Hα-detected experiments and the CON, due to heavy overlap, Cα/Cβ shifts were far more difficult to read from the HNCACB and CBCA(CO)NH spectra. Indeed, a precise Cα/Cβ chemical shift for 66% of the residues was obtained, Fig. 3 a–c Schematic presentation of magnetization transfer pathway during the 4D iHACANCO d, 4D HACACON e and 4D (HACA) CONCAHA f experiments. Red arrows indicate direct transfer pathway from 1Hα(i) to 15N(i) or 15N(i + 1) whereas green arrows indicate a nested CαC′ZNZ →NZCα transfer in d) known as the intraresidual filter (Permi 2002; Brutscher 2002) or highly selective CαC′ZNZ →Cα transfer in e). Arrows indicate out-and-back type magnetization transfer, whereas one-way arrows represent coherence transfer route which is unidirectional. One-letter codes above the arrows indicate time points in the pulse sequence. d Intraresidual iHACANCO experiment to correlate 1Hα(i), 13Cα(i), 13C′(i) and 15N(i) chemical shifts, e the HACACON experiment, which correlates chemical shifts of 1Hα(i), 13Cα(i), 13C′(i) and 15N(i + 1) resonances. f The (HACA) CONCAHA experiment for correlating 1Hα(i), 13Cα(i), 13C′(i) and 15N(i + 1) resonances. Narrow and wide filled bars on 1H and 15N channels correspond to rectangular 90° and 180° pulses, respectively, applied with phase x unless otherwise stated. All 13C pulses are bandselective shaped pulses, denoted by filled narrow bars (90°) and filled and unfilled half ellipsoids (180°). Unfilled bars are applied on-resonance. The 1H, 15N, 13C′, and 13Cα carrier positions are 4.7 (water), 118 (center of 15N spectral region), 174ppm (center of 13C′ spectral region), and 56ppm (center of 13Cα spectral region). The 13C carrier is set initially to the middle of 13C′ region (174ppm), shifted to 13Cα region (56ppm) prior to 90° 15N pulse ϕ1 in scheme d). In scheme e) and f), the carrier is initially at 56ppm and shifted to 174ppm prior to 90° 13C pulse ϕ2, and shifted back to 56ppm before 90° 13C pulse ϕ4. The first band-selective 180° 13C pulse, refocusing 13Cα magnetization (56ppm, denoted with an asterisk) had duration of 788μs at 800MHz. Other band-selective 90° and 180° pulses for 13Cα (56ppm) and 13C′ (174 ppm) were applied with durations of 240.0 μs and 192.0μs at 800MHz, respectively. Band-selective 90° and 180° pulses for 13C′/13Cα have the shape of Q5 and Q3 (Emsley and Bodenhausen 1992) and duration of 240.0μs and 192.0μs at 800MHz, respectively. The adiabatic 180° Chirp broadband inversion pulse for inverting 13Cα and 13C′ magnetization in the middle of t1 period had duration of 500μs at 800MHz (Böhlen and Bodenhausen 1993). The Waltz-65 sequence (Zhou etal. 2007) with strength of 4.17kHz was employed to decouple 1H spins. The GARP (Shaka etal. 1985, 1987) with field strength of 4.55 kHz was used to decouple 13C during acquisition. Delay durations: τ = 1/(4JHC) ~ 1.7ms; τ2 = 3.4ms (optimized for nonglycine residues) or 2.2–2.6ms (for observing both glycine and nonglycine residues); ε = duration of GH + field recovery ~ 0.4ms; 2TC = 1/ (2JCαC′) ~ 9.5 ms; TCA = 1/(6JCαC′) ~ 3.3ms; TA = 1/(4JC′N) ~ 16.6 ms; TC′ = TC + TCC; TCC = 1/(JCαCβ)–1/(4JC′N)–1/(2JCαC′) ~ 0–2.5 ms; TNC ~ 14 ms; TCN ~ 14 ms; TN ~ 14 ms. Maximum t1, t2 and t3 are restrained in scheme d, t2,max < 2.0*TC′, t3,max < 2.0*TCN, in scheme e, t2,max < 4.0*TA, t3,max < 2.0*TCA, in scheme f, t1,max < 2.0*TA, t2,max < 2.0*TNC, t3,max < 2.0*TCN. Frequency discrimination in 15 N and 13C′ dimensions is obtained using the States-TPPI protocol (Marion etal. 1989) applied to ϕ1 and ϕ2, respectively, whereas the quadrature detection in 13Cα dimension is obtained using the sensitivity-enhanced gradient selection (Kay etal. 1992; Schleucher etal. 1994). The echo and antiecho signals in 13Cα dimension are collected