Many diseases are caused by mutations that reduce the stability or compromise the folding of key proteins. Examples include some of the pathologically relevant mutations of the tumor suppressor p53 (Bullock and Fersht, 2001
), the F508 deletion in CFTR which results in cystic fibrosis (Boucher, 2002), and various amyloidogenic mutations in PrP that are associated with familial spongiform encephalopathies or increased susceptibility to transmissible forms (Prusiner, 1998). Also, many proteins of potential therapeutic importance are insufficiently stable for reliable administration (Bishop et al., 2001). Both for such biomedical reasons and in the pursuit of a better fundamental understanding of protein stability, there have been many studies of the effects of mutations on protein stability. Nevertheless, we still have only a qualitative understanding of the effects of point mutation or more substantial core repacking (Richards, 1997). Certainly, there is as yet no computational model that can faithfully predict the thermodynamic effects of mutation on protein stability (Guerois et al., 2002).
Without doubt, part of the reason for our inability to predict the effects of mutation is that our notions of the underpinnings of protein stability come from a relatively small number of mutations to any particular protein architecture. (By ‘small number’, we mean relative to the size of sequence space, which is 20N for N sites, and thus is vast.) Also, the effects of mutation on protein stability are obscured by the fact that most protein architectures are complicated, making it hard to understand the results of site-directed mutations. One particularly fruitful method for gleaning insight into the basis of protein stability has been the systematic modification of model proteins, such as lysozyme (Matthews, 1995), 
repressor (Sauer et al., 1990), barnase (Fersht, 1993) and Rop (Munson et al., 1996). These proteins have typically been chosen for the ease of preparation and crystallization (e.g. lysozyme) or the ability to select for binding even if further biophysical characterization is more difficult or protein architecture is relatively complex (e.g. repressor). In general, the scope of systematic studies is limited by the fact that it is not easy to screen large numbers of protein variants directly for biophysical properties.
It is sometimes possible to screen libraries of mutants of a protein for function, using visible phenotypes (e.g. chromogenicity), or metabolic or antibiotic resistance selections, that link the function of the protein mutant to a cell-based property (fluorescence, survival, etc.). We believe that a general solution to the problem of ‘combinatorial biophysics’ can be achieved by combining the use of a well behaved, simple model protein of regular architecture with a functional assay. Structural characterization and systematic mutations of a model protein can be used to guide the construction of targeted libraries of protein variants, and a screen can then interrogate those libraries for active molecules. Since one can know from the outset that one is mutating structurally important residues (e.g. hydrophobic core residues), one can effectively screen for structure or stability. Such approaches have been applied in elegant combinatorial experiments on repressor (Lim and Sauer, 1989, 1991), and have also recently been exploited in other contexts. For example, chorismate mutase catalytic activity has been used as a reporter of structural mutations (Taylor et al., 2001), and phage-displayed IgG-binding domains of protein L (Gu et al., 1995) and protein G (Distefano et al., 2002) have been used to screen for structured variants of the parent proteins by binding to immobilized IgG.
Rop is an antiparallel, homodimeric four-helix bundle protein that has been studied extremely thoroughly: the crystal (Banner et al., 1987) and solution structures (Eberle et al., 1991) are known, and systematic mutations have been made to the RNA-binding (Predki et al., 1995), hydrophobic core (Munson et al., 1994a, 1996) and turn residues (Nagi and Regan, 1997; Nagi et al., 1999). Moreover, Rop has an activity: the modulation of the copy number of ColE1 plasmids in Escherichia coli by facilitation of the binding of the inhibitory RNA I to the priming RNA II of the ColE1 origin (Figure 1) (Tomizawa and Som, 1984). The simplicity of the architecture and the depth of characterization make it possible to construct rational, targeted libraries of Rop variants. However, no simple in vivo screen for Rop function has been reported. Castagnoli et al. (1994) previously fused Rop to the DNA-binding domain of the repressor, creating a complex but selectable system wherein sufficient Rop homodimerization resulted in immunity to infection. However, there is no way to ‘tune’ this selection, and selections make it difficult to analyze the ‘negatives’, because the cells containing these proteins are dead. Also, Cesareni et al. (1982, 1984) fused ß-galactosidase to the first 110 nt of RNA II, which results in diminution of ß-galactosidase activity in response to Rop activity. However, this is a negative screen for Rop function, and it is not entirely clear how this activity is related to Rop’s ability to reduce plasmid copy number. An electrophoretic mobility shift assay for the binding of Rop to ‘kiss complexes’ composed of isolated stem–loops from RNA I and RNA II has been developed (Eguchi and Tomizawa, 1990, 1991; Gregorian and Crothers, 1995). Although this assay allowed some details of the protein–RNA interaction to be elucidated (Predki et al., 1995; Munson et al., 1996; Lee and Crothers, 1998), it is a decidedly low-throughput approach and an in vitro simplification of the actual binding problem that Rop faces in vivo.
Here, we demonstrate a novel cell-based screen for Rop function that makes it possible to screen large numbers of Rop variants quickly and reliably. The screen relies on modulation of the expression of a reporter molecule [here, green fluorescent protein (GFP)] as a result of modulation of the copy number of the ColE1 plasmid from which GFP is expressed. By altering the expression scheme for GFP, both ‘positive’ and ‘negative’ screens are possible, wherein either Rop activity or inactivity can be reported by cellular fluorescence. By screening Rop variants that were previously subjected to gel-shift assay for activity in vitro, we show that binding to the small RNA stem–loops is not sufficient for activity in vivo. We also have engineered a Rop expression vector that makes it possible to readily clone libraries of Rop variants for screening. Currently, we are using this screen to interrogate libraries of Rop core variants in a rigorous, statistical manner, an approach that we believe will significantly expand the database of knowledge from which potential functions can be created for modeling protein stability (T.J.Magliery and L.Regan, manuscript in preparation).
Answers to all your Biology Questions
