PEDS Advance Access originally published online on May 10, 2008
Protein Engineering Design and Selection 2008 21(7):413-424; doi:10.1093/protein/gzn016
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Using T7 phage display to select GFP-based binders
1Biosciences Division 2MPA-CINT, Los Alamos National Laboratory, Los Alamos, NM, USA 33M Drug Delivery Systems, St. Paul, MN, USA 4Unite' de Biochimie, Universite' Catholique de Louvain, Louvain-la-Neuve, Belgium
1 To whom correspondence should be addressed. E-mail: amb{at}lanl.gov
 |
Abstract |
|---|
 Top Abstract IntroductionResultsDiscussionMaterials and methodsFundingReferences |
|---|
Filamentous phage do not display cytoplasmic proteins very effectively. As T7 is a cytoplasmic phage, released by cell lysis, it has been prospected as being more efficient for the display of such proteins. Here we investigate this proposition, using a family of GFP-based cytoplasmic proteins that are poorly expressed by traditional phage display. Using two single-molecule detection techniques, fluorescence correlation spectroscopy and anti-bunching, we show that the number of displayed fluorescent proteins ranges from one to three. The GFP derivatives displayed on T7 contain binding loops able to recognize specific targets. By mixing these in a large background of non-binders, these derivatives were used to optimize selection conditions. Using the optimal selection conditions determined in these experiments, we then demonstrated the selection of specific binders from a library of GFP clones containing heavy chain CDR3 antibody binding loops derived from normal donors inserted at a single site. The selected GFP-based binders were successfully used to detect binding without the use of secondary reagents in flow cytometry, fluorescence-linked immunosorbant assays and immunoblotting. These results demonstrate that specific GFP-based affinity reagents, selected from T7-based libraries, can be used in applications in which only the intrinsic fluorescence is used for detection.
Keywords: affinity reagents/GFP/phage display/T7 phage
|
Introduction |
|---|
|
TopAbstractIntroductionResultsDiscussionMaterials and methodsFundingReferences |
|---|
The ability to display a correctly folded protein for biopanning constitutes a significant challenge in phage display, and the demonstration that a fusion protein is present on the phage surface (using western blotting, for example) does not guarantee correct folding. Filamentous phage are most commonly used for phage display, with many publications describing the display of peptides (Scott and Smith, 1990
; Greenwood et al., 1991; Markland et al., 1991) and antibody fragments such as scFvs (Marks et al., 1991; Vaughan et al., 1996; Sblattero and Bradbury, 2000) or Fabs (Hoogenboom et al., 1991; de Haard et al., 1999). In fact, the selection of antibodies is widely regarded as the most successful application of phage display, with one human antibody selected by phage display approved by the FDA, and many others in clinical trials (Reichert et al., 2005). The similarities between the secretion of proteins such as antibodies into the eukaryotic endoplasmic reticulum, and the translocation of phage proteins used for display across the inner bacterial membrane into the periplasm, both of which are governed by stringent requirements (Bendtsen et al., 2004), may account for the success of this technique with secreted or membrane proteins. However, although antibody fragments can be selected from filamentous phage libraries, their use in high throughput applications has been problematic due to their poor expression in bacteria and their relatively poor stability and storage properties. This has led to a growing number of alternative protein scaffolds proposed as affinity reagents (for a review see, Binz and Pluckthun, 2005). In general, these alternative scaffolds have attempted to recapitulate the binding of antibodies, with regions of diversity concentrated on one face of the protein. Many of these proteins are secreted or membrane-bound, and phage display using filamentous phage has been relatively successful (Hansson et al., 1999; Nuttall et al., 1999; Schlehuber et al., 2000). This is in contrast to the use of filamentous phage display for cytoplasmic proteins (Choo and Klug, 1994; Jamieson et al., 1994; Abedi et al., 1998; Malabarba et al., 2001; Binz et al., 2004), which has been less successful. For example, it has proved impossible to display affinity reagents based on ankyrins using filamentous phage display, unless an alternative [signal recognition peptide (SRP)] leader is used (Steiner et al., 2006). Furthermore, a direct comparison between the use of lambda and g3p or g8p of filamentous phage for the display of a hepatitis C cDNA library (Santini et al., 1998) clearly showed the superiority of the display afforded by the cytoplasmic phage for the display of the predominantly cytoplasmic proteins. T7, an extensively studied double-stranded DNA phage assembled in the cytoplasm and released by cell lysis (Stroud et al., 1981), is the cytoplasmic phage most widely used for display purposes, in large part because it is available commercially for this purpose. Compared with filamentous phage, fewer publications describe the use of T7 for display. However, the range of displayed proteins is broad (Bukanov et al., 2000; Hansen et al., 2001; Han et al., 2002; Houshmand and Bergqvist, 2003; Kurakin et al., 2004; Lehmann et al., 2004; Takakusagi et al., 2005; Tan et al., 2005; Nowak et al., 2006), and a comparison between T7 and g3p of filamentous phage for the display of peptides showed significantly less bias with T7 (Krumpe et al., 2006). Interestingly, although T7 has been used for many different proteins, its first description as a display vector was never formally published, and as a result, there has been little of the systematic study initially carried out for filamentous phage display.
All affinity reagents proposed to date have no function beyond binding, and subsequent detection always requires the use of tags and secondary binding reagents. An affinity reagent with intrinsic detection capability, such as fluorescence, would provide significant advantages, by being able to provide a real-time indication of binding. Green fluorescent protein (GFP) (Tsien, 1998), originally isolated from Aequorea Victoria, is an intrinsically fluorescent 27-kDa protein with high thermodynamic and chemical stability. GFP is not fluorescent unless correctly folded (Reid and Flynn, 1997), allowing a direct assessment of the effects of amino acid insertions or replacements upon protein function. This is not possible with alternative scaffolds in which function, and hence folding, cannot be assessed by direct visualization. GFPs containing linkers or random peptides have been created and used to identify peptides mediating localization to specific cellular organelles (Abedi et al., 1998; Peelle et al., 2001). However, most such insertions render the GFP either non- or weakly fluorescent. This is reflected in the fact that although GFP has been displayed on the surface of bacteria (Shi and Wen Su, 2001), fluorescent GFP-based affinity reagents have not been selected from such libraries. Superfolder GFP (sfGFP) is a form of GFP evolved for increased stability and folding that contains six mutations scattered throughout the sequence (Pedelacq et al., 2006). We were recently able to show that sfGFP could accommodate antibody binding loops at three individual loop-insertion sites, with minimal loss of fluorescence (Kiss et al., 2006), while there was significant loss of fluorescence at two other sites tested. We were also able to transform GFP into a fluorescent lysozyme-binding protein by inserting the third complementarity determining region (CDR3) from an anti-lysozyme VHH antibody (Desmyter et al., 1996) into one of the GFP loops, analogous to the work carried out with Neocarzinostatin (Nicaise et al., 2004). Yellow fluorescent protein (YFP) has also been transformed into an affinity reagent by inserting specific binding peptides selected within the context of OmpA, a bacterial outer membrane protein, into a YFP loop (Bessette et al., 2004), although we have been unable to replicate this with sfGFP (data not shown). However, in both of these cases, the binding elements were defined elsewhere and subsequently grafted into the fluorescent protein. In the work described here, in contrast, we have selected specific fluorescent affinity reagents directly from T7 display libraries in which random CDR3s have been inserted into single GFP loops, and show that such selected binders can be used to detect binding without the use of secondary reagents.
