Harnessing Gene Conversion in Chicken B Cells to Create a Human Antibody Sequence Repertoire

Transgenic chickens expressing human sequence antibodies would be a powerful tool to access human targets and epitopes that have been intractable in mammalian hosts because of tolerance to conserved proteins. To foster the development of the chicken platform, it is beneficial to validate transgene constructs using a rapid, cell culture-based method prior to generating fully transgenic birds. We describe a method for the expression of human immunoglobulin variable regions in the chicken DT40 B cell line and the further diversification of these genes by gene conversion. Chicken VL and VH loci were knocked out in DT40 cells and replaced with human VK and VH genes. To achieve gene conversion of human genes in chicken B cells, synthetic human pseudogene arrays were inserted upstream of the functional human VK and VH regions. Proper expression of chimeric IgM comprised of human variable regions and chicken constant regions is shown. Most importantly, sequencing of DT40 genetic variants confirmed that the human pseudogene arrays contributed to the generation of diversity through gene conversion at both the Igl and Igh loci. These data show that engineered pseudogene arrays produce a diverse pool of human antibody sequences in chicken B cells, and suggest that these constructs will express a functional repertoire of chimeric antibodies in transgenic chickens.


Introduction
Monoclonal antibodies (mAB) are an important pillar in the treatment of multiple disorders such as cancer, inflammatory diseases, and orphan diseases [1][2][3]. With the development of hybridoma technology, it became possible to produce mAB in mice [4]. Because of their murine origin, however, these antibodies are immunogenic in humans [5,6]. To reduce immunogenicity, chimeric antibodies, humanized antibodies and fully human antibodies from phage display libraries were created using recombinant DNA techniques [7][8][9][10]. Another attempt to solve this problem was to create transgenic animals carrying human immunoglobulin loci in order to produce human sequence antibodies directly without further manipulation [11][12][13][14][15]. The animal-based approaches are all limited by the fact that some antigens, especially human tumor antigens, are not well recognized in mammals because of the close evolutionary relationship to humans. To date, all of the transgenic animals producing human antibodies have been mammalian species, but a non-mammalian host such as chicken would access a much wider set of epitopes, since chickens have not shared a common ancestor with humans in at least 300 million years. The complex genetic modifications necessary to produce human antibodies in chickens (knockout of endogenous immunoglobulins and insertion of human transgenes) can be accomplished in cultured primordial germ cells, leading to the creation of fully transgenic birds. [16,17].
The chicken B cell line DT40 expresses a normal surface IgM receptor and continues to diversify its immunoglobulin loci by the process of gene conversion, a type of homologous recombination [18]. Gene conversion generates sequence diversity in the functional light and heavy chain variable regions by using upstream pseudogenes as the sequence donors in a templatedriven, unidirectional process to mutate the single rearranged V region in each locus [19]. Wild type DT40 cells have been used to generate antigen-specific antibodies from the endogenous immunoglobulin loci in vitro but the variable regions remained chicken sequence [20][21][22]. The ability of DT40 cells to promote gene conversion has been applied to exogenous genes such as GFP, which was inserted into the immunoglobulin light chain locus [23,24]. The application of gene conversion to exogenous genes requires that the gene of interest be inserted in an immunoglobulin locus, as the gene conversion machinery preferentially acts at these loci over other loci [25][26][27], and it requires that pseudogenes be present to serve as sequence donors. Although the DT40 gene conversion machinery could be used directly for the diversification of human immunoglobulin variable regions that could be used in antibody discovery programs, we believe an in vivo immune system with affinity maturation will generate higher affinity antibodies with higher efficiency [28]. However, DT40 cells can still serve an important role in validating transgene constructs prior to insertion into transgenic chickens.
Here, we demonstrate production of a repertoire of human V region sequences by gene conversion, using a DT40 cell line with a double knock out of the chicken immunoglobulin light (Igl) and heavy (Igh) chain loci. The double knock-out cell line was used for a targeted, site-specific integration of functional human V K and V H genes, thereby replacing the chicken functional V L and V H genes (Igl huVK , Igh huVH ). In order to diversify the inserted huV K and huV H by gene conversion, we included synthetic human sequence pseudogene arrays upstream of the functional huV K and huV H . By using a stop-codon reversion assay we show that Igl huVK , Igh huVH DT40 cells diversified the functional human heavy and light chain genes by gene conversion, suggesting that these transgenes, when inserted into fully transgenic chickens, will create a diverse repertoire of human antibodies in B cells in vivo. This demonstration of simultaneous in vitro molecular evolution of two genes in the same cell line can be generalized to provide a method for creating Figure 1. Knockout of the chicken V L and V H genes. Chicken V L and V H loci are shown and the homology regions of the knockout constructs flanking the deleted region are marked by dotted lines. Targeting constructs for the V L and V H in chicken DT40 cells are shown. a) For the V L single knockout, the knockout construct consists of a b-actin-eGFP (eGFP), a CAG-puromycin (puro) selectable marker cassette and the 59 and 39 homology regions. To create the double knockout of V L and V H , an alternative targeting construct for V L has a b-actin-hygromycin (hygro) selectable marker cassette. The homology regions are the same for both V L knockout constructs. b) For the V H knockout, the targeting construct consists of a 59 and 39 homology region, a b-actin-eGFP and a CAG-puromycin selectable marker cassette. In all constructs, the selectable markers are flanked by the HS4 insulator from chicken beta-globin (H), and loxP sites [39]. Downstream of the 39 loxP site a promoterless neomycin (neo) or blasticidin (Bsr) gene in opposite orientation together with an attP site for targeted insertion is included. c) Knockout of the V L and V H was detected with gene specific primers on the 59 side as well as with primers in the deleted region (black arrow heads: VL 59, primers 1 and 2; VL deletion region, primers 3 and 4; VH 59, primers 5 and 2; VH deleted region, primers 6 and 7). b-actin served as a quality control for the genomic DNA. Scale bar equals 1 kb. doi:10.1371/journal.pone.0080108.g001 libraries of proteins whose sequence are defined by the pseudogene arrays.

