VλJλ rearrangements obtained from genomic DNA of individual IgM+ B cells from human fetal spleen were analyzed. A nonrandom pattern of λ gene rearrangements that differed from the adult Vλ repertoire was found. The Vλ distal genes 8A and 4B were absent from the nonproductive fetal repertoire, whereas 2E and 3L were overrepresented and 1B was underrepresented in the productive fetal repertoire. Positive selection of the Vλ gene, 2E, along with Vλ rearrangements employing homologous VλJλ joins were observed in the fetal, but not in the adult Vλ repertoire. Overrepresentation of Jλ distal cluster C genes rearranging to the Vλ distal J segment, Jλ7, in both productive and nonproductive fetal repertoires suggested that receptor editing/replacement was more active in the fetus than in adults. Numerous identical VλJλ junctions were observed in both the productive and nonproductive repertoire of the fetus and adult, but were significantly more frequent in the productive repertoire of the fetus, suggesting expansion of B cells expressing particular λ-light chains in both stages of development, with more profound expansion in the fetal repertoire. Notably, B cells expressing identical λ-light chains expressed diverse heavy chains. These data demonstrate that three mechanisms strongly influence the shaping of the human fetal λ-chain repertoire that are less evident in the adult: positive selection, receptor editing, and expansion of B cells expressing specific λ-light chains. These events imply that the expressed fetal repertoire is shaped by exposure to self Ags.
In the adult human, B cells are generated in the bone marrow. During prenatal development, however, fetal liver transiently plays the major role in production of B lineage cells. By ∼12 wk of gestation, production of B lineage cells shifts to the bone marrow, which then assumes the major role throughout life (1).
B cell development in bone marrow requires successful rearrangement of the Ig H and L chain gene loci and the surface expression of the B cell Ag receptor (BCR)4 (2). These properties define the stages of B cell maturation during bone marrow development. At the pre-B cell stage, cells have successfully completed functional rearrangement of the heavy chain and it pairs with surrogate light chain to express the pre-B cell receptor (pBCR) on the surface (3). When pre-B cells undergo successful light chain rearrangement, they develop into immature B cells and express the mature BCR (4). Immature B cells produced in the bone marrow migrate into the periphery to complete their maturation process (5).
Expression of the pBCR and BCR affords the possibility that the primary B cell repertoire is shaped and directed by continuous clonal selection beginning at the pre-B cell stage of differentiation (6). Analysis of VH and Vκ gene expression in different stages of development, in both humans and mice, has suggested developmentally controlled nonrandom use of V(D)J genes and unique patterns of junctional diversity (7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17). With better understanding of BCR signaling, recent studies have provided direct evidence that bias in the Ab repertoire is a result of selection mediated through pBCR and BCR (18, 19, 20). It has also been demonstrated that the migration of immature B cells into the periphery and competition for entry into the peripheral immune system are active processes mediated by signals through the BCR. The net result of these events is a marked narrowing of the Ig repertoire (3, 21). However, the ligands that drive and select B cells during these stages of differentiation are still not known. It has been suggested that certain levels of BCR signaling by interaction with self Ag might be essential for survival and maturation of B cells (6, 22).
The fetal microenvironment is different from that of the adult, because it is not exposed to external Ags, but presumably abundantly exposed to self Ags. Therefore, the repertoire that develops during fetal life, the preimmune repertoire, may be considerably different from that found in adults. Mechanisms that restrict and shape the human adult primary Ig repertoire have been studied intensively. However, molecular and selective mechanisms involved in shaping the human preimmune repertoire have not been examined in detail. Analysis of the fetal B cell repertoire could, therefore, lead to a better understanding of the impact of the different influences that shape the expressed Ig repertoire.
In the current study, detailed Ig λ gene repertoire analysis was conducted using genomic DNA obtained from individual IgM+ B cells from fetal spleen. The goal was to analyze the preimmune repertoire in humans to understand in greater detail the impact of the molecular mechanisms and selective influences that shape it, and to compare this with the adult human Vλ repertoire. The technique allowed us to analyze both productive and nonproductive rearrangements without introducing an activation-related bias. Using this methodology, molecular mechanisms and selective influences that shape the human λ gene repertoire in peripheral IgM+ B cells from normal adult donors have recently been assessed (23). This adult Vλ database was used for comparison with the fetal preimmune repertoire. The results indicate that there was a nonrandom rearrangement pattern in the λ gene repertoire at the fetal stage of development. Positive selection of individual Vλ genes and expansion of B cells expressing particular Vλ rearrangements with homology-mediated joining independent of heavy chain expression strongly influenced the fetal repertoire, resulting in predominance of particular VλJλ junctions in the productive repertoire. Evidence for more active peripheral receptor editing in the fetal, compared with the adult Vλ repertoire was also found. These events occur before exposure of the fetus to exogeneous Ag and, therefore, must be mediated by intrinsic developmental signals and/or self Ags.
