The creation of large phage antibody libraries has become an important
goal in selecting antibodies against any antigen. Here we describe a method
for making libraries so large that the complete diversity cannot be accessed
using traditional phage technology. This involves the creation of a primary
phage scFv library in a phagemid vector containing two nonhomologous lox sites.
Contrary to the current dogma, we found that infecting Cre recombinase−expressing
bacteria by such a primary library at a high multiplicity of infection results
in the entry of many different phagemid into the cell. Exchange of Vh
and Vl genes between such phagemids creates many new V
h/Vl combinations, all of which are functional. On the basis
of the observed recombination, the library is calculated to have a diversity
of 31011. A library created using this method was validated
by the selection of high affinity antibodies against a large number of different
protein antigens.antibody engineeringphage displayrecombinationCre recombinasefilamentous phagesingle-chain Fv (scFv)
Phage display was recently introduced as a means of making antibodies in
vitro1,
2,
3,
4,
5. In general, the affinity of the antibodies
isolated is proportional to the initial size of the library used for selection6,
7. Thus large libraries have become important as sources of high-affinity
antibodies to virtually any antigen3,
7,
8,
9. Most such libraries
have been made using cloning, and when cloning procedures are optimized, can
have diversities of >1010, from which subnanomolar affinity
antibodies have been isolated7,
9.
Recombination has been proposed as an alternative to cloning for creation
of large libraries10,
11,
12. Lambda recombinase has been used
to recombine Fabs10, and Cre recombinase to recombine both Fabs12 and single-chain Fv's (scFv)11. For making large
phage antibody libraries, however, there has been only one report describing
the use of Cre to recombine Fabs9. This library, derived from
synthetic libraries of 108 Vh and 810
5 Vl, had a final estimated diversity of 6.510
10. Although antibodies with nanomolar affinities were isolated, the
system is difficult to use, since deletion of antibody genes occurs at relatively
high frequencies because the library is formed in phage.
In all the recombination systems described above, Vh genes are
cloned in one vector, Vl in another, and recombination is used to
create a third vector capable of displaying functional antibodies. When the
products of recombination are selected by the use of newly created antibiotic
resistances10,
11, all plasmids with the appropriate resistance
should be recombined correctly. This reflects an underlying strategy to render
recombination irreversible, since in all three cases, a significant proportion
of the vectors present after recombination are derived either from nonfunctional
parental plasmids or from undesired products of recombination.
Here we describe a method that uses a single vector to exploit the reversibility
of Cre-catalyzed recombination. First we created a relatively small primary
library (7107 was used here) in a phagemid vector in
which the Vh and Vl genes are separated by two nonhomologous
lox sites. The Vh and Vl genes in this primary library were
then recombined by infecting the phagemid into Cre-expressing bacteria at
high multiplicity of infection (MOI). Under these conditions, many different
phagemids enter a single bacterium, and the Vh and Vl genes
were exchanged between different phagemids, creating many new Vh/V
l combinations, all of which are functional.
Results A phagemid antibody display vector with incorporated lox sites.
To use Cre recombinase to shuffle V genes in the scFv format,
a lox site must be placed between the Vh and Vl genes, which
requires a translated lox site as a protein linker. By examining the six possible
frames for the wild-type loxP site and the mutated loxP511 site (which will
not recombine with the wild-type loxP (15)),
we identified a translation of loxP511 (ITSYNVYYTKL) that had only a single
basic amino acid (to reduce the possibility of proteolysis), lacked stop codons,
and was the least hydrophobic. The ability of this sequence to act as an scFv
linker was tested following the construction of a new phage display vector,
pDAN5 (Fig. 1A), in which the described translation
of the loxP511 site was used as a protein linker between Vl and V
h. Three different monoclonal antibodies (mAbs) (D1.3, antilysozyme16; Y13.259, anti-p21ras (17); and
GL30, antigliadin), were cloned into pDAN5. In each case, scFvs with the loxP511
linker were able to recognize the appropriate antigen specifically and gave
ELISA signals comparable to those obtained using display vectors with the
standard glycine-serine linker (Fig. 1B).
Figure 1. (A) Map of the display vector pDAN5 with an scFv cloned.
