IgM Heavy Chain Complementarity-Determining Region 3 Diversity Is Constrained by Genetic and Somatic Mechanisms Until Two Months After Birth

Sample collection

Liver and spleen from fetuses of 7–28 wk gestation were obtained from the Laboratory of Embryology of the University of Washington over a 10-yr period. Cord blood was obtained from 10 normal term infants. Peripheral blood was obtained from 62 infants ranging in age from 2 wk to 18 mo and from 5 patients with X-linked agammaglobulinemia (XLA) ranging in age from 5 to 32 yr of age. Fetal bone marrow (FBM) cells were flushed from the long bone specimens of a 19 and a 22 wk gestation fetus (determined by fetal foot length), adult bone marrow (ABM) was derived from the resected ribs of two kidney donors (30 and 48 yr old), and XLA bone marrow was obtained from the superior iliac crest of a 5-yr-old patient with XLA. All six patients with XLA have been shown to have mutations that abrogate function of the btk gene (18). All tissue collections and distributions were performed in accordance with policies established by the Human Use Institutional Review Board of the University of Alabama (Birmingham, AL).

Due to sample size, fetal tissue was directly used for RNA extraction in samples of ≤11 wk gestation. In fetal samples of age >12 wk, mononuclear cells were purified by Ficoll-Hypaque gradient centrifugation. Liver and splenic tissue samples were minced and bone marrow cells were flushed from the ribs and the long bones. Buffy coat samples were obtained from cord blood and peripheral blood samples of infants and young children. In selected cases, FACS was used to isolate B cell progenitors from the bone marrow lymphoid cells (see below).

FACS and separation of B lymphocyte progenitor subpopulations

Purified bone marrow mononuclear cells were divided into two aliquots, one of which was stained with 4G7 anti-CD19 (γ₁κ), (Becton Dickinson, Mountain View, CA) and 8G12 anti-CD34 (γ₁κ), (Becton Dickinson). The other aliquot was stained with the anti-CD19 and anti-secretory (s)IgM (Southern Biotechnology Associates, Birmingham AL). In initial experiments, cells were sorted into CD19⁺,sIgM⁻ and CD19⁺,sIgM⁺ subpopulations on a FACStar^Plus instrument (Becton Dickinson). In later studies, the following subpopulations were sorted: CD34⁺/CD19⁺, CD19^high/sIgM⁻, CD19⁺/sIgM^low, and CD19⁺/sIgM⁺ (19). Postsort analysis (FACScan instrument) revealed >97% purity.

RNA isolation and cDNA synthesis

Total RNA was isolated using Tri-Reagent (Molecular Research Center, Cincinnati, OH), according to the manufacturer’s protocol (20). Using standard protocols (21), one-third of each RNA sample was used to synthesize first strand cDNA with avian myeloblastosis virus reverse transcriptase (Boehringer Mannheim, Indianapolis, IN). For each cDNA preparation, a control synthesis reaction was performed without reverse transcriptase to rule out genomic DNA contamination of the RNA.

PCR amplification for IgH transcripts

Rearranged D7-JCμ, D3-JCμ, and VDJCμ transcripts were amplified from 1 μl of the cDNA reaction product from each sample. PCR was performed using a Perkin-Elmer (Norwalk, CT) Cetus model 9600 thermal cycler. Each PCR amplification was performed under the following reaction conditions: 31.7 μl dH₂O, 5 μl 10× PCR buffer with 15 mM MgCl₂ (Boehringer Mannheim), 1.5 μl 50 mM MgCl₂, 8 μl 1.25 mM dNTPs, 2 μl 20 μM each primer, and 1.25 units Pwo DNA polymerase (Boehringer Mannheim). Each reaction underwent one initial cycle of denaturation at 95°C for 1 min, followed by 30 cycles of amplification with denaturation at 95°C for 30 s, primer annealing for 1 min, and extension at 72°C for 1.5 min. A final extension was performed at 72°C for 5 min. Primer annealing temperatures were 55°C for D7-JCμ, 60°C for D3-JCμ, and 56°C for VDJCμ amplifications. Primer combinations used to amplify cDNA, in separate reactions, were: D7-JCμ (5′-GAG CTG AGA ACC ACT GTG-3′ and 5′-CCC GGG TGC TGC TGA TGT CAG-3′) (22, 23), D3-JCμ (5′-GAG GTC TGT GTC ACT GTG-3′ and 5′-GGA GAA AGT GAT GGA GTC GGG-3′ (24) or 5′-GGA CAT CCC (AG)GG TTT CCC CAG-3′ and 5′CCC GGG TGC TGC TGA TGT CAG-3′) (25), and VD-JCμ (5′-CGC AAG CTT ACA C(TGA)G C(CT)G-3′ and 5′-CCC GGG TGC TGC TGA TGT CAG-3′) (26). (The first set of D3 primers amplified members of the D2, D3, and D6 families. The second set was specific for D3.) Actin mRNA was analyzed as a control for the reliability of the cDNA synthesis, using the primers 5′-GTG GAC TTG GGA GGA GGA CTC TGG G-3′ and 5′-GCGGGA AAT CGT GCG TGA CAT T-3′ (27).

