Introduction

Specific labeling of methyl groups (Ruschak et al. 2010; Rosenzweig and Kay 2014; Kerfah et al. 2015a), combined with optimized NMR spectroscopy (Tugarinov et al. 2003; Amero et al. 2009), provides a powerful tool to study structure, dynamics, interaction and mechanism of large protein assemblies (Sprangers and Kay 2007; Rosenzweig et al. 2013; Macek et al. 2017; Mas et al. 2018). Protonation of methyl probes in a perdeuterated protein background enables the detection of long range NOEs (nuclear overhauser effects) between methyl groups up to 12 Å apart in high molecular weight proteins (Sounier et al 2007; Kerfah et al. 2015b; Chi et al. 2018; Ayala et al. submitted). These distance restraints are particularly useful to determine the structure of large proteins and protein complexes (Tugarinov et al. 2005a; Gauto et al. 2019). If the tertiary structure of a protein is previously known, the network of NOE connectivities can also be exploited for automated NMR assignment of methyl resonances (Xu and Matthews 2013; Chao et al. 2014; Pritišanac et al. 2017; Monneau et al. 2017; Pritišanac et al. 2019). In protein complexes, special care should be taken to separate the large number of intra-chain NOEs from the few inter-chain signals, as any erroneous interpretation could significantly impact NMR assignment or structure determination (Werner et al. 1997). Separation of intra-chain and inter-chain NOEs is particularly challenging in homo-oligomeric complexes.

Standard approaches to identify inter-chain NOEs in protein complexes use asymmetric isotope labeling of the different subunits with either 15 N or 13C in order to specifically filter/edit inter-chain NOEs (Handel and Domaille 1996; Zwahlen et al. 1997; Caffrey et al. 1998; Anglister et al. 2016). Such strategies have proven to be very efficient for unambiguous extraction of inter-molecular distance restraints, but they suffer from fast proton transverse relaxation, and are therefore limited to complexes with a molecular weight of less than about 35–40 kDa. Perdeuteration combined with asymmetric isotope labeling of methyl groups has also been proposed to identify intermolecular NOEs. Veglia and co-workers have used the fact that isoleucine-δ1, and leucine-δ/valine-γ methyl groups can be labeled separately, by supplementing culture media with different keto-acid precursors (Gardner and Kay 1997; Goto et al. 1999; Tugarinov et al. 2006), to introduce an asymmetry between protomers (Traaseth et al. 2008). By reconstituting a homo-pentamer from a mix of protomers labeled either on isoleucine-δ1 or leucine-δ/valine-γ methyl groups, the authors were able to unambiguously identify inter-chain NOEs between I and V/L methyl groups. While this approach allows to extract NOEs between distant methyl groups in large protein complexes, only inter-chain NOEs between I–V and I–L pairs can be identified among the 36 possible pairs of methyl containing residues, reducing significantly the number of structurally meaningful inter-molecular distance restraints.

Herein, we introduce an alternative approach that allows to distinguish intra-chain NOEs from intermolecular connectivities. The method is based on the labeling of the different subunits with either 12C–13CH3 or 13C–13CH3 isotopomers, and the acquisition of 3D-NOESY experiments, in which both methyl 13C frequencies are edited in constant time mode. In hetero-oligomeric complexes, this strategy enables simultaneous editing of both intra-chain and inter-chain NOEs, with an opposite sign. This allows immediate and unambiguous identification of the ‘precious’ inter-molecular distance restraints between methyl groups of alanine, isoleucine, leucine, methionine, threonine and valine. For the even more challenging case of large homo-oligomeric assemblies, we demonstrate that our strategy allows to suppress signals of inter-chain NOEs in a high-resolution 3D NOESY spectrum, in order to detect exclusively intra-chain NOEs between methyl groups of alanine, valine, leucine and isoleucine. Principles, limits and possibilities of this approach are presented together with applications to the 87 kDa hetero-hexameric prefoldin complex and the 53 kDa homo-tetramer formed by prefoldin β-subunits.

Material and methods

Protein expression and purification

E. coli BL21 (DE3) cells transformed with pET23c plasmids encoding either for the α- or β-subunit of Prefoldin (denoted here PFD but also known as co-chaperonin) from Pyrococcus horikoshii (denoted Ph), were progressively adapted to M9/2H2O media in three stages over 24 h (M9 in H2O, M9 in 50% H20 + 50% 2H2O, M9 in 2H2O). In the final culture the bacteria were grown at 37 °C in M9 media prepared with 99.85% 2H2O (Eurisotop), 2 g/L of deuterated d-glucose and 1 g/L15NH4Cl (CIL).