Sites used for V gene cloning are in bold. BssHII,
BspEI, SalI, XhoI, KpnI, NheI. (B) ELISA signals
with scFvs derived from monoclonal antibodies. Vh and Vl
genes cloned into pDAN5 in the scFv format were tested for binding to their
own antigens (D1.3: lysozyme; Y13-259: p21ras peptide; GL30: gliadin), as
well as to the other tested antigens and a control antigen: human serum albumin
(HSA).
Recreating a functional scFv by shuffling the V genes. The ability to shuffle Vh and Vl genes in vivo to create
functional antibodies was tested using an scFv derived from the antilysozyme
mAb D1.3 (16). Two scFvs that contained
either D1.3 Vh or D1.3 Vl with irrelevant partner chains,
X and Y, were created (Vl/X-Vh/D1.3 and Vl/D1.3-V
h/Y). Recognition of lysozyme by D1.3 scFv was shown to require the presence
of both D1.3 heavy and light chains. Single D1.3 chains associated with irrelevant
partner chains were nonfunctional by ELISA. Phagemid containing these scFv
genes were allowed to infect Escherichia coli expressing Cre recombinase
at an MOI of 20:1. If recombination was successful, each bacterium would contain
four different scFv genes (Vl/D1.3-Vh/D1.3; Vl/D1.3-V
h/Y; Vl/X-Vh/D1.3 and Vl/X-Vh/Y; Fig. 2). Phagemid were rescued from such bacteria and coupled
to the appropriate scFv protein by passage through normal E. coli infected
at MOI<1 (Fig. 2). In the presence of Cre
recombinase, recombination was demonstrated in 25% of phagemid by PCR and
17% by ELISA (Table 1), whereas in wild-type bacteria,
no recombination was observed. These results indicate that recombination induced
by Cre recombinase can be used to shuffle Vh and Vl genes
between different phagemids, thereby giving rise to new specificities. Furthermore,
the recombination reaction appears to go to equilibrium, as Vl/D1.3-V
h/D1.3 comprises approximately 25% of the total.
Figure 2. scheme of D1.3 recombination experiment.
Two phagemid containing Vl/X-Vh/D1.3 and Vl/D1.3-V
h/Y (where X and Y represent irrelevant V genes) are added to either
Cre-expressing bacteria or wild-type DH5F at an MOI of 20:1. After
overnight growth, phagemid are made, reinfected into DH5F' at
a phagemid:bacteria ratio of <1 (to couple genotype and phenotype), and
tested by PCR and ELISA for the presence of functional Vl/D1.3-V
h/D1.3.
Creating a large recombined antibody library in pDAN5. A small scFv phagemid library consisting of 7107
independent clones was created in pDAN5 by cloning PCR-assembled scFv derived
from peripheral blood V, V, and V genes *(see Experimental
Protocol for details). This primary library was used to infect bacteria expressing
Cre recombinase following the scheme illustrated in Figure
2, except that instead of the two D1.3 phagemids, the library of 710
7 phagemids was used, and the MOI was 200:1 instead of 20:1. This
procedure results in bacteria containing multiple phagemids, each of which
encodes different Vh and Vl genes, which can be recombined
by the Cre recombinase. Following recombination, phagemid were derived from
these bacteria and used to infect bacteria not expressing Cre (DH5F')
at MOI<1 to couple phenotype and genotype, as indicated in Figure 2.
In order to determine the number of different phagemid entering a single
bacterium, and the degree of recombination occurring between them, single
bacteria that had undergone infection and recombination were plated out to
isolate single colonies. Such colonies contain the complement of Vh
and Vl genes present in the single bacteria before plating out. These
colonies were individually amplified in liquid culture, used to prepare phagemid,
and subsequently rescued as individual colonies. The scFvs present in all
the individual colonies represent those found in the original cell giving
rise to the single colony that was amplified. The scFvs present in a number
of these colonies were then characterized by either sequencing or separately
amplifying Vh and Vl genes and fingerprinting them with
the restriction enzyme BstNI. Results obtained with the two methods
were essentially identical, with different fingerprint patterns always representing
different Vh and Vl genes.
On the basis of either sequencing or fingerprint patterns, the different
Vh and Vl genes arising from a single cell were numbered
for five different individual colonies, with 40−80 V genes analyzed
for each. Results for these were very similar, with 12−18 different
Vh and 12−15 different Vl genes identified for each
cell (one colony is shown in Table 2). These were
present in 19−30 different scFv combinations, with the vast majority
of scFvs containing shared V genes. In the cell illustrated, 11 examples of
all four possible combinations of Vh/Vl gene pairs were
identified (shown in underlined bold copy in Table 2),
indicating that recombination had been extensive. Only one Vh/V
l pair did not appear to have participated in recombination in this cell.
