Spliceosome component PHD finger 5A is essential for early B lymphopoiesis Free

In most mammals, B cell development primarily originates from hematopoietic stem cells in the fetal liver or bone marrow (BM) post-birth (Hardy and Hayakawa, 2001). The process of B cell development in the BM, spanning from fractions (Fr.) A to F, as classified by Hardy (Hardy et al., 2007), is characterized by a stepwise recombination of the immunoglobulin (Ig) gene. Ig heavy (IgH) chain gene assembly follows a specific sequence. First, the diversity (D_H) gene segment pairs with the joining (J_H) gene, followed by the variable (V_H) gene segment joining the rearranged D_HJ_H genes (Alt et al., 1984). Upon successful in-frame recombination, a functional IgH chain pairs with a surrogate light chain, composed of VpreB and λ5, to form a pre-B cell receptor (pre-BCR) (Martensson et al., 2007). Signal transduction mediated by the pre-BCR ensures the survival of pre-B cells and their further maturation. Ultimately, a fruitful rearrangement of the Ig light chain (IgL) variable gene (V_L) to the joining gene (J_L) at the IgL chain gene locus leads to the formation of a functional IgL chain (Alt et al., 1980). The IgH and IgL chains together form the completed B cell receptor (BCR), displayed on the surface of immature B cells (Melchers, 2015). After an extensive negative selection that eliminates those immature B cells with self-reactive BCRs, the surviving B cells that express functional and non-self-reactive BCRs migrate to peripheral lymphoid tissues and become mature B cells capable of recognizing foreign antigens (Wardemann et al., 2003).

The regulation of V(D)J gene rearrangement involves numerous factors. First, the trans-factor recombination-activating gene (RAG) endonuclease plays a crucial role. It initiates Igh V(D)J recombination in progenitor B cells by binding to a recombination signal sequence (RSS) within the DJ_H recombination center (RC) to generate DNA cleavage (Dai et al., 2021). In addition, the cis-elements in the Igh locus, such as intergenic control region 1 (IGCR1), D_H segments, intronic enhancer (iEμ) and 3′ ends of Igh (3′RR), form the CCCTC-binding factor (CTCF)-binding element (CBE)-anchored three-dimensional (3D) chromatin loop domains (Zhang et al., 2019). These loop extrusion processes enhance the accessibility of V_H to RC, thereby promoting RAG chromatin scanning (Zhang et al., 2022). Furthermore, the 3D organization is an important transcriptional regulator governing the transcription of germline transcript of Igh genes, which controls their chromatin accessibility to the RAG recombinase, further facilitating the V(D)J rearrangement (Cuartero et al., 2023; Afshar et al., 2006).

Recent studies have demonstrated that post-transcriptional modifications, such as RNA splicing and A to I editing, play crucial regulatory roles in B cell development (Ruan et al., 2022; Chen et al., 2022). Pre-mRNA splicing is conducted by a large ribonucleoprotein complex, known as the spliceosome, consisting of five small nuclear ribonucleoproteins (snRNPs) and many non-snRNP proteins. A plethora of studies have affirmed the crucial involvement of spliceosome components in the development, differentiation and function of diverse immune cells (Pioli et al., 2014; Preussner et al., 2012; Ergun et al., 2013). Impaired spliceosome assembly and RNA splicing are implicated in the onset of autoimmune diseases and cancers (Katsuyama et al., 2019; Bonnal et al., 2012; Yang et al., 2021). The assembly of the spliceosome is stringently dependent on conserved sequence elements within introns, namely the 5′ splice site (5′SS), branch point sequence (BPS) and 3′ splice site (3′SS) (Shi, 2017). Precise and dynamic assembly of the spliceosome is essential for the accurate excision of introns. Splicing primarily occurs alongside transcription (Gomez Acuna et al., 2013). Studies have shown that histone modifications recruit splicing factors and induce chromatin modifications, which in turn aid the spliceosome in identifying exons amidst extensive noncoding intronic regions. This process interlinks splicing, transcription and chromatin structure (Gomez Acuna et al., 2013; Haque and Oberdoerffer, 2014). Furthermore, the spliceosome is instrumental in preventing R-loop accumulation and modulating the expression of genes that are essential for genome stability, thus safeguarding genomic integrity (Tam and Stirling, 2019).

