A Role for DNA Polymerase μ in the Emerging DJ_H Rearrangements of the Postgastrulation Mouse Embryo

Abstract

The molecular complexes involved in the nonhomologous end-joining process that resolves recombination-activating gene (RAG)-induced double-strand breaks and results in V(D)J gene rearrangements vary during mammalian ontogeny. In the mouse, the first immunoglobulin gene rearrangements emerge during midgestation periods, but their repertoires have not been analyzed in detail. We decided to study the postgastrulation DJ_H joints and compare them with those present in later life. The embryo DJ_H joints differed from those observed in perinatal life by the presence of short stretches of nontemplated (N) nucleotides. Whereas most adult N nucleotides are introduced by terminal deoxynucleotidyl transferase (TdT), the embryo N nucleotides were due to the activity of the homologous DNA polymerase μ (Polμ), which was widely expressed in the early ontogeny, as shown by analysis of Polμ^−/− embryos. Based on its DNA-dependent polymerization ability, which TdT lacks, Polμ also filled in small sequence gaps at the coding ends and contributed to the ligation of highly processed ends, frequently found in the embryo, by pairing to internal microhomology sites. These findings show that Polμ participates in the repair of early-embryo, RAG-induced double-strand breaks and subsequently may contribute to preserve the genomic stability and cellular homeostasis of lymphohematopoietic precursors during development.

The adaptive immune system is characterized by the great diversity of its antigen receptors, which result from the activities of enzymatic complexes that cut and paste the genomic DNA of antigen receptor loci. The nonhomologous end-joining (NHEJ) machinery is then recruited to repair the double-strand DNA breaks (DSBs) inflicted by the products of the recombination-activating genes (RAGs) (45, 65). Within B cells, each immunoglobulin (Ig) receptor represents a singular shuffling of two heavy (H) and two light (L) chains, which are derived from the recombination of V, D, and J gene segments of the IgH locus and of V and J for IgL (71). Besides these combinatorial possibilities, most Ig variability derives from extensive processing of the coding ends, including exonucleolytic trimming of DNA ends, together with the addition of palindromic (P) nucleotides templated by the adjacent germ line sequence and of nontemplated (N) nucleotides secondary to the activity of the terminal deoxynucleotidyl transferase (TdT), a lymphoid-specific member of family X of DNA polymerases (reviewed in reference 56). During B-lineage differentiation, IgH rearrangements occur before those of the IgL locus, and D-to-J_H rearrangements precede V-to-DJ_H rearrangements (62). DJ_H joints are formed in any of the three open reading frames (ORFs). ORF1 is predominantly used in mature Igs, ORF2 is transcribed as a Dμ protein that provides negative signals to the B-cell precursors, and ORF3 frequently leads to stop codons (32, 33, 37). Germ line V, D, and J gene segments display short stretches of mutually homologous nucleotides (SSH), which are frequently used in gene rearrangements during perinatal periods, when N additions are absent (27, 32, 55, 57). The actual Ig V-region repertoires represent both the results of the NHEJ process associated with genomic VDJ recombination and those of antigen-independent and -dependent selection events. Although the core NHEJ components (Ku-Artemis-DNA-PK and XLF-XRCC4-DNA ligase IV) are by themselves able to join RAG-induced, incompatible DNA ends, family X DNA polymerases can be recruited to fill gaps created by imprecise coding ends with 3′ overhangs (DNA polymerase μ [Polμ] and Polλ) and/or to promote diversity through the addition of N nucleotides (TdT) (34, 56).

The lymphoid differentiation pathways and clonotypic repertoires are developmentally regulated and differ between the embryo-fetal and adult periods (2, 44, 68). The perinatal B cells result from a wave of B lymphopoiesis occurring during the last third of mouse gestation (13, 14, 21, 70). Perinatal V_H gene usage differs from that predominating in the adult (1, 69), and the former VDJ joints rarely display N additions, leading to V-region repertoires enriched in multi- and self-reactive specificities (36, 40). The program of B-cell differentiation starts at embryonic days 10 to 11 (E10 to E11) in embryo hematopoietic sites, after the emergence of multipotent progenitors (at E8.5 to E9.5) (18, 19, 23, 31, 51, 73). DJ_H rearrangements were detected in these early embryos, whereas full VDJ_H sequences were not observed before E14 (14, 18, 51, 66), when VJκ rearrangements were also found (63). The earliest mouse DJ_H/VDJ_H Ig sequences analyzed to date corresponded to late fetuses (E16) (14, 53). We reasoned that the true baseline of the Ig rearrangement process occurs in midgestation embryos, when the first DJ_Hs are not yet transcribed and, consequently, not subjected to selection and are conditioned only for the evolutionarily established and developmentally regulated usage of distinct NHEJ machineries.