Table 2. Assessing the diversity of Cre bacteria infected by multiple phagemid.
One concern with Cre/lox systems is their perceived instability. We analyzed
full-length scFvs by PCR (data not shown) and found that 4 (4%) of 96 from
the primary library and 6 (6%) of 96 from the recombined library contained
genes that had either the Vh or Vl gene deleted. The lox
site was preserved in all sequenced scFvs, including those that did not appear
to participate in recombination. Therefore, these deletions are likely to
have occurred during library construction, probably as a result of spurious
priming during PCR assembly.
Testing the large phage antibody library. The library
was tested by selection on a number of different protein antigens (Table 3A, and data not shown). Antibodies were obtained against all 15 antigens attempted,
with a range of 3 to 11 antibodies per antigen and a mean of over 6 antibodies
per antigen. Antibodies specifically recognized their antigens and were not
polyreactive. Eight of these scFvs were purified by affinity chromatography
using the His tag and were characterized by gel filtration. All scFvs were
monomeric and did not show the 'diabody' peak associated with
scFv containing shorter linkers18. Four of the monomeric scFv
peaks were purified, and their affinities were assessed by surface plasmon
resonance. All had affinities <90 nM, with the best having an affinity
of 15 nM (Table 3B).
Table 3. Antigens against which antibodies have been selected.
Variable gene sequence analysis. The different V genes
found in the primary library, the recombined library and a number of selected
scFvs were analyzed by sequencing 24−40 V genes for each category. After
sequencing, the origin of the V gene and complementarity determining region
3 (CDR3) length were determined using the V BASE immunoglobulin V gene database19. The primary and recombined libraries were diverse, with almost
all families represented (Fig. 3A). These were not evenly
distributed, with Vh genes having a predominance of Vh4
genes, and Vl genes having more V1 and V3 genes. There
was no great difference between the primary and recombined libraries, indicating
that diversity does not appear to be compromised by recombination.
Figure 3. V gene use and CDR3 length in primary, recombined, and selected scFvs.
V gene family (A) and CDR3 length (B) were determined using V BASE19, and the percentage of V genes deriving from a particular V gene
family and the CDR3 length are indicated for scFvs derived from the primary
or recombined library, and for selected scFvs recognizing specific antigens.
Upon selection, the V gene distribution changed somewhat, as shown previously9. Whereas the primary and recombined libraries had an excess of V
h4 genes, the V genes found in antibodies recognizing selected antigens
appeared to be more widely distributed, with Vh1, Vh3, V
h5, and Vh6 being the Vh gene classes found most frequently.
Among the Vl genes, 38% of all selected antibodies contained a gene
of the Vl3 family, even though this accounts for <10% of the primary
and recombined libraries. The CDR3 length, on the other hand (
Fig. 3B), does not appear to change very much between the three populations
of antibodies, with a wide distribution of all lengths found, ranging between
six and 22 amino acids for the Vh genes and seven and 13 for the
Vl genes.
Discussion Although phage antibody libraries have been used to isolate antibodies
against a large number of antigens3,
5,
7,
8,
9,
20,
21, the
procedure is still not widely used, possibly because of the difficulty of
making large phage antibody libraries, which require that at least 10
9 independent clones be derived from a good source of diverse V
h and Vl genes. In general, such libraries are made by carrying
out a large number of ligations and transfections, and once made, become a
limited resource as amplification cannot be carried out without a potential
reduction in diversity. The method described here overcomes this problem,
as diversity is regenerated each time recombination is used to create each
new secondary library. In fact, given sufficient resources, library adequate
for 107 selections can be made from an initial primary library
without losing diversity. This is likely to be an important factor in the
application of phage antibodies to functional genomics.
Other recombination methods used or proposed for making antibody libraries3,
10,
11,
12 use two plasmids to generate the library. The V
h genes are cloned on one plasmid, and the Vl genes on another.
