An in vivo study of the impact of deficiency in the DNA repair proteins PAXX and XLF on development and maturation of the hemolymphoid system

Repair of DNA double-strand breaks by the nonhomologous end joining pathway is central for proper development of the adaptive immune system. This repair pathway involves eight factors, including XRCC4-like factor (XLF)/Cernunnos and the paralog of XRCC4 and XLF, PAXX nonhomologous end joining factor (PAXX). Xlf−/− and Paxx−/− mice are viable and exhibit only a mild immunophenotype. However, mice lacking both PAXX and XLF are embryonic lethal because postmitotic neurons undergo massive apoptosis in embryos. To decipher the roles of PAXX and XLF in both variable, diversity, and joining recombination and immunoglobulin class switch recombination, here, using Cre/lox-specific deletion to prevent double-KO embryonic lethality, we developed two mouse models of a conditional Xlf KO in a Paxx−/− background. Cre expressed under control of the iVav or CD21 promoter enabled Xlf deletion in early hematopoietic progenitors and splenic mature B cells, respectively. We demonstrate the XLF and PAXX interplay during variable, diversity, and joining recombination in vivo but not during class switch recombination, for which PAXX appeared to be fully dispensable. Xlf/Paxx double KO in hematopoietic progenitors resulted in a shorter lifespan associated with onset of thymic lymphomas, revealing a genome caretaking function of XLF/PAXX.

vating genes 1 and 2 (RAG1/2) during somatic DNA rearrangement (V(D)J recombination) of variable (V), diversity (D), and joining (J) elements of the T cell receptor and immunoglobulin genes in T and B cells in the thymus and bone marrow (BM), respectively (2,3). In addition, immunoglobulin class switch recombination (CSR), which occurs during the terminal maturation of B lymphocytes in the spleen upon antigen triggering and results in exchange of the IgM constant region C for a downstream C H gene (␥, ⑀, or ␣) also proceeds through introduction of prDSBs by the activation-induced cytidine deaminase at switch regions (S, S␥, etc.) flanking C H genes (4,5).
The nonhomologous end joining (NHEJ) machinery is the DNA repair pathway that copes with these lymphoid-specific prDSBs. Briefly, the NHEJ machinery is composed of several factors, including Ku70/80, DNA PKCs, Artemis, DNA ligase IV, XRCC4, MRI, XLF, and PAXX, allowing recognition, processing, and religation of broken DNA ends (6,7). Loss of function of many of these core factors leads to severe combined immunodeficiency in mice and humans because of arrest of B and T cell maturation as a result of abortive V(D)J recombination (8,9). In the case of XRCC4 or ligase IV, it results in late embryonic lethality caused by massive apoptosis of postmitotic neurons (10 -13). CSR is also affected to various extents by defective NHEJ (14,15). XLF and PAXX appear to be an exception because their loss of function only marginally affects V(D)J recombination; Xlf or Paxx KO mice present with a mild immunophenotype characterized by a slightly reduced number of B and T lymphocytes (16 -21). We and others identified functional relationships between XLF and other actors of V(D)J recombination that account for this peculiarity, such as ataxia telangiectasia-mutated (ATM) (22,23) and the C terminus part of RAG2 (24). Indeed, Xlf/Atm or Xlf/Rag2 cc double-mutant mice lack B and T lymphocytes because of impaired V(D)J recombination (22,24). The synthetic embryonic lethality between Paxx and Xlf also argues for a functional link between these two factors (19,20). Although Paxx Ϫ/Ϫ mice do not present an overwhelming phenotype in general and in the immune system in particular, the increased sensitivity of Paxx-deficient cells to ionizing radiations in various settings corroborates implication of PAXX in DSB repair (19,21,25,26). Moreover, E18.5 fetuses recovered from Xlf/Paxx double KO (DKO) mice revealed complete block of B and T cell development (19).
To better understand the role of PAXX in DSB repair during lymphocyte maturation and function and its interplay with Xlf in  cro ACCELERATED COMMUNICATION these processes in an in vivo setting, we developed two models of Xlf conditional KO in Paxx Ϫ/Ϫ mice to bypass the embryonic lethality of the DKO. Cre expression under the iVav and CD21 promoters allowed tissue-specific deletion of Xlf in hematopoietic precursors or only in splenic mature B cells, respectively.

