Terminal Deoxynucleotidyl Transferase Expression During Neonatal Life Alters D_H Reading Frame Usage and Ig-Receptor-Dependent Selection of V Regions¹

The ability of the immune system to respond to the vast array of potential pathogens encountered during the life of an organism depends on the diverse repertoire of Ag receptors expressed by B and T lymphocytes. Ag-receptor diversity is created in large part by the process of V(D)J recombination (1), in which separate DNA elements are imprecisely joined to form the variable domains of Ag-receptor genes. The enzymatic machinery underlying V(D)J recombination is being elucidated. Key molecules in this process are the recombination-activating genes RAG-1 and RAG-2, as well as other molecules not solely specific for V(D)J recombination, including the DNA-dependent protein kinase (DNA-PK) complex and XRCC4 (2, 3). Terminal deoxynucleotidyl transferase (TdT)³ is an auxiliary enzyme that is not essential for recombination but is responsible for the majority of nontemplated (N) nucleotide additions between the joining DNA segments (4, 5, 6). Generation of functional Ig heavy (H) and light (L) chain genes through V(D)J recombination occurs in a stepwise process during B cell development in the primary lymphoid organs (7) .

A large body of evidence has accumulated indicating that the preimmune repertoire of Ig receptors is nonrandom in terms of both gene segment usage and characteristics of the DNA junctions (8, 9, 10, 11, 12, 13, 14, 15). Bias in the recombination process and/or cellular selection through Ig receptors are thought to account for the nonrandom nature of the Ig repertoire; however, the details of these selection mechanisms and the relative impact of each type of selection mechanism on the Ig repertoire have not been determined. The Ig H chain repertoire displays two interesting nonrandom characteristics that have been particularly well studied: 1) unequal usage of D_H reading frames (RF), and 2) overusage of the V_H81x gene segment.

D_H elements can be read in three different frames (designated RF1, RF2, and RF3), and in both forward and inverted orientations depending on how they join with the J_H element. However, examination of DJ_H and VDJ_H rearrangements made in vivo showed that most joins use RF1 in the forward orientation (12, 14, 16). Two nonexclusive hypotheses have been proposed to explain this bias in RF usage. Because DJ_H rearrangements in RF2 can give rise to expression of truncated μ protein (Dμ; 17 , one hypothesis is that cells making DJ_H rearrangements in RF2 are counterselected through a Dμ protein-surrogate L chain (SLC)-receptor complex (18, 19, 20, 21). Together with the fact that DJ_H rearrangements in RF3 frequently encode stop codons, this cellular selection mechanism could explain the predominant usage of RF1 in productive VDJ_H rearrangements. The second hypothesis suggests that the presence of short sequence homologies in the 3′ end of D_H elements and the 5′ end of J_H elements directs the choice of recombination sites such that RF1 predominates (14, 16, 22). Because the influence of sequence homologies on recombination site choice is most readily apparent in the absence of TdT (22), this mechanism might be expected to play a greater role in determining D_H RF usage during fetal life, where TdT is absent, than adult life, where TdT is present (23).

The V_H segment called V_H81x has been shown to be highly overused in B cell precursors, but rarely used in mature B cells. V_H81x is used in 20–40% of VDJ_H rearrangements isolated from early B cell precursors, but in only 4–4.5% of rearrangements from mature, peripheral B cells (10, 24). During adult B cell development, this decline in usage of V_H81x is accompanied by a progressive decrease in the ratio of productive to nonproductive V_H81xDJ_H rearrangements (25, 26, 27, 28), suggesting that V_H81x-encoded H chains are removed through cellular selection mechanisms. Consistent with this hypothesis, V_H81x-encoded H chains often fail to associate with SLC proteins to form the pre-B cell receptor (pre-BCR) (29, 30), which provides signals essential for B cell differentiation (31, 32). Interestingly, this selective disfavoring of productive V_H81xDJ_H rearrangements does not appear to operate during fetal/neonatal B cell differentiation, where a high productive to nonproductive (P/NP) ratio of V_H81xDJ_H rearrangements is still observed in mature B cell populations (27, 33, 34).

Here we report studies designed to assess the impact of TdT synthesis on molecular and cellular selection mechanisms operating during the Ig repertoire development at the endogenous IgH locus. We use TdT-transgenic (tg) mice (35) to determine the effect of enforced TdT synthesis on the neonatal repertoire of V_H81xDJ_H rearrangements. We provide evidence that the presence of N addition in the neonate interferes with both molecular selection processes influencing D_H RF usage and cellular selection processes influencing the P/NP ratio of V_H81xDJ_H rearrangements. Our results suggest that TdT has a wider influence in repertoire development than was previously appreciated; moreover, CDR3 seems to play a critical role in determining association with the pre-BCR and in shaping the repertoire.