Assessment of HCDR3 length by spectrotype

A 1-μl aliquot of first-round VDJCμ product underwent a second round of PCR amplification using the original Framework 3 consensus region V primer (5′-CGC AAG CTT ACA C(TGA)G C(CT)G-3′) in association with an internal, nested Cμ primer (5′-AAT TCT CAC AGG AGA CGA G-3′). This latter primer was [γ-³²P]ATP end-labeled with polynucleotide kinase (Boehringer Mannheim), as per the manufacturer’s recommendations. The reaction conditions were the same as the first round, with the exception that amplification proceeded for only five cycles. A 10-μl aliquot of second round PCR product was mixed with 2 μl of loading dye (21), boiled for 2 min in a water bath, quick cooled, and a 6-μl aliquot of the mix was loaded on 7.5% polyacrylamide sequencing gel. Each gel included a set of lanes with sequenced M13 mp18, and one lane containing an amplified HCDR3 interval from a VDJCμ clone of known sequence as controls for length and amplification. After electrophoresis, the gel was exposed to x-ray film (Kodak, Rochester, NY) for an average of 4 h, and to an imaging plate (Fuji Photo Film, Japan) for an average of 30 min. The image plate was scanned with a Fujix Bas 1000 Phosphoimager, and the generated images were analyzed using MacBass V2X software (Fuji). After subtraction for background, the intensity of each section of a given lane corresponding to a three-nucleotide interval was quantified and divided by the total intensity of the lane that corresponds to lengths of 6–30 codons. This percentage contribution was multiplied by the corresponding HCDR3 length in codons. Finally, the contribution of each codon was added to determine the average length of the HCDR3 intervals in each lane.

Sequence analysis of DJ and VDJ transcripts

A total of 5 μl of the initial PCR products were reamplified using: D7 (5′-TTT AAG CTT GAG CTG AGA ACC ACT GTG-3′), D3 (5′-CGG AAG CTT GAG GTC TGT GTC ACT GTG-3′), or V (5′-CGC AAG CTT ACA C(TGA)G C(CT)G-3′) primers in association with a nested Cμ primer (5′-GGA GAA AGT GAT GGA GTC GGG-3′). There is an internal EcoRI restriction site in Cμ just upstream of this latter primer. Reamplification was performed under the following conditions: 60.5 μl dH20, 10 μl 10× buffer with 15 mM MgCl₂ (Boehringer Mannheim), 16 μl 1.25 mM dNTPs, 4 μl 20 μM primers, 1.5 U Pwo DNA Polymerase (Boehringer Mannheim), and 5 μl from the first round. Second-round PCR products were digested with EcoRI and HindIII, cloned into either pBluescript SK (Stratagene, San Diego, CA) or pZero (Invitrogen, Carlsbad, CA), and screened with a Cμ probe (AAT TCT CAC AGG AGA CGA G). DNA sequencing was performed by the dideoxy termination technique (28).

HCDR3 is defined as the interval between the conserved cysteine (amino acid 92) at the carboxy terminus of Framework 3 and the conserved tryptophan (amino acid 103) at the amino terminus of Framework 4 (9, 12). To classify the D_H germline sequence (25) within the HCDR3 interval of each of the 253 sequences analyzed, the criteria of a minimum of either seven nucleotides of uninterrupted identity with germline sequence, including P junctions, or eight nucleotides with one mismatch with at least two nucleotides of identity at both the 3′ and 5′ end of the candidate D_H gene segment was used.

A total of 37 HCDR3 intervals, of which 28 were in-frame, were cloned from 15 different fetal liver samples of 7–10 wk gestation. This compilation includes 13 previously reported intervals (8). Fetal liver sample size precluded sorting cells into subpopulations; thus, these intervals derive from a mixture of B cells and B cell progenitors. A total of 32 second trimester fetal liver HCDR3 intervals, 26 in-frame, were also analyzed. Of these, 26 HCDR3 intervals from 2 fetal liver cDNA libraries of 15 and 19 wk gestation (23, 26) have been previously reported. Six additional cDNAs were isolated from the same 15-wk cDNA library (23). Mononuclear cells from 19-wk FBM were sorted on the basis of surface expression of CD19 and IgM. Twenty-two HCDR3 intervals from CD19⁺,sIgM⁻ pro and pre-B cells, 20 in-frame, and 63 HCDR3 intervals from CD19⁺,sIgM⁺ B cells, all in-frame, were obtained. These sequences were compared with a previously published set of 99 HCDR3 sequences cloned from the blood of 6 normal adults (17).

