Nucleolar localization of RAG1 modulates V(D)J recombination activity

Significance

Vertebrate immune systems can respond to many infections and insults. This ability relies on a diverse binding repertoire of antigen receptors. Antigen receptor diversity is created through a process called V(D)J recombination in which arrayed gene segments are shuffled to form functional receptors. This process introduces breaks in chromosomal DNA catalyzed by the RAG1-RAG2 protein complex and requires strict regulation to guard genome integrity. Here we demonstrate a mode of RAG1 regulation by nucleolar sequestration. RAG1’s nucleolar localization is dynamically regulated and is disrupted during a B cell’s transition to a prorecombination state, leading to increased recombination.

Abstract

V(D)J recombination assembles and diversifies Ig and T cell receptor genes in developing B and T lymphocytes. The reaction is initiated by the RAG1-RAG2 protein complex which binds and cleaves at discrete gene segments in the antigen receptor loci. To identify mechanisms that regulate V(D)J recombination, we used proximity-dependent biotin identification to analyze the interactomes of full-length and truncated forms of RAG1 in pre-B cells. This revealed an association of RAG1 with numerous nucleolar proteins in a manner dependent on amino acids 216 to 383 and allowed identification of a motif required for nucleolar localization. Experiments in transformed pre-B cell lines and cultured primary pre-B cells reveal a strong correlation between disruption of nucleoli, reduced association of RAG1 with a nucleolar marker, and increased V(D)J recombination activity. Mutation of the RAG1 nucleolar localization motif boosts recombination while removal of the first 215 amino acids of RAG1, required for efficient egress from nucleoli, reduces recombination activity. Our findings indicate that nucleolar sequestration of RAG1 is a negative regulatory mechanism in V(D)J recombination and identify regions of the RAG1 N-terminal region that control nucleolar association and egress.

A vast diversity of molecular specificity is needed to mediate recognition and interaction between host and pathogen in the jawed vertebrate adaptive immune system. This diversity is created in part by combinatorial gene rearrangements carried out in developing B and T cells during the process of V(D)J recombination, in which discrete V (variable), D (diversity), and J (joining) gene segments are stochastically combined to form a functional antigen receptor gene (1). V(D)J recombination is initiated by Recombination Activating Gene (RAG) proteins 1 and 2 (2, 3). Together, RAG1/2 carry out V(D)J recombination as a heterotetramer that binds and cleaves DNA at specific recombination signal sequences (RSSs) flanking exon segments. After cleavage, these exon segments are chaperoned into the nonhomologous end joining DNA repair pathway by the RAG complex (reviewed in refs. 4–6). While the RAG proteins enable the generation of a diverse B and T cell receptor repertoire, the DNA breaks generated during V(D)J recombination are inherently genotoxic and can lead to harmful translocations and subsequent lymphocytic malignancies (7–10). In light of this, understanding the mechanisms underlying RAG targeting and regulation are of great interest.

Regulation of RAG2 at the protein level has been well characterized. Spatial regulation of RAG2 away from the nuclear periphery is thought to contribute to allelic exclusion and the ordered rearrangement of the TCRβ loci (11), while CDK2-dependent degradation of RAG2 limits V(D)J recombination to the G₁ phase of the cell cycle (12, 13). In contrast, little is known regarding the regulation and localization of the RAG1 protein, with most work focusing on RAG1 transcriptional regulation (14–16).

Identifying proteins involved in the targeting, regulation, and repair of RAG-mediated DNA breaks has been hindered by the lack of methodologies amenable to probing the protein–protein interactions of RAG. RAG1, the major agent of DNA binding and cleavage, is a 1,040 amino acid (aa) protein that is largely insoluble and difficult to extract from the nucleus (17, 18). As such, much of the biochemical and structural characterization of RAG1 has been done on a truncated “core” version spanning residues 384 to 1,008. While core RAG1 retains catalytic activity, its in vivo recombination efficiency and fidelity are reduced compared to full-length RAG1 (FLRAG1) and its binding to the genome is more promiscuous (19–24). The evolutionarily conserved 383 aa N-terminal domain (NTD) missing from core RAG1 is predicted to harbor multiple zinc-binding motifs including a Really Interesting New Gene (RING) domain (aa 287 to 351) capable of ubiquitylating multiple targets, including RAG1 itself (23, 25–27). Although this ubiquitylation activity has been characterized in vitro, its in vivo relevance to V(D)J recombination remains unclear. Also contained within the NTD is a region (aa 1 to 215) that mediates interaction with DCAF1, causing degradation of RAG1 in a CRL4-dependent manner (28, 29). The NTD also contributes to chromatin binding and genomic targeting of the RAG complex (20, 24). Despite a growing body of evidence highlighting the importance of RAG1’s NTD, our understanding of its functional contribution to V(D)J recombination is far from complete. In addition, because of its low-level expression, microscopy of FLRAG1 in a cellular context has been extremely limited, leaving many questions unanswered regarding RAG1 localization and trafficking.

Many proteins are regulated by their localization or sequestration within distinct cellular compartments. The nucleolus is a phase-separated, nonmembrane bound nuclear organelle that is the site of ribosome biogenesis. However, recent efforts to map the nucleolar proteome have revealed a plethora of proteins with roles beyond canonical nucleolar processes, including DNA repair and apoptosis (30–32). Further work has also shown the nucleolus as a dynamic hub capable of regulating protein function in response to specific stimuli, including DNA double-strand breaks (DSBs) (33–35).

