Introduction

In comparison to mouse and man, lymphoid differentiation and somatic recombination in several domestic animals, including rabbit, chicken and sheep, yield a relatively small B-cell pool with a restricted antibody repertoire.1 In these latter species, fetal and young animal gut-associated lymphoid tissue hosts an actively proliferating population of B cells, which generates the peripheral B-cell pool and the final primary (pre-immune response) antibody repertoire.2, 3 In ruminants, the amplification of naive B cells and primary repertoire expansion is thought to take place in the ileal Peyer’s patch (IPP).4, 5 Somatic hypermutation (SHM) and gene conversion (GC) have been suggested to expand the limited antibody repertoire in ruminants after recombination but before encountering external antigens,6, 7, 8, 9 although evidence for GC in ruminants is not as strong as in rabbit10, 11 or chicken.12, 13

SHM and GC are well-characterized as secondary diversification processes mediated by activation-induced cytidine deaminase (AID) upon encountering an external antigen. AID attacks single-stranded DNA during transcription and deaminates cytosines to uracils.14 AID preferentially targets C residues in the third position in WRCY hotspot motifs (W=A/T, R=A/G, Y=C/T).15, 16, 17, 18 If the resulting uracil is not recognized and removed by base excision or mismatch repair pathways, it may be used as a template during replication. The outcome is often a mutation regardless of the mechanism.19, 20, 21, 22 Mutations caused by AID activity are targeted to a region extending about 2 kb downstream from the Ig promoter, including the VDJ-coding exon and the J-region intron.23 In SHM, U:G mismatches recruit error-prone DNA repair machinery, which increases the mutation rate by up to a million fold, as recently reviewed by Storck et al.24 The number of mutations is significantly higher in the complementarity-determining regions (CDRs) than in the framework regions (FRs). Mutations in the CDRs enable the generation of high-affinity antibodies.23, 25

To our knowledge, this is the first study in fetal ruminants that directly links AID to IPP or to primary antibody formation. mRNA expression of bovine AID and its mutagenic activity has been analyzed in neonatal and adult lymphoid tissues, although not in IPP.9 In this paper, we demonstrate AID expression in bovine fetal IPP and also in fetal spleen before exposure to exogenous antigens. The mutational profiles of immunoglobulin heavy chain variable (IGHV) genes in fetus are compared with those generated by antigen-induced AID-mediated SHM.

Results

AID is expressed in bovine fetal IPP and spleen

The expression of AID mRNA was analyzed by reverse transcription-quantitative PCR (RT-qPCR) in the fetal ileum, spleen, lymph node and bone marrow, with the muscle as a negative control. The AID expression level [as mean of glyceraldehyde 3-phosphate dehydrogenase (GAPDH)-normalized cycle threshold values±s.d.] was 11.0±3.8 in the ileum (n=21), 9.9±3.3 in the spleen (n=20), 12.8±2.3 in the lymph node (n=16), 16.2±1.5 in the bone marrow (n=14), and 16.5±3.7 in the muscle (n=10). AID expression in the ileum and spleen differed significantly from the expression in the bone marrow and muscle, as assessed using nonparametric Kruskal–Wallis analysis of variance (χ2=35.5, P<2 × 10−16) followed by multiple comparisons using Nemenyi–Damico–Wolfe–Dunn test at α=0.01. Looking at the lymphoid tissues at various stages of development, no statistically significant differences in AID expression were observed in any tissue, apart from the high level of expression in the adult lymph node (Figure 1).

Figure 1
figure 1

Activation-induced cytidine deaminase (AID) mRNA expression analyzed by reverse transcription-quantitative PCR. Negatives of GADPH (glyceraldehyde 3-phosphate dehydrogenase)-normalized cycle threshold values (–ΔCt, mean±s.d.) are shown in ileum, spleen, bone marrow and lymph node for three fetal stages and for adults. The dashed lines represent mean±s.d. of AID expression in the muscle used as the negative control (N=12). gd, gestational days.