separately by inverting the sign of the GC gradient pulse together with the inversion of ψ, respectively. Phase cycling: ϕ1 = x, − x; ϕ2 = 2(x), 2(− x); ϕ3 = 4(x), 4(− x); ϕ4 = x; ψ = x; rec. = x, 2(− x), x, − x, 2(x), − x. Selective 180° pulse for 13Cα in the middle of delay 2TA induces a Bloch-Siegert shift to 13C′ magnetization, a careful adjustment of phase (bsp) of the last 13C′ 90° (phase y) pulse is necessary in scheme d). Gradient strengths and durations: GC = 13 k G/cm (1.6 ms), GH = 13k G/cm (0.4ms). The pulse sequences code and parameter file for Bruker Avance system are available from authors upon request ◂ 152 Journal of Biomolecular NMR (2020) 74:147–159 1 3 while for Hα, resolved peaks for 87% of the residues were observed (Suppl. Fig. S2). To resolve these ambiguities and to extend the number of accurate chemical shifts, we resorted to 4D NMR spectroscopy. In order to bypass the Cα overlap problem in 3D HN-detected experiments and to establish direct connectivities between Hα and Cα, and to provide dispersion to solve ambiguities arising from occasional overlap of C′, Hα resonances encountered in the 3D Hα-detected experiments, we devised Hα-detected 4D experiments with Cα as an additional dimension. In addition, we developed and employed new 4D Hα-start, HN-detect experiments which can bridge stretches over single prolines similar to 3D experiments described in Hellman etal. (2014), but which have two NH dimensions to enhance peak dispersion in a sequential walk through C′, NH and HN, and to provide for more accurate NH/ HN chemical shifts than those that could be obtained from the crowded regions of the 1H, 15N HSQC. 4D iHACANCO, HACACON and(HACA)CONCAHA experiments The proposed 4D iHACANCO, HACACON and (HACA) CONCAHA experiments are extensions of their established 3D counterparts (Mäntylahti etal. 2010, 2011) with additional sampling on the fourth 13CA dimension (Fig.3). The coherences flow through the 4D iHACANCO, HACACON and (HACA)CONCAHA experiments in Eqs.1, 2, 3: (1) 1 H𝛼(i) 2𝜏 (1 JH𝛼C𝛼 ) �������������������������������������→ 13C𝛼(i) 2TC (1 JC𝛼N, 1 JC𝛼N, 1 JC𝛼C� ) ��������������������������������������������������������������������������������→ 13 C�(i)[2TA−t2;1JC�N,1JC𝛼N,2JC𝛼N] →13C𝛼(i)[2TC�−t2;1JC�𝛼N,2JC𝛼N,1JC𝛼C�]→15N(i)[t1 ] →13C𝛼(i) [ 2TN−t3;1JC�𝛼N,2JC𝛼N ] 4𝜏(1JH𝛼C𝛼) �������������������������������������→ 1H𝛼(i)[t4 ] (2) 1 H𝛼(i) 2𝜏(1JH𝛼C𝛼) �������������������������������������→ 13C𝛼(i) 2TC(1JC𝛼C�) �����������������������������������������→ 13C�(i) 2TA(1JC�N) ��������������������������������������→ 15 N(i+1)[t1]→13C�(i)[2TA−t2;1JC�N] →13C𝛼(i) [ 2TC�−t3;1JC𝛼C� ] 4𝜏(1JH𝛼C𝛼) �������������������������������������→ 1H𝛼(i)[t4] (3) 1 H𝛼(i−1) 2𝜏(1JH𝛼C𝛼) �������������������������������������→ 13C𝛼(i−1) 2TC(1JC𝛼C�) �����������������������������������������→ 13 C�(i−1)[2TA−t1;1JC�N] →15N(i)[2TNC −t2;1JC𝛼N,2JC𝛼N,1JC�N] →13C𝛼(i)[2TA−t2;1JC�N] →13C𝛼(i) [ 2TCN −t3;1JC𝛼N,2JC𝛼N ] 4𝜏(1JH𝛼C𝛼) �������������������������������������→ 1H𝛼(i)[t4 ] respectively. All experiments start with the 1Hα(i)→ 13Cα(i) transfer, and the density operator immediately after the ϕ3 pulse is described as Hαz(i)Cαz(i) (time point a). Subsequently, the magnetization is transferred to the 13C′ spin followed by the labeling of 13C′ chemical shift in t1 or t2. The relevant density operator after the ϕ2 pulse in all experiments is described as Cαz(i)C′y(i) (time point b). Next, the desired coherence is transferred to the 15N spin of the sequential residue in HACACON and (HACA)CONCAHA experiments, described with the density operator Cαz(i)C′z(i) Ny(i + 1) (time point c). In the iHACANCO experiment, the magnetization is solely transferred to the 15N spin within the residue, described with the density operator