translate

[lorenzo]

translate

Phage display: practicalities and prospects

William G.T. Willats

Department of Biochemistry and Molecular Biology; University of Leeds, Woodhouse Lane, LS2 9JT, UK (Tel: +44 (0)113 3433168; Fax: +44 (0)113 3433144; E-mail: [email][email protected][/email])

Accepted 20 August 2002

Key words: antibody microarrays, immunomodulation, molecular evolution, Phage display, protein interactions, recombinant antibodies

Abstract

Phage display is a molecular technique by which foreign proteins are expressed at the surface of phage particles. Such phage thereby become vehicles for expression that not only carry within them the nucleotide sequence encoding expressed proteins, but also have the capacity to replicate. Using phage display vast numbers of variant nucleotide sequences may be converted into populations of variant peptides and proteins which may be screened for desired properties. It is now some seventeen years since the first demonstration of the feasibility of this technology and the intervening years have seen an explosion in its applications. This review discusses the major uses of phage display including its use for elucidating protein interactions, molecular evolution and for the production of recombinant antibodies.
Abbreviations: scFv ? single chain variable fragment of an antibody; dsFv ? disulphide stabilised scFv; Fab ? antigen binding antibody fragment; PCR ? polymerase chain reaction; RG II ? rhamnogalacturonan II; CDR ? complimentarity determining region; VH ? variable region of an antibody heavy chain; VL ? variable region of an antibody light chain; HG ? homogalacturonan; AceHG ? acetylated HG; ELISA ? enzyme linked immunosorbent assay; GFP ? green fluorescent protein.
Introduction or antibody fragments at the surface of phage par¬
ticles (Smith, 1985; Winter et al., 1994; Kay and Phage display technology has had a major impact on Hoess,1996). This is accomplished by the incorpora¬immunology, cell biology, drug discovery and phar-tion of the nucleotide sequence encoding the protein macology and is increasingly gaining importance in to be displayed into a phage or phagemid genome as plant science. The aim of this review is to provide a fusion to a gene encoding a phage coat protein. This a practical survey of the principles and applications fusion ensures that as phage particles are assembled, of phage display. The emphasis will be on the rela-the protein to be displayed is presented at the surface tive merits of this technology for addressing diverse of the mature phage, while the sequence encoding it is biological problems and the practicalities of what is in-contained within the same phage particle (Figure 1a volved, rather than a detailed exposition of molecular and 1b). This physical link between the phenotype technique or a comprehensive review of the literature. and genotype of the expressed protein and the replica-
Phage display is an extremely powerful tool for tive capacity of phage are the structural elements that selecting peptides or proteins with specific binding underpin all phage display technology (Figure 1b). properties from vast numbers of variants. Its utility Using phage display, libraries of variant nucleotide lies principally in generating molecular probes against sequences with diversities of millions or billions may specific targets and for the analysis and manipulation be converted into populations of displayed variant pro¬of protein/ligand interactions. Put at its most simple, teins which can then be conveniently screened for phage display is the expression of peptides, proteins

Figure 1. The principle of phage display. (a) The simplified hypothetical bacteriophage shown has a genome that contains an origin of replication (Or) and genes (g1 and g2) that encode two types of coat protein -p1 and p2, respectively. A foreign protein pχ, that is encoded by gene gχ, may be displayed at the phage surface by the fusion of gχ to one of the phage coat protein genes. The number of copies of pχ displayed is related to which phage coat protein (p1 or p2) is chosen as a fusion partner. (b) This principle can be applied to the expression of natural or random peptides, protein domains or whole proteins and antibody fragments. (c) Using phage display, a library of variant nucleotide sequences can be converted into a library of variant peptides or proteins. The phage display library may then be screened in order to isolate phage displaying peptides or proteins with desired properties.

desirable properties (Figure 1c). Screening of phage display libraries is usually accomplished by an affin¬ity selection (or bio-panning) process during which phage populations are exposed to targets in order to selectively capture binding phage (Hoogenboom, 1997; Hoogenboom et al., 1998; Sparks et al., 1996). Throughout successive rounds of binding, washing, elution and amplification, the originally very diverse phage population is increasingly enriched with phage with a propensity to bind to the target in question. Ul¬timately, monoclonal phage populations with desired specificities can be selected. This procedure of DNA manipulation to create a library of variants, packaging into phage, and subsequent bio-panning is the basic protocol for all phage display and has been coined the ?phage display cycle? (Hoogenboom et al., 1998) (Figure 2). Because the genotype of each protein phe¬notype is carried within phage particles, once proteins of interest have been isolated the sequence encoding them can be readily determined and altered in order to manipulate or refine binding properties.
Phage display WWW resources
A wealth of information and physical resources relat¬ing to phage display is now available via the World Wide Web. Table 1 is a list of the URLs for some of the major sources of phage display libraries and information.
Why use phage display?
As a system for the high throughput analysis of protein interactions phage display is complimentary to, rather than a substitute for, other methods such as yeast hy¬brid systems (Drees, 1999; Mendelsohn and Brent, 1999; Uetz, 2001) (see also pages X ? X of this issue) and each have their advantages and limitations. One advantage of phage display is the enormous diversity of variant proteins that can be represented. For exam¬ple, phage display antibody libraries with diversities as high as 1010 are routinely constructed (Hoogen¬boom et al., 1998) (see also Section 7). Phage display is highly flexible and selection may be performed in vivo or in vitro. (Johns et al., 2000; Sparks et al., 1996; McCafferty and Johnson, 1996). In vitro selec¬tion enables phage displayed proteins to be screened not only against a wide range of biological targets but also inorganic ones (Whaley et al., 2000). Yeast hybrid systems have the potential advantage that protein in¬teractions are assessed under physiological conditions ? but only a limited range of conditions are avail¬able, namely those within yeast cells (Drees, 1999). In contrast, phage display screening formats can be read¬ily modified to manipulate selection conditions and stringencies (Watters et al., 1997; Rodi et al., 2001; Hoogenboom et al., 1998). Both yeast hybrid systems and phage display provide a means of rapidly screen¬ing large numbers of proteins against potential binding partners but phage display has the higher throughput. As a rough guide, billions of clones can be screened within a week using phage display while with yeast two-hybrids millions of clones can be screened in two to four weeks (Rodi et al., 2001). An attractive aspect of phage display is that provided appropriate libraries can be obtained (Table 1), the technique is simple, cheap, rapid to set up and requires no special equipment.