Materials and Methods
Cell preparation and sorting
Single cell preparations were made from three fetal spleens at 18 wk of gestation by mechanical disruption of tissue fragments, followed by filtration through nylon mesh. Fetal bone marrow cells were flushed from long bone specimens of an 18-wk fetus. All tissue collections and processing were done in accordance with policies established by the institutional review board for human experimentation at the University of Texas Southwestern Medical Center (Dallas, TX). Mononuclear cells were enriched by Ficoll-Hypaque density-gradient centrifugation, as described (24). The cells were then stained with a PE-labeled anti-CD19 mAb (Becton Dickinson, Mountain View, CA) and a FITC-labeled anti-human IgM mAb (Caltag, Burlingame, CA). The CD19+/IgM+ B cells were sorted using a FACStarPlus flow cytometer (Becton Dickinson, San Jose, CA) outfitted with an automated single cell deposition unit, and one cell was deposited into each well of a 96-well PCR plate assembled on a microAmp base, as described previously (24, 25). Each well contained 10 μl of PCR buffer (50 mMKCl, 10 mM Tris-HCl, pH 9, 0.1% Triton X-100).
PCR amplification included an initial primer extension preamplification (24) and subsequent nested PCR steps (23). Single cells in 10 μl of PCR buffer were incubated with 0.4 mg/ml proteinase K (Sigma, St. Louis, MO) for 1 h at 55°C, and the enzyme was inactivated by heating at 95°C for 10 min. Primer extension preamplification employing random 15 mers and 60 rounds of amplification with Taq polymerase (Promega, Madison, WI) was performed to produce sufficient DNA for multiple subsequent DNA amplifications. Rearranged VλJλ and VHDJH genes were then amplified, as described previously (23, 24).
PCR products were separated by electrophoresis on a 1.5% Seakem agarose gel (FMC Bioproducts, Rockland, ME) and purified using GenElute agarose spin columns (Supelco, Bellefonte, PA). Purified products were directly sequenced using the ABI Prism Dye Termination Cycle Sequencing kit (Perkin-Elmer, Foster City, CA) and analyzed with an automated sequencer (ABI Prism 377; Applied Biosystems, Foster City, CA). For identification of the germline Vλ and VH gene segments, the V BASE Sequence Directory (26) was used in conjunction with the software programs Sequencher (Gene Codes, Ann Arbor, MI) and Gene Works (release 2.45; IntelliGenetics, Mountain View, CA). Vλ, Jλ, VH, D, JH gene nomenclature was adopted according to the V BASE Sequence Directory.
A rearrangement was considered productive if the VλJλ junction maintained the reading frame into the Jλ segment. Rearrangements that failed to maintain the reading frame into the Jλ gene segment (n = 66), and those rearrangements that introduced a stop codon at the junction during the rearrangement process (n = 4) were considered as nonproductive. Rearrangements that involved pseudogenes were always considered nonproductive (n = 9). At the junctions, sequences were considered to be germline if at least two contiguous nucleotides matched the germline sequence. Junctional additions were scored as either inverted repeats at full-length coding ends (P nucleotides; P), inverted repeats at nucleolytically processed coding ends (Pr nucleotides; Pr), or nontemplated junctional additions (N nucleotides; N). In cases in which a nucleotide could be scored as either a P or a Pr, it was scored as a P. When junctions without N additions contained repeated nucleotides that could not be unequivocally assigned to either coding end, these were assigned as junctional microhomologies (H).
A total of 184 fetal sequences was obtained from three fetal spleens all at 18 wk of gestational age. For comparison, 227 adult sequences that had been published previously (23) were used. In brief, adult sequences were obtained from two healthy normal male donors (26 and 45 years old) and included two populations, CD19+/IgM+/CD5+ or CD5− B cells. In the analysis, clonally expanded rearrangements in the productive repertoire were considered as the same sequence. A total of 154 fetal sequences, including 84 nonproductive and 70 productive rearrangements, and a total of 201 adult sequences, including 146 productive and 55 nonproductive rearrangements, were analyzed. The numbers derived from the adult sequence analysis were calculated after clones were removed, and hence are different from what was previously published (23).
Estimation of Taq-polymerase error rate
The maximal PCR error rate for this analysis has been documented to be 1 × 10−4 (23). Thus, few, if any, of the nucleotide changes encountered in this analysis can be ascribed to PCR amplification errors.
