Probing Protein-Protein Interactions Using Asymmetric Labeling and Carbonyl-Carbon Selective Heteronuclear NMR Spectroscopy

Protein-protein interactions (PPIs) regulate a plethora of cellular processes and NMR spectroscopy has been a leading technique for characterizing them at the atomic resolution. Technically, however, PPIs characterization has been challenging due to multiple samples required to characterize the hot spots at the protein interface. In this paper, we review our recently developed methods that greatly simplify PPI studies, which minimize the number of samples required to fully characterize residues involved in the protein-protein binding interface. This original strategy combines asymmetric labeling of two binding partners and the carbonyl-carbon label selective (CCLS) pulse sequence element implemented into the heteronuclear single quantum correlation (1H-15N HSQC) spectra. The CCLS scheme removes signals of the J-coupled 15N–13C resonances and records simultaneously two individual amide fingerprints for each binding partner. We show the application to the measurements of chemical shift correlations, residual dipolar couplings (RDCs), and paramagnetic relaxation enhancements (PRE). These experiments open an avenue for further modifications of existing experiments facilitating the NMR analysis of PPIs.


Introduction
Biological processes rely primarily on protein-protein interactions (PPIs) to mediate a cellular function [1]. Historically biochemical techniques (co-immunoprecipitation, yeast-two hybrid, pull-down assays, etc.) measuring parameters intrinsic to the whole complex have been used to characterize these PPIs [2,3]. Recently, advances in nuclear magnetic resonance (NMR) spectroscopy have provided the means to characterize PPIs at an atomic resolution, which offers fine details of individual macromolecules participating within the complex [4][5][6]. In addition to allowing the characterization of these complexes at an atomic resolution, NMR is well-suited for studying dynamic, transient (~100 µM K D ), and low-populated states of complexes [7][8][9].
to 13 C to antiphase magnetization. This antiphase magnetization contains components in both the x-direction and y-direction. The π 1 H and 13 C pulses applied at the end of the 15 N evolution convert the y-component, 4H z N y C z , to an unobservable multiple quantum coherence, 4H y N z C y , while the x-component, 4H y N z C' y , is de-phased by the G2 gradient ( Figure 1A). As a result, signals from 1 H-15 N groups coupled to 13 C are suppressed while signals from 1 H-15 N groups coupled to 12 C are unaffected. The suppression spectrum can then be subtracted from the reference spectrum, which leaves the U- 15 N, 13 C species observable ( Figure 1B).
We tested the sensitivity of the CCLS method by comparing a reference CCLS-HSQC spectrum and a conventional HNCO spectrum of the 20-kDa protein U-13 C, 15 N-Ubiquitin at 10 • C, 20 • C, 30 • C, and 40 • C corresponding to average T 2 values of 27 ms, 33 ms, 40 ms, and 47 ms, respectively [32]. The slower tumbling rates at lower temperatures lead to longer rotational correlation times (τ C ) and faster relaxation results in broader linewidths. We found the reference CCLS-HSQC experiment was more sensitive compared to the HNCO experiment for lower temperatures, which demonstrates that the shorter time delay (T NC ) allows for increased sensitivity for large proteins or protein-protein Molecules 2018, 23,1937 3 of 12 complexes. The sensitivity enhancement gained from optimal T NC values compensates for the decrease in S/N observed upon subtraction of the suppression spectrum from the reference spectrum.
Furthermore, we applied this technique to resolve assignment ambiguities on the 41 kDa catalytic subunit of cAMP-dependent protein kinase A (PKA-C) [33,34]. PKA-C is the prototypical Ser/Thr kinase and, until relatively recently, had remained unexplored by NMR due to its size and presence of conformational exchange effects on the µs-ms timescale [35][36][37][38]. Advances in pulse sequence design and sample preparation have since made it possible to investigate this system using NMR [39][40][41][42]. We successfully implemented the CCLS-HSQC pulse sequence to assist in the assignment of multiple catalytically relevant residues of PKA-C. Furthermore, recent