Phage Display vehicles
The two key physical elements of phage display are the libraries of nucleotide sequences encoding peptides or proteins (e.g. gene fragments, random oligonucleotides, cDNA populations or antibody gene sequences) and the phage vehicles on which these se¬quences are expressed. The simplest way of achieving the expression of a foreign protein is to simply cre¬ate a fusion between the nucleotide sequence to be expressed and a coat protein gene within the viral genome (Figure 3a). Using this direct approach all the copies of the chosen coat protein become fusion pro¬teins (Winter et al., 1994). This can be advantageous in terms of numbers of expressed foreign proteins but if the functionality of the chosen coat protein is com¬promised by the fusion then phage viability may be affected, especially since no wild type versions of the coat protein are retained. This is avoided if hybrid phage are produced in which some versions of a given coat protein are wild type and some are fused to a for¬eign protein (Figure 3b and 3c). In some hybrid phage systems the gene fusion is an additional element of the phage genome so that a wild type copy of the coat protein gene is retained and phage particles express both wild type and fusion proteins (Figure 3b) (Sidhu, 2001). Alternatively, hybrid phage may be created us¬ing a phagemid-based system and this approach has been widely adopted (Figure 3c). Sequences encod¬ing fusion proteins are carried by phagemids (plasmids with a phage origin of replication) while the majority of the genes required for the formation of phage parti¬

Figure 2. The phage display cycle (a) A library of variant DNA sequences encoding peptides or proteins is created and (b) cloned into phage or phagemid genomes as fusions to a coat protein gene (see also Figure 3). (c) The phage library displaying variant peptides or proteins is exposed to target molecules and phage with appropriate specificity are captured. (d) Non binding phage are washed off ? although some non-specific binding may also occur. (e) Bound phage are eluted by conditions that disrupt the interaction between the displayed peptide or protein and the target. (f) Eluted phage are infected into host bacterial cells and thereby amplified. (g) This amplified phage population is in effect a secondary library that is greatly enriched in phage displaying peptides or proteins that bind to the target. If the bio-panning steps (c) to (f) are repeated the phage populated becomes less and less diverse as the population becomes more and more enriched in the limited number of variants with binding capacity. (h) After several (usually three to five) rounds of bio-panning monoclonal phage populations may be selected and analysed individually.

Table 1. Phage display resources on the WWW.

Source URL Comments
Smith Lab. University of Missouri www.biosci.missouri.edu/smithgp/Pha- A source of peptide phage display libraries
geDisplayWebsite/ PhageDisplayWebsite¬ and associated information
Index.html
MRC Centre for Protein Engineer¬ www.mrc-cpe.cam.ac.uk/∼phage/ The Winter group has previously provided
ing. The Winter Group home page. synthetic antibody libraries. Although not
distributed at present, copies of these li¬
braries may be obtained from existing users.
This site also contains useful phage display
information.
S. Dübel, at the University of Hei¬ www.mgen.uni- A comprehensive source of recombinant an¬
delberg, Molecular Genetics heidelberg.de/SD/SDscFvSite.html tibody resources
The Queen?s University of Belfast www.qub.ac.uk/bb/awpage/faq.htm Extensive information about obtaining phage
display libraries and associated reagents.
General information about phage display
protocols
New England BioLabs www.neb.com/neb/frame_cat.html Information about the PhDTM phage display
peptide libraries as well as useful general
phage display information
MRC Centre for Protein Engineer¬ www.mrc-cpe.cam.ac.uk/imt- A database of human antibody genes
ing. V Base (MRC) doc/public/INTRO.html
Philipps-Universität Marburg http://aximt1.imt.uni- Phage display and general filamentous phage
marburg.de/ rek/AEPphage.html information
University of Nijmegen http://baserv.uci.kun.nl/ jraats/links1.html Comprehensive phage display links

cles are carried by helper phage that are co-infected together with phagemids into host bacteria (Sidhu, 2001) (Figure 3c).
Hybrid phage systems have the potential disad¬vantage that the average number of displayed fusion proteins is reduced because of competition for incor¬poration into the phage particle between wild type and fusion coat proteins (Winter et al., 1994; McCafferty, 1996). However, low valency can be used as a strategy to select for high avidity binders during bio-panning. If coat protein functionality is not completely com¬promised by fusion to a foreign protein, then valency can be increased in phagemid systems by the use of modified helper phage (such as M13gIII) that lack the gene for the chosen coat protein (Winter et al., 1994; Rondot, 2001; Griffiths et al., 1993). More¬over, the choice of coat protein fusion partners has been extended recently by the development of new mutant variants of coat proteins and even completely artificial coat proteins (Sidhu, 2001). The number of expressed proteins therefore depends on the coat pro¬tein chosen as a fusion partner, the display system used (phage or phagemid) and, if a phagemid system is used, the choice of helper phage. A refinement of some phage display systems is the insertion of an am¬ber stop codon between the sequences encoding the coat protein and the displayed foreign protein. This allows a soluble (i.e. non-phage bound) version of the foreign protein to be produced if the phage are propa¬gated in an appropriate non-suppressing strain of host bacteria (Winter et al., 1994). Peptide tags such as c¬myc and poly-histidine are routinely incorporated into displayed proteins for ease of subsequent purification and detection.
Many types of phage have been used as vehi¬cles for phage display including Ff filamentous phage, Lambda and T7 (Rodi and Makowski, 1999; Danner and Belasco, 2001). Each of these has advantages and disadvantages with respect to each particular applica¬tion. The Ff phage family (M13 and its close relatives fd and fl) are excellent cloning vehicles because their size is not constrained by the DNA contained within them. The insertion of foreign sequences within their genome is accommodated simply by the assembly of longer phage particles. On the other hand, the non-lytic propagation mechanism of Ff phage requires that