|
Results |
|---|
|
TopAbstractIntroductionResultsDiscussionMaterials and methodsFundingReferences |
|---|
GFP and GFP variants fused to gp10 are displayed on T7 phage
In a previous paper (Kiss et al., 2006), we were able to show that the insertion of the third complementarity determining region (CDR3) from a VHH domain binding lysozyme, between amino acids 23/24 of sfGFP (GFP-lys) was able to confer lysozyme binding to the modified sfGFP with an affinity of approximately 1.34 µM. For this study, we created an additional construct, termed GFP-myc, which contained the linear epitope recognized by the anti-tag monoclonal antibody, 9E10, also inserted between amino acids 23/24 of sfGFP. This protein was also fluorescent, and was specifically recognized by 9E10 (Fig. 1). sfGFP, GFP-lys and GFP-myc were cloned into the T7Select10-3b vector (Novagen), which purports to display 5–15 copies of the displayed protein on each T7 phage fused to the C-terminus of the T7 capsid protein, 10B. Display is carried out by growing the T7 phage in Escherichia coli strain BLT5403, which provides wild-type capsid protein 10 in trans to the modified 10B in the phage. In the absence of wild-type capsid protein 10, T7 phage is unable to assemble. Correct folding of the GFP-10B fusion proteins should lead to fluorescent phage if they are incorporated into the T7 capsid. Figure 2A–C show that fluorescent phage could be obtained for all constructs at 25°C, and for sfGFP and GFP-myc at 37°C, although at lower levels than those obtained at 25°C, reflecting the frequent observation that recombinant proteins are better expressed at lower temperatures. The GFP-myc and GFP-lys constructs contained the SV5 tag (Hanke et al., 1992) at their C termini, while the sfGFP construct was made without this tag, allowing T7 phage containing the former constructs, but not those containing sfGFP, to be visualized by western blot with SV5 (Fig. 2D and E). The observed molecular weight of the fusion protein is consistent with the calculated molecular weight of 66 kDa.
|
|
Although these experiments showed that recombinant T7 phage containing the GFP constructs could be produced, it did not show that these were displayed in a form which allowed recognition of their targets. This was demonstrated by flow cytometry, using specific and non-specific antigens coupled to microspheres. Streptavidin-coated microspheres were incubated with either biotinylated lysozme, 9E10 or no antigen, followed by T7 phage displaying GFP-myc or GFP-lys. Microspheres with the specific antigens were also used to control for non-specific binding for the non-cognate antigen. T7 phage at different titers was added to each of the microspheres and analyzed by flow cytometry using a FASCaliber. Figure 3 shows that at a titer of 1013 phage, both T7 GFP-lys and T7 GFP-myc phage bind their targets, with binding visible down to 1012 phage, although the signal-to-noise ratio with T7 GFP-myc was far greater, perhaps reflecting the 4–40-fold higher affinity for its antigen (Schiweck et al., 1997
; Hilpert et al., 2001), when compared with GFP-lys.
|
In order to determine the number of fluorescent proteins displayed per phage, all three T7 preparations were evaluated using two different single molecule detection methods: fluorescence correlation spectroscopy (FCS; Dai et al., 2007
and references therein) and photon antibunching (Basche et al., 1992). FCS measures and autocorrelates the fluorescence fluctuations that arise when individual fluorescent entities (for example, a dye, fluorescent protein or fluorescent phage) enter and exit an optical probe volume. The amplitude of the autocorrelation curve yields information regarding the average occupancy of the probe volume, whereas the temporal decay of the curve yields information regarding the time required for the molecule to diffuse into and out of the optical volume element. The product of the amplitude of the correlation function and the average count rate during the measurement yields the brightness per fluorescent entity, which can be used to ‘count’ the number of emitters present on a fluorescent entity if the brightness per emitter is known. FCS has been previously applied to baculovirus (Toivola et al., 2002), where it was used to show that it was possible to display approximately three GFP molecules per virus, and has also been applied to T7 phage (Slootweg et al., 2006), although not for the purposes described here. Figure 4A shows the autocorrelation curves (normalized by the concentration) of a dye used as a standard (Rhodamine 110), free GFP-myc protein and T7 phage displaying GFP-myc. The other protein-phage pairs were measured in an identical fashion and gave similar results (Table I). All three free proteins (GFP, GFP-myc and GFP-lys) had essentially the same size and brightness at the single-molecule level, with the hydrodynamic radii between 3 and 4 nm, consistent with dimers when compared with the Azami green series of defined fluorescent proteins (Dai et al., 2007). The sizes of the phages were between 30 and 45 nm, consistent with previous measurements obtained using other methods (Stroud et al., 1981; Ronto et al., 1983). The phage brightness differed from preparation to preparation, reflecting differences in the number of displayed proteins in different preparations, and/or the protein environment. By comparing the brightness and the diffusion time (autocorrelation curves) of phage displaying fluorescent proteins with the free fluorescent proteins, we were able to determine that each phage displays only one to three monomeric fluorescent protein equivalents. The exception was the fluorescence of the GFP-lys phage. This was significantly less than that of the other displayed fluorescent proteins, even though the fluorescence of the purified protein was similar.
|
|
In addition to FCS, we counted the number of fluorescent emitters on the phage by photon antibunching, which can be used to determine the stoichiometry of molecular complexes (Sykora et al., 2007
). Photon antibunching is a correlation between consecutively emitted photons from the molecule and can be used to count the number of independent fluorescent emitters on a fluorescent entity. In photon antibunching, the fluorescence emission is split 50/50 between two photon counting detectors, and the time that elapses between photons detected on each of the detectors is recorded. For a single fluorescent emitter, the probability of simultaneously detecting two photons on both detectors is zero (neglecting background photons), as after emitting a photon, it takes a finite period of time to re-excite the molecule for the next emission cycle. If the molecule under study contains more than one independent emitter, then photons can arrive simultaneously on both detectors. The probability of photon pair detection at zero delay or the antibunching dip, a, increases according to the relationship (a = 1–1/m), where m is the number of independent emitters. In the absence of a background, for a molecule containing 1, 2, 3 and 4 independent emitters, the parameter a will have a value of 0, 0.5, 0.67 and 0.75, respectively. In practice (as was the case here), values are elevated above these theoretical predictions, as the presence of a fluorescence background creates apparently correlated photons detected at zero delay. Figure 4B shows antibunching curves for the standard (Rhodamine 110) and the T7 phage carrying the greatest number of fluorescent proteins (GFP-myc). The antibunching dip (parameter a) for Rhodamine 110 molecules was 0.33 ± 0.12. This gives a value of 
1.5 for the number of independent emitters, neglecting background considerations. However, as noted earlier, the presence of a background elevates the level of the antibunching dip away from zero, and thus for Rhodamine 110 these measurements inform us there is only one independent emitter per molecule, as expected for this standard fluorescent dye. The T7-GFP-myc phage had an antibunching dip of 0.55 ± 0.22, corresponding to slightly more than two independent emitters, while for T7-GFP, the antibunching dip was 0.45 ± 0.22. These antibunching measurements provide estimates of the number of proteins displayed per phage which are in the same range (one to two per phage) as those obtained by the FCS measurements.