More than 20 y ago, RAG1 was reported to localize to the nucleolus when overexpressed in a nonlymphoid cell line (36). We are not aware of subsequent studies to determine whether this occurs at physiological levels of RAG1 expression in its normal cellular context or whether it might have functional relevance. Here, we demonstrate that RAG1 harbors a nucleolar localization signal (NoLS) motif in its NTD and that RAG1 function is regulated by nucleolar localization. During Ig (Ig) κ-gene recombination and in response to nucleolar stress, we observe that RAG1 egresses from the nucleolus and forms small, bright puncta in a manner dependent on aa 1 to 215. These findings delineate a repressive function for nucleolar localization of RAG1 and set the stage for further work examining the role of the nucleolus in the regulation of RAG1 and V(D)J recombination.

Results

Biotin Identification Identifies Multiple Nucleolar Proteins Proximal to Full-Length RAG1.

To identify RAG1-associated proteins, we used proximity-dependent biotin identification (BioID) (37), which makes use of a promiscuous Escherichia coli biotin ligase (BirM) to biotinylate lysine residues on proximal proteins. We generated various truncations of RAG1 fused to BirM (Fig. 1A) and tested these fusion proteins for biotinylation and recombination activity in HEK293T and the commonly used pre-B cell model system, Abelson murine leukemia virus-transformed (vAbl) cells (SI Appendix, Fig. S1). vAbl cells are developmentally arrested at the pre-B stage via expression of a constitutively active form of the Abelson kinase. Upon addition of the Abelson kinase inhibitor STI-571, the cells exit cell cycle in the G₁ phase and activate RAG expression and Igκ locus recombination (38). We utilized a doxycycline-inducible system to express the RAG1-BioID constructs in stably retrovirally transduced vAbl cells, allowing us to initiate V(D)J recombination and RAG1 interactome labeling synchronously by addition of STI-571, doxycycline, and biotin. After 24 h of labeling, cells were lysed and biotinylated proteins were enriched, digested, and the resulting peptides analyzed by liquid chromatography tandem mass spectrometry (LC-MS/MS).

Fig. 1.

BioID reveals nucleolar association of RAG1. (A) Schematic of the three BioID constructs N-terminally fused to RAG1 and the control construct fused to a NLS. (B) Volcano plot comparing enriched proteins between core RAG1 and Δ215. P value is on a −Log₁₀ scale. Nucleolar proteins are shown as red squares. (C) Bar graph showing the number of nucleolar proteins found enriched with each RAG1 construct compared to the control. (D) Confocal images of HEK293T cells transiently transfected with mCherry fused to either Δ215 or core RAG1, imaged 24 h after transfection, showing colocalization of Δ215 with the nucleolar marker GFP-fibrillarin. Representative of three independent experiments.

We collected interactome data on four constructs in duplicate (39) (Fig. 1A): BirM fusion proteins of full-length murine RAG1 (FLRAG1), a 215- to 1,040-aa truncation omitting the region responsible for RAG1 degradation (Δ215), the minimal core catalytic unit required for V(D)J recombination (core RAG1), and a control construct (BirM with an added nuclear localization signal [NLS]) used to assess background biotinylation. Out of 820 proteins identified in both a RAG1 sample and the control sample, 205 were significantly enriched in one or more of the RAG1 samples compared to the control (SI Appendix, Table S1). A further 140 proteins were identified solely by a RAG1 construct (SI Appendix, Table S2). Overlap between the proteins identified with the three BirM-RAG1 fusion proteins is depicted in SI Appendix, Fig. S2A, while the high reproducibility of the data between biological duplicates is illustrated in SI Appendix, Fig. S2B. Several proteins known to be involved in V(D)J recombination, including RAG2 and HMGB1, were enriched in RAG1-containing constructs compared to the control.

Upon comparing the interactomes of different forms of RAG1, we found a striking pattern of enrichment of nucleolar proteins, with Δ215 and FLRAG1 exhibiting a five-fold and three-fold increase in nucleolar partners, respectively, compared to core RAG1 (Fig. 1 B and C and SI Appendix, Fig. S3). Multiple nucleolar proteins were among the most highly enriched when comparing Δ215 to core RAG1 (Fig. 1B) or when comparing FLRAG1 to core RAG1 (SI Appendix, Fig. S3A). This observation suggested that RAG1 localizes to the nucleolus, and does so in a manner dependent on amino acids 215 to 384. As an initial test of this idea, we cotransfected HEK293T cells with expression constructs for mCherry-tagged Δ215 and core RAG1 as well as a GFP-tagged nucleolar marker, fibrillarin. Tagged RAG1 colocalized with fibrillarin in the nucleoli in a manner dependent on amino acids 215 to 384 (Fig. 1D).

RAG1 Harbors a NoLS in Its NTD.