  • For production of the U-[2H], A-[13CH3]β, I-[13CH3]δ1, L-[13CH3]δ1, V-[13CH3]γ1 PFD-β subunit, 2 g/L of d-glucose-d7 (Sigma Aldrich) was used as carbon source. When the O.D. at 600 nm reached 0.7, a solution containing 2-[2H3]methyl-2,4-[13C2]-acetolactate (NMR-Bio) was added for the stereoselective labeling of proR Leuδ1 and Valγ1 (Mas et al. 2013) methyl groups (240 mg/L). 40 min later 3-[13C]-2-[2H]-L-Alanine and (S)-2-hydroxy-2-(2′-[13C],1′-[2H2])ethyl-3-oxo-4-[2H3]-butanoic acid (NMR-Bio) were added to a final concentration of 250 mg/L and 100 mg/L respectively (Kerfah et al. 2015b) for the simultaneous labeling of Ileδ1 and Alaβ methyl groups, and 20 min afterwards protein production was induced by IPTG. U-[2H], A-[13CH3]β, I-[13CH3]γ1, L-[13CH3]δ2, V-[13CH3]γ2, T-[13CH3]γ2 PFD-β subunit sample was produced using a similar protocol with 2-[13C]methyl-4-[2H3]-acetolactate (NMR-Bio) for the stereoselective labeling of proS (Leuδ2 and Valγ2; Gans et al. 2010) and addition of 50 mg/L of 2,3-(2H) 4-(13C)-l-Threonine (Ayala et al. submitted) together with the other precursors 20 min before induction.

  • For production of the U-[2H, 13C], A-[13CH3]β, I-[13CH3]δ1, L-[13CH3]δ1, V-[13CH3]γ1 PFD-β and PFD-α subunits, 2 g/L of U-[2H,13C]-glucose was used as carbon source (CIL). When the O.D. at 600 nm reached 0.7, a solution containing 2-[2H3]methyl-1,2,3,4-[13C4]-acetolactate (NMR-Bio) was added (240 mg/L), and 40 min later U-[13C]-2-(2H)-l-Alanine and (S)-2-hydroxy-2-(1′,1′-[2H2],1′,2′-[13C2])ethyl-3-oxo-1,2,3-[13C3]-4,4,4-[2H3]butanoic acid were added to a final concentration of 250 mg/L and 100 mg/L, respectively (Kerfah et al. 2015b). 20 min afterwards protein production was induced by IPTG.

The expression was performed at 37 °C for 3 h, then cells were collected by centrifugation at 7000×g for 15 min at 4 °C. The purification protocol of prefoldin subunits was adapted from (Okochi et al. 2002). The cells, corresponding to 1 L of culture, were resuspended in 50 ml of buffer A containing 50 mM Tris pH 8.0, 0.1 mM EDTA, and complemented with 0.025 mg/mL RNAse (Euromedex), 0.025 mg/mL DNAse (Sigma Aldrich) and 1 anti-protease tablet (cOmplete™), and disrupted by sonication on ice. After removal of cell debris by centrifugation (40,000×g, 30 min, 4 °C), the supernatant was heated to 80 °C for 30 min and precipitated proteins were removed by centrifugation. Then, the supernatant was loaded on a Q Sepharose—XK 26/20 column (Sigma-Aldrich), pre-equilibrated with buffer A. With a linear gradient the protein was eluted at 20% of buffer B (50 mM Tris pH 8.0, 0.1 mM EDTA, 500 mM NaCl) and the combined fractions were loaded on a Superdex 200 gel filtration column (Sigma Aldrich) equilibrated with the NMR buffer (50 mM Tris, 100 mM NaCl, pH 8.5).

  • For preparation of U-[2H], A-[13CH3]β, I-[13CH3]γ1, L-[13CH3]δ1, V-[13CH3]γ1 PFD-β4 sample (denoted β4-13CH3), the gel filtration fractions containing PFD subunits were concentrated in NMR buffer using Amicon® Centrifugal Filter Units (Merck) to 200 μM (concentration of the β4 oligomer). A 200 µL aliquot was lyophilized and re-suspended in 200 µL of ultra-pure D2O before loading in a 4 mm shigemi tube.

  • For preparation of the β4 mixed-labeled sample (denoted β4-mix), the β-subunits with the two type of labeling (i.e. U-[2H], A-[13CH3]β, I-[13CH3]δ1, L-[13CH3]δ1, V-[13CH3]γ1 and U-[2H,13C], A-[13CH3]β, I-[13CH3]δ1, L-[13CH3]δ1, V-[13CH3]γ1) were combined in a 1:1 ratio, concentrated to 200 μM (total concentration of the β4 oligomer), and incubated at 70 °C for 3 days to ensure mixing, and therefore a statistic distribution of both types of labeling schemes in the resulting β4 oligomers. This sample was lyophilized and re-suspended in ultra-pure D2O before loading in a 4 mm shigemi-tube.

  • For preparation of PFD-α2β4 sample, purified U-[2H, 13C], A-[13CH3]β, I-[13CH3]δ1, L-[13CH3]δ1, V-[13CH3]γ1 α-subunits and purified U-[2H], A-[13CH3]β, I-[13CH3]δ1, L-[13CH3]δ2, V-[13CH3]γ2, T-[13CH3]γβ-subunits were combined in a ratio of 1:1 and incubated at 80 °C for 30 min. Excess α2 subunits were removed by gel-filtration on a Superdex200 column, the fractions containing PFD (α2β4) were combined, concentrated, lyophilized and re-suspended in ultra-pure D2O before loading in a 4 mm shigemi tube. The final concentration of PFD (α2β4) was 225 µM.

