Detection of intermolecular NOE interactions in large protein complexes

Accepted Manuscript Detection of intermolecular NOE interactions in large protein complexes Jacob Anglister, Gautam Srivastava, Fred Naider PII: DOI: Reference:

S0079-6565(16)30024-3 http://dx.doi.org/10.1016/j.pnmrs.2016.08.002 JPNMRS 1427

To appear in:

Progress in Nuclear Magnetic Resonance Spectroscopy

Received Date: Accepted Date:

29 May 2016 7 August 2016

Please cite this article as: J. Anglister, G. Srivastava, F. Naider, Detection of intermolecular NOE interactions in large protein complexes, Progress in Nuclear Magnetic Resonance Spectroscopy (2016), doi: http://dx.doi.org/ 10.1016/j.pnmrs.2016.08.002

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Detection of intermolecular NOE interactions in large protein complexes. Jacob Anglister1, Gautam Srivastava1 and Fred Naider2,3.

1

Department of Structural Biology, Weizmann Institute of Science, Rehovot 76100,

Israel. 2

Department of Chemistry and Macromolecular Assembly Institute, College of Staten

Island of the City University of New York, Staten Island, New York 10314, USA and the Ph.D. Programs in Biochemistry and Chemistry, The Graduate Center of the City University of New York, New York, NY 10016. 3

On sabbatical leave as the Erna and Jakob Michael Visiting Professor in the

Department of Structural Biology, Weizmann Institute of Science, Rehovot 76100, Israel.

Correspondence: J. Anglister, Department of Structural Biology, Weizmann Institute of Science, Rehovot 76100, Israel Fax: 972-8-9344136 Phone: 972-8-9343394 E-mail: [email protected] Keywords: NMR, proteins, NOE, TRNOE, intermolecular interactions Running Title: Intermolecular NOEs in large complexes 1

ABSTRACT Intermolecular NOE interactions are invaluable for structure determination of biomolecular complexes by NMR and they represent the “gold-standard” amongst NMR measurements for characterizing interfaces. These NOEs constitute only a small fraction of the observed NOEs in a complex and are usually weaker than many of the intramolecular NOEs. A number of methods have been developed to remove the intramolecular NOEs that interfere with the identification of intermolecular NOEs. NMR experiments used to observe intermolecular NOE interactions in large protein complexes must cope with the short T2 relaxation time of the protons and heteronuclei in these complexes because they result in severe losses in sensitivity. The isotopeedited/isotope-filtered experiment is a powerful method for extraction of intermolecular NOEs in biomolecular complexes. Its application to large protein complexes is limited because of severe losses in signal-to-noise ratio caused by delays in the pulse sequence necessary for the multiple magnetization transfer steps between protons and heteronuclei. Isotope-edited/isotope-edited experiments, in which one protein is usually labeled with

13

C and the other is labeled with

15

N, reduce possible artifacts in the

filtering experiments and improve somewhat the sensitivity of these experiments. Sensitivity can also be improved by deuteration of the components of the complex in order to replace either or both of the filtering or editing steps. Asymmetric deuteration, where aromatic residues in one protein and non-aromatic amino acids in the other are reverse protonated, can eliminate the editing and the filtering steps altogether, thus maintaining high sensitivity even for large proteins complexes. Difference spectroscopy

2

and the use of 2D NOESY experiments without using editing or filtering steps can significantly increase the signal-to-noise ratio in experiments aimed at observing intermolecular NOEs. The measurement of NOESY spectra of three different preparations of a heterodimeric complex under investigation in which one or neither of the components is uniformly deuterated, and calculation of a double difference spectrum provides information on all intermolecular NOEs of non-exchangeable protons. Recent studies indicate that many protein-protein interactions are actually between a protein and a linear peptide recognition motif of the second protein, and determinants represented by linear peptides contribute significantly to the binding energy. NMR is a very versatile method to study peptide-protein interactions over a wide range of binding affinities and binding kinetics. Protein-peptide interactions in complexes exhibiting tight binding can be studied using single and/or multiple deuteration of the peptide residues and measuring a difference NOESY spectrum. This difference spectrum will show exclusively intra- and intermolecular interactions of the peptide protons that were deuterated. Transferred nuclear Overhauser spectroscopy (TRNOE) extends NMR to determine interactions within and between a weakly-bound rapidly-exchanging peptide and its protein target. TRNOE, together with asymmetric deuteration, is applicable to complexes up to ~100 KDa and is highly sensitive, taking advantage of the long average T2 of the peptide protons. Among the methods described in this review, TRNOE has the best potential to determine intermolecular NOEs for the upper molecular weight limit of proteins that can be studied in detail by NMR.

3

Contents 1.

Introduction

2.

The isotope-edited/isotope-filtered experiment

3.

Isotope-edited/isotope-edited

experiment

with

asymmetric

labeling

of

heteronuclei 4.

Isotope-editing experiments combined with uniform deuteration

5.

Observation of intermolecular NOE interactions using asymmetric deuteration

6.

Observation of intermolecular NOE interactions using double difference spectra and uniform deuteration

7.

Protein-peptide intermolecular NOE interactions in complexes exhibiting tight binding using deuteration and difference spectroscopy

8.

Identification of intermolecular protein-peptide NOEs using transferred NOE 8.1

Introduction

8.2

TRNOE difference spectroscopy

8.3

T1ρ-filter

8.4

TRNOE in combination with asymmetric deuteration

9.

NOESY difference spectra using a spin-labeled ligand

10.

Assignment of intermolecular NOEs to specific residues in large protein complexes.

11.

Conclusions and future perspectives

4

1.

Introduction

Biological processes such as signal transduction, immune response, enzymatic catalysis and transcription of the genetic code are mediated by specific interactions between biomolecules. NMR spectroscopy can be used to study such interactions at atomic resolution. Compared to X-ray crystallography, which is primarily applicable to tightly bound complexes, an advantage of NMR is that it is applicable to studies of interactions for complexes exhibiting a wide range of dissociation constants from weak (KD ≈ 1 mM) to very strong (KD < 1 M) binding. The most informative data on such interactions can be extracted from intermolecular proton-proton NOE measurements, which provide distances for pairwise short range interactions (< 6 Å) between components of protein-protein, protein-nucleic acid and protein-small ligand complexes. NOESY spectra of protein complexes contain a very large number of cross peaks. These spectra become increasingly crowded as the molecular weight of the complex increases, due to the larger number of hydrogens in the molecule and the concomitant broadening of the individual proton resonances caused by enhanced relaxation. The NOEs in such spectra are from both intramolecular and intermolecular proton-proton interactions. Intermolecular NOEs constitute only a small fraction of the observed NOEs and they are usually weaker than many of the intramolecular NOEs because of the r-6 dependency of the NOE on the internuclear distance. Moreover, the intensity of the intermolecular NOE can be affected by the strength of binding and to a large extent by the exchange rate of the complex (k ex). When considering NMR linewidths, the exchange rate, kex, is defined as slow (kex< Δω), medium (kex ~ Δω) or fast (kex > Δω) depending on the difference in resonance frequencies (Δω) between the

5

bound and free state for the individual protons [1, 2]. It is a widely held view that observation of intermolecular NOE interactions is usually limited to complexes with medium to strong binding (KD < 10 µM) and slow exchange rates [3]. In this review we will demonstrate that intermolecular interactions can also be observed for complexes with weak binding and fast exchange rates. Our article will focus primarily on intermolecular NOE interactions in large protein complexes (> 35 kDa) and we will describe methods to observe them in such complexes exhibiting either tight or weak binding. The limitations and advantages of the various approaches in terms of their applicability to large proteins, the level of information about intermolecular interactions that they provide, and their dependence on the strength of binding and the ligand exchange rate will all be assessed. A special focus in this review will be on protein-protein interactions and proteinpeptide interactions, although some of the techniques may also be applicable to complexes involving nucleic acids. Recent studies indicate that many protein-protein interactions are actually between a protein and a linear peptide recognition motif of the second protein [4, 5]. Such protein-peptide recognition events, now known to be quite common, have characteristics that differ from the recognition between two globular proteins, and have been the subject of a variety of docking and modeling efforts [4, 5]. Determination of pairwise interactions between peptide and protein residues will greatly aid the computation of more refined models of the peptide-protein complexes. Many of the methods for elucidating intermolecular interactions use what have been termed X-isotope filtered, X-isotope half-filtered and isotope-filtered/isotopeedited or isotope-edited/isotope-filtered experiments (see Fig. 1). These are applied to

6

complexes in which one component is labeled with

13

C or

15

N and the other one is

unlabeled. Such methods have been described thoroughly elsewhere [6, 7], and therefore in this review we will focus mainly only on their applications to large molecular weight complexes. A common feature of all isotope-edited/isotope-filtered pulse sequences is the incorporation of the INEPT or HMQC pulse train for transfer of magnetization between the protons and their bonded heteronuclei and vice versa. The INEPT and HMQC sequences contain delays that are on the ms timeframe. Thus, pulse sequences involving multiple INEPT or HMQC elements can be lengthy and thereby result in severe signal loss as the molecular weight increases and the T2 relaxation of proton and heteronuclear spins in the proteins becomes shorter. We will address this in more detail when we present the specific edited or filtered experiment. One way to shorten the experiment is to eliminate the editing and/or the filtering step. This can be done by “chemical editing” using deuteration [8]. Such experiments are also included in this review. An important goal of this review is to evaluate the use of transferred NOE (TRNOE) for determining intermolecular contacts. Intermolecular TRNOE is defined in Section 8.1. Elegant theoretical research dating back to the pioneering reports of Clore and Gronenborn in 1982 established the power of TRNOE for determining the structure of ligands bound to proteins [9, 10]. Although the earliest studies involved non-peptide ligands, later investigations showed that TRNOE could be used to determine the structure of peptides bound to proteins [11]. The analysis by Clore and Gronenborn concluded that intermolecular contacts between protons of the bound ligand and

7

protons of the protein could be observed by TRNOE if the chemical exchange rate was fast relative to the total spin-lattice relaxation-rate . However, no data were provided that demonstrated this theoretical conclusion. Pioneering studies by Bothner-By, Cohn and James very nicely demonstrated the applicability of TRNOE for studying intermolecular interactions when biomolecular NMR was still in its infancy [12-14]. Later, using carefully designed experiments we showed that information on intermolecular contacts between a bound peptide ligand and its cognate binding protein could be obtained using 2D NOESY experiments [15, 16]. Such experiments have the potential to observe all intermolecular NOE interactions in peptide complexes with proteins exhibiting a fast exchange rate. Despite the experimental successes and the existing theoretical treatment, there are almost no recent reports of the use of TRNOE to systematically study intermolecular pairwise interactions. We feel that the increased use of TRNOE is warranted and will discuss strategies that improve the ability of NMR spectroscopists to use TRNOE to determine intermolecular interactions in weakly binding protein-peptide complexes with molecular weights from 10 to 100 kDa.