Transfection of DT40 Cells
For each transfection, 5610 6 cells were collected, pelleted and resuspended in V-buffer (Lonza Walkersville Inc., Walkersville, USA) with 10 mg DNA of each construct for a total volume of 100 ml. The cell-DNA suspension was transferred to a 2 mm cuvette and subjected to 8 square wave pulses of 350 V/125 msec (BTX 830 electroporator). The transfected cells were resuspended in medium and plated into a 96 well plate. Selection with either 0.5 mg/ml puromycin, 5 mg/ml neomycin, 100 mg/ml blasticidin, or 2 mg/ml hygromycin was started 24 hours following transfection. As soon as single colonies were identified they were transferred to new wells and expanded for further analysis. as well as huV K insertion (Igl huVK ), huV H insertion (Igh huVH ) and huV K -huV H (Igl huVK , Igh huVH ) double insertion cell lines were analyzed for expression and signaling of immunoglobulin receptors. a) 1610 6 cells were lysed and immunoglobulin heavy chain expression determined by Western blotting using goat-anti-chicken-IgM-AP and immunoglobulin light chain expression by rabbit-anti-chicken-IgY-AP. Mouse-anti-b-actin followed by goat-anti-mouse-AP was used to detect bactin. b) The cell lines from above and Igl huVK cells with a stop codon (Igl huVK-Stop ) in CDR1 cultured for four weeks were stained with mouse-antichicken-IgM followed by goat-anti-mouse-Ig-Cy5. All cell lines except wild type express eGFP from the selectable marker cassette used in the knockouts. Fluorescence signal was visualized using a Beckman Coulter FC-500. c) 1610 6 wild type DT40 cells, non-green Igl KO , Igh KO cells and nongreen Igl huVK , Igh huVH DT40 cells were labeled with FLUO-4-AM and incubated with 10 mg/ml goat-anti-chicken-IgM starting from the time point indicated by arrows. The change in fluorescence intensity was measured for a total of 300 sec using a Beckton Dickinson LSRII Fortessa. One of three representative experiments is shown. d) The same cell lines (DT40 wt grey, Igl KO , Igh KO red, Igl huVK , Igh huVH blue) were stained with goat-anti-humankappa-RPE. Fluorescence was measured using a Beckman Coulter FC-500. doi:10.1371/journal.pone.0080108.g002

Construction of the Puromycin/Neomycin/eGFP Selectable Marker Cassette
The enhanced green fluorescent protein (eGFP) driven by the chicken b-actin promoter was cloned with the puromycinresistance gene driven by the CAG promoter. The eGFP-puro cassette was flanked by two sets of duplicated HS4 insulators. A 46 bp attP site (GTGCCCCAACTGGGGTAACCTTT-GAGTTCTCTCAGTTGGGGGCGTA) was cloned upstream of a promoterless neomycin resistance gene using annealed oligos, and a 34 bp loxP site (ATAACTTCGTATAGCATACATTA-TACGAAGTTAT) was inserted downstream of the neo gene using annealed oligos. The attP-neo-loxP was then inserted into the eGFP-puro cassette. A second loxP site was added at the 59 end of the cassette by ligating annealed oligos to restriction enzyme-digested plasmid. A sequence from HA (AAGCG-TAATCTGGAACATCGTATGTA) was also included as a primer binding site for the PCR genotyping assay.

Construction of the Hygromycin/Blasticidin Selectable Marker Cassette
A hygromycin resistance gene driven by the chicken b-actin promoter was obtained from H. Arakawa (plasmid 257) and cloned in between two sets of duplicated HS4 insulators from the chicken b-globin locus. The promoterless blasticidin resistance gene (Bsr) was amplified from a Bsr-containing plasmid (597, a gift from H. Arakawa) using primers that included a 51 bp attP site 59 of the Bsr gene, and a 34 bp loxP site at the 39 end. (Primers: 59-TTACGTAGTGCCCCAACTGGGGTAACCTTT-GAGTTCTCTCAGTTGGGGGCGTAGGTCATTTTTGCA-GAAATCGGAGGAAG-39 and 59-CCATGCATA-TAACTTCGTATAGCATACATTATACGAAGTTATG-GATCCAGACATGATAAGATACA-39).
A second loxP site was added at the 59 end of the cassette by ligating annealed oligos to restriction enzyme-digested plasmid. The HA sequence from above was also included as a primerbinding site for the PCR genotyping assay in transfected cells. Integrase-mediated insertion of human V L and V H to replace the corresponding chicken genes. The Igl KO and Igh KO cell lines, as well as the Igl KO , Igh KO line, were used to insert human V K or V H into the chicken loci. Co-transfection of WC31 integrase and the shown a) huV K or b) huV H constructs resulted in recombination of the attP and attB sites leading to an insertion of the human V constructs and creating attL and attR sites. A b-actin promoter was integrated in front of the neomycin (neo) or blasticidin (Bsr) gene, and cells were selected with the indicated drug for stable integration of the huV K or huV H . In the case of huV K insertion into the chicken Igl KO single knockout, the selectable marker cassette was the same as for the heavy chain locus shown in b). c) To test for proper integration of the human genes, genomic DNA was isolated and a constructspecific PCR was performed with the indicated primers (primers are indicated by black arrowheads with primer orientation: VK+VH 59, primers 12 and 9; VK 59 Bsr, primers 8 and 9; VK 39, primers 10 and 11; VH 39, primers 13 and 14). Primers are placed on the 59 and 39 side of the integration showing the correct integration versus the knock out or wild type DT40 cell lines. b-actin was used as a quality control for the genomic DNA. Scale bar equals 1 kb. doi:10.1371/journal.pone.0080108.g003