  • U-[2H], A-[13CH3]β, I-[13CH3]δ1, M-[13CH3]ε Malate Synthase G (denoted MSG) sample was produced and purified as described previously (Kerfah et al. 2015b). For the specific labeling of methyl groups, deuterated l-methionine specifically 13CH3-labeled on ε position, 3-[13C]-2-[2H]-l-Alanine and (S)-2-hydroxy-2-(2′-[13C],1′-[2H2])ethyl-3-oxo-4-[2H3]-butanoic acid (NMR-Bio) were added 1 h prior to IPTG induction. The MSG protein was concentrated to 0.5 mM in 100% 2H2O buffer containing 25 mM MES (pH 7.0 uncorrected), 20 mM MgCl2, 5 mM DTT. 200 µL were loaded in a 4 mm shigemi tube.

NMR spectroscopy

All the 3D 13C edited-NOESY experiments were acquired on a Bruker Avance III HD spectrometer operating at a 1H frequency of 950 MHz and equipped with a cryogenic probe. For all experiments, the interscan delay was set to 1.25 s, the heteronuclear 1H- > 13C transfer delay was set to 4 ms (0.5/1JHC) and the acquisition times in 13C indirect dimension were fixed to 23.5 ms (t1max and t2max) and in the 1H detection dimension (t3max) to 60 ms. The 3D HMQC-NOESY-HMQC experiment (Tugarinov et al. 2005b) was implemented using EBURP2 (Geen and Freeman 1991) as 90° excitation pulse and time reverse EBURB2t was used to flip-back 1H magnetization on the z-axis before the NOE mixing period (Lescop et al. 2010). The NOE mixing period was set to 600 ms to detect a maximum number of long range NOEs. 1H chemical shift evolution during the HMQC was refocused using a pair of high power band-selective BIP-720-50-20 1H pulses (Kupce and Freeman 1994; Amero et al. 2009). The 3D CT2-HMQC-NOESY-HMQC pulse scheme (Fig. 1a) was derived by adding a pair of 180° 13C hard pulses during the constant time 13C editing period (Powers et al. 1991) fixed to 25 ms. The constant time editing allows to selectively invert the sign of 13CH3 multiquantum coherences depending on the nature of the covalently bound carbon (12C or 13C). The acquisition time for 3D NOESY experiments varied from 2 to 3 days. All the NMR data were acquired at 70 °C for PFD samples from hyperthermophilic P. horikoshii or 37 °C for MSG sample.

Fig. 1
figure 1

Specific editing of intra vs. inter-chain methyl–methyl NOEs. a 3D CT2-HMQC-NOESY-HMQC pulse sequence. Excitation of proton magnetization was achieved using 1.2 ms selective 90° EBURP2 pulses covering a bandwidth of 4 ppm (Geen and Freeman 1991) and was followed by a delay δ = 4 ms for heteronuclear magnetization transfer. Methyl carbon frequencies (F1 and F2) were encoded in constant time mode during the two multi-quantum evolution periods (Δ = 25 ms) flanking the NOE mixing period (tmix = 600 ms). 1H chemical shift evolution during each HMQC element was refocused using a pair of high power band-selective BIP-720-50-20 1H pulses (Kupce and Freeman 1994; Amero et al. 2009). 1.2 ms time reverse EBURB2 pulse (Lescop et al. 2010) was used to flip-back 1H magnetization on the z-axis before the NOE mixing period. Narrow and wide bars indicate nonselective 90° and 180° 13C pulses. 13C decoupling during 1H signal detection (F3) was achieved using waltz65 decoupling pulse sequence (Zhou et al. 2007). All the pulses were applied along x axis unless indicated otherwise on the figure. Phase cycling was performed as follows: Φ1 = x, − x; Φ2 = x, y; Φ3 = x, x, − x, − x; Φ4 = x, x, y, y; and receiver phase was cycled with Φrec. = x, − x, − x, x. Recycling delay was set to 1.25 s and 200 ms gradients were applied on z-axis with magnitude G1 = 19 G/cm; G2 = 7 G/cm; G3 = 13 G/cm and G4 = 11 G/cm. b Schematic representation of inter-subunit NOE sign encoding principles. This application is restricted to hetero-oligomeric complexes and is enabled by reconstituting the complex from the two different subunits labeled either with a 13C–13CH3 isotopomer (α-subunits) or with 12C–13CH3 groups for the β-subunits. Such a sample was used to acquire a 3D 13C-edited NOESY experiment with the pulse scheme presented on Fig. 1a. NOEs between two methyl groups of α-subunits or two methyls of β-subunits are edited as positive signals, while inter-subunit NOEs between methyl groups of α- and β-subunits have a negative sign due to the evolution of the 1JCC coupling during only one of the two CT-HMQC elements. c Schematic representation of the principles of specific editing of intra-chain NOEs. This application is restricted to homo-oligomeric complexes and is enabled by reconstituting the complex from a mix in a ratio 1:1 of the same subunit fully enriched with either 13C–13CH3 or 12C–13CH3 isotopomers. NOEs between two methyl groups of the same chain are positive whatever is the labeling scheme of the chain. Inter-chain NOEs between two methyl groups are either positive or negative due to 1JCC evolution and cancel each other, leading to the suppression of inter-chain NOEs

Full size image

Data processing and analysis

All data were processed and analyzed using nmrPipe/nmrDraw (Delaglio et al. 1995) and CcpNMR (Vranken et al. 2005). For all 3D spectra, the number of points in both 13C indirect dimensions were doubled using forward–backward linear prediction. A cos2 apodization function was applied on all FIDs, before zero-filling and Fourier transform. For the identification of NOEs in all the 3D NOESY experiments, a S/N threshold of 3 was used.