2. The isotope-edited/isotope-filtered experiment The

isotope-edited/isotope-filtered

technique

and

the

reverse

order

isotope-

filtered/isotope-edited experiment have been widely used to study intermolecular interactions in biomolecular complexes, especially for protein complexes smaller than 30 kDa. In the isotope-edited/isotope-filtered method, initially magnetization of protons bonded to

13

C or

15

N is selected, as illustrated in Fig. 1a. This is referred to as the

editing step. In the next step following the NOESY mixing period, the magnetization on

8

13

protons bonded to to

12

C and

such as

13

C and

15

N is eliminated, leaving only signals from protons bonded

14

N. The elimination of the magnetization of protons bonded to heteronuclei

C and

15

N is referred to as the filtering step (13C- or

15

N-filtering). As a result

of the editing and the filtering steps, the detected signal arises from magnetization transfer between the 13C and 15N bonded protons of protein A to the

12

C and 14N bonded

protons of protein B. Kay and coworkers developed an isotope-edited/isotope-filtered 3D

13

C-HMQC-

NOESY experiment that minimizes delays and the required phase cycle steps using several gradient pulses to eliminate experimental artifacts [17]. This pulse sequence was applied to the 37 kDa trp-repressor dimer in complex with unlabeled operator DNA and was found to be very useful in eliminating ambiguities between inter- and intramolecular NOE interactions. Measurements were carried out on a 500 MHz spectrometer equipped with a conventional probe (not a cryoprobe) using 1 mM dimer solution in D2O and at 37 ºC, suggesting that much lower concentrations can be examined if higher field strengths and a cryoprobe are employed. In the different amino acids a wide range of 3JCH coupling constants occurs between different

13

C nuclei and their bonded 1H protons. This range of J values may

impede the complete filtering (rejection) of NOEs involving protons bonded to

13

C in

protein complexes. In general, this will cause the filtering experiment to be less reliable than the editing step. To address this concern, adiabatic pulse sequences have been used to cover a wide range of coupling constants (Fig. 1b) [18]. In this experiment the filtering step preceded the editing step. This enabled incorporation of the WATERGATE water suppression pulse scheme into the final HSQC step and measurements in H2O

9

solution without the need for additional delays that [19] may further attenuate the signal. The power of this improved pulse sequence was demonstrated on a ~35 kDa peptide/RNA complex at 25 ºC.

Fig. 1. Detection of intermolecular NOEs by isotope-editing/isotopefiltering

and

by

isotope-

filtering/isotope-editing experiments.

a)

representation scheme

A

of

and

schematic

the the

labeling observed

interactions. An isotope editing experiment selects and detects (1H)

protons

bonded

to

heteronuclei (13C/15N). An isotope filtering

experiment

rejects

1

H

bonded to 13C/15N. Only 1H bonded to

12

C/14N is detected b) A pulse

sequence

for

an

filtered/isotope-edited

isotopeNOESY-

HSQC experiment designed for an H2O sample that includes adiabatic pulses to cover a range of 3JHC coupling constants [18]. In this experiment preceded

the the

filtering

step

editing

step.

Experimental details are given in the cited publication.

10

Clore, Gronenborn and co-workers used 3D

13

C-edited/13C-filtered experiments to

determine the structure of several large complexes including the 44 kD SIV gp41 homotrimer (2.5 mM at 45 ºC) [20], the 40 kDa phosphoryl transfer complex between the N-terminal domain of enzyme I and HPr (1 mM 40 ºC) [21], and the 42 kDa Oct1.Sox2.DNA ternary complex [22] (30 ºC, 0.5-1 mM samples). The Zwahlen and Kay isotope-filtered/isotope-edited sequence was used by Yang and co-workers to observe intermolecular interactions in hemoglobin, a 65 kDa tetramer of two α chains and two βchains [23]. The concentration of the complex was 1 mM and it was measured on a 800 MHz spectrometer at 30 ºC. Recently this technique was used to determine the structure of the mitochondrial translocator protein (TSPO) in complex with a diagnostic ligand. TSPO is an 18 kDa membrane protein, and the structure of the complex was determined in DPC micelles that considerably increased the apparent molecular weight of the complex to ~80 kDa and consequently decreased the T2 relaxation times of its nuclei [24]. Only intermolecular interactions involving methyl protons were observed in this study. There is no doubt that the isotope-edited/isotope-filtered technique is the most powerful and most commonly used technique to observe intermolecular interactions in protein complexes. When conditions are favorable, such as with highly soluble nonaggregating protein complexes that allow studies at high protein concentrations, and particularly when measurements can be carried out at elevated temperatures, intermolecular interactions in complexes as large as 80 kDa can be detected using isotope-edited/isotope-filtered experiments. At this upper molecular weight limit it is

11

likely that only interactions involving methyl protons, which give more intense signals, will be observed.

3.

Isotope-edited/isotope-edited

experiment

with

asymmetric

labeling

of

heteronuclei Replacing the filtering step by another editing step may alleviate some of the artifacts in the isotope-edited/isotope-filtered experiments and may lead to simpler pulse sequences with an overall shorter time required for the magnetization transfer steps. Asymmetric labeling of the heteronuclei is then necessary to replace the filtering step as shown in Fig. 2.

12

Fig. 2. A schematic representation of the labeling scheme used for isotopeediting/isotope-editing experiments with 13

asymmetric

15

C and

observing

N labeling, for

intermolecular

interactions. a) Uniform

13

protein A and uniform

15

NOE

C labeling of

N labeling of

protein B. Only interactions of bonded to

13

C with 1H bonded to

detected. results

The in

asymmetric detection

1

H

15

N are

labeling of

only

intermolecular interactions. b) Uniform 13

C labeling of protein A and uniform 15N

labeling of protein B in combination with uniform deuteration of protons bonded to

12

C in protein B. c) The methyl groups

of Ile, Leu and Val residues of protein A 13

are protonated and background

of

C labeled in a

uniformly

deuterated

protein (2H,ILV–13CH3 labeling). Protein B

is

uniformly

deuterated.

The

15

N

labeled

amides

are

and back

protonated. d) The methyl of Ile residues of protein A and Leu and Val residues of Protein B are protonated as well as

13

C

labeled in a background of uniform deuteration of the proteins. Groups shown in parentheses are not seen in the spectra either because they are not NMR active or because they are not selected by the pulse sequence.

13

To study intermolecular interactions in a homodimer, Handel and Domaille [25] described a novel scheme to asymmetrically label their homodimer. In their method one of the proteins, protein A, is uniformly labeled with protein B, is labeled with

15

N. A

15

N-

13

13

C and the interacting protein,

C- separated 4D NOESY of this complex should

show only intermolecular NOEs between the protons bonded to

13

C in protein A and the

amide protons of protein B (Fig. 2a). One disadvantage of this strategy is that on forming

the

asymmetrically

labeled

“homodimer”

two

symmetrically

labeled

“homodimers” are formed as well, thereby effectively lowering the concentration of the magnetically active species and decreasing sensitivity. Furthermore, asymmetric labeling results in NOE detection in one direction also leading to signal loss. To address this problem increased sensitivity was obtained by using 2D and 3D versions of this 4D experiment. Application of this approach to the 17.5 kDa homodimer of the monocyte chemoattractant protein-1 (MCP) at 35 ºC resulted in the definition of 65 intermolecular NOEs which were combined with a large number of intramolecular NOEs, chemical shift data and J-coupling constraints to determine the three dimensional structure of the complex [25]. The detection of intermolecular NOEs relied only on isotope-editing steps and thus overcame any imperfections in the filtering schemes that were considered problematic before the use of adiabatic pulses was introduced by Kay and coworkers [18]. Although this method is very attractive for studying protein complexes up to 30 KDa, for larger complexes highly efficient proton-proton relaxation pathways decrease T2 resulting in severe line broadening. Perdeuteration of protein B and back exchange of its amide hydrogens with protons should increase the T2 relaxation time of the amide

14

protons, thereby resulting in narrower line widths and extending the application of this method to larger complexes [8] (Fig. 2b). Selective

13

CH3 labeling of Ile, Leu and Val residues in an otherwise deuterated

background, and [2H]ILV–13CH3 labeling [26] of one of the proteins in the complex in a deuterated protein background while the second protein is

15

N labeled and uniformly

deuterated (except for the amide hydrogens) may further expand the applicability of the method to protein complexes in the 80-100 kDa range. This was previously demonstrated for the monomeric 82 kDa enzyme malate synthase G [27] (Fig. 2c). An elegant example of

13

C-editing/13C-editing employed selective methyl-labeling

of Ile, Leu, and Val residues in an otherwise perdeuterated background to detect interprotomer cross peaks in homo-oligomers of phospholambin (PLN) [28]. The homopentamer of the ~5500 Da membrane protein was formed from asymmetrically methyl labeled protomers (one labeled on the 13CH3 of Ile and the other labeled on the 13CH3 of Leu and the 13CH3 of Val residues) (Fig. 2d). Intermolecular NOEs for this leucine-isoleucine zipper stabilized complex could be identified and distinguished from intramolecular I-I L-L V-V and V-L contacts based on the unique

13

C chemical shifts of

the Ile methyls compared with the Val and Leu methyls. This method is appropriate for determining methyl-methyl interactions at the interfaces of membrane protein helices. A positive outcome of the perdeuteration of the protein is that the elimination of many spin diffusion pathways and T2 relaxation mechanisms by deuteration, allows measurements of NOESY experiments with longer mixing times thereby increasing the sensitivity. The interaction distances that can be detected are also extended up to 12 Å. The fact that the study was conducted in 300 mM dodecylphosphocholine suggests it is amenable for

15

large complexes. [2H]ILV–13CH3 labeling is a very powerful technique that has been applied for complexes up to a molecular weight of 1 MDa, although NOE interactions have not been studied at the upper molecular weight limit [26, 29]. If sequential assignment for the methyl 1H and