PowerPoint slide

Full size image

In the ileum, immunohistochemistry showed strong positive staining for AID in the IPP B-cell follicles (Figure 2a,b). The follicles were AID positive from the stage they become detectable at approximately 195 gestation days (gd). In the fetal lymph nodes, AID-positive follicles were observed only occasionally (Figure 2c,d). In the fetal spleen and bone marrow, the distribution of AID protein could not be reliably analyzed due to non-specific binding of the primary antibody to non-lymphoid cells. Adult lymph node was used as a positive control and displayed the characteristic germinal center staining pattern (not shown).26

Figure 2
figure 2

Activation-induced cytidine deaminase (AID) expression and B lymphocyte proliferation in fetal bovine lymphoid tissues. (a) AID expression in the ileum of a 238–gestational days (gd) fetus. (b) Higher magnification of the area indicated by the box. (c) AID expression in the lymph node of a 250–gd fetus. (d) Higher magnification of the area indicated by the box. (e–g) Double immunofluorescence for the B-cell marker CD79α (green) and the proliferation marker Ki67 antigen (red) in the ileum of a 260–gd fetus (e), in a lymph node of a 250–gd fetus (f), and in the spleen of a 227–gd fetus (g). Bars=100 μm.

PowerPoint slide

Full size image

To assess the level of B-cell production in the various lymphoid tissues, we also performed double immunofluorescence staining for the B-cell marker CD79α+ and for the proliferation marker Ki67 antigen (Figure 2e–g). CD79α+ B cells proliferate more actively in the fetal IPP (median 75% Ki67+; s.d. 9.4%; n=7) than in the lymph nodes (cortical follicles; 9.0±6.7%; n=7; P=0.002) and spleen (15±10%; n=9; P=0.002).

Fetal IGHV cDNAs show signs of hypermutation and selection, but the overall mutational load is lower than in postnatal animals

In order to characterize the mutation patterns in fetal IGHV sequences, we constructed IGHV cDNA libraries from the ileum, spleen, and bone marrow of two fetuses (240 and 270 gd). To compare the naive fetal antibody repertoire to that after secondary antibody responses, we also generated cDNA libraries from the ileum of a 51-day-old calf and from a peripheral lymph node of an immunized calf. The germline reference sequences were obtained from the available bovine genome sequencing data27 and by sequencing the functional IGHV genes in four additional animals to a total of 35 functional IGHV genes. The fetal cDNAs aligned to the FR1–FR3 region of the germline sequences, with an average pairwise identity of 99.6%. For the 51-day calf and the immunized calf, the average identities were 97.4% and 92.4%.

Mutation pattern characteristic of SHM was observed in cDNAs derived from the fetal IPP and spleen as well as from the postnatal samples, although fetal sequences contained fewer mutations (Figure 3 and Table 1). The total mutation load was 5–9 mutations per kilobase in fetal samples from the IPP and spleen, 26 per kilobase in the IPP of the 51-day calf, and 80 per kilobase in the lymph node of the 2-year-old immunized calf. The ratio of mutations in CDRs vs. FRs was 5.6 in fetuses (average of all ileal and splenic samples), 6.7 in the 51-day calf, and 3.4 in the immunized calf. The ratio of replacement over silent mutations was 3.3 in fetuses, 3.1 in the 51-day calf, and 2.1 in the immunized calf. In the fetal bone marrow, few mutations were observed, and the mutations were not concentrated in the CDR regions (Figure 3 and Table 1).

Figure 3
figure 3

Mutational loads in framework (FR) and complementarity-determining (CDR) regions of immunoglobulin heavy chain variable cDNA sequences from fetal and calf lymphoid tissues. Bars represent mutations per 1 kb FR or CDR sequence analyzed. No CDR mutations were observed in the bone marrow of the 270-day fetus. For numbers of cDNA sequences analyzed, see Table 1.

PowerPoint slide

Full size image
Table 1 Replacement and silent mutations in FR1-FR3 regions of bovine IGHV sequences
Full size table

Negative selection had occurred in FRs, whereas positive selection had occurred in the CDRs in both the fetus and calf data sets, based on statistical tests developed specifically to estimate selection in Ig sequences (Table 2).28, 29 In the bone marrow, selection could not be reliably analyzed owing to the small numbers of mutations.

Table 2 Assessing selection in bovine IGHV sequences
Full size table

Fetal mutations correlate with AID expression and are targeted to well-characterized AID hotspot motifs

In the fetal tissues analyzed (ileum, spleen, and bone marrow at 240 and 270 gd), AID expression as measured by RT-qPCR correlated significantly and positively with the overall mutational load (r=0.922, P=0.027, n=6) and specifically with the numbers of CDR mutations (r=0.946, P=0.016) and missense mutations (r=0.917, P=0.027). AID expression also correlated with silent mutations (r=0.845, P=0.034), which are not affected by selection on the protein level.