Cαz(i)C′z(i) Ny(i) (time point c). After labeling the 15N chemical shifts in t1 (or t2), the magnetization is transferred to the 13Cα coherence after the ϕ4 pulse. The relevant density operators (time point d) are Cαy(i)Nz(i) for iHACANCO, Cαy(i)C′z(i) for HACACON and Cαy(i)Nz(i) for (HACA)CONCAHA schemes. The 13Cα chemical shift is labeled during the t3 period between time points d–e. While the HACACON (Fig.3b) is the conventional out-and-back experiment, the (HACA)CONCAHA and iHACANCO utilize the intraresidual filter for the selective 13Cα(i)→ 15N(i) transfer (Permi 2002; Mäntylahti etal. 2010, 2011). Especially in the iHACANCO experiment, the magnetization transfer is nested and further clarification is delivered in the following. After converting magnetization to the Hαz(i)Cαz(i) coherence (time point a), the 1JCαC′, 1JCαN, 2JCαN and 1JCαCβ couplings are active during the time interval (2TC + 2TA + 2TC′ + TCC) = 52–57ms, which converts it to the Cαz(i)Ny(i) coherence (time point c). However, during the delay 2TCC, that can be selected to be 0–5ms based on the relaxation properties of 13Cα spins, only 1JCαCβ is active. This is to maximize the transfer efficiency during the (2TC + 2TA + 2TC′ + TCC) delay. Of note, to avoid chemical shift evolution of 13Cα–13C′ multiple-quantum coherence during 2TA –t2, an additional but opposite frequency labeling period for 13Cα has been implemented in the 2TC′ period. Efficiently, only the chemical shift evolution of 13C′ will take place during t2, but the attainable resolution is limited by 2TC′ (= 2TC + TCC) i.e. t2,max is 19–22ms, depending on the setting of 2TCC (0–5ms). Thus, the 4D iHACANCO, HACACON and (HACA) CONCAHA experiments yield correlations at ωHA(i), ωCA(i), ωC′(i), ωN(i); ωHA(i), ωCA(i), ωC′(i), ωN(i+1), and ωHA(i), ωCA(i), ωC′(i–1), ωN(i) frequencies, respectively. Given that frequency labeling of 13Cα chemical shifts is implemented in a constant-time manner, without lengthening the actual pulse sequence and incorporating sensitivity enhanced gradient 153Journal of Biomolecular NMR (2020) 74:147–159 1 3 echo in t3, there is no sensitivity loss involved in increasing the dimensionality of these Hα-detected experiments. Hence, the coherence transfer efficiencies provided with the corresponding 3D experiments by Mäntylahti etal. (2010, 2011) are directly comparable to 4D implementations shown in Fig.3. Indeed, by taking into account typical values of one-bond couplings 1JCαC′ = 53Hz, 1JC′N = 15Hz, and 1JCαCβ = 35Hz, and the average random coil values for one-bond (1JCαN = 10.6Hz) and two-bond (2JCαN = 7.5Hz) couplings between backbone 13Cα and 15N spins (Delaglio etal. 1991), as well as transverse relaxation times (T2) for 13Cα (= 100ms), 13C′ (= 200ms) and 15N (= 200ms) spins, we can estimate coherence transfer efficiencies for these experiments in IDPs (Mäntylahti etal. 2010, 2011). The HACACON is superior in sensitivity (I ~ 0.28) in comparison to iHACANCO and (HACA)CONCAHA experiments, with coherence transfer efficiencies of 0.22 and 0.18, respectively. Particularly, for the assignment of prolines, sensitivities of iHACANCO and HACACON are superior to the (HACA)CONCAHA scheme, which yields coherence transfer of 0.026 for proline residues. The sensitivity loss is associated with the 15N(i)→ 13Cα(i) transfer, 2TNC, during which the 15N magnetization is further modulated by 1JNCδ coupling interaction in prolines. Figure4 compares 3D Hα-detected with the new 4D spectra: problems associated with multifold overlap and ambiguities in choosing the right sequential connection when using 3D spectra can be surpassed, and the assignment procedure expedited by extending frequency labeling to Cα. With the help of these Hα-detected 4D experiments, 76% of the α correlations