TdT-tg mice were generated as described (35) and were maintained by brother-sister mating of heterozygotes at the Ontario Cancer Institute Animal Facility. As the generation of TdT-tg mice proved to be difficult, these data are based on a single TdT-tg line. Individual spleens were harvested at 1–2 days after birth, and DNA was prepared and typed for the presence or absence of the transgene by PCR, as described (35). Examination of the B lineage populations in neonatal and adult TdT-tg spleens by flow cytometry revealed no significant differences from controls (data not shown). DNA from two tg or three non-tg pups were pooled. We obtained μ membrane exon-targeted (μmT) mice (32) from the laboratory of Dr. Klaus Rajewsky (Institute for Genetics, Cologne, Germany) through Dr. Len Schultz (Jackson Laboratories, Bar Harbor, MN). Timed matings of homozygous μmT mice were conducted and fetal livers were harvested at day 16 of gestation (morning after mating = day 0). DNA was prepared from pooled livers from a single pregnancy. DNA was also prepared from bone marrow cells obtained from an adult (8-wk-old) μmT mouse.

We isolated VDJ gene segments from neonatal spleens of the TdT mice and from fetal livers of the μmT mice although it would have been as informative to keep the perinatal tissue the same. Because of breeding problems we were not able to obtain timed pregnancies for the TdT mice, whereas the μmT mice, which were maintained as a homozygous strain, proved to be relatively simple to obtain. However, because the information extracted from the perinatal μmT V_H81x rearrangements depends only on the absence of N addition (determined by using perinatal tissue) and the absence of receptor-mediated selection (determined by the μmT mutation), we were satisfied that the information should not be influenced by which perinatal tissue was used.

V_H81xDJ_H rearrangements were amplified from the indicated DNA samples as described (34). Briefly, total VDJ_H rearrangements were amplified for 30 cycles with a degenerate V_H primer (V_Hall) and a J_H4 primer, diluted 1/500 and then reamplified for 25 cycles with a V_H81x-specific primer and an internal J_H4 primer (J_H4IN). The amplified V_H81xDJ_H rearrangements were then cloned using the TA cloning kit (Invitrogen, San Diego, CA), according to the manufacturer’s protocols. Randomly picked clones were then sequenced using the T7 sequencing kit (Pharmacia, Piscataway, NJ) in conjunction with a sequencing primer specific for the 3′ end of V_H81x (5′-GACAATACCAAGAAGACC-3′).

D_H elements were identified by comparison to a list of published D_H sequences. D_H reading frames were classified according to the standard nomenclature (29, 31). Predicted amino acid sequences were determined using the DNA Strider software package (Service de Biochime, Gif-sur-Yvette, France).

To determine the effect of transgenic expression of TdT on the neonatal repertoire, we examined VDJ_H rearrangements in TdT-tg and control neonatal spleen. Spleens were harvested from TdT-tg and non-tg littermates at 1–2 days after birth, and genomic DNA was extracted. VDJ_H rearrangements using the V_H81x gene segment and the J_H4 gene segment were amplified from these DNA samples using PCR. The V_H81x gene segment was chosen for these studies because its genomic sequence is known (10), allowing the accurate identification of N regions, and because productively rearranged V_H81xDJ_H alleles are subject to interesting selection processes (25, 26, 28, 33, 34, 36), which we speculated might be influenced by TdT expression. The amplified rearrangements were cloned into a plasmid vector, and randomly selected clones were sequenced (Fig. 1).

Examination of the V_H81xDJ_H4 junctional sequences revealed the expected presence of N addition in the Tg samples and near absence of N addition in the non-tg samples (Table I and Fig. 1). While the frequency of N addition in the TdT-tg neonatal spleens is similar to that observed in adult tissues, the average length of the additions is approximately twofold lower (Table I).

In VDJ_H rearrangements isolated from both neonatal and adult mouse tissues, D_H elements are most often found to be joined to the J_H element such that they would be read in the RF designated RF1 (12, 14, 16). We compared D_H RF usage in V_H81xDJ_H rearrangements isolated from TdT-tg and non-tg neonatal spleen (Fig. 1). Rearrangements isolated from non-tg neonatal spleen show greater than 80% RF1 usage, confirming previous findings. However, the V_H81xDJ_H4 rearrangements from TdT-tg neonatal spleen show a significantly different pattern of D_H RF usage, with <30% using RF1 (Fig. 1). This result suggests that the presence of TdT interferes with the process(es) which establish biased D_H RF usage during neonatal life.