χ2 tests were used to compare the gene usage and junctional nucleotide differences found in the productive and nonproductive repertoire as well as between fetal and adult repertoires. The goodness-of-fit χ2 test was used to assess differences between the observed frequencies of Vλ genes as would be expected by random usage based on the number of Vλ genes known to be in the genome. Values of p ≤ 0.05 were considered to be statistically significant.
Fetal spleen consists of mostly IgM+ B cells
In the fetal bone marrow, CD19+ IgM− pro/pre-B cells formed the major population, comprising 52% of the B cell population, whereas in fetal spleen, CD19+ IgM+ B cells constituted the major B cell population, comprising 95% of the total B cell population (Fig. 1). The ratio of IgM− pro/pre-B cells to IgM+ B cells was 1.1:1 in the fetal bone marrow compared with 1:19 in the spleen. For the current study, the IgM+ B cell population from fetal spleen was sorted for Ig repertoire analysis.
Nonrandom pattern of λ rearrangements in the fetal repertoire
Because nonproductive rearrangements fail to give rise to functional proteins, the frequency of gene segments observed in the nonproductive repertoire should reflect the relative frequency of rearrangement of each individual gene segment in the absence of any putative selection. To determine whether the rearrangement process itself is different between fetus and adult, the distribution of nonproductive VλJλ rearrangements in fetal splenic B cells was compared with that in adult donors (Fig. 2, upper panel). Although the usage of individual Vλ gene segments in the fetus and the adults showed some differences, the general rearrangement pattern of nonproductive Vλ gene segments was very similar between the fetus and the adult. Of a total of 51 functional Vλ genes and pseudogenes, 19 were detected in the fetal repertoire compared with 18 in the adult repertoire. These included 2 pseudogenes in the fetal repertoire (2A1, 5A) and 4 pseudogenes (2A1, 3A2, 7C, and 5A) in the adult repertoire. Of the functional Vλ genes detected, 2A2, 2B2, and 1G were significantly overrepresented in the nonproductive repertoire, in both the fetus and the adult, compared with the expected frequency based on random usage. However, 4B, which was overrepresented in the adult, was not detected in the fetus, and 6A, which was not overrepresented in the adult, was overrepresented in fetus. When particular Vλ gene segment usage was compared, only one gene, 4B, which is the most distal Vλ gene from the Jλ-Cλ region, was rearranged significantly less frequently in the fetus compared with the adult (p = 0.005). Jλ segment use in the nonproductive repertoire showed similar patterns in both fetus and adult (Fig. 3). Jλ7 was used most often (61.4% vs 60%), followed by Jλ2/3 (35.7% vs 34.5%), and Jλ1 (2.9% vs 5.5%). These results demonstrate that a similar repertoire bias is introduced in VλJλ rearrangements in both the fetus and the adult, and nonrandom use of Vλ and Jλ gene segments, as shown in the adult, is already present at the fetal stage of development.
Chromosomal position influences Vλ gene rearrangement during the fetal stage of development
To investigate whether chromosomal position can affect the rearrangement of Vλ genes during the different stages of development, we analyzed Vλ gene usage in nonproductive rearrangements according to Vλ gene clusters. The Vλ genes are divided into three clusters, clusters A, B, and C, with cluster A being most Jλ proximal and cluster C being most Jλ distal (27). Frequencies of cluster A, B, and C Vλ genes in the nonproductive repertoire of fetus were comparable with the adult (Fig. 3). When individual Vλ gene segment usage was analyzed according to Vλ gene clusters, rearrangement of Vλ genes in the fetus occurred throughout the entire Vλ locus, as shown in adult, except for the two most distal genes in cluster C, 8A and 4B (Fig. 2, upper panel). Observations that the two most distal Vλ genes in cluster C were less frequently rearranged in the fetus suggest that some positional effects influence Vλ gene rearrangement during fetal stages of development.
The expressed fetal λ gene repertoire is restricted and shaped by positive selection
The distribution of productive (and thus potentially selectable) VλJλ rearrangements was analyzed to define differences in the expressed repertoires between the fetus and the adult (Fig. 2). The distribution of Vλ gene segments in the productive repertoire was found to differ considerably between the fetus and the adult. Relative restriction of Vλ gene segment usage was noted in the fetal repertoire. Of the functional Vλ genes expressed, five Vλ genes, 2E, 2A2, 1C, 1G, and 3L, were found more often than expected in the fetal productive repertoire, whereas seven genes, 3R, 2A2, 3H, 2B2, 1C, 1G, and 1B, were found more often than expected in the adult. A number of Vλ genes were significantly less frequent in the fetal productive repertoire compared with the adult, including 3R and 1B, whereas 2E and 3L were more frequent in the fetus. Certain Vλ gene segments within each Vλ family were noted most frequently in both fetus and adult. However, 1G was the most frequent gene in the Vλ1 family, 2E in the Vλ2 family, 3L in the Vλ3 family, and 4A in the Vλ4 family in the fetus, compared with 1B, 2A2, 3H, and 4B, respectively, in adult. Overall, a few Vλ genes were overexpressed by a large number of cells in the productive repertoire of the adult, and this tendency was more marked in the fetus.