Construction of the Chicken V L Targeting Constructs
The 59 homology region (1 kb) was amplified from chicken genomic DNA using primers: 59-GCCCATCACTCAGGGAGGAGA-39 59-TAGCTCTGAAGTCTCCATCCT-39 The unique restriction sites AscI and NheI were used to clone the 59 homology region upstream of the selectable marker cassette The 39 homology region (7.2 kb) was amplified from chicken genomic DNA in two steps, and the products were combined by ligation via a naturally occurring BssSI restriction site. The 59 and 39-most primers delineating the 39 homology region are: 59-CTGAACTAGTGCTGACTCTGCA-39 59-GCCGCACGTTGCACAGCTGT-39 The unique restriction sites SpeI and NotI were used to clone the 39 homology region downstream of the selectable marker cassette.

Construction of the Chicken V H Targeting Construct
The 59 homology region (1.1 kb) was amplified from DT40 genomic DNA using primers: 59-CTCAGAGCCCCTAATAAGTG-39 59-TCTGCGCTGAGTTCTTTGA-39 The unique restriction sites AscI and NheI were used to clone the 59 homology region upstream of the selectable marker cassette.
The 39 homology region (4.3 kb) was amplified from DT40 genomic DNA using primers: 59-TGGCGGTGTAGGGGAAAATGTC-39 59-AGCCCCTAATAACCGTAAT -39 The unique restriction sites SpeI and NotI were used to clone the 39 homology region downstream of the selectable marker cassette.

Knock Out of Chicken V L and V H
To create the Igl KO cell line with a knockout of the chicken V L , DT40 cells (see above) were transfected with the EGFP-puro containing construct shown in Fig. 1a and single colonies were selected for puromycin resistance and eGFP expression. To create the Igh KO cell line with a knockout of the chicken V H , cells were transfected with the targeting construct shown in Fig. 1b and single colonies were selected for puromycin resistance and eGFP expression. To create the Igl KO , Igh KO cell line with a double knockout of V L and V H , a stable pool of DT40 Igh KO cells was transfected with the alternative V L targeting construct containing the hygro gene shown in Fig. 1a and single colonies were selected for hygromycin/puromycin resistance and eGFP expression. Clonal populations were expanded after every selection and genomic DNA was isolated using DNeasy Blood&Tissue Kit (Qiagen, Valencia, USA). FIREPol PCR Mix 7.5 (Solis BioDyne, Tartu, Estonia) was used according to the manufacturer's protocol

Selection of Functional huV K and huV H
The functional human V K and V H genes were selected after screening a small library of genes from the V K 3 and V H 3 gene families cloned from human B cell DNA for a V K /V H pair that would express at high levels in HEK293 cells. A V K 3-15 gene and a V H 3-23 gene were chosen as the functional variable regions for expression in DT40. The sequences of the V K and V H functional genes served as the starting template for the design of the human pseudogenes.

Construction of the SynVK Insertion Constructs
For the SynVK-12 pseudogene array, huV K pseudogenes were designed based on the DNA sequence encoding the mature huV K 3-15 polypeptide from the signal peptide cleavage site to the CDR3 region without any JK sequence. Framework regions (FW) and CDRs were defined using the IMGT domain system [30]. Each position in CDR1, CDR2 and the first 6 positions of CDR3 were individually substituted with either Tyrosine (Y) or Tryptophan (W), and the FW are identical to the functional huV K gene. Each pseudogene contains at most one amino acid change in each of its three CDRs.
For the SynVK-C array, the sequence of the functional huV K gene was altered to incorporate AID hotspots throughout the V region (nucleotides WRC/GYW) without changing the amino acids encoded. The pseudogenes, where homologous to the functional huV K , contained the same changes to maintain the homology. For the pseudogene array, variable region sequences were obtained from the NCBI human EST database and a diverse set of CDR1, CDR2 and CDR3 sequences were selected. FW diversity was also included in some of the pseudogenes, while the remaining pseudogenes contained FW identical to the functional V K . FW and CDRs from different ESTs were assembled together into the pseudogenes, such that each pseudogene could be a combination of 3-6 EST sequences. Spacer sequences (100 bp between each pseudogene) were derived from the chicken YV L pseudogene locus. The pseudogenes were synthesized (BioBasic, Markham, Ontario, Canada) as individual genes and ligated together sequentially into the array. The pseudogenes were all in a direct repeat orientation (except for the most distal pseudogene in SynVK-12), and in the opposite orientation relative to the functional huV K . Plasmids containing the huV K pseudogene arrays were propagated in HB101 cells (MCLab, South San Francisco, USA) for stability.
The remainder of the light chain locus in the SynVK insertion vectors was derived from the chicken light chain by PCR. The chIgL promoter, J-C intron, and Cl constant region were assembled with the rearranged human V K gene and the human leader exon. The chicken leader intron was included in the gene synthesis of the human V K gene. The human pseudogene array was cloned upstream of the 2.4 kb promoter fragment, putting the most proximal pseudogene in a position nearly identical to that in the endogenous light chain locus. For insertion into the attP site at the knockout locus, an attB site (either 45 or 70 bp) was included adjacent to the chicken b-actin promoter that will drive expression of the promoterless blasticidin or neomycin genes in the knockout. The HS4 insulator was included upstream of the b-actin promoter.