We report here the sequence profiles of the earliest embryo E10 to E12 DJ_H joints. Unexpected frequencies of embryonic DJ_H joints bearing N nucleotides, in the absence of detectable TdT expression, were found. Moreover, the embryo DJ_H joints lacking N nucleotides (N⁻) used fewer SSH to recombine than newborn DJ_Hs, and these SSH were widely dispersed along the embryo D sequences, in contrast to the most joint-proximal ones, which predominated in newborn DJ_Hs. Considering that Polμ is the closest relative of TdT (42% amino acid identity) (22), which is able to introduce N nucleotides in vitro (4, 22, 34, 39, 49) and to join DNA ends with minimal or even null complementarity (17, 58), and that it is expressed in early-embryo organs, we decided to investigate its putative contribution to the first embryo DJ_H joints. The DJ_H joints obtained from Polμ^−/− embryos (48) showed a significant reduction of N nucleotides compared to wild-type (WT) embryos. Moreover, highly preserved DJ_H joints (with <3 deleted nucleotides) were selectively depleted in the Polμ^−/− mouse embryos, while the remaining DJ_Hs preferentially relied upon longer stretches of homology for end ligation. These findings support the idea that Polμ is active during early-embryo DJ_H rearrangements and that both its template-dependent and -independent ambivalent functions may be used to fill in small nucleotide gaps generated after asymmetric hairpin nicking and also to extend coding ends via a limited TdT-like activity.

MATERIALS AND METHODS

Embryo microsurgery and cell purifications.

BALB/c, BALB/c.RAG-2^−/−, and BALB/c × 129/SV Polμ^−/− (nine generations backcrossed to BALB/c) mice were maintained in the animal facilities of the Instituto de Salud Carlos III, Madrid, Spain, and the Centro de Biología Molecular Severo Ochoa, Madrid, Spain. The gestational age was determined by the vaginal plug after overnight mating (E0). Eggs containing the embryo attached to the placenta were washed by sequential transfers through 50-ml phosphate-buffered-saline tubes. After removal of the placenta, the embryo proper and yolk sac were separated and washed out again by passages through 20-ml petri dishes (51). All animal studies were performed in accordance with Spanish animal protection laws. The cells were incubated with anti-CD19-phycoerythrin (1D3), anti-CD45R/B220-allophycocyanin (RA3-6B2), anti-CD45-allophycocyanin (30F11), anti-cKit-fluorescein isothiocyanate (2B8), and anti-IgM-fluorescein isothiocyanate (331.13) antibodies (BD Pharmingen). Nonspecific background was eliminated by incubations with 10% mouse serum and Fc-Block (BD Pharmingen) and controlled with isotype-matched irrelevant antibodies. Cell debris and dead cells were discarded by adjusting light parameters and by propidium iodide staining. Labeled cells were purified in a DIVA cell sorter (BD Pharmingen), the degree of purity was controlled in a FACSCalibur with Cellquest software (BD Pharmingen), and only samples more than 95% pure were processed.

RT-PCR, genomic PCR, cloning, and sequencing.

Genomic DNA and total RNA were extracted, and oligo(dT)-primed cDNA samples were prepared with avian myeloblastosis virus reverse transcriptase (RT), as described previously (51). PCR amplifications were performed with 1 U of AmpliTaq Gold DNA polymerase (Roche Molecular Systems) in a PTC-200 DNA Engine cycler (Bio-Rad). DJ_H rearrangements were detected with a genomic amplification assay by using the Faststart PCR amplification kit (Roche Molecular Systems). DNA templates corresponding to 10³ cells were amplified. For embryo-derived DJ_H rearrangements, 1 μl of the first PCR amplification reaction mixture was subjected to a nested PCR for 20 additional cycles. The primers and PCR conditions were as indicated in Table S1 in the supplemental material. In the case of newborn-derived samples and to facilitate cloning of the larger DJ_H structures, nested amplifications were performed by using JH intron-specific oligonucleotides as 3′ primers. The products of the primary (for J_H4) or nested (J_H1 to -3) PCRs obtained from adult and newborn samples, containing different DJ_H rearrangements in each of the J_H-specific bands of the same size, were cloned using the pMBL-T vector kit (Dominion MBL) and transformed into JM109 Escherichia coli competent cells (Promega). After being plated, colonies were amplified, their products were separated electrophoretically on 2% agarose gels, and the bands were visualized with ethidium bromide and cleaned or gel purified with either PCR clean-up kits or gel spin kits (MoBio Laboratories), respectively. The purified bands were sequenced in an ABI7000 automatic sequencer with the BigDye sequencing mixture (Applied Biosystems).

The relative expression levels of TdT, Polμ, Polλ, and GαS genes were calculated by real-time PCR, performed on the LightCycler 2.0 system, by using the LightCycler FastStart DNA Master SYBR green I kit (Roche). The cycling steps were as follows: 1 cycle of 95°C for 10 min, followed by 45 cycles of 95°C for 10 s, 60°C for 10 s, and 72°C for either 15 s (TdT, Polμ, and Polλ) or 30 s (GαS), with a final cycle of 72°C for 5 min, before a melting curve was performed. LightCycler software 4.0 was used to calculate the threshold cycle (C_T) of each reaction, and the relative amount of specific cDNA on each sample was determined using the 2^−ΔΔCT method (46) by using GαS as the internal control gene. The values obtained for newborn samples were used as calibrators to determine the changes in the specific gene expression levels.

In vitro terminal-transferase assay.