Figure 3. Phage display formats (a)?(c) Strategies for the expression of proteins at the surface of a simplified hypothetical bacteriophage (see also Figure 1). (a) The simplest format for the expression of a peptide or protein is to fuse the gene (gχ) encoding the foreign protein (pχ) to one of the phage coat protein genes (e.g., g1) (see also Figure 1). This strategy produces phage particles in which all the copies of chosen phage coat protein are fusion proteins (p1/pχ). (b) Hybrid phage may be created by incorporating the gene fusion (g1/gχ) as an additional element in the phage genome. With this arrangement, two versions of the phage coat protein chosen as the fusion partner are encoded -one by the native gene (p1) and one by the fusion gene (p1/pχ). As phage particles are assembled both p1 and p1/pχ are incorporated into the phage coat. (c) Phagemid based systems are also widely used to construct hybrid phage. However, instead of being present on a single genome, the genes encoding wild type coat protein and fused protein are carried by helper phage and phagemid respectively. Host bacteria contain both phagemid and helper phage DNA and both genomes contribute to the synthesis of hybrid phage particles. (d) M13 bacteriophage are widely used as vehicles for phage display. The pIII coat protein can be used as a fusion partner for a limited number (maximum of five) of proteins while thousands of proteins can be expressed at the phage surface if pVIII is used as a fusion partner. The approximate number of copies of each M13 coat protein is indicated.
the all the components of the phage coat be exported proteins at high density (Zucconi et al., 2001). In some through the bacterial inner membrane prior to the cases it may even be advantageous to combine dif¬assembly of the mature phage particle. As a conse-ferent phage types in one experiment. This approach quence, only proteins that are capable of withstanding was used by Castillo et al. (2001) in order to select this export may be displayed (Danner and Belasco, anti-peptide single chain antibody fragments (scFvs). 2001). This limitation may be avoided by using the Whilst the peptide targets were displayed on T7, the lytic phage Lambda and T7, in which capsid assembly scFvs were selected from an M13 display library. occurs entirely in the cytoplasm prior to cell lysis. Fur-Despite some limitations, the Ff bacteriophage thermore, recent studies have shown that unlike T7, provide a robust and highly flexible platform for dis-Lambda phage can tolerate the display of relative large play and have been widely adopted. These long (about 1 µm) phage particles consist of single stranded DNA packaged into a coat consisting of five different types of coat protein, all of which have been used for the display of foreign proteins (Figure 3d) (Sidhu, 2001; Hoogenboom et al., 1998; Rodi and Makowski, 1999). Each of the coat proteins have their relative merits as fusion partners with respect to the number of fusion proteins displayed per phage, the effects of expressed fusion proteins on phage viability, and stability of the fusion proteins (Sidhu, 2001; Rodi and Makowski, 1999). In general terms, large numbers of smaller pro¬teins may be displayed if pVIII is chosen as a fusion partner, whilst pIII is a suitable partner for smaller numbers of larger proteins.