Generation of conditional Xlf KO in Paxx ؊/؊ mice
To analyze the combined role of PAXX and XLF during V(D)J recombination and CSR, we crossed Paxx Ϫ/Ϫ /Xlf Ϫ/flox mice with transgenic mice expressing the Cre recombinase under the iVav and CD21 promoters, respectively. In the iVav model, Cre recombinase is first expressed in fetal liver and then in the bone marrow, spleen, thymus, and lymph nodes at postnatal stages (27), allowing specific deletion of Xlf in all hematopoietic lineages. In the ⌬CD21 model, the Cre recombinase is expressed only in mature B lymphocytes (28), resulting in B cell-specific deletion of Xlf. For the two models, mice were born at a Mendelian ratio and did not present any morphological defects at birth, confirming bypass of the embryonic lethality seen in the constitutive Paxx/Xlf DKO. From this point on, the Paxx Ϫ/Ϫ Xlf Ϫ/fl iVav-Cre and Paxx Ϫ/Ϫ Xlf Ϫ/fl CD21-Cre models will be indicated as Xlf⌬iVav and Xlf⌬CD21, respectively. PCR analyses confirmed the effective deletion of Xlf in purified splenic mature B cells from Xlf⌬CD21 or BM from Xlf⌬iVav mice (Fig. 1, A and B). Xlf deletion in Xlf⌬CD21 B cells was further confirmed by Western blotting (Fig. 1C).

PAXX and XLF are redundant during VDJ recombination
DNA repair is a key mechanism for development of the immune system, and previous studies reported that mice with a single Paxx or Xlf deletion present overall normal B and T cell development (16 -21). However, DKO of these factors leads to total blockage of the V(D)J recombination process both in vitro and in vivo (E18 recovered embryos), suggesting redundancy in their function (15,19,20,29). Although only a slight but significant decrease in splenic and thymic cellularity was observed in Xlf Ϫ/Ϫ mice compared with Paxx Ϫ/Ϫ and WT mice (Fig. 1D), Xlf⌬iVav mice were lacking a visible thymus, and splenic cellularity was strongly diminished (Fig. 1D). Likewise, although flow cytometry analysis of T (CD4ϩ and CD8ϩ) and B (IgMϩ/B220ϩ) cells in either spleen or thymus revealed an indistinguishable phenotype in Paxx Ϫ/Ϫ and Xlf Ϫ/Ϫ mice compared with the WT (Fig. 1, E and F), Xlf⌬iVav mice presented a severe defect in both lymphoid populations. Moreover, BM from Xlf⌬iVav mice was also devoid of mature B cells and revealed arrest at the pre-proB cell stage (Fig. 1E). These results recapitulate the profound immune deficiency caused by a block in the V(D)J process noted previously in E18.5 Paxx/Xlf DKO fetuses (19), reinforcing the absolute requirement of the combined function of PAXX and XLF for proper development of the adaptive immune system.