RAG1’s NTD has been reported to influence its nuclear import via interaction with KPNA1 (40), but its contribution to subnuclear localization has not been carefully examined. To identify sequences that could dictate nucleolar localization, we performed an in silico analysis of the RAG1 amino acid sequence using a NoLS prediction tool (41), which identified a large basic region from aa 211 to 260 as a potential nucleolar localization motif (SI Appendix, Fig. S4A). A similar region was identified in the 1995 study by Spanopoulou et al. as contributing to nucleolar localization of RAG1 when overexpressed in nonlymphoid cells (36). Taking into account previous mutagenesis experiments done on this region (42–45) and analyzing evolutionary conservation (SI Appendix, Fig. S4), we chose to focus on a short motif spanning amino acids 243 to 249. Based on this, we mutated the basic residues of this motif to neutral mimetics (RRDRRKR→QQDQQIQ) in the context of FLRAG1 to generate the dNOL RAG1 protein. Various mCherry-tagged murine RAG1 proteins (Fig. 2A) were inducibly expressed in vAbl cells. Despite roughly comparable expression levels among the mCherry-RAG1 proteins (SI Appendix, Fig. S5), those containing the NoLS motif (FL and Δ215) showed strong nucleolar localization, while those lacking a NoLS (dNOL and core RAG1) showed pannuclear localization (Fig. 2 B, Left). Quantitative analysis of these experiments is described below.

Fig. 2.

Nucleolar localization of RAG1 is dynamic and dependent on the recombination state of the cell. (A) Schematic of RAG1 fusion proteins used for analysis. (B) Confocal images from vAbl cells constitutively expressing GFP-fused fibrillarin and doxycycline-inducible mCherry-fused RAG1. Cells were treated with doxycycline to induce RAG1 expression for 16 h and, when stated, treated with 5 µM STI-571 for 4 h prior to being mounted and fixed. Representative of three independent experiments. (C) Colocalization analysis between RAG1 and fibrillarin (Fbl) before and after STI-571 treatment. Pearson correlation was calculated from individual cells and plotted with whiskers at 10th to 90th percentile. Statistical significance was determined by ANOVA. ****P < 0.0001; NS, not significant. (D) Quantitation of nucleolar size by pixel area before and after STI-571 treatment. Individual nucleoli plotted with whiskers at 10th to 90th percentile. Statistical significance was determined by ANOVA. ****P < 0.0001.

STI-571 Treatment of vAbl Cells Leads to RAG1 Egress from Nucleoli in a Manner Dependent on aa 1 to 215.

To determine the effect of inducing a prorecombination state on RAG1 localization, we treated the mCherry construct-containing vAbl cells with STI-571 and doxycycline and assessed localization of mCherry-RAG1 relative to GFP-fibrillarin. When induced with STI-571, we observed that FLRAG1 egressed from nucleoli and formed small, bright puncta (Fig. 2 B, Right). Puncta were observed with FLRAG1 and dNOL but not Δ215 or core RAG1, arguing that their formation in response to STI-571 depends on aa 1 to 215. Unlike FLR1, Δ215 showed little discernible change in nucleolar association in response to STI-571, remaining strongly colocalized with GFP-fibrillarin. This suggests that nucleolar egress requires RAG1 aa 1 to 215. Localization of core RAG1 did not change discernibly in response to STI-571, remaining pannuclear. dNOL exhibited less overlap with GFP-fibrillarin than did FLRAG1, as predicted by its mutated NoLS. We similarly assessed RAG2 subnuclear localization and observed little overlap between mCherry-RAG2 and GFP-fibrillarin and no apparent change in RAG2 localization in response to STI-571 (SI Appendix, Fig. S6 A and B).

To quantitate the colocalization of RAG1 with fibrillarin, we developed an analysis pipeline that allowed calculation of a Pearson correlation coefficient between the mCherry-RAG1 and GFP-fibrillarin fluorescent signals in large numbers of individual cells. This analysis indicated that FLRAG1 colocalized with fibrillarin to a significantly greater extent under cycling, recombination-repressed conditions than after STI-571 treatment, while Δ215 strongly colocalized with fibrillarin under both cycling and STI-571–treated conditions (Fig. 2C). In contrast, core RAG1 showed low colocalization with fibrillarin before and after STI-571 treatment. Under cycling conditions, dNOL exhibited a striking absence of colocalization with fibrillarin, far lower than any of the other forms of RAG1, with colocalization increasing but remaining low upon STI-571 treatment. This increase might reflect the dNOL protein’s combination of the egress-permitting 1 to 215 domain and a disrupted NoLS. In addition to affecting the nucleolar localization of forms of RAG1 containing aa 1 to 215, STI-571 triggered a significant decrease in nucleolar size and change in nucleolar morphology, irrespective of the form of RAG1 expressed (Fig. 2D). Human FLRAG1 tagged with mCherry expressed in vAbl cells also exhibited substantial colocalization with fibrillarin that was reduced upon STI-571 treatment (SI Appendix, Fig. S6C), suggesting that this behavior is not restricted to mouse RAG1. Together, these data demonstrate that FLRAG1 localizes to nucleoli in a manner dependent on a basic patch from aa 243 to 249 and dissociates from nucleoli upon STI-571 treatment in a manner dependent on aa 1 to 215.

Disruption of Nucleolar Function via RNA Polymerase I Inhibition Leads to RAG1 Egress, Nucleolar Contraction, and Increased V(D)J Recombination.