Finding the needle in the haystack: screening phage display libraries
Once a phage display library has been constructed or acquired the task is to screen the library in such a way that the original very high diversity of the library is reduced to a manageable number of clones which can then be analysed in detail. Most screening procedures are based on affinity selection and involve the follow¬ing fundamental steps: 1, A library is amplified and phage particles produced; 2, phage particles are ex¬posed to a target for which a binding protein is sought; 3, non-binding phage are removed by washing and 4, binding phage are eluted, infected into host bac¬teria and thereby amplified. These bio-panning rounds are then repeated, typically three to six times. In the following sub-sections some general considerations involved in the various steps of phage display library screening are discussed.
Library amplification
Although libraries with very high diversities are avail¬able, some expressed sequences are incompatible with phage propagation, whilst others are highly suscepti¬ble to proteolysis during propagation. These factors impose constraints on effective diversity and it is therefore desirable to start with a library that is as diverse as possible (Sparks et al., 1996). The possibil¬ity that some expressed sequences may be somewhat deleterious to phage propagation can be militated to some extent by including a growth step in each pan¬ning round that creates less competitive growth condi¬tions, for example by growing on solid media rather than exclusively in liquid culture. It is also important to empirically check the diversity of libraries before starting any screen because of the possibility that in¬stabilities in libraries can lead to loss of inserts. A quick check of diversity can be made by simply plating out a representative portion of the library, selecting a number of individual clones and then using PCR to check what proportion of clones contain inserts.
Bringing phage and targets together
One of the strengths of phage display is that screen¬ing protocols can readily be tailored to the particular requirements of many different target molecules. The simplest and most widely used approach is to immo¬bilise target molecules to a support and then to expose solutions containing phage to the immobilised target. Many variations to this theme have been successfully used including immobilisation onto coated tubes or plates, within columns or on the surface of magnetic beads. Immobilisation of many bio-molecules can be achieved by passive adsorption onto polystyrene tubes with an appropriate surface modification such as MaxiSorpTM (Nissim et al., 1994). Passive adsorption has the convenience that a wide range of molecules can be immobilised without any prior treatment. How¬ever, passive adsorption relies on establishing a large number of relatively weak bonds between target and support which can result in the immobilised mole¬cule being forced out of its functional configuration (Wilson and Nock, 2001). Clearly, this is undesirable if protein ligands are sought to functional versions of targets. A solution to this is to create one, or a limited number of tighter interactions between sup¬port and target ? for example by using biotinylating targets (Hoogenboom et al., 1998). More innovative screening methods have also been employed includ¬ing panning against whole fixed or living cells, tissue sections or even within living animals (Watters et al., 1997; Johns et al., 2000). Screens may also be de¬signed such that specific complexes can be selected for. One example of this is infectivity screening based on phage bearing truncated, non-infective fusion coat proteins. Infectivity is restored only if a complex is formed with a binding partner that has the capacity to restore infective functionality to the truncated coat protein.
Washing and elution
The basic purpose of washing is to remove non¬binding phage from the selection process so that binding phage are selectively enriched. However, this step is worth some consideration because a balance is required between specificity and avidity of selected clones. Most phage display libraries of whatever sort are likely to contain clones with a spectrum of avidities for any particular target. Some may be strong binders with low specificities, others the reverse. If washing is too stringent then highly specific, but weak binders may be lost. If washing is not stringent enough then populations of selected clones may be dominated by strong binders with low specificity. In practice this bal¬ance is achieved by adjusting washing times, detergent concentrations and using regimes in which washing stringencies are progressively increased. A number of treatments can be used to elute bound phage from tar¬gets. Dramatically lowering or increasing pH is often employed, or reducing agents may be used to disrupt disulphide-based links between supports and targets. A more subtle approach using enzymatic cleavage can be used where there are concerns about the effects on phage integrity of harsh elution conditions. Enzyme cleavage sites can be incorporated into the fusion pro¬tein, for example a trypsin cleavage site can be inserted between M13 pIII and the displayed fusion protein (Rondot et al., 2000).
Re-infection into host cells
It is usually assumed that following elution of bound phage, it is always essential to then amplify the re¬covered phage population before the next round of bio-panning, and indeed virtually all protocols include this step. However, this dogma may be worth careful examination since some reports indicate that directly using eluted phage without amplification may reduce background problems and help reduce the number of non-specific phage that are inevitably carried through the panning process. The rational is that during am¬plification, phage with inferior avidities for the target but better growth characteristics may be preferentially amplified. This has some important practical implica¬tions. The in vivo amplification steps are the most time consuming part of phage display library screening and if they could be avoided the time required for each screen would therefore be greatly reduced. Moreover, without the in vivo steps it is much easier to envisage how the whole screening process could eventually be completely automated (Hoogenboom et al., 1998).
What can be expressed at the surface of phage?
The first incarnation of phage display involved the en¬richment of just one expressed protein against a wild type phage population (Smith, 1985). In the inter¬vening seventeen years the scale and scope of phage display has vastly increased. Natural and synthetic peptides, proteins and protein domains and synthetic antibodies are now all routinely displayed on phage (Winter, 1998a; Winter et al., 1994; Kay and Hoess, 1996).

Phage display of peptides and proteins
The starting point of much peptide phage display work is the generation of random combinatorial libraries that provide a pool of variants from which peptides can be isolated by affinity selection. The peptides dis¬played in these libraries typically range in length from 5 to 20 amino acids and in some cases the conforma¬tional flexibility of displayed peptides is constrained by cyclisation. This is likely to afford some protection against proteolysis and may yield peptides with higher affinities. Cyclisation can be achieved by an amide bond between the N-alpha group and the side chain of the last residue or by a disulphide bridge between cysteine residues positioned at the N-and C-termini. A number of peptide libraries are freely available from the Laboratory of George Smith, University of Mis¬souri (Table 1). Random peptide libraries are a source of binding partners for a wide range of targets and in some cases the objective of phage display is to sim¬ply use isolated peptides directly as molecular probes or agonists. However, peptides may also be isolated with sequence homology to the natural protein bind¬ing partners of targets and such ?convergent evolution? studies are a powerful application of peptide phage display (Kay and Hoess, 1996).
Convergent evolution
The theory of convergent evolution of peptides is that by affinity selection, peptides can be isolated from a diverse starting pool that interact with a given target. Furthermore, the isolated peptides may have sequence homology to the natural binding partners of the target. Therefore, if the genome of the organism in ques¬tion has been sequenced to a significant extent, then the sequences from selected phage displayed peptides can be used to identify their natural counterparts by

Figure 4. Convergent evolution The natural protein binding partners of a given target may be identified by isolating phage displayed peptides that bind to the target and comparing them to a database of native sequences. (a) A population of variant nucleotide sequences are package into phage to generate a phage library displaying variant peptides (b). (c) Phage are screened against a target and binding peptides isolated (d). (e) A database for the organism in question is then searched for sequences with homology to the sequences encoding peptides carried by binding phage. Genes containing sequences with such homology may then be considered as candidates that encode the natural binding partner of the