13

C nuclei of Ile, Leu and Val

residues is available for both proteins in the complex, computer programs for structure determination such as CYANA [30] can, in principle, use the NOE data without the prior classification of intra- and intramolecular NOE interactions and assign these interactions automatically. In a very elegant application of this approach Kalodimos and coworkers determined the structure of complexes of Trigger Fragment Chaperone (TF) with unfolded alkaline phosphatase [31]. Complexes of domains of both proteins with molecular weight up to 50 kDa were used to obtain the structures. The study identified multiple domains on TF that interact with unfolded alkaline phosphatase. The complex formation involved a large number of interactions of Ile, Leu, and Val residues and [2H]ILV–13CH3 labeling proved to be invaluable for obtaining a large number of both intra- and intermolecular interactions. Given the fact that alkaline phosphatase is unfolded, its interaction with TF is not amenable to X-ray analysis and NMR is uniquely suitable to study the structural details of this system. The NMR editing methods described in this section are powerful approaches for identifying intermolecular contacts between proteins and between proteins and peptides. All of the above strategies have certain strengths and deficiencies. The

15

NH-

13

CH editing used for the MCP-1 protein only reveals backbone NH to side-chain

connectivities. It is somewhat rare to see such interactions at protein-protein interfaces where most of the contacts involve side chains. In principle the need to run 4D

16

experiments lowers the efficiency and sensitivity of the method, but better sensitivity can be achieved by running a set of 3D experiments. 2H,I–13CH3 labeling of one protein and 2

H,LV–13CH3 labeling of the other reveals only a very specific subset of interactions of

Ile methyls with Leu methyls and Ile methyls with Val methyls. However, this labeling scheme together with methyl-TROSY makes it useful for large protein complexes [32]. 4. Isotope-edited experiment combined with uniform deuteration The

isotope-editing/isotope-filtering

and

the

isotope-editing/isotope-editing

approaches discussed in the previous sections will result in a considerable decrease in the signal-to-noise ratio for large protein complexes. To effectively shorten the total duration

of

the

isotope-edited/isotope-filtered

or

isotope-edited/isotope-edited

experiment by 50%, one of the filtering or editing steps can be replaced by specific asymmetric deuteration of one of the proteins in the complex. This approach has been termed “chemical editing” [8]. Wagner and co workers [33] employed chemical editing where one of the proteins in the complex was uniformly labeled with

15

N and was also perdeuterated.

After protein expression the amides were reverse protonated by dialysis against a protonated buffer. The second protein in the complex was unlabeled. A 3D

15

N-edited

NOESY pulse sequence detects the intermolecular interactions between the amide protons of the perdeuterated protein and all possible protons of the other protein (Fig. 3a). This spectrum also contains intramolecular interactions between amide protons of protein A.

17

Fig. 3. A schematic representation of the observed intermolecular NOEs and the labeling scheme used for NOESY experiments with one editing step and no additional filtering or editing step. a) Protein A is uniformly

15

N and 2H labeled while

protein B is unlabeled. Only interactions between 1H connected to

15

N of protein A

and protons of protein B, as well as interactions with other amide protons of protein A, are detected. b) Methyl groups of Ile, Leu and Val residues of protein A are protonated and

13

C labeled in a background of uniform deuteration (2H,ILV–13CH3

labeling) while protein B is unlabeled.

Another method developed by Wagner and co-workers employed

13

C/1H labeling

of the methyl protons of Ile, Leu and Val residues of one protein in an otherwise perdeuterated protein (2H,ILV–13CH3 labeling). The second protein in the complex was unlabeled. A

13

C HMQC-NOESY spectrum of the complex revealed the intermolecular

interactions of Ile, Leu, and Val methyl protons with aromatic and aliphatic protons of the second protein in the complex (Fig. 3b). The method was applied to a 35 kDa

18

elf4E/elf4G complex in the presence of CHAPS which probably increased the apparent molecular weight by at least 7 kDa [34]. The structure of the complex was determined with a backbone RMSD of 0.95 Å. The same two methods were used by this group to determine the solution structure of a 41 kDa Calcineurin A in complex with a peptide ligand [35]. An interesting variation of editing and asymmetric labeling was used in a study of a 27 kDa homotrimer of the MHC Class II invariant chain. In this case the NOESY spectrum for a homotrimer sample where all chains were

13

C-edited

13

C labeled was

compared with a similarly acquired spectrum of a homotrimer sample where one protomer was labeled with

13

C and the other two chains were labeled with 2H [36]. The

first spectrum contained both intra- and interprotomer NOEs whereas the latter contained only intrachain NOEs. The second and the third methods described in this section, containing only one editing step, in general allow detection of a considerably larger number of interactions in comparison with the methods described in section 3 that use two editing steps. The second method (Fig. 3b) described in this section detects NOE connectivities between the methyl protons of Ile, Leu and Val residues of protein A and all carbon bonded protons of protein B. Such side-chain/side-chain interactions are very common in protein complexes and the method can be easily applied to large protein complexes. The method used with the MHC-Class II invariant chain [36] can in principle provide even more complete information about intermolecular interactions between protons bonded to carbon in homoprotomeric complexes. The method required measurement of two spectra of two different preparations of the complex and then

19

visual comparison to pinpoint the intermolecular interactions. Because it depends on visual comparison, this method is limited to homoprotomeric complexes consisting of small subunits in which the number of different protons is considerably smaller than in a large heterodimeric complex. Although this last approach can be generally applied, it might benefit from calculating a difference spectrum. Our experience with difference NOESY spectroscopy indicates that by normalizing the spectrum very accurate subtraction can be obtained (vide infra).

5. Observation of intermolecular NOE interactions using asymmetric deuteration A further improvement in the sensitivity of intermolecular NOE measurements would be obtained if all the INEPT transfer steps were eliminated altogether from the pulse sequence used to select for intermolecular NOE interactions. This can be achieved if “deuteration” is used to filter out the intramolecular interactions within each component of the complex instead of using editing or filtering techniques. Aromatic and nonaromatic non-labile protons resonate in different frequency windows, and the upper left and the lower right quarters of the NOESY spectrum show interactions between these two types of protons. We have developed an asymmetric deuteration approach that has enabled elimination of all cross peaks due to intramolecular NOEs in these two quarters of the NOESY spectrum (Fig. 4). What remains in this region are only intermolecular interactions between the components of the complex [37].

20

Fig. 4. A schematic representation of the observed intermolecular NOEs and the labeling scheme used for NOESY experiments on asymmetrically deuterated complexes. A) In protein A the aromatic amino acids are unlabeled (H-Aromatic) while those with non-aromatic side chains are deuterated (D-Aliphatic). In protein B the non-aromatic amino acids are unlabeled (H-Aliphatic) while the aromatic ones are deuterated (D-Aromatic).

To eliminate the intramolecular NOEs, all aromatic amino acids in one protein in the complex are reverse protonated in an otherwise perdeuterated background. In the second protein in the complex a select group of aliphatic amino acids is reverse protonated in an otherwise perdeuterated background. Reverse protonation is achieved by supplying the expression cells with excess of specific protonated amino acids in an otherwise fully deuterated media [37]. The two proteins can also be labeled in exactly the opposite way with the first protein reverse protonated on the aliphatic residues and the second protein reverse protonated on the aromatic residues, both in perdeuterated backgrounds. This labeling scheme enabled us to observe intermolecular interactions between aromatic protons of interferon IFNα2 with non-aromatic protons of its receptor IFNAR2 and vice versa. Using NOESY measurements on a 800 MHz spectrometer equipped with a cryoprobe at 32 ºC and a 0.2-0.3 mM solution of this 44 kDa complex,

21

24 intermolecular NOEs with an excellent signal-to-noise ratio were observed and assigned to the corresponding protons (Fig. 5).

a)

b)

22

Fig. 5. Inter-molecular NOE interactions detected in the asymmetrically deuterated IFNAR2/IFNα2 complex [37]. a) Overlay of 2D NOESY spectra in D2O of IFNAR2(HFW)/IFNα2(KRLAM) (black) and IFNα2(KRLAM) (red). b) Overlay of 2D NOESY

spectra

in

D2O

of

IFNAR2(IVLTMAK)/IFNα2(HFWY)

(black)

and

IFNAR2(IVLTMAK) (red). The amino acid types in parenthesis are unlabeled and the rest are deuterated. Black cross-peaks not overlaid with red cross-peaks and not marked by arrows originate from inter-molecular interactions and are labeled according to the assignment of the aliphatic proton. Red cross-peaks originate from intra-molecular interactions due to residual protonation in the deuterated proteins. Arrows indicate small changes in the positions of the intra-molecular cross-peaks between the spectra of the free molecule and the complex. Vertical lines indicate spin systems of cross-peaks originating from the same assigned aromatic proton and are labeled accordingly. IFNAR2 and IFNα2 protons are labeled by r2 and a2, respectively. Light blue boxes indicate the few cross peaks, which appeared after urea induced partial denaturation probably due to carbamylation of side-chain amines of a lysine residue.

NOESY measurements were carried out in D2O to avoid using the WATERGATE pulse-scheme for water suppression [19], since the long pulses and delays applied in the WATERGATE pulse train result in severe relaxation of protons with short T2 values. For the suppression of the HDO signal in the D2O protein solution, presaturation was applied with minimal power. Although the proteins were expressed in D 2O, during their isolation and purification, aqueous solutions were used and many of the amide deuterons exchanged with protons. This required back exchange to deuterium by dialysis. Slowly exchanging amide protons were replaced by deuterium by partial unfolding of the proteins in a urea buffer and refolding them in a D2O buffer. The

23

unfolding and refolding in the presence of urea may cause carbamylation of lysine residues, leading to the blue cross peaks in Fig. 5. Special measures have to be taken to minimize this undesired reaction. The main advantages of the asymmetric deuteration technique are the excellent signal-to-noise ratio obtained in a relatively short measurement time (1-2 days) even for dilute protein solutions (~0.2 mM), the very good resolution obtained after filtering out the intramolecular interactions within each protein, and the ease of implementation for proteins that can be expressed in E. coli. We believe this method can be extended to study protein complexes in the ~100 kDa range. A drawback of the technique is that it provides information only on interactions of aromatic protons with non-aromatic protons. Although aromatic residues are overrepresented at binding interfaces in protein complexes [38, 39], interactions that do not involve aromatic residues may also be important. Furthermore, assignment to the interacting residues is done only on the basis of proton chemical shifts, which may be a source of ambiguity. Assignments can be improved in an iterative manner by initial reference to low-resolution docking models that are based on the input of constraints from biochemical, mutagenesis and other physical chemical methods. Another option which has not been explored is to use 1

H,13C-labeled amino acids, in particular for methyl groups, in place of the unlabeled

ones and to record

13

C-separated 3D NOESY spectra. Such spectra will provide the

13

C

chemical shifts of the carbon atoms bonded to the protons and may facilitate unambiguous assignment. However, this will result in reduced sensitivity and may be applicable only to methyl protons.