Because the transition to a prorecombination state induced by STI-571 led to the egress of RAG1 from the nucleolus, we hypothesized that the nucleolus acts as an inhibitory depot for RAG1 to repress V(D)J recombination. To explore this possibility, we used actinomycin D (actD) at low concentrations to inhibit RNA polymerase I (RNAP-I) and perturb nucleolar function (46). We performed recombination assays using the chromosomally integrated fluorescent V(D)J recombination reporter pMGInv (47) (Fig. 3A) and vAbl cells made from WT or core RAG1 mice, allowing comparison of proteins that do (FLRAG1) or do not (core RAG1) localize to nucleoli. Strikingly, actD triggered a significant, dose-dependent increase in recombination activity in WT but not core RAG1 vAbl cells (Fig. 3B). ActD did not alter levels of RAG1 expression (Fig. 3C) or cell viability (SI Appendix, Fig. S7) in either WT or core RAG1 vAbl cells, consistent with the idea that actD enhances RAG1 activity rather than expression. As actD has been observed to alter nucleolar morphology in HeLa cells (48), we predicted that actD treatment would affect the nucleoli of vAbl cells. Indeed, in a manner similar to STI-571 treatment, treatment with actD of WT vAbl cells expressing mCherry-FLRAG1 and GFP-fibrillarin led to contraction of nucleoli, RAG1 puncta formation, and a visible decrease in RAG1-fibrillarin colocalization (Fig. 4A). Quantification of these microscopy data confirmed significantly decreased RAG1-nucleolar colocalization and nucleolar size in actD-treated vAbl cells (Fig. 4 B and C). Because RAG2 is largely absent in cycling vAbl cells, this phenomenon is likely to be RAG2 independent. These data demonstrate that a disruption of nucleolar homeostasis can lead to decreased RAG1 localization to nucleoli and increased recombination activity.

Fig. 3.

Recombination efficiency is regulated by nucleolar localization of RAG1. (A) Schematic of pMGInv inversional GFP reporter. Black and white triangles represent 23RSS and 12RSS, respectively. (B) Recombination efficiency assay using pMGInv reporter in WT and core RAG1 vAbl cells. Cells were induced for 48 h with 5 µM STI-571 with varying amounts of actD and analyzed via flow cytometry for GFP⁺ cells. Statistical significance was determined by ANOVA. ****P < 0.0001; NS, not significant. (C) Western blot of RAG1 from vAbl cells treated with 5 µM STI-571 and/or 2 nM actD for 48 h. ActD does not influence RAG1 expression levels. Representative of two independent experiments.

Fig. 4.

Nucleolar morphology is affected by actD treatment. (A) Confocal images from vAbl cells constitutively expressing GFP-fused fibrillarin and doxycycline-inducible mCherry-fused RAG1. Cells were treated with doxycycline to induce RAG1 expression for 16 h and with 2 nM actD for 4 h prior to being mounted and fixed. Representative of three independent experiments. (B) Colocalization analysis between RAG1 and fibrillarin (Fbl) before and after actD treatment. (C) Quantitation of nucleolar size by pixel area before and after actD treatment. Individual nucleoli plotted with whiskers at 10th to 90th percentile. Statistical significance was determined by ANOVA. ****P < 0.0001.

Nucleolar Sequestration of RAG1 Corresponds to Reduced V(D)J Recombination.

The hypothesis that RAG1 sequestration in the nucleolus suppresses V(D)J recombination leads to the prediction that the Δ215 protein, which is trapped in the nucleolus, should be less active than FLRAG1, while the opposite should be true of the dNOL protein, which is excluded from the nucleolus. To test this, RAG1^−/− vAbl cells were infected with the pMGInv recombination substrate along with a vector driving constitutive RAG2 expression. The cells were then transduced with vectors constitutively expressing FLRAG1, core RAG1, Δ215, or dNOL to compare their ability to catalyze recombination in cycling cells using a common clonal progenitor cell line. In agreement with our hypothesis, Δ215 supported the lowest level of recombination (Fig. 5A) despite being highly expressed (Fig. 5B), reflecting its strong nucleolar sequestration and inability to egress. Conversely, dNOL supported the highest efficiency of recombination, consistent with its diminished nucleolar sequestration as compared to FLRAG1. Core RAG1 showed a diminished level of recombination compared to FLRAG1, as expected given its previously reported recombination inefficiency in the context of extrachromosomal substrates and endogenous antigen receptor loci (22, 42, 45). We speculate that the recombination defect associated with the absence of all noncore regions in core RAG1 masks the boost in recombination expected from the lack of nucleolar localization for this protein.

Fig. 5.

Functional comparison of RAG1 truncations reveals distinct roles for different parts of the RAG1 NTD. (A) Recombination efficiency from a clonal RAG1^−/− RAG2^+/+ vAbl cell line transduced with different forms of RAG1. At 72 h postinfection, cells were analyzed via flow cytometry for GFP⁺ cells. Statistical significance was determined by ANOVA. ****P < 0.0001. (B) Western blot showing expression levels of RAG1 from infected RAG1-complementation lines. WT vAbl cells uninduced and induced with STI-571 shown as control. Representative of three independent experiments.

Nucleolar Disruption Stimulates V(D)J Recombination in Ex Vivo Pre-B Cell Cultures.