Automated methyl assignment was performed using MAGIC software (Monneau et al. 2017) with the 3D structure of the PFD-β4 complex, built by homology modeling using MODELLER (Šali and Blundell 1993) implemented in the NMRBox server (Maciejewski et al. 2017). The homology model is based on the X-ray crystallography structure of the PFD-β4 complex from Thermococcus sp. strain KS-1 (Kida et al. 2008, PDB:2ZQM) whose β-subunit is 83.4% identical and 94.9% similar to the β-subunit of Ph.PFD. Input NOE lists for MAGIC were created with CcpNMR. MAGIC was run with a score threshold factor of 1 and distance thresholds of 7–10 Å using all inter methyl NOEs detected.

Results

In large perdeuterated proteins, selectively labeled on methyl groups, long-range methyl–methyl NOEs are optimally detected using 3D or 4D HMQC-NOESY-HMQC-type experiments (Tugarinov et al. 2005b; Kerfah et al. 2015b) preserving the methyl-TROSY effect (Tugarinov et al. 2003). The 3D (13C, 13C, 1H)-edited CT2-HMQC-NOESY-HMQC experiment used here is shown in Fig. 1a. The pulse sequence uses selective 1H pulses covering only methyl resonances (Amero et al. 2009) together with constant time (CT) editing of methyl 13C frequencies (Powers et al. 1991) before and after the NOE mixing period. The CT period Δ was set to 25 ms in order to invert the sign of 13C-bound methyl groups (i.e. 13C–13CH3 isotopomers) with respect to 12C-bound methyl (i.e. 12C–13CH3 isotopomers). As the 13C-frequencies are edited twice in this 3D experiment, all diagonal and NOE cross-peaks between methyl groups labeled with the same isotopomer are positive (Fig. 1b), while negative NOE cross peaks are detected between methyl groups that are labeled with different methyl isotopomers (one bound to 12C, the other one to 13C).

Sign encoding of intra- and inter-chain NOEs in hetero-oligomeric complexes

First, we have used the above described 3D CT2-HMQC-NOESY-HMQC experiment to identify interchain NOEs in Pyrococcus horikoshii (Ph) prefoldin (PFD), a co-chaperonin forming a hetero-hexameric assembly of 87 kDa with 2α and 4β subunits arranged in a ring-like structure (Fig. 2a). The two α-subunits are structurally and spectroscopically equivalent, both preceded and followed by β-subunits. The four β-subunits have the same sequence, but are not structurally equivalent, as they can either be preceded or followed by an α-subunit in the assembly, and therefore give rise to two different sets of NMR signals (denoted β and β’). Combinatorial labeling of several methyl containing amino acids represents an obvious route for increasing the number of NMR probes in the sample. To date, only alanine, isoleucine, leucine and valine are available in the 13C–13CH3 isotopomeric form (Gans et al. 2010; Mas et al. 2013; Miyanoiri et al. 2013; Kerfah et al. 2015a). Therefore, we have labeled the methyl groups of alanine (Ala-β), isoleucine (Ile-δ1), leucine (Leu-δ1) and valine (Val-γ1) in the α-subunits with 13C–13CH3 isotopomers, while in the β-subunits alanine (Ala-β), isoleucine (Ile-δ1), leucine (Leu-δ2), valine (Val-γ2) and threonine (Thr-γ) were enriched with 12C–13CH3 isotopomers. Stereo-selective isotope labeling of methyl groups in valine, isoleucine, and leucine provided improved spectral resolution, and enabled the detection of long-range NOEs (Gans et al. 2010; Kerfah et al. 2015a, b).

Fig. 2
figure 2

Sign encoding of inter-subunit NOEs in prefoldin. a Structure of prefoldin from Pyrococcus hirokoshii (Ohtaki et al. 2008), an 87 kDa heterohexameric co-chaperonin. The two α-subunits (dark blue) are structurally and spectroscopically equivalent, whereas the β-subunits (cyan) adopt a different structure depending on their position in the ring-like assembly, resulting in a different set of NOE peaks, here named β and β’. Zooms around the δ1-methyl groups from Ile-43 (β and β’) and Ile-52 (β), are showing nearby methyl groups. b Examples of 2D extracts from the 3D CT2-HMQC-NOESY-HMQC experiment (Fig. 1a) acquired using a NOE mixing time tmix = 600 ms on a perdeuterated sample of prefoldin in which 13C–13CH3 isotopomers were introduced in Ala, Ile, Leu, and Val side chains of the α-subunits and 12C–13CH3 isotopomers were introduced in Ala, Ile, Leu, Thr and Val side chains of the β-subunits (Fig. 1b). 2D planes extracted at the 1H frequencies of δ1-methyl from Ile-52 (α and Ile-43(β), both located at the interface between of the α- and β-subunits, show both positive and negative NOEs cross-peaks, corresponding to intra- and inter-subunit NOEs respectively. Only positive NOEs are detected in the 2D slice corresponding to δ1-methyl from Ile-43 (β’), which is surrounded by methyls from β and β’ subunits