target.
homology comparison (Kay et al., 2000) (Figure 4). One obvious danger is that such screens will isolate not peptides with homologous sequences, but ?mimo¬topes? -peptides that bind just as tightly to the target as the natural binding partner, but have no resem¬blance to it at the sequence level. However, various screening strategies have been developed to minimise this outcome (Rodi and Makowski, 1999) and conver¬gent evolution has proved to be a powerful strategy for unravelling protein interaction networks. Never¬theless, it perhaps seems surprising that short peptides can mimic so closely the interactive characteristics of often much larger and more complex natural counter¬parts. The explanation lies in the fact that for many proteins only a small subset of residues account for most of the change in free energy that mediates bind¬ing. For example human growth hormone consists of 217 resides but eight of these account for 85% of the binding energy (Rodi and Makowski, 1999).
Directed evolution
Once a population of peptide or protein ligands has been isolated, further layers of modification and selec¬tion can be applied in order to enhance or manipulate binding properties or affinities (Figure 5). The strategy of directing a population of peptides or proteins to¬wards specific properties by creating random sequence variation is known as directed evolution. In contrast to rational approaches for manipulating the properties of proteins, directed evolution has the advantage that pro¬teins can be manipulated without the need for a prior knowledge of molecular structure, or of the details of molecular action. Using directed evolution it has been possible to identify stronger binding ligands to recep¬tors, and to produce novel enzyme inhibitors and DNA binding proteins (Lowman and Wells, 1993; Dennis and Lazarus, 1994; Choo et al., 1994). The products of convergent evolution experiments can be a fruitful source of variants upon which further diversity can be imposed. Using the sequences encoding isolated peptides as a starting point, a second combinatorial library may be generated that is varied around selected sequences. The starting point for directed evolution can also be a protein of which the function is already known and characterised. A number of strategies are employed to introduce limited variation, including er¬ror prone PCR, the amplification of phage populations in mutator strains of host bacteria and DNA and fam¬ily shuffling. A recent example of this approach is the creation of variant forms of phytocystatin protease inhibitors (McPherson and Harrison, 2001). It has been demonstrated that protease inhibitors expressed in plants under the control of appropriate promoters can confer resistance to plant parasitic nematodes. Of the more than 60 phytocystatin sequences now known, eleven were chosen and subjected to family shuffling. The library of variants is now being screened with the intention of isolating phytocystatins with more potent inhibitory characteristics.

Figure 5. Directed evolution Phage display is a powerful tool for molecular evolution. A phage library displaying peptides or proteins is screened against a target and the binding properties of selected peptides or proteins are assessed by an appropriate assay. The nucleotide sequences encoding the selected peptides or proteins are then altered, for example by error prone PCR or DNA shuffling to create a new population of variant nucleotides. The phage display, bio-panning and analysis steps are then repeated in the hope of finding peptides or

proteins with altered or improved binding properties.
Directed evolution is also an important tool for the manipulation of enzyme characteristics and displayed variants of a given enzyme may be rapidly screened for altered properties. This approach is illustrated by the selection of lipase variants from an M13 phage library (Danielsen et al., 2001). Nine amino acids close to the active site of lipase from Thermomyces lanuginosa were targeted for randomisation by cassette mutage¬nesis and three rounds of selection were performed against a biotinylated inhibitor. Analysis of 84 active clones did not identify enzymes with greater activity than wild type but sequencing of the diversified region did provide insights into the mode of action of this enzyme.
An adaptation of phage display ? substrate phage, may also be used for the analysis of protease sub¬strate specificities. With this technique, the displayed moieties consist of peptides that are potential protease substrates. The peptides are sandwiched between a phage coat protein and a tag (such as c-myc) that serves to anchor the phage particle to a support. When exposed to a particular protease, only phage displaying a cleavable peptide are released into solution, whilst phage displaying non-cleavable peptides remain im¬mobilised to the support. The released phage may be retrieved and amplified and the sequencing of the inserts from recovered phage provides information about the substrate specificity of the protease used (Matthews 1996).

Phage display of antibodies
The worth of antibodies in plant research is well es¬tablished (Willats et al., 2002b). In addition to their uses for detection and isolation of cellular compo¬nents, antibodies have the unique capacity when used as immunocytochemical probes to provide contextual information at the sub-cellular level about defined epi¬tope structures (Willats et al., 1999; Willats et al. 2000). Although the number of antibodies directed against plant epitopes has grown steadily over recent decades they still cover only a minute fraction of the molecular structures involved in plant growth and de¬velopment and this shortfall is reflected by gaps in our understanding. Antibody phage display not only greatly extends our capacity to generate antibodies but also extends their potential applications for the direct functional analysis of epitopes. A further major advan¬tage of antibody production by phage display is that in many cases the whole process can be performed in vitro, thereby negating the requirement for target anti¬gens to be immunogenic. The range of feasible target antigens is therefore extended considerably because a major limitation for hybridoma antibody production is the lack of immunogenicity of potential targets. This is particularly true of glycan epitopes and is a factor that has seriously hampered antibody produc¬tion against carbohydrate plant cell wall components (Willats et al., 2000). The amount of target antigen re¬quired for antibody phage display is much less than is typically required for hybridoma antibody production (micrograms compared to milligrams) and the time re¬quired to generate monoclonal antibodies is also much reduced (a few weeks compared to several months). Because immunisation is by-passed (if single-pot li¬braries are used) the ethical and financial burdens of animal use are also avoided and phage display anti¬body production is relatively simple and cheap and requires no special facilities.
The principle of antibody phage display
Both conventional hybridoma and phage display an¬tibody production exploit the vast diversity of the mammalian antibody repertoire. The fundamental dif¬ference is that with hybridoma antibody production this diversity is harnessed by the immortilisation of antibody producing B-cells, while with phage display it is the genes that encode antibody variable regions (V-genes) that that are immortilised. (Winter et al., 1994; Marks et al., 1991; Clackson et al., 1991; Hoogenboom et al., 1998).
The principles of antibody phage display are iden¬tical to those discussed above in relation to the display of peptides and proteins. However, with antibody phage display the sequences encoding the displayed proteins are derived from genes encoding the key el¬ements of natural antibodies that determine binding. The procedures of affinity selection and screening for desired specificity are essentially the same as those de¬scribed for peptide libraries and to some extent mimic the processes of clonal selection and expansion in the mammalian immune system that underpin natural anti¬body production. Using directed evolution the binding properties of phage antibodies can be further biased towards a given target -a process analogous to affinity maturation in mammals.
Antibody phage display libraries
The mammalian V-genes that encode antibody vari¬able domains provide the raw materials for phage antibody library construction. Libraries essentially fall into two categories depending on whether these genes are derived from non-immunised animals or animals immunised with the target antigen ? single-pot and post immunisation libraries respectively (Figure 6a? c).
Post-immunisation libraries. In the construction of post-immunisation libraries, IgG sequences are gen¬erally derived from the spleen B-cells of immunised animals (Figure 6a). The repertoires of isolated V-genes are manipulated and packaged into phage li¬brary vectors. The rational is that some selection and affinity maturation of sequences with specificity for the antigen will have already occurred in vivo.Post¬immunisation libraries may therefore be pre-biased towards containing antibody fragments with desirable specificities and affinities (Hoogenboom et al., 1998). This approach has been used to create a number of valuable antibody probes against plant cell compo¬nents. Williams et al. (1996) generated a Fab fragment with specificity for the rhamnogalacturonan II (RG II) domain of the pectic cell wall matrix (Williams et al., 1996) while Shinohara et al. (2000) isolated a scFv from an post-immunisation library with specificity for the hemicellulosic fraction of Zinnia cell walls. Although high affinity antibodies can be produced using post-immunisation phage display libraries this approach has several drawbacks. Most serious is the necessity to construct a new library for every antigen so that the logistical, financial and ethical burdens of animal use associated with hybridoma antibody pro¬duction are not avoided. Moreover, because of the in vivo stage this approach requires target antigens to be immunogenic.
Single-pot libraries. Because of the limitations out¬lined above, a major goal over the past decade has been to create highly diverse, universal, antigen-unbiased libraries from which antibody fragments with specificities for a wide range of targets can be isolated. Such single-pot libraries completely avoid immunisation, library construction for all but the first user, and any immunogenic requirement. For these reasons single-pot libraries have been widely adopted and used to generate highly specific antibodies to a wide range of targets.
Two types of single-pot library have been devel¬oped ? naïve and synthetic (Figure 6b and c). In un¬immunised animals the primary, unselected antibody repertoire is dominated by IgMs with a specificity for a variety for antigens. For naïve library construction, v-gene sequences that have undergone some in vivo