Each Vλ gene was analyzed for differences in its distribution between the productive and nonproductive repertoires to determine whether they were influenced by selection (Fig. 2). In the fetus, no functional gene was found at a significantly lower frequency in the productive than nonproductive repertoire. Thus, no Vλ gene was negatively selected in the fetal repertoire, whereas 4C was negatively selected in adult (p = 0.032). In the adult, no positive selection was observed. However, in the fetus, 2E was found at a significantly higher frequency in the productive than nonproductive repertoire (p = 0.025), suggesting that it was positively selected.
Junctional diversity in the fetal repertoire is influenced by positive selection
There were no significant differences in the complexity of the VλJλ junctions assessed by the frequency of nucleotytic processing at the coding ends, P nucleotide addition, and N nucleotide addition between the fetal and the adult nonproductive repertoires (Fig. 4). When productive and nonproductive junctions in the fetal repertoire were compared (Fig. 4), evidence of selection based on combinatorial diversity was obtained. Junctions with untrimmed ends, P nucleotides, Pr nucleotides, and N nucleotides, were all comparable in the fetal productive and nonproductive repertoires. However, in junctions without N additions, microhomology-mediated joining (H joining) was observed significantly more frequently in the productive rearrangements in fetus compared with nonproductive rearrangements of the fetus (77.5% vs 28.6%; p = 0.00002), suggesting positive selection. No similar difference was observed between productive and nonproductive rearrangements in the adult (55.7% vs 47.3%).
Both fetal and adult IgM+ B cells are enriched in cells expressing identical λ-light chains
When fetal and adult VλJλ junction sequences from productive rearrangements were analyzed in detail, it was noteworthy that numerous identical junctions in both the fetus and adult were observed (Figs. 5 and 6). In the fetus, 19% of the total productive rearrangements were represented more than once. It was particularly striking that nine identical versions of the 3L-Jλ7 rearrangements, 5-3.3.1H, were detected. In the adult, although the frequency was lower than in the fetus, 13.7% of the total productive rearrangements were found more than once. Notably, however, 75% of the identical junctions in the adult were found within the CD5+ population, whereas the frequency of identical λ-light chain sequences was much lower in the CD5− population. The frequent occurrence of identical Vλ sequences in the productive repertoire was in contrast to their significantly lower frequency in the nonproductive repertoires in both the fetus and the adult (5.7% in each) (Fig. 7,A). Notably, 43.8% of the productive rearrangements in the fetus with identical junctions used H joining compared with 15% in adult (p = 0.018) (Fig. 7,B). This compared with 0/4 and 1/3 of the nonproductive rearrangements in the fetus and adult, respectively, that employed identical junctions with H joining. These findings indicate that fetal B cells expressing identical productive Vλ-chains are much more likely to use H joining. To determine whether B cells with identical λ-chains were truly clonal, rearranged heavy chain genes were amplified from the same cells that exhibited identical λ-chains. In eight of sixteen sets of fetal B cells with identical λ-light chains, we were able to amplify productive VHDJH rearrangements from at least two members of the set. Although λ-light chains were identical, heavy chain rearrangements were different in seven of eight. The exception was 5-3.3.5D and 5-3.3.6G, which both employed VH3-07/D7-27/JH3 along with Vλ3L/Jλ7 (Fig. 8). However, five additional B cells expressing an identical arrangement involving Vλ3L employed different VHDJH rearrangements. These findings indicate that the fetal IgM+ B cells with common Vλ rearrangements are not clonal with regard to heavy chain use.
Complementarity-determining region 3 (CDR3) length is tightly regulated in λ rearrangements
The average CDR3 length of λ rearrangements was similar in the fetus and the adult. The mean (±SEM) length of the CDR3 of productively rearranged VλJλ genes in the fetus was 31 ± 0.3 bp compared with 31.5 ± 0.2 bp in the adult. The CDR3 length distribution also showed a similar Gaussian distribution pattern in both the fetus and the adult productive rearrangements, with a range of 24–39 bp in both groups (Fig. 9, upper panel). In the nonproductive rearrangements, the distribution of CDR3 length maintained a similar Gaussian distribution pattern as was seen in the productive rearrangements, but they were somewhat more broadly distributed, with a range of 14–45 bp in the fetus and 23–45 bp in the adult (Fig. 7, lower panel). The mean (±SEM) length of the CDR3 of nonproductively rearranged VλJλ genes was 31.5 ± 0.6 bp in the fetus and 32.5 ± 0.6 bp in the adult, not significantly different from that of the productive rearrangements.