After recombination, four plasmid populations are generated, only one of which
is correct (containing heavy and light library chains). The others contain
copies of either single library chains in combination with dummy chains, or
dummy chains alone. Although these extra products should not be packaged into
phage, we have found that plasmids that lack an Ff origin of replication may
still be packaged, and as a result will contaminate the library. The use of
a single plasmid to generate diversity ensures that all recombination products
are functional. It is generally believed that one bacterium cannot be infected
with more than one phage or phagemid, because p3 expression is believed to
inhibit pilus synthesis. We have overcome this problem by inhibiting p3 synthesis
with glucose. This does not, however, explain the maintenance of so many different
phagemids with the same origin of replication within a single cell. The mechanisms
of this are presently under investigation.
Although assessing the degree of diversity created in a single cell is
difficult, we found similar results for five colonies analyzed. In the colony
shown, 13 different Vh genes and 12 different Vl genes were
identified. These were recombined in a total of 30 different combinations,
with all except one scFv containing Vh or Vl genes also
found in other scFvs. One Vl was found with six different Vh
's, one Vh with six different Vl's, and 11 overlapping
cases of all four combinations of a Vh/Vl pair were identified
(Table 2), indicating that recombination had been extensive.
Interestingly, in all five cells analyzed, no single scFv dominated the
analysis, with the most abundant being present in only five copies (14%),
indicating that all scFv appear to have similar probabilities of remaining
within the cell. The single sequenced scFv that did not appear to have participated
in recombination was normal, with a functional lox site present. Thus, its
apparent nonparticipation was probably due to not having analyzed enough clones
rather than an intrinsic problem of the scFv itself. The minimum diversity
generated by a single cell can be estimated to be the number of recombined
scFv actually observed: 26−30, which gives a minimum library size estimate
of 31011. However, if the 13−18 different V
h genes rescued from a single colony are matched by 13−18 different
Vl genes, the potential diversity identified in this small sample
is 169−324 (132−182), giving a maximum estimate that approaches
the 500−700 copy number of pUC-based plasmids22, and the
potential for a library at least 10-fold larger.
Previous libraries created using the Cre/lox system9 have
suffered from instability as a result of deletion of antibody V genes. This
has been attributed to the fact that the library was constructed in a filamentous
phage vector, although the possibility that the lox site in some way confers
instability cannot be excluded. In this study, we found that the scFv genes
containing the lox site are remarkably stable, with the percentage of full-length
scFv genes remaining essentially identical in the passage from primary (4%)
to recombined library (6%). These figures probably represent the growth advantage
of preexisting deleted clones23 during the growth cycles required
for the recombination process. Furthermore, in all full-length scFvs, the
lox site was always found to be present by sequencing.
The possibility that the recombination process itself may induce bias in
the library was examined by sequencing V genes from the primary and recombined
libraries. An increase in the representation of V3 and V2 genes,
and a reduction in Vh1 and V4 genes in the recombined library
compared with the primary was found, but in general the ratio of the different
V genes was remarkably well preserved. Interestingly, neither the primary
nor the recombined library was dominated by single V gene families, as has
been found in other naive libraries, and representative genes from almost
all Vh and Vl gene families were found. This may be because
of the V gene primer set used24, which was specifically designed
to be able to amplify all known germline V genes, or because amplification
of V genes was carried out with individual primers, rather than mixtures.
The diversity of V gene families in the primary and recombined libraries was
also reflected in the V genes of scFvs selected for antigen binding. Although
V3 genes are frequently found, as shown for scFvs selected from other
published naive libraries7,
9, there is no predominance of other
V gene families, and members of almost all V gene families can be found. Of
22 different selected scFvs sequenced, 14 different Vh genes, and
13 different Vl genes were found, with no example of identical V
genes being found in two scFvs recognizing different antigens. By comparing
V gene distribution in the recombined library to that found in selected scFvs,
we found that some V genes (e.g., Vh4, V1, and V3)
were recovered more frequently in the library than in selected scFvs, whereas
others (e.g., Vh1, Vh3, V1, and V3) were
recovered more often in selected scFvs than in the recombined library. Whether
this represents V genes that, in general, are more likely to recognize antigens
of interest, or reflects the fact that relatively few antigen-binding scFvs
have been analyzed, awaits further analysis.
The affinities of the antibodies isolated were all better than 90nM, with
the best having an affinity of 15 nM. This is lower than the best affinities
reported for the larger libraries, which in some cases were subnanomolar7,
9. However, the highest affinities reported in these papers were
all obtained by selection in the soluble phase using biotinylated antigen
and magnetic streptavidin beads, or alternatively by selecting on haptens.