AID is known to preferentially target WRCY and its complementary RGYW motif in the genome.17, 30 In particular, Cs at the third position of WRCY and the corresponding Gs at the second position of RGYW are targeted. We searched for these hotspot motifs at the FR1–FR3 region and analyzed mutations within the motifs. As shown in Figure 4, the target G/C nucleotides in RGYW/WRCY motifs were preferentially mutated as compared with G/C nucleotides outside the motifs in both fetal and postnatal calf sequences.

Figure 4
figure 4

Targeting of mutations to activation-induced cytidine deaminase (AID) hotspot motifs (RGYW/WRCY) in immunoglobulin heavy chain variable cDNA sequences from fetal and calf lymphoid tissues. Numbers of mutated AID target G/C nucleotides and numbers of mutated G/C nucleotides outside AID target motifs are shown per 1,000 target and non-motif G/C nucleotides analyzed. A total of 11,039 target G nucleotides, 10,931 target C nucleotides, 63,078 non-target G nucleotides, and 68,486 non-target C nucleotides were analyzed.

PowerPoint slide

Full size image

Discussion

In ruminants, the fetal IPP is characterized by SHM of Ig genes in the absence of an external antigen.6 To our knowledge, no molecular studies addressing the mechanism have been reported. We studied the role of AID in this process. AID is an obligatory factor in all known mechanisms that shape the antibody repertoire in response to antigenic challenge, e.g., SHM, GC, and class switch recombination.31, 32, 33

We report that AID is strongly expressed in the fetal bovine IPP and spleen (Figures 1 and 2). The mutational profile in IGHV cDNA sequences derived from these tissues is similar to the profile generated by AID-mediated SHM in secondary antibody response. Mutations are preferentially located in CDRs (Figure 3), generally resulting in an amino-acid change (Table 1), and are enriched in the AID hotspot motifs (Figure 4). By contrast, AID expression is low in the fetal bone marrow, correlating with small numbers of IGHV mutations. The data are highly suggestive of AID-dependent SHM in the fetal bovine IPP and spleen before exposure to exogenous antigens and imply a specific molecular mechanism to diversify the limited germline Ig repertoire.27, 34

Statistical analysis indicated strong selection of the mutated IGHV sequences in the fetal IPP and spleen (Table 2). Negative selection on the FRs can be explained by the adverse effects replacement mutations have on protein structure. The CDR sequences were positively selected, although the fetus is sterile and there is no appreciable external antigen present to account for the selective forces. In sequences derived from the fetal IPP and spleen, mutations were ∼5.6 times as frequent in CDRs as in FRs. The RGYW/WRCY preference of AID explains at most about 34% of this difference, as the frequency of these motifs in CDRs was ∼1.9 times that in FRs. The remainder can be attributed to selection, although other components of the mutation machinery may also contribute, such as the WA/TW mutator DNA polymerase η recruited by AID-induced U:G mismatches.17, 30, 35 The positive selection may be driven by unknown endogenous antigen(s). Even if not antigen driven, the products of unmutated IGHV genes may be unable to provide for proliferative signals as efficiently as those encoded by genes carrying favorable mutations in their CDR-encoding regions.

Unknown IGHV germline sequences potentially present in the analyzed animals or technical sequence errors did not likely affect our conclusions. To complement the limited information on the structure of the bovine Ig heavy chain locus, we recently characterized the bovine IGHV repertoire based on the available genome sequencing data.27 In addition, we now cloned and sequenced functional IGHV genes in individual animals analyzed in this work. To improve the accuracy of cDNA data sets, short or low quality cDNA sequences were excluded from the analyses. The combined error rate for PCR and sequencing, based on analysis of the vector-derived sequences, was 0.784 per kilobase or about 5% of the average observed mutational load. We also tested for the generation of chimeric IGHV sequences in the PCR reactions, using a 1:1 mixture of two known IGHV clones. We cloned and sequenced 109 PCR products out of which only one clone was chimeric, indicating that the polymerase used was highly processive.

The bovine genome contains a limited number of functional Ig genes, which are restricted to few families or subgroups.27, 34 Antibody-diversifying mechanisms in addition to recombination are therefore likely to exist. Previous studies have suggested that these may include SHM,6, 36 GC,7 and junctional diversity.37 Our study provides extensive molecular and mutational data suggesting that AID-associated SHM and selection are taking place in the IPP and spleen of bovine fetuses before exposure to external antigens. These processes are apparently spatially separated from de novo B lymphopoiesis and recombination, which occur in the fetal bovine bone marrow and lymph nodes,38, 39 as AID expression and/or the numbers of IGHV mutations are low in these tissues.