Recent research has highlighted the U4/U6.U5 triple small nuclear ribonucleoprotein (tri-snRNP) as a key player in early B lymphopoiesis (Ruan et al., 2022). This finding leads to the possibility that other spliceosomal snRNPs may have analogous roles in the early stages of B cell development. The PHD finger protein 5A (Phf5a), known as Rds3 in yeast, was initially identified as a component of spliceosome U2 snRNP within the spliceosome. Studies have shown that Rds3 is necessary for the appropriate assembly of the yeast U2 snRNP and for maintaining spliceosome stability (Wang and Rymond, 2003). In human spliceosomal machinery, Phf5a also functions as a component of the U2 snRNP subcomplex splicing factor 3b (SF3b). Cryo-electron microscopic study of the U2 snRNP structure reveals Phf5a as a central scaffolding protein that interacts with SF3B1, SF3B3, SF3B5 and the intron region of the pre-mRNA (Teng et al., 2017). Specifically, Phf5a works in tandem with SF3B1 to identify the BPS, a crucial step for correct splicing (Cretu et al., 2016). Phf5a, which is characterized by a typical PHD domain, is highly conserved across eukaryotic species and is hypothesized to be crucial for chromatin-associated transcription regulation (Trappe et al., 2002). Previous studies have indicated the importance of Phf5a in RNA elongation for genes governing pluripotency and differentiation, and its regulatory influence on tumorigenesis (Strikoudis et al., 2016; Wang et al., 2019; Chang et al., 2021; Zheng et al., 2018).

Recent findings have implicated Phf5a as a pivotal regulator of DNA repair during IgH chain gene class switch recombination (CSR) in B cells (Begum et al., 2021). Studies showed that Phf5a plays a crucial role in non-homologous end joining (NHEJ)-dependent DNA repair, as it preserves chromatin structure, thereby facilitating an effective DNA damage response and the subsequent assembly of NHEJ machinery at the S region. Phf5a stabilizes the p400 histone chaperone complex at this locus, promoting the incorporation of H2A variants H2AX and H2AZ. These variants are essential for initiating the early DNA damage response and the NHEJ process. Nevertheless, the precise functions of Phf5a in early B cell development remain to be fully elucidated.

In this study, we have systematically assessed the role of Phf5a in early B cell development by specifically deleting Phf5a in the B cell lineage. Our results show that Phf5a is essential for early B lymphopoiesis. Furthermore, we unravel that Phf5a facilitates Igh V-to-DJ gene rearrangement by orchestrating RAG activity and modulating Igh chromatin interaction in a spliceosome-dependent manner.

To determine the specific stage at which Phf5a may play a role in early B cell development, we first analyzed the expression of Phf5a across various B cell subsets using the Immunological Genome Project RNA-seq dataset (Ergun et al., 2013). We observed pronounced Phf5a expression in early B cell subsets, notably in pre-pro-B (Fr. A) and pro-B cells (Fig. 1A), indicating that Phf5a could be involved in early B lymphopoiesis. To investigate its physiological function in B cell development, we generated B cell-specific Phf5a-deficient mice (Phf5a^f/fMb1^Cre/+; hereafter, Phf5a bKO) by breeding Phf5a^f/f mice with Mb1^Cre/+ mice (Fig. S1A). The Mb1-Cre is known to express from the Fr. A B cell stage onwards, with significantly higher expression in pro-B cells and beyond, thus ensuring efficient deletion of floxed genes during early B lymphopoiesis. Our results showed that Mb1-Cre-mediated deletion of the Phf5a gene achieved ∼50% efficiency in Fr. A (pre-pro-B) cells (Fig. S1B, S1C), and the deletion efficiency increased to 80% in the more mature Fr. B (early pro-B) cells, as confirmed by both Phf5a mRNA and protein levels measured by qRT-PCR and western blot analysis, respectively (Fig. S1D,E).