A total of 55 fetal and 33 adult D3-J and 55 fetal and 22 adult D7-J transcripts were analyzed. A preliminary evaluation of a subset of these DJ transcripts has been previously reported (15). The 8-wk fetal liver samples yielded 14 D3-J transcripts and 14 D7-J transcripts. A 12-wk fetal liver sample yielded 16 D3-J and 22 D7-J transcripts. CD19⁺,sIgM⁻ B cell progenitors from a 19-wk FBM sample yielded 25 different D3-J and 19 D7-J transcripts. CD19⁺,sIgM⁻ B cell progenitors from 2 different ABM samples yielded 17 D3-J and 10 D7-J transcripts. An additional 16 D3-J and 12 D7-J transcripts were obtained from unsorted bone marrow of a third adult donor. Lastly, 31 D3-J transcripts were obtained from the peripheral blood and bone marrow of 6 different patients with XLA. Due to message abundance, only two D7-J transcripts were obtained from the same samples. Each of these reported sequences are unique; repeat sequences were not included in this analysis.

Statistical analysis

Statistical analysis was performed using the program JMP (SAS Institute, Cary, NC). Average values are reported with the SEM.

Molecular modeling

All molecular modeling was performed on a Silicon Graphics IRIS/4D220GTX computer system using the QUANTA modeling software (Polygen, Waltham, MA). The crystallographic structure of mouse J539 Fab (pdb file 2FBJ (29); obtained from the Brookhaven National Protein Database (30, 31)) was chosen for homology-based modeling because of its use of the κ isotype and its use of a V_H gene segment from the same evolutionary clan as human V3-23. The mouse κ L chain variable domain was first altered to form a humkv325-Jκ1. This mutation was followed by 100 cycles of steepest descents minimization to clear the structure of any bad amino acid contacts caused by mutation. The J539 V_H segment was then mutated to the appropriate sequence. The complete structure was subjected to extensive minimization, which included 100 cycles of steepest descents minimization followed by at least 500 cycles of conjugate gradient minimization with the objective of gaining an rms force applied under 0.6 for 10 consecutive cycles.

During ontogeny there is an inverse relationship between D_H and J_H utilization and chromosomal position relative to the boundary between D_H and J_H

The single member of the D7 (DQ52) family, D7-27, contributed to one-half of first trimester, one-third of second trimester, but only one-fiftieth of the adult HCDR3 intervals examined (Fig. 1⇓). This gene segment is located immediately adjacent to the J_H locus. The five members of the D3 (DXP) family (D3-3, D3-9, D3-10, D3-16, and D3-22) are scattered 20–50 kb upstream of J_H (25). These gene segments contributed to one-twentieth of first trimester, one-tenth of second trimester, and nearly one-third of adult sequences (p < 0.0001, χ²). Among VDJCμ transcripts, J_H4 was the most commonly used J_H gene segment at all stages of ontogeny. However, J_H2 was the second most commonly utilized J_H gene segment among first trimester transcripts, J_H3 was the second most common among second trimester transcripts, and J_H6 was the second most common among adult transcripts (p < 0.0001, χ²).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 1.

HCDR3 diversity during human ontogeny. Top row, First trimester fetal liver VDJCμ⁺ transcripts. Middle row, Second trimester fetal liver and bone marrow VDJCμ⁺ transcripts. Bottom row, Adult blood VDJCμ⁺ transcripts from the published work of Yamada et al. (17). Left column, Percentage of transcripts that utilize members of the designated D_H families. A “?” designates transcripts where the D_H family cannot be determined. The D7 (DQ52) gene segment contributed to 19 of 37 first trimester, 40 of 117 second trimester, and 2 of 99 adult VDJCμ transcripts; whereas D3 (DXP) gene segments contributed to 2 of 37, 13 of 117, and 29 of 99 transcripts, respectively (p < 0.0001 χ²). Middle column, Distribution of the lengths of the CDR 3 intervals of the transcripts (residues 93–102 (8;9;12)) divided into 3 residue intervals (e.g., <9, 10–12, 13–15, 16–18, 19–21, 22–24, and >24 codons). Right column, Percentage of transcripts that utilize the designated J_H gene segment. For each developmental stage, the number of transcripts analyzed is indicated by the value of n. The arrows are intended to emphasize the major differences between fetal and adult transcripts.