Although the effect on RF usage is apparent among nonproductive rearrangements (Fig. 1), it is still possible that TdT is influencing selection at the level of the Dμ receptor expressed from DJ_H rearrangements in RF2 (17, 18, 19, 20, 21). To distinguish Ig-receptor-independent effects from receptor-dependent effects we compared D_H RF usage in V_H81xDJ_H4 rearrangements isolated from μmT fetal liver or bone marrow cells (Figs. 2 and 3). These populations provide a sample of rearrangements generated in the absence or presence of TdT, respectively, that have not been influenced by Ig-receptor-mediated selection processes. It was found that >80% of the rearrangements from μmT fetuses use RF1 (Fig. 3), suggesting that establishment of D_H RF bias during fetal/neonatal life is not dependent on receptor-mediated selection processes. In contrast, rearrangements isolated from μmT bone marrow show a more random pattern of D_H RF usage (Fig. 3), as shown previously (18). These data indicate that TdT activity results in the alteration of D_H RF usage by an Ig-receptor-independent mechanism.

Selection of productive V_H81xDJ_H rearrangements differs in fetal and adult B cell progenitors in that fetal cells positively select these rearrangements while adult cells negatively select these rearrangements (27, 33, 34). Because TdT represents one gene known to be differentially expressed in fetal and adult B cell progenitors (23), we hypothesized that TdT may be a critical factor determining this differential selection of productive V_H81xDJ_H rearrangements. If this were the case, we would expect the P/NP of V_H81xDJ_H rearrangements in TdT-tg neonatal spleen to be different from in non-tg littermates. Indeed, when we compared the P/NP of V_H81xDJ_H4 rearrangements in non-tg and TdT-tg neonatal spleens, a striking difference was found (Fig. 4). Rearrangements from non-tg neonatal spleens have a high P/NP (2.4), as observed previously (33, 36). However, V_H81xDJ_H rearrangements from TdT-tg neonatal spleens exhibit a significantly lower P/NP (0.32), indicating a disruption of the positive selection for productive rearrangements. Thus, synthesis of TdT does appear to be a critical factor in determining the selection of functional V_H81xDJ_H rearrangements.

We next examined whether the effect of TdT synthesis on selection of productive V_H81xDJ_H rearrangements is due to an Ig-receptor-dependent or Ig-receptor-independent mechanism. We thus examined V_H81xDJ_H rearrangements isolated from μmT fetal liver or bone marrow to determine the P/NP generated in the absence or presence of TdT without the influence of Ig-receptor-mediated selection (Fig. 4). It was found that the rearrangements from μmT fetal liver and bone marrow both had a P/NP close to that which would be expected from random joining (0.3–0.5), indicating that the presence of TdT activity does not significantly alter the P/NP ratio of V_H81xDJ_H structures in the absence of Ig-receptor-mediated selection. Rearrangements from μmT fetal liver had a significantly lower P/NP than in non-tg neonatal spleen (Fig. 4), indicating that the high P/NP in normal neonates is dependent on Ig-receptor-mediated selection. In contrast, the P/NP observed in TdT-tg neonatal spleen was similar to that in μmT fetal liver, indicating that the Ig-receptor-mediated selection occurring in the neonatal spleen is largely abrogated by TdT activity. Together, these results suggest that TdT activity results in an alteration the P/NP ratio of V_H81xDJ_H structures by influencing an Ig-receptor-dependent selection process.

The observations on the effect of TdT activity on receptor-dependent selection of productive V_H81xDJ_H rearrangements suggest that ability of V_H81x to generate functional receptors is determined by the CDR3 sequence in such a way that N-less sequences more frequently give rise to functional receptors. Therefore, we examined the productive V_H81xDJ_H rearrangements positively selected in the neonatal spleen to determine whether they exhibit any restrictions in CDR3 amino acid sequence that could be attributed to receptor-mediated selection (Fig. 5). We observed a clear conservation in the four amino acids immediately adjacent to the V_H81x gene segment (positions 95–98 of the H chain; 37 . All (100%) of the sequences contained a histidine residue at the first position adjacent to the V_H81x segment (position 95). The amino acids at positions 96–98 also show striking conservation, with hydrophobic amino acids being nearly absent and glycine, tyrosine, asparagine, and serine comprising approximately 80% of the amino acids. Notably, positions 96 and 98 consist mainly of glycine, tyrosine, and serine, while position 97 differs in that 41% of sequences contains asparagine residues with the remainder mainly containing tyrosine or serine. By tabulating the most frequent amino acids at positions 95–98, a degenerate consensus sequence can be derived (Fig. 5). Over half (58.8%) of the sequences from non-tg neonatal spleen conform to this consensus sequence (marked with an asterisk in Figure 5) with most of the remaining sequences containing only a single nonconsensus amino acid. As expected, this consensus sequence is present at a lower frequency (16.7%) in TdT-tg neonatal spleen (Table II).

