Selectively infective phage (SIP) technology: scope and limitations

https://doi.org/10.1016/S0022-1759(99)00143-XGet rights and content

Abstract

We review here the selectively infective phage (SIP) technology, a powerful tool for the rapid selection of protein–ligand and peptide–ligand pairs with very high affinities. SIP is highly suitable for discriminating between molecules with subtle stability and folding differences. We discuss the preferred types of applications for this technology and some pitfalls inherent in the in vivo SIP method that have become apparent in its application with highly randomized libraries, as well as some precautions that should be taken in successfully applying this technology.

Introduction

The selectively infective phage (SIP) technology was developed for selecting interacting protein–ligand pairs (Dueñas and Borrebaeck, 1994; Gramatikoff et al., 1994; Krebber et al., 1995). It has also been called selection and amplification of phage (SAP) (Dueñas and Borrebaeck, 1994) or direct interaction rescue (DIRE) (Gramatikoff et al., 1994). While SIP is related to phage display, it has the advantage of directly coupling the productive protein–ligand interaction with phage infectivity and amplification, without the need of an elution step from a solid matrix (Fig. 1).

SIP exploits the modular structure of the gene-3-protein (g3p), which consists of three domains, N1, N2 and CT, which are connected by glycine-rich linkers and possess different functions for the phage life cycle (Fig. 1) (Armstrong et al., 1981; Stengele et al., 1990). The g3p is present most likely in five copies on the phage, reflecting the five-fold symmetry of the phage coat and the pilus (Marvin, 1998). The N-terminal N1 domain of g3p consists of 68 amino acids and is absolutely essential for Escherichia coli infection (Armstrong et al., 1981; Jakes et al., 1988; Stengele et al., 1990; Holliger and Riechmann, 1997; Krebber et al., 1997). The 132 amino acid sized N2 domain, which forms a complex with N1 on the phage (Lubkowski et al., 1998), specifically interacts with the E. coli F-pilus (Jakes et al., 1988; Stengele et al., 1990). This pilus interaction, however, is not absolutely required for infection, as an alternative, albeit less effective, direct infection pathway exists (Russel et al., 1988; Krebber et al., 1997), which will be described later. The CT domain consists of 149 amino acids (including the C-terminal transmembrane anchor), forms part of the phage coat and is absolutely essential for phage morphogenesis (Nelson et al., 1981; Crissman and Smith, 1984).

In SIP, the basic infectivity of the M13 filamentous phage is destroyed by deleting from the phage genome either the N1 domain or the N1 and N2 domains of the g3p. A peptide or protein library is fused N-terminally to some or all copies of the CT domain or the N2–CT domains of g3p, and no w.t. g3p must be present on the phage. The infectivity of the phage can now only be restored by adding the N1 or the N1–N2 complex, as the N1 domain is absolutely required for infection. These domains are themselves fused or chemically coupled to a ligand which binds to the peptide or protein displayed on the phage. These infectivity restoring molecules will be referred to as the “adaptors”, and the consequences of choosing different adaptors, consisting of either N1 or N1–N2, will be discussed later.

There are two routes to selectively restoring the infectivity of the phage: in vivo and in vitro SIP (Fig. 2). For in vitro SIP, both components — the phage displaying the protein and the N1 adaptor or N1–N2 adaptor with the ligand coupled to it — are separately purified and combined in defined amounts in vitro to yield infective phages, provided the ligand binds to the protein. Consequently, the adaptor is encoded on an expression plasmid and the ligand can be either genetically fused to it or, in case of a small organic molecule such as a hapten, chemically coupled to the purified N1–N2 (Gao et al., 1997; Krebber et al., 1997).

In contrast, in the in vivo SIP approach the ligand has to be a protein or peptide genetically fused to N1 or N1–N2, and this fusion protein is encoded on the phage genome. During in vivo phage production, the N1–ligand or N1–N2–ligand adaptor is exported to the bacterial periplasm, while the CT–peptide or CT–protein fusion is also transported to the periplasmic space but remains anchored to the inner membrane through the C-terminal transmembrane helix of CT, before it is incorporated into the budding phage. In case of a tight interaction in the periplasmic space between the polypeptides fused to the adaptor or to the CT domain, respectively, the infectivity of the phage is restored.

The major advantage of SIP in comparison to phage display is the strict coupling of the selection and the infection process, which occur simultaneously. Two further important advantages are apparent for the in vivo SIP approach. First, in identifying an interacting peptide or protein partner to a specific protein, this protein does not have to be first expressed and purified as in phage display. Instead, its DNA is all that is needed, and only very small quantities have to be functionally expressed in the selection system. Nevertheless, it obviously does have to be compatible with transport to and folding in the periplasmic compartment. Second, the in vivo SIP strategy would in principle also be suitable for “library-vs.-library” selections, which are not possible in a direct manner in traditional phage display. However, current limitations in the efficiency of selection, leading to only a limited effective library size, and some unresolved issues in adaptor exchange between phages (see below) have so far not lead to a practical realization of this strategy. On the other hand, progress has been made in developing methods how such “two-dimensional” libraries can in principle be constructed conveniently, as under some circumstances filamentous phages can pack two single-stranded vectors, which may each encode one of the potentially interacting proteins (Rudert et al., 1998).