To determine whether RAG1 nucleolar sequestration also suppresses V(D)J recombination in nontransformed lymphocytes, we tested the effect of nucleolar disruption in ex vivo primary pre-B cell cultures. Bone marrow was harvested from WT or core RAG1 mice and cultured for 1 wk in the presence of IL7, which expands a population of large, cycling pre-B cells in which recombination is repressed (49). Cells were then infected with the pMGInv reporter, split into IL7-positive and -negative cultures containing varying concentrations of actD, and analyzed by flow cytometry for recombination-mediated GFP expression (Fig. 6A). Removal of IL7 promotes transition to a prorecombination, small pre-B cell state in which cells undergo recombination of the Igκ locus. ActD treatment of WT pre-B cells increased recombination of the reporter in a dose-dependent manner in recombinationally repressed IL7-positive cultures but not in prorecombination IL7-negative cultures (Fig. 6 B, Left). Similar results were obtained when cleavage of the Igκ locus was assayed with droplet digital PCR (ddPCR) using primers spanning Jκ2 and its flanking RSS (Fig. 6C), where cleavage at Jκ2, or a recombination event that results in Jκ2 deletion from the cell, prevents PCR amplification. The lack of strong stimulation in the absence of IL7 suggested the possibility that IL7 removal is sufficient to release nucleolar inhibition of RAG1, thereby rendering actD treatment largely redundant in IL7-negative conditions. Consistent with this, removal of IL7 reduced nucleolar volume nearly to the same extent as addition of actD (Fig. 6D). In agreement with our findings with vAbl cells, primary pre-B cells from core RAG1 mice showed no increase in recombination when treated with actD whether in the presence or absence of IL7 (Fig. 6 B, Right). Notably, levels of RAG1 protein expression were not affected by IL7 withdrawal or by actD treatment (SI Appendix, Fig. S8), indicating that under our culture conditions, increased RAG1 protein does not explain the enhancement of recombination caused by withdrawal of IL7 or addition of actD, and suggesting that regulated release of RAG1 from a repressive nucleolar compartment can activate V(D)J recombination in primary pre-B cells.

Fig. 6.

Recombination efficiency in primary pre-B cells is altered by actinomycin D. (A) Schematic showing culture conditions during assay. (B) pMGInv recombination assay of homogenized pre-B cells from WT and core RAG1 mice. Cells were analyzed for GFP⁺ cells after 48 h with/without IL7 in various concentration of actD. Representative of two independent experiments. Statistical significance was determined by ANOVA. **P < 0.01; ***P < 0.001; ****P < 0.0001; NS, not significant. (C) ddPCR assay for intact Jκ2, as assessed by the decrease in PCR amplification resulting from Igκ cleavage and recombination. The signal for intact Jκ2 alleles was measured, normalized to an amplicon from the gene RPP30, and then subtracted from 2 to yield the apparent level of Jκ2 cleavage. Genomic DNAs from kidney and B220⁺ splenocytes were used as no cleavage and high cleavage controls, respectively. For IL7 culture data, each data point represents the mean of either two or eight technical replicates performed for two independent samples. For spleen and kidney, each data point represents the mean of four technical replicates for two independent samples. Statistical significance was determined by ANOVA. **P < 0.01; ****P < 0.0001. (D) Quantitation of nucleolar size by pixel area in pre-B cells untreated and treated with IL7 and/or 2 nM actD for 48 h. Individual nucleoli plotted with whiskers at 10th to 90th percentile. Statistical significance was determined by ANOVA. ****P < 0.0001.

Discussion

A 1995 study detected RAG1 in the nucleolus when overexpressed in a nonlymphoid cell line (36), but to our knowledge this observation has not been elaborated upon, nor has a functional or regulatory role been demonstrated for RAG1 nucleolar localization until now. Using BioID in a pre-B cell line, we find that RAG1 resides in proximity to numerous nucleolar proteins in a manner strongly facilitated by RAG1 aa 216 to 384. Nucleolar localization of RAG1 was supported by fluorescence microscopy in both HEK293T and vAbl cells using mCherry-RAG1 fusion proteins. It is unlikely that the mCherry fusion partner is responsible for nucleolar localization, given the weak or negligible nucleolar localization seen with the mCherry-core RAG1 or mCherry-dNOL proteins and given that a different fusion partner was used in the BioID analysis. We cannot exclude the possibility that overexpression of the BirM- and mCherry-RAG1 fusion proteins contributes to the observed nucleolar localization. It is clear, however, that mere overexpression of a RAG1 protein is not sufficient for nucleolar localization, as BirM-core-RAG1 and mCherry-dNOL are expressed at higher levels than their FLRAG1 counterparts (SI Appendix, Figs. S1A and S5). Hence, a specific feature(s) of RAG1 contributes to nucleolar localization. Furthermore, the functional analyses reported here, whose results correspond well with predictions stemming from the localization experiments, were performed in large part with endogenous, untagged RAG1 protein expressed at levels characteristic of vAbl or primary cultured pre-B cells. We have not been able to detect endogenous RAG1 protein reliably above background by immunofluorescence microscopy, and confirmation of nucleolar localization of endogenous RAG1 might require a more sensitive approach such as superresolution microscopy.