24

6. Observation of intermolecular NOE interactions using double difference spectra and uniform deuteration All the methods described above, except the isotope-edited/isotope-filtered technique described in section 2, provide information regarding a subset of the intermolecular NOE interactions in protein complexes. As already mentioned, the isotopeedited/isotope-filtered experiment is usually limited to tight binding complexes with molecular weights of up to 35 kDa. For cases where the studies can be carried out at elevated temperature (~40 °C) and high protein concentrations (≥1mM), higher molecular weight complexes can also be investigated. Spec

Spec

Spec

Double

[1H-A/1H-B]

[1H-A/2H-B]

[2H-A/1H-B]

difference Spec

Intra-A

-f1*

Intra-A

-f2*

--------

=

--------

Intra-B

--------

Intra-B

--------

Inter- A/B

--------

--------

Inter- A/B

Fig. 6. A schematic illustration of the double-difference NOESY technique to extract intermolecular interactions in protein complexes. The proteins in the complex are designated A and B. The observed interactions in each spectrum (Spec) are designated: intra for intramolecular interactions and inter for intermolecular interactions. The coefficients f1 and f2 are the multipliers used for optimal subtraction [40]. We recently developed a double-difference NOESY spectrum method that has the potential to reveal intermolecular interactions between all non-exchangeable protons in large protein complexes [40]. This method was applied to the 44 kDa complex of interferon IFNα2 with its receptor IFNAR2. The filtering of intramolecular interactions in each of the components of the complex is achieved through the combination of uniform deuteration and subtraction, as schematically illustrated in Fig. 6. NOESY spectra are

25

measured for three samples of the complex with different deuteration combinations. The first spectrum is measured for the complex in which the two proteins are unlabeled. This spectrum shows the intramolecular interactions within each of the proteins as well as the intermolecular interactions between the two. This spectrum is very crowded and contains very few resolved NOE cross peaks. NOESY spectra in which protein A is unlabeled and B is deuterated and vice versa are then measured. The spectrum in which protein B is deuterated contains only intramolecular interactions within protein A and the second spectrum contains only intramolecular interactions in protein B. Neither spectrum contains intermolecular interactions. Sequential subtraction of these two spectra from the spectrum of the unlabeled protein complex reveals only intermolecular interactions. The subtraction procedure is not complicated but requires that the three NOESY spectra be measured at exactly the same sample conditions and temperature and similar concentrations of the complex (not necessarily identical). The procedure also involves careful adjustment of the phases of the measured spectra in the acquisition dimension as well as of the factor used for subtraction, which may differ slightly from the ideal factor of 1. It has been explained thoroughly in the original publication [40]. Using this strategy a well-resolved double difference spectra with an excellent signal-to-noise ratio was obtained, as shown in Fig. 7.

26

Fig. 7. A double difference NOESY spectrum of the IFN2/IFNAR2 complex showing intermolecular interactions between the two proteins [40]. Positive cross-peaks are due to intermolecular interactions and are in black. The observed negative crosspeaks are in red and result from intra-residue interactions involving residue F27 of IFN2, which occupies a binding pocket in IFNAR2. As a result of the deuteration of IFNAR2, protons of F27 of IFN2 have fewer relaxation pathways than in the complex in which both proteins are unlabeled. Therefore, the intraresidue NOEs within F27 in the IFN2 complex where IFNAR2 is deuterated are more intense and, in this special circumstance, residual signal from intraresidue NOEs of F27 are seen in the difference spectrum as negative cross peaks. Vertical lines indicate crosspeaks originating from the same aromatic proton and are labeled at the top of the spectrum according to the assignment of the specific proton. The aliphatic proton assignments are marked for each cross-peak in the spin system. IFNAR2 and IFN2 residues are labeled with superscripts r2 and a2, respectively. 27

A total of 97 intermolecular interactions were observed and assigned to the corresponding protons of interferon and its receptor [40]. Comparison of these NOEs with the intermolecular NOEs found on the same complex using asymmetric deuteration [37] (Section 5) showed high congruity. These 97 NOEs and the structures of uncomplexed interferon and interferon receptor were used as input in the calculation of the structure of the complex. This method led to a refined model for INF bound to its receptor that had RMSDs of 0.77 ± 0.12 Å and 1.08 ± 0.10 Å for all backbone and all heavy atoms, respectively. This NMR derived model was found to be in a very good agreement with the crystal structure of IFN/IFNAR1/IFNAR2 ternary complex that was published later on [41]. The double difference approach can be applied readily to complexes of proteins that can be expressed and perdeuterated in E. coli.

7. Protein-peptide intermolecular NOE interactions in complexes exhibiting tight binding using deuteration and difference spectroscopy. NMR has been extensively used to study small and medium size protein-peptide complexes exhibiting a wide range of binding affinities. However, there are only a few studies of intermolecular interactions in protein-peptide complexes exceeding 40 kDa molecular weight. Unlike proteins, short peptides can be synthesized and labeled, e.g. with 2H, at specific positions using solid-phase peptide synthesis. Such site-specific deuteration makes assignment very easy. In addition, like proteins, peptides also can be expressed in E. coli as fusion proteins, and thereby uniformly labeled and subsequently released from the carrier protein. As a result, all the methods described in the previous

28

sections for protein-protein complexes exhibiting tight binding can be equally applied for tightly bound protein-peptide complexes, and their description will not be repeated.

Fig. 8. A schematic illustration of the NOESY difference approach when a single residue (here A) of a peptide in a complex with a protein is deuterated. The inter- and intra-molecular NOE interactions that are not observed when the peptide residue is deuterated are indicated and marked as cancelled (X). The difference between the NOESY spectrum of the natural abundance (unlabeled) peptide in complex with the protein and the NOESY spectrum of the peptide/protein complex when a single peptide residue is deuterated reveals only the inter- and intramolecular NOEs of this residue. All other NOEs are cancelled out by subtraction [42].

As has already been mentioned above, deuteration can be used as a chemical filter to extract intermolecular interactions. The use of site-specific deuteration and difference spectroscopy further extends the applicability of this chemical filtering 29

approach. We developed a difference spectroscopy approach to study the 53 kDa Fab complex of a HIV-1 neutralizing antibody with a tightly bound 24-residue HIV-1 gp120 V3 peptide antigen, RP135 [42-44]. Six different RP135 molecules, in each of which a different single residue was deuterated, were prepared by chemical synthesis [42]. These included I7, I9, P13, A16, F17 and V18 of the peptide (the numbering is according to the entire V3 sequence). Initially, for simplicity, only amino acids that did not require side chain protection were selected for deuteration. Two NOESY spectra were measured in D2O, one for the Fab complex with the natural abundance (unlabeled) peptide and the other for the Fab complex with the peptide deuterated at a single position. The NOESY difference spectrum showed several cross peaks with excellent signal-to-noise ratio [42]. Only NOE crosspeaks involving the deuterated residue were retained in the difference spectrum. All other crosspeaks resulting from intra- and intermolecular interactions of the nondeuterated protons were removed by the subtraction. A schematic illustration of this procedure is shown in Fig. 8. The chemical shifts of the six residues that were deuterated were assigned to the corresponding protons by computing six 1D difference spectra from spectra recorded of the unlabeled complex and of a complex of the Fab bound to the peptide specifically deuterated at a given type of residue. These difference spectra showed the backbone and sidechain resonances only of the residue that was deuterated. These chemical shift assignments enabled us to easily assign those NOE cross peaks due to intramolecular interactions. In addition to the intramolecular NOEs, we observed NOE interactions involving aromatic protons that could not be assigned to the only aromatic residues of RP135 (F17 and H8). These NOEs were assigned to intermolecular interactions with

30

aromatic residues of the Fab [42]. In a follow-up study ten residues were deuterated in a single peptide molecule (P1053; RKSIRIQRGPGRAFVTIG), which is a shorter version of RP135 [43]. As shown in Fig. 9, multiple deuteration in one peptide molecule provided a spectrum showing a large number of intermolecular NOE interactions between the aromatic protons of the antibody and the peptide residues. In other experiments amino acid-specific deuteration of the Fab aromatic residues was used to assign the NOE interactions to the specific aromatic amino acid types in the antibody. Specific labeling of either the light or heavy chain of the antibody further enabled the assignment of the observed NOEs to a specific antibody polypeptide chain. The data that were obtained were used to derive a model for the antibody complex with the peptide antigen [44]. The measurements were carried out on a Bruker 500 MHz spectrometer equipped with a conventional probe and considering the size of the complex the identification of pairwise interactions between the peptide and the antibody was a major achievement at that time (late 1990s).

31

Fig. 9. A NOESY difference spectrum showing the interactions of aromatic protons of the Fab fragment of the HIV-1 neutralizing antibody 0.5 and Phe-17 of a gp120 V3 peptide (P1053) with non-aromatic protons of the peptide [44]. The difference shown here is that between the NOESY spectrum of 0.5 Fab complex with P1053 (RKSIRIQRGPGRAFVTIG) and the spectrum of the Fab complex with P1053 in which Ile-7, Ile-9, Gln-10, Gly-12, Pro-13, Gly-14, Ala-16, Phe-17, Val-18, and Ile20 were deuterated. The assignments of the cross peaks to the peptide protons involved is marked in F1 and the assignment to amino acid type of the antibody (i.e. Y, H, F) and to phenylalanine of the peptide is marked in F2. Cross peaks denoted “Fab” in F1 have been assigned to Fab protons but could not be assigned to amino acid type.