Terminal-transferase-mediated incorporation of a single deoxynucleotide (ddTTP) was assayed as described previously (39) on double-stranded DNA molecules differing in the number of 3′ protruding nucleotides. The strand providing the 3′ overhang (5′-CGCAAGTCAGCGCTACGGG[T]_0-5) was 5′ labeled with [γ-³²P]ATP and T4 polynucleotide kinase and hybridized to the unlabeled complementary strand (5′-CCCGTAGCGCTGACTTGCG) in the presence of 50 mM Tris-HCl (pH 7.5) and 0.3 M NaCl. The reaction mixture contained 50 mM Tris-HCl (pH 7.5), 1 mM dithiothreitol, 1 mM MnCl₂, 5 nM of labeled DNA, 200 nM of either human Polμ or TdT, 100 μM ddTTP, 4% glycerol, and 0.1 mg/ml bovine serum albumin in 12.5 μl. After incubation (37°C; 30 min), reactions were stopped by adding gel-loading buffer (95% formamide, 10 mM EDTA, 0.1% [wt/vol] xylene cyanol, and 0.1% [wt/vol] bromophenol blue). The +1 extension products were analyzed by 8 M urea-20% polyacrylamide gel electrophoresis, detected by autoradiography, and quantitated with a phosphorimager.

Statistical analyses.

GraphPad Prism 3.0 software was used. Means, standard deviations, and standard errors of the means (SEM) were calculated, and the normality of the distributions was determined with the Kolmogorov-Smirnov test. Comparisons of means were performed with the two-tailed unpaired t test with the Welch correction, from which P values were derived. For nonparametric samples, one-way analysis of variance with the Kruskal-Wallis test was performed. Contingency tables were used to compare frequencies among experimental groups, using chi-square and Fisher's exact tests to calculate P values. Deletion profiles were obtained by Gaussian kernel density estimation (The R Project for Statistical Computing [http://www.r-project.org]).

Nucleotide sequence accession numbers.

The sequences analyzed in this study are posted as supplemental material (see Fig. S1 to S5 in the supplemental material) and have been deposited in EMBL with the following accession numbers: FM162406 to FM162492 (WT embryos); AM998795 to AM998802, FM161938 to FM161957, and FM161968 to FM161985 (newborns); FM161991 to FM162014 and FM162405 (adults); FM162395 to FM162397 and FM162493 to FM162508 (RAG2^+/− embryos); and FM162509 to FM162556 (Polμ^−/− embryos).

RESULTS

The first DJ_H rearrangements of the mouse embryo.

A genomic PCR simultaneously amplified DJ_H1 to -4 gene rearrangements and revealed DJ_H-specific bands in hematopoietic organs of synchronized E10 to E12 BALB/c mouse embryos, as described previously (51). Fifty-eight percent of the PCRs performed in samples derived from a mean of 10 embryo explants per experiment were positive, and of these, two-thirds showed a unique DJ_H band, suggesting that they represented the earliest-appearing rearrangements during postgastrulation periods (Table 1; see Fig. S1 in the supplemental material). From E13 onward, the DJ_H rearrangements logarithmically expanded in the liver (14, 51). Control DJ_H groups recovered from 1-day-old livers and spleens and from 3-month-old spleens were included (see Fig. S2 and S3 in the supplemental material). DFL16.1 was the most utilized D gene, although this preference was more pronounced among newborn DJ_Hs. All J_H genes were similarly used in embryo and newborn samples, but most adult sequences contained the smallest DJ_H4 rearrangement, which was more easily cloned when the four DJ_H rearrangements were present in the PCR samples. Both adult and newborn DJ_Hs preferentially utilized ORF1, whereas E10 to E12 embryo DJ_Hs showed roughly similar frequencies of the three ORFs, indicating their unselected character.

TABLE 1.

Characteristics of DJ_H joint sequences in BALB/c mouse embryos, newborns, and adults

Groupa	No. of sequences
	Embryo (n = 92)	Newborn (n = 57)	Adult (n = 25)
P-Sp/AGM	20
Liver	29	39
YS	33
Blood	10
Spleen		18	25
E10/11/12	24/50/18
DFL16b	28	45	7
DSP2	64	12	18
JH1	10	15	0
JH2	21	14	0
JH3	28	12	2
JH4	33	16	23
ORF1	40	36	14
ORF2	23	8	2
ORF3	29	13	9

^aP-Sp/AGM, para-aortic splanchnopleura/aorta-gonad-mesonephros region; YS, yolk sac.

^bTwo embryo DJ_H sequences and one adult DJ_H sequence used the DFL16.2 gene and were not included in the experimental analyses. The DFL16/DSP2 ratios were as follows: embryo, 0.36; newborn, 3.75; adult, 0.39.

E10 to E12 DJ_H coding joints include N nucleotides.

The processing of DJ_H joints is conditioned by at least four mechanisms: (i) usage of SSH in the recombination of two coding ends, (ii) exonucleolytic nibbling of germ line sequences, (iii) inclusion of templated P nucleotides, and (iv) random addition of N nucleotides by TdT. TdT is highly expressed in adult pro-B cells and hardly or not at all expressed in perinatal lymphoid progenitors (8, 44). In agreement with this, most of the adult spleen DJ_Hs (80%) and only 2 out of 57 newborn DJ_Hs (3.5%) showed N nucleotides. Unexpectedly, 51% of E10 to E12 DJ_Hs included N nucleotides (47 out of 92; P < 0.001 and P < 0.01 with respect to newborn and adult DJ_Hs, respectively) (Fig. 1, left). The lengths of embryo N stretches per N⁺ sequence were less than those of adult DJ_Hs (3.2 ± 0.3 and 4.6 ± 0.7 for embryo and adult DJ_Hs, respectively; mean ± SEM; P = 0.051). Considering that each exsanguinated embryo (and its dissected organs) was sequentially diluted in large buffer volumes (well over 10⁶ times its own volume), the possibility of surgical contamination with B cells from the adult mice was low. Alternatively, DJ_H joints could represent mother-derived chimerism, although only a transient wave of maternal macrophages has been detected in E7.5 to E9.5 embryo yolk sacs and no other cell chimerism in fetuses up to E16 (6, 52). Additional support for the embryo origin of the N⁺ DJ_Hs came from their unbiased ORF usage, while ORF2 was absent among adult N⁺ DJ_Hs (P < 0.01 compared to embryo ORF2 usage) (Fig. 1, right); the slight preference for ORF1 in total embryo DJ_Hs (Table 1) was actually accounted for by their N⁻ sequences (data not shown). However, to definitively discard a maternal origin of the N⁺ DJ_Hs, we sequenced DJ_Hs from embryos derived from BALB/c.RAG2^−/− females mated to BALB/c males. They also displayed an unselected distribution of ORFs, and 35% of these DJ_Hs showed stretches of N nucleotides (Fig. 1, left; see Fig. S4 in the supplemental material), confirming that the E10 to E12 N⁺ DJ_H sequences were of embryonic origin.

