Animal immunization

One male macaque was immunised with the A chain of ricin (RCA-A). After five RCA-A injections, the ELISA titre towards whole ricin as an antigen, was equal to 250 000.

Library construction and isolation of scFv specific for ricin

Isolation of anti-ricin scFvs

While the library was obtained from a macaque immunized with the RCA-A, it was panned with ricin since the whole toxin was our final target molecule. During panning, the number of eluted phages rose from 10⁵ (first round of panning), to 5 × 10⁵ (second round) and finally to 5 × 10⁶ phages (third and last round). This increase typically indicated enrichment in phages interacting specifically with the antigen. A phage-ELISA utilising ricin and RCA-A as antigens was performed to assess reactivity of the selected phages. Before panning, the signals on both antigens were at the background level, and both signals increased to five-fold over background after the first and the second rounds of panning. According to the phage technology, such signal increases correspond to enrichment in high-affinity binders. After a third round, the signal on RCA-A further increased as expected and it was 15-fold higher than the background, and 25-fold higher on ricin, which would be in agreement with the fact that whole ricin had been utilized for panning. The phagemidic DNA was extracted from the library after its third round of panning, and transformed into E. coli HB2151 [25]. Fifty clones were isolated and each monoclonal phagemidic DNA was extracted for sequencing. In parallel, these clones were induced for expression.

Ricin neutralization in cell-based assays, scFv stability

Of the 50 clones isolated after the panning, only 19 possessed full-sized, non-redundant sequences that were expressed for further testing in vitro. At the highest scFv concentration (1 μg/ml), clone 9RCA showed only 30% inhibition of cellular toxicity even though the molar ratio [scFv 9RCA]/[ricin] was in 150-fold excess. All other clones had neutralizing properties either equal, or less than, clone 9RCA. Exceptions included clones 43RCA, 38RCA, and 44RCA only. By using the cell-based assay, the IC₅₀ of clone 44RCA was 484 ± 10 ng/ml and that of clone 38RCA, 93 ± 5 ng/ml. The best clone, 43RCA, had an IC₅₀ equal to 23 ± 3 ng/ml (Figure 1), representing a molar ratio [43RCA]/[ricin] equal to 4. This clone was sufficiently expressed (around 100 μg/culture liter) to allow further study.

In order to determine scFv stability, the scFvs were incubated in culture media for 24 h (37°C). The reactivity of clones 43RCA, 44RCA, and 38RCA was not affected by incubation as assessed by ELISA. The stability of these three scFvs generally differed from other scFvs that showed less neutralization capacity, as the latter lost between 0 and 75% of their ability to bind toxin in an ELISA after they were exposed to the same conditions.

Ricin neutralization in cell-free assays

In order to appreciate whether anti-ricin scFv could function as an effective therapeutic against ricin intoxication, 43RCA was tested for its ability to neutralize ricin biological activity using a cell-free translation assay. The scFv 43RCA neutralized 89% of ricin activity at 40 μg/ml, and even concentrations of 2 μg/ml neutralized 60% of ricin activity when compared to controls containing ricin only (Figure 2). 1.5 μg/ml of 43RCA (50 nM) neutralized 50% of the activity of ricin, present at a 4 nM concentration in this assay, thus the corresponding molar ratio [43RCA]/[ricin] is equal to 12, in the same order of magnitude than for the cell-based assay (4). The activity of 43RCA was compared with the neutralizing capacity of anti-ricin IgGs purified by protein A from sera of mice hyper-immunised with the deglycosylated ricin A chain (anti-dgRCA), the most neutralizing antibodies available to us. By using the same concentrations, the anti-dgRCA IgGs showed ~20% neutralization at 5 μg/ml, and none at 2 μg/ml. The molecular weight of a scFv is 25 kDa with one paratope, while the molecular weight of an IgG is 150 kDa with two paratopes. As a consequence, the molecular weight of a paratope in the form of an IgG is three times larger than that of a scFv, and an IgG concentration of 5 μg/ml represents approximately the same molar paratope concentration as the scFv concentration at 2 μg/ml. At this same concentration of paratopes, the scFv neutralizes ricin three times more efficiently than the polyclonal IgG.