32

8. Identification of intermolecular protein peptide NOEs using transferred NOE 8.1. Introduction The methods and approaches described above have been mostly used for systems exhibiting medium to tight binding and slow exchange. It is somewhat surprising to the authors of this review that many NMR spectroscopists assume it is almost impossible to detect intermolecular interactions for complexes exhibiting weak binding and fast exchange [3]. This assumption is even less understandable in view of the fact that the first reports on the observation of intermolecular interactions in macromolecular complexes by NMR, and of the assignment of these to specific protons of the bound ligand, were for protein/small ligand complexes exhibiting fast exchange. The groundbreaking experiments of Bothner–By, Cohn and James date back to the early 1970s and preceded the development of 2D NMR spectroscopy and modern NMR spectrometers [12-14]. As a matter of fact, these seminal studies were conducted using CW spectrometers making them spectacular achievements for that time. Although these early publications did not use the term transferred NOE (TRNOE), Clore and Gronenborn justifiably referred to them as the first examples of TRNOE [9]. We find it very odd that, despite its far-reaching potential, the application of TRNOE to the study of intermolecular interactions in large protein complexes exhibiting fast off-rates has been neglected for so long. To date, in the vast majority of investigations, TRNOE has been applied to study the conformation of ligands bound to large proteins. Accordingly, the definition of TRNOE that has been used by most investigators has been restricted to the observation of cross relaxation between protons in the bound ligand observed through NOE cross 33

peaks between free ligand protons. However, in a broader definition Clore and Gronenborn noted that the term TRNOE may also be applied to the detection of interactions between a protein and a small bound ligand through the observation of intermolecular NOE cross peaks via the free ligand protons [9]. The phenomenon occurs because magnetization transferred from protons of the protein to those of the bound ligand is carried with the ligand when it exchanges to the free state during the NOE mixing period. The intermolecular TRNOE cross peaks are characterized by the chemical shifts of the protein protons and the free ligand protons (or the average between the chemical shift of the free and bound ligand protons). These cross peaks are broad in one dimension and sharp in the second dimension. This is unlike intramolecular TRNOE cross peaks, which are sharp in both dimensions. As a result, the sensitivity gained by using TRNOE for large molecules in complex with small ligands is much larger for the intra-ligand TRNOEs than for the intermolecular TRNOE cross peaks. Intermolecular TRNOE experiments are conducted using a 4 to 10-fold molar excess of the peptide ligand with respect to the protein. Under such conditions, for a protein/peptide complex exhibiting fast exchange relative to the inverse of the T1 relaxation time of the protein and peptide protons, and when the length of the mixing period allows sufficient exchange between bound and free ligand, the NOESY spectrum contains the following types of cross peaks: a) numerous cross peaks due to intramolecular interactions within the protein, b) TRNOE cross peaks due to intramolecular interactions within the bound peptide (these are observed through the free peptide), c) TRNOE cross peaks due to intermolecular interactions between the

34

protein and the peptide (these are observed through the free peptide), and d) NOEs between protons in the free peptide [15]. For peptides shorter than 15 residues long and for measurements at temperatures higher than 30 ºC the NOE interactions within the free peptide are negligible. The TRNOE cross peaks involving the peptide protons have resonance frequencies of the free peptide protons or of the average between the resonance frequencies of the free and bound forms. In the case where the chemical exchange rate is slow in comparison with the change in chemical shift upon binding, the NOESY spectra also will contain exchange cross peaks linking free and bound signals of the peptide ligand, as well as TRNOE cross peaks due to interproton contacts within the peptide-protein complex [15].

8.2. TRNOE difference spectroscopy NOESY spectra of high molecular weight complexes are characterized by severe cross peak overlap and poor resolution. To remove the numerous cross peaks due to intramolecular interactions within the protein we have developed two different approaches [15, 16]. In the first approach one must measure two NOESY spectra. The first spectrum is of the protein and the peptide in a 1:1 complex. This spectrum contains NOE cross peaks due to intramolecular interactions within the protein, intramolecular interactions in the bound peptide and intermolecular interactions between the protein and the bound peptide. When the free protein concentration is significantly above the KD and there is no peptide excess, effectively there is no measurable free peptide and therefore no transferred NOE (vida infra). Accordingly, all of these cross peaks exhibit broad linewidths in both dimensions. The second spectrum is of the protein in the

35

presence of a large (4-5 fold molar) excess of the peptide ligand, which is in fast exchange on the T1 scale. Subtraction of the first spectrum from the second spectrum, in principle, eliminates the numerous cross peaks due to intramolecular interactions within the protein. Also removed are those due to intermolecular peptide-protein interactions or to intramolecular interactions involving any bound peptide proton resonances exhibiting slow exchange on the chemical-shift scale. The resulting difference NOESY spectrum thus exhibits only TRNOE cross peaks due to intramolecular interactions within the bound peptide and due to intermolecular NOE interactions between the protein and the bound peptide as revealed through the free peptide resonances (Fig. 10). The intrapeptide cross peaks will have proton chemical shifts corresponding to population-weighted averages of the free and bound peptide forms, which as the peptide is in large excess will be close to those of the free form, and will have narrow linewidths in both dimensions. Many of these cross peaks, especially those due to intraresidue interactions, are very intense. The intermolecular TRNOE peaks will have proton chemical shifts corresponding to population-weighted averages of the free and bound peptide forms in one dimension, and of the protons of the protein in the second dimension. These will be much broader in the protein dimension.

36

Fig. 10. The aromatic region of the difference between the NOESY spectrum of the Fab fragment of the anti-cholera toxin peptide antibody TE33 saturated with the cholera toxin peptide CTP3 and the NOESY spectrum of the Fab in the presence of a 4-fold excess of CTP3 [45]. Neither the Fab nor the peptide are labeled by deuteration. Interactions between aromatic protons of the antibody and protons of the peptide are shown. Assigned antibody residues are marked by capital letters and arbitrary numbers; peptide protons are marked by lower case letters and numbers denoting the location in the peptide sequence of the residues to which the protons belong. Additional cross peaks marked by h8 are of the peptide his 8 and are due to exchange between bound and free peptide or intraresidue interactions in the bound peptide.

The cross peaks due to intermolecular TRNOE interactions are considerably less intense than those due to intramolecular TRNOE interactions within the bound peptide.

37

To avoid losing these cross peaks in the background of the multiple and intense NOEs due to intramolecular interactions within the peptide, a smaller peptide to protein molar ratio should be used in comparison with that used for observing intramolecular TRNOE interactions within the peptide. With a lower peptide excess the T2s of the bound peptide protons are shortened in comparison with those of the free peptide and this alleviates problems due to T1 noise and truncation artifacts caused by using an acquisition time in the indirect dimension (F1) that is too short for the decay of the magnetization of the peptide protons, which thereby become truncated. This point needs to be considered because of the very pronounced difference in the T2 values of the protein and peptide protons. It should be realized that the decreased excess does impact the signal-to-noise ratio of the TRNOE phenomenon and also the resolution of the spectra. Therefore, given the opposite effects that peptide concentration has on the truncation artifacts and signal-to-noise and resolution in the TRNOE experiment, compromises must be made as to the conditions chosen. For practical reasons, in our applications of TRNOE difference spectroscopy, we chose to focus on the interactions of antibody aromatic residues with all residues of the peptide antigen. The cross peaks representing these interactions are observed in the left half of the NOESY spectrum, which is relatively free from T1 noise and truncation artifacts caused by the non-aromatic protons of the peptide, especially of the methyl protons, when a large excess of the peptide is present. The peptide antigen we studied (cholera toxin peptide antigen–CTP3) contained only one aromatic residue (his8; the peptide residues are labeled with lower case letters see Fig. 10), the resonance of which could be easily identified. All other cross peaks in the left half of the NOESY

38

spectrum were due to intermolecular interactions between the aromatic residues in the Fab fragment of the antibody and the residues of the peptide immunogen. By selectively deuterating the Trp and Phe residues of the antibody we identified the contribution of antibody Tyr residues to binding. Similarly we identified the contributions of Trp and Phe residues to CTP3 binding by deuteration of Tyr and Phe, or of Tyr and Trp, respectively [15]. The NOESY difference spectrum with the assignment to specific amino acids type is presented in Fig 10. In collaboration with Michael Levitt, models of the combining site of the Fab fragment of the antibody were computed [45]. Together with the NMRderived knowledge of the interacting aromatic residues obtained by difference TRNOE, this model was used to manually fit the Fig. 11. Superposition of the five best-fit calculated models for the polypeptide

backbone

of

the

segment VPGSQHID of the peptide CTP3 bound to TE33 obtained on the

basis

intramolecular

of

inter

NMR

and derived

distance constraints [11].

peptide into the binding site, and to obtain residue specific assignments for the 35 intermolecular cross peaks to protons of the antibody [11]. The chemical shift assignments of the interacting protons of the peptide were based on NOESY and HOHAHA spectra. These

39

standard 2D spectra were measured using excess peptide. An additional 17 medium and long range constraints on intramolecular interactions within the bound peptide were used in energy minimization calculations to create high-resolution models for three Fab/peptide antibody complexes (Fig. 11) [11]. Subsequently, an X-ray structure of one of these antibodies, TE33 [46], showed excellent agreement with the NMR derived structure, with backbone RMSDs of 0.85 Å for the antibody and of 1.6 Å for the bound peptide antigen [47]. As mentioned above, our intermolecular TRNOE investigations were limited to studying intermolecular interactions involving aromatic amino acids of the antibody. These studies led to the conclusion that aromatic amino acids residues make a major contribution to the binding of peptide antigens, a conclusion that was later confirmed for other antibody/antigen complexes. Recent bioinformatics investigations indicate that aromatic amino acid residues are important for intermolecular interactions in many protein complexes and this is now becoming widely accepted by the protein chemistry community [39].

40

8.3. T1ρ-filter

Fig. 12. The pulse sequence for the

F1

T1ρ-filtered

NOESY

experiment [16].