FIG. 1.

N nucleotides in DJ_H joints derived from E10 to E12 BALB/c embryos. On the left, the frequencies of N⁺ DJ_H joints in adults (black), newborns (gray), WT embryos (hatched), and RAG2^+/− embryos derived from BALB/c male × BALB/c.RAG2^−/− female matings (crosshatched) are shown. The distribution of ORFs (circled numbers) among N⁺ DJ_H joints of adults and embryo DJ_H joints is displayed on the right. *, P < 0.05; **, P < 0.01; ***, P < 0.001, ns, not significant.

Developmental shifts of SSH utilization, D recombination sites, and nucleotide deletions in DJ_H joints.

It is thought that the SSH present in D and J_H gene segments promote recombination whenever TdT is absent and N nucleotides are not introduced (28, 57). In agreement with this, 12% of adult and up to 78% of newborn DJ_Hs used such microhomologies. In contrast, both total and N⁻ embryo DJ_Hs showed lower usage of SSH-related recombination sites than the equivalent DJ_Hs of newborns (P < 0.001 and P < 0.01, respectively) (Fig. 2A). The analysis of D recombination sites also revealed distinctive patterns in both embryo and newborn DJ_Hs. The most joint-proximal nucleotide was preferentially used to ligate both embryo and newborn DJ_Hs, yet this recombination point was much more frequent in newborn than in embryo DJ_Hs (30/57 and 35/108 DJ_H joints, respectively, ligating at position 1 versus total DJ_H joints [BALB/c and RAG^+/− embryos are considered together hereafter]; P < 0.01). Compared to those of newborns, the embryos’ SSH-related recombination sites were distributed along the whole D segment sequence. When these SSH were classified between joint-proximal (first four D nucleotides) and joint-distal groups, the latter appeared more frequently in embryo than in newborn DJ_Hs (P < 0.01) (Fig. 2B, left). The coding-end sequences may influence the DJ_H joint-processing events (3, 24, 29, 43, 74); however, the observed embryo/newborn differences applied equally to DJ_Hs using the single DFL16.1 gene (P < 0.01) and to the DSP2 family members (P < 0.05) (Fig. 2B, right). The reduced numbers of DSP2 gene-specific newborn DJ_Hs precluded a gene-by-gene comparison with the equivalent embryo DJ_Hs. The increased utilization of joint-distal D recombination positions in embryo DJ_Hs resulted in larger DJ_H nucleotide deletions: Both DFL16.1⁺ and DSP2⁺ newborn DJ_H coding ends were more “protected” than the equivalent embryo DJ_Hs (P < 0.01) (Fig. 2C). We conclude that (i) embryo N⁻ DJ_H joints used lower frequencies of SSH than newborn N⁻ DJ_Hs, (ii) embryo DJ_Hs frequently rearranged to joint-distal SSH of D segments, and (iii) embryo DJ_Hs supported larger nucleotide deletions than newborn DJ_Hs.

FIG. 2.

Developmental shifts of SSH usage, of D recombination sites, and of nucleotide deletions in DJ_H joints. (A) Frequencies of usage of SSH-related recombination sites in total and N⁻ embryo and newborn DJ_Hs. (B) Embryo and newborn DJ_H joints recombine at SSH⁺ D sites. The DJ_H joints are arbitrarily located over the most joint-proximal SSH nucleotide; the progressively darker shades correspond to DJ_H joints using SSH 1, 2, 3, and >3 nt long; joint-proximal DJ_Hs are delimited between the two vertical dotted lines (upper left). The two upper-right bar graphs represent the distributions of D recombination sites in DFL16.1⁺- and DSP2⁺-specific SSH⁺ DJ_Hs (the nucleotide sequences are shown on the vertical axes, and for DSP2 genes, the numbers indicate variable nucleotide positions among the various gene family members). The bars below show the frequencies of joint-proximal DJ_Hs in total, as well as in D family-specific, embryo and newborn SSH⁺ DJ_Hs. (C) Degrees (means plus SEM) of nucleotide deletion in DFL16.1- and DSP2-specific embryo, newborn, and adult DJ_H joints. The symbols are as in Fig. 1, except for the shading of the upper bars in panel B.

High expression of Polμ in the early mouse embryo.