The distribution of HCDR3 lengths is restricted in the fetus and neonate

The average length of HCDR3 length distributions also changes during ontogeny. By sequence analysis, although the average length was the same in first and second trimester sequences, it was approximately five codons longer in the blood of adults (12.2 ± 0.5, 12.5 ± 0.2, and 17.7 ± 0.5 codons, respectively; p < 0. 0001, χ²). The similarity in average HCDR3 length between first and second trimester samples led us to assess the relationship between average HCDR3 length and the maturation of the IgM repertoire. By spectrotype analysis, the average HCDR3 length distribution was found to remain relatively unchanged in fetal liver of 8- to 28-wk gestation, exhibiting a mean of 13.2 ± 0.2 codons (Fig. 2⇓). Fetal liver is a hematopoietic organ, enriched for B cell progenitors, whereas spleen is a lymphoid organ that contains a higher proportion of mature B cells as the fetus develops (32). Unlike liver, where the HCDR3 length was similar in first and second trimester tissues, by spectrotype analysis, HCDR3 intervals from first trimester spleen were smaller than those from second trimester (12.6 ± 0.2 vs 13.3 ± 0.2 codons, respectively; p < 0.02, Student’s t test) (Fig. 2⇓). In the cord blood, which contains primarily naive sIgM⁺/sIgD⁺ B cells, the average HCDR3 was two residues longer by spectrotype analysis than either second trimester spleen or liver (15.4 ± 0.1 codons) (Fig. 2⇓). By spectrotype analysis, a steady state HCDR3 average of 17.3 ± 0.2 was reached at 2 mo of age (p < 0.0001, χ²).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 2.

HCDR3 length distribution during human ontogeny. Shown are average HCDR3 lengths in VDJCμ transcripts from: A, 34 fetal liver samples ranging in age from 8 to 28 wk gestation, B, 35 fetal spleen samples of which 17 derive from the first trimester of gestation and 16 from the second trimester, and C, 72 samples of cord blood and peripheral blood from infants ranging in age from birth to 18 mo of age. The dotted lines denote 5% and 95% confidence limits for the distribution of lengths over time.

D7- and D3-containing sequences were analyzed to elucidate the contribution of individual gene segments to HCDR3 length

D7 and D3 gene segments were chosen because they best represent the change in D_H utilization between fetus and adult (Fig. 1⇑). Among our database of 253 VDJCμ sequences, 14 fetal D3-, 59 fetal D7-, 28 adult D3-, and 2 adult D7-containing HCDR3 intervals were identified. The adult D7-HCDR3 sequences were both out-of-frame and of insufficient numbers for comparison. Two of the fetal D3 sequences that had undergone rearrangement by inversion were also excluded from the analysis. In the adult, D3-containing HCDR3 intervals had the same length as the average for the population of HCDR3 intervals as a whole (17.7 ± 0.7 codons vs 17.5 ± 0.5 codons, respectively, by sequence analysis).

D_H and J_H gene segment lengths differ and might be expected to influence the length of HCDR3 (Fig. 3⇓). D7-27 is the shortest D_H, consisting of only 11 nucleotides. With a length of 31 nucleotides, D3-3, -10, -11, and -22 are among the longest conventional D_H gene segments (25). J_H4 is the shortest human J_H, contributing only 14 nucleotides to HCDR3; whereas J_H3 can contribute 16; J_H5, 17; J_H1, 18; J_H2, 19; and J_H6 can contribute up to 29 nucleotides, respectively. When limited to germline sequence, D3-HCDR3 intervals can achieve lengths of 17 (J_H4) to 22 (J_H6) codons, whereas D7-containing HCDR3 intervals can only achieve lengths of 11–16 codons, respectively. To attain lengths longer than 16 or 22 codons, respectively, P junctions or N nucleotides must be added.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 3.

A dissection of the components contributing to HCDR3 length in D3- and D7-containing H chain variable domains. The potential contribution of the germline sequence of V_H gene segments, P junctions, N region addition, D_H gene segments, and J_H gene segments to HCDR3 length is illustrated. The actual contribution of these components to 28 adult D3-containing HCDR3 intervals, 12 fetal D3-containing intervals, and 59 fetal D7-containing intervals is shown below. Also shown beside the D_H gene segment is the average 5′ and 3′ loss of sequence for D3 and D7 gene segments. All components are shown to scale.