Using hematopoietic chimeras, we have previously shown that peripheral B lymphocytes in young cattle are generated from a self-sufficient pool of developing B cells independent of bone marrow hematopoietic stem cells.40 In these animals, the ratio of donor cells among circulating B lymphocytes stably reflected that in the IPP. These observations and our current data suggest two phases of B-cell generation: first, differentiation in the bone marrow and lymph nodes; second, migration to the IPP for additional diversification of the limited germline repertoire by SHM. In addition to IPP, SHM in cattle fetuses occurs in the spleen, and occasional AID-positive follicles were also observed in the fetal lymph nodes (Figure 2c,d). The proliferation rate of B cells is markedly higher in the IPP than in the other lymphoid tissues (Figure 2e–g),41 whereas the total number of lymphocytes in IPP is roughly comparable with the spleen and lymph nodes.42, 43 Thus, our observations are compatible with previous studies indicating that IPP produces the majority of the diversified peripheral B-cell pool.44

Methods

RNA extraction. Tissue samples from 15 bovine fetuses (190–290 gd) were collected from a local abattoir, snap frozen in liquid nitrogen, and conserved at −80 °C. Total RNA was extracted from the ileum, spleen, lymph node, bone marrow, and muscle using Eurozol (EuroClone, Milano, Italy) according to the manufacturer’s instructions. Briefly, 100–300 mg of frozen tissue was ground in a pre-cooled mortar, suspended in 1 ml Eurozol, and homogenized using Polytron PT1200 homogenizer (Kinematica AB, Lucerne, Switzerland). In all, 2.5 M lithium chloride was used to precipitate RNA followed by resuspension in RNase-free water. Remaining genomic DNA was removed with DNase treatment (Turbo DNase, Life Technologies, Carlsbad, CA) according to the manufacturer’s instructions. The RNA concentration was determined on Nanodrop spectrophotometer (Thermo Scientific, Asheville, NC) and the quality was assessed using RNA 6000 Nano kit on Agilent 2100 Bioanalyser (Agilent Technologies, Santa Clara, CA).

RT-qPCR. First-strand cDNA was synthesized from 1 μg of DNase-treated total RNA using RevertAid M-MuLV Reverse Transcriptase (Fermentas, St Leon-Rot, Germany). cDNAs were prepared as previously described.38 The cDNA mixture was diluted 1:2 in RNase-free water and 2 μl of this was used as a template in the qPCR reaction.

qPCR was performed using the SYBR green technology. The primers bAIDfw3, bAIDrev2 and QGAPDfw, QGAPDrev (Table 3) were designed on the basis of known sequences of bovine AID and bovine GAPDH (GeneBank accession number NM_001038682.1 (AID) and NM_001034034.1 (GAPDH)). Each reaction contained 1 × SYBR Premix Ex Taq mixture (TaKaRa, Otsu, Japan), 0.2 μM of forward and reverse primers and 1 × ROX II Reference Dye in a final volume of 20 μl. PCR conditions were adjusted with serial dilutions (107–101) of plasmid templates confirming that the PCR efficiency was satisfactory (90–110%). Templates used in these standard curves were cloned into pBluescript II SK (AID) vector (Fermentas) and pSTBlue-1 (GAPDH) vector (Life Technologies). The cDNA was transcribed from total RNA obtained from a young calf. The cloning primers were Aicda 5′ and Aicda 3′ yielding a 651 bp product (AID) and QGAPDfw and QGAPDrev yielding a 145 bp product (GAPDH; Table 3). Amplicons from the templates and cDNA samples were sequenced. Amplification was carried out using Stratagene’s Mx3005P real-time PCR system (Agilent Technologies). Cycling conditions were: 95 °C for 10 s, followed by 40 cycles of 95 °C for 5 s, 60 °C for 35 s and 80 °C for 15 s. The fluorescence signal was collected at 80 °C, with two replicates per each sample. The dissociation curves of the final products were analyzed with following conditions: 95 °C for 60 s, 55 °C for 30 s and 95 °C for 30 s, where the data were collected during all points from 55 °C to 95 °C.

Table 3 PCR primers
Full size table

Cycle threshold (Ct) values were determined with MxPro software based on a threshold line that was automatically defined above the non-informative data. To control the possible variation in the amount of cDNA in different samples, the level of gene expression of the target genes was normalized in relation to the reference gene GAPDH.