The expression of TdT, Polμ, and Ku80 genes in B-cell progenitors during mouse ontogeny was analyzed by conventional RT-PCRs in purified CD19⁺ IgM⁻ cells from adult bone marrow (BM) and newborn liver, together with cKit⁺ CD45⁺ cells from embryo hematopoietic sites, which included a small subset of CD19⁺ B-cell precursors (5%) (19). TdT was present in adult BM, expressed at highly reduced levels in newborn pre-B cells, and undetected in embryo cells (even when a nested RT-PCR assay was used) (data not shown). In contrast, Polμ transcripts were more conspicuous in embryo cKit⁺ CD45⁺ cells than in both adult and newborn samples. The NHEJ core Ku gene was also well expressed in adult, embryo, and newborn samples (Fig. 3A). To better quantify the relative expression of X family DNA polymerase genes, we undertook real-time PCRs for TdT, Polμ, and Polλ genes in the above-mentioned populations of adult BM, newborn liver, and embryo liver cells. Compared to newborn values, the adult TdT expression was 50- to 100-fold increased, while it was completely absent in the embryo samples. In contrast, Polμ relative transcript levels were 5- to 10-fold higher in the embryo than in the newborn and adult cell samples. Polλ expression was also slightly increased in the embryo samples (Fig. 3B).

FIG. 3.

Differential expression of TdT, Polμ, Polλ, and Ku80 in lymphohematopoietic progenitors during mouse ontogeny. (A) CD19⁺ IgM⁻ B-cell precursors were purified from adult BM and newborn liver, and cKit⁺ CD45⁺ cells were purified from the indicated embryo organs (YS, yolk sac; AGM, aorta-gonad-mesonephros region). Semiquantitative RT-PCRs were done as described in Materials and Methods. GαS gene expression was used as a reference for the total mRNA per sample. The negative and positive control samples were non-cDNA and adult thymus, respectively. (B) Normalized TdT, Polμ, and Polλ transcript levels relative to newborn 2^−ΔΔCT. The expression of each specific gene was normalized to that of the control GαS gene, and the values obtained for both adult and embryo samples are relative to those obtained for newborn samples. The data shown are the means plus standard deviations of at least three different samples per experimental group. The symbols are as in Fig. 1.

Polμ and TdT polymerases show different size preferences for extending double-stranded DNA with 3′ overhangs.

The finding of N additions in the embryo DJ_Hs when TdT was absent but Polμ was highly expressed prompted us to analyze the differential efficiencies of both family X polymerases at introducing N nucleotides in vitro into different double-stranded DNA substrates, as they might be found in vivo in V(D)J rearrangements. As seen in Fig. 4, blunt ends were inefficient substrates for both enzymes. TdT also could not extend short protruding ends (1 or 2 nucleotide [nt] residues) but was active on >2-nt-long 3′ protruding ends. Polμ showed the highest efficiency on DNA ends with small 1- or 2-nt overhangs, later declining to its lowest level at 4- or 5-nt overhangs, those in which the TdT activity was maximal. In absolute terms, the terminal-transferase activity of Polμ on its optimal substrates was about threefold lower than that of TdT on ends protruding 4 nt. A mechanistic explanation is that the formation of a precatalytic ternary complex for terminal-transferase activity is a rate-limiting process in the case of Polμ, which would thus avoid excessive N additions and consequently favor templated insertions (P. Andrade, F. J. Lopez de Saro, and L. Blanco, unpublished data).

FIG. 4.

Differential in vitro terminal-transferase activities of Polμ and TdT. The assays were done on various double-stranded DNA substrates, either blunt ended or having increasing numbers of 3′ protruding T residues. The percentage of extended DNA of each type was related to the length of the blunt-ended molecule and is represented in a combined histogram (black and gray bars for Polμ and TdT, respectively; means plus SEM; n = 3 independent experiments).

Polμ contributes to the processing of mouse embryo DJ_H joints.

The lack of TdT, together with the potential of Polμ to add N nucleotides in DSBs (see above) (34, 39, 58) and its involvement in V(D)J rearrangements (4, 5), prompted us to investigate its putative activity in embryo DJ_H joints by analyzing Polμ^−/− embryo mice. Fewer and late-appearing DJ_H rearrangements were detected in Polμ^−/− compared to WT embryos (25 and 67% of E11 DJ_H⁺ samples, respectively), suggesting that the V(D)J rearrangement process is less efficient in the mutant mice. Polμ^−/− embryo DJ_H joints were equally distributed between DFL16.1- and DSP2-bearing sequences, utilized all the JH gene segments, and showed an unbiased ORF distribution (see Fig. S5 in the supplemental material). They showed a significant decrease in N nucleotides relative to those of WT embryos, defined both as N⁺ DJ_H frequencies (13%; P < 0.001) and as the lengths of N stretches per N⁺ sequence (1.6 ± 0.4; P < 0.01) (Fig. 5A). Interestingly, the group of Polμ^−/− DJ_Hs included many embryos obtained from matings of mutant males with RAG2^−/− females to eliminate the maternal contribution. Only 2 out of 26 of these DJ_Hs showed N nucleotides (P < 0.05 with respect to the equivalent group of WT embryos derived from RAG2^−/− females).

FIG. 5.