Given the difference in lengths, we expected that fetal HCDR3 intervals containing D3 gene segments would be longer than those containing D7. However, by sequence analysis, D3-HCDR3 and D7-HCDR3 intervals exhibited nearly identical average lengths of 12.8 ± 0.7 and 12.7 ± 0.3 codons, respectively (Fig. 3⇑).

Differences in fetal and adult HCDR3 length distributions were the product of D-specific preservation of D_H sequence

Fetal and adult HCDR3 sequences containing D3 and D7 gene segments were dissected to assess the role of D_H length, J_H utilization, and N region addition in controlling HCDR3 length (Fig. 3⇑). One very significant factor was the greater resistance of D7 gene segments to the loss of terminal nucleotides. D7 gene segments lost only an average of 0.2 ± 0.2 nucleotides at the 3′ terminus and 2.2 ± 0.2 at the 5′ terminus, whereas D3 gene segments lost 8.3 ± 0.9 and 11.5 ± 0.9 nucleotides at each terminus, respectively (p < 0.0001, Student’s t test) (Fig. 3⇑). D7 elements also contained more P junctions. There was an average of 0.5 P nucleotides added to the 3′ end of the D7 gene segments, whereas D3 elements had no P junctions (p < 0.04, Student’s t test). J_H length and N addition appeared to play lesser roles. (Fig. 3⇑). The J_H portion of HCDR3 and the extent of N addition between D and J were slightly greater in D7- vs D3-HCDR3 intervals, but neither of these trends achieved statistical significance (J_H length: 11.5 ± 0.5 vs 9.9 ± 1.2 nucleotides, respectively; p = 0.08, Student’s t test. N addition: 4.3 ± 0.5 vs 2.8 ± 1.0 nucleotides, respectively; p = 0.25, Student’s t test). (Fig. 3⇑).

In the adult, preservation of D_H sequence was a major factor contribution to the increase in the average length of HCDR3. When compared with the fetus, adult D3-containing sequences retained eight additional nucleotides at the 5′ end (loss of 3.7 ± 0.9 vs 11.5 ± 0.9 nucleotides in adult and fetus, respectively; p < 0.0001, Student’s t test) (Fig. 3⇑). Loss of these eight terminal nucleotides in fetal μ transcripts reduces the likelihood of incorporating the only charged amino acid encoded by the 5′ end of D3 gene segments, aspartic acid. N addition and the contribution of J_H sequence to HCDR3 length was enhanced in the adult, but these differences again failed to achieve statistical significance (Fig. 3⇑).

Fetal constraints on HCDR3 length are already present in non-Ag selectable DJ transcripts

The change in HCDR3 length distributions between fetus and adult could reflect differences in the interaction between the recombinase and the IgH locus, Ag receptor-influenced selection of the final products, or both. Human DJ joins, the intermediate products of H chain rearrangement, are transcribed but lack the conserved upstream translation start codon that allows translation to a Dμ protein product in the mouse (25, 33). Analysis of DJ transcripts allows assessment of the site of 3′ D_H rearrangement, N region addition between D and J, patterns of J_H utilization, and the site of J_H rearrangement without concern for selection for a protein product.

Within each D_H family, processing of the 3′ coding ends was similar in fetus and adult. Between families, however, recombination led to a greater loss of 3′ D coding sequence in D3-J transcripts than in D7-J transcripts (5.3 ± 3.9 vs 1.3 ± 3.2 nucleotides at the 3′ terminus, respectively; p < 0.0001, Student t test). Fewer D3-J transcripts incorporated P junctions than D7-J transcripts (1 of 88 vs 27 of 77 sequences with P junctions, respectively; p < 1 × 10⁻⁶, χ²).

J_H utilization in fetal and adult DJ transcripts differed by both age and D gene segment

Mirroring the pattern seen in VDJ rearrangements, sequence analysis of the DJ transcripts revealed that J_H4 was the predominant J_H gene segment in both D3-J and D7-J transcripts at all ages studied (Fig. 4⇓). Use of J_H6 increased from one-seventh of fetal D3-J transcripts to one-third of adult D3-J transcripts (8 of 55 vs 11 of 33, respectively; p < 0.02, χ²). Among D7-J transcripts, J_H6 increased from none in the fetus to one-seventh of the adult D7-J transcripts (0 of 54 vs 3 of 22, respectively; p = 0.03, χ²). Between D gene segments, in the fetus, J_H6 was used in none of the D7-J transcripts and in one-seventh of the D3-J transcripts (0 of 54 vs 8 of 55; p < 0.001, χ²). In the adult, differences in the prevalence of J_H6 in D3-J and D7-J transcripts did not reach statistical significance (p = 0.24, χ²).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 4.