Peripheral receptor editing is apparent in the fetal repertoire
To address whether receptor editing played a role in shaping the fetal repertoire, the relationship between the usage of the Vλ cluster and Jλ segments was analyzed. Receptor editing/replacement would be expected to produce bias in the frequency of association of the most 5′ cluster C Vλ genes to the most 3′ Jλ gene, Jλ7, more reflected in the nonproductive rearrangement compared with the productive rearrangement. In the fetus, 90% of cluster C Vλ genes were associated with the Jλ7 gene segment in the nonproductive rearrangements and 80% in the productive rearrangement, whereas in the adult, 66.7% of cluster C Vλ genes associated with Jλ7 in the nonproductive rearrangements and 88.2% in the productive rearrangements (Fig. 3). These data are consistent with the conclusion that there is more extensive receptor editing/replacement in the fetus compared with the adult.
Ab repertoire development during fetal life has unique characteristics compared with postnatal life, because it is a period when exposure to external Ag has not yet occurred and immune recognition is mainly directed to autologous Ags. Three distinct characteristics of the fetal repertoire have been described, namely restriction in Ag specificities, low avidity, and multireactivity to self Ags. Previous reports have indicated that restrictions in Ag specificities are achieved by limited and preferential usage of certain VH and Vκ gene segments (12, 28). Preferential gene use reflected the proximity of gene segments in chromosomal position (8, 9, 12). Fetal diversity was also reported to reflect limited diversification of the CDR3 region during early ontogeny. This was related to a relative paucity of N region additions (7, 29), a high frequency of homology-directed recombination (17), and generation of shorter CDR3 domains during the fetal life (30). It has been suggested that the restricted diversity of the fetal Ig repertoire predisposes to the generation of multireactive, low affinity self-reactive Abs at this stage of development (31). The physiologic impact of self-reactive, germline-encoded Abs remains obscure, but it has been suggested that self-reactivity of these Abs is an essential component for B cell survival and maturation (6), and mediates positive selection (22). In addition, cross-reactivity with exogenous microbes may play a role in host defense to the small number of pathogens that may threaten the survival of an individual during the perinatal period (32).
Most of the information on the fetal Ig repertoire derives from the analysis of the murine repertoire. Limited analyses of human VH and Vκ repertoires suggested that there is also restricted Ag specificity and a high degree of self-reactivity, although some of the mechanisms leading to these features may differ from those observed in the mouse. However, no information is available regarding the development and selection of the VλJλ repertoire in the human fetus. The current data provide a sufficiently detailed analysis of the VλJλ rearrangement in fetal spleen to assess the characteristics of the repertoire and delineate some of the influences that shape it.
By comparing the distribution of B cells in the bone marrow and the spleen of an 18-wk fetus, we were able to define that during the second trimester of gestation, the pro/pre-B cell population was mostly found in the bone marrow, whereas less than 5% of splenic B cells were this B cell precursor population. Most of the B cells found in the fetal spleen at this stage were surface IgM+ B cells. Our data confirm that very little, if any, B cell generation takes place in the second trimester fetal spleen, but rather, B cells generated in the bone marrow migrate to this organ to complete their maturation process (33, 34). By the second trimester of fetal development, the spleen functions mainly as a peripheral lymphoid organ as in postnatal life. Fetal spleen, therefore, serves as a good source for sampling the fetal B cell repertoire.
By analyzing nonproductive rearrangements from IgM+ B cells from fetal spleen and comparing them with the adult, we were able to assess possible differences in the rearrangement process involved in the development of the fetal VλJλ repertoire compared with the adult in a relatively unbiased way. One of the recombination mechanisms contributing to the generation of the rearranged VλJλ repertoire is preferential rearrangement of Vλ and Jλ gene segments. Analysis of the fetal nonproductive repertoire revealed that the nonrandom ultilization of the Vλ and Jλ segments in the fetal rearrangements was similar to that noted in the adult (23). Overrepresentation of 2A2, 2B2, 1G Vλ genes in the adult nonproductive repertoire was also observed in the fetal repertoire. Jλ7 was most frequently used of the four functional Jλ segments in both repertoires, followed by Jλ2/3 and Jλ1. However, notable differences were observed when individual Vλ genes were analyzed. Although the rearrangement of Vλ genes in the fetus occurred throughout the entire Vλ locus, we found that Vλ genes, 4B and 8A, were not rearranged in the nonproductive repertoire of the fetus. In the adult nonproductive repertoire, the 4B gene was not only detected, but it was the second most frequently rearranged gene. Cluster C genes are the most Jλ distal, located more than 600 kb upstream of the Jλ1 locus, and 8A and 4B are located at the most distal region of this cluster, about 800 kb upstream of Jλ1 locus (27, 35). The finding that the two most distal Vλ genes were not detected in the fetal nonproductive repertoire suggests that, as in the mouse, proximity of gene segments in the chromosome position plays a role in preferential VλJλ rearrangements early in human ontogeny. However, it is different in that this positional effect is not as profound as in mouse (8, 9, 36), because rearrangement of Vλ genes in the human fetus occurred throughout the entire Vλ locus, except the most distal two genes. These findings indicate that nonrandom gene segment use largely reflects a regulatory process that is intrinsic to the specific gene element and is independent of the stage of B cell maturation. Imposed upon this is a modest maturation-dependent influence of chromosomal location. Presumably, because of immaturity in the accessibility or promoter influence of Vλ genes, rearrangement to the most Jλ distal Vλ genes was restricted in the fetus, contributing to a modest developmental bias in the distribution of Vλ genes in the fetal nonproductive repertoire.