Differential features of Polμ^−/− embryo DJ_H joints. (A) Frequencies of N⁺ DJ_H joints (left) and numbers of N nucleotides per N⁺ sequence (right) in WT (thickly hatched) and Polμ^−/− (thinly hatched) embryos. (B) Frequencies of usage of SSH⁺ DJ_H joints (left) and of ≥3-nt-long SSH (right) in WT and Polμ^−/− embryos. (C) Deleted nucleotides per DJ_H joint in WT and Polμ^−/− embryos and in newborns (a logarithmic scale is used in the x axis on the left). The arrows indicate the mean values for each group. The distribution profiles of DJ_H joints related to the extents of nucleotide deletions in WT and Polμ^−/− embryos and in WT newborns as defined by Gaussian kernel density estimations are shown in the right histograms (a linear scale is used in the x axis). The vertical lines delimit the highly preserved DJ_Hs (≤2 deleted nucleotides). (D) Frequencies of highly preserved DJ_H joints detected in perinatal pre-B/B, adult B1, and adult pre-B/B cells from healthy mice in an extensive literature review (n = 1,572 DJ_H joints from 27 series in 19 papers). Each bar corresponds to an independent series of DJ_Hs. The numbers above the bars refer to the original articles, which are listed in Table S2 in the supplemental material. The horizontal lines denote the mean of each group.

Embryo Polμ^−/− DJ_Hs relied more on the utilization of SSH for joining than those of WT embryos (P < 0.01), and the former SSH tended to be distinctively longer (39% and 19% of Polμ^−/− and WT embryo SSH, respectively, were >3 nt long; P = 0.081) (Fig. 5B). Also, fewer joint-distal SSH were used among Polμ^−/− than WT embryo DJ_Hs, which therefore more closely resembled WT newborn DJ_Hs (P = 0.09; data not shown). The average numbers of nucleotide deletions did not differ for WT and Polμ^−/− embryo DJ_H joints. However, we observed that a subset of highly preserved DJ_H joints (≤2 deleted nucleotides), which accounted for 17% and 25% of WT embryo and newborn DJ_Hs, respectively, were selectively depleted from Polμ^−/− embryo DJ_Hs (P < 0.05 and P < 0.01 compared to WT embryos and newborns, respectively) (Fig. 5C, left). The distribution profiles of DJ_H nucleotide deletions further revealed that the WT embryo DJ_Hs showed a bimodal pattern, with peaks centered on 5- and 11-nt-long deletions, whereas newborn DJ_Hs were predominantly limited to the less deleted subset. The Polμ^−/− DJ_H profile resembled that of newborn DJ_Hs in the rarity of highly deleted joints, although it differed from the other two in the selective absence of highly preserved joints (Fig. 5C, right). To elucidate whether the latter sequence subset was a consistent feature of every DJ_H repertoire, we undertook an extensive review of published DJ_H/VDJ_H joints from WT adult and perinatal mice (n = 1,572 DJ_H joints). All the analyzed series included a fraction of highly preserved DJ_H joints (15.2% ± 1.5%) (Fig. 5D) (the original references are shown in Table S2 in the supplemental material), suggesting that highly preserved DJ_Hs represent an intrinsic result of the V(D)J rearrangement process and that the activity of Polμ is essential for their generation. Taken together, the above-mentioned findings show that Polμ participates in DJ_H coding-end ligation in the mouse embryo by (i) introducing a limited number of N nucleotides, (ii) filling in small gaps in a template-dependent manner, and (iii) contributing to recombination with end-distal SSH sites of the D sequence.

DISCUSSION

The Ig V region repertoires differ between the perinatal and adult periods of life, with the former preferentially expressing D-proximal V_H genes (1, 69), showing distinct recombination sites (41), and, more importantly, lacking the diversifying potential of the N-nucleotide-producing TdT polymerase (8, 27, 32, 44, 55, 57). In the absence of TdT activity, the usage of germ line SSH in the recombination between most coding gene partners is predominant. B-lymphoid differentiation, however, starts much earlier, in the midgestation mouse embryo (14, 19, 51, 73). We have shown here that the embryo DJ_H joints have unusual characteristics, such as (i) short stretches of N nucleotides in the absence of detectable TdT expression, (ii) reduced utilization of SSH in embryo N⁻ DJ_Hs compared to those of newborn N⁻ DJ_Hs, and (iii) frequent DJ_Hs rearranging with joint-distal D-gene SSH, which results in extensive DJ_H nucleotide deletions. The embryo N⁺ DJ_Hs were also found among those recombined in ORF2 (which are strongly counterselected for in the adult) and, more importantly, in DJ_H RAG2^+/− embryos derived from RAG2^−/− mothers, demonstrating that they were of embryonic origin. Most adult-derived N nucleotides are introduced by TdT (30, 42), although a few N nucleotides remained in TdT^−/− mice and low levels of N⁺ V(D)J joints were also detected in non-TdT-expressing perinatal life periods (21, 63). In agreement with others, we could not detect any TdT transcript signal in embryo hematopoietic cells, which might account for the N nucleotides found. Also, when a limited pool of TdT^−/− embryo and adult DJ_Hs (n = 23 and 13, respectively) were sequenced, N nucleotides were not found in the adult, whereas a subset of N⁺ DJ_Hs were detected in the embryos, supporting the TdT independence of the embryo DJ_H N additions (M. A. R. Marcos and M. L. Gaspar, unpublished data). There is evidence of selection against the utilization of N nucleotides in neonatal B and T cells (7, 11, 12), as well as of positive selection for N⁻ IgHs among marginal-zone B cells (10) and for N⁻ public (26) and highly shared T-cell receptor β chains (38), suggesting that the levels of N⁺ V(D)J joints under TdT⁻ conditions (as is the case in the perinatal periods) might even be higher in the absence of antigen selection and that there is room for the activities of other polymerases (30).