To determine the extent to which the sequence conservation identified is dependent on receptor-mediated selection, we determined the frequency with which the consensus sequence occurs among productive V_H81xDJ_H sequences from μmT fetal liver or bone marrow (Fig. 6,A). As another measure of conformity to the consensus sequence, the percent of amino acids deviating from the consensus sequence defined in Fig. 5 was also determined for the various samples (Fig. 6,B). None of the sequences from μmT bone marrow contained the consensus CDR3 sequence, suggesting that generation of this sequence is rare in the presence of TdT. In contrast, the consensus sequence was present among the sequences from μmT fetal liver; however, the frequency of rearrangements containing the consensus appears lower than that in the normal neonatal spleen (33.3% vs 58.8% of rearrangements; Fig. 6,A), and the percent of amino acids deviating from the consensus is significantly higher (39% vs 15%; Fig. 6 B). In addition, only 66.7% of the productive sequences from μmT fetal liver encode histidine at position 95 in contrast with 100% in the normal neonatal spleen. Thus, while the consensus sequence is more frequently generated in the absence of TdT activity than in the presence of TdT activity, there also appears to be a receptor-mediated selection for this sequence in the neonatal spleen.

Compilation of published V_H81xDJ_H4 sequences from adult bone marrow pre-B cells or adult spleen (25, 26, 27, 28, 33, 34) provides further evidence for receptor-mediated selection of the consensus sequence. Productive V_H81xDJ_H4 sequences derived from normal bone marrow pre-B cell populations confirm that the consensus sequence is rarely generated in the presence of TdT (0/8 sequences). However, 40% (2/5) of the productive sequences from adult spleen have the consensus sequence, despite the rare occurrence of these rearrangements in the bone marrow. The overall percent of amino acid deviations from consensus is also lower in the adult spleen sequences than in the bone marrow pre-B cell sequences (30% vs 56.3%). Together with our results, these data indicate that receptor-mediated selection results in a skewing of CDR3 sequences immediately adjacent to the V_H segment toward a hydrophilic consensus sequence.

It is well established that TdT plays an important role in generation of the Ig repertoire by giving rise to nontemplate-encoded nucleotides at the junctions between joining V_H, D_H and J_H segments (4, 5, 6). The differential regulation of N addition during fetal and adult B cell development thus provided compelling evidence for a specific molecular mechanism to “program” different Ig repertoires during fetal and adult life (12, 16). Here we demonstrate that this differential TdT expression has far reaching implications for Ig H chain repertoire generation, which go beyond the direct effect of catalyzing N nucleotide addition. With the caveat that the data emerged from a single mouse line, our results indicate that TdT activity affects the selection process at at least two stages: at the stage of V(D)J joining and at the stage of receptor-mediated selection.

Our study uses mice expressing TdT under the control of the N-myc promoter and the Eμ enhancer (35), which has been shown to give high level expression in lymphoid cells throughout their development (38). However, the overall level of N addition in the TdT tg neonatal spleen (Table I) and TdT-tg fetal liver (35) is significantly less than observed in normal adult spleen. Receptor-mediated selection against N-region-containing sequences cannot account for this difference because it is apparent among both productive and nonproductive rearrangements (Fig. 1). Thus, this lower level on N addition appears to result from lower TdT activity. It remains to be determined whether this is simply due to lower expression of TdT in the tg B cell progenitors or to differential posttranscriptional regulation of TdT activity in fetal vs adult B cell progenitors.