In order to determine the efficiency of selection, specific phage at titers ranging from 105 to 1012 were incubated with 0.1 µg biotinylated antigens, selected using streptavidin coated beads, washed and eluted by the addition of E. coli. In order to control for non-specific binding, biotinylated lysozyme was used as a negative control for GFP-myc, while biotinylated 9E10 was used to control for GFP-lys. The results in Fig. 5 show that for both targets, the number of phages eluted after binding to the specific antigen was always greater than the number of phages eluted with the non-specific antigen, particularly at low phage titers, where the difference was as much as 104. At higher phage input titers, the difference between specific and non-specific binding was less, particularly for GFP-lys, presumably reflecting non-specific binding. Interestingly, even at the best enrichment ratios, no more than one in one hundred specific phages was purified, and this was reduced to one in 104 at higher phage titers.
|
Having demonstrated that T7 phage bearing specific fluorescent binding proteins could be specifically enriched, we examined the effect of changing the concentration of the selection antigen. Although Fig. 5C and D show that each antigen has a different optimum selection concentration (GFP-lys optimum is 1 µg/ml, GFP-myc optimum is 0.1 µg/ml), in terms of the number of specific phage enriched, the differences between 0.1 µg/ml or 1 µg/ml were no more than 2-fold, and for this reason we chose 1 µg/ml for further experiments. This reduction in phage enrichment with increasing concentration of 9E10 is somewhat counterintuitive. 9E10 is approximately 10-fold larger than lyszoyme, and is biotinylated to correspondingly higher levels (16:1 for 9E10 versus 1.5:1 for lysozyme). The selection was carried out by interacting the phage with the biotinylated antigen in solution, and subsequently purifying bound phage with the magnetic streptavidin beads. The most likely explanation for the reduction in titers with increased amounts of 9E10, is that at the higher amounts not all the biotinylated 9E10 is able to bind to the streptavidin beads, with the result that although specific phage may bind to the 9E10, these phages are not subsequently purified by the magnetic streptavidin beads and remain in the supernatant. This could be either because 1 µg of 9E10 exceeds the capacity of the beads and/or the large size of the antibody causes steric hindrance to bead binding. A similar effect appears to occur with the lysozyme, but at the highest concentration used, which would be consistent with this interpretation.
Mock selections, mimicking selection from a true library, were carried out by mixing 103 T7 GFP-lys or GFP-myc phage with 1011 T7 GFP phage and adding either biotinylated lysozyme or biotinylated 9E10. For the GFP-lys phage, selection on lysozyme represented specific selection, while that on 9E10 represented non-specific selection, with the situation reversed for GFP-myc. The use of 103 specific phage at a dilution of 10–8 represents the expected frequency of specific phage in a real library at a sufficient number to allow selection to occur given the purification ratios found above. T7 phage bound to the biotinylated target antigens were harvested with streptavidin-coated magnetic beads. Selection was repeated for a total of four rounds. The GFP-lys and GFP-myc phages contained the SV5 tag, while the GFP phage lacked this tag. This allowed a straightforward screen for the enrichment of specific and non-specific phage. Aliquots of diluted phage plated on top agar containing BLT5403 resulted in plaques which could be transferred to nitrocellulose membranes and probed with the SV5 mAb. The number of SV5 positive plaques were compared with the total number of plaques, to give the percentage of positive phage after each round. As shown in Table II, for each of the specific selections, specific phage could be seen by the second round, and were abundant by the third and fourth rounds, while the non-specific selections remained consistently negative. The phage titer of the specific eluted phage selections also increased from 104 after the first round to 107pfu/ml after the fourth, reflecting a successful ongoing selection.
|
Selection from GFP libraries containing random HCDR3s
Having demonstrated that specific GFP-based binders could be isolated in mock selections, we constructed three T7 phage libraries in which random human CDR3s (HCDR3s) derived from pooled human peripheral blood lymphocytes were inserted into the three sites in GFP previously identified to be most permissive (23/24, G51/K52, 173/174) (Kiss et al., 2006). The libraries were created by ligating HCDR3 libraries into the pETCK3 GFP vectors containing SacB at each site, as previously described (Kiss et al., 2006). However, rather than transforming the ligation, it was reamplified directly, cleaved with EcoRI/HindIII and ligated into the T7 arms. After in vitro packaging, the number of independent plaque forming units for each library was approximately 2x108. Although this represents the maximum possible library diversity, the true diversity is likely to be less, due to limitations in the primary diversity, and the presence of duplicated HCDR3s (Huang et al., 1996; Kiss et al., 2006). The amplified T7 libraries were all fluorescent when grown at 25°C, and PCR using primers flanking the loop-insertion sites (data not shown) demonstrated that HCDR3 inserts were present at the appropriate sites.
The three libraries were mixed together at equal titers and selections carried out on biotinylated lysozyme, alcohol dehydrogenase (ADH), beta-galactosidase (B-G), ubiquitin, myoglobin, glutathione-S-transferase (GST), human transferrin (HTR) and glucose oxidase (GO). Total phage output titers increased with each successive round of panning. After four rounds, the selected T7 phages were amplified by PCR, cloned into pETCK3 (Kiss et al., 2006), a pET-based vector (Rosenberg et al., 1987) and transformed into E. coli BL21 (DE3) Gold. Auto-induction media (Studier, 2005) was used to induce the expression of 94 colonies from each selection in deep well plates. After induction, bacteria were lysed in POP culture (Novagen) and tested for binding by ELISA, using detection with SV5. A total of 38 clones for lysozyme, 16 clones for ADH, 5 clones for UBQ, 15 clones for myogloblin, 17 clones for GST, 7 clones for HTR and 6 clones for GO showed binding by ELISA (Table III). Protein was purified from those clones which gave specific binding signals and re-analyzed by flow cytometry, as previously described (Kiss et al., 2006). Positivity in this assay required both fluorescence and binding activity to reside in the same molecule. No positive clones could be confirmed by flow cytometry for myoglobin, ubiquitin, HTR and GO, while the number of positives confirmed by flow cytometry for the others is shown in Table III, and the results for some of the best clones are shown in Fig. 6. Positive clones were sequenced, and found to have the inserts shown in Table III. These are characteristic HCDR3 sequences, with the (cys)ser-ala-arg at the N-terminus and trp-gly at the C-terminus (the PCR method replaces the cys found in HCDR3s with ser). Binding clones had HCDR3 inserts at either 23/24 or 173/174. No positive clones were found with inserts at the 51/52 position.