When selections were performed as described here (protein antigens coupled
to immunotubes), the affinities obtained were similar (see Sheets et al. for
a full discussion of this point). Furthermore, the bacterial elution method
used here has recently been shown to be far less efficient than more stringent
methods (e.g., 100 mM HCl, pH 1.1, or 100 mM triethylamine) at eluting high-affinity
antibodies25, suggesting that the antibodies with the highest
affinities may well have been left on the immunotube.
Despite these caveats, the affinities of antibodies selected from the recombined
library still exceed those selected from libraries with diversities similar
to the primary library used (e.g., the 3107 library
described by Marks et al.21), indicating that antibodies of
high affinity can be selected from initial small libraries when recombination
is carried out. This is not surprising, given that the large libraries created
using traditional methods do not have greater numbers of different Vh
and Vl genes, but a greater proportion of the possible combinations
of these genes. This is reflected in the recent trend to first make small
Vh and Vl libraries and then combine these by cloning7,
8,
9, a procedure that is carried out far more efficiently by in
vivo recombination.
While there have been practical demonstrations that there is an affinity
advantage to creating larger libraries, the search for larger and larger libraries
reaches a practical limit in the volume required to perform a selection. This
is fixed at approximately 1013 phage in a volume of 1 ml, indicating
that diversity >1012 probably remains untapped (assuming a
display level of 10%26). It is unlikely that standard methods
of library creation could reach this level of diversity, given that libraries
100 times smaller already require hundreds of transfections to create. However,
it should be possible to tap this level of diversity using the method we describe
here by increasing the volume of bacteria in the preparation of the final
library to 3−5 l. With still higher volumes, one can contemplate
the creation of far larger libraries, which could be used at the industrial
level.
By including cycles of recombination between rounds of selection, the effective
diversity accessed would be expected to approach the theoretical maximum diversity
of the library. Binding V regions could be shuffled either against one another
or against the starting library. In both cases, this would be the equivalent
of shuffling the chains of all binding antibodies in parallel, in a fashion
similar to that which has been carried out in series for single antibodies
with notable increases in affinity4,
27. We expect the affinities
of antibodies obtained in this way to be higher than those reported here.
This approach could also be applied to affinity maturation, with mutations
introduced simultaneously in both heavy and light chains, giving the advantage
that all potential combinations of mutations can be sampled simultaneously.
BS1365: BS591 F' kan (BS591: recA1 endA1 gyrA96 thi-1 lacU169
supE44 hsdR17 [lambda imm434 nin5 X1-cre]28)
Creating pDAN5 (scFv display vector) and pDAN5-scFv derivatives. To make
pDAN5, a new polylinker was cloned into pUC119 by overlap PCR of two long
oligonucleotides. This introduced a bacterial leader sequence, a polycloning
cassette, the SV5 tag29, a His6 tag, and an amber
stop codon (Fig. 1). Gene 3 was subsequently cloned
by PCR amplification from fdtet, with the 5' end downstream of the amber
stop codon, and the 3' end upstream of a wild-type lox sequence. The
Y13-259 scFv was assembled (in the order Vl-Vh) from Y13-259
scFv (Vh-Vl order)30 using murine V region−specific
primers based on those previously described31 in which the standard (gly4ser)
3 scFv linker was replaced with the loxP511 site. This was cloned into
pDAN5, to create pDAN5-Y13-259, using BssHII and NheI. Other
pDAN-scFvs were cloned, either as assembled scFv PCR fragments, or as individual
Vh and Vl genes amplified to incorporate BssHII and
SalI for Vl and XhoI and NheI for Vh.
Primary library construction. Total RNA was prepared
from 40 different samples of human peripheral blood lymphocytes purified by
Ficoll Hypaque (Amersham Pharmacia Biotech, UK). cDNA was synthesized using
random hexamers and reverse transcriptase following standard protocols. IgM
V regions were amplified using an IgM 3' primer (GGA AAA GGG TTG GGG
CGG AT) and the 5' Vh primers and methods as previously described24. The primers described in this paper were also used to amplify
Vl and V genes from random primed cDNA, and Vh genes
from gel-purified IgM V genes. Vh and Vl genes were reamplified
to add a region of overlap in the scFv linker as well as long tails to facilitate
restriction enzyme digestion. The scFv were assembled by mixing equimolar
amounts of Vh and Vl genes and performing assembly as previously
described31. The scFv obtained were cloned into BssHII/
NheI-cut pDAN5−Y13-259 to obtain a primary library of 710
7 clones.