Full size image

Using the above described labeling scheme, a high-resolution 3D CT2-HMQC-NOESY-HMQC spectrum was acquired on the PFD-α2β4 sample at 950 MHz 1H frequency in an experimental time of 2 days. Examples of 2D slices extracted for methyl residues located either in the core of the subunits or at the α/β interface are presented in Fig. 2. It can be seen that I43-δ1 in the β’ subunit only gives rise to positive NOEs. Because this residue is far from the α/β interface, only NOEs with methyl groups in β’-type subunits are observed. I43-δ1 in the β-subunit however is located in sufficient proximity to 13C–13CH3 labeled methyl groups in the α-subunit to give also rise to negative inter-chain NOEs. Another example showing intra-chain (positive) and inter-chain (negative) NOEs is I52-δ1 in the α-subunit. Complete analysis of the 3D NOESY spectrum allowed the extraction of 237 NOEs, 20 of them corresponding to negative cross peaks. These structurally meaningful inter-chains NOEs can be distinguished easily from the large number of intra-chain connectivities without requiring a tedious analysis of the 3D NOESY connectivity matrix. Such a set of NOE-derived distance restraints is particularly useful to characterize the structure of protein complexes using solution NMR.

As the sign inversion of inter-chain NOEs with respect to intra-chain signals relies on the possibility to label different subunits with distinct isotope labeling schemes, only NOEs between the α and β chains are selectively edited as negative signals, while NOEs between two homologous chains (two α or two β/β’ chains) remain positive and cannot be easily distinguished from intra-chain signals.

Selective intra-chain NOE detection in homo-oligomeric complexes

In order to selectively detect intra-chain NOEs in homo-oligomeric complexes, we have again exploited the possibility of modulating the sign of NOE cross peaks with the CT2-HMQC-NOESY-HMQC experiment using an appropriate labeling scheme. For this purpose, a 13C–13CH3 labeled sample was mixed in a 1:1 ratio with a 12C–13CH3 labeled sample, both in an otherwise deuterated background. The reconstituted homo-oligomeric complex contains a statistic distribution of each type of isotopomer. Figure 1c shows the expected outcome of a CT2-HMQC-NOESY-HMQC performed on such a specifically labeled homo-oligomer. Positive intra-chain NOEs are detected whatever the labeling of the interacting methyl groups. For inter-chain connectivities, the NOE can be either positive if the two subunits are enriched with the same isotopomer, or negative if the two subunits are enriched with different isotopomers. Therefore, in a perfect 1:1 mixture, positive and negative signals cancel each other, and no inter-chain NOEs are detected in the final spectrum. In order to increase the amount of detected NOE connectivities, a maximum number of methyl-containing amino acids should be labeled, while stereospecific labeling of isoleucine, leucine and valine is used again to reduce spectral overlap and favor the detection of long range NOEs between remote amino acids (Gans et al. 2010; Kerfah et al. 2015b).

We have applied this strategy to the 53 kDa homo-tetrameric prefoldin 4-complex. In absence of α-subunits, the β-subunits of archaeal prefoldin form a symmetric homo-tetramer as shown for the β4-complex from Thermococcus KS1 (Kida et al. 2008). Due to 94.9% sequence similarity, the P. horikoshii β4-PFD is expected to behave similarly, and its structure (Fig. 3a) was calculated by homology modeling using Thermococcus KS1 PFD-β4 as a reference. For this application, we have chosen to label the 40 Ala-β, Ile-δ1, Leu-δ1 and Val-γ1 methyl positions, as these residues can be labeled simultaneously with both type of isotopomers (Gans et al. 2010; Mas et al. 2013; Miyanoiri et al. 2013; Kerfah et al. 2015a). In a first sample 13CH3 groups were incorporated at the 40 methyl positions in a 12C, 2H background (β4-13CH3 regular methyl labeled sample). The second β4-PFD sample was reconstituted by mixing an equimolar ratio of two types of β-subunits labeled on Ala-β, Ile-δ1, Leu-δ1 and Val-γ1 positions with either 12C–13CH3 or 13C–13CH3 isotopomer (β4-mixed labeled sample).