When the B cell development was assessed using flow cytometric analysis, we observed a significant decrease, approximately sixfold, in both the percentage and the absolute number of total B220⁺ B cells in the BM of Phf5a bKO mice compared with wild-type mice (Fig. 1B,C). A more-detailed analysis of various B cell subsets in the BM revealed that both the percentages and absolute numbers of B220⁺CD43⁻ late-stage B (LSB) and B220^highCD43⁻ recirculating B cells were significantly diminished in Phf5a bKO when compared with the wild-type mice (Fig. S2A,B). In contrast, early-stage B (ESB) cells, identified as B220⁺CD43⁺, were comparable between Phf5a bKO and wild-type mice (Fig. S2A,B). Further analysis of ESB cells, using Hardy's classification, revealed a marked increase in both the frequency and the absolute number of BP1⁻CD24⁻ pre-pro-B (Fr. A) cells in Phf5a bKO compared with wild-type mice (Fig. 1D,E). In contrast, both the percentages and the absolute numbers of BP1⁻CD24⁺ early pro-B (Fr. B), BP1⁺CD24^high late pro-B (Fr. C) and BP1⁺CD24^low large pre-B (Fr. C′) were significantly reduced in Phf5a bKO mice compared with wild-type mice (Fig. 1D,E). In addition, the populations of B220⁺CD43⁻IgM⁻IgD⁻ small pre-B (Fr. D) and B220⁺CD43⁻IgM⁺IgD⁻ immature B (Fr. E) cells were drastically diminished in Phf5a bKO compared with wild-type mice (Fig. 1F,G).

Consistent with the severe impairments in B cell development in the BM, the B220⁺ splenic B cell population was also dramatically diminished in Phf5a bKO mice (Fig. 1H,I). Furthermore, IgD⁺ mature B cells were virtually absent in the spleen of the mutant mice, as revealed by confocal microscopy (Fig. 1J). Additionally, the spleens of Phf5a bKO mice were noticeably smaller in size compared with wild-type mice (Fig. S2C), indicating that Phf5a deficiency leads to severe B lymphopenia in the periphery.

We also carried out a bromodeoxyuridine (BrdU) incorporation assay and a CaspGLOW fluorescein active caspase staining assay to assess the proliferation and survival of the Phf5a bKO pre-pro-B cells. The frequencies of BrdU⁺ and Caspase⁺ pre-pro-B cells were comparable between Phf5a-deficient and wild-type B cells (Fig. S3), suggesting that the increase in the proportion of pre-pro-B cells in the mutant mice is not due to altered proliferation or to survival of these progenitor B cells in the absence of Phf5a. Our findings underscore the essential role of Phf5a in B cell development, with its deficiency leading to a severe block at the transition from pre-pro-B to early pro-B cell stage during early B lymphopoiesis.

To elucidate the mechanistic underpinnings of the developmental block in B cells of Phf5a bKO mice, we subsequently performed a bulk RNA sequencing (RNA-seq) analysis on sorted pre-pro-B cells. We identified 163 differentially expressed genes (DEGs) in the mutant B cells, with ∼95% showing deregulation (Fig. 2A). Gene ontology (GO) enrichment analysis of the downregulated DEGs indicated a strong association with the Ig complex, BCR signaling and binding functions (Fig. 2B). A closer examination of the RNA-seq data revealed that the mutant pre-pro-B cells exhibited a global reduction in IgH chain gene expression, especially within the V segment genes, in comparison with wild-type cells (Fig. 2C). Thus, our findings suggest that Phf5a plays a vital role in IgH chain gene expression during early B lymphopoiesis. Its absence causes a severe developmental block at the transition from pre-pro-B to early pro-B cell stage, where the expression of IgH chains commences during early B lymphopoiesis.

We proceeded to assess the diversity of IgH chain repertoires by examining the clonality of the variable complementarity determining region 3 (CDR3) of IgH chains using RNA-seq data. The diversity of IgH clonotypes in Phf5a-deficient B cells was significantly decreased compared with wild-type B cells, as illustrated by the rarefaction curve (Fig. 3A). Similarly, the Chao1 index, a measure of the clonal diversity of the IgH chain, also showed a marked decrease in the mutant B cells compared with wild-type B cells (Fig. 3B). These results suggest that the deficiency of Phf5a significantly reduces the diversity of IgH chain repertoires.