PAXX is dispensable for CSR in vivo
We showed previously that loss of PAXX does not impair B cell development (19). The functional relationship between PAXX and XLF during late stages of B cell maturation was analyzed through CD21-Cre-mediated deletion of Xlf in Paxx Ϫ/Ϫ mature B cells (Xlf⌬CD21 mice). The overall development of both T and B cells was not affected in this setting ( Fig.  1, D-F). Histological analysis of the spleen performed by hematoxylin and eosin staining revealed a conserved splenic architecture with the presence of germinal centers compared with WT or Paxx Ϫ/Ϫ mice, despite slightly decreased cellularity in Xlf Ϫ/Ϫ and Xlf⌬CD21 mice ( Fig. 2A). We then analyzed the consequences of the concomitant Paxx and Xlf inactivation on CSR. CD43-negative sorted mature B cells from WT, Paxx Ϫ/Ϫ , Xlf Ϫ/Ϫ , and Xlf⌬CD21 mice were activated during 4 days in vitro to induce CSR toward the IgG1, IgG2b, or IgG3 isotypes. The B-cell proliferative capacity, as determined by Cell-Trace dilution, was not affected under any of the tested conditions (Fig. 2B). Although CSR to the three tested isotypes was not impaired in Paxx Ϫ/Ϫ B cells, as documented previously in CH12-B cells induced to switch to IgA in vitro (15), a significant decrease in the rate of switched cells toward all Ig isotypes was apparent for Xlf Ϫ/Ϫ B cells (Fig. 2, C and D), as reported previously (17). Interestingly, a similar CSR defect was observed in Xlf⌬CD21 B cells (Fig. 2, C and D). The absence of synthetic CSR dysfunction in Xlf⌬CD21 mice argues for the absence of PAXX contribution during CSR, at least quantitatively, even in the absence of XLF, as first noted in vitro (17).

PAXX and XLF participate in the maintenance of the lymphoid progenitor pool in the bone marrow
We then took advantage of Xlf⌬iVav mice to analyze the possible cooperation between PAXX and XLF for maintenance of the hemopoietic potential in these mice. We first measured the blood parameters in 7-to 10-week-old and 20-to 24-weekold mice. A shown in Fig. 3A, no differences were noticed in red blood cell and platelets counts for the various genotypes in the two age groups, indicating that myelopoiesis is not overtly affected by combined Paxx and Xlf deficiency. In contrast, a statistically significant decrease in total white blood cell and lymphocyte counts was observed in Xlf Ϫ/Ϫ mice, as described previously (16,17), and in Xlf⌬iVav mice compared with Paxx Ϫ/Ϫ mice (Fig. 3A). Consistent with the major V(D)J recombination defect in Xlf⌬iVav mice described above, the decrease in lymphocyte numbers was more severe in Xlf⌬iVav animals than in Xlf simple KO animals. The rise in lymphocyte numbers in aged Xlf⌬iVav probably reflects lymphocyte homeostatic proliferation to some extent in aged mice. Analysis of bone marrow (Figs. 3, B and C) revealed that, although the total LSK pool was not altered in young and aged Paxx Ϫ/Ϫ and Xlf⌬iVav mice, the common lymphoid progenitor (CLP) pool was significantly decreased in young Xlf⌬iVav mice compared with Paxx Ϫ/Ϫ mice, with a concomitant increase in the short-term HSC pool (Fig. 3C). The variation of CLPs/short-term HSCs was further aggravated in aged mice. These results indicate a functional redundancy between PAXX and XLF at early stages during hematopoietic development, beyond the V(D)J recombination process.

The survival of mice is impaired by Xlf deletion in iVavexpressing cells
During a 1-year follow up, no difference in the survival of Paxx Ϫ/Ϫ and Xlf Ϫ/Ϫ mice was observed, which contrasted with ACCELERATED COMMUNICATION: Combined PAXX and Xlf deficiency the sharply increased mortality of Xlf⌬iVav mice starting around 100 days (Fig. 4A). Aged Xlf⌬iVav mice showed decreased weight and moderate to severe alopecia (Fig. 4B). Moreover, three of five analyzed mice developed a thymic mass (Fig. 4C) whose histology revealed total loss of the thymus cortical and medullar architecture compared with the Paxx Ϫ/Ϫ thymus (Fig. 4C, top panels). Higher magnification revealed neoplastic cells with a large nucleus and dense cytoplasm, suggesting the tumorigenic nature of these cells (Fig. 4C, bottom  panels, arrowheads). Thymic cell suspension analysis (Fig. 4D) showed a lack of activated T cells, indicated by negative TCR␤ϩ CD69ϩ staining and a shift in proliferative doublepositive CD44ϩ CD62Lϩ cells for Xlf⌬iVav mice. These results are consistent with development of thymic lymphomas in aging Xlf⌬iVav mice, the possible cause of their accelerated death.