work from our group demonstrates the ability of this pulse sequence to simultaneously detect PKA-C in complex with an endogenous inhibitor known as the heat-stable protein kinase A inhibitor (PKIα) [34,43], which gives the possibility to detect the mutual effect of PKA-C and PKIα interaction ( Figure 1C-unpublished data). Together, these applications underscore the ability of CCLS to simultaneously detect PPIs and emphasizes the performance of this pulse sequence with high molecular weight systems. Furthermore, we applied this technique to resolve assignment ambiguities on the 41 kDa catalytic subunit of cAMP-dependent protein kinase A (PKA-C) [33,34]. PKA-C is the prototypical Ser/Thr kinase and, until relatively recently, had remained unexplored by NMR due to its size and presence of conformational exchange effects on the μs-ms timescale [35][36][37][38]. Advances in pulse sequence design and sample preparation have since made it possible to investigate this system using NMR [39][40][41][42]. We successfully implemented the CCLS-HSQC pulse sequence to assist in the assignment of multiple catalytically relevant residues of PKA-C. Furthermore, recent work from our group demonstrates the ability of this pulse sequence to simultaneously detect PKA-C in complex with an endogenous inhibitor known as the heat-stable protein kinase A inhibitor (PKIα) [34,43], which gives the possibility to detect the mutual effect of PKA-C and PKIα interaction ( Figure 1Cunpublished data). Together, these applications underscore the ability of CCLS to simultaneously detect PPIs and emphasizes the performance of this pulse sequence with high molecular weight systems. It can be assumed, unless otherwise indicated, that all rectangular pulses are applied along the x-axis. 90° and 180° flip angles are represented by narrow bars and wide bars, respectively. The carrier frequency for 1 H is set on resonance with water at 4.77 ppm. The carrier frequency for 15 N is set in the center of the amide region at 121.8 ppm. The 13 C offset is set to 174.8 for the C′ region. The reference spectrum is recorded with the shaped pulse for 13 C′ (open rectangle) at the position while the suppression spectrum is recorded with this pulse in position b. A 3-9-19 Watergate pulse scheme is used in the reverse INEPT transfer. GARP1 decoupling with a field strength of 1 kHz is used during the acquisition of 15 N. Delay durations: Δ = 2.4 ms, δ = 0.11 ms, TNC′ = 16.5 ms. Phase cycling: φ1 = x, −x, φ2 = x, x, −x, −x, φrec = x, −x. A second FID is acquired for each increment by changing the φ1 phase to y, −y in order to accomplish States quadrature detection for the 15 N indirect dimension. The φ1 and φrec phases are also incremented by 180° every other 15 N increment for States-TPPI acquisition. The gradients use the Wurst shaped z-axis gradients of 1 ms. Gradient strengths (G/cm): G1: 5, G2: 7, G3: 17. The CCLS-HSQC pulse sequence is based on the fast HSQC experiment [44] to preserve water magnetization. (B) Example spectra representing the reference spectrum, the suppression spectrum, and the resulting subtraction spectrum followed by insets from the PKA-C/PKIα complex displaying the separation of resonances from each species. The blue and red species are present in the reference CCLS-HSQC while A second FID is acquired for each increment by changing the ϕ 1 phase to y, −y in order to accomplish States quadrature detection for the 15 N indirect dimension. The ϕ 1 and ϕ rec phases are also incremented by 180 • every other 15 N increment for States-TPPI acquisition. The gradients use the Wurst shaped z-axis gradients of 1 ms. Gradient strengths (G/cm): G1: 5, G2: 7, G3: 17. The CCLS-HSQC pulse sequence is based on the fast HSQC experiment [44] to preserve water magnetization. (B) Example spectra representing the reference spectrum, the suppression spectrum, and the resulting subtraction spectrum followed by insets from the PKA-C/PKIα complex displaying The suppression spectrum (middle, blue) suppresses a signal from the 13 C labeled PKIα, which shows only peaks from 12 C labeled PKA-C (S/N = 50). Upon subtraction of the suppression spectrum from the reference spectrum, a sub-spectrum is obtained containing only peaks from the PKIα (right, red) (S/N = 15). (All figures were cited with permission of Springer Nature).