Affinity determination

Affinity measurements focused on the best neutralizing scFvs. The affinity of 43RCA against ricin was initially evaluated in standard conditions, using ~600 seconds elution steps, and resulting in an excellent affinity of approximately 40 pM. The K_D of 38RCA measured under the same conditions yielded 60 pM. In contrast, the affinity of 44RCA was 4.9 nM, or 100-fold less favourable, although considered high in absolute terms. Because 43RCA neutralisation activity was four times higher than 38RCA, we focused our work upon this more efficient scFv. A more precise measurement of 43RCA's affinity, performed by lengthening the elution time and by using a small scFv immobilization level (100 RU) (Figure 3), gave values of 40.8 pM after a 3 hr elution and 40.9 pM after a 4 hr elution. Measured after a 4 hr elution step, the K_on was equal to 3.34 10⁵ M ^-1 S ^-1 and the K_off, 1.36 10^-5 s^-1. The quality of these affinity measurements was assessed by the internal consistency test, which had numerical values of 0.56 (3 hr elution measurement) and 0.626 (4 hr elution measurement) when these values have to be inferior to 3 for the measurements to be valid.

Computational analysis

Of the 50 isolated clones, only 24 full-sized sequences coding for scFvs could be obtained. On these sequences, 3 were present in 2 occurrences (6 RCA and 37 RCA, 13 RCA and 32 RCA, 14 RCA and 44 RCA) and one was present in 3 occurrences (34 RCA, 41 RCA, 42RCA). Such repetitions may be informative since they could signal scFvs selected due to their high affinity, as was the case with 44RCA. The 19 non-redundant sequences coding the VH and VL domain were automatically analyzed using the IMGT/V-QUEST software. Human germline V, (D) or J alleles, found most similar to these 19 sequences, are shown on tables 2 and 3. In parallel, those sequences were analysed using a new online tool http://www.phylogeny.fr/, and this analysis was presented in the form of a phylogenetic tree (Figure 4). The initial root of the tree separated the scFvs in two sub-trees, corresponding to two clusters. Cluster A, corresponding to the sub-tree presented on the upper part of the figure, comprised three subgroups (scFvs 48 RCA, 11 RCA, 34 RCA, 9 RCA, 47 RCA; scFvs 13 RCA, 30 RCA; scFvs 37 RCA, 7 RCA, 15 RCA), whose light chains were of the λ isotype. The 13RCA and 30RCA pair differed only by two nucleotides (A/G41 located in the middle of FR1; and T/C116, the first residues of FR2) and probably resulted from the same parental IgG by somatic hypermutations, which did not lead to a significantly better neutralization effect. Beside their VLλ, 13 RCA and 30 RCA were not related to the rest of cluster A, and were removed from the remainder of the study. Analysis of the [V, (D) and J] genes arranged to code for the 6 remaining scFvs composing cluster A permitted an evaluation of its diversity. Typically, the VH of these scFvs were coded by a unique combination of a single IGHV gene (IGHV4-b) re-arranged with a single IGHD gene (IGHD2-8) and with a single IGHJ gene (IGHJ4) (tables 2 and 3), except for one occurrence in IGHV4-34 (scFv 9RCA) and two occurrences in IGHD1-7 (scFv 37RCA and 48RCA) combined with IGHJ5. The VL of these 6 scFvs were coded by IGLV1-36 and IGLV3-21 on three occurrences each, while another IGLV gene (IGLV1-47) was used twice. The IGLJ genes coding VL were rather constant, with IGLJ1 genes on six occurrences, in addition to one using of IGL3 and IGL6 genes. Overall, the 6 main scFvs in cluster A showed limited diversity and encompassed none of the three best clones. In contrast, cluster B, composed of 8 scFvs (scFvs 1, 23, 4, 43, 35, 44, 38, 31), contained more variability. Six different IGHV genes (1–69, 3–11, 3–23, 3–74, 4–28 and 4–34) were combined with 5 different IGHD genes (3–10, 1–1, 2–8, 6–13, 5–12) and two different IGHJ genes (4 and 5). Similarly, the 8 κ light chains of cluster B were coded by a combination of 7 IGKV genes (1–9, 1–16, 1–17, 1–18, 1–39, 1-NL1, 3–15) and 4 IGKJ genes (1, 2, 3, 4). Thus, detailed analysis of the genes re-arranged to code for the scFvs constituting clusters A and B showed that B had a higher diversity than A. This higher diversity of cluster B was also evidenced by the longer branches of the B subtree as compared to A, and this diversity very possibly allowed cluster B to encompass all the best clones. Such a difference of diversity between clusters is much more easily observed from the tree and subtrees than analysed as detailed above. The presentation of the diversity of antibody sequences in the form of trees could be of further interest when analyzing antibody fragments isolated from a library, even though these trees were not initially designed for this purpose.