Recombination and the creation of the secondary library. To induce recombination, BS1365 (which expresses Cre-recombinase constitutively)
was grown in 20 ml 2TY, 100 mg/ml kanamycin, 1% glucose at 37°C
to OD550 0.5. Phagemid were added at MOI 20:1 for the D1.3 shuffling
experiment and 200:1 when the library was created. Phagemid were left for
1 h without shaking at 37°C to allow infection, and ampicillin was then
added to 100 mg/ml. Recombination was allowed to continue by shaking overnight
growth at 30°C.
After recombination, bacteria were diluted 1/20 in the same growth medium,
and grown to OD550 0.5 at 37°C. M13K07 helper phage were added
at an MOI of 20:1, and left without shaking for 1 h at 37°C before further
growth (6 h to overnight). Additional kanamycin was not added because BS1365
expresses kanamycin resistance. Phagemid were prepared by centrifugation and
polyethylene glycol precipitation as previously described4.
As these phagemid arise from bacteria containing many different scFvs, there
is no coupling between phenotype and genotype. This is overcome by using the
isolated phagemid to infect DH5F' grown to OD550 0.5
at MOI1. Phagemid used for selection were prepared from overnight
cultures of bacteria derived at this low MOI using standard techniques. When
making the library, 20 ml of Cre-expressing bacteria were used. These were
diluted into 400 ml to prepare phagemid, and 51011 of
these phagemid were used to infect 1 L of DH5F' (510
11 bacteria). When recombining D1.3, all volumes used were 1 ml with
appropriate dilutions.
Assessment of diversity in individual cells. After
overnight growth in Cre-expressing bacteria, the culture was plated out to
isolate individual colonies. These contain multiple recombined V genes. A
number of these colonies were grown in 10 ml 2TY, 1% glucose, and 100g/ml
ampicillin at 37°C to OD550 0.5. M13K07 helper phage were added
at an MOI of 20:1, the culture was left for 1 h at 37°C without shaking,
and then left to grow 2−4 h at 30°C shaking (250 r.p.m.). Phagemids
were prepared from these cultures as described above. All phagemids present
in each individual culture represent all the different scFv combinations that
have arisen within the original starting cell. Colonies were derived from
these individual phagemid by infection at MOI1 into DH5F',
and individual Vh and Vl chains present in each phagemid
were identified by sequencing or BstNI fingerprinting.
Proteins. Proteins used for selection were kindly provided
by Min Park (Rad52), Scott Peterson (cyclin D, cdk2, and cdc25A), Tom Peat
(phosphoglycerate dehydrogenase). Human serum albumin was purchased from Sigma
(St. Louis, MO).
Selection of phage antibodies. Phage antibody selection
was performed essentially as previously described4, with proteins
coupled to immunotubes (Nunc, Rochester, NY) at 10 mg/ml overnight, blocked
in 2% nonfat milk−phosphate buffered saline (MPBS) and incubated with
the phage antibody library (also blocked in MPBS) for 1−2 h. Washing
after the first cycle involved five PBS and five PBS−0.1% Tween-20 washes.
Phage were eluted by the addition of 1 ml DH5F at OD550
0.5. Following elution, bacteria were amplified and phage prepared for further
cycles of selection. Subsequent washes were more stringent, and phage antibodies
were tested for reactivity by ELISA after the second or third cycle.
Phage ELISA testing of antibody binding specificity. Phage ELISA was used to identify positive lysozyme binding D1.3 scFvs created
by recombination. Phage ELISAs were also performed on individual phage clones
isolated following selection on target proteins. Positive clones gave signals
at least three times the background signal of 0.08. These were fingerprinted
using V gene primers23 for amplification and BstNI to
identify the number of different positive clones.
Sequencing. Sequencing was carried out using the Epicentre
Sequitherm Excel II kit (Alsbyte, Mill Valley, CA) and analyzed using specific
labeled primers annealing within the SV5 tag region and the leader sequence.
Sequences were analyzed on a Li-Cor 4000L automatic sequencer (Lincoln, NE).