The embryo N⁻ DJ_H joints did not show a predominant usage of SSH-related recombination sites (as happened in the newborn DJ_Hs), suggesting that SSH were initially used stochastically and that only afterwards were SSH⁺ DJ_H joints preferentially expanded, because SSH increased the efficiency of DJ_H rearrangements and/or because SSH⁺ DJ_Hs were positively selected for, as previously proposed (59). Whereas most SSH⁺ newborn DJ_Hs used joint-proximal D nucleotides, thus preserving the genomic sequence, a significant fraction of embryo DJ_Hs ligated to joint-distal SSH, resulting in large germ line nucleotide deletions. These highly deleted DJ_H joints were infrequent in later ontogenic periods, but they have been described in Bcl-XL transgenic mice, suggesting that B-cell precursors bearing “inefficient” V(D)J products were normally eliminated by apoptosis (25). Interestingly, the few V(D)J rearrangements rescued in DNA-PK-deficient mice also displayed significant nucleotide deletions that were joined in embryo but not in adult V-Jλ1 joints, implying that the NHEJ machineries acting in the two stages of life are different (43, 66). Both embryo and perinatal DJ_H⁺ B-cell precursors might represent independent compartments in which distinctive NHEJ complexes act. Alternatively, we are tempted to suggest that neonatal DJ_Hs are a fraction selected from those emerging in the postgastrulation mouse embryo, after negative selection of most ORF2-expressing N⁺ and highly deleted DJ_H joints and/or promotion for ORF1⁺ SSH⁺ DJ_Hs. As the embryo DJ_H patterns (particularly the presence of N nucleotides) were already undetectable in E16 DJ_Hs when IgH and the λ5/V preB-encoded surrogate L chain were present (51, 63), it is possible that differential pairing of the emerging IgHs with the surrogate L chain is involved in filtering out the embryos’ “inadequate” IgH chains during late fetal periods (54, 72) and subsequently gives rise to a select perinatal IgH repertoire.

The processing of V(D)J coding joints can be sustained by the core NHEJ components (Ku-Artemis-DNA-PK and XLF-XRCC4-DNA ligase IV) (34), but the family X polymerases (Polμ, Polλ, and TdT) further contribute, especially when the nick leaves 3′ overhangs, which are good substrates for them (45, 49, 58). There is a gradient of template dependence in the activities of X polymerases (56, 58), as Polλ requires an extensive alignment of the ends to exclusively catalyze template-directed gap filling (60, 61), whereas the lymphoid-specific TdT activity is incompatible with the template strand due to interference with its rigid loop 1 domain (20) and so is restricted to catalyzing N additions in V(D)J reactions. In contrast, Polμ has a more flexible loop 1, which allows ambivalent strategies (39): (i) it preferentially inserts templated nucleotides at small gaps created from overhangs with minimal or even null complementarities and (ii) it may also add nucleotides in a terminal-transferase-like fashion to create a “connector” that provides de novo microhomology. The latter N nucleotides would then be initially masked but, if elongated with a second N addition or with a templated insertion or used for promiscuous ligation, would be detected as bona fide N nucleotides. Polμ is biased to insert pyrimidines versus purines, but it is important to consider that any untemplated insertion of Cs and Gs would be favorably selected for upon end joining, due to the greater strength of G·C than A·T pairs. By analyzing Polμ^−/− embryos, Polμ was found to be responsible for most N nucleotides detected in the embryonic WT DJ_Hs, although a few N nucleotides still remained, which might be accounted for by either Polλ (also expressed in the embryo) or other activities. Some in vitro insights into the Polμ potential to catalyze N nucleotides (including the experiments shown here) have been produced; the results obtained in the embryo DJ_H joints represent (to our knowledge) the first in vivo evidence of the TdT-like Polμ activity. Polμ was also required for the formation of highly preserved DJ_H sequences (<3 deleted nucleotides), which are always present in every DJ_H repertoire observed in healthy mice. Another feature of Polμ^−/− embryo DJ_Hs was the relative reduction in sequences with large nucleotide deletions, which were observed in WT embryos. These simultaneous decreases in highly deleted and in highly preserved DJ_Hs (plus reduced N additions) in Polμ^−/− embryos explained why the mean deletion rates were similar in WT and Polμ^−/− embryo DJ_Hs, despite the fact that the latter became a more homogeneous group. Finally, more and larger recombination-related SSH were used in the Polμ^−/− than in the WT embryo DJ_Hs, likely representing a backup mechanism, as is observed when the classical NHEJ process is altered (15). It has been reported that the VJκ joints of adult Polμ^−/− mice showed increased nucleotide deletions (4). The authors did not observe any change in V(D)J_H processing, presumably due to insufficient Polμ expression in IgH-rearranging adult BM pro-B cells. The apparent discrepancy with our work may be accounted for by the different ontogenic origins of the samples and, in particular, by the higher expression of Polμ in embryo DJ_H-rearranging precursors, which could thus reach the threshold level required to participate in DJ_H joining.