Construction of the SynVH Insertion Constructs
The SynVH pseudogenes were designed based on the functional huV H gene, and extend from the start of the mature protein to the CDR3 sequence, without any JH sequence. For the SynVH-B pseudogene array, all FW were identical to the functional huV H , and the CDRs contained individual substitutions of Y, W or Serine (S) residues. Each CDR in each pseudogene has a single amino acid substitution relative to the functional huV H . Several pseudogenes contained CDRs identical to the functional V H because of naturally occurring Y, W or S residues in those positions. Spacer sequences between pseudogenes (100 bp) were obtained from the chicken YV H pseudogene locus. The pseudogenes were synthesized and cloned into an array by sequential rounds of ligation.
The non-coding sequences (2.1 kb chicken V H promoter, and 275 bp of the chicken J-C intron) were amplified from chicken genomic DNA and assembled with the rearranged human V H region and human leader exon. The chicken V H leader intron was included in the gene synthesis of the human functional V H . The SynVH-B array was then cloned upstream of the chV H promoter, with the pseudogenes all in opposite orientation to the functional huV H . The b-actin-attB cassette for selection of insertions was the same as in the SynVK insertion vectors. For the SynVH-A7 array, the pseudogenes have the same basic structure as those in the SynVH-B array, with sequences extending from the start of the mature protein coding sequence to the end of CDR3, and similar spacer sequences between the pseudogenes. Every CDR is composed entirely of varying combinations of Y, W and S residues. Whereas the CDRs in the SynVH-B array had single substitutions in each CDR, the CDRs in the SynVH-A7 array had all positions substituted. The FW are identical to the functional huV H , and the remainder of the construct is identical to the SynVH-B insertion vector.

Integrase-mediated Insertion of huV K and huV H
For insertion of the huV K and huV H constructs, Igl KO and Igh KO single knockout cell lines (see above) were transfected as described above with 10 mg WC31 integrase and 10 mg of the huV K -SynVK-12, huV K -SynV K -C, huV H -SynV H -B or huV H -SynV H -A7 construct. 24 hours after transfection G418 selection was started.
Integration of the huV K and huV H constructs was tested by PCR on genomic DNA samples from the transfected cells. To verify the integration on the 59 side for all of the single insertion constructs, the primers 59-CTCTGCTAACCATGTTCATGCCTTC-39 (primer #9) and 59-AGTGACAACGTCGAGCACAGCT-39 (primer #12) were used. The integration on the 39 side was confirmed for the huV K using the primers 59-CGCACACGTA-TAACATCCATGAA-39 (primer #10) and 59-GTGTGAGATG-CAGACAGCACGC-39 (primer #11) and for the huV H 59-TTTGACTACTGGGGCCTGG-39 (primer #13) and 59-GCCCAAAATGGCCCCAAAAC-39 (primer #14). To generate a cell line expressing both huV K and huV H , Igl KO , IgH KO double knockout cells (see above) were first transfected with 10 mg huV H -SynV H -B and 10 mg WC31 integrase. After selection for neomycin/puromycin/hygromycin resistance and eGFP expression, the integration of the huV H -SynV H -B construct was confirmed by the same PCR used for the single insertion of the huV H . These cells were then transfected with 10 mg WC31 integrase and 10 mg of the huV K -SynV K -12 or huV K -SynV K -C construct. 24 hours after transfection cells were selected for neomycin/puromycin/hygromycin/blasticidin resistance and eGFP expression. The resulting cell lines were tested for the integration of the huV K construct and re-confirmed for the integration of the huV H -SynV H -B construct as above. Because the selectable marker cassette used for the huV K construct differs from the one used for the single integration of the huV K a different primer set was needed to verify the 59 integration: 59-AATTGCCGCTCCCACATGATG-39 (primer #8) and 59-CTCTGCTAACCATGTTCATGCCTTC-39 (primer #9).