|
|
The affinities of two selected anti-lysozyme clones, G7 and A2 were determined by flow cytometry, as previously described (Kiss et al., 2006). Figure 7 shows that the approximate affinity of these clones is 470 nM and 740 nM, respectively, somewhat better than the positive control, GFP-lys. Although the screening involved determining binding to a non-specific antigen, ADH, we further examined the specificity of G7 by flow cytometry and western blotting. Figure 8A shows that G7 does not recognize five other antigens by flow cytometry, while Figure 8B and C show that G7 can be used in western blotting to recognize lysozyme, down to 0.1 µg, and that there is no recognition of a number of non-specific antigens, including an overloaded E. coli extract.
|
|
F’ cell lysate. The arrow indicates the expected molecular weight of lysozyme. |
Discussion |
|---|
|
TopAbstractIntroductionResultsDiscussionMaterials and methodsFundingReferences |
|---|
Monoclonal antibodies (mAbs) are stable proteins, of high affinity and specificity, which can be used in many research and clinical applications, but the procedure to generate them is time consuming, labor intensive and requires mouse immunization. This precludes their use on the genomic scale, which requires the ability to generate specific affinity reagents against thousands of gene products in a high throughput fashion (Bradbury et al., 2003
; Taussig et al., 2007). Ten years ago it seemed as if mAbs would be replaced by single chain Fvs (scFvs) or Fabs selected from large naïve phage display libraries (Hoogenboom et al., 1991; Marks et al., 1991; Vaughan et al., 1996; de Haard et al., 1999; Sblattero and Bradbury, 2000), which appeared to offer the advantages of diversity, high affinity and specificity in a potentially high throughput format, without the use of animals and the problems of poor immunogenicity. While scFvs have been very successful in some cases, it has been found that their use beyond simple ELISAs is often limited by low production levels, relatively poor stability and the need for additional labeling steps. It is as a result of a failure to effectively and consistently use scFvs that we, in common with others (Binz and Pluckthun, 2005), have examined alternative scaffolds. All such protein scaffolds developed to date are similar to antibodies, in that they are ‘silent’; that is they have no function beyond binding, and detection of such binding always requires the use of secondary reagents. In an attempt to overcome this problem, we have attempted to develop specific affinity reagents based on the use of GFP as a scaffold, where fluorescence, intimately associated with binding activity, would be used to detect binding. With appropriate engineering, GFP fluorescence (or FRET) has been modulated by changes in voltage (Siegel and Isacoff, 1997), ß-lactamase inhibitory protein concentration (Doi and Yanagawa, 1999), pH (Miesenbock et al., 1998) or metal ions (Baird et al., 1999). Although GFP has been localized to specific cellular organelles by random peptides (Abedi et al., 1998; Peelle et al., 2001), beyond our previously cited publication (Kiss et al., 2006), there have been no reports of specific binding to a target of interest mediated by insertions, rather than N- or C-terminal fusions. Having demonstrated that some GFP loops, of a particularly stable form of GFP (Pedelacq et al., 2006), were permissive to the insertion of HCDR3s, with minimal effects on fluorescence (Kiss et al., 2006), we show in this publication that it is possible to select specific fluorescent GFP-based binders using T7 display.
Many papers have been published using T7 phage display (Bukanov et al., 2000; Hansen et al., 2001; Han et al., 2002; Houshmand and Bergqvist, 2003; Kurakin et al., 2004; Lehmann et al., 2004; Takakusagi et al., 2005; Tan et al., 2005; Nowak et al., 2006). However, it was originally introduced as a commercial kit, and the parameters for its use have never been formally described in a publication. In this paper, we first examine some of these parameters as applied to GFP and two variants containing specific binding inserts. These experiments show that GFP and these variants can be displayed on T7 in a form that allows specific selection. The display levels of these different proteins were determined by two different single-molecule methods: FCS and photon anti-bunching, both of which gave consistent results showing that only one to three proteins are displayed per phage. This is significantly less than the 5–15 expected (and advertised), and explains the poor selection efficiencies obtained (Figs 3 and 5). Had the display levels been higher, avidity effects would have expected to result in far stronger binding, as has been previously shown with multivalent filamentous phage display (Huie et al., 2001; Rondot et al., 2001; Baek et al., 2002; Soltes et al., 2003; Chasteen et al., 2006). As shown here, and previously for baculovirus (Toivola et al., 2002), single-molecule fluorescence techniques can be used to count the number of fluorescent proteins per phage and thus determine display levels. Moreover, these methods are able to distinguish between soluble and phage-bound fluorescent proteins, through a direct measurement of the hydrodynamic radius of the fluorescent species present in the sample. The single-molecule approach to quantifying display levels is thus more accurate than bulk methods, which are complicated by the presence of non-infectious phage and soluble protein. The only drawback of this method is that since they are non-fluorescent, non-displaying phages are invisible, as a result of which the display estimates of one to three fluorescent proteins per phage only apply to those phages that display, and not to the whole population of phages. Although we expect display levels with other proteins to be similar to those found here, it is also likely that different proteins will have different display levels, reflecting the levels of correctly folded 10B fusion protein.
While we have been able to display sfGFP using standard filamentous phage vectors, it has proved impossible to display GFP-lys or GFP-myc in fluorescent forms that are able to bind their targets. Standard filamentous phage display uses the Sec leader, responsible for translocating proteins across the inner membrane post-translationally in an unfolded form (Driessen et al., 1998; Fekkes and Driessen, 1999), suggesting that these modified forms of GFP must fold in the cytoplasm to be functional. There are a number of ways in which this can be carried out. One solution is to use a cytoplasmic phage, such as T7 or lambda, as described here. An alternative is to use part of a filamentous phage protein, such as the C-terminus of p6, which folds in the cytoplasm (Jespers et al., 1995; Hufton et al., 1999), or to uncouple the folding of the displayed protein from the display protein. This was recently carried out for a GFP circular permutant (cpGFP), using the TAT leader to translocate the cytoplasmically folded GFP into the periplasm, where it associated with g3p by a coiled-coil interaction (Paschke and Hohne, 2005), although in this case the cpGFP was not fluorescent. The ease with which we have been able to display GFP-lys and GFP-myc using T7 shows the importance of allowing cytoplasmic proteins to fold in the cytoplasm prior to display. This corroborates experiments carried out with hepatitis C (Santini et al., 1998), showing that the display of random hepatitis C cDNA fragments using lambda, another cytoplasmic phage, is far more efficient than the use of filamentous phage using Sec leaders. In the experiments reported here, we demonstrated that T7 can functionally display GFP-lys and GFP-myc. Although we were aware that these proteins bind their targets with relatively low affinities, we had expected the avidity effects of 5–15 displayed binders to overcome the low affinities. The discovery, using FCS and antibunching, that each T7 displayed only one to three fluorescent proteins explains why we were unable to purify more than 1% of purified phage preparation: at these low display levels, and given the geometry of the phage, binding will be monomeric, rather than multimeric. The low fluorescence of GFP-lys when displayed on phage, when compared with its fluorescence as a free protein, was somewhat surprising, and we assume due to quenching caused by the local environment, perhaps as a result of interaction between the binding loop and phage coat proteins. Whether this is specific to this particular molecule, or will occur with others is not clear at this time. However, the fact that the free protein is significantly more fluorescent than the phage suggests that this may not be a serious concern.