Although the results obtained using the TRNOE difference spectrum were useful and led to reliable models for the complex of antibodies and cholera toxin peptides, the method can become laborious since two samples need to be prepared and measured and the phasing of the two spectra needs to be carefully adjusted for best results. To simplify the process of extracting intermolecular TRNOEs in large protein complexes from NOESY spectra we developed the T1ρ-filtering approach [16]. This takes advantage of the large difference in the transverse relaxation times T2 and T1ρ of the protons of the peptide relative to those of the large protein. As shown in Fig. 12, a spin lock pulse with a duration that was adjusted to eliminate the magnetization of the protein protons was placed immediately following the first 90º pulse. This spin lock pulse leads to some attenuation of the peptide magnetization. However, most of the peptide proton magnetization is retained after the spin-lock pulse and it is frequency labeled during the t1 evolution period. During the mixing period, protons of the peptide molecules that were bound to the protein transferred magnetization to nearby protein protons. After the mixing, magnetization was found on the peptide protons and on those protein protons that had received magnetization from the peptide. The T 1ρ-filtered NOESY experiment revealed intramolecular interactions within the peptide as well as intermolecular

41

interactions between the protein and the peptide. The spectrum was symmetric with respect to intramolecular interactions within the peptide. However, the cross peaks due to intermolecular interactions appeared just on one side of the diagonal because the filtering of the protein proton resonances prevents transfer from these protons to the peptide. The T1ρ relaxation time of the peptide is the average of the T1ρ values in the bound and free states. To increase the difference between the T 1ρ of the peptide and that of the protein, a larger excess of the peptide needs to be used in comparison with the difference method. This is required in order to minimize the loss in the peptide magnetization when the spin-lock pulse is applied. The T1ρ filter is very easy to implement, and gives spectra that do not suffer from subtraction artifacts and have better baselines than those obtained with the difference techniques. However, the T 1ρfiltered NOESY spectrum still contains intraprotein NOEs from protein protons that are mobile and thereby have longer T2 and T1ρ relaxation times. The NOE cross peaks due to intramolecular interaction between mobile protein protons can be distinguished from intermolecular interactions between the peptide and protein protons on the basis of their chemical shifts. The cross peaks due to intermolecular interactions should have the peptide frequency in one dimension. When applied to the study of the anticholera toxin antibodies interacting with CTP3 peptides, in general, there was an excellent agreement between the TRNOE difference spectrum results and the T 1ρ filtered NOESY results [16]. When the peptide contains several aromatic residues and the intermolecular interactions of these residues are investigated, the T1ρ filter can be placed just before the acquisition instead of just

42

after the first 90º pulse. This improves resolution and allows the characterization of the interaction of the peptide’s aromatic protons (see below).

8.4. TRNOE in combination with asymmetric deuteration TRNOE spectra obtained using the 2D difference technique or the T 1ρ-filter show both TRNOE cross peaks due to intramolecular interactions within the bound peptide as well as cross peaks due to interactions between the peptide and the protein. As mentioned above, the TRNOE due to intramolecular interactions within the peptide are much stronger than those due to intermolecular interactions and can mask these crucial indicators of structure of the protein-peptide complex. Most specifically, when the peptide ligand contains several aromatic residues, the intramolecular interactions of the peptide’s aromatic protons with its non-aromatic protons can mask interactions of peptide aromatic protons with non-aromatic protein protons, as well as interactions of protein aromatic protons with peptide non-aromatic protons. The asymmetric deuteration approach that we developed for studying intermolecular interactions in interferon complexed with its receptor [37] is a very powerful and direct way to eliminate cross peaks due to intramolecular interactions (see above). This approach can be used in combination with TRNOE measurements to eliminate intramolecular interactions involving aromatic residues with non-aromatic protons in both the protein and the peptide. Recently, we used this approach to identify the interactions of the aromatic protons of the four tyrosine residues of a 27-residue peptide derived from the N-terminal extracellular segment of the CCR5 chemokine receptor [Nt-CCR5(1-27)] with the gp120 envelope protein of HIV-1 (unpublished

43

results). Biochemical and genetic analyses suggested that the putative binding site for Nt-CCR5 contains many methyl containing amino acids, especially Ile residues. TRNOE studies were carried out to determine whether these methyl protons of gp120 interact with the Nt-CCR5 tyrosine residues. Since Nt-CCR5(1-27) contains three isoleucine and two valine residues, intramolecular methyl-aromatic interactions involving peptide residues were a potential complication. A T 1ρ-filtered NOESY spectrum of a gp120 in complex with the CD4-mimic peptide CD4M33 [48] in the presence of a 10-fold molar concentration of Nt-CCR5(1-27) shows that the spectrum obtained is dominated by intramolecular interactions between the four tyrosine residues and the methyl protons of Ile and Val residues of the peptide. To eliminate the intramolecular interactions involving aromatic protons in both gp120 and Nt-CCR5(1-27), we prepared a gp120 molecule in which all aromatic residues were deuterated. In addition, the isoleucine and valine residues of Nt-CCR5(1-27) were deuterated. This labeling scheme eliminated all intramolecular interactions involving aromatic protons within gp120, as well as the intramolecular interactions within Nt-CCR5(1-27) involving the isoleucine and valine methyl protons. As a result of this labeling scheme, the spectral region (6.0 to 8.0 ppm by 0.5 to 1.1 ppm) showing interactions between aromatic protons and methyl protons should be free from intramolecular interactions in both gp120 and Nt-CCR5(1-27). The NOESY spectrum of the asymmetrically labeled gp120/CD4M33 complex with excess Nt-CCR5(1-27) showed several cross peaks in the aromatic/methyl region. The intensity of the cross peaks and the chemical shift of the Nt-CCR5(1-27) aromatic protons was dependent on the peptide excess, providing evidence that these are indeed TRNOE cross peaks. All these cross peaks disappeared in samples where the Leu, Ile, Val and

44

Thr residues of gp120 were also deuterated. Deuteration of gp120 IIe residues was sufficient to cause very pronounced reduction in the cross peak intensities, suggesting that Ile residues of gp120 are the major contributors to the observed intermolecular interactions. A T1ρ filter applied just before the acquisition period helped remove residual intensity from cross peaks due to intramolecular interactions in the gp120/CD4M33 part of the ternary complex and considerably improved the baseline. Our results on the gp120/CD4M33/Nt-CCR5 complex demonstrate that the TRNOE in combination with asymmetric deuteration of proteins and peptide ligands can be a powerful tool to observe intermolecular interactions in large protein complexes in the ~100 kDa range. In our preliminary studies the method was used to identify interactions between aromatic protons of the Nt-CCR5(1-27) peptide ligand and methyl protons of the gp120 protein. Bacterial expression of the protein and the peptide ligand will enable study of intermolecular interactions between all aromatic residues of the ligand and all non-aromatic protons of the protein and vice-versa. The technique can pinpoint crucial binding interactions in biological systems, and should provide valuable information for the development of drugs targeting interaction surfaces in large protein complexes.

9. NOESY difference spectra using a spin-labeled ligand Our analysis of NMR theory and the literature leads us to conclude that homonuclear 2D NOESY experiments can be used to study intermolecular as well as intramolecular interactions in large protein complexes in the binding site region, if methods are used to extract sub-spectra showing these inter and intramolecular interactions by eliminating most or all NOEs due to intramolecular interactions within the binding protein.

45

Paramagnetic relaxation enhancement (PRE) is the result of a dipole-dipole interaction between the nuclei of the protein and an unpaired electron of a spin label or a paramagnetic center [49, 50]. Such interactions lead to enhanced relaxation rates and thereby disappearance of signals from the NMR spectrum. The PRE effect has a r-6 dependency where r is the distance between the nucleus and the unpaired electron. The magnetic dipole of the electron is very large and therefore PREs are seen over large distances (>30 Å). When PREs are operative, signals are considerably broadened.

Therefore,

in

combination

with

difference

spectroscopy,

PRE

measurements can provide information about interacting nuclei either within one molecule or between molecules, for example, such as those occurring at the ligand binding site of a protein when the ligand is spin-labeled. This approach has been used in concert with NOESY measurements to refine models of protein complexes.

46

Fig. 13. Intra- and intermolecular interactions

in

the

TE33

Fab

complexed with the CTP3 peptide [53]. The 2D spectrum shown is the difference

between

the

NOESY

spectrum of the Fab in the presence of a five-fold excess CTP3 peptide and the NOESY spectrum of the Fab in the presence of equimolar ratio of the spin-labeled peptide. Antibody tryptophan

and

phenylalanine

residues are perdeuterated, while tyrosine residues are deuterated at the

δ

positions

positions,

leaving

unlabeled.

the

ε

Assigned

antibody residues are marked by capital letters and arbitrary numbers; peptide protons are marked by lower case letters and numbers denoting the location in the peptide sequence.

47

Two dimensional difference spectra were calculated from the NOESY spectra of flavodoxin in complex with reduced and oxidized riboflavin 5’-monophosphate (FMN). The NOESY difference spectrum of this 15 kDa complex contains NOE cross peaks of protons that are up to 15 Å away from the paramagnetic center [51]. Subspectra containing NOEs from smaller shells around the paramagnetic center can be obtained by varying the concentration of the bound semiquinone radical if the ligand is in fast exchange. Some assignments to intramolecular NOE interactions were obtained in this pioneering study of a protein complex using NOESY spectra. A similar approach was used by Hilbers and co-workers to study the binding site of the single-stranded DNA binding protein IKeGVP [52]. Although a number of residues and interactions at the binding site were identified, the observation of specific intermolecular interactions was not reported. One of the challenges in studying large protein complexes is determining interactions between residues in the binding site of the protein, because these are often very broad and obscured by background interactions between similar residues throughout the protein. Spin labels broaden and often eliminate NMR signals from protons within 15 Å of the label. As shown in Fig. 13, we used this fact to calculate the 2D difference NOESY spectrum from the NOESY spectrum of a Fab fragment of an antibody in the presence of a 5 fold molar excess of CTP3 and the NOESY spectrum of the Fab in the presence of a 1:1 ratio of CTP3 spin labeled at its N-terminus [53]. The attachment of the spin-label had no effect on the binding constant of the peptide. All of the aromatic residues in the CTP3 binding site were very much broadened by the spinlabel. Since in this experiment the Tyr residues were deuterated on the δ position of the

48

ring and protonated on the ε position, sharp lines in the difference spectrum were obtained for the ε protons of the Tyr residues. This was the result of both elimination of the (δ, ε) J coupling and removal of a T2 relaxation pathway. Therefore, the difference spectrum involving the spin-labeled peptide antigen allowed us to observe all the interactions of the protons of Tyr residues of the antibody with protons of the peptide. Moreover, we were able to assign seven inter-residue interactions involving aromatic residues in the antibody binding site. These interactions helped us to refine the model of the antibody and to reduce the RMSD of the NMR restraint violations from 0.32 Å to 0.27 Å, and the RMSD between the NMR model for CDR3 and the crystal structure from 1.45 Å to 1.14 Å [53].

10.