Figure 6. An overview of phage display antibody library construc¬tion (a) Post-immunisation libraries are constructed using antibody gene sequences derived from animals that have been immunised with the target of interest. This approach capitalises on in vivo an¬tibody production processes, such as affinity maturation and may produce high affinity antibodies. However, a significant disadvan¬tage is that a new library must be constructed for each antigen. In contrast, single-pot libraries (b) and (c) may be used as a universal resource for the selection of antibodies against a wide range of targets. Naïve libraries (b) are constructed using V-genes sequences that have undergone some natural rearrangement, for ex¬ample sequences derived from IgM mRNA. Synthetic libraries (c) are constructed ?from scratch? using un-arranged germline V-gene sequences. (d) Antibody gene sequences are arranged and pack¬aged to produce expressed antibody fragments in various formats including single chain antibody fragments (scFvs), Fab fragments and disulphide stabilised scFvs (dsFvs).
rearrangement are derived from the IgM mRNA of an un-immunised animal (Figure 6b). This need not be an invasive process since mRNA can be sourced from peripheral blood lymphocytes (Marks et al., 1991).
Synthetic libraries are ?built? in vitro from un¬rearranged antibody gene segments with some crit¬ically positioned additional random sequences (Fig¬ure 6c). The design of synthetic libraries is based on knowledge of the key CDR (complementarity de¬termining region) sequences that shape the antigen combining site and are therefore critical for bind¬ing (Winter, 1998b; Hoogenboom and Winter, 1992). Broadly speaking the success of naïve libraries relies on their sheer size while with synthetic libraries the contents and overall diversity can be designed and controlled. Indeed, synthetic antibody libraries are be¬ing constructed that are tailor made for given epitopes (Kirkham, et al., 1999; Winter, 1998b).
Two M13-based libraries produced by the Win¬ter group at the MRC Centre for Protein Engineering (U.K.) have been widely used and have yielded high affinity scFvs to diverse targets. Both the Synthetic scFv Library (#1) and Human Synthetic VH+VL scFv Library libraries have been made available to the scientific community (Table 1). Using the Syn¬thetic scFv Library (#1) we have isolated scFvs with specificity for both protein and carbohydrate targets. Antibody PAM1 binds specifically to un-esterified ho¬mogalacturonan (HG, a component of the cell wall pectic matrix) (Willats et al., 1999) while PAM5 binds specifically to the tobacco GATA transcription fac¬tor TGAF (unpublished results). Using the Human Synthetic VH+VL scFv Library we have generated a panel of antibodies with specificity for acetylated HG. In this case phage display was successfully used after a hybridoma-based approach failed to yield anti-acetylated HG antibodies.
Antibody formats
Displayed antibody fragments can be configured in a variety of formats (Figure 6d). In the simplest arrangement, scFvs consist simply of a linear chain of natural-antibody derived heavy (VH) and light chain (VL) domains joined by additional flexible linker se¬quence. Fragments can also be designed with an engi¬neered intermolecular disulphide bond that stabilises the VH-VL pair (dsFvs). The display of antibody Fab fragments can be achieved by fusing one chain to the C-terminus of pIII and expressing the other chain un¬fused and secreted into the periplasmic space of host cells where the two chains then associate (Hoogen¬boom et al., 1998). A similar approach can be used to form bivalent and bispecific antibody fragments. Be¬cause of the small size of the inserts scFv libraries tend to be more genetically stable then Fab libraries. On the other hand scFvs are prone to dimer-and trimerisa¬tion which can hamper selection and characterisation of specificity.
Finally, the tantalising prospect of phage particles that are not based on, but mimic antibody binding has been raised by a strategy known as landscape phage display. This approach does not rely on anti¬body derived sequences at all but involves the display of thousands of copies of a peptide that cover as much as 50% of the phage surface (for example by fusion to the pVIII coat protein of filamentous fd-tet) (Petrenko and Smith, 2000). The spatial limitations imposed by the packed phage surface serve to constrain the dis¬played peptides into a defined organic surface that collectively has the capacity to bind to targets with high affinity and specificity. Since each ?landscape? varies according to the displayed peptide, a high di¬versity of specificities may potentially be generated.