Generating eGFP Negative Cell Lines
In order to create the eGFP-negative Igl KO , Igh KO and Igl huVK ,Igh huVH cell lines, cells were transfected with 10 mg pCAG-Cre (gift from Connie Cepko, Harvard Medical School, Boston, USA) by electroporation. Cells were monitored for the loss of eGFP expression over several days following the transfection and as soon as non-green cells were observed, 50 cells were plated on a 96 well plate to obtain single cell clones. Non-green clones were identified after days and were expanded for further study. Flow Cytometry Analysis 1610 6 cells were pelleted and washed with cold labeling buffer (PBS containing 0.5% bovine serum albumin). Cells were resuspended in 500 ml cold labeling buffer containing a monoclonal anti-chicken-IgM antibody or polyclonal anti human-kappa- Figure 8. Creating huV K diversity by gene conversion using a human light chain pseudogene array with naturally occurring CDRs, AID optimization and framework changes. The chicken V L was replaced by a huV K including 16 human synthetic pseudogenes containing naturally occurring CDR sequences from an EST database (NCBI), the SynVK-C array. In addition the pseudogenes and the functional V L were AID optimized. Some of the pseudogenes also differ in their framework regions compared to the functional huV K . The inserted huV K had a stop codon and a HpaI restriction enzyme site in CDR1 or CDR3 so that there was no IgM expression unless the stop codon was repaired due to gene conversion. RPE antibody (Southern Biotech, Birmingham, USA). Cells were incubated for 30 min in the dark at 4uC. After washing two times with cold labeling buffer the anti-chicken-IgM antibody was detected by a polyclonal anti-mouse-IgG antibody (Bethyl Laboratories Inc, Montgomery, USA) conjugated to Cy5. After incubation for another 30 mins, the cells were washed twice with cold labeling buffer and resuspended in labeling buffer. Fluorescence was measured using a Beckman-Coulter FC500.  Ca 2+ Signaling 1610 6 cells were washed with labeling buffer (calcium/ magnesium free PBS containing 2% fetal bovine serum and 1 g/L glucose) at room temperature. Cells were then resuspended in labeling buffer and 0.5 mM FLUO-4 (Life Technologies, Carlsbad, USA) and incubated for 20 min in the dark at room temperature. Afterwards cells were washed three times with the labeling buffer and resuspended in 500 ml labeling buffer. Background fluorescence was measured for 20 sec on a Beckton Dickinson LSRII, Fortessa followed by cross-linking the immunoglobulin receptor with a polyclonal anti-chicken-IgM antibody at a final concentration of 10 mg/ml. Change in fluorescence was measured for additional 220 sec.

Amplification and Cloning of Gene-converted huV K and huV H Regions
To select for gene-converted sequences by PCR, a reversion assay was designed. A HpaI restriction site, containing a stop codon, was included in CDR1 or CDR3 of the transcribed huV K and huV H transgenes. These constructs were otherwise identical to those described above. DT40 cells were tranfected with these constructs as described above, resistant clones were confirmed by PCR, and the huV K and huV H sequences were analyzed after growing these cells at least four weeks to allow accumulation of mutations. Genomic DNA was isolated using the DNeasy Blood&Tissue Kit (Qiagen, Valencia, USA) according to the manufacturer's protocol. Genomic DNA was digested with HpaI over night. After digestion, the huV K region was amplified using primers 59-ACTGCACCGGAGAAATTGTCTTG-39 and 59-GAGACGAGGTCAGCGACTCAC-39. The huV H region was amplified using primers 59-GACGTGCAGTTGGTGGAGTC-39 and 59-GTTGAAGACTCACCTGAGGAGACGG-39. The PCR was performed using GoTaq (Promega, Madison, USA) according to the manufacturer's protocol. The PCR product was again digested with HpaI and the remaining product was gel purified using the Ultra Sep Gel Extraction Kit (Omega Bio-tek, Norcross, USA) according to the manufacturer's protocol. The purified product was cloned into pCR2.1 using the TOPO/TA cloning kit (Life Technologies, Carlsbad, USA). 96 w plasmid preparations were performed using the Wizard SV 96 Plasmid Purification System (Promega, Madison, USA). Isolated DNA was sent for sequencing (MCLab, South San Francisco, USA).

Analyzing Gene Conversion Events
A reference database of all human CDRs and framework regions used in the pseudogenes in the different arrays was created. Sequences from the amplified huV regions were aligned using Lasergene (DNAStar Inc., Madison, USA) and exported to a FASTA database. These two databases were compared by using CD-HIT [31,32] to identify full-length gene conversion events. All shorter events, point mutations and deletions were verified manually.

Knock Out of Chicken V L and V H in the Chicken B Cell Line DT40 Results in Loss of Surface IgM Expression
To replace the rearranged chicken V L and V H in DT40 cells with functional huV K and huV H , the chicken genes must first be knocked out. Targeting constructs for the V L and the V H were designed ( Fig. 1a and b). The constructs consisted of a selectable marker cassette and 59 and 39 homology regions. The homology regions for the V L targeting constructs were 1.1 kb on the 59 side and 7.2 kb on the 39 side (Fig. 1a). For the V H targeting construct, the homology regions were 1.1 kb on the 59 side and 4.3 kb on the 39 side (Fig. 1b). To generate cell lines with a single knockout of either Igl or Igh, the selectable markers consisted of a puromycin resistance gene and an eGFP gene. For double knock-out Igl KO , Igh KO cell lines, a hygromycin resistance cassette was used in the V L targeting construct. No eGFP was included since the V H targeting construct already expresses eGFP (Fig. 1b). The selectable marker cassettes also included an attP site for later site-specific integration using WC31 integrase of an insertion construct carrying an attB site. Linked to the attP site was a promoterless drug-resistance gene, either neomycin or blasticidin, in opposite orientation. By inserting a promoter into the attP site in front of this drugresistance gene, clones with the correct insertion can be selected with the appropriate drug selective media. All cassettes were flanked by loxP sites to loop out the selectable markers by treatment with Cre recombinase. After transfection with the single targeting constructs, clonal populations were selected and analyzed for the correct targeting by PCR (Fig. 1c). Primers binding in the deleted region can also bind on the unrearranged allele but since the PCR extension time was kept short no PCR product was observed. The targeting efficiency for both the chicken V L and V H was ,15%. The chV H knock-out cell line (Igh KO ) was transfected with the alternative V L targeting construct to obtain a double knock out of V L and V H (Igl KO , Igh KO ) in the same cell line. After selecting these cells with puromycin/ hygromycin, the double knock out was confirmed by PCR (Fig. 1c). The targeting efficiency here was ,10%.
Cell lysates of the knock-out cells were prepared and analyzed by Western blot. The chicken Cm was detected by a polyclonal anti-chicken-IgM antibody, and the chicken immunoglobulin light chain (IgL) was detected by a polyclonal anti-chicken-IgY antibody. Since DT40 cells express IgM and are not able to class switch, this antibody only detects IgL in DT40 cells. Wild-type DT40 cells show a double band at the expected molecular weight for the heavy chain at 74 kDa and for the light chain at 22 kDa. The Igl KO did not show any band with the IgY antibody but still shows a single band at 74 kDa for heavy chain. The Igh KO cells express chicken V L but not V H . In the Igl KO , Igh KO cell line, the bands for heavy chain and light chain are absent (Fig. 2a). Cell surface staining of the immunoglobulin receptor with a monoclonal anti-chicken-Cm antibody confirmed that the Igl KO , the Igh KO and the Igl KO , Igh KO cell lines are completely negative for surface IgM (Fig. 2b top row).