Fig. 3
figure 3

Selective editing of intra-chain NOEs in the 53 kDa PFD-β4 homotetramer. a Homology model of the tetramer formed by only four β-subunits of PFD from Pyrococcus horikoshii. Each one of the four β-subunit copies are represented by different colors. b Zoom around the δ1-methyl from Ile-46, showing nearby methyl groups belonging to the same chain or a different β-subunit chain. c 2D planes extracted at the 1H frequency of δ1-methyl from Ile-46 from a 3D CT2-HMQC-NOESY-HMQC experiment (Fig. 1a) acquired with a NOE mixing period tmix = 600 ms on either regular methyl labeled (left), or mixed-labeled PFD-β4 samples (right). d 2D plane extracted at the 1H frequency of Ile-46 δ1-methyl from a 3D HMQC-NOESY-HMQC experiment acquired with a NOE mixing time tmix = 600 ms on the regular methyl labeled sample

Full size image

3D spectra were acquired for both samples using the CT2-HMQC-NOESY-HMQC pulse sequence of Fig. 1a) on a spectrometer operating at a 1H frequency of 950 MHz. 2D strips extracted from these 3D spectra for I46-δ1 are shown in Fig. 3c. I46-δ1 is located at the interface between the two β-subunits. This methyl group is part of a methyl cluster involving 3 methyl containing residues belonging to a different subunit (V59-γ1, V64-γ1 and L62-δ1) distant by less than 8 Å from the I46-δ1 methyl group (Fig. 3b). A total of 8 NOEs were detected between I46-δ1 and other methyl groups using standard β4-13CH3 labeled PFD sample. As expected, the 3 inter-chain NOEs are missing in the 3D spectrum acquired on the β4-mix labeled sample (Fig. 3c). Full analysis of the 3D CT2-HMQC-NOESY-HMQC spectra allowed the identification of 99 NOE cross peaks, corresponding to 60 different pairs of methyl groups, for the β4-13CH3 sample, while 27 of these NOEs, corresponding to 14 inter-chain methyl pairs, are missing in the spectrum recorded for the β4-mixed labeled sample. Interestingly, inspection of the spectrum acquired using the β4-mixed labeled sample at very low contour level does not reveal any residual signal at the positions of the missing inter-chain NOEs (Fig. S1), demonstrating the efficiency of the proposed approach to suppress selectively inter-chain NOEs.

Filter efficiency for homo-oligomeric complexes

The selective cancellation of inter-chain NOEs in homo-oligomeric complexes relies on the inversion of 50% of the NMR signal corresponding to particular methyl isotopomers (Fig. 1c). Several factors may affect the efficiency of this filter: (i) J-mismatch. The theoretical CT period Δ for a perfect sign inversion of 13C-bound methyl group signals is Δ = 1/1JCC = 28.6 ms, with 1JCC ≈ 35 Hz for the methyl groups. As shown in Fig. 4a, the filter efficiency to suppress signals corresponding to inter-chain NOEs does not significantly vary with the value of Δ, and a variation of the constant time period by 5 ms still allows to reduce the intensity of inter-chain NOEs by more than two orders of magnitude. Intensities of intra-chain NOEs depend more strongly on Δ and should be optimized taking into account relaxation losses. Calculation indicates that the optimal CT period Δ is on the order of 24–25 ms (Fig. S2) for methyl groups characterized by average transversal relaxation rates of 25 Hz (corresponding to the expected average methyl relaxation rates of a protein with a τC of 40 ns). Therefore, for applications to large proteins, we propose to use a constant time delay Δ of 25 ms, preserving signal intensities of selected intra-chain NOEs in large proteins and providing a good suppression of inter-chain connectivities. (ii) Imperfect sample preparation. This can be due to an imbalance between the two labeling schemes, or an incomplete incorporation of precursors. While an imbalance in the two differently labeled (12C–13CH3 or 13C–13CH3) protein chains does not affect the intensity of intra-chain NOEs (Fig. 4b), incomplete labeling of methyl groups will affect the intensity of both, intra-chain and inter-chain NOEs (Fig. 4c). Figure 4b, c shows the calculated residual intensity of inter-chain NOEs due to imperfect isotope labeling. In the extreme case of an isotope incorporation level of only 66%, or a 1:2 imbalance of the two isotopomers in the complex, the intensities of inter-chain NOEs are still reduced by a factor of 10, demonstrating the robustness of this filter with respect to errors in sample preparation. In case of doubts, the acquisition of a conventional 3D HMQC-NOESY-HMQC on the same sample, should allow to discriminate if a weak signal corresponds to the residual intensity of a strong inter-chain or a weak intra-chain NOE.

Fig. 4
figure 4

Filter efficiency. a Effect of constant time delays (Δ) on the efficiency of the 3D CT2-HMQC-NOESY-HMQC experiment to preserve signals of intra-chain NOEs and to suppress signals of inter-chain NOEs. b Effect of an imbalance in the ratio of the two labeled chains. The blue curve represents the efficiency of the suppression of inter-chain NOEs as function of xA: the fraction of chain labeled with 12C–13CH3 isotopomers (1 − xA being the fraction of chain enriched with 13C–13CH3 isotopomers). The red curve depicts the effect on intra-chain NOEs. c Effect of imperfect isotope labeling. A low incorporation of 12C–13CH3 or 13C–13CH3 isotopomers (IA) in one type of labeled chains will not only lead to lower intra-chain NOE intensities (red curve), but also to partial suppression of inter-chain NOEs (blue curve)

Full size image

NOE-based methyl NMR assignment of the prefoldin β4 complex

A particularly interesting application of methyl NOEs is their use, together with an available 3D structure, for NMR assignment of methyl resonances, a prerequisite for functional, interaction and dynamic studies of a target protein. Here, we have used the measured long-range NOEs (Fig. 5a) together with the 3D structure of prefoldin-β4 (Fig. 3a) to assign Ala-β, Ile-δ1, Leu-δ1 and Val-γ1 methyl groups of the tetrameric complex aided by the software MAGIC (Monneau et al. 2017). 39 NMR signals are detected in a 2D methyl-TROSY spectrum (Tugarinov et al. 2003; Amero et al. 2009) of PFD-β4, indicating that only one methyl group is not detected or not visible due to peak overlap. In a first step, we identified the amino-acid-type of each methyl group giving rise to a NMR signal using a 3D HCC experiment (Ayala et al. 2009, 2012). While the isoleucine-δ1 methyl groups can be unambiguously identified from their particular chemical shifts (Fig. 5b), alanine-β, valine-γ and Leucine-δ methyl groups were identified based on the characteristic frequencies of the neighboring Ala-Cα, Val-Cβ and Leu-Cγ carbon frequencies (Fig. S3). With this residue-type information in hand, the MAGIC input was created from the full set of detected (inter- and intra-chain) NOEs and the β4 structure. A particular strength of the MAGIC algorithm is to provide information about the ambiguity of the calculated assignment and to propose multiple assignments in case of insufficient information. In our case, only 15% of the methyl resonances could be unambiguously assigned by the software, five were left without assignment (three of those due to missing NOE cross-peaks), and the remaining ones were given multiple assignments. Taking the software-generated assignment as a starting point, the assignment was completed manually. At this point, the information regarding inter-chain and intra-chain connectivities was introduced. In a first step, it was used to distinguish residues at the interface of the different β-chains from residues further away from the interface. Furthermore, this additional information was particularly useful to resolve assignment ambiguities at the interface region, the upper part of the coiled coil helices and the β-barrel. For instance, some pairs of residues, such as Val-72 and Val-59-γ1, or Leu-62-δ1 and Leu-75-δ1, are characterized by a similar NOE connectivity network. These ambiguities could be sorted out, as one of the methyl groups is only connected by NOEs identified as intra-chain, whereas for the other one the respective connections were identified as inter-chain NOEs. Starting from unambiguously assigned methyl groups, the assignment was iteratively extended to other connected methyl groups, resulting in the unambiguous assignment of 36 out of the 39 observable methyl resonances (Fig. 5b–e). Finally, two of the three methyl resonances, for which no NOE cross-peaks were observed, Ile63-δ1 and Leu24-δ1, were assigned by exclusion principle, as only one possibility remained after all other peaks of the same methyl type were assigned. The last unassigned alanine resonance can correspond either to Ala-10 or Ala-116. Taking into consideration that this alanine methyl group gives rise to an intense signal and is connected in the 3D HCC spectra to a single Cα signal with a chemical shift characteristic of a random coil structure, this signal has been tentatively assigned to Ala-116 in the C-terminal unfolded segment of the protein.

Fig. 5
figure 5

NOEs based assignment of the methyl resonances in β4-prefoldin. a 2D representation of the PFD-β4 residue pairs for which inter-methyls NOEs are observed. The secondary structure elements of the protein are presented below the matrix. Top left part of the matrix corresponds to the 67 methyl pairs for which NOEs (113 cross-peaks) were detected using the standard 3D HMQC-NOESY-HMQC experiment acquired with a regular methyl labeled PFD-β4 sample. Bottom right part of the matrix presents in red the intra-chain NOEs (71 cross-peaks, 47 pairs) detected using 3D CT2-HMQC-NOESY-HMQC with regular and mixed labeled PFD-β4 sample. Inter-chain NOEs in black (28 cross-peaks, 14 pairs) are only detected when the 3D CT2-HMQC-NOESY-HMQC experiment is acquired with the regular sample. b–d 2D SOFAST-methyl-TROSY sub-spectra of PFD-β4 corresponding to Ile-δ1 (b), Val-γ1 (c), Ala-β (d) and Leu-δ1 (e) resonances. Assignment was obtained using MAGIC software (Monneau et al. 2017) based on the 3D structure of PFD-β4 and experimental methyl-methyl NOEs (a)

Full size image

Discussion

Spectral editing of intra- and inter-chain NOEs by the CT2-HMQC-NOESY-HMQC experiments relies on the use of two CT 13C editing periods. While this allows the recording of highly resolved, isotopomer-edited NOESY spectra for both, homo- and hetero-oligomeric complexes, the additional CT delays of about 25 ms will result in relaxation-induced signal loss. In order to quantify this signal loss for large proteins, we have used a sample of deuterated, methyl-labeled Malate Synthase G, an 82 kDa monomeric protein extensively studied by NMR (Tugarinov and Kay 2005a) and characterized by a rotational correlation time of τC = 39.2 ns at 37 °C in D2O buffer (Tugarinov and Kay 2005b). We have previously shown that methyl–methyl long range NOEs between probes up to 12 Å apart are detectable in such a protein, suitably labeled using the standard 3D-HMQC-NOESY-HMQC experiment (Kerfah et al. 2015b; Ayala et al. submitted). Using the same MSG sample and comparable acquisition and processing parameters, we acquired 3D CT2-HMQC-NOESY-HMQC (Fig. 1) and standard 3D-HMQC-NOESY-HMQC (Tugarinov et al. 2005b; Kerfah et al. 2015b) spectra. Comparing the signal intensities measured in these two spectra, we observed an average reduction of 65% of the signal intensities (Fig. S4A), in agreement with an estimated average MQ relaxation rate of 24 Hz (Fig. S4B). Taking into account that the NOE intensity decreases with the sixth power of the inter-methyl distance, we estimate that the two extra CT-periods in the 3D CT2-HMQC-NOESY-HMQC experiment will reduce the maximum distance for which a NOE cross peak can be detected by 15% for this protein (with a τc = 40 ns). Therefore, inter-methyl NOEs between methyl probes separated by up to 10 Å should be detectable with the 3D CT2-HMQC-NOESY-HMQC pulse scheme within 3 days on a modern high field NMR spectrometer (equipped with a cryogenic probe) using a 1 mM sample of well behaving protein such as the 82 kDa Malate Synthase G. Lower protein solubility, or presence of conformational exchange in the micro- to millisecond time scale, as observed for prefoldin samples (data not shown), affect the quality of the NMR spectra and reduces the theoretical distance for which inter-methyl NOEs can be detected. Full analysis of the 3D CT2-HMQC-NOESY-HMQC spectrum acquired with PFD-β4 sample indicates that NOEs can still be detected between methyl probes distant by up to 8 Å (Fig. S5). However, to extract distance restraints, the contribution of spin diffusion cannot be neglected as a relatively long mixing period (600 ms) was used to detect a maximum number of long range NOEs in these highly deuterated proteins. Full relaxation-matrix analysis is required to extract accurate distance restraints from such 3D NOESY spectra (Sounier et al. 2007; Vögeli 2014; Nichols et al. 2018).

For the applications presented here, we have selectively labeled methyl groups of Alanine, Isoleucine, Leucine and Valine, representing on average ca. 80% of methyl containing amino-acids in proteins. Current protocols only allow to selectively label Threonine with a 13CH3 group in an otherwise U-[12C, 2H]-labeled amino-acid (Velyvis et al. 2012; Ayala et al. submitted), but the approaches presented here could be further extended to Threonine residues when suitably labeled Threonine precursors will become available. Commercially available deuterated Methionine and Threonine, specifically enriched with 13CH3 groups, can be incorporated together with Alanine, Isoleucine, Leucine and Valine in protein enriched with 12C–13CH3 isotopomers, in order to increase the number of intra-chain and inter-chain NOEs (of opposite sign) in hetero-oligomeric complexes. It has been previously reported that the mixing of two proteins, labeled on different types of methyl-containing residues can be used to selectively detect inter-chain methyl-methyl NOEs between Isoleucine-δ1 and Valine-γ, as well as between Isoleucine-δ1 and Leucine-δ methyl groups (Traaseth et al. 2008). The approach proposed in this article allows to distinguish intra-chain and inter-chain NOEs simultaneously between all the methyl groups of Alanine, Isoleucine, Leucine and Valine, thus significantly increasing the number of both intramolecular and intermolecular distance restraints between remote methyl probes.

The selective detection of intra-chain NOEs in a homo-oligomeric complex requires the preparation of a mixed labeled sample with half of the subunits perdeuterated and labeled with 13C–13CH3 isotopomers and the remaining perdeuterated subunits labeled only with 13CH3 groups. Production of such a mixed labeled sample, with a statistic distribution of both labeling schemes in the complex, requires that the subunits are in exchange with a transient soluble monomer or between two distinct oligomers. Otherwise, a more demanding sample preparation, including unfolding of the native complex before mixing of both labeling schemes in a 1:1 ratio and re-oligomerization, need to be implemented (Macek et al. 2017). Such a sample can be used with the CT2-HMQC-NOESY-HMQC experiment (Fig. 1) to detect exclusively ‘precious’ intra-chain NOEs, enabling assignment of methyl resonances based on a previously known structure of symmetric oligomers (Xu and Matthews 2013; Chao et al. 2014; Pritišanac et al. 2017; Monneau et al. 2017; Pritišanac et al. 2019).

In homo-oligomeric protein complexes, the identification of inter-chain NOEs was obtained by comparing two 3D CT2-HMQC-NOESY-HMQC spectra acquired on mixed methyl labeled and regular methyl labeled samples prepared at the same concentration (Figs. 3, 5a). Alternatively, in order to work with a single sample, a standard 3D 13C-edited NOESY experiment (Tugarinov et al. 2005b; Kerfah et al. 2015b) and a 3D CT2-HMQC-NOESY-HMQC spectra, both acquired on the mixed labeled sample, can be used to identify both intra-chain and inter-chain NOEs in homo-oligomeric complexes. The conventional 3D NOESY (i.e. non-constant-time) should be recorded with a maximum acquisition time of c.a. 10 ms, in both 13C dimensions, in order to limit peak shape distortion due to the presence of a 1JCC coupling for 50% of the signal. A comparison of the NOE pattern allows to identify unambiguously intra-chain NOEs in the CT2-HMQC-NOESY-HMQC experiment, and potential inter-chain NOEs as the supplementary signals present only in the conventional 3D 13C-edited NOESY experiment.