The nanomolar affinities of the binders selected against lysozyme are relatively high when considered within the context of peptide binders, where the usual affinities are in the micromolar range (Saggio and Laufer, 1993; Yayon et al., 1993; Sparks et al., 1994; Craig et al., 1998; Zwick et al., 1998). Although we were able to select specific fluorescent binders recognizing lysozyme, ADH, GST and β-galactosidase from a naïve library of single random HCDR3s, the inability to confirm fluorescent binders by flow cytometry against the other targets tested (myoglobin, ubqiutin, HTR and GO) suggests either that the diversity of the library is not sufficiently broad or that the use of single HCDR3s as a source of diversity is only appropriate for particular kinds of antigens. With regard to the diversity, the HCDR3s used here were derived from PCR of the total immunoglobulin gene pool of 40 donors. Sequencing of over 300 of these showed that 15% were duplicated (Kiss et al., 2006). By reducing the number of PCR cycles, and using the IgM, rather than the total immunoglobulin pools, the effective diversity would probably be increased. However, even with large random peptide libraries, where the source diversity is derived from degenerate oligonucleotides, the inability to select binders against all targets is a common finding, usually overcome by screening targets against multiple different libraries (Bonnycastle et al., 1996). This implies that for single-loop GFP libraries, the best way to increase the success rate is to screen against additional libraries made using CDRs derived from both different binding loops and different species, as well as degenerate random oligonucleotides of different lengths. However, even if such a broad array of single-loop libraries was available, the affinities would never approach those routinely obtained using protein scaffolds with diversity at multiple sites (Binz and Pluckthun, 2005). It was interesting that no specific binders were selected from libraries in which insertions were placed between positions 51/52, suggesting that if this site is to be used in the future, alternative forms of diversity should be explored.
Although the affinities of the selected binders were low when compared with other protein scaffolds, they are relatively high compared with those usually obtained from random peptide libraries, and are similar to experiments in which single binding loops are inserted within the context of thioredoxin (Colas et al., 1996; Geyer et al., 1999). This probably reflects the fact that HCDR3s are derived from antibodies, and hence enriched for peptides that routinely mediate binding activity. The experiments described here provide proof of principle for both the use of CDRs as sources of diversity as well as the utility of fluorescent proteins as potential affinity reagents. Even within the context of the experiments described here, the advantage of intrinsically fluorescent affinity reagents is clear. Quite apart from the use of fluorescence to detect binding (Figs 3, 7 and 8), the presence of intrinsic fluorescence has proved extremely useful in the determination of optimum display conditions (Fig. 2), the determination of display levels (Fig. 4) and affinity (Fig. 7). In Fig. 2, for example, the higher fluorescence (and hence display) of the phage grown at 25°C compared with those grown at 37°C, is easily determined visually without any further experimentation, whereas similar experiments with any silent scaffold would require western blotting at a minimum, without any indication of whether the scaffold was correctly folded or not. In order to use such fluorescent affinity reagents routinely, it is likely that the affinity/avidity would have to be increased. This could be carried out most easily by fusion to multimerizing domains, such as coiled coils, streptavidin or toxin molecules, as previously described for scFvs and derivatives (de Kruif and Logtenberg, 1996; Kipriyanov et al., 1996; Terskikh et al., 1997; Zhang et al., 2004). Once fully validated, we expect such fluorescence affinity reagents to be extremely useful in diagnostics, detection and proteomics.
|
Materials and methods |
|---|
|
TopAbstractIntroductionResultsDiscussionMaterials and methodsFundingReferences |
|---|
Reagents
All restriction enzymes and DNA modifying enzymes were purchased from New England BioLabs (Beverly, MA). The T7Select10-3b cloning kit, E. coli strain BLT5403 and the primers T7Select Up and T7Select Down were purchased from Novagen (Madison, WI). Other oligonucleotides were supplied by MWG. NHS-LC-LC-Biotin, Streptavidin-AP and Biotinlyted 9E10 were purchased from Pierce. Proteins were biotinylated using NHS-LC-LC-biotin according to the manufacturer’s instructions, with the efficiency of biotinylation determined by EZTM biotin quantification kit (Pierce). The biotin:protein ratios were about 1.5:1 for lysozyme and 16:1 for the 9E10 mAb.
Cloning GFP and GFP containing inserts into T7 phage
GFP-myc was created by annealing the two oligonucleotides myc-S (P- CC CGC GAA CAG AAA CTG ATT AGC GAA GAA GAT CTG AAC GAT TAT TGG GG) and myc-AS (P-CCA ATA ATC GTT CAG ATC TTC TTC GCT AAT CAG TTT CTG TTC GCG GGC A). This creates the myc epitope with overhanging ends able to ligate to the pET-CK3-sfGFP-SacB-loop1 vector (Kiss et al., 2006) digested with BpmI. This flanks the myc epitope with human heavy chain CDR3 sequences (ser/ala/arg at the N-terminus and asp/tyr/trp/gly and the C-terminus) and inserts the whole into sfGFP between amino acids 23 and 24. GFP-lys, GFP-myc and the previously created single-loop CDR3 libraries (Kiss et al., 2006) were amplified with psfGFPEcoRI (AAA TTT GAA TTC AGG CGC GCA TGC CGC ACT) and psfGFPHindIII (ACG TAT ATA AGC TTA TGG TGA TGG TGA TGG TGA GTA). sfGFP, which lacks the SV5 tag, was amplified with psfGFPEcoRI and psfGFPNOSV5HindIII (ACG TAT ATA AGC TTC TCA CTA ATC CAG GCC CAG CAG CAG TGG GTT G). All PCR products were digested with HindIII and EcoRI at 37°C for 8 h, gel purified and ligated into EcoRI/HindIII predigested T7Select10-3b vector arms (Novagen, Madison, WI). The ligation mixtures were packaged in vitro using the T7Select packaging extract. The ligation, phage packaging, amplification and titration were carried out according to the manufacturer’s instructions (Novagen, Madison, WI).
After selections, the T7 phage output was used as a template for PCR using T7 select up and down primers, and cloned into pETCK3 (Kiss et al., 2006), a pET derivative using, BssHII and NheI.