During recombination, diversity can be introduced at the VλJλ junctions by addition and deletion of a number of nucleotides (37). Analysis of the VλJλ junctions in the nonproductive rearrangements revealed no significant differences in the complexity of the VλJλ junctions between the fetal and adult repertoires. However, a trend toward restriction of the complexity of the VλJλ junctions was observed in the fetal repertoire. In the fetal nonproductive VλJλ junctions, biochemical events that occur at coding ends during recombination resulted in use of more germline-encoded sequences, in that there was less nucleotide processing at both the 5′ and 3′ coding ends, more use of germline-templated P nucleotides, and less use of nongermline-templated N regions. Although VλJλ rearrangements containing N segments were less frequent in the fetus compared with adult, 50% of fetal rearrangements contained N segments.
These observations indicate that the basic molecular mechanism governing the recombination of the λ-light chain genes is not significantly different between the fetus and the adult. Although there was some degree of restriction in diversity introduced during the VλJλ recombination in the fetus, this bias was very modest, implying that recombination mechanisms governing the adult V(D)J recombination is already established at the second trimester of fetal life.
The productive rearrangements, which encode functional protein, can potentially be influenced by selection. Comparison of the productive repertoire to the nonproductive repertoire allows for a determination of positive and negative selective influences that shape the Ig repertoire. Because of differences in the environment between the prenatal and the postnatal period, it is possible that different selection mechanisms are operative at these stages of life. It has been clearly demonstrated that negative selection events shape the primary repertoire of Ag-reactive B cells (38, 39, 40). Negative selection occurs upon encounter with self Ag and ensures elimination of autoreactive B cells that might have emerged during recombination. However, B cells with self-reactivity do occur in the primary Ab repertoire, although such processes seem to be restricted to early ontogeny and perhaps to a particular B cell subset (B-1 cells) (41, 42). Based on these findings, it has been suggested that B cells can be subjected to positive selection and maintained on the basis of their autoreactivity (22, 29, 42), but the role of positive selection in shaping the primary Ig repertoire is still unclear. During prenatal life, autoreactive Abs are abundantly generated, and this reactivity may serve as the basis of positive selection rather than negative selection. The current data demonstrate that positive selection is likely to contribute to shaping the fetal Vλ repertoire, whereas negative selection mainly shapes the adult Vλ repertoire.
In the adult, seven functional Vλ genes were overrepresented in the productive repertoire, but positive selection was not evident. Rather, negative selection of Vλ gene 4C was observed. These findings implied that this particular gene may have a propensity for autoreactivity and, thus, be deleted from the expressed repertoire (23). However, in the fetus, a smaller number of Vλ genes were overrepresented compared with the adult, but positive selection was apparent. The Vλ gene, 4C, which was negatively selected in the adult, was not apparently influenced by negative selection in the fetus, and the 2E gene, which was not overrepresented in the productive repertoire of the adult, was significantly overrepresented and also positively selected in the fetus. Because restriction of the repertoire was not found in the nonproductive repertoire, relative restriction of the fetal repertoire appears to result from selection. Thus, it is possible that restricted Vλ gene usage and selective use of certain Vλ genes in the fetus predispose to the generation of multireactive, low affinity, self-reactive Abs required for immunologic needs at this stage of development, and through positive selection, these autoreactive Abs may be maintained.