Consistent with our BioID data and the prior 1995 study (36), colocalization of murine RAG1 with the nucleolar protein fibrillarin in pre-B cells is dependent on RAG1 aa 243 to 249. This strongly basic motif is embedded in a larger basic region from aa 211 to 260 that we identified computationally as a potential NoLS (41). While the aa 243 to 249 motif is necessary for efficient nucleolar localization in our experiments, it might not be sufficient, with flanking basic residues potentially also contributing. Our data are consistent with the idea that the nucleolar localization of RAG1 is dictated by a mechanism similar to that of other nucleolar proteins, such as ARF or VHL, being largely driven by charge-based interactions with RNA (50). However, for many nucleolar proteins, including RAG1, a static protein-RNA interaction is not sufficient to explain the dynamic regulation of subnuclear localization seen in response to specific stimuli. Previous work has shown that transcription of specific noncoding RNAs from the intergenic region of the rDNA loci can explain stimulus-dependent nucleolar sequestration or release of target proteins (34). Because inhibition of RNAP-I by actD treatment leads to RAG1 egress from nucleoli, this is an attractive potential mechanism for RAG1 regulation, though it remains speculative.

Our findings suggest that nucleolar localization and egress of RAG1 are dictated by nonoverlapping portions of the RAG1 NTD and have the potential to be regulated processes. While nucleolar localization is dictated by aa 216 to 384, likely localized around aa 243 to 249, efficient nucleolar egress requires RAG1 aa 1 to 215. This region contains several conserved clusters of cysteine and histidine residues (45) and coordinates two zinc atoms (51), but has not been characterized structurally. The only well-defined function for aa 1 to 215 is interaction with DCAF1 (also known as VprBP) (28), which limits RAG1 protein levels and restrains V(D)J recombination activity by mediating RAG1 degradation in an E3-ligase/proteasome-dependent manner (29). Hence, our data suggest that the aa 1 to 215 region of RAG1 serves as both a positive (nuclear egress) and negative (RAG1 degradation) regulator of V(D)J recombination. Whether these two functions are in any way related or make use of overlapping RAG1 residues remains to be determined. Our finding of a positive regulatory function for aa 1 to 215 provides a potential explanation for the previous finding that deletions or point mutations in this region can strongly compromise V(D)J recombination activity (45).

Multiple lines of evidence presented here support the hypothesis that nucleolar sequestration of RAG1 represses V(D)J recombination: 1) treatments that disrupt nucleolar integrity (e.g., actD) and reduce colocalization of RAG1 with fibrillarin result in enhanced V(D)J recombination by FLRAG1 but not by core RAG1 (which does not localize strongly to nucleoli); this is true in both vAbl cells and cultured primary pre-B cells, and in the latter was demonstrated with both an integrated recombination reporter and at the Igκ locus; 2) activation of recombination in vAbl cells with STI-571 triggers disruption of nucleoli and reduces RAG1-fibrillarin colocalization; 3) withdrawal of IL7 from cultures of primary pre-B cells disrupts nucleoli and creates a prorecombination state that cannot be further stimulated by addition of actD; and 4) deletion of the region of RAG1 that controls nuclear egress (∆215) reduces V(D)J recombination activity while the opposite is observed when NoLS function is disrupted (dNOL). Hence, we propose that RAG1 function is regulated through both its entry into and exit from the nucleolus.

ActD is a nonspecific transcriptional inhibitor that could perturb multiple parameters of cell function. However, the actD concentrations used in our experiments, which are sufficient to disrupt nucleoli, dissociate RAG1 from fibrillarin, and stimulate V(D)J recombination, are about 1,000-fold lower than required for inhibition of RNAP-II (46) and have no discernible effect on cell viability (SI Appendix, Fig. S7). This suggests that the ability of actD to stimulate V(D)J recombination is likely mediated by its effects on nucleoli and RAG1.

In addition to RAG1 nucleolar egress, STI-571 treatment of vAbl cells leads to the formation of RAG1 puncta in a manner dependent on the first 215 aa of RAG1. The physiological relevance of these puncta is unclear. We cannot rule out that puncta formation is a consequence of overexpression of forms of RAG1 that contain aa 1 to 215, though it is clear that puncta do not arise through overexpression of RAG1 proteins in general, as they are not seen without STI-571 or actD treatment or with core RAG1 and Δ215, the most highly expressed RAG1 proteins. It remains to be determined whether these puncta represent phase-separated bodies and if their formation relates to the functions of the aa 1 to 215 region in nucleolar egress or RAG1 degradation through interaction with DCAF1.

We do not understand the signaling pathways that lead STI-571–treated vAbl cells and IL7-depleted primary pre-B cells to undergo nucleolar contraction similar to that seen with actD treatment. It will now be important to determine whether this phenomenon is recapitulated in recombinationally active B and T lineage cells in the bone marrow and thymus. Identifying links between V(D)J recombination and nucleolar homeostasis might provide clues as to how RAG1 localization is regulated and whether its presence in the nucleolus has a broader influence on other nucleolar processes.

V(D)J recombination is tightly regulated during B and T cell development. This requires precise activation of RAG when appropriate (in G₀/G₁ phase of the cell cycle) and rapid cessation upon generation of DSBs by RAG. Following the resolution of a RAG-mediated break, RAG activity often needs to be reinitiated to salvage a nonproductive rearrangement event or a developing lymphocyte expressing an autoreactive or useless antigen receptor (52–54). Release of RAG1 from nucleolar stores could provide a rapid mechanism for up-regulation of recombination activity in these circumstances.