	VH			VL
scFv	V	D	J	V	J
RCA7	IGHV4-b*01	IGHD2-8*02	IGHJ4*02	IGLV3-21*01	IGLJ1*01
RCA9	IGHV4-34*13	IGHD2-8*02	IGHJ4*02	IGLV1-36*01	IGLJ1*01
RCA11	IGHV4-b*01	IGHD2-8*02	IGHJ4*02	IGLV1-36*01	IGLJ1*01
RCA15	IGHV4-b*01	IGHD2-8*02	IGHJ4*02	IGLV3-21*02	IGLJ6*01
RCA34	IGHV4-b*01	IGHD2-8*02	IGHJ4*02	IGLV1-47*02	IGLJ1*01
RCA37	IGHV4-b*01	IGHD1-7*01	IGHJ5*02	IGLV3-21*01	IGLJ1*01
RCA47	IGHV4-b*01	IGHD2-8*02	IGHJ4*02	IGLV1-36*01	IGLJ1*01
RCA48	IGHV4-4*01	IGHD1-7*01	IGHJ5*02	IGLV1-47*02	IGLJ3*02
RCA13	IGHV3-11*01	IGHD3-22*01	IGHJ4*02	IGLV1-51*02	IGLJ1*01
RCA30	IGHV3-11*01	IGHD3-22*01	IGHJ4*02	IGLV1-51*02	IGLJ1*01

Human germline genes were retrieved by IMGT/V-QUEST analysis.

	VH			VL
scFv	V	D	J	V	J
RCA1	IGHV1-69*10	IGHD3-10*01	IGHJ4*02	IGKV1-NL1*01	IGKJ1*01
RCA-8	IGHV4-b*01	IGHD1-7*01	IGHJ5*02	IGKV1-18*01	IGKJ2*01
RCA4	IGHV3-23*01	IGHD3-10*01	IGHJ4*02	IGKV1-17*01	IGKJ4*01
RCA23	IGHV1-69*04	IGHD1-1*01	IGHJ4*02	IGKV3-15*01	IGKJ2*03
RCA31	IGHV4-34*13	IGHD2-8*02	IGHJ4*02	IGKV1-39*01	IGKJ2*01
RCA35	IGHV3-23*01	IGHD3-10*02	IGHJ5*02	IGKV1-9*01	IGKJ3*01
RCA38	IGHV4-28*01	IGHD6-13*01	IGHJ5*02	IGKV1-16*01	IGKJ1*01
RCA43	IGHV3-11*01	IGHD5-12*01	IGHJ4*02	IGKV1-17*01	IGKJ4*01
RCA44	IGHV3-74*01	IGHD3-10*02	IGHJ5*02	IGKV1-9*01	IGKJ3*01

Human germline genes were retrieved by IMGT/V-QUEST analysis.

The 43RCA scFv was the most efficient scFv for neutralization of ricin biological activity, and had very similar human germline counterparts as identified by IMGT/V-QUEST software. The overall identity between 43RCA eight framework regions and their most similar peptidic sequences, coded by human germinal genes, was 90%. In the IMGT Collier de Perles representation of 43RCA (Figure 5), black squares indicated amino acids that differ from the closest human germline genes IGHV3-11*01 and IGHJ4*02 for VH and IGKV1-17*01 and IGKJ4*01 for VL. By using IMGT amino acid classes [26], an analysis of the physicochemical properties of these differing 43RCA framework residues was performed and its results summarized in table 4. In particular, the proximity of 43RCA with sequences coded by human germline genes was further appreciated by the fact that only four residues, three located in VH (S40>D and Y55>R in FR2-IMGT, T122>V in FR4-IMGT) (human>macaque) and one in VL (T101 > V in FR3-IMGT) on a total of 180 framework residues, were rated as very dissimilar.

Animal immunization

A cynomolgus macaque (Macaca fascicularis) (6 kg) was subcutaneously injected with the non toxic A-chain of ricin (RCA-A) purchased from Sigma, Saint Louis, Missouri. The first injection was RCA-A (100 μg per injection) mixed with complete Freud's adjuvant (Sigma) and then RCA-A mixed with incomplete Freud adjuvant for the remaining injections.

Animal housing was performed in facilities accredited by veterinary services. Animals were provided with NHP feed, water ad libitum, and maintained on a 12-h light cycle. The protocol was approved by the CRSSA ethical committee for animal care and use.

