Optimization of Protein Solubility and Stability for Protein Nuclear Magnetic Resonance

Stefan Bagby; Kit I. Tong; Mitsuhiko Ikura

Introduction

Nuclear magnetic resonance (NMR) spectroscopic techniques and hardware for the study of biomacromolecular structure and function have developed to the point where we can envisage obtaining high quality spectra of biomacromolecules and biomacromolecular assemblies of greater than 100 kDa molecular mass. 1 This will permit structure determinations of larger proteins that cannot be crystallized and allow studies of many intermolecular interactions in solution.2,3

For those NMR laboratories focusing on a particular target or type of target for structural analysis and to a lesser extent those pursuing a structural genomics approach,4 there remains, however, the prosaic but fundamental and often difficult problem of generating suitable samples for detailed NMR study: one of the major bottlenecks in the analysis of protein structure and function in solution by high resolution NMR methods is generating protein samples that are stable and soluble. NMR studies require the protein to be stable in the magnet for several weeks (unless the researchers have the time, energy, and funds to prepare numerous batches of sample) at high concentrations (ideally 1 mM or higher). This problem has been exacerbated by the move toward study of larger proteins by NMR with their greater tendency to aggregate.

Here we review methods that have been developed to optimize the polypeptide construct, facilitate initial screening of structural integrity, and assess aggregation state. We consider additives that may be used to improve protein stability and solubility at high concentrations without affecting the structure of the protein and protocols that have been developed to allow screening of a wide range of solution conditions for protein NMR studies using small amounts of protein.

Polypeptide Constructs: Defining Domain Boundaries and Segmental Isotope Labeling

Large proteins typically comprise several smaller domains, with most domains falling in the range of 100–250 amino acids. To date technical limitations have forced structural biologists using NMR to tackle such multidomain proteins by a “divide and conquer” strategy whereby single domains are studied in isolation. Because most full-length proteins are only marginally stable at physiological temperature, and selection of the start and end points for subcloning of domains has often been carried out in the absence of concrete information on the domain boundaries, it is not surprising that isolated fragments are frequently partially unfolded and/or prone to aggregation. Subcloning sites may typically be selected using secondary structure prediction and alignment of multiple sequences.

Structural information that permits identification of domain boundaries and therefore assists in selection of suitable sites for subcloning can alternatively be obtained from limited proteolysis, N-terminal sequencing, and matrix-assisted laser desorption/ionization (MALDI) or electrospray ionization (ESI) mass spectrometry.5,6 The principle of this method for designing constructs is that amino acid residues within a folded domain are protected from proteolysis whereas solvent-exposed, flexible amino acid residues are susceptible to rapid cleavage. The fragments generated by limited proteolysis are separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS–PAGE) or high-performance liquid chro-matography (HPLC) and characterized by N-terminal sequencing and MALDI or ESI mass spectrometry. This method has been used to define the domain boundaries of a number of proteins, including Max5 and an NAD+-dependent DNA ligase.6

An alternative to the divide and conquer approach for studies of multidomain proteins is to generate multidomain polypeptides in which only one of the domains is labeled with NMR-active isotope(s) such as 15N and/or 13C.7–10 Techniques for joining together protein segments,7–9,11 based on protein splicing, permit such domain-selective labeling and potentially allow structure determination by NMR of a single domain within the context of the full-length protein. Yamazaki and co-workers7 have developed a trans-splicing approach to segmental labeling of proteins for NMR studies which involves a denaturation step. A mild chemical ligation procedure for joining together folded recombinant domains which does not require a denaturation step has been demonstrated by chemical ligation of the SH3 and SH2 domains of the Abelson tyrosine kinase with the SH2 domain 15N-labeled.9 The peaks in the 1H–15N heteronuclear single quantum coherence (HSQC) spectrum of the SH2 domain in this chemically produced fusion coincided almost exactly with those of the recombinant SH3–SH2 construct in which both domains were 15N-labeled. Both trans-splicing and chemical ligation approaches can be extended to allow three recombinant protein segments to be regioselectively linked together: the feasibility of joining three segments by chemical ligation has been demonstrated in a model synthetic peptide system12; trans-splicing has been used for selective isotope labeling of a central segment of maltose binding protein8 and can be used to label selectively any segment between structurally flexible residues.

Polypeptide Folding

Once the polypeptide construct has been decided, the usual sequence of events would involve protein expression, protein purification, and then qualitative assessment of the structural integrity of the pure protein by recording a fingerprint spectrum such as 1H–15N HSQC; the backbone amide cross peaks in such a spectrum will cluster around 8 ppm if the protein is denatured. This lengthy and labor-intensive process is often fruitless, particularly if the construct boundaries are selected using alignment of multiple sequences or data from secondary structure prediction rather than the more rigorous limited proteolysis/mass spectrometry method discussed above. Structural integrity of the protein can instead be assessed rapidly by expression of the protein in 15N-labeled minimal medium, removal of the cell debris, and acquisition of a 1H–15N HSQC spectrum on the crude cell lysate. This was illustrated for two proteins, interleukin-1β and a double mutant of the B1 immunoglobulin (Ig) binding domain of streptococcal protein G, both of which comprised 15–25% of total expressed cellular protein.13 In these cases, 15NH4Cl was used as the sole nitrogen source throughout the growth of the cells. In cases where the protein of interest is expressed at levels corresponding to 5–10% of total cellular protein, 14NHaCl can be used as the nitrogen source until just prior to induction when the medium is changed to one that contains 15NH4Cl as the sole nitrogen source. If the peak dispersion observed in the 1H–15N HSQC spectrum indicates that the protein or protein fragment is folded, then it is obviously worth proceeding with further purification and spectral analysis.

This fast and simple method to assess the structural integrity of overexpressed proteins and domains may not be applicable to proteins that are very sensitive to solution conditions. It is also only applicable to proteins that are expressed in a soluble form, i.e., not packaged into inclusion bodies. These limitations have been tackled by the design of expression vectors specifically for the purpose of rapid screening by NMR. The vectors reported by Huth et al., 14 for example, encode the immunoglobulin-binding domain of streptococcal protein G (GB1 domain) fused to the N terminus of the relevant protein or protein fragment. The presence of the GB1 domain enhances expression and improves the chances of expression in a soluble form, and its small size (56 amino acid residues) means that NMR spectra can be acquired without separating the GB1 domain from the protein of interest. This last point represents a considerable advantage for rapid screening over expression systems that encode fusions with larger proteins such as glutathione transferase and maltose binding protein, where the fusion must be cleaved before structural integrity can readily be assessed.

Typically, 0.1–1.0 liter cultures are required for screening and the GB1 fusions offer the choice of recording a 1H–15N HSQC spectrum on the crude cell lysate or purifying the fusion protein using Ni2+ or IgG Sepharose affinity chro-matography prior to acquisition of the 1H–15N HSQC spectrum.14 Proteins greater than 30 kDa molecular weight may require use of 1H–15N transverse relaxation-optimized spectroscopy (TROSY)15,16 instead of a standard 1H–15N HSQC for screening of structural integrity since the TROSY technique provides superior spectral resolution and improved effective sensitivity for larger proteins.

As an alternative to NMR spectroscopy, circular dichroism (CD) spectroscopy can be used to assess the structural integrity of the polypeptide. Many secondary structure motifs in proteins, such as the α helix, β sheet, and β turn, give rise to characteristic CD spectra, and CD spectroscopy can be used to estimate the percentage secondary structure composition of polypeptides.17 The sample for CD spectroscopy must be free of contaminating proteins and other optically active impurities such as nucleotides and also free...