of were successfully assigned (Suppl. Fig S2). Considering that some N, C′ resonances are separated by less than 0.05ppm in the CON spectrum, e.g. corresponding resonances within73/113KDAISE78/118 differ only by 0.03–0.04ppm, all peaks were not expected to be resolved in spectra with spectral resolutions of 0.11 (15N) and 0.05 (13C) ppm. The Hα, Cα shifts were as ineffective in providing the needed dispersion, the smallest peak separations being comparable, e.g. the 76/116Ile alpha peaks differ by only 0.01 and 0.05ppm in 1H and 13C, respectively, and those of 77/117Ser even less (Suppl. Fig. S1). 4D (HACA)CON(CA)NH and(HACA)N(CA)CONH experiments Analogously to 4D Hα-detected experiments described above, the 4D HN-detected experiments (Fig.5), with an isolated proline assignment enhancement, are based on their A134Ha-C’-A135N A97Ha-C’-A98N 177.9 178.0 15N 124.9 ppm 15N 124.70 ppm 15N 124.60 ppm 3D HA(CA)CON3D iHA(CA)NCO 3D iHA(CA)NCO 3D iHA(CA)NCO 4D (HACA)CONCAHA 4D (HACA)CONCAHA4D HACACON 4D HACACON 13C (ppm) A135Ha-Ca-N-A134C’ A98Ha-Ca-N-A97C’ 52.5 53.0 A97Ha-Ca-N-E96C’ A94N-A93Ha-Ca-C’ A135N-A134Ha-Ca-C’ A98N-C’A97Ha-Ca4.204.25 E96Ha-Ca-A97N C’- 56.5 57.0 57.5 1 H (ppm) 13C (ppm) 1513 N 122.92 ppm,C 177.90 ppm 15 13 N 122.92 ppm,C 177.90 ppm 15 13 N 124.64 ppm,C 176.53 ppm 15 13 N 124.73 ppm,C 176.60 ppm 1H (ppm) 15N 122.92 ppm 4.204.25 4.204.25 4.204.25 4.204.25 177.8 4.30 4.204.25 4.30 4.204.25 4.30 4.154.204.25 Fig. 4 Problematics in resonance assignment with 3D Hα-detected experiments can be resolved with 4D HACACON and 4D (HACA) CONCAHA experiments. The upper row of 2D planes from 3D HA(CA)CON and iHA(CA)NCO spectra shows that the sequential walk from 97Ala to 96Glu is ambiguous because the 97Ala Ha, C′ shift pair observed in the HA(CA)CON can be found in several planes of the iHA(CA)NCO and is overlapping with other peaks. While there is overlap in the 4D spectra also, the planes are far more easily interpreted and allow for unambiguous assignment of 98Ala-97Ala-96Glu. Grey labels mark peaks with maximum in adjacent 15N or 13C planes 154 Journal of Biomolecular NMR (2020) 74:147–159 1 3 3D counterparts (Hellman etal. 2014). Again, sampling of an additional 15N dimension can be implemented in without introducing sensitivity loss thanks to the gradient enhanced coherence order selective coherence transfer (COS-CT) (Kay etal. 1992; Schleucher etal. 1994). Magnetization transfer through (HACA)CON(CA)NH and (HACA)N(CA)CONH experiments are briefly described in Eqs.4 and 5, respectively: (4) 1 H𝛼(i−1) 2𝜏 (1 JC𝛼H𝛼 ) �������������������������������������→ 13C𝛼(i−1) 2TC (1 JC𝛼C� ) �����������������������������������������→ 13 C�(i−1)[2TC�N−t1;1JC�N]→15N(i)[t2] →13C𝛼(i−1) 2TCAN (1JC𝛼N,2JC𝛼N) �����������������������������������������������������������������→ 15N(i−1) [ 2TNCA −t 3 ;1JC𝛼N,2JC𝛼N ] 4Δ(1JNH ) ���������������������������������→ 1HN(i−1)[t 4] Like in the Hα-detected experiments (vide supra), the magnetization is first transferred from 1Hα to 13Cα spin (time point a), and further to either 13C′(i–1) spin in (HACA)CON(CA)NH (density operator Cαz(i–1)C′y(i–1)) or selectively to 15N(i) spin in (HACA)N(CA)CONH (density operator Cαz(i)Ny(i)) at time point b. This is followed by the frequency labeling of 13C′(i–1) and 15N(i) chemical shifts during t1 in (HACA)CON(CA)NH and (5) 1 H𝛼(i) 2𝜏 (1 JC𝛼H𝛼 ) �������������������������������������→ 13C𝛼(i) 2TCN (1 JC𝛼N, 2 JC𝛼N ) ��������������������������������������������������������������→ 15N(i)[t1 ] →13C𝛼(i) 2TCAN ,2TC(1JC𝛼N,2JC𝛼N,1JC𝛼C�) �����������������������������������������������������������������������������������������������������→ 13C�(i) [ 2TC−t2;1JC𝛼C�]→15N(i+1)[2TNC −t3;1JC�N] 4Δ(1JNH ) ���������������������������������→ 1HN ( i + 1 )[ t 4] ab c d