Construction and screening of the anti-ricin antibody gene library

Bone marrow (5 ml) was sampled several times after the last boost in order to isolate RNA using Tri Reagent (Molecular Research Center Inc., Cincinnati, USA) according to the manufacturer's instructions. RNA was retro-amplified, then primers intended for the amplification of DNA encoding macaque Fd and VLκ were utilised as formerly exposed [22]. Later, a set of human λ primers (table 1) was utilised under the same PCR conditions.

PCR products were first cloned in the pGemT vector (Promega, Madison, Wiscontin), according to the manufacturer's instructions, yielding three antibody gene sub-libraries encoding the heavy (Fd fragment) or light (κ or λ) chains. The pGemT cloned PCR products were reamplified using two oligucleotide primer sets for introducing restriction sites for library cloning. A set of macaque κ oligonucleotide primers [21] or of human λ oligonucleotide primer (table 1) were used as forward (annealing to the 5' end of VH or VL) oligonucleotide primers. Only macaque-specific oligonucleotide primers were used as reverse oligonucleotide primers ([21] and table 1). The secondary PCRs were carried with separated forward oligonucleotide primers in order to maintain diversity. Each PCR was performed in 100 μl using 100 ng purified PCR reaction of the pGemT cloned cDNA, 4 U Red Taq (Sigma, Hamburg, Germany), 200 μM dNTPs each and 200 nM of each oligonucleotide primer for 20 cycles (30 s 94°C, 30 s 57°C, 30 s 72°C), followed by 10 min 72°C. The PCR products were separated by a 1.5% (w/v) agarose gel, then cut out and purified using Nucleospin Extract II Kit (Macherey-Nagel, Düren, Germany) according to the manufacturer's instructions.

Construction of the library was performed in two steps. First, the VLκ or VLλ PCR products were cloned to pHAL14 [21, 28, 29] and second, the VH PCR products were cloned to pHAL14 containing the VLκ or VLλ repertoire. The pHAL14 (5 μg), as well 2 μg VL (VLκ or VLλ), were digested using 50 U MluI and 50 U NotI (NEB, Frankfurt, Germany) in a 100 μl reaction volume (2 h at 37°C). The digest was inactivated by heating at 65°C for 10 min. Afterwards, 0.5 U calf intestinal phosphatase (MBI Fermentas) was added to the digested pHAL14 and incubated for an additional 30 min. This step was repeated once. The vector was purified using the Nucleospin Extract II Kit and 270 ng VL were cloned into 1 μg of the dephosporylated pHAL14 by incubating overnight (16°C) with 1 U ligase (Promega, Mannheim, Germany). The ligation reactions were precipitated with ethanol plus sodium acetate, and the pellet washed two times with 70% ethanol. These reactions were electroporated (1.7 kV) in 25 μl XL1-Blue MRF' (Stratagene, Amsterdam, Netherlands). The transformed bacteria were plated onto SOB agar plates (25 cm petri dishes) supplemented with 100 μg/mL ampicillin and 100 mM glucose. These colonies were harvested by suspending in 40 mL SOB media. Plasmids from the VLκ or VLλ library were isolated using the Nucleobond Plasmid Midi Kit (Macherey-Nagel) according to the manufacturer's instructions. Five micrograms of each VL chain library, as well 1.5 μg of the VH fragments, were digested using 50 U NcoI and 100 U HindIII (NEB) in a 100 μl reaction volume (overnight at 37°C). The following steps were performed as described for VL with the following modification: 250 ng of the digested and purified VH repertoire was used for ligation. In total 2 transformations were performed. The harvested bacteria, containing the final antibody gene libraries, were aliquoted and stored at -80°C.

For library packaging, 500 mL 2xTY medium [30] containing 100 μg/mL ampicillin and 100 mM glucose were inoculated with a 1 mL aliquot of the antibody gene library. Bacteria were grown (37°C, 250 rpm rotary shaker incubator) to an O.D.₆₀₀ of 0.4 – 0.5. Twenty-five millilitres of bacteria (~1,25 × 10¹⁰) were infected with 2,5 × 10¹¹ Hyperphage [31–33], incubated at 37°C for 30 min without shaking, followed by 30 min at 250 rpm. The infected cells were harvested by centrifugation (10 min, 3,220 × g). The pellet was resuspended in 200 mL 2xTY containing 100 μg/mL ampicillin and 50 μg/mL kanamycin. The phages were produced at 30°C and 250 rpm for 16 h. Cells were pelleted by centrifugation for 10 min (10,000 × g). The supernatant containing the phages was precipitated with 1/5 volume of 20% (w/v) polyethylene glycol (PEG)/2.5 M NaCl solution (1 h on ice) with gentle shaking and then pelleted by centrifugation for 1 h at 10,000 × g (4°C). The precipitated phage were resuspended in 10 mL phage dilution buffer (10 mM TrisHCl pH7.5, 20 mM NaCl, 2 mM EDTA) and filtered through a 0.45 μM filter. Phage precipitation was repeated once more. The phage pellet was diluted in 1 mL phage dilution buffer and cell debris was pelleted by additional centrifugation (5 min. at 15,400 × g, 20°C). The supernatant containing the phage were stored at 4°C. Each library packaging was controlled by tittering, and subsequently tested by immunoblot according to [28, 34].