For all selections biotinylated antigens were blocked with 5% BSA in PBS (phosphate-buffered saline), and incubated with 1011 T7 phage in 100 µl for 1 h at room temperature. For model selections, T7-GFP-lys and T7-GFP-myc were each mixed with T7-GFP at 1:108 ratio (1000 in 1011). To select antigen-specific binders, a library of T7 phage displaying single HCDR3 loops at three different sites (N23/G24, D102/D103 and D173/G174) of sfGFP were used. A Kingfisher magnetic bead handling robot (Thermo Electron, Vantaa, Finland) was used for the selection with 10 µl of streptavidin-coated magnetic beads (Dynabeads, Dynal). Unbound phage particles were removed by washing the beads three times with PBST (containing 0.1% Tween 20) and three times with PBSLT (containing 0.01% Tween 20). The bound phage particles were harvested by adding the washed beads to 5 ml log-phase E. coli BLT5403. The culture was incubated at 37°C for 30 min then at 25°C for 8 h or until cell lysis occurred. Phages were purified by PEG precipitation and titered. After amplification, 1011 output phages were used for the next round of selection.
In the mock selections, to investigate the efficiency of selection, output phages after each round were diluted and mixed with BLT5403 bacteria and molten agarose and plated on LB/Carb agar plates. Those dilutions giving approximately 400 plaques were used. A nitrocellulose membrane was placed onto the plate and incubated for 5 min at room temperature to lift the phages. The membrane was dried in air for 20 min then blocked with 5% BSA/PBS, and incubated with anti-SV5 antibody (Hanke et al., 1992) and (1:1000 dilution) anti-Mouse Alkaline phosphatase conjugate (DAKO, Glostrup, Denmark) was used. PBST and PBS wash was performed between each step. Positive plaques were revealed with 1-step BCIP (Pierce) and positive plaques counted and expressed as a percentage of total plaques.
Determination of affinity and binding by flow cytometry
Affinity determination was carried out as previously described by Kiss et al., 2006. Fifty microliters of streptavidin-coated beads (Spherotec) were incubated with either 1 ng of biotinylated target in a final volume of 100 µl PBS, for 1 h at room temperature. Hundred microliters of 5% BSA was added to block the bead surface and incubated for a further hour at room temperature. Beads were washed once in PBST and resuspended in 150 µl of PBS. GFP containing the anti-lysozyme CDR3 with an initial concentration of 0.6 mg/ml was 2-fold serially diluted and 5 µl of antigen-coated beads added to 50 µl diluted protein per well. After 1 h incubation at room temperature the supernatant was removed by washing once in PBST and the beads were analyzed by flow cytometry using a FACSAria instrument (BD Biosciences). For each dilution, the mean fluorescence intensity of the non-specific binding of GFP-lys to the beads coated with an irrelevant target was subtracted from the specific fluorescence of the lysozyme coated ones. The resulting fluorescence values at each dilution were fitted to a logistic function using Origin (Microcal Software, Inc., Northampton, MA) and the affinity determined as the concentration at which half maximal fluorescence was obtained (Sklar et al., 2002).
Protein expression and purification
Individual colonies were picked and grown overnight in liquid 2XTY/Kan/Glucose at 37°C. One microliter of confluent culture was used to inoculate 50 ml 2XTY/Kan/IPTG (0.1 M) in 250 ml shaker flasks for expression at 22°C for 48 h. For purification, cultures were harvested by centrifugation, sonicated and resuspended in lysis buffer (10 mM Tris/300 mM NaCl/10% glycerol/10 mM Imidazole, pH 8.0), and recentrifuged at 3000g for 30 min at 4°C. The supernatant was applied to IMAC columns pre-equilibrated with lysis buffer. The flow-through was reapplied three additional times and washed with 20 bed volumes of wash buffer (10 mM Tris/300 mM NaCl/10% glycerol/20 mM Imidazole, pH 8.0) before addition of elution buffer (10 mM Tris/300 mM NaCl/10% glycerol/300 mM Imidazole). The buffers were exchanged from the eluted proteins by spin filtration, using 10 000 MWCO Amicon Ultra centrifugal filtration devices at 4°C. The desalted proteins were diluted into prechilled 25 mM Tris pH 8.0 and stored at 4°C preceding further evaluation.
Antigen was coated onto MaxiSorb plates (Nunc) by incubating 1 µg target protein per well at 4°C overnight or at 37°C for 1 h. The antigen bound plate was washed with PBS and blocked with 200 µl of 5% BSA for 1 h at room temperature. Bacteria were grown in 96 deep well plates (Thompson) in 1 mL of autoinduction media for 48 h at 22°C, after which bacteria were lysed with 100 µl of POPculture (Novagen). Cell lysate was diluted with 1 ml PBS and 20 µl of the diluted cell lysate was added to the antigen-coated wells using a Tecan Robot system, incubated for 1 h and washed three times with PBST and three times with PBS using a PowerWash 384 plate washer. Bound GFP clones were revealed with monoclonal anti-SV5 antibody, which recognizes the C-terminal SV5 tag (Hanke et al., 1992), and alkaline phosphatase conjugated anti-mouse antibody (Dako Corp).
Different amounts of lysozyme (1 µg, 100 ng, 10 ng and 1 ng) were separated by SDS–polyacrylamide gel electrophoresis. The proteins were transferred to nitrocellulose membrane (Schleicher and Schull), and the membrane blocked with 5% BSA in PBS. Ten microliters 5 µg/ml (that is 0.16 µM) LG7 protein in 1% BSA was added and incubated for 20 min at room temperature. The membrane was then washed with PBST for 5 min and fluorescence detected with a digital camera under blue light.
FCS was performed on a home-built confocal microscope described in detail in (Dai et al., 2007) and experimental FCS curves were fitted using 3D diffusion equation (Eq. 2 in Dai et al., 2007). Hydrodynamic radii and the brightness of the samples (from the fits) are given inTable II. For anti-bunching measurements, the fluorescence signal was split 50/50 and detected with two single photon counting avalanche photodiodes in a classical Hanburry–Brown Twiss interferometer (Hanbury Brown and Twiss, 1956). The output pulses of the detectors were fed to the SYNC and CFD inputs of the SPC-630 card (Becker & Hickl GmbH, Berlin, Germany) with appropriate delay between signals so that the anti-bunching dip was observed in the center of the 50 ns time-to-amplitude converter window.
|
Funding |
|---|
|
TopAbstractIntroductionResultsDiscussionMaterials and methodsFundingReferences |
|---|
ARMB is grateful to the DOE GTL program for funding.
|
Footnotes |
|---|
Edited by Jim Marks
|
References |
|---|
|
TopAbstractIntroductionResultsDiscussionMaterials and methodsFundingReferences |
|---|
Abedi M.R., Caponigro G., Kamb A. Nucleic Acids Res. (1998) 26:623–630.
Baek H., Suk K.H., Kim Y.H., Cha S. Nucleic Acids Res. (2002) 30:e18.