Restoration of Surface IgM Expression by Insertion of Human V K and V H in Chicken Igl KO and Igh KO DT40 Cell Lines
Human V K and V H genes were inserted into the chicken V L and V H loci to produce chimeric antibodies consisting of human V regions and chicken constant regions. Since the constant regions are critical for a functional B cell receptor-signaling complex and for downstream effector functions in vivo, they must remain chicken sequence for optimal function. Insertion constructs included a human pseudogene array comprised of 7 to 17 designed pseudogenes. These are referred to as pseudogenes because they lack promoters and splice sites and are designed for gene conversion, not because they are related to the endogenous human VK and VH pseudogenes found in the human genome. Downstream of the pseudogene array, the huV K constructs had a rearranged, functional human kappa variable region (consisting of V and J segments), the chicken J-C intron and the chicken lambda constant region gene followed by an attB site and the chicken bactin promoter (Fig. 3a). The constructs for the huV H insertion also had a pseudogene array and a rearranged, functional human V H gene (consisting of V, D and J segments) followed by the attB site and a b-actin promoter (Fig. 3b). WC31 integrase-mediated recombination of the attB site in the inserting constructs and the attP site in the selectable marker cassette led to targeted integration into the immunoglobulin loci and placement of the b-actin promoter upstream of the neomycin or blasticidin resistance markers. Only cells with a targeted insertion into the immunoglobulin locus could be selected with neomycin or blasticidin.
To confirm the correct insertion on the 59 side, PCR primers spanning from the b-actin promoter of the insertion vector into the neomycin or blasticidin resistance genes were used. The PCR from b-actin to neomycin gave a single band for the huV K insertion into the chicken Igl KO , the huV H into the chicken Igh KO , and the huV H insertion into the chicken Igl KO , Igh KO (shown as VK+VH 59 in Fig. 3c). PCR from b-actin into blasticidin gave a band only for the huV K insertion into the Igl KO , Igh KO because blasticidin was used as a selectable marker (Fig. 3c).
Integration on the 39 side was also confirmed by PCR. Primers for the V L detected the non-rearranged wild-type allele in DT40 as well as the rearranged modified allele. Because of the resulting attR site after recombination of attP and attB, a double band was seen for correctly integrated clones (Fig. 3c). Correct integration of the huV H on the 39 side was confirmed with primers spanning from the huV H into the chicken J-C intron (Fig. 3c).
Restoration of IgM expression in Igl huVK and Igh huVH single insertions and double Igl huVK , Igh huVH DT40 cell lines was analyzed. Lysates of PCR-positive clones were collected and Cm or IgL were detected by Western Blot (Fig. 2a). The single huV K insertion gave a strong double band at the expected molecular weight for light chain, comparable to wild type DT40 cells. The single huV H insertion showed a single band around 74 kDa expected for heavy chain. The double insertion gave a band that is fainter than the wild type band at 75 kDa. The IgL blot shows a strong band for the huV K integration at 25 kDa, a strong double band comparable to the wild type for the huV H integration and a faint double band at the same size as the wild-type bands for the huV K -huV H insertion.
Surface IgM expression was analyzed by flow cytometry using a monoclonal anti-chicken-Cm antibody. All cells with a huV K and wild type chicken heavy chain, as well as all cells with the double insertion Igl huVK , Igh huVH , showed surface IgM expression similar to wild type DT40 cells. In contrast, only a subpopulation (30.7%) of the cells with a huV H insertion and wild type chicken light chain showed surface IgM expression (Fig. 2b bottom row). The Igl huVK , Igh huVH cells were also stained with a polyclonal anti-human-kappa antibody. Production of human V regions was supported by the shift of the Igl huVK , Igh huVH cells compared to wild type cells and Igl KO , Igh KO cells (Fig. 2d).
A Human-chicken Chimeric IgM Receptor on Chicken DT40 Cells is able to Release a Calcium Signal after Crosslinking EGFP-positive Igl KO , Igh KO cells, as well as Igl huVK , Igh huVH cells, were transiently transfected with CAG-Cre to remove the selectable marker cassettes by Cre-lox recombination. After 72 h 35% to 45% eGFP-negative cells were observed. Limiting dilution was performed and clonal non-green populations of double knockout cells and double insertion cells were grown and analyzed for calcium signaling. Cells were labeled with FLUO-4 calciumdependent dye and IgM surface receptors were cross-linked using a polyclonal anti-chicken-IgM antibody. After cross-linking, an increase in fluorescence was measured for DT40 wild-type and Igl huVK , Igh huVH cells by flow cytometry. The increase in fluorescence was higher for Igl huVK , Igh huVH cells compared to DT40 wild type cells. By contrast, the addition of polyclonal anti-chicken IgM antibody to Igl KO , Igh KO knock out cells did not change fluorescence (Fig. 2c).