Junctional diversity of the fetal repertoire was also influenced by selection. In the nonproductive junctions, no significant differences in the complexity of VλJλ junctions were observed between fetus and adult. However, there was a trend toward use of more germline-encoded sequences at the junctions in the fetal repertoire. In the fetal productive repertoire, junctions showed a greater tendency toward diversification. The 5′ nucleotide processing was more frequent, P nucleotides were less frequent, and N nucleotide addition was more frequent in the productive repertoire compared with the nonproductive repertoire of the fetus. A striking finding was that junctions employing microhomology-mediated joining were significantly more frequent in the productive rearrangements of the fetus compared with fetal nonproductive repertoire, indicating that these junctions were positively selected. This selection was not observed in the adult. These results imply that the expressed fetal repertoire is both positively and negatively selected based upon junctional diversity. Selection of rearrangements with H joining causes major constraints on the junctional diversity mechanism observed in the productive repertoire, which appears to be the major outcome of selection in the fetus (17, 43).
One of the remarkable findings of this study was the identification of numerous identical VλJλ junctions in both fetus and adult. In the fetus, 19% of the total VλJλ productive rearrangements were represented more than once. Surprisingly, however, when rearranged heavy chain genes were analyzed from the same fetal B cells that expressed identical VλJλ junctions, most were disparate. Therefore, these fetal B cells were clonal at the λ-light chain locus, but obviously not typical clones when the heavy chain rearrangements were considered. A similar finding was made when the adult productive repertoire was analyzed with B cells identified that expressed identical λ-light chains but disparate heavy chain rearrangements (data not shown). These λ-light chain clones were significantly more frequent in the productive compared with nonproductive repertoire in the fetus and the adult, and significantly more frequent in fetal compared with adult productive repertoires, especially when only CD5− adult B cells were considered. These results imply that selection rather than bias in recombination played a major role in the appearance of the expanded λ-light chain clones, and this selection bias was more apparent in the fetus. These data imply that B cells expressing specific Vλ-light chain rearrangements are positively selected in fetal B cells and adult CD5+ B cells, regardless of heavy chain expression. These results suggest a unique Vλ-specific mode of B cell selection that expands B cells presumably by reacting with autoantigen. Precedence for such a mechanism of B cell selections derives from an analysis of avian B cells that exhibit extreme restriction of junctional diversity with predominance of one VL-JL joint as a result of developmental selection that takes place in the embryonic spleen (20). This embryonic selection process is not restricted to avian B cells, because similar junctional restriction was shown in Ig genes of neonatal murine B cells (7, 17). Based on the current data, developmental selection and expansion of certain VλJλ junctions also occur in the human fetus, and strikingly appear to be strictly based on sequences in the CDR3 domain of the λ-light chain, irrespective of the heavy chain. Because this occurs in the fetal spleen, the current data indicate that B cells undergo selective expansion based on the λ-light chain CDR3 independent of external Ag, and probably, therefore, mediated by self Ag (22, 44).
Selective B cell expansion based on only the λ-light chain in the fetus may reflect the nature of unique Ag-Ab interactions in the fetal life, potentially reflecting self Ag-mediated interactions essential for survival and maintenance of B cells and the need for creation of primitive immune response in the immediate postnatal period. In the adult, although the frequency was lower than in the fetus, expansion of B cells expressing specific λ-light chains was found in 13.7% of the total productive rearrangements. However, most of these cells were found within the CD5+ population in the adult. Previously, in adult mice, B cells with canonical receptors containing particular VH-D-JH/Vκ-Jκ junctions were reported in association with the B1 phenotype (22, 45, 46). In addition, selection of certain VH-Jλ-bearing clones into the mature long-lived B cell pool (46), and BCR-directed selection of VH81x-Vκ1C and VHS107-Vκ24 clones into the splenic marginal zone were recently reported (47). It is possible that λ-light chain clones found in IgM+ B cells in the fetus may persist to the adult stage as a long-lived B cell pool, and mainly remain restricted to the special subset of B cells expressing CD5 in adult life.
There were some differences between the expanded B cell populations with identical λ-light chains in the fetus and the adult. The first was the greater use of H joining in the fetus. H joining was found to be used significantly more frequently in the B cells with identical VλJλ rearrangements in the fetus compared with that of the adult. When the amino acid introduced at the junctions was analyzed (data not shown), it was noteworthy that two of the fetal B cells with identical VλJλ junctions introduced proline at the junctions, which was not found in the adult B cells with the same VλJλ junctions. Notably, the two fetal junctions that introduced proline were the result of joining by microhomology. By contrast, in the adult, three of the identical VλJλ rearrangements introduced glycine into the CDR3, which was not found in the fetal rearrangements. Introduction of proline at a junction causes constraints on ligand binding by tightly limiting the rotational degree of freedom at the Ag binding site, whereas glycine allows free rotations at the Ag binding site. Selective expansion of λ-light chains using H joining in the fetus may suggest more tightly regulated Ag reactivity during the fetal period of the development. This restriction of Ag reactivity in early life was shown to be essential for the acquisition of complete functional adult repertoire (48).