Since its first proof-of-principle experiments with antibody Fab and scFv fragments as well as with coiled-coil peptides (Dueñas and Borrebaeck, 1994; Gramatikoff et al., 1994; Krebber et al., 1995), progress in understanding the underlying mechanisms has been made, and this knowledge has lead to the construction of improved in vitro and in vivo SIP phage vectors, which have been successfully applied to the selection from various synthetic scFv libraries.

New insight has been gained into the structural requirements of fusions to N1 and N2 through the solution of the N1–N2 structure by X-ray crystallography (Lubkowski et al., 1998). Both domains consist mainly of β-sheet and show a striking similarity in their core folds, which suggests an evolutionary origin by domain duplication. Between the N1 and N2 domains exists a large contact interface formed by two β-strands of N2 that participate in the N1 β-sheet. Nevertheless, there is some flexibility in the relative orientation of N1–N2 (Holliger et al., 1999), and N1 alone has the same structure as in the complex, as determined by NMR (Holliger and Riechmann, 1997). In the infection process, the N2 domain binds to the E. coli F-pilus, and while the pilus is “withdrawn”, the N1 domain is brought into contact with the C-terminal domain of TolA (Click and Webster, 1997; Riechmann and Holliger, 1997; Click and Webster, 1998; Deng et al., 1999). This interaction appears to be absolutely critical, as no infection is possible at all without either the N1 domain or in the absence of TolA, while the pilus and the N2 domain both merely improve infectivity, but are not indispensable. The crystal structure of the complex of N1 and TolA was solved recently (Lubkowski et al., 1999), and it clearly shows that TolA displaces the N2 domain, which had been proposed from biochemical experiments (Riechmann and Holliger, 1997), even though both bind with very different geometry. Thus, the flexible linkers connecting N1, N2 and CT are an integral part of the rearrangements necessary in the infection process. It is at present not clear what the further fate of the domains is in the infection process nor which further E. coli proteins may interact with them. It follows that there may be geometric restrictions in the protein–ligand pairs compatible with SIP, and the affinity threshold (see below) may also be related to the infection mechanism.

Section snippets

Model systems

A thorough study of infection properties of different g3p fusion modules has brought some further understanding of the infection process, especially of the in vitro SIP method (Krebber et al., 1997). In this study, β-lactamase was inserted at different positions within g3p, and also different fusions of a scFv fragment to the phage have been investigated in conjunction with different adaptor constructs. It could be shown that N1 is absolutely required for infection under all circumstances,

Troubleshooting SIP:pitfalls and countermeasures

While SIP has been shown to be able to select tight binders from libraries in a single round, as well as to be a very powerful technique for the enrichment of the best binder and folder from a library of similar molecules, we have discovered a few pitfalls, which the user needs to be aware of in order to take the appropriate countermeasures for making optimal use of the technology. The selection for tight binding is so powerful that covalent bonds between the adaptor and the phage are strongly

Conclusions

While the in vivo SIP technology is especially convenient, as no protein at all needs to be expressed and purified for the selection of binding partners, it is important to understand the potential side reactions which can result in false positives: spurious cysteines, leading to covalently disulfide-linked adaptor–phage complexes, and rare genetic recombinations which regenerate N1–N2–CT rearrangements. Recombination events can be efficiently eliminated by recloning of the correct-sized g3p

Acknowledgements

We thank René Hermann (ETH Zürich) for the electron microscopy experiments.

References (38)

Cited by (36)

  • Antibody production, design and use for biosensor-based applications

    2009, Seminars in Cell and Developmental Biology
  • Prokaryotic expression of antibodies and affibodies

    2004, Current Opinion in Biotechnology
  • The Potential Of Protein-Detecting MicroArrays For Clinical Diagnostics

    2004, Advances in Clinical Chemistry
    Citation Excerpt :

    Disadvantages of both phage and mRNA display are the requirement for numerous rounds of selection and amplification, as well as the need to express and purify target proteins. Selectively infective phage (J3) and bacterial (H4, J2) and yeast (Y2) two-hybrid methods can overcome these obstacles because they are one-step screening assays with in vivo expressed target proteins. For the selectively infective phage technique, the N-terminal domains of the pIII coat protein is replaced with peptides from a ligand library, resulting in noninfective phage particles.

  • Core-directed protein design

    2001, Current Opinion in Structural Biology
View all citing articles on Scopus
View full text