While mechanisms have been described that contribute to the control of RAG activity during DSB generation and repair, they have largely focused on transcriptional regulation. For example, ATM kinase activity has been shown to be important for suppressing additional Igκ rearrangements following a RAG-mediated DNA DSB (55). This has been linked to ATM-dependent down-regulation of RAG transcription; however, the kinetics of this repression are relatively slow with roughly 50% of RAG1 protein still present 1 h postbreak (15). Sequestration of RAG1 in the nucleolus could act in concert with transcriptional repression to quench recombination quickly following a DNA DSB, thereby helping to protect genome integrity and enforce allelic exclusion. It will be important to test this idea in mice expressing mutant RAG1 proteins with defects in nucleolar localization or egress and to determine whether RAG1 from other species has the capacity to undergo regulated nucleolar sequestration and release.

Materials and Methods

Cell Culture.

vAbl-transformed pre-B cells were generated as described previously (56) and cultured in RPMI media supplemented with 10% (vol/vol) FBS (Gibco), 0.1% (vol/vol) 2-mercaptoethanol (Sigma), and penicillin-streptomycin-glutamine (Life Technologies). Cells were incubated at 37 °C in 5% CO₂.

Primary Pre-B Cell Culture.

Bone marrow cells were harvested from 4- to 6-wk-old Bcl2 transgenic mice, cultured in 15% FBS vAbl media supplemented with 5 ng/mL IL7 (BioLegend, 577804), and grown for 5 to 7 d at a density of 4 to 5 × 10⁶ cells/mL. For IL7 withdrawal, cells were washed twice with 15% vAbl media before being plated with varying concentrations of actD (Sigma) in 15% FBS vAbl media. All animal use was carried out in accordance with a protocol approved by the Animal Care and Use Committee of Yale University.

Generation of Cell Lines.

WT A70.2 vAbl pre-B cells (56) were retrovirally transduced to generate RAG1-fusion protein-expressing lines for BioID and microscopy experiments. Retroviruses were generated in Plat-E cells (Cell Biolabs). The core RAG1 vAbl cell line was made by infecting a bone marrow IL7 culture from core RAG1 × Tg(BCL2)36Wehi (002321, The Jackson Laboratory) mice with vAbl virus (21, 57). For recombination assays, WT (A70.2), core RAG1, and RAG1^−/− (11-3-4) vAbl lines (56) were transduced with retrovirus containing an inverted GFP reporter substrate (pMGInv) as described by Hung et al. (47). RAG1 expression was induced by treatment with 5 µM STI-571 (Selleckchem) for 24 or 48 h and in doxycycline-inducible systems, with 10 to 100 ng/mL doxycycline (Sigma) for 4 to 24 h.

Generation of DNA Plasmids.

The full-length murine RAG1 DNA sequence was used as the template for all subsequent RAG1 modification. BirM constructs were created by cloning the murine RAG1 open reading frame (ORF) into the following larger fusion protein construct: V5 tag (GKPIPNPLLGLKST)-BirM-10 aa linker sequence (GGSGGSGGSR)-RAG1. Once this larger ORF was constructed, FLRAG1 and subsequent truncations were cloned into the pRetroX-TRE3G (Clontech)-inducible vector using In-Fusion cloning (Takara). mCherry constructs were generated by insertion of mouse FLR1, core RAG1, ∆215, and dNOL ORFs into the mCherry2-C1 vector (54563, AddGene) using In-Fusion. The complete mCherry-RAG1 ORF was then cloned into the pRetroX-TRE3G–inducible vector for use in vAbl cells. RAG1 complementation constructs were cloned by insertion of the mouse FLR1, core RAG1, ∆215, and dNOL ORFs into the murine stem cell virus (MSCV)-internal ribosome entry site (IRES)-dTomato vector (89716, AddGene), respectively, using In-Fusion cloning. The mouse RAG2 ORF was cloned into a MSCV-IRES-blasticidin (based on 20672, AddGene) vector using In-Fusion cloning.

BioID in vAbl Cells.

RAG1-BirM fusion proteins (FLRAG1, cRAG1, and Δ215 RAG1) were inducibly expressed in RAG1^−/− vAbl cells. Treatment with STI-571, doxycycline, and biotin induced RAG1 expression while initiating promiscuous biotinylation of RAG1-proximal proteins as described by Roux et al. (37). A construct in which BirM was fused to the SV40 nuclear localization sequence was constitutively expressed in WT vAbl cells and treated in the same manner. Transduced cells were selected via puromycin and subsequently grown in bulk to a density of 7.5 × 10⁵ cells/mL and induced with 5 µM STI-571 for 3 h before initiating labeling with the addition of 50 µM biotin (Sigma) and 600 ng/mL doxycycline. Labeling was carried out for 24 h under normal growth conditions. After labeling, 30 × 10⁶ cells per sample were lysed and enriched on streptavidin beads (Pierce Streptavidin Magnetic Beads, Thermo) as described by Hung et al. (58). After washing, the beads were subjected to an on-bead trypsin digest as described by Turriziani et al. (59). Trypsinized peptides were desalted before being run on an Orbitrap-Elite LC-MS/MS to generate interactome datasets corresponding to each of the BirM-RAG1 constructs and the NLS-BirM control. Biological replicates were performed for a total of eight samples. Analysis of the LC-MS/MS data were performed using the MaxQuant/Perseus software packages (60, 61). MS/MS spectra were searched using the Andromeda algorithm against the Mus musculus proteome (UP000000589). Search parameters were for a trypsin digest with a maximum of two missed cleavages and a fixed carbamidomethyl modification. Methionine oxidation, lysine biotinylation, and N-terminal acylation were included as variable modifications. First search peptide tolerance was 20 ppm while the main search was 4.5 ppm with a protein false discovery rate (FDR) of 0.01. Protein identifications required at least two unique peptides with a minimum peptide length of seven amino acids. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE (62) partner repository with the dataset identifier PXD016221 and 10.6019/PXD016221.