Immunostaining. AID immunohistochemistry was performed using the anti-human AID monoclonal antibody EK2-5G9 (Ascenion, Munich, Germany).26 The immunogenic peptide sequence of the anti-AID antibody is 100% identical to the bovine AID, and it produced an expected staining pattern in the adult lymph nodes used as positive controls. Paraffin sections of paraformaldehyde-fixed tissues were subjected to heat-induced antigen retrieval in 10 mM Tris-HCl pH 9.5, 1 mM EDTA pH 8.0, and blocked for endogenous peroxidase and biotin. The primary antibody was detected using a biotinylated goat anti-rat Ig secondary antibody (Jackson Immunoresearch, Suffolk, UK) and tyramide signal amplification as described previously.45

Double immunofluorescence for CD79α and the Ki67 antigen was performed using the mouse anti-human CD79α monoclonal antibody HM57 (Dako, Glostrup, Denmark) and the rabbit anti-human Ki67 antigen monoclonal antibody SP6 (Thermo Scientific). Antigen retrieval was performed as above. The primary antibodies were detected using an Alexa Fluor 488–conjugated goat anti-mouse Ig secondary antibody and an Alexa Fluor 546–conjugated goat anti-rabbit Ig secondary antibody (Life Technologies). The sections were counterstained with DAPI (4,6-diamidino-2-phenylindole) and the formaldehyde-induced autofluorescence was suppressed by staining in 0.1% Sudan Black B in 70% ethanol for 20 min.

The immunostained sections were observed and photographed using a Leica DM4000 microscope (Leica Microsystems, Wetzlar, Germany) equipped with an Olympus DP70 (Olympus, Tokyo, Japan) camera. CD79α single-positive and CD79α/Ki67 double-positive cells were counted using the ImageJ software (Bethesda, MD).46

Preparation and sequencing of fetal and calf IGHV cDNA libraries Samples of ileum were collected from fetuses of 240 and 270 gd and from a 51-days-old calf. Spleen and bone marrow samples were also collected from the fetuses. First-strand cDNA was synthesized using SuperScript III First-Strand Synthesis SuperMix (Life Technologies) according to the manufacturer’s instructions. cDNA was primed with target-specific primers (IgH rev1 for ileum of 240-days-old fetus) or with equal amounts of oligo(dT)20 and random primer mixes for the rest of the samples. Phusion High-Fidelity PCR Master Mix (Thermo Scientific) was used for PCR. Reaction contained 1 × Phusion Master Mix and 0.5 μM of forward and reverse primers (IgH fwd1 and IgH rev1 for samples from the ileum and IgH fwd2 and IgH rev2 for samples from the spleen and bone marrow, Table 3) in a final volume of 50 μl. Primers were designed to span the IGHV region from leader sequence to IGHμ region. In all, 2 μl of first-strand cDNA was used as the template. Cycling conditions consisted of an initial denaturation of 98 °C for 30 s, followed by 18 cycles of 98 °C for 10 s, 60 °C for 30 s, 72 °C for 20 s, and a final extension of 72 °C for 7 min. PCR products were then electrophoresed and purified. Approximately 100 ng of the purified PCR product was ligated into the pCR 4Blunt-TOPO vector and transformed into TOP10 Escherichia coli (Life Technologies). The TOP10 E. coli were grown overnight at 37 °C on LB-ampicillin (100 μg ml−1) plates. In all, 1 ml of LB-ampicillin medium was inoculated with a single bacterial clone.

For the 240 gd ileum, a total number of 384 single clones were isolated for sequencing. Purified DNA was amplified for 35 cycles using Dynazyme II polymerase (Thermo Scientific) with T7 and T3 primers (Oligomer, Helsinki, Finland). The PCR products were sequenced at the DNA sequencing and Genomics laboratory (Institute of Biotechnology, University of Helsinki, Finland).

For the rest of the samples, 48–144 single colonies were picked, purified, and sequenced. Plasmid purification and sequencing was conducted by GATC Biotech AG (Konstanz, Germany).