Using phage display antibodies
In broad terms, phage display and hybridoma-derived natural antibodies may be used in the same range of applications, for example ELISAs, western blots and immunocytochemistry. However, phage antibod¬ies have some particular limitations and advantages. One limitation of hybridoma antibodies is that binding is usually most effective at mammalian physiologi¬cal conditions which can be a disadvantage for some plant research applications. For example, a poten¬tially powerful application of antibodies is to directly disrupt target antigens in vivo However, for many hybridoma antibodies the conditions required for bind¬ing, or even solubility, are not compatible with plant growth. With phage display it is possible to regulate screening conditions such that antibodies with binding capacity under defined conditions are isolated.
Discussed below are two aspects of phage antibody use in relation to plant science, their use as molecu¬lar probes, and for the in vivo immunomodulation of targets.

Figure 7. The use of phage display antibodies ? whole phage verses antibody fragments Both whole phage particles (a) and isolated anti¬body framents (b) may be used as immunological probes in a similar range of applications as hybridoma antibodies. The use of whole phage particles has the disadvantage that their large size leads to poor resolution. (c) In this example of immunogold labelling of tomato pericarp cell walls, the homogalacturonan epitope recog¬nised by the M13-based antibody PAM1phage is restricted to the middle lamella between adjacent cells. Arrowheads indicate the approximate extent of the middle lamella. However, PAM1phage particles are in the order of 1 µm long and covered in the pVIII coat protein that is the epitope recognised by the gold-conjugated secondary antibody. The resulting labelling therefore extends far beyond the position of the epitope since the whole length of phage particles is visualised (indicated by arrows). (d) Similarly, when PAM1phage are used for immunofluorescent microscopy their large size results in poor resolution ?fuzzy? images, as shown by this im¬age of PAM1phage binding to cells of tobacco stem parenchyma.
(e) In contrast, if the relatively small (∼ 30 KD) free PAM1scFvs are used as probes resolution is greatly increased, as shown by this image of PAM1scFv binding to cells of tobacco stem parenchyma similar to those shown in (d). PAM1scFv was detected via an N-terminal poly-histidine tag. Thanks to Dr Carolina Orfila (KVL, København) for the image shown in (c).

Figure 8. Antibody and antigen microarrays Microarrayed antibodies can be used to detect antigens, while microarrayed antigens can be used to detect antibodies. (a) Antibody microarrays are analogous to DNA microarrays in that they can be used to determine, in parallel, the relative abundance in complex mixtures of molecules isolated from two or more sources, for example mutant and wild type tissues. Differentially dye labelled extracted molecules are captured by microarrayed natural antibodies (i), antibody fragments (ii) or a mixture of antibody types (iii). The contribution of each dye to the total signal collected from a given spot is a measure of the relative abundance in the samples tested of the molecule recognised by the antibody immobilised on that spot. (b) Antigen microarrays arrays may be used to assess the binding of antibodies to immobilised antigens, for example to determine antibody specificity. Depending of the microarray platform used a wide range of antigens may be immobilised including proteins (i) carbohydrates (ii), or glycoproteins (iii). Binding to microarrayed antigens may be detected using fluorophore conjugated secondary antibodies (iv and v) or by GFP tagging or dye labelling of antibodies (vi and vii). (c) We have used antigen microarrays to assess the binding of phage display antibodies with specificity for acetylated homogalacturonan (AceHG) domains of cell wall pectic polysaccharides. Ten replicates (I-X) of ten different (1?10) AceHG samples were microarrayed onto polystyrene MaxiSorpTM treated slides (Nunc, Denmark). The differential binding of three (i?iii) different phage display monoclonal antibodies is shown. Antibody binding was detected using anti-M13pVIII /FITC secondary antibodies.