Except for the fetal B cells with identical rearrangements, CDR3 length of the λ-light chain was not selected. The average length and also the length distribution of the CDR3 region in both the productive and nonproductive repertoire remained unchanged during the ontogeny. Molecular mechanisms appear to impose strict limits on the CDR3 length of the λ-chain by tightly regulating the mechanisms governing the junctional diversity. It was shown in the human heavy chain that the length of the CDR3 region was considerably shorter in fetal B cells compared with those in adult (31) and a shorter CDR3 length was selected to restrict diversity in the fetal repertoire. However, this restriction in fetal CDR3 length was not profound in human κ-light chain (28) and clearly is not a feature of the λ-light chain repertoire. The difference found between the CDR3 restriction in the heavy chain and the light chains may be related to the greater range of lengths available to the CDR3 region in the heavy chain, because heavy chain rearrangement involves D segments in addition to V and J segments, and the very tight molecular regulation of CDR3 length seen in κ- and λ-light chains evident in the nonproductive rearrangements (23, 49).
During development, self-reactive B cells can be rescued from negative selection by up-regulating RAG-1 and RAG-2 and replacement of autoreactive BCR by secondary Ig gene rearrangements, i.e., receptor editing/replacement (50, 51, 52, 53). To maintain a B cell pool in the face of a changing environment continuously challenged with autoantigen as in fetal life, it is possible that receptor editing is actively revising the Ag receptor specificity to escape negative selection (54, 55). Evidence of receptor editing can be reflected in the Ig repertoire in the following two ways. First, because receptor editing is achieved by rearrangements of V genes 5′ to the initially rearranged productive VJ rearrangements to 3′ J elements (50, 52), bias in the association of the most 5′ V gene to most 3′ J gene would be evidence of secondary rearrangements. Because only about one-third of the secondary rearrangements are expected to be successful in producing productive joints, this bias would be reflected both in yielding and nonproductive repertoire, with more bias in the nonproductive repertoire if receptor editing were actively occurring. We found bias in the association of the most Jλ distal cluster C Vλ genes with Jλ7 in both the productive and nonproductive repertoires in the fetus, with more apparent bias in the nonproductive repertoire. Ninety percent of cluster C Vλ genes were associated with Jλ7 in nonproductive rearrangements, and 80% in the productive repertoire of the fetus, reflecting evidence of receptor editing in the fetus. However, this bias was not as apparent in the adult, in which 66.7% of cluster C Vλ gene associated with Jλ7 in nonproductive rearrangements and 88.2% in productive rearrangements. Appearance of secondary nonproductive rearrangements would depend on another secondary rearrangement that would introduce a productive light chain, usually a κ rearrangement. It is possible that secondary κ rearrangement is less efficient in the adult, so that bias in cluster C Vλ gene to Jλ7 association was seen mostly in the productive repertoire of the adult, whereas it was observed in both productive and nonproductive rearrangements in the fetus.
Second, evidence of receptor editing can be reflected in the productive to nonproductive ratio. Theoretically, if both alleles undergo ordered rearrangement, the productive to nonproductive ratio in the detected sequences should be close to 2.5:1. Because rearrangements are usually one-third productive and two-thirds nonproductive, any additional light chain rearrangement in which the nonproductive rearrangements can be retained will bring the productive to nonproductive ratio below the expected 2.5:1 ratio. We found that the productive to nonproductive ratio in the adult was close to the expected ratio (2.6:1), whereas the ratio in the fetus was 1:1, significantly lower than the expected ratio. This implies that receptor editing was very active in the fetus, leading to the introduction of additional nonproductive rearrangements that appeared in the repertoire presumably because of additional secondary productive rearrangement of the κ locus. This finding implies that receptor editing is more actively revising the Ag receptor specificity in the fetus compared with the adult to escape negative selection and maintain a functional B cell repertoire.
The current data demonstrate that three mechanisms strongly influence the shaping of the fetal λ-chain repertoire in human: receptor editing, positive selection, and expansion of B cells expressing specific λ-light chains. These events occur before exposure of the fetus to exogeneous Ag and suggest, therefore, that the fetal repertoire is strongly influenced by exposure to autoantigens.
We thank Angie Mobley and Michelle McGuire for excellent technical assistance. We also thank Dr. E. A. Padlan for helpful discussions.
This work was supported by the Harold C. Simmons Arthritis Resaerch Center. J.L. is a recipient of Korea Science and Engineering Foundation Grant 98-12-34-2. The nucleotide sequence data reported in this paper will appear in the EMBL, GenBank, and DDBJ Nucleotide Sequence database under the accession numbers AF247204 through AF247344.
Abbreviations used in this paper: BCR, B cell Ag receptor; CDR, complementarity-determining region; pBCR, pre-BCR.