Assignment of intermolecular NOEs to specific residues in large protein

complexes To this point we have presented powerful methods to extract NOESY data on intermolecular interactions in large protein complexes. These methods overcome the severe overlap in the NOESY spectra of large protein complexes and the reduction in signal-to-noise ratio found in practically all NMR experiments due to the enhanced relaxation of the 1H,

13

C and

15

N nuclei as the molecular weight of the complexes

increases. To make use of the observed intermolecular interactions one must assign them to specific protons of the residues involved in these interactions. However, the chemical shift assignments of the side-chain protons of individual residues of large proteins present a challenging task.

49

Complete or nearly complete sequential and side-chain assignment of

13

C and

15

N nuclei for proteins in the range of 35-100 kDa can be achieved using uniform

deuteration of the proteins of interest [27]. Once these assignments are obtained, sequential and side chain assignment for the protons can be achieved using the method developed by Yang and coworkers [23]. This method relies on the most sensitive triple resonance experiments, namely

15

N and

13

C separated NOESY, HNCA and MQ-HCCH-

TOCSY, and uses computer programs to cluster the spin systems of individual residues and to find sequential connectivities. In favorable cases the method can be applied for proteins up to 45 kDa molecular weight and to protein complexes up to 65 kDa when one of the proteins in the complex is made NMR silent by deuteration or is not labeled with 15N and 13C. The methyl labeling techniques and the methyl TROSY experiments developed by Kay and coworkers enable the assignment of side-chain methyl-protons to specific residues of the protein [26, 32, 54]. When used in the background of otherwise highly deuterated proteins, molecules up to 100 kDa and perhaps larger may be studied by this approach [55]. However, the obtained structures are based on the use of minimal restraints. Using SAIL (Stereo array isotopic labeled) amino acids may also significantly extend the size limit for which sequential backbone and side chain proton assignment can be obtained [56]. However, this approach requires protein expression in cell-free systems and is very expensive. As we have alluded to in reviewing the literature, inclusion of specifically deuterated amino acids in the expression system is a powerful way to identify the types of amino acid involved in particular interactions. This is especially useful in

50

distinguishing methyl protons of Val, Ile or Leu residues when these cannot be directly assigned solely on the basis of their proton chemical shifts. Once the amino acid type has been identified, the identification of specific residues at the binding interface of large proteins can be addressed using site-directed mutagenesis. The double mutant cycle approach [57, 58], which provides information about pairwise interactions in protein complexes, is also complementary to NMR investigations and can help pin-point interacting pairs. Crystal structures of unliganded proteins or their models are then combined with the NMR and biochemically derived restraints to provide a highresolution map of the binding interface. Docking of the two interacting molecules using floating assignment is a useful approach for resolving ambiguities in the assignment of intermolecular NOEs (HADDOCK, XPLOR-NIH) [59, 60].

11.

Conclusions and future perspectives

The isotope-edited/isotope-filter technique is probably the most commonly used NMR technique to study intermolecular interactions in protein complexes. One of the main advantages of this technique is that, in principle, it can be used to determine all intermolecular interactions in the complex. However, this technique suffers from poor signal-to-noise ratios with increasing size of the complex. Homonuclear 2D-NOESY difference techniques provide the highest sensitivity for the detection of intermolecular interactions. To eliminate the background due to intramolecular NOE interactions, the double difference homonuclear NOESY technique can be used to provide a map of all intermolecular interactions of non-exchangeable protons in the complex. It is possible to apply this method for nearly any complex of two proteins that can be expressed in E.

51

coli. If one of the interacting species is a small peptide ligand, chemical deuteration of individual residues in combination with difference spectroscopy can be used for identification. It remains surprising to us that TRNOE has not been more widely used to study intermolecular interactions protein-peptide complexes, as it can in principle be applied for practically any such system where chemical shift titrations can be used to map the binding interface, or for systems in which STD can be used to determine the peptide residues involved in binding. While it is obvious that even in the TRNOE experiment the binding protein is still characterized by a short T2, it is also true that the peptide ligand, which is present in large excess, exhibits much longer T2 values. This will enable the extension of the applicability of the combination of isotope-edited/isotope-filter technique with TRNOE to much larger proteins. Asymmetric deuteration together with isotopeedited/isotope-filtered or TRNOE is also a powerful approach to eliminate NOEs due to intramolecular interactions and thereby reveal those that originate from intermolecular TRNOE interactions. An important advantage of TRNOE measurements over the editing experiments is that it can be used for systems exhibiting weak binding. Such systems are often not amenable to X-ray crystallographic analysis because the weakly bound peptide ligand results in a poor electron density map. Thus, we posit that the TRNOE experiment can fill a special niche wherein NMR has a unique opportunity to provide structural information at atomic resolution that cannot be obtained by other techniques.

Acknowledgment

52

We thank Prof. Lewis Kay for valuable comments and for critical reading of the manuscript. This study was supported by the Minerva Foundation with funding from the Federal German Ministry for Education and Research, by the Comisaroff Family Trust and by the Kimmelman Center. J.A. is the Dr. Joseph and Ruth Owades Professor of Chemistry. F.N is the Leonard and Esther Kurtz Term Professor at CSI of CUNY. F.N. was supported by the Erna and Jakob Michael Visiting Professorship while on Sabbatical at the Weizmann Institute of Science.

53

References [1]

H.M. McConnell, Reactions Rates by Nuclear Magnetic Resonance, J. Chem. Phys. (1958) 430-431.

[2]

O. Jardetzky, G.C.K. Roberts, NMR in Molecular Biology, Academic Press, New York, 1981.

[3]

C. Gobl, T. Madl, B. Simon, M. Sattler, NMR approaches for structural analysis of multidomain proteins and complexes in solution, Prog. Nucl. Magn. Reson. Spectrosc. 80 (2014) 26-63.

[4]

N. London, B. Raveh, O. Schueler-Furman, Peptide docking and structure-based characterization of peptide binding: from knowledge to know-how, Curr. Opin. Stru. Biol. 23 (2013) 894-902.

[5]

N. London, D. Movshovitz-Attias, O. Schueler-Furman, The Structural Basis of PeptideProtein Binding Strategies, Structure 18 (2010) 188-199.

[6]

G. Otting, K. Wuthrich, Heteronuclear Filters in 2-Dimensional [H-1, H-1] NmrSpectroscopy - Combined Use with Isotope Labeling for Studies of Macromolecular Conformation and Intermolecular Interactions, Quart. Rev. Biophy. 23 (1990) 39-96.

[7]

A.L. Breeze, Isotope-filtered NMR methods for the study of biomolecular structure and interactions, Prog. Nucl. Magn. Reson. Spectrosc. 36 (2000) 323-372.

[8]

P.C. Driscoll, in: Protein NMR Spectroscopy: Practical Techniques and Applications, eds G.C.K. Roberts, L.-Y. Lian, Wiley, 2011.

[9]

G.M. Clore, A.M. Gronenborn, Theory and applications of the transferred nuclear Overhauser effect to the study of the conformations of small ligands bound to proteins, J. Magn. Res. 48 (1982) 402-417.

54

[10] G.M. Clore, A.M. Gronenborn, Theory of the Time-Dependent Transferred Nuclear Overhauser Effect - Applications to Structural-Analysis of Ligand Protein Complexes in Solution, J. Magn. Res. 53 (1983) 423-442. [11] T. Scherf, R. Hiller, F. Naider, M. Levitt, J. Anglister, Induced peptide conformations in different antibody complexes: molecular modeling of the three-dimensional structure of peptide-antibody complexes using NMR-derived distance restraints, Biochemistry 31 (1992) 6884-6897. [12] P. Balaram, A.A. Bothner-By, J. Dadok, Negative nuclear Overhauser effects as probes of macromolecular structure, J. Am. Chem. Soc. 94 (1972) 4015-4017. [13] T.L. James, M. Cohn, Role of Lysyl Residue at Active-Site of Creatine-Kinase - Nuclear Overhauser Effect Studies, J. Biol. Chem. 249 (1974) 2599-2604. [14] T.L. James, Binding of Adenosine 5'-Diphosphate to Creatine-Kinase - Investigation Using Intermolecular Nuclear Overhauser Effect Measurements, Biochemistry 15 (1976) 47244730. [15] J. Anglister, R. Levy, T. Scherf, Interactions of antibody aromatic residues with a peptide of cholera toxin observed by two-dimensional transferred nuclear Overhauser effect difference spectroscopy, Biochemistry 28 (1989) 3360-3365. [16] T. Scherf, J. Anglister, A T1 rho-filtered two-dimensional transferred NOE spectrum for studying antibody interactions with peptide antigens, Biophys J. 64 (1993) 754-761. [17] W. Lee, M.J. Revington, C. Arrowsmith, L.E. Kay, A pulsed field gradient isotope-filtered 3D 13C HMQC-NOESY experiment for extracting intermolecular NOE contacts in molecular complexes, FEBS Lett. 350 (1994) 87-90.

55

[18] C. Zwahlen, P. Legault, S.J.F. Vincent, J. Greenblatt, R. Konrat, L.E. Kay, Methods for measurement of intermolecular NOEs by multinuclear NMR spectroscopy: Application to a bacteriophage lambda N- peptide/boxB RNA complex, J. Am. Chem. Soc. 119 (1997) 6711-6721. [19] M. Piotto, V. Saudek, V. Sklenar, Gradient-tailored excitation for single-quantum NMR spectroscopy of aqueous solutions, J. Biomol. NMR 2 (1992) 661-665. [20] M. Caffrey, M. Cai, J. Kaufman, S.J. Stahl, P.T. Wingfield, D.G. Covell, A.M. Gronenborn, G.M. Clore, Three-dimensional solution structure of the 44 kDa ectodomain of SIV gp41., EMBO J. 17 (1998) 4572-4584. [21] D.S. Garrett, Y.J. Seok, A. Peterkofsky, A.M. Gronenborn, G.M. Clore, Solution structure of the 40,000 Mr phosphoryl transfer complex between the N-terminal domain of enzyme I and HPr., Nat. Struct. Biol. 6 (1999) 166-173. [22] D.C. Williams, Jr., M. Cai, G.M. Clore, Molecular basis for synergistic transcriptional activation by Oct1 and Sox2 revealed from the solution structure of the 42-kDa Oct1.Sox2.Hoxb1-DNA ternary transcription factor complex, J. Biol. Chem. 279 (2004) 1449-1457. [23] Y. Xu, Y. Zheng, J.S. Fan, D. Yang, A new strategy for structure determination of large proteins in solution without deuteration, Nature Methods 3 (2006) 931-937. [24] L. Jaremko, M. Jaremko, K. Giller, S. Becker, M. Zweckstetter, Structure of the Mitochondrial Translocator Protein in Complex with a Diagnostic Ligand, Science 343 (2014) 1363-1366.