A scenario for the variations in the V(D)J rearrangement process during ontogeny might be proposed as follows. RAG activity is switched on in the postgastrulation mouse embryo when it induces the first DSBs at immune-specific loci. During early embryo ontogeny, somatic cells proliferate very actively and are notably sensitive to genotoxic stresses. Subsequently, robust machineries are required to ligate imprecise and highly processed coding joints in order to preserve cell viability and genome stability (35, 64). The evolutionarily ancient X polymerase Polμ, which is part of the cellular response to DSBs (9, 50), is well equipped to connect these embryo DSBs and to avoid collateral genomic damage (Fig. 6). First, Polμ can use templated nucleotide insertion to allow complete or almost complete conservation of the germ line sequence (Fig. 6, middle left) (17, 58). In good agreement with this, we observed that highly preserved DJ_H joints disappeared from Polμ^−/− embryos. Secondly, Polμ's terminal-transferase activity can selectively target short DSB overhangs arising after hairpin nicking and opening (2-nt-long overhangs are the shortest nonblunt ends) (47), where it can introduce a few Ns, which may contribute to the pairing of DNA ends (Fig. 6, left) (17), particularly in the cases of C and G additions. Alternatively, after end bridging by Polμ, N nucleotides could also be inserted due to an imprecise orientation of the templating base that leads to an insertion error/point mutation, ready for ligation (Fig. 6, middle right). Finally, exhaustive processing of DNA ends, resulting in deletion and extensive 3′ overhangs, can still be handled by Polμ through its primer-realigning capacity (67, 75) and avidity for internal 5′ P, subsequently promoting pairing to internal SSH sites (16) at the cost of germ line sequence losses (Fig. 6, right). Challenged by strong evolutionary pressures to resolve cell-stressing DSBs in the early embryo, Polμ thus may use all its skills to repair the coding ends, even if many of the resulting products are “useless” and will be counterselected for later on. During the last third of gestation, N⁻ RF1⁺ DJ_H/VDJ_H rearrangements, which preferentially used genome-preserving joint-proximal SSHs, selectively expanded, a process that may be secondary to changes in the NHEJ complexes used and/or to selection for pairing to surrogate light-chain and/or (self) antigens. Both in the newborn and in the adult, Polμ will concentrate on the preservation of germ line coding ends by filling up small nucleotide gaps at those DSBs where it is recruited (IgL chain-encoding genes in the adult BM) (67, 75). The most recently evolved TdT will then be upregulated in adult pro-B cells, and its exclusive template-independent activity will be devoted to the generation of the extensive junctional diversity of IgH antibody chains. Finally, Polμ and TdT are not only compartmentalized during ontogeny and in different lymphoid differentiation stages, they also target distinct DSB DNA ends related to the extent of the 3′ overhangs produced by the nicking step, an event that might also influence the V(D)J repertoires during ontogeny.

FIG. 6.

A model for nucleotide insertions catalyzed by Polμ during embryonic DJ_H rearrangements. The Polμ structure (in gray) is depicted as two ellipses, the large one representing the polymerization core and the small one indicating the 8-kDa domain, which has a single-stranded DNA binding cleft for the 5′ P binding site. The triangle represents mobile loop 1 of Polμ (39, 56). The deoxynucleotide triphosphate (dNTP) binding site is represented by a square (magenta). The hatched areas are “out-of-frame” J segments. (a) Germ line genomic region with D and J segments of the IgH chain. (b) RAG-induced hairpinned coding ends at both D and J segments. A nicking site is indicated by a small triangle at each end. (c) After nicking by Artemis-DNA-PK, a 3′ overhang (2 nt in this case), which contains a P nucleotide, is generated at both ends. An internal 5′ phosphate is indicated by a small circle. (d) Binding of Polμ to DNA ends can occur in two different ways: an enzyme-DNA complex can be formed in which the 3′ protruding end is oriented as a primer strand whose primer terminus (P nucleotide) is close to the dNTP binding site (left); alternatively, the 5′ P of one DNA end is bound by the Polμ 8-kDa domain next to the dNTP binding site (center and right). (e to j, left) Terminal-transferase N addition occurs before end joining (e). The inserted N nucleotide can serve as a connector for end joining if it is complementary to the other DNA end and the synapsis is stabilized by loop 1 (f). The N nucleotide, which is now inapparent as a template-independent insertion, is further extended by insertion of a templated dXTP (g). Once this strand is ligated, Polμ is adjusted to the remaining gap, coupled to a conformational change of loop 1 (h). After gap filling and ligation (i), the end joining is completed and produces an “out-of-frame” J segment (hatched area) due to the two extra P nucleotides added at the junction. An “in-frame” product (shown below) could be obtained by the addition of two N nucleotides before the end-joining step. In this case, one of the Ns would be evident (j). (f to j, middle left) Polμ's loop 1-mediated synapsis without requiring N nucleotides. If the two P nucleotides are complementary, their extension (X) will be template directed at both gaps, and the resulting product will not include N nucleotides (an “out-of-frame” J segment; one P nucleotide added). (f to j, middle right) If the two P nucleotides are not complementary and the templating base (located in front of the dNTP binding site) is not properly adjusted, Polμ can add N nucleotides. One or more N nucleotides could be immediately ligated and detected in the final recombination product. (c to j, right) When the ends are heavily processed and have large single-stranded overhangs, Polμ can bind the internal 5′ P and attract a second 3′ overhang to its vicinity, based on its capacity to accept microhomologies. After gap filling and further processing of the flapped strand, a largely deleted product is obtained that could contain either “out-of-frame” or “in-frame” J segments.