Screening of the library was performed as described elsewhere [35, 36], except that 10, 15, and 30 washes were performed for each successive round of panning. Ricin was utilized as the antigen and PBS-Tween 20 0.1% as the washing buffer.

scFv production, ELISA testing

Phagemid DNA isolated after the panning process was used to transform the non-suppressor E. coli strain HB2151 such that it expressed the soluble scFv fragment. Single colonies of randomly chosen transformants were used to inoculate 5 ml of SB (Super Broth) medium supplemented with carbenicillin (50 μg.ml^-1) and 1% glucose. In parallel (see below), the clones were sequenced so that redundant clones were not re-analyzed. For expression, cultures were incubated overnight (30°C) with vigorous shaking (250 rpm). Five hundred milliliters of SB medium supplemented with carbenicillin and 0.1% glucose were then inoculated with 500 μl of each culture. The cultures were grown at 30°C until the 600 nm optical density (OD) reached a value of one. In order to induce gene expression, IPTG (1 mM) was added and the cultures incubated overnight (22°C). The cells were harvested by centrifugation at 2500 × g for 15 min, 4°C. The scFvs were extracted with polymyxin B sulphate [37] and purified using a nickel column (Ni-NTA spin column, Qiagen, Valencia, CA) according to the manufacturer's instructions.

For ELISAs, RCA-A or ricin (Sigma, 10 μg/ml PBS) was coated onto a 96 well microtitre plate (Maxisorp, Nunc, Danemark). When sera were tested, the negative control was pre-immune serum and secondary antibody reporter utilized polyclonal anti-macaque IgG, Fc specific tagged with HRP (Nordimmune, Tilburg, The Netherlands). The antibody titre was measured as the reciprocal of the highest dilution of immune serum giving a signal three times stronger than the negative control, at the same dilution. When scFv were tested, they were detected by an anti-his tag antibody (Qiagen, Courtaboeuf, France).

Ricin neutralization in cell-based assays

Soluble scFvs, eluted after the library panning and expressed individually, were tested for their neutralization capacity using a cell-based viability assay. In this assay, cells are put in contact with ricin, in such conditions that all cells die. The scFv to be tested is added, at different concentrations, to ricin and may inhibit its toxic activity. At the end of the assay, cell viability is assessed and plotted against scFv concentration. This assay replicates the two main steps of ricin inhibition (internalization of the scFv bound to ricin and inhibition of its toxic activity). More precisely here, J774A.1 cells (ATCC-LGC, Molsheim, France) were plated at a density of 14,000 cells/well (200 μl/well), cultured (37°C with 5% CO₂) for 24 hours in DMEM supplemented with 10% SVF. Ricin, at a concentration of 15 ng/ml (corresponding to 10 LD₅₀, data not shown), was pre-incubated with scFvs, or with control serum for 1 h (37°C) and then added to the cells. After a 24 h incubation (37°C, 5% CO2), cell viability was determined by using the Cytotox 96 Non-radioactive Cytotoxicity Assay (Promega, Madison, Wi), as suggested by the manufacturer. Each assay was corrected for 100% cell viability (control wells with no toxin and no scFv) and 0% viability (control wells with ricin and no scFv). The scFv concentration corresponding to 50% viability is defined as the IC₅₀ (inhibitory concentration 50%). These assays were performed three times in triplicate. All experiments were conducted on J774A.1 cells sub-cultured less than 15 times after receiving them from the vendor.

ScFv stability

The scFv stability was estimated by determining the percentage of scFv active after a 24 h incubation at 37°C in DMEM medium. ScFvs (50 μg/ml in PBS) were incubated in triplicate and then tested in an ELISA, utilizing a freshly thawed aliquot for control.