Creation of Diversity in huV K and huV H Sequences by Gene Conversion in DT40 Cells
Human pseudogene arrays containing synthetically-designed V regions (SynV genes) were integrated upstream of the functional huV K and huV H in DT40 cells. Two different types of SynV pseudogenes were designed. In the SynVK-12 and SynVH-B arrays, each of the pseudogenes contained complementarity determining regions (CDRs) with individual tyrosine or tryptophan substitutions at every position in each CDR (Figs. 4 and 5). This array design was based on evidence that minimalist CDRs consisting of appropriately spaced tyrosine and tryptophan residues are capable of producing high affinity antigen-specific antibodies [33,34]. In contrast, the SynVK-C pseudogenes consisted of naturally occurring human CDR sequences from expressed, full-length V regions (Fig. 6), which provides a larger pool of diverse sequences with which to mutate the functional V. For some of the SynVK-C pseudogenes, the framework regions between the CDRs were different from the frameworks used in the functional huV K , in order to provide further sequence diversity.
To select for gene-converted sequences by PCR, a reversion assay was designed. A HpaI restriction site, containing a stop codon, was included in CDR1 or CDR3 of the transcribed huV K and huV H transgenes. These constructs were identical to those used for the expression analysis in Fig. 2 except for the HpaI site and stop codon. DT40 cells were tranfected with these constructs, resistant clones were confirmed by PCR, and the huV K and huV H sequences were analyzed after growing these cells at least four weeks to allow accumulation of mutations. Genomic DNA was isolated and digested with HpaI. Only huV mutations that eliminated the HpaI site (either by gene conversion or other type of mutation) could then serve as template in the PCR reaction. To ensure that intact, unmutated huV amplicons were not cloned, the PCR product was digested again with HpaI, cloned and sequenced. In most cases we used the unsorted bulk culture to analyze all of the Hpa-resistant sequences, but in the case of the SynVK-12 array with a stop in CDR1 (Fig. 4a) we first sorted the cells based on surface IgM expression (Fig. 2b, Igl huVK-Stop ).
In the SynVK-12 construct, most pseudogenes were used frequently for gene conversion of CDR1, except for pseudogenes 3, 9 and 12, which were only used once or not at all (Fig. 7). There was a bias towards pseudogene 1, which is the most proximal to the functional huVK (Fig. 7b). Only two long gene conversion events extending from CDR1 to CDR2 were observed (Fig. 7a). For CDR3, a high frequency of point mutations was observed: 42 point mutations out of 62 sequences analyzed for the SynVK-12 array (Fig. 7c). Only pseudogenes 4, 5, 7 and 10 were used for gene conversion of CDR3 (Fig. 7d). Also, some gene conversion events for CDR1 and CDR2 independent from the selection for CDR3 were detected (Fig. 7c).
In the SynVK-C construct, the sequence of the functional huVK was modified to include hotspots for activation-induced cytidine deaminase (AID), and the pseudogenes were also similarly modified, to retain sequence homology with the functional V. Most SynVK-C pseudogenes were used for gene conversion of CDR1 except for pseudogenes 21, 26, 29 and 30 (Fig. 8b). Long events from CDR1 to CDR2 were observed for the SynVK-C array (Fig. 8a). For CDR3, 78 out of 96 sequences showed point mutations and no recognizable gene conversion events. In 18 of the sequences, gene conversion was detected for CDR3 with the SynVK-C array, yet in every case it came from pseudogene 23 (Fig. 8c).
Analysis of gene conversion of the huV H gene in DT40 cells revealed that the SynVH-B array was used very efficiently for gene conversion in CDR1 (Fig. 9). All pseudogenes except 21 and 31 were used for gene conversion in CDR1. An apparent bias toward pseudogene 27 was observed (Fig. 9b). Long events going from CDR1 to CDR2 were detected (Fig. 9a). Analysis of the sequences for CDR3 showed only one sequence pattern (Fig. 9c). It was not possible to determine the origin of this gene conversion event since all pseudogenes except 22 and 30 have the same sequence at the beginning of CDR3 at the position where the HpaI site is located. Point mutations were also observed.
To analyze if it is possible to generate more diversity for CDR3 in huV H , an array with seven pseudogenes called SynVH-A7 was used. Every amino acid position in each CDR in the SynVH-A7 array was substituted with serine, tyrosine or tryptophan (Fig. 10). Analyzing gene conversion events from this array in CDR3 showed 23 partial gene conversion events replacing the HpaI site, out of 91 sequences. 4 different pseudogenes out of seven were used (Fig. 11).