Fingerprinting Three Binding Partners Using One Sample
Masterson et al. applied the CCLS pulse sequence element to deconvolute PPIs in a ternary mixture simultaneously [45]. The dual carbon label selective (DCLS) 1 H-15 N HSQC experiment requires three labeled binding partners with the first species U-15 N labeled, the second 15 N, 13 C labeled, and the third U-13 C, 15 N labeled. The deconvolution of these spectra follows the same spin-echo filtering theory as CCLS with additional filtering of Cα coupled spins ( Figure 2A). Cα suppression requires a longer T NCα delay due to both inter-residue and intra-residue 1 H-13 Cα J coupling [46]. Increasing the T NCα delay nullifies protein backbone conformation dependency of 1 J NCα and 2 J NCα since it completely suppresses the signal from 1 J NCα while inverting the residual signal intensities of 2 J NCα .
This and the previously introduced pulse sequence rely on selective labeling of individual binding partners. Asymmetric selective labeling schemes to study PPIs in a multiple component sample are increasing in popularity [16,[47][48][49] both for solution and solid-state NMR spectroscopy. For instance, Anglister and coworkers have demonstrated the application of asymmetric deuteration in combination with transferred nuclear Overhauser spectroscopy to study intermolecular nuclear Overhauser effects (NOEs) of large, fast exchanging protein complexes [50][51][52]. With respect to CCLS and DCLS, selective labeling of 13 C can be accomplished in recombinant proteins using either 15 N-and 13 C -labeled amino acids or 1-13 C pyruvate and 13 C-labeled NaHCO 3 as the sole carbon sources [53][54][55][56][57]. Selective 13 Cα labeling is achieved by using 2-13 C glucose as the sole carbon source [54].
The DCLS experiment requires the acquisition of three interleaved experiments in parallel ( Figure 2B). A reference data set is collected observing all three species simultaneously, which is followed by the first suppression data set where amide resonances adjacent to 13 C are undetected. This is identical to the CCLS suppression spectrum. Lastly, a second suppression data set is collected where amide resonances coupled to 13 Cα are not detected. Deconvolution of the spectra is obtained by a linear combination of the data set. Subtraction of the second suppression spectrum from the reference spectrum provides a sub-spectrum containing only resonances from the U-13 C, 15 N labeled species. The subtraction of the first suppression spectrum from the second suppression spectrum provides an additional sub-spectrum containing only resonances from the U-15 N, 13 C labeled species. In this manner, sub-spectra are obtained from a single sample for each individual component of the ternary mixture and all resonances can be resolved. As a proof of concept, Masterson et al. applied this labeling scheme and pulse sequence to three non-interacting proteins, which includes maltose binding protein (MBP), Kemptide, and ubiquitin. By applying DCLS, the authors obtained sub-spectra corresponding to each individual component of the ternary mixture displaying the potential of this approach to study protein-protein interactions with a single sample.
A second FID is acquired for each increment by changing the ϕ 1 phase to y, −y in order to accomplish states quadrature detection for the 15

Measuring Residual Dipolar Coupling (RDC) of Complexes Using One Sample
Residual dipolar coupling (RDC) allows orientation specific data to be derived through dipole-dipole interactions. The orientation restraints provided by RDC have proven useful in protein structure determination, nucleic acid structure, domain orientation, and more recently PPIs [58,59]. We implemented CCLS and DCLS to sensitivity-enhanced TROSY or anti-TROSY spin-state selection to record the simultaneous measurement of RDCs [58,[60][61][62][63] for the relative orientations of multiple proteins within a single sample ( Figure 3A,B). RDC measurements are susceptible to experimental conditional variations, which alter alignment tensors, making direct correlations of orientational constraints obtained from different samples more difficult. Our approach, together with specific isotopic labeling, eliminates the need for multiple samples and, therefore, removes errors associated with sample inconsistencies [58].
Similar to DCLS ( Figure 3C), we applied this pulse sequence to a non-interacting mixture of U-2 H, 15 N MBP, 15 N-Ser 5 , 13 C -Ala 4 Kemptide, and, U-13 C, 15 N ubiquitin [64]. Following the same linear subtraction scheme reported for DCLS, we were able to measure RDCs for each individual component in a ternary mixture. Furthermore, these RDC values were in agreement with back calculated values determined from already solved crystal structures of MBP [65] and ubiquitin [66], which confirms that the backbone conformational space of these proteins along with their relative alignment tensors were sufficiently defined.