Ricin neutralization in a cell-free translation assay

In this assay, a cell-free translation system is put in contact with ricin, resulting in translation inhibition. The scFv is added, at different concentrations, to ricin and may inhibit its toxic activity. At the end of the assay, translation activity is plotted against scFv concentration. This assay replicates the intra-cellular step of ricin inhibition. More precisely here, the neutralizing activity of the best scFv was determined using a microtitre cell-free translation assay that detects luciferase translation from luciferase m-RNA [38]. By using phosphate buffered saline (PBS), antibodies were diluted in a 96 well microtitre plate and a constant amount of ricin (4 nM final concentration) was added to the antibody dilutions. Dilutions of ricin alone were included in each assay for generation of a standard. In addition, anti-ricin IgGs, purified from serum of mice that had been hyperimmunized with the dgRCA-A to result in antibodies with a high neutralization activity of ricin, were utilized as a benchmark of neutralization. The plate was placed on a microplate shaker for 15 min (25°C), and then 5 μl transferred to a v-shaped microtitre plate. The rabbit reticulocyte lysate, RNasin, amino acids complete, and luciferase m-RNA (Promega, Madison, WI) were mixed together and 25 μl added to each well of the v-shaped plate. Plates were incubated (37°C) for 90 min. Five microliters of the translation lysate were transferred to a black microtitre plate (Sigma Chemicals, St. Louis, MO). After the addition of 45 μl reaction buffer (Luciferase assay reagent, Promega, Inc.), luminescence was measured as counts per second (CPS) on a Victor multiplate reader (PerkinElmer Wallac, Boston MA). Data was expressed as the percent of control [% control = (CPS treated/CPS control) × 100].

Affinity measurements

Affinities were measured by surface plasmon resonance (SPR) with a BIAcore™ × (Biacore, Uppsala, Sweden) instrument. The scFv 43RCA was immobilized at a maximum of 100 RU on a CM5 chip (Biacore) via amine coupling, according to the manufacturer's instructions. A 30 μl/min flow rate was maintained during measurement. For each measurement, eight ricin dilutions were prepared in HBS-EP buffer (Biacore) (concentrations ranging from 1 to 250 nM) and were tested at various elution times from 600 sec to 14,500 sec. For the precise measurement of very high affinities, we were advised (C. Quetard, Biacore, France) to use long elution times with the highest tested ricin concentration. After each ricin dilution, the chip was regenerated with 1.5 M glycine buffer (Biacore), run 10 μl/min for 30 seconds. Constants were calculated [39], and verified with internal consistency tests [40] as formerly described.

Nucleic acid analysis of ricin-specific scFv clones

In parallel to expression, transformed HB2151 bacteria were cultivated to isolate the phagemidic DNA (Nucleobond AX, Macherey-Nagel). Variable region sequences of the light and heavy chains were determined by Genome Express (Meylan, France) using the primers Mkmyc and MkpelB, respectively [41]. The sequences were analyzed on line, using the International ImMunoGeneTics information system ^® (IMGT) [42]http://imgt.cines.fr and its nomenclature. In particular, these sequences were compared with those of the human germline immunoglobulin genes by using the IMGT/V-QUEST [43] and IMGT/JunctionAnalysis softwares [44]. Peptidic sequence comparisons utilized new IMGT tools [26], which classify and compare three separate items; hydropathy, volume, and chemical properties, for each amino acid. This three item analysis was summarized as follows: when no significant difference between two residues was recognised in the three items, residues were regarded as highly similar. When one difference was found, the residues were regarded as similar but then regarded as dissimilar in the case of two differences, and very dissimilar in the case of three differences.

In addition, the similarities between all analysed scFv sequences were analysed using a new on-line tool http://www.phylogeny.fr/, which builds phylogenetic trees. Phylogenetics in the strict sense apply to immunoglobulins because, due to somatic hypermutations that are the support for affinity maturation, unmutated rearranged germline sequences are ancestors of those coding for high affinity antibodies. After library panning however, only high affinity, mutated antibody fragments are retained. Thus, in this study, we utilised this tool as a multiple alignment software that displays results in the form of phylogenetic trees or sub-trees, and applied it to the sequences coding scFvs selected from the library. In such trees, each node, where certain branches of a tree meet, indicates the consensus shared by the different sequences displayed at the extremity of corresponding branches. The length of each branch, between each node and each sequence, is in proportion with the differences between the corresponding consensus and sequence.

Nucleotide sequences

The Macaca fascicularis 43RCA-H and 43RCA-L sequences (VH and VL domain sequences, respectively) are accessible in the data banks under the accession numbers [EMBL: FJ178346 and EMBL: FJ178347].