Discussion
We analyzed, for the first time in vivo, the PAXX and XLF interplay during V(D)J recombination and CSR through design of two conditional KO models: Xlf⌬iVav and Xlf⌬CD21. These two models circumvent the embryonic lethality resulting from combined lack of PAXX and XLF, as observed in previous studies (19,20). Xlf⌬CD21 mice recapitulate the synthetic dysfunction of Paxx and Xlf in V(D)J recombination, manifested by the T-B-severe combined immunodeficiency phenotype. Xlf⌬CD21 mice confirmed the dispensability of PAXX in CSR, first proposed in vitro using the CH12 model (15), as shown by the absence of an additive effect of Paxx deficiency over Xlf KO alone. Overall, these results are in line with previous studies and suggest the existence of two distinct mechanisms for repair of programmed DNA breaks induced during V(D)J and CSR. They suggest that, in contrast to what happens during V(D)J recombination, XLF and PAXX do not complement each other for DNA end synapsis during CSR, in line with recent findings in vitro, proposing that Ku and XRCC4/DNA-Lig4 can initiate synapse formation on their own without the combined contribution of PAXX and XLF, at least on blunt DNA ends (30). Likewise, PAXX failed to rescue the defect in precise DNA end joining resulting from Xlf deficiency in a CRISPR/Cas9 model of DNA repair (31). It is thus proposed that classical NHEJ factors such as XLF or PAXX are involved in precise (without indels) DNA blunt end ligation but that their redundancy may vary depending on the DSB context. In that respect, it is worth mentioning that DSBs that occur during CSR are very unique, as they occur within large regions of repetitive sequences (called S for switch regions).
Last, the conditional Xlf⌬iVav model enabled analysis of the impact of Paxx/Xlf deficiency on the hematopoietic potential of aging mice, which was not possible with models used in previously published studies. For the first time, we could observe a shortened lifespan and the appearance of spontaneous thymic lymphomas without the need of a tumor accelerator background such as TP53 KO in Xlf⌬iVav aged mice, suggesting an additional role of PAXX/XLF as a genome caretaker in the context of RAG1/2-induced breaks because those thymic lymphomas were not observed in aged Xlf⌬CD21 mice. Although new studies are needed for full characterization of these hemopathies, we suggest the possible implication of aberrant oncogenic rearrangements to explain their appearance, given the previous demonstration of chromosomal translocations followed by production of fusion oncogenic genes in mouse cells in the context of classical NHEJ factor deficiencies (24,(32)(33)(34)(35).

Generation of conditional Xlf KO mice on a PAXX KO background
C57/Bl6 Paxx Ϫ/Ϫ Xlf ϩ/Ϫ -CD21-cre or Paxx Ϫ/Ϫ Xlf ϩ/Ϫ -iVavcre mice were intercrossed with Paxx Ϫ/Ϫ Xlf flox/flox mice to generate Paxx Ϫ/Ϫ XlfϪ/ flox CD21-cre or Paxx Ϫ/Ϫ Xlf Ϫ/flox iVav-cre mice and to specifically delete the Xlf gene under the expression of Cre recombinase along with the CD21 or iVav promoters. All mice used for this work were 7-10 weeks old (considered young mice) or 20 -24 weeks old (considered old mice), kept under pathogen-free conditions, and sacrificed humanely. Mouse experiments were performed with approval from the local ethics committee and the French Ministry of Education and Research.