We further investigated the impact of Phf5a deficiency on IgH gene rearrangement. Since pairing V_H and J_H segments is a crucial aspect of V(D)J rearrangement, we first examined the V_H-J_H pairing pattern using Circos plots. The Phf5a bKO B cells displayed significantly fewer V_H-J_H junctions than wild-type B cells, which manifested distinct polyclonal parings, notably IghV1-26 with IghJ4, and IghV5-9 with IghJ2 (Fig. 3C). We also accessed the clonotype frequency distribution of the CDR3 regions. Unlike the broad CDR3 length distribution in wild-type B cells, Phf5a-deficient B cells predominantly exhibited a CDR3 length of 42 bp (Fig. 3D). Moreover, we employed TRUST4 further to evaluate the V(D)J rearrangement. The analysis revealed that the Phf5a bKO B cells exhibited a higher incidence of out-of-frame IgH CDR3 than wild-type cells (Fig. 3E), suggesting that Phf5a deficiency led to a substantial increase in out-of-frame V(D)J rearrangement of IgH genes.

Next, we directly assessed IgH V(D)J rearrangement in sorted Phf5a bKO pre-pro-B and early pro-B cells using semi-quantitative PCR. Our analysis revealed that the D-J_H rearrangement was comparable between Phf5a bKO and wild-type pre-pro-B cells (Fig. 3F). However, V_H-DJ_H rearrangement was more substantially reduced in Phf5a-deficient early pro-B cells than in wild-type controls (Fig. 3G). These findings suggest that Phf5a is crucially required for V_H-DJ_H but not for D-J_H rearrangement.

To ascertain whether the severely compromised early B cell development was due to impaired IgH V(D)J gene rearrangement, we introduced a pre-arranged IgH transgene into the Phf5a bKO mice to assess whether it could rescue the defective B cell development in the mutant mice. To this end, we crossed Phf5a^f/fMb1^Cre/+ mice with MD4^Tg/+ mice, which express a hen egg lysozyme-specific transgenic BCR, to generate Phf5a^f/fMb1^Cre/+MD4^Tg/+ (Phf5a Tg bKO) mice. Flow cytometric analysis revealed that the expression of MD4 BCR transgene could partially but significantly increase the number of CD19⁺ B cells in the BM of Phf5a Tg bKO mice (Fig. 3H,I). These results suggest that Phf5a is essential for early B cell development through regulating IgH V-DJ gene rearrangement.

We next attempted to delve deeper into the molecular mechanism whereby Phf5a regulates IgH chain gene arrangement. The cis-interaction within the Igh locus and the resultant chromatin conformational changes are crucial for V(D)J rearrangement (Guo et al., 2011; Hu et al., 2015). A chromatin rosette structure at the Igh locus, which involves the interactions of IGCR1, 3′RR and the Eμ enhancer in the RAG-bound recombination center, orchestrates the sequential D-J_H and V_H-DJ_H rearrangements. We first determined whether the chromatin conformation at the Igh locus was affected by the absence of Phf5a. To this end, we performed a high-throughput chromosome conformational capture (Hi-C) analysis to examine chromatin looping at the Igh locus.

Our Hi-C analysis showed that the interaction of 3′RR and Eμ was completely disrupted in Phf5a-deficient pre-pro-B cells, as the interaction intensity between two loci was not detected in these cells (Fig. 4A, region A), indicating a compromised interaction of the RC with other loops at the 3′ Igh locus. Additionally, we noted that the 3′RR-V_H81X interaction was slightly inhibited, whereas the IGCR1-V_H81X interaction was enhanced in Phf5a-deficient B cells (Fig. 4A, regions C and D). These findings suggest that the interaction between the V_H81X and the RC was impaired in the mutant B cells. Given that the V_H81X is a highly used V_H gene segment during V(D)J gene rearrangement (Marshall et al., 1996), the aberrant joining of V_H81X to RAG-bound RC in Phf5a-deficient pre-pro-B cells indicates that the V_H to D_HJ_H gene rearrangement is compromised during the transition from the pre-pro-B to the pro-B stage. These results suggest that Phf5a is instrumental in facilitating proper chromatin interactions at the Igh locus during V_H-DJ_H gene rearrangement.