J_H utilization among DJ transcripts during human ontogeny. Depicted are the percentage of transcripts that use the designated J_H gene segment. Left column, D3-J transcripts. Right column, D7-J transcripts. Shown by row from top to bottom are transcripts from 8-wk fetal liver, 12-wk fetal liver, 19-wk FBM, ABM, and the blood and bone marrow of patients with XLA, respectively. For each category, the number of transcripts analyzed is indicated by the value of n. The arrows are intended to emphasize the major differences between fetal and adult transcripts.

The site of rearrangement in J_H also differed by D gene segment. Due to greater abundance, this was best demonstrated among those transcripts that contained J_H4. On average, J_H4 lost one more nucleotide at the 5′ terminus in D7-J_H4 transcripts vs D3-J_H4 transcripts (5.4 ± 0.4 vs 4.2 ± 0.4 nucleotides, respectively; p < 0.02, Student’s t test). The presence or absence of N nucleotides had no effect on the site of J_H rearrangement. The greater loss of J_H sequence in D7-J_H4 rearrangements was unexpected, given that on average, D7-HCDR3 intervals contained more nucleotides derived from J_H sequence than D3-HCDR3 intervals (Fig. 3⇑).

The extent of N addition differed by both age and D gene segment

By sequence analysis, the extent of N addition between D and J was similar in adult D3-J and D7-J transcripts, but differed markedly between fetal D3-J and D7-J transcripts and between fetal and adult D-J transcripts in general (Fig. 5⇓). Fetal D7-J transcripts contained five times more N addition than fetal D3-J transcripts (5.0 ± 0.6 vs 0.9 ± 0.3 nucleotides, respectively; p < 0.0001, Student’s t test) (Fig. 5⇓). Between fetus and adult, N region addition in D7-J transcripts increased by 70% (5.0 ± 0.6 vs 8.5 ± 1.2 nucleotides, respectively; p < 0.01, Student’s t test), whereas there was an 800% increase of N addition in D3-J transcripts (0.9 ± 0.3 vs 7.8 ± 1.0 nucleotides, respectively; p < 0.0001, Student’s t test) (Fig. 5⇓). Although the proportion of sequences containing N addition increased from 7% at 8 wk to 25% at 19 wk, in the adult 6% (2 of 33) of the D3-J transcripts still lacked N addition. (Fig. 5⇓). Thus, even in the adult, TdT did not add N nucleotides to all rearranging gene segments.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 5.

Patterns of N addition during human ontogeny. Distribution of the extent of N addition in D7- and D3-containing DJ transcripts from normal fetal and adult B lineage cells, and from the blood and bone marrow of patients with XLA. Statistically, there was no difference in N addition between D7-J and D3-J transcripts in the adult (8.5 ± 1.2 vs 7.8 ± 1.0 nucleotides, respectively; p = 0.67, Student’s t test).

J_H utilization appeared to vary between those transcripts that contain N addition vs those that did not. Of the 14 D3-J sequences from first trimester tissues, only one used J_H6, and this sequence was the only one with N addition. One-third (4 of 13) of all the fetal D3-J sequences that contain N addition used J_H6, whereas only one-tenth (4 of 41) of the sequences without N addition used J_H6. Thus, control of N addition and J_H utilization may be linked.

Absence of N addition had no apparent effect on D_H reading frame preference

Unlike the mouse, where preferential rearrangement at sites of sequence identity between D and J helps dictate use of D_H reading frame 1 (rf1) (6, 7, 34), absence of N addition had a minimal effect on reading frame preference. One-half of the first trimester D3-J transcripts, one-fourth of the second trimester D3-J transcripts, and one-twentieth of the adult D3-J transcripts had D-J rearrangements that occurred at sites of identity between D and J (14 of 30, 9 of 33, and 2 of 33, respectively). Following the pattern seen in mouse DFL and DSP sequences, rf1 in human D3 gene segments is enriched for tyrosine, serine, and glycine, rf2 encodes hydrophobic amino acids, and rf3 contains many termination codons (8, 25). Among the 25 transcripts with overlapping sequence, reading frame utilization appeared random with nine using rf1, nine using rf2, and seven using rf3. Among the transcripts that contained N addition, one-half used rf1, one-fourth used rf2, and one-sixth used rf3 (a ratio of 36:16:11; p < 0.0001, binomial distribution). Among all of the D3-J transcripts, one-half used rf1, one-fourth used rf2, and the remaining one-fifth used rf3 (a ratio of 45:25:15; p < 0.0003, binomial distribution). There was no significant difference in reading frame preference between fetal and adult D3-J transcripts, or in fetal and adult D7-J transcripts. A reading frame preference was also apparent among D7-J transcripts, with one-third using rf1, one-fourth using rf2, and one-half using rf3 (a ratio of 23:20:34; p = 0.02, binomial distribution). A preference for rf1 in D3-HCDR3 and for rf3 in D7-HCDR3 intervals has been previously reported (25). Similarity in reading frame preference by both precursor DJ and mature VDJ transcripts suggests that the pattern of reading frame utilization is established early in B cell development.

