Extending the size limit of protein nuclear magnetic resonance.

In the past 15 years, nuclear magnetic resonance (NMR) spectroscopy has emerged as one of the principle techniques of structural biology (1, 2). It not only is capable of solving protein structures to atomic resolution but also has the unique ability to accurately measure the dynamic properties of proteins and to probe the process of protein folding (3, 4). However, a major drawback of macromolecular NMR is its size limitation caused by two technical barriers. First, larger molecules have slower tumbling rates and shorter NMR signal relaxation times. This reduces the sensitivity of the complicated pulse sequences that often use long delays for the necessary coherence transfer steps. The increased molecular weight also introduces more complexity to a given spectrum, simply because there are more NMR-active nuclei and, therefore, more interactions among them. The current size limit of protein NMR is '35 kDa, but recent advances in both hardware and experimental design promise to allow the study of much larger proteins (2). The future is even brighter with the development of novel strategies for isotopic labeling of proteins that are synergistic with the new NMR techniques. One such strategy is the segmental protein isotopic labeling scheme described by Xu et al. in this issue of the Proceedings (5). Xu et al. have successfully ligated together two independently expressed, folded protein domains under mild conditions in vitro, making it possible to selectively label a given fragment or domain of modular proteins (5). They have done so by taking advantage of protein splicing (6, 7). A small number of proteins are made as precursors that contain motifs called inteins. During the maturation process, inteins are excised from these precursors, and the two resulting fragments or exteins then rejoin. This process has been called protein splicing presumably because it is conceptually similar to RNA splicing. Much has been learned about the mechanism of protein splicing (Fig. 1). As a first step, the N-terminal cysteine of the intein attacks the C-terminal amide bond of the N-extein, resulting in an N-S acyl shift and the formation of a thioester at the N-terminal splice site. In a transesterification reaction, the N-extein is ligated to the Nterminal cysteine of the C-extein. The sidechain amide of a conserved Asn residue at the C terminus of the intein then attacks the main chain amide bond to form a succinimide group, excising the intein. The ligated exteins undergo an S-N acyl rearrangement to form a native amide bond in the final step. Certain intein mutants are defective in the cleavage of the C-terminal splice site but still capable of cleavage at the N terminus, leading to accumulation of the thioester intermediate between the intein and N-extein (8). Recently, the properties of these mutant inteins have attracted much attention from protein chemists. For example, commercial protein expression vectors have been constructed to allow the production of protein of interest as an intein-chitin-binding domain fusion, which can be purified on a chitin affinity column (9). Because of mutations within the intein, the fusion protein is trapped as a thioester between the protein of interest and the intein. The intein-chitinbinding domain then can be cleaved off by the addition of reducing agents, such as DTT. Based on similar principles, Muir et al. have developed a technique termed ‘‘expressed protein ligation’’ (10, 11). Instead of reducing the thioester intermediate with DTT, they showed that the thioester generated by the splicing process can react with peptides containing a cysteine at their N-termini. Using this technique, they ligated synthetic peptides to recombinant proteins through native peptide bonds (10, 11). Xu et al. have now taken this approach one step further (5). They postulated that similar principles should work for the ligation of two folded proteins. To accomplish this, they made a simple yet elegant extension to their existing scheme (Fig. 1). In their earlier work with peptide ligation, they mixed the thioester form of the fusion protein on chitin beads directly with peptides. For the reactions to go to completion, they used high concentrations of peptides. Because it is difficult to obtain a similarly high molar concentration of proteins, the ligation reaction between two proteins would have been inefficient. Their solution to the problem was to incubate the beads with a small thiol compound, ethanethiol, which led to the liberation of an ethylthioester derivative into solution. The ethyl-thioester reacted efficiently with a second protein. Using this seemingly simple modification, they were able to join together the SH3 and SH2 domains of the Abelson tyrosine kinase with a remarkable 70% yield. Furthermore, the final products were characterized thoroughly by using analytical techniques such as mass spectrometry to confirm their chemical structures. They also obtained SH3SH2 proteins with only the SH2 portion labeled with 15N. Comparison of the 1Hy15N heteronuclear single quantum correlation (HSQC) spectrum of the SH3-15N-SH2 with that of uniformly 15N-labeled SH3-SH2 confirmed that the SH2 domain retained its tertiary fold after the ligation reaction. As the authors pointed out, this strategy permits three protein domains to be joined together in vitro. Evans et al. also developed a method to trap the thioester intermediate of a different intein with 2-mercaptoethanesulfonic acid (12), although they only attempted the ligation of the trapped thioester with synthetic peptides in their semisynthesis of RNase A and HpaI enzymes. Other strategies for generating regioselective isotopically labeled proteins in vitro include a trans-splicing scheme reported by Yamazaki et al., in which two protein fragments each containing part of the intein and the target protein were individually expressed (13). These two fragments then were mixed to allow for the excision of the intein in a trans-splicing process. A serious limitation of this method, however, is its requirement of a protein denaturation-refolding sequence. It is also not as versatile as that presented by Xu et al. because it cannot be extended to ligate three protein fragments. During its relatively short history, protein NMR already has undergone several transformations that have extended its size limit (Fig. 2) (1). These transformations were brought about by technical advances in NMR spectroscopy and by progress in protein labeling schemes. The assignment of resonance and identification of nuclear Overhauser effects (NOEs) initially were accomplished by analyzing two-dimensional homonuclear spectra, limiting the size of proteins suitable for NMR studies under 10 kDa because of spectral complexity. The subsequent availability of uniformly 15Ny13C-labeled proteins produced in bacteria led to the development of the so-called triple resonance