Baird G.S., Zacharias D.A., Tsien R.Y. Proc. Natl Acad. Sci. USA (1999) 96:11241–11246.
Basche T., Moerner W.E., Orrit M., Talon H. Phys. Rev. Lett. (1992) 69:1516–1519.[CrossRef][Web of Science][Medline]
Bendtsen J.D., Nielsen H., von Heijne G., Brunak S. J. Mol. Biol. (2004) 340:783–795.[CrossRef][Web of Science][Medline]
Bessette P.H., Rice J.J., Daugherty P.S. Protein Eng. Des. Sel. (2004) 17:731–739.
Binz H.K., Pluckthun A. Curr. Opin. Biotechnol. (2005) 16:459–469.[CrossRef][Web of Science][Medline]
Binz H.K., Amstutz P., Kohl A., Stumpp M.T., Briand C., Forrer P., Grutter M.G., Pluckthun A. Nat. Biotechnol. (2004) 22:575–582.[CrossRef][Web of Science][Medline]
Bonnycastle L.L., Mehroke J.S., Rashed M., Gong X., Scott J.K. J. Mol. Biol. (1996) 258:747–762.[CrossRef][Web of Science][Medline]
Bradbury A., Velappan N., Verzillo V., Ovecka M., Chasteen L., Sblattero D., Marzari R., Lou J., Siegel R., Pavlik P. Trends Biotechnol. (2003) 21:275–281.[CrossRef][Web of Science][Medline]
Bukanov N.O., Meek A.L., Klinger K.W., Landes G.M., Ibraghimov-Beskrovnaya O. Funct. Integr. Genomics (2000) 1:193–199.[CrossRef][Medline]
Chasteen L., Ayriss J., Pavlik P., Bradbury A.R. Nucleic Acids Res. (2006) 34:e145.
Choo Y., Klug A. Proc. Natl Acad. Sci. USA (1994) 91:11163–11167.
Colas P., Cohen B., Jessen T., Grishina I., McCoy J., Brent R. Nature (1996) 380:548–550.[CrossRef][Medline]
Craig L., Sanschagrin P.C., Rozek A., Lackie S., Kuhn L.A., Scott J.K. J. Mol. Biol. (1998) 281:183–201.[CrossRef][Web of Science][Medline]
Dai M., Fisher H.E., Temirov J., Kiss C., Phipps M.E., Pavlik P., Werner J.H., Bradbury A.R. Protein Eng. Des. Sel. (2007) 20:69–79.
de Haard H.J., van Neer N., Reurs A., Hufton S.E., Roovers R.C., Henderikx P., de Bruine A.P., Arends J.W., Hoogenboom H.R. J. Biol. Chem. (1999) 274:18218–18230.
de Kruif J., Logtenberg T. J. Biol. Chem. (1996) 271:7630–7634.
Desmyter A., Transue T.R., Ghahroudi M.A., Thi M.H., Poortmans F., Hamers R., Muyldermans S., Wyns L. Nat. Struct. Biol. (1996) 3:803–811.[CrossRef][Web of Science][Medline]
Doi N., Yanagawa H. FEBS Lett. (1999) 453:305–307.[CrossRef][Web of Science][Medline]
Driessen A.J., Fekkes P., van der Wolk J.P. Curr. Opin. Microbiol. (1998) 1:216–222.[CrossRef][Web of Science][Medline]
Fekkes P., Driessen A.J. Microbiol. Mol. Biol. Rev. (1999) 63:161–173.
Geyer C.R., Colman-Lerner A., Brent R. Proc. Natl Acad. Sci. USA (1999) 96:8567–8572.
Greenwood J., Willis A.E., Perham R.N. J. Mol. Biol. (1991) 220:821–827.[CrossRef][Web of Science][Medline]
Han Z., Xiong C., Mori T., Boyd M.R. Biochem. Biophys. Res. Commun. (2002) 292:1036–1043.[CrossRef][Web of Science][Medline]
Hanbury Brown R., Twiss R.Q. Nature (1956) 177:27–29.[CrossRef]
Hanke T., Szawlowski P., Randall R.E. J. Gen. Virol. (1992) 73:653–660.
Hansen M.H., Ostenstad B., Sioud M. Int. J. Oncol. (2001) 19:1303–1309.[Web of Science][Medline]
Hansson M., Ringdahl J., Robert A., Power U., Goetsch L., Nguyen T.N., Uhlen M., Stahl S., Nygren P.A. Immunotechnology (1999) 4:237–252.[CrossRef][Web of Science][Medline]
Hilpert K., Hansen G., Wessner H., Kuttner G., Welfle K., Seifert M., Hohne W. Protein Eng. (2001) 14:803–806.
Hoogenboom H.R., Griffiths A.D., Johnson K.S., Chiswell D.J., Hudson P., Winter G. Nucleic Acids Res. (1991) 19:4133–4137.
Houshmand H., Bergqvist A. Biochem. Biophys. Res. Commun. (2003) 309:695–701.[CrossRef][Web of Science][Medline]
Huang S.C., Jiang R., Glas A.M., Milner E.C. Mol. Immunol. (1996) 33:553–560.[CrossRef][Web of Science][Medline]
Hufton S.E., Moerkerk P.T., Meulemans E.V., de Bruine A., Arends J.W., Hoogenboom H.R. J. Immunol. Methods (1999) 231:39–51.[CrossRef][Web of Science][Medline]
Huie M.A., Cheung M.C., Muench M.O., Becerril B., Kan Y.W., Marks J.D. Proc. Natl Acad. Sci. USA (2001) 98:2682–2687.
Jamieson A.C., Kim S.H., Wells J.A. Biochemistry (1994) 33:5689–5695.[CrossRef][Web of Science][Medline]
Jespers L.S., Messens J.H., De Keyser A., Eeckhout D., Van Den Brande I., Gansemans Y.G., Lauwereys M.J., Vlasuk G.P., Stanssens P.E. Biotechnology (N Y) (1995) 13:378–382.[CrossRef][Medline]
Kipriyanov S.M., Little M., Kropshofer H., Breitling F., Gotter S., Dübel S. Protein Eng. (1996) 9:203–211.
Kiss C., et al. Nucleic Acids Res. (2006) 34:e132.