Confocal Microscopy.

Confocal imaging was conducted with a Nikon-Ti microscope combined with UltraVox spinning disk (PerkinElmer). Colocalization studies were performed using fixed cell imaging of vAbl pre-B cells cotransduced with GFP-tagged fibrillarin and various doxycycline-inducible mCherry-RAG1 expression constructs. Prior to fixation, cells were induced to express their respective mCherry-RAG1 proteins with 10 ng/mL doxycycline for 24 h and, when noted, treated with 2 nM actD, or 5 µM STI-571 for 4 h. Cells were then washed and resuspended in cold PBS + 1% FBS at a concentration of 1 × 10⁶ cells/mL. The 3 × 10⁵ cells were mounted using a Cytospin centrifuge before being fixed in 4% paraformaldehyde in PBS for 15 min at 4 °C. Slides were then washed with PBS three times. ProLong Diamond Antifade Mountant with DAPI (Thermo) was applied prior to coverslip sealing. The slides were imaged as Z-stacks with spacing of 0.2 µm. Laser power and exposure time for each channel were kept consistent between slides in each dataset.

Colocalization Data Analysis.

The three-channel Z-stacks collected for each image were merged into a single two-dimensional (2D) projection using the channel merge and SD Z-projection functions in Fiji/ImageJ (63) using a custom-written macro. A pipeline was created in CellProfiler 3.0 (64) to identify nucleoli and RAG1 puncta in discrete cells and generate masks of cells containing both. Application of each image’s respective mask in Fiji/ImageJ defined the region of interest for subsequent colocalization analysis of individual cells. The Coloc 2 colocalization analysis in Fiji/ImageJ was then performed using the Costes statistical significance test with 50 iterations and approximate point spread function of five to obtain the Pearson correlation coefficient (r) for each image (65). Statistical significance of colocalization was determined by one-way ANOVA.

Flow Cytometry Analysis.

Cells were washed twice with PBS + 1% FBS and stained with DAPI (Thermo) and, when appropriate, anti-mouse Thy1.2 (53-2.1; eBioscience). Cells were then washed twice, filtered, and resuspended in PBS + 1% FBS. Data were collected on either a BD Biosciences LSR II or Stratedigm STD-13L and analyzed in FlowJo 10 (FlowJo).

vAbl Recombination Assays.

RAG activity was assessed in various vAbl cell lines using pMGInv as described by Hung et al. (47). Following RAG induction by 5 µM STI-571 treatment for 48 h, recombination levels were quantified via flow cytometry as a percentage of GFP-positive cells. In nucleolar disruption experiments, cells were treated with varying concentrations of actD for 48 h concurrent with STI-571 treatment. In a typical experiment, a culture of vAbl cells or primary pre-B cells in IL7 was divided into separate wells, each of which was treated as indicated in the relevant figure to generate the data points shown. Data shown are representative of two to three such independent experiments.

RAG1 Complementation Assays.

The pMGInv reporter and MSCV RAG2-IRES-blasticidin expression vector were integrated into RAG1^−/− vAbl cells. With the complementation of RAG1 protein, this line allowed for recombination of the integrated reporter without STI-571 treatment. These cells were clonally transduced with one of several constructs for constitutive expression of FL or mutant RAG1 proteins with dTomato as a selection marker for infection. At 72 h after infection, cells were analyzed for the ratio of number of GFP⁺ dTomato⁺ cells divided by number of dTomato⁺ cells via flow cytometric analysis.

ddPCR Jκ2 Loss Assay.

To quantify presence of intact Jκ2 alleles, primers were designed to amplify the intact 23RSS of IGκJ2 where a FAM-labeled IGκJ2-specific probe lies within the amplified region (forward [F] primer: 5′-GCCTGCCCTAGACAAACCTT; reverse [R] primer: 5′-GCTTGGTCCCCCCTCCGAAC; probe: 5′-FAM-CTCGGTGCTCAGACCATGCTCAGTTTTTGT). Murine RPP30 served as an internal standard reference gene with a HEX-labeled probe (F primer: 5′-CCAGCTCCGTTTGTGATAGT; R primer: 5′-CAAGGCAGAGATGCCCATAA; probe: 5′-HEX-CTGTGCACACATGCATTTGAGAGGT). ddPCR primers and probes were designed and ordered through Bio-Rad. Genomic DNA samples were obtained using phenol-chloroform extraction of whole cell lysates. A total of 25 ng genomic DNA was used per 20 µL ddPCR reaction, in which both Jκ2 and RPP30 reactions occurred simultaneously. Probe-based ddPCR mixes were made and droplets were generated as previously described (66). Copy number of intact Jκ2 loci was determined by calculating the ratio of the Jκ2 copy number to the RPP30 copy number. The normalized copy number for intact Jκ2 alleles was subtracted from 2 to yield the apparent level of Jκ2 cleavage. The IL7 culture ddPCR data points for Fig. 6C derive from two independent experiments with two or eight technical replicates. The kidney and splenocyte data were collected using two biological replicates with four technical replicates collected for each.