During B cell development in the adult bone marrow, it has been shown that D_H RF bias, and specifically suppression of RF2, is dependent on signaling through the Dμ protein/SLC complex (18, 19, 20, 21). In contrast, our data indicates that, during fetal B cell development, D_H RF bias is established independently of expression of μ-related proteins on the cell surface. Instead, it appears that during the fetal/neonatal period D_H RF bias is established at the level of V(D)J recombination by a mechanism that requires the absence of TdT. These results provide supporting evidence for the homology-directed joining model as the mechanism determining biased RF usage during fetal/neonatal life (14, 16, 22). Homology-directed joining has been shown to be essential for the establishment of “canonical” junctional sequences in the invariant γ/δ TCR generated during fetal life, and generation of such receptors is largely abrogated in the presence of TdT (39). Thus, D_H RF bias appears to be a second clear example of the influence of homology-directed recombination in generation of the restricted fetal repertoire. Given the existence of the Dμ-receptor-dependent mechanism for maintaining biased RF usage, it is surprising that this mechanism does not appear to maintain biased RF usage when homology-directed joining is disrupted in TdT-tg neonates. This could indicate that the Dμ-receptor-dependent mechanism is only operative during adult B cell differentiation; however, additional experiments will be required to clarify this issue. To determine whether the finding that premature TdT expression interferes with normal neonatal D_H RF bias may be generalizable to other V_H genes, we reexamined our data of TdT-tg fetal liver sequences reported in (35). Of 16 informative V_HJ558DJ_H sequences, eight had N additions, and eight did not. Of those without N, seven are in RF1 (88%); of those with N, three are in RF1 (38%). Thus, it may be a general finding that N addition changes the bias for RF1 usage.

Several studies have indicated a striking difference in positive selection of V_H81x-encoded H chains by fetal and adult B cell progenitors (25, 27, 34). We previously put forward two hypotheses to explain this difference (34): 1) fetal and adult B cell progenitors have different requirements for Ig-receptor-mediated selection, or 2) selection of V_H81x-encoded H chains is dependent on the V_H-D_H-J_H junctional sequence (CDR3 region) in such a way that the absence of N addition during fetal life (23) more frequently generates V_H81x-encoded H chains which have the structure required for positive selection of B cell progenitors. A recent study (40) using an in vitro assay to assess the impact of various transgenic H chains on fetal vs adult B cell development drew the conclusion that fetal and adult B cell precursors differ in their H chain selection requirements, consistent with the first hypothesis. However, the present data provide strong in vivo evidence in support of the second hypothesis, showing that positive selection of V_H81x-encoded H chains during fetal life is virtually abrogated by transgenic expression of TdT. And, as discussed above, homology directed joining is likely to be the mechanism responsible for those CDR3 regions that are productive. Thus, in the case of V_H81x, it appears that the fetal vs adult difference in TdT synthesis is sufficient to determine the fetal vs adult difference in selection of H chains.

Recent studies examining the ability of various μ H chains to form a pre-BCR complex (29, 30, 41) provide a probable mechanism for the CDR3-dependent positive selection of cells bearing V_H81x-encoded H chains. The initial repertoire of H chains generated in the bone marrow was found to contain both H chains capable of forming a pre-BCR complex and H chains incapable of forming this complex, in approximately equal proportions (29). However, the large majority of H chains isolated from later stages of pre-B cell development are capable of forming a pre-BCR, indicative of a positive selection based on pre-BCR formation (29). Several V_H81x-encoded H chains were tested for the ability to form a pre-BCR complex, and it was found that all N-region-containing V_H81x-encoded H chains tested fail to form a complex with SLC (29, 30). To date, the only two V_H81x-encoded H chains (42) found to form a pre-BCR complex are N-less joins isolated from fetal cells. Importantly, a transgene encoding a V_H81x H chain that has no N addition and that can form a pre-BCR (H.-M. Jack, unpublished observation) was shown to support B cell maturation (42), while a transgene encoding a V_H81x H chain that cannot form a pre-BCR failed to support B cell differentiation (30, 41). Thus, it appears plausible that the differential positive selection of V_H81x-encoded H chains with simple vs complex CDR3 regions is due to differential association with SLC.

Interestingly, the capacity of different V_H81x-encoded H chains to bind SLC appears to correlate with the ability to bind conventional Ig L chains (41). This indicates that pre-BCR formation may serve as a broad screen for “functionality” of H chains in terms of their ability to fold correctly to form Ig-like receptors (30, 41). In light of this view, it is interesting that the “functionality” of H chains encoded by V_H81x appears to be extremely sensitive to variations in the CDR3 region. We speculate that H chain proteins encoded by V_H81x may be inherently unstable and only form a “functional” Ig fold when joined to sequences encoded by simple, conserved DJ_H joins that stabilize the Ig domain. This instability could be due to one or more of the unusual amino acid substitutions found in V_H81x (10). Thus, the V_H81x segment may be structurally specialized for selective expression in the TdT-negative repertoire generated early in life. It remains to be determined whether other such structurally specialized V_H segments exist in the mouse or human.

We thank Queenie Lam and Stacy Olson for technical assistance, and other members of the Wu and Paige laboratories for their companionship and helpful discussions. We thank Dr. Hans-Martin Jäck for discussions and for revealing data before publication.