Preparation and sequencing of a post-immunisation IGHV cDNA library An inactivated whole-cell Streptococcus uberis vaccine was used for vaccination of a calf, which developed a strong antibody response. The draining lymph node of the vaccination site was removed at slaughter 9 months after vaccination, cut into small pieces, and stored in RNAlater (Life Technologies). Total RNA was extracted from the lymph node using Trizol (Life Technologies) according to the manufacturer’s instructions. RNA was DNase treated (DNase I Recombinant, Roche, Basel, Switzerland) and first-strand cDNA was synthesized using the iScript Select cDNA synthesis kit (Bio-Rad, Hercules, CA) with equivalent concentrations of the Oligo(dT)20 and random primer mixes. The bovine VH repertoire was amplified in PCR reactions containing 20 ng cDNA, 0.5 μM of each of the three reverse primers (Bov VH-R1–Bov VH-R3, Table 3) in combination with 0.5 μM of each of one of the eight forward primers (Bov VH-F1–Bov VH-F8), 0.2 μM each dNTP, 2.5 mM MgCl2, 0.05 U Platinum Taq DNA Polymerase (Life Technologies), and 1 × PCR buffer in a final volume of 20 μl. Cycling conditions were: 95 °C for 10 min, followed by 30 cycles of 94 °C for 30 s, 55 °C for 30 s, and 72 °C for 45 s followed by a final 10 min extension at 72 °C. The eight heavy chain reactions were combined and column purified using the DNA Clean and Concentrator kit (Zymo Research, Irvine, CA) before digestion with AscI and SpeI. The digested DNA was ligated into the similarly digested vector JSC (GenBank Accession EU109715). DH5α-E ElectroMax (Life Technologies) were transformed with the ligated DNA and plated on 2YT-amp plates (100 μg ml−1 ampicillin) containing 2% glucose. There were 4.3 × 1010 primary transformants of which 88% contained an insert. A total of 384 clones were isolated from the library for sequencing using primers specific to plasmid JSC (JSCmcsFwd and JSCmcsRev1, Table 3) and analyzed as above.

Cloning and sequencing of bovine germline IGHV genes. Muscle genomic DNA was extracted from three fetuses (aged 182, 240, and 270 gd) and a 51-days-old calf using GenElute Mammalian genomic DNA Miniprep kit (Sigma-Aldrich, St Louis, MO). Primers GenIGHV fwd1 and GenIGHVrev1 (Table 3) were used with Phusion High-Fidelity PCR Master Mix. Purified PCR product was then cloned into pCR 4 Blunt-TOPO vector as above. A total number of 384 colonies was picked, purified, and sequenced by GATC Biotech AG, yielding 35–70 sequences per animal. Sequences that were observed at least twice were accepted, to a total of 214 reads.

Analysis of mutations. The sequence data from the cDNA libraries were analyzed with Geneious Pro software version 5.6.6 (Biomatters, Auckland, New Zealand) and R.47 High-quality cDNA sequences spanning the FR1–FR3 region were matched to the bovine IGHV gene sequences and aligned on the best matching genes using the “map-to-reference” function of Geneious with high sensitivity. Replacement and silent mutations in the cDNA alignments and specifically in the AID-target motifs RGYW/WRCY17, 30 in the IGHV genes were detected in Geneious. RGYW and WRCY target motifs were detected separately in the forward strand. The numbers of mutations in FR and CDR regions, mutations of the AID target nucleotides (the second G in motif RGYW and the third C in motif WRCY), and mutations of G/C nucleotides outside the AID motifs were counted in R.47 The ratios of mutated FR/CDR/AID target/non-target nucleotides were then calculated as

where mi=the number of mutated FR/CDR/AID target/non-target nucleotides in the ith cDNA sequence, ni=the total number of FR/CDR/AID target/non-target nucleotides in the corresponding reference gene sequence, and k=the number of cDNA/reference gene sequence pairs analyzed from the tissue sample.

Statistical analysis. RT-qPCR data from different fetal tissues were compared in R using the nonparametric approximate Kruskal–Wallis one-way analysis of variance followed by multiple comparisons using the nonparametric Nemenyi–Damico–Wolfe–Dunn test at α=0.01.

Cell counts from immunofluorescence analyses were compared using the nonparametric Mann–Whitney U test in the IBM SPSS Statistics software version 20 (IBM Corporation, Armonk, NY).

For detecting selection, the BASELINe framework (Bayesian estimation of antigen-driven selection) was applied.28, 29 The tests were run at the web server at http://selection.med.yale.edu/baseline/ using focused selection statistics and CDR and FR boundaries as defined by Kabat et al.48

The Pearson correlations between AID expression data and the mutation counts were calculated in IBM SPSS Statistics version 20.

P values for cell counts and correlations were corrected for multiple testing using Holm’s method.49