Flow cytometry analysis
For the immunophenotype, the thymus, spleen and bone marrow were harvested. Thymic populations were identified by anti-CD8 and anti-CD4, and the double negative population was stained with anti-CD44 and anti-CD25 mouse antibodies (all from Sony Biotechnologies). Thymic lymphomas were additionally stained with anti-CD69, anti-CD62L, and anti-TCR␤. Splenic T cells were stained with anti-CD3, anti-CD4, and anti-CD8, whereas B cells were identified by anti-B220 and anti-IgM (all from Sony Biotechnologies). Bone marrow was obtained from one femur and stained with anti-IgM, anti-B220, and anti-CD43 (all from Sony Biotechnologies). Flow cytometry was performed on a BD-LSR Fortessa (BD Biosciences).

Blood measurement
Blood from mice was collected by intracardiac puncture after anesthesia by injection of a mixture of ketamine and xylazine. A, quantification of red blood cell, platelet, white blood cell, and lymphocyte concentrations in the blood of 7-to 10-week-old and 20-to 24-week-old Paxx Ϫ/Ϫ , Xlf Ϫ/Ϫ , and Xlf⌬iVav mice. Results are indicated as mean Ϯ S.E. Statistical analyses were performed by one-way ANOVA. Significant differences are indicated as follows: *, p Ͻ 0.05; **, p Ͻ 0.01; ns, not significant. B, gating strategy for analysis of hematopoietic populations in BM. Analysis were performed in a LinϪ population based on staining with c-kit, Sca-1, Flt3, CD34, and IL7RA antibodies for assessment of LinϪc-kitϩScaϪ1ϩ cells (LSK), short-term (ST) hematopoietic stem cells HSCs, and CLPs. C, quantification of flow cytometry analyses for the hematopoietic compartment of 7-to 10-week old (young) and 20-to 24-week-old (old) Paxx Ϫ/Ϫ and Xlf⌬iVav mice. The percentages of total LSK, short-term HSC, and CLP populations are represented as mean Ϯ S.E., and statistical analyses were performed by Student's t test. Significant differences are indicated as follows: *, p Ͻ 0.05; **, p Ͻ 0.01.

ACCELERATED COMMUNICATION: Combined PAXX and Xlf deficiency
Blood was then analyzed on a Procyte Dx Hematology analyzer (IDEXX Bioanalytics).

Histological analysis
The thymus and spleen were harvested, fixed overnight in 4% paraformaldehyde, washed with PBS, paraffin-embedded, and sectioned at 4-m thickness. Sections were then stained with hematoxylin and eosin for morphological analysis.

Purification and activation of splenic mature B cells in vitro
Mature B cells were negatively sorted from the spleen using CD43ϩ magnetic beads following the manufacturer's instruc-tions (Miltenyi). The negative fraction containing CD43Ϫ mature B cells was harvested and labeled with Cell-Trace (Thermo Fisher). For each condition, 500 000 cells were plated in triplicate in a 6-well plate in Iscove's modified Dulbecco's mediumϩ Glutamax medium supplemented with 10% fetal bovine serum, 1% penicillin-streptomycin, 1% sodium pyruvate, 1% HEPES, 1% nonessential amino acids, and 0.1% ␤-mercaptoethanol (all from Thermo Fisher). Cells were incubated for 4 days with 25 g/ml LPS from Escherichia coli (Sigma) ϩ 20 ng/ml IL-4 (R&D Systems) for IgG1 switching or with 25 g/ml LPS for IgG3 and IgG2b switching. At the end of incubation, cells were collected and labeled with anti-mouse B220-PE ACCELERATED COMMUNICATION: Combined PAXX and Xlf deficiency (Sony Biotechnologies) and anti-mouse biotin-coupled IgG1, IgG2b, and IgG3 (R&D Systems), followed by incubation with allophycocyanin-streptavidin (Thermo Fisher). Cells were analyzed on a BD LSR-Fortessa (BD Bioscience).

Statistics
All statistics were determined using Prism (GraphPad Software). Groups were analyzed by one-way ANOVA or Student's t test as indicated, and the difference was considered statistically significant at p Ͻ 0.05.