56

[25] T.M. Handel, P.J. Domaille, Heteronuclear (H-1, C-13, N-15) NMR assignments and solution structure of the monocyte chemoattractant protein-1 (MCP-1) dimer, Biochemistry 35 (1996) 6569-6584. [26] N.K. Goto, K.H. Gardner, G.A. Mueller, R.C. Willis, L.E. Kay, A robust and cost-effective method for the production of Val, Leu, Ile (delta 1) methyl-protonated 15N-, 13C-, 2Hlabeled proteins, J. Biomol. NMR 13 (1999) 369-374. [27] V. Tugarinov, R. Muhandiram, A. Ayed, L.E. Kay, Four-dimensional NMR spectroscopy of a 723-residue protein: chemical shift assignments and secondary structure of malate synthase g, J. Am. Chem. Soc. 124 (2002) 10025-10035. [28] N.J. Traaseth, R. Verardi, G. Veglia, Asymmetric methyl group labeling as a probe of membrane protein homo-oligomers by NMR spectroscopy, J. Am. Chem. Soc. 130 (2008) 2400-2401. [29] R. Rosenzweig, L.E. Kay, Bringing Dynamic Molecular Machines into Focus by MethylTROSY NMR, Ann. Rev. Biochem. Vol 83 83 (2014) 291-315. [30] P. Guntert, L. Buchner, Combined automated NOE assignment and structure calculation with CYANA, J. Biomol. NMR 62 (2015) 453-471. [31] T. Saio, X. Guan, P. Rossi, A. Economou, C.G. Kalodimos, Structural Basis for Protein Antiaggregation Activity of the Trigger Factor Chaperone, Science 344 (2014) 597-+. [32] V. Tugarinov, P.M. Hwang, J.E. Ollerenshaw, L.E. Kay, Cross-correlated relaxation enhanced H-1-C-13 NMR spectroscopy of methyl groups in very high molecular weight proteins and protein complexes, J. Am. Chem. Soc. 125 (2003) 10420-10428.

57

[33] K.J. Walters, H. Matsuo, G. Wagner, A simple method to distinguish intermonomer nuclear overhauser effects in homodimeric proteins with C-2 symmetry, J. Am. Chem. Soc. 119 (1997) 5958-5959. [34] J. Anglister, S. Grzesiek, H. Ren, C.B. Klee, A. Bax, Isotope-edited multidimensional NMR of calcineurin B in the presence of the non-deuterated detergent CHAPS, J. Biomol. NMR 3 (1993) 121-126. [35] K. Takeuchi, M.H. Roehrl, Z.Y. Sun, G. Wagner, Structure of the calcineurin-NFAT complex: defining a T cell activation switch using solution NMR and crystal coordinates, Structure 15 (2007) 587-597. [36] A. Jasanoff, An asymmetric deuterium labeling strategy to identify interprotomer and intraprotomer NOEs in oligomeric proteins, J. Biomol. NMR 12 (1998) 299-306. [37] I. Nudelman, S.R. Akabayov, E. Schnur, Z. Biron, R. Levy, Y. Xu, D. Yang, J. Anglister, Intermolecular interactions in a 44 kDa interferon-receptor complex detected by asymmetric reverse-protonation and two-dimensional NOESY, Biochemistry 49 (2010) 5117-5133. [38] J.L. Asensio, A. Arda, F.J. Canada, J. Jimenez-Barbero, Carbohydrate-Aromatic Interactions, Acc. Chem. Res. 46 (2013) 946-954. [39] I.S. Moreira, P.A. Fernandes, M.J. Ramos, Hot spots-A review of the protein-protein interface determinant amino-acid residues, Prot. Stru. Func. Bioinform. 68 (2007) 803-812. [40] I. Nudelman, S.R. Akabayov, T. Scherf, J. Anglister, Observation of intermolecular interactions in large protein complexes by 2D-double difference nuclear Overhauser enhancement spectroscopy: application to the 44 kDa interferon-receptor complex, J. Am. Chem. Soc. 133 (2011) 14755-14764.

58

[41] C. Thomas, I. Moraga, D. Levin, P.O. Krutzik, Y. Podoplelova, A. Trejo, C. Lee, G. Yarden, S.E. Vleck, J.S. Glenn, G.P. Nolan, J. Piehler, G. Schreiber, K.C. Garcia, Structural linkage between ligand discrimination and receptor activation by type I interferons, Cell 146 (2011) 621-632. [42] A. Zvi, I. Kustanovich, Y. Hayek, S. Matsushita, J. Anglister, The principal neutralizing determinant of HIV-1 located in V3 of gp120 forms a 12-residue loop by internal hydrophobic interactions, FEBS Lett. 368 (1995) 267-270. [43] A. Zvi, D.J. Feigelson, Y. Hayek, J. Anglister, Conformation of the principal neutralizing determinant of human immunodeficiency virus type 1 in complex with an anti-gp120 virus neutralizing antibody studied by two-dimensional nuclear magnetic resonance difference spectroscopy, Biochemistry 36 (1997) 8619-8627. [44] A. Zvi, V. Tugarinov, G.A. Faiman, A. Horovitz, J. Anglister, A model of a gp120 V3 peptide in complex with an HIV-neutralizing antibody based on NMR and mutant cyclederived constraints, Eur. Jour. Biochem. 267 (2000) 767-779. [45] R. Levy, O. Assulin, T. Scherf, M. Levitt, J. Anglister, Probing antibody diversity by 2D NMR: comparison of amino acid sequences, predicted structures, and observed antibodyantigen interactions in complexes of two antipeptide antibodies, Biochemistry 28 (1989) 7168-7175. [46] M. Shoham, Crystal structure of an anticholera toxin peptide complex at 2.3 A, J. Mol. Biol. 232 (1993) 1169-1175. [47] M. Shoham, T. Scherf, J. Anglister, M. Levitt, E.A. Merritt, W.G. Hol, Structural diversity in a conserved cholera toxin epitope involved in ganglioside binding, Prot. Sci 4 (1995) 841-848.

59

[48] L. Martin, F. Stricher, D. Misse, F. Sironi, M. Pugniere, P. Barthe, R. Prado-Gotor, I. Freulon, X. Magne, C. Roumestand, A. Menez, P. Lusso, F. Veas, C. Vita, Rational design of a CD4 mimic that inhibits HIV-1 entry and exposes cryptic neutralization epitopes, Nat. Biotech. 21 (2003) 71-76. [49] R.W. Wien, J.D. Morrisett, H.M. McConnell, Spin-label-induced nuclear relaxation. Distances between bound saccharides, histidine-15, and tryptophan-123 on lysozyme in solution, Biochemistry 11 (1972) 3707-3716. [50] I.D. Campbell, R.A. Dwek, N.C. Price, G.K. Radda, Studies on the interaction of ligands with phosphorylase b using a spin-label probe, Eur. Jour. Biochem. 30 (1972) 339-347. [51] C.T.W. Moonen, R.M. Scheek, R. Boelens, F. Muller, The Use of Two-Dimensional NmrSpectroscopy and Two-Dimensional Difference Spectra in the Elucidation of the ActiveCenter of Megasphaera-Elsdenii Flavodoxin, Eur. Jour. Biochem. 141 (1984) 323-330. [52] E.A.M. Dejong, C.A.A. Claesen, C.J.M. Daemen, B.J.M. Harmsen, R.N.H. Konings, G.I. Tesser, C.W. Hilbers, Mapping of Ligand-Binding Sites on Macromolecules by Means of Spin-Labeled Ligands and 2d Difference Spectroscopy, J. Mag. Res. 80 (1988) 197-213. [53] T. Scherf, R. Hiller, J. Anglister, NMR observation of interactions in the combining site region of an antibody using a spin-labeled peptide antigen and NOESY difference spectroscopy, Faseb J. 9 (1995) 120-126. [54] M.K. Rosen, K.H. Gardner, R.C. Willis, W.E. Parris, T. Pawson, L.E. Kay, Selective methyl group protonation of perdeuterated proteins, J. Mol. Biol. 263 (1996) 627-636. [55] V. Tugarinov, P.M. Hwang, L.E. Kay, Nuclear magnetic resonance spectroscopy of highmolecular-weight proteins, Ann. Rev. Biochem. 73 (2004) 107-146.

60

[56] M. Kainosho, P. Guntert, SAIL - stereo-array isotope labeling, Quart. Rev. Biophy. 42 (2009) 247-300. [57] A. Horovitz, M. Rigbi, Protein-Protein Interactions - Additivity of the Free-Energies of Association of Amino-Acid Residues, J. Theo. Biol. 116 (1985) 149-159. [58] A. Horovitz, Double-mutant cycles: a powerful tool for analyzing protein structure and function, Fold Des 1 (1996) R121-126. [59] S.J. de Vries, M. van Dijk, A.M. Bonvin, The HADDOCK web server for data-driven biomolecular docking, Nat. Protoc. 5 (2010) 883-897. [60] C.D. Schwieters, J.J. Kuszewski, N. Tjandra, G.M. Clore, The Xplor-NIH NMR molecular structure determination package, J. Magn. Res. 160 (2003) 65-73.

61

Glossary CYANA: Combined Assignment and Dynamics Algorithm for NMR Application HADDOCK: High Ambiguity Driven Protein-protein Docking HMQC: Heteronuclear Multiple Quantum Correlation Spectroscopy HOHAHA: Homonuclear Hartmann-Hahn transfer spectroscopy HSQC: Heteronuclear Single Quantum Correlation Spectroscopy INEPT: Insensitive Nuclei Enhanced by Polarization Transfer NOE: Nuclear Overhauser Effect NOESY: NOE Spectroscopy PRE: Paramagnetic Relaxation Enhancement RMSD: Root-mean squared deviation TOCSY: Total Correlation Spectroscopy TRNOE: Transferred Nuclear Overhauser Spectroscopy WATERGATE: Water Suppression by Gradient Tailored Excitation

62

Graphical abstract

63

Highlights     

Methods revealing intermolecular contacts in large protein complexes are discussed. 13C-edited/13C-filtered NMR experiments to illuminate intermolecular NOEs. 2D NOESY difference spectroscopy to elucidate intermolecular interactions. Asymmetric deuteration to determine intermolecular interactions. Transferred NOE spectroscopy to determine intermolecular interactions.

64