The IgH V(D)J gene rearrangement commences with the recognition of the RSSs, which flank the V, D and J gene segments, by V(D)J recombinase RAG, a multi-protein complex comprising proteins such as RAG1 and RAG2 (Spicuglia et al., 2006). The RAG recombinase generates double-stranded DNA breaks between the coding gene segments and their adjacent RSS, creating a hairpin loop in the coding segment and a blunt end in the signal segment. These hairpin loops are subsequently processed and joined by the DNA repair machinery NHEJ (Lieber, 2010). To delve deeper into the molecular mechanism underlying the defective IgH chain gene rearrangement, we investigated whether RAG-mediated recombination was defective in the absence of Phf5a.

We designed a report assay to assess the RAG-mediated recombination by constructing a retroviral fluorescent reporting vector MIGR1-23>/12>RSS-mCherry, in which the reporter gene mCherry was inserted in the opposite orientation and flanked by a 23 RSS and a 12 RSS, and transducing the Phf5a-deficient BM cells with the reporter construct (Fig. S4A). A functional RAG-mediated recombination would result in the inversion of the mCherry gene, thereby establishing the expression of mCherry fluorescent protein. The RAG-mediated recombination can be quantitatively assessed by measuring the proportion of mCherry⁺ cells among the GFP⁺ transduced BM cells. We also constructed a MIGR1-12>/12>RSS-mCherry reporter gene, which can not be recognized by the RAG recombinase, as a negative control (Fig. S4B). Upon transduction of wild-type BM cells with the MIGR1-23>/12>RSS-mCherry reporter gene, ∼20% mCherry⁺ cells were detectable among the total GFP⁺ cells. In contrast, no mCherry⁺ cells were detected when wild-type BM cells were transduced with the MIGR1-12>/12>RSS-mCherry negative control construct (Fig. 4B,C). Notably, no mCherry⁺ cells were detected when Phf5a bKO BM cells were transduced with the MIGR1-23>/12>RSS-mCherry reporter (Fig. 4B,C), suggesting that RAG-mediated recombination is defective in the absence of Phf5a. We further showed that the expression of RAG1 and RAG2 was unaffected by Phf5a deficiency at either mRNA or protein levels (Fig. 4D,E, Fig. S4C), indicating that the defective RAG-mediated recombination is not due to compromised expression of RAG1 and RAG2. Considering that the transduced reporter construct is most likely not constrained by the complex structure at the IgH locus, our findings imply that Phf5a might regulate IgH V(D)J rearrangement through mechanisms beyond regulating chromatin interactions at the Igh locus.

Phf5a acts as a scaffold protein within the U2 snRNP complex, which is crucial for recognizing intron branch sites during pre-mRNA splicing (Cretu et al., 2016). We next assessed whether the RNA splicing is compromised in Phf5a-deficient B cells. By extracting typical 5′SS (9 bp) and 3′SS (23 bp) sequences from splicing junctions in RNA-seq data and performing sequence motif analysis, we determined that both wild-type and Phf5a-deficient B cells displayed similar sequence motifs at the 5′SS and the 3′SS (Fig. 5A,B). These results suggest that the splice site recognition is unaffected in the Phf5a-deficient B cells.

We also examined the splicing efficiency in Phf5a-deficient B cells by evaluating the catalytic activity of the spliceosome. We constructed a retroviral reporter plasmid, MIGR1-CD8-i8a, which contains the coding region and the first intron of the murine Cd8a gene (Fig. S5A). Proper spliceosome functions will result in the excision of the inserted intron, leading to cell surface expression of CD8a, which we can measure by the presence of CD8⁺ cells among the GFP⁺ transduced BM cells. Using this assay, we found that both wild-type and Phf5a-deficient pre-pro-B cells exhibited ∼90% splicing efficiency (Fig. 5C,D). Conversely, the cells transduced with negative control, MIGR1-CD8a-δi1 carrying a G to C mutation in the conserved 5′SS sequence in the first intron of Cd8a gene, demonstrated only ∼5% splicing efficiency (Fig. 5C). These results suggest that the absence of Phf5a does not affect the basal pre-mRNA splicing activity of the spliceosome.