The identity of the different V genes was analyzed by submitting the sequence
to V BASE19. Twenty-two scFvs were sequenced for the primary
and 26 for the recombined libraries; 22 different selected scFvs were sequenced;
and in the single cell analysis, 35 different scFvs were sequenced.
scFv purification. Periplasmic extracts were made from
500 ml volumes of bacteria induced with 0.5 mM isopropylthiogalactoside at
an OD550 of 0.6. The scFv were purified using the Ni-NTA kit (Qiagen,
Valencia, CA) and checked by 12% PAGE. Gel filtration was carried out on a
Superdex 75 column (Amersham Pharmacia Biotech, UK) to analyze and isolate
the monomeric scFv fraction.
Affinity determination. We calculated scFv dissociation
equilibrium constants (Kd) from the association (kon)
and dissociation (koff) rate constants determined using surface
plasmon resonance in a BIACORE 2000 instrument (Biacore AB, Uppsala, Sweden).
Calculation of Kd values was performed by fitting the data according
to a single-site model, using the BIAevaluation 3 software (Biacore AB). Approximately
200 RU of each antigen were immobilized on CM5 sensor chips (Biacore AB).
The scFv were purified by immobilized metal-ion affinity chromatography (IMAC)
and gel filtration as described above and used at 50−300 nM.
Received 16 August 1999; Accepted 10 November 1999
REFERENCES
Barbas, C.F., Kang, A.S., Lerner, R.A. & Benkovic, S.J. Assembly of combinatorial antibody libraries on phage surfaces: the gene III site. Proc. Natl. Acad. Sci. USA88, 7978−7982 (1991). | PubMed | ChemPort |
Griffiths, A.D. et al. Human anti-self antibodies with high specificity from phage display libraries. EMBO J. 12, 725−734 (1993). | PubMed | ISI | ChemPort |
Griffiths, A.D. et al. Isolation of high affinity human antibodies directly from large synthetic repertoires EMBO J.13, 3245−3260 (1994). | PubMed | ISI | ChemPort |
Marks, J.D. et al. By-passing immunization: building high affinity human antibodies by chain shuffling. Bio/Technology. 10, 779−783 (1992). | Article | PubMed | ISI | ChemPort |
Nissim, A. et al. Antibody fragments from a 'single pot' phage display library as immunochemical reagents. EMBO J. 13, 692−698 (1994). | PubMed | ISI | ChemPort |
Perelson, A.S. & Oster, G.F. Theoretical studies of clonal selection: minimal antibody repertoire size and reliability of self non-self discrimination. J. Theor. Biol.81, 645−670 (1979). | PubMed | ISI | ChemPort |
Vaughan, T.J. et al. Human antibodies with sub-nanomolar affinities isolated from a large non-immunized phage display library. Nat. Biotechnol.14, 309−314 (1996). | Article | PubMed | ISI | ChemPort |
de Haard, H.J. et al. A large non-immunized human Fab fragment phage library that permits rapid isolation and kinetic analysis of high affinity antibodies. J. Biol. Chem. 274, 18218−18230 (1999). | Article | PubMed | ISI | ChemPort |
Sheets, M.D. et al. Efficient construction of a large nonimmune phage antibody library; the production of panels of high affinity human single-chain antibodies to protein antigens. Proc. Natl. Acad. Sci. USA95, 6157−6162 (1998). | Article | PubMed | ChemPort |
Geoffroy, F., Sodoyer, R. & Aujame, L. A new phage display system to construct multicombinatorial libraries of very large antibody repertoires. Gene151, 109−113 (1994). | Article | PubMed | ISI | ChemPort |
Tsurushita, N., Fu, H. & Warren, C. Phage display vectors for in vivo recombination if immunoglobulin heavy and light chain genes to make large combinatorial libraries. Gene172, 59−63 (1996). | Article | PubMed | ISI | ChemPort |
Waterhouse, P., Griffiths, A.D., Johnson, K.S. & Winter, G. Combinatorial infection and in vivo recombination: a strategy for making large phage antibody repertoires. Nucleic Acids Res. 21, 2265−2266. (1993). | PubMed | ISI | ChemPort |
Hoess, R., Wierzbicki, A. & Abremiski, K. The role of the loxP spacer region in P1 site-specific recombination. Nucleic Acids Res. 14, 2287−2300 (1986). | PubMed | ISI | ChemPort |
Mariuzza, R.A. et al. Preliminary crystallographic study of the complex between the Fab fragment of a monoclonal anti-lysozyme antibody and its antigen. J. Mol. Biol. 170, 1055−1058 (1983). | PubMed | ISI | ChemPort |
Furth, M.E., Davis, L.J., Fleurdelys, B. & Scolnic, E.M. Monoclonal antibodies to the p21 products of the transforming gene of Harvey murine sarcoma virus and of the cellular ras gene family. J. Virol.43, 294−304 (1982). | PubMed | ISI | ChemPort |
Holliger, P., Prospero, T. & Winter, G. "Diabodies": small bivalent and bispecific antibody fragments. Proc. Natl. Acad. Sci. USA. 90, 6444−6448 (1993). | PubMed | ChemPort |
de Kruif, J., Boel, E. & Logtenberg, T. Selection and application of human single chain Fv antibody fragments from a semi-synthetic phage antibody display library with designed CDR3 regions. J. Mol. Biol.248, 97−105 (1995). | Article | PubMed | ChemPort |
Marks, J.D. et al. By-passing immunization. Human antibodies from V-gene libraries displayed on phage. J. Mol. Biol.222, 581−597 (1991). | PubMed | ISI | ChemPort |
Sambrook, J., Fritsch, E.F. & Maniatis, T. Molecular cloning: a laboratory manual(Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY; 1989).