In the adult, individual D3 gene segment utilization was similar between DJ transcripts and VDJ transcripts (p = 0.23, χ²), with D3-22 most common, followed by D3-3, D3-9, and D3-10 in that order. In the fetus, D3-3 rather than D3-22 was the most commonly used D_H in DJ transcripts (p = 0.03, χ²).

Absence of btk expression can influence the extent of N addition

Examination of DJ transcripts from patients with XLA confirmed that enhancement of N addition could be associated with prolonged exposure to TdT in vivo. Patients with XLA have loss of function mutations in the cytoplasmic tyrosine kinase btk (35, 36). These mutations lead to a block in B cell maturation at the pro-B/pre-B cell stage. In normal individuals, TdT is detected in <1% of cytoplasmic Cμ⁺ cells. In XLA, however, TdT expression is prolonged with up to one-third of cytoplasmic Cμ⁺ cells still expressing this enzyme (37, 38). By sequence analysis, XLA D3-J transcripts were found on average to contain almost twice as many N nucleotides as D3-J transcripts from normal ABM (13.1 ± 1.9 vs 7.8 ± 0.9 nucleotides, respectively; p = 0.01, Student’s t test) (Fig. 5⇑). The HCDR3 lengths of VDJCμ transcripts in the peripheral blood of some of the XLA donors were then evaluated by spectrotype analysis to see if long DJ rearrangements dictated the presence of long HCDR3s. Of the four patients evaluated, VDJCμ transcripts could not be detected in one, a second patient had a single band likely representing only one productive rearrangement, a third patient exhibited three such bands, and the fourth had eight. Presuming each of these 12 bands represented a unique transcript, an average HCDR3 length was calculated and found to be similar to the average for cord blood (14.8 ± 0.9 vs 15.4 ± 0.1 codons, respectively).

Convergence of fetal D3- and D7-HCDR3 lengths is likely the result of selection

In the fetus, J_H utilization and the extent of N addition in VDJ transcripts was not representative of that observed among DJ progenitors. Among fetal D3-J transcripts, the extent of N addition at the DJ junction increased 3-fold between DJ and VDJ (0.9 ± 0.3 vs 2.8 ± 1.0 N nucleotides, respectively; p = 0.01, Student’s t test) (Figs. 3⇑ and 5⇑). In the adult, the differences in N addition were not significant (7.8 ± 0.9 vs 6.5 ± 1.3 N nucleotides, respectively; p = 0.4, Student’s t test) (Figs. 3⇑ and 5⇑). Similarly, among fetal sequences using the D7 gene segment, there was little difference between DJ transcripts and VDJ transcripts (5.0 ± 0.6 vs 4.3 ± 0.5 N nucleotides, respectively; p = 0.3, Student’s t test) (Figs. 3⇑ and 5⇑). Note that for those sequences using the short D7 gene segments, N addition between DJ and VDJ transcripts did not enhance length; and that for sequences using the longer D3 gene segments, N addition was more likely to increase length (Fig. 3⇑).

Trimming of D_H and use of alternative J_H gene segments appear to be the two most important factors in allowing the length distribution of D3- and D7-HCDR3 intervals to converge. When compared with DJ transcripts, fetal D3 gene segments in VDJ transcripts lost twice as many nucleotides at the 3′ terminus (4.7 ± 0.5 vs 8.3 ± 0.9 nucleotides, respectively; p < 0.006, Student’s t test), trimming the average length by one codon. There was no significant difference in the loss of 3′ nucleotides between adult DJ and VDJ transcripts. In contrast, D7-HCDR3 intervals preserved 3′ D_H sequence (loss 1.3 ± 0.4 nucleotides in DJ transcripts vs 0.2 ± 0.2 nucleotides in VDJ transcripts; p = 0.02, Student’s t test).