Measuring Long-Range Distances and Transient Complexes Using CCLS for Paramagnetic Relaxation Enhancements (PRE)
Paramagnetic relaxation enhancements (PRE) have been used extensively to obtain long-distance restraints for structure calculation and to study PPIs for both stable and transient complexes [67][68][69][70][71][72]. In the standard PRE experiment that involves two interacting proteins, the intra-molecular or inter-molecular effects of a paramagnetic center are detected for only one of the binding partners in each independent NMR experiment (Figure 4Ai). To accurately probe these interactions, a minimum of four samples with differing spin label positions as well as reversed labeling schemes are required. Recently, we incorporated the CCLS pulse sequence in the traditional 1 H N -Γ 2 ( 1 H N -Γ 2 -CCLS) [69] that, together with an asymmetric labeling scheme, enables the detection of both intra-molecular and inter-molecular paramagnetic relaxation enhancements (PREs) simultaneously using only one sample [10] (Figure 4B). In this newly proposed strategy, one of the two binding partners must be U-15 N labeled and the second U-15 N, 13 C labeled (Figure 4Aii). We also tested the proposed pulse sequence on the non-covalent, transient dimerization of ubiquitin. Specifically, we studied the complex formed between U-15 N, 13 C wild-type ubiquitin and the U-15 N-spin labeled the K48C mutant. We were able to discriminate intra-molecular and inter-molecular interactions detecting the structural and dynamics changes intrinsic to ubiquitin upon dimerization ( Figure 4C). The Γ 2 rates obtained with the new pulse sequence were confirmed to be identical among standard experiments. This work demonstrates that the Γ 2 -CCLS PRE experiment is suitable for identifying structural changes occurring in both binding partners upon formation of transient and permanent interactions using a reduced number of samples.

Measuring Long-Range Distances and Transient Complexes Using CCLS for Paramagnetic Relaxation Enhancements (PRE)
Paramagnetic relaxation enhancements (PRE) have been used extensively to obtain longdistance restraints for structure calculation and to study PPIs for both stable and transient complexes [67][68][69][70][71][72]. In the standard PRE experiment that involves two interacting proteins, the intra-molecular or inter-molecular effects of a paramagnetic center are detected for only one of the binding partners in each independent NMR experiment (Figure 4Ai). To accurately probe these interactions, a minimum of four samples with differing spin label positions as well as reversed labeling schemes are required. Recently, we incorporated the CCLS pulse sequence in the traditional 1 HN-Γ2 ( 1 HN-Γ2-CCLS) [69] that, together with an asymmetric labeling scheme, enables the detection of both intra-molecular and inter-molecular paramagnetic relaxation enhancements (PREs) simultaneously using only one sample [10] (Figure 4B). In this newly proposed strategy, one of the two binding partners must be U-15 N labeled and the second U-15 N, 13 C labeled (Figure 4Aii). We also tested the proposed pulse sequence on the non-covalent, transient dimerization of ubiquitin. Specifically, we studied the complex formed between U-15 N, 13 C wild-type ubiquitin and the U-15 N-spin labeled the K48C mutant. We were able to discriminate intra-molecular and inter-molecular interactions detecting the structural and dynamics changes intrinsic to ubiquitin upon dimerization ( Figure 4C). The Γ2 rates obtained with the new pulse sequence were confirmed to be identical among standard experiments. This work demonstrates that the Γ2-CCLS PRE experiment is suitable for identifying structural changes occurring in both binding partners upon formation of transient and permanent interactions using a reduced number of samples. the standard experiment for the detection of intra-molecular and inter-molecular PRE. In this case, four different samples are needed. The first sample for the intramolecular PRE is prepared with an asymmetric labeling scheme using the first binding partner uniformly 15 N labeled with a conjugated spin label (SL) and the second is NMR silent (unlabeled). A sample with a reversed labeling scheme is necessary to detect the intra-molecular PRE for the second binding partner (top panel A). For the inter-molecular PRE, two additional samples are required: one NMR silent with the conjugated SL and a second NMR active (e.g., 15 N or 13 C labeled) (lower panel intra-molecular and inter-molecular PRE. In this case, four different samples are needed. The first sample for the intra-molecular PRE is prepared with an asymmetric labeling scheme using the first binding partner uniformly 15 N labeled with a conjugated spin label (SL) and the second is NMR silent (unlabeled). A sample with a reversed labeling scheme is necessary to detect the intra-molecular PRE for the second binding partner (top panel A). For the inter-molecular PRE, two additional samples are required: one NMR silent with the conjugated SL and a second NMR active (e.g., 15 N or 13 C labeled) (lower panel A). Simultaneous detection of inter-molecular and intra-molecular PRE using 1 H N -Γ 2 -CCLS experiment. One species is uniformly 15 N labeled while the other is 13 C and 15 N labeled (b) allowing for the simultaneous detection of intra-molecular and inter-molecular PREs. As reported before, reverse positioning of the SL is required for obtaining a complete characterization of the complex. (B) The 1 H-Γ 2 -CCLS pulse sequence for PRE Γ 2 measurements. The narrow and wide bars represent 90 • and 180 • hard pulses, respectively. The three 13 C 180 • shaped pulses are 256 µs long Q3 pulse, the first two and the last one shaped pulses are applied to 13 C and 13 C α , respectively. The 13 C 180 • shaped pulse may be at either position a or b. When it is at position a, the 1 J NC is decoupled and reference spectra are acquired. When it is at position b, the 1 J NC is present and 13