de Bruin, R., Spelt, K., Mol, J., Koes, R. & Quattrocchio, F. Selection of high-affinity phage antibodies from phage display libraries Nat. Biotechnol.. 17, 397−399 (1999). | Article | PubMed | ISI | ChemPort |
Sblattero, D. & Bradbury, A. A definitive set of oligonucleotide primers for amplifying human V regions. Immunotechnology3, 271−278 (1998). | Article | PubMed | ISI | ChemPort |
Schier, R. & Marks, J.D. Efficient in vitro affinity maturation of phage antibodies using BIAcore guided selections. Hum. Antib. Hybrid. 7, 97−105. (1996). | ChemPort |
Clackson, T. & Wells, J.A. In vitro selection from protein and peptide libraries. Trends Biotechnol.12, 173−184. (1994). | Article | PubMed | ISI | ChemPort |
Schier, R. et al. Isolation of high-affinity monomeric human anti-c-erbB-2 single chain Fv using affinity-driven selection. J. Mol. Biol.255, 28−43 (1996). | Article | PubMed | ISI | ChemPort |
Sauer, B. & Henderson, N. The cyclization of linear DNA in Escherichia coli by site-specific recombination. Gene. 70, 331−341 (1988). | Article | PubMed | ISI | ChemPort |
Hanke, T., Szawlowski, P. & Randall, R.E. Construction of solid matrix-antibody-antigen complexes containing simian immunodeficiency virus p27 using tag-specific monoclonal antibody and tag-linked antigen. J. Gen. Virol.73, 653−660 (1992). | PubMed | ISI | ChemPort |
Werge, T.M., Bradbury, A., Di Luzio, A. & Cattaneo, A. A recombinant cell line, expressing a form of the anti-p21ras antibody, Y13-259, which binds protein A and may be produced as ascites. Oncogene7, 1033−1035 (1992). | PubMed | ISI | ChemPort |
Krebber, A. et al . Reliable cloning of functional antibody variable domains from hybridomas and spleen cell repertoires employing a reengineered phage display system. J. Immunol. Methods201, 35−55 (1997). | Article | PubMed | ISI | ChemPort |
Acknowledgments We are grateful to Roberto Marzari, Jianlong Lou, and Chonglin Yang for
helpful discussions; to Francesco Tedesco for RNA from human lymphocytes;
to Gabriella Rossi and Jessica Franzot for excellent technical help; and to
the following for the proteins they kindly provided for selection: Min Park
(Rad52), Scott Peterson (cyclin D, cdk2, cdc25A), Tom Peat (phosphoglycerate
dehydrogenase). We would also like to thank Brian Sauer for providing BS1365
as well as many useful discussions on Cre recombinase. This work was partly
financed by Fondo Trieste.
1011. A library created using this method was validated
by the selection of high affinity antibodies against a large number of different
protein antigens.
F at an MOI of 20:1. After
overnight growth, phagemid are made, reinfected into DH5
, V
, and V
genes *(see Experimental
Protocol for details). This primary library was used to infect bacteria expressing
Cre recombinase following the scheme illustrated in 
(lacZYA-argF)U169 deoR (
80
dlac
1. Phagemid used for selection were prepared from overnight
cultures of bacteria derived at this low MOI using standard techniques. When
making the library, 20 ml of Cre-expressing bacteria were used. These were
diluted into 400 ml to prepare phagemid, and 5