Noncoding germline transcripts (GLTs) at the Igh locus also play a crucial role in rendering the locus accessible, which in turn facilitates V(D)J rearrangement by making the chromatin more amenable to the RAG recombinase (Bolland et al., 2004). RNA splicing and polyadenylation are essential for GLT formation (Teng and Schatz, 2015). To investigate whether the impaired V(D)J rearrangement could be due to the compromised germline transcription at the Igh locus, we measured the levels of representative GLTs at the Igh locus. Our analysis showed that the levels of both spliced and unspliced transcripts of the antisense transcripts of PAIR4 and PAIR6 were comparable in wild-type and Phf5a bKO B cells (Fig. 5E, Fig. S5B). Likewise, the levels of Cμ and Iμ GLTs were not significantly different between wild-type and Phf5a bKO B cells (Fig. 5E, Fig. S5B). These data suggest that Phf5a deficiency does not impact the expression or splicing of GLTs at the Igh locus, and, consequently, the impaired V_H to DJ_H rearrangement in Phf5a-deficienct B cells is not due to altered expression or to splicing of GLTs.

A previous study has shown that Phf5a can regulate gene transcription through its PHD domain (Strikoudis et al., 2016). To further dissect the mechanisms whereby Phf5a governs early B cell development, we sought to determine which function – the transcriptional or the U2 snRNP assembly – is primarily responsible for its role in B cells. To this end, we first generated a D47A mutant by replacing the aspartic acid at position 47 with alanine. This mutation is known to abolish the transcriptional activity of Phf5a without affecting its U2 snRNP assembly function (Zheng et al., 2018). We also constructed a C72A mutant, which harbors a point mutation that converts the cysteine at position 72 into alanine in the Phf5a protein. The C72A mutation is known to disrupt the interaction of Phf5a with other components of U2 snRNP, leading to impaired spliceosome assembly without affecting its transcriptional activity (Begum et al., 2021). We then conducted BM chimera experiments by mixing 20% of wild-type BM cells, retrovirally transduced with an empty vector, or 20% of Phf5a bKO BM cells, retrovirally transduced with an empty vector, wild-type Phf5a, Phf5a-D47A or Phf5a-C72A constructs, with 80% of BM cells from the μMT mice. These mixed BM cells were then transferred into the sub-lethally irradiated wild-type recipient mice. After 5 weeks of reconstitution, we analyzed the B cells derived from successfully transduced (GFP⁺) BM cells in the chimeric mice using flow cytometry. We found that the chimeric mice reconstituted with Phf5a bKO BM cells carrying wild-type Phf5a displayed a comparable percentage of B cells in their spleens to the mice reconstituted with wild-type BM cells transduced with an empty vector (Fig. 5F,G). Almost no GFP⁺B220⁺ B cells were detected in the chimeric mice reconstituted with Phf5a bKO BM cells transduced with an empty vector, mirroring the B lymphopenia phenotype seen in Phf5a bKO mice. Importantly, the chimeric mice reconstituted with Phf5a bKO BM cells transduced with Phf5a-D47A mutant exhibited a comparable percentage of GFP⁺B220⁺ B cells in their spleens compared with the mice reconstituted with wild-type Phf5a-transduced Phf5a bKO BM cells (Fig. 5F,G). In contrast, the Phf5a-C72A mutant failed to restore the function of wild-type Phf5a, as evidenced by the negligible presence of GFP⁺B220⁺ B cells in the spleens of chimeric mice reconstituted with Phf5a bKO BM cells transduced with the Phf5a-C72A mutant construct (Fig. 5F,G). These results suggest that the role of Phf5a in early B cell development is dependent on its association with U2 snRNP.