Phage antibodies as immunocytochemical probes
Phage antibody binding can be detected by the use of secondary antibodies with specificity for phage coat proteins. For example PAM1phage (Figure 7a) can be detected using secondary antibodies with specificity for the M13 pVIII coat protein. The use of anti¬pVIII secondary antibodies effectively amplifies scFv binding because of the approximately 2,700 copies of pVIII that coat the M13 phage particles (see Fig¬ure 3d). However, the large size of M13 is a disad¬vantage for immunocytochemical localisation studies because of the diffuse signal resulting from secondary antibody binding to the multiple copies of pVIII dis¬tributed along the phage particle (Figure 7c and d). In order to overcome this it is necessary to use solu¬ble (non-phage bound) scFvs. These can be produced by amplifying PAM1 in a non-suppressing host al¬though we found that this approach was inefficient. Instead, we cloned the PAM1 coding sequence from the phagemid (pHEN1, Nissim et al., 1994) into a bacterial expression vector and at the same time added a poly-histidine tag to the scFv. The soluble form of PAM1 (known as PAM1scfv, Figure 7b) can be readily isolated to high purity using a nickel resin column. PAM1phage and PAM1scfv have identical specificities although the detection limit of PAM1scfv is less than that of PAM1phage because the poly-histidine tag pro¬vides more limited binding possibilities for each sec¬ondary antibody compared to the numerous pVIII coat proteins. However, when used for immunocytochem¬ical labelling the small size of PAM1scfv provides much superior resolution compared to PAM1phage (Figure 7e).
Recently, the immunocytochemical applications of phage display antibodies have been extended by the production of scFv/GFP fusions. Functional analysis has established that in many cases such fusions can be made in which both the scFv and GFP moieties retain their original activities (Morino et al., 2001; Casey et al., 2000). Apart from providing convenient probes for quick and simple one-step immunocytochemistry, the exciting possibility is also raised of being able to express scFv/GFP fusions in planta for real time intracellular labelling of given epitopes during devel¬opmental processes. Moreover, panels of co-expressed fluophor tagged scFvs could also be used for the in vivo analysis of molecular interaction using fluores¬cence resonance energy transfer (FRET) (Truong and Ikura, 2001; Gadella et al., 1999).
Immunomodulation
Several powerful approaches are available for the dis¬ruption of plant processes and molecules at the gene level. However, if the genetic pathways controlling a particular process or molecule are not characterised an alternative strategy is to directly disrupt gene prod¬ucts in order to elucidate their functions. One such direct approach is immunomodulation -the disruption of antigen function by the action of antibody binding (Smith and Glick, 2000). This can be achieved either by micro-injection of antibodies into cells, incorpora¬tion of antibodies into plant or plant cell growth media, or by the expression of antibodies in plants. This last approach has the most practical potential for func¬tional analysis of a wide range of intracellular antigens in vivo.
One problem associated with the expression of whole antibodies or Fab fragments in plants is that the intracellular environment is not conducive to cor¬rect antibody assembly. In this regard scFvs, with their relatively undemanding folding requirements, are particularly well suited for this role and have been suc¬cessfully used to immunomodulate a variety of plant antigens (De Jaeger et al., 2000). Immunomodula¬tion of the activity of the plant hormones abscisic acid (Strauß et al., 2001) and gibberellin (Shimada et al., 1999) and the receptor protein phytochrome (Owen et al., 1992) has been demonstrated. Antibod¬ies have been targeted to the cytosol, the endoplasmic reticulum, and apoplast but in theory any cellular compartment can be targeted. However, immunomod¬ulation using scFvs is not always straightforward and many expressed scFvs are unstable. One solution is to develop more stable scFv scaffolds, another is to build antibodies free of disulphide bonds (Hoogenboom et al., 1998). Another interesting approach exploits the unusual antibodies of the Camelidae (Hamers-Casterman et al., 1993). In addition to four chain antibodies, the Camelidae produce antibodies con¬taining only heavy chains (Hamers-Casterman et al., 1993). Simple single-domain fragments (VHH) de¬rived from these heavy chain antibodies may be a valuable resource for immunomodulation (De Jaeger et al., 2000).
Microarrays for characterising and using phage display antibodies
The development of DNA microarrays has been one of the most significant bio-technological advances in recent years and is set to revolutionise the high throughput analysis of gene expression (Lander, 1999; Debouck and Goodfellow, 1999). Microarray tech¬nology is increasingly applied to the analysis of pro¬tein interactions and the analysis of antibody binding (Tomlinson and Holt, 2001; Kodadek, 2001). Arrays of antibodies (antibody arrays) can be used to detect antigens, whilst arrays of antigens (antigen arrays) can be used to detect antibodies (Haab et al., 2001).
Antibody arrays (Figure 8a) can be used for pro¬teome profiling. The rational is to isolate ligands from complex mixtures on the basis of their binding to immobilised antibodies. A typical experiment in¬volves the isolation of mixtures of proteins from say, experimental and control cells or tissue. Each pro¬tein mixture is then bulk labelled with distinguishable markers (such as Cy3 and Cy5) and exposed to the immobilised antibodies. The contribution of each dye to the total signal collected from a given spot is a measure of the relative abundance of the molecule recognised by the antibody immobilised on that spot. This parallel analysis is analogous to typical DNA mi¬croarray experiments. Antigen arrays (Figure 8b) are essentially very high throughput versions of ELISA or dot-blot assays in which the binding capacity of a particular protein ligand or antibody is assessed by its binding to a series of microarrayed potential binding partners. Binding can be detected directly if fluorophore-coupled proteins or antibodies are used or by using fluorescently labelled secondary antibodies (Figure 8b).
In our experience the time limiting step in phage antibody production is the detailed analysis of each monoclonal phage population. Conventional assays, such as ELISAs and immuno-dot-assays have the dis¬advantages that only a relatively small number of sam¬ples can be tested simultaneously, and large amounts of antibody are required for each assay. Using pro¬tein antigen microarrays, many thousands of potential binding partners can be assessed simultaneously us¬ing a very small (< 100 µl) amount of phage solution amplified simultaneously in microtitre plates. Slide surfaces coated with polylysine or super-aldehydes are available for the microarray deposition of proteins but equivalent surfaces are not available for the prepara¬tion of carbohydrate microarrays. To address this, a novel polymer microarray slide has recently been de¬veloped. The slide surface has capacity to immobilise structurally and chemically diverse glycans without any derivatisation of the slide surface or the need to create reactive groups on the glycans prior to immo¬bilisation (Willats et al., 2002a). The slides are made of polystyrene and have a surface modification known as MaxiSorpTM (NUNC A/S, Denmark) ? a surface that has been widely used in a microtitre plate for¬mat for ELISAs for many years. Using these slides we have generated carbohydrate microarrays and used them to characterise phage display antibodies with specificities for plant cell wall pectic polysaccharides (Figure 8c).

Conclusions
Phage display is a multi-purpose tool for amongst other things, molecular evolution, analysis of pro¬tein/ligand interactions and the generation of antibod¬ies. However, the relative scarcity of plant-specific examples of some applications of phage display re¬flects the fact that this technology has much still to offer plant research. As the post-genomic era pro¬gresses the emphasis of research is likely to focus increasingly on making sense of the biological con¬texts of gene products. In this respect many of the applications of phage display outlined here will be valuable tools.
Acknowledgements
Thanks to Iain Manfield, Sue Marcus, Lesley McCart¬ney, Jürgen Denecke, Carolina Orfila, Jørn Dalgaard Mikkelsen, NUNC A/S and Eva Maria Klein.

[emmebar]

Ciao lorenzo, ma che succede? problemi con l'inglese?

Spesso anche io ma non credo che il posto giusto sia questo Forum .

Vabbè, l'altra l'ho passata ad un traduttore automatico, questa vedi di sistemarla da solo che è un po' tardi e ch'ho sonno.

Ah, dimenticavo non farlo più altrimenti ci passa il buonumore ()