A shift in J_H utilization between DJ and VDJ helped lengthen D7-HCDR3 intervals in the fetus, but no such shift was detected in adult. Among 19 first trimester D7-VDJ transcripts, 5 contained J_H2, whereas none of 14 D7-J transcripts contained J_H2 (p < 0.05, Fisher’s exact test) (Fig. 4⇑). On average, D7-HCDR3 intervals with J_H2 were longer by six codons than those using J_H4 (15.6 ± 0.8 vs 9.0 ± 0.6 nucleotides, respectively; p < 0.001, Student’s t test). The number of adult D7-HCDR3 intervals and fetal D3-HCDR3 intervals was not sufficient for statistical evaluation of J_H utilization. Among adult D3-HCDR3 intervals, J_H utilization mirrored that seen in DJ transcripts (Fig. 4⇑).

Selection of HCDR3 length is associated with the surface expression of μ H chain

Trimming of D_H and the shift in J_H utilization supports the view that HCDR3 intervals could be actively selected for length. Studies showing a decrease in HCDR3 length distribution in B cell progenitors compared with B cells in ABM (39), enhanced length in nonfunctional vs functional rearrangements (40), and the decrease in length distribution between first trimester fetal liver and spleen in this work suggest that selection in the fetus could occur during the process of differentiation from a pre-B cell to B cell. We had previously amplified V_H3-containing VDJCμ transcripts from the bone marrow of a 22-wk fetus and a 47-yr-old adult. We reamplified these products and analyzed the distribution of HCDR3 lengths by spectrotype (Fig. 6⇓). In both fetus and adult, the average HCDR3 length decreased during B cell differentiation. (Fig. 6⇓). The most dramatic alteration of lengths was observed between those cells expressing only cytoplasmic μ H chain and those expressing a low level of surface IgM (Fig. 6⇓).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 6.

Distribution of average HCDR3 lengths of V_H3DJCμ transcripts from FBM and ABM as a function of B cell differentiation.

HCDR3 length is also regulated in the mouse

Barring additional experiments of nature such as XLA, analysis of the mechanism by which HCDR3 length is selected remains a primarily descriptive endeavor. To determine whether a similar process occurs in a species more amenable to experimental manipulation, we evaluated the HCDR3 length distributions of previously reported VDJCμ transcripts from newborn liver, newborn spleen, and adult spleen of mouse (7) (Fig. 7⇓). With a length of 11 nucleotides, mouse DQ52 is 2 codons shorter than the 17-nucleotide long members of the DSP family. In turn, 23-nucleotide long DFL16.1 contains 2 more codons than DSP gene segments. On average, neonatal liver DQ52 HCDR3 intervals were three codons shorter than DSP HCDR3s; and DSP HCDR3s were two codons smaller than DFL16.1 HCDR3s (7.2 ± 0.8, 10.9 ± 0.6, and 13.0 ± 0.5 codons, respectively; p < 0.0002, one way ANOVA) (Fig. 7⇓). In spleen, HCDR3 lengths appear to have been adjusted by selection. Although still exhibiting unique length distributions, DQ52 HCDR3 intervals were lengthened, DSP intervals retained a similar length, and DFL16.1 intervals were shortened when compared with liver (8.0 ± 0.5, 9.7 ± 0.3, and 11.1 ± 0.6 codons, respectively; p < 0.0002, one way ANOVA) (Fig. 7⇓). The difference between an average length for neonatal liver and spleen DFL16.1-containing HCDR3 intervals was significant at p < 0.05 (Student’s t test). In adult spleen, DQ52, DSP, and DFL16.1 HCDR3 intervals converged to the same average length (10.6 ± 0.6, 11.1 ± 0.5, and 11.8 ± 0.6; DSP codons, respectively; p = 0.4, one way ANOVA) (Fig. 7⇓). Although DFL16.1 lengths remained the same between neonatal and adult spleen, the increase in length for splenic DQ52- and DSP-containing HCDR3 intervals was significant at p < 0.05 (Student’s t test).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 7.

The distribution of HCDR3 lengths is controlled in mouse. The distribution of HCDR3 lengths in VDJCμ⁺ transcripts from neonatal liver, spleen, and adult spleen (7). Shown is the average length of sequences using the DFL16.1 and DQ52 gene segments, and members of the DSP family, as well as the average length of HCDR3 intervals for the population as a whole.

Even in the presence of D_H gene segments of human length, HCDR3 length was preserved in the mouse. We evaluated the distribution of HCDR3 intervals from mice with a transgenic human IgH minilocus containing representatives of all of the human D_H gene segments, as well as the complete J_H locus (41). Among 32 in-frame transcripts, the average HCDR3 length was 12.1 ± 0.6 codons, which is similar to the average for fetal liver and bone marrow in human, but 5 codons less than that seen in the peripheral blood of children and adults (see above).