Krumpe L.R., Atkinson A.J., Smythers G.W., Kandel A., Schumacher K.M., McMahon J.B., Makowski L., Mori T. Proteomics (2006) 6:4210–4222.[CrossRef][Web of Science][Medline]
Kurakin A., Wu S., Bredesen D.E. Methods Mol. Biol. (2004) 264:47–60.[Medline]
Lehmann D., Sodoyer R., Leterme S. Vet. Microbiol. (2004) 104:1–17.[CrossRef][Web of Science][Medline]
Malabarba M.G., Milia E., Faretta M., Zamponi R., Pelicci P.G., Di Fiore P.P. Oncogene (2001) 20:5186–5194.[CrossRef][Web of Science][Medline]
Markland W., Roberts B.L., Saxena M.J., Guterman S.K., Ladner R.C. Gene (1991) 109:13–19.[CrossRef][Web of Science][Medline]
Marks J.D., Hoogenboom H.R., Bonnert T.P., McCafferty J., Griffiths A.D., Winter G. J. Mol. Biol. (1991) 222:581–597.[CrossRef][Web of Science][Medline]
Miesenbock G., De Angelis D.A., Rothman J.E. Nature (1998) 394:192–195.[CrossRef][Medline]
Nicaise M., Valerio-Lepiniec M., Minard P., Desmadril M. Protein Sci. (2004) 13:1882–1891.[CrossRef][Web of Science][Medline]
Nowak J.E., Chatterjee M., Mohapatra S., Dryden S.C., Tainsky M.A. Biotechniques (2006) 40:220–227.[Web of Science][Medline]
Nuttall S.D., Rousch M.J., Irving R.A., Hufton S.E., Hoogenboom H.R., Hudson P.J. Proteins (1999) 36:217–227.[CrossRef][Web of Science][Medline]
Paschke M., Hohne W. Gene (2005) 350:79–88.[CrossRef][Web of Science][Medline]
Pedelacq J.D., Cabantous S., Tran T., Terwilliger T.C., Waldo G.S. Nat. Biotechnol. (2006) 24:79–88.[CrossRef][Web of Science][Medline]
Peelle B., Lorens J., Li W., Bogenberger J., Payan D.G., Anderson D.C. Chem. Biol. (2001) 8:521–534.[CrossRef][Web of Science][Medline]
Reichert J.M., Rosensweig C.J., Faden L.B., Dewitz M.C. Nat. Biotechnol. (2005) 23:1073–1078.[CrossRef][Web of Science][Medline]
Reid B.G., Flynn G.C. Biochemistry (1997) 36:6786–6791.[CrossRef][Web of Science][Medline]
Rondot S., Koch J., Breitling F., Dubel S. Nat. Biotechnol. (2001) 19:75–78.[CrossRef][Web of Science][Medline]
Ronto G., Agamalyan M.M., Drabkin G.M., Feigin L.A., Lvov Y.M. Biophys. J. (1983) 43:309–314.[Web of Science][Medline]
Rosenberg A.H., Lade B.N., Chui D.S., Lin S.W., Dunn J.J., Studier F.W. Gene (1987) 56:125–135.[CrossRef][Web of Science][Medline]
Saggio I., Laufer R. Biochem. J. (1993) 293:613–616.[Web of Science][Medline]
Santini C., Brennan D., Mennuni C., Hoess R.H., Nicosia A., Cortese R., Luzzago A. j. Mol. Biol. (1998) 282:125–135.[CrossRef][Web of Science][Medline]
Sblattero D., Bradbury A. Nat. Biotechnol. (2000) 18:75–80.[CrossRef][Web of Science][Medline]
Schiweck W., Buxbaum B., Schatzlein C., Neiss H.G., Skerra A. FEBS Lett. (1997) 414:33–38.[CrossRef][Web of Science][Medline]
Schlehuber S., Beste G., Skerra A. J. Mol. Biol. (2000) 297:1105–1120.[CrossRef][Web of Science][Medline]
Scott J.K., Smith G.P. Science (1990) 249:386–390.
Shi H., Wen Su W. Enzyme Microb. Technol. (2001) 28:25–34.[CrossRef][Web of Science][Medline]
Siegel M.S., Isacoff E.Y. Neuron (1997) 19:735–741.[CrossRef][Web of Science][Medline]
Sklar L.A., Edwards B.S., Graves S.W., Nolan J.P., Prossnitz E.R. Annu. Rev. Biophys. Biomol. Struct. (2002) 31:97–119.[CrossRef][Web of Science][Medline]
Slootweg E.J., Keller H.J., Hink M.A., Borst J.W., Bakker J., Schots A. Nucleic Acids Res. (2006) 34:e137.
Soltes G., Barker H., Marmai K., Pun E., Yuen A., Wiersma E.J. J. Immunol. Methods (2003) 274:233–244.[CrossRef][Web of Science][Medline]
Sparks A.B., Quilliam L.A., Thorn J.M., Der C.J., Kay B.K. J. Biol. Chem. (1994) 269:23853–23856.
Steiner D., Forrer P., Stumpp M.T., Pluckthun A. A Signal sequences directing cotranslational translocation expand the range of proteins amenable to phage display. Nat. Biotechnol (2006) 24:823–831.[CrossRef][Web of Science][Medline]
Stroud R.M., Serwer P., Ross M.J. Biophys. J. (1981) 36:743–757.[Web of Science][Medline]
Studier F.W. Protein Expr. Purif. (2005) 41:207–234.[CrossRef][Web of Science][Medline]
Sykora J., Kaiser K., Gregor I., Bonigk W., Schmalzing G., Enderlein J. Exploring fluorescence antibunching in solution to determine the stoichiometry of molecular complexes. Anal. Chem. (2007) 79:4040–4049.[Medline]
Takakusagi Y., Kobayashi S., Sugawara F. Bioorg. Med. Chem. Lett. (2005) 15:4850–4853.[CrossRef][Medline]
Tan G.H., Yusoff K., Seow H.F., Tan W.S. J. Med. Virol. (2005) 77:475–480.[CrossRef][Web of Science][Medline]
Taussig M.J., et al. Nat. Methods (2007) 4:13–17.[CrossRef][Web of Science][Medline]
Terskikh A.V., Le Doussal J.M., Crameri R., Fisch I., Mach J.P., Kajava A.V. Proc. Natl Acad. Sci. USA (1997) 94:1663–1668.
Toivola J., Ojala K., Michel P.O., Vuento M., Oker-Blom C. Biol. Chem. (2002) 383:1941–1946.[CrossRef][Web of Science][Medline]
Tsien R.Y. Annu. Rev. Biochem. (1998) 67:509–544.[CrossRef][Web of Science][Medline]
Vaughan T.J., Williams A.J., Pritchard K., Osbourn J.K., Pope A.R., Earnshaw J.C., McCafferty J., Hodits R.A., Wilton J., Johnson K.S. Nat. Biotechnol. (1996) 14:309–314.[CrossRef][Web of Science][Medline]
Yayon A., Aviezer D., Safran M., Gross J.L., Heldman Y., Cabilly S., Givol D., Kathcalski-Katzir E. Proc. Natl Acad. Sci. USA (1993) 90:10643–10647.
Zhang J., Tanha J., Hirama T., Khieu N.H., To R., Tong-Sevinc H., Stone E., Brisson J.R., MacKenzie C.R. J. Mol. Biol. (2004) 335:49–56.[CrossRef][Web of Science][Medline]
Zwick M.B., Shen J., Scott J.K. Curr. Opin. Biotechnol. (1998) 9:427–436.[CrossRef][Web of Science][Medline]
Received March 10, 2008; revised March 10, 2008; accepted March 13, 2008.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||