Discussion
Here we describe a novel cell-based system to diversify exogenous genes of interest, and applied it to the diversification of human variable region genes. The system was used to validate the constructs that will be inserted into a line of transgenic chickens expressing chimeric immunoglobulins. A two-step approach to replace the endogenous chicken V regions with human V regions was taken: first, knockouts of the chV L and chV H , both individually and together, were created and analyzed, followed by site-directed insertion of functional human Vs in conjunction with synthetic human pseudogene arrays. This is the first description of a knockout of the functional heavy chain V in chicken cells.
The targeting vectors used to knock out the chicken Ig loci inserted selectable marker cassettes and attP sites into the chicken Igl and Igh loci. By co-transfection with wC31 integrase and an insertion cassette containing an attB site, complex constructs can now be efficiently inserted into the Ig loci [16,35]. The Igl KO , Igh KO knockout cell line thus represents a ''universal acceptor'' that can be harnessed to diversify any given exogenous gene or pair of genes by gene conversion [24]. The design of the pseudogene arrays dictates how genes can be diversified: which domains or regions of the target genes are to be mutated, and which residues are available to the gene conversion process with which to mutate them.
While it was possible to restore surface IgM expression by expressing huV K in combination with the chV H in the Igl huVK cell line, the converse did not lead to surface IgM expression on every cell. Western blot analysis suggested that the huV H and chV L do not pair efficiently, leading to protein degradation in the cells. Efficient pairing was restored by inserting both huV K and huV H into the same cells. Since the chicken constant regions are the same in all the insertion cell lines, the difference in the functional V enabled the assembled IgM receptor to express on the surface at levels indistinguishable from wild-type DT40 cells. The difference in the molecular weight of the chimeric heavy chain in the double insertion cell line compared to wild type chicken heavy chain may be explained by a different glycosylation pattern of the huV H compared to the chV H . After cross-linking the surface IgM with a polyclonal anti-chicken-IgM antibody a strong calcium signal was detected indicating that the human-V-chicken-constant-region chimeric receptor assembled correctly and was functional. The calcium release observed in Igl huVK , Igh huVH cells was stronger than the signal detected in wild-type cells, as shown by the higher proportion of cells that shift upon stimulation. One possible explanation for this difference is that we have inadvertently selected a high-signaling variant after 5 rounds of clonal selection to produce the Igl huVK , Igh huVH cells (2 knockouts, 2 insertions, and limiting dilution after Cre transfection).
Rates of productive gene conversion that repaired the HpaI/ stop codon and restored surface IgM were between 1 and 3% for the huV K and the huV H with a stop codon in CDR1, and approximately 0.5% for the stop codon in CDR3, after culturing the cells for four weeks (Fig. 2b and data not shown). Gene conversion of endogenous V genes occurs at a rate of 1% in DT40 cells [21], consistent with the rates seen for our transgenes. Observed frequencies in vivo appear much higher, with 3-7 gene conversion events found in each sequence analyzed from bursal B cells at 3 weeks of age [36], suggesting that we will obtain much higher frequencies in our transgenes in vivo. Analysis of gene conversion in surface IgM-restored SynVK-12 cells showed that two pseudogenes were rarely used for CDR1, indicating that the sequences in those pseudogenes were not compatible with expression (Fig. 7b). All other analysis was performed on unsorted cells to capture all of the potential sequence diversity regardless of expressability. The observed biases in pseudogene usage in unsorted cells cannot be easily explained. Neither proximity to the functional V nor sequence homology correlate with frequency. Sequence analysis of the huV K and huV H showed that the lower frequency of gene conversion observed in CDR3 is consistent with previous data published for the chicken V L in DT40 cells [37]. In contrast to endogenous chicken B cells [36], DT40 cells show a strong bias towards point mutations and short gene conversion events in CDR3 [18].
The frequencies of point mutations in CDR3 for most of the arrays were correspondingly high: the SynVK-12 and SynVK-C arrays showed point mutations of the huV K CDR3 at a rate of 67% and 81% of the sequences analyzed, respectively, and the SynVH-A7 array showed a rate of 63% for huV H . In contrast, the SynVH-B array showed point mutations at a lower rate of 16%. This might be due to the design of CDR3 in the pseudogenes of the SynVH-B array. CDR3 of the SynVH-B pseudogenes 21,22, 29, 30 and 33 is identical with the sequence downstream of the HpaI site in the functional huV H (Fig. 5), and this sequence homology may have increased the gene conversion frequency with a concomitant reduction in point mutation frequency in the sequences analyzed. Reynaud et al. have shown that for CDR3, the preferred pseudogenes were those that display homology on the 39 side towards CDR3 in the functional V [36].
For the purpose of increasing gene conversion efficiency, the sequence of the functional V K for the SynVK-C array and the SynVK-C pseudogenes was AID optimized. Without changing the protein sequence, the nucleotide sequence was changed where possible to WGCW [38]. The analysis of the CDR1 HpaI-stop codon restriction site reversion assay showed a significant increase towards long gene conversion events compared to the SynVK-12 array without AID optimization. With the SynVK-C array, 24% of the gene conversion events for CDR1 were long events, whereas for SynVK-12, only 1% of the gene conversion events for CDR1 were long events extending to CDR2. Interestingly, the enhanced gene conversion rate was not seen for SynVK-C in CDR3.
Taken together, these data demonstrate that the diversification machinery of DT40 cells can be harnessed to accomplish in vitro molecular evolution. It will be interesting to compare sequences obtained from in vivo studies with these same human transgenes to determine the extent to which gene conversion and somatic hypermutation diversify the human V regions both in the preimmune repertoire and in the sequences of antigen-specific antibodies. It remains to be seen whether the level of diversity introduced by gene conversion and point mutations in our transgenes will lead to antibodies of high affinity in transgenic birds. The lack of multiple gene families and combinatorial diversity of these human transgenes should not be a limitation, since normal chickens produce a highly diverse pre-immune repertoire from single functional V regions by gene conversion. The method provides the ability to circumscribe the boundaries of the canonical variants through the design of the pseudogene array. These results illustrate the utility of DT40 cells for testing and validating human immunoglobulin constructs that will be used to create a transgenic chicken platform for the discovery and development of novel human antibody therapeutics.