Improving Sensitivity with the G5 Pulse
Advances in NMR methodology (TROSY, deuteration, selective labeling) have allowed for studies of protein-protein complexes approaching 1 MDa [73,74]. However, these studies lack the ability to distinguish one species from another without the preparation of multiple samples. A recent technological advance that can improve nearly any pulse sequence is the universal triply compensated π pulses for high field spectrometers [75,76], which we have incorporated into the CCLS pulse sequence ( Figure 5). All inversion and refocusing pulses in the 1 H and 15 N channel were replaced with G5 pulses except the 15 N refocusing pulse in the middle of 3-9-19 water suppression. We were able to improve the signal intensity from 6% to 23% compared to the regular CCLS version. These experiments were performed on the Bruker 900 MHz AVIII spectrometer at 298 K and this enhancement will only be more significant in GHz spectrometers.
Molecules 2018, 23, x FOR PEER REVIEW 8 of 12 A). Simultaneous detection of inter-molecular and intra-molecular PRE using 1 HN-Γ2-CCLS experiment. One species is uniformly 15 N labeled while the other is 13 C and 15 N labeled (b) allowing for the simultaneous detection of intra-molecular and inter-molecular PREs. As reported before, reverse positioning of the SL is required for obtaining a complete characterization of the complex. (B) The 1 H-Γ2-CCLS pulse sequence for PRE Γ2 measurements. The narrow and wide bars represent 90° and 180° hard pulses, respectively. The three 13 C 180° shaped pulses are 256 μs long Q3 pulse, the first two and the last one shaped pulses are applied to 13 C′ and 13 Cα, respectively. The 13 C′ 180° shaped pulse may be at either position a or b. When it is at position a, the 1 JNC′ is decoupled and reference spectra are acquired. When it is at position b, the 1 JNC′ is present and 13

Improving Sensitivity with the G5 Pulse
Advances in NMR methodology (TROSY, deuteration, selective labeling) have allowed for studies of protein-protein complexes approaching 1 MDa [73,74]. However, these studies lack the ability to distinguish one species from another without the preparation of multiple samples. A recent technological advance that can improve nearly any pulse sequence is the universal triply compensated π pulses for high field spectrometers [75,76], which we have incorporated into the CCLS pulse sequence ( Figure 5). All inversion and refocusing pulses in the 1 H and 15 N channel were replaced with G5 pulses except the 15 N refocusing pulse in the middle of 3-9-19 water suppression. We were able to improve the signal intensity from 6% to 23% compared to the regular CCLS version. These experiments were performed on the Bruker 900 MHz AVIII spectrometer at 298 K and this enhancement will only be more significant in GHz spectrometers.

Conclusions and Perspectives
In this paper, we demonstrate that the CCLS/DCLS pulse sequences enable the study of PPIs through simultaneous inter-leaved detection of all components in a single sample. As we have illustrated, the CCLS and DCLS pulse sequence blocks can be applied to a multitude of well-established experiments (RDC and PRE). Extrapolating from this integration into existing NMR experiments, could NOESY be the next step? The possibility of observing multiple species in a single sample for NOESY experimentation is viable since Anglister et al. [50] has reviewed different spectroscopy and its application to 2D NOESY experiments. However, the pulse sequences are, therefore, limiting sensitivity, which the CCLS/DCLS pulse blocks show promise toward combating. Therefore, reflecting upon the versatility of the CCLS/DCLS pulse block and the associated advantages afforded, we envisage the insertion into other existing NMR experiments to study a wide range of multicomponent systems.
Funding: This research was funded by the Foundation for the National Institutes of Health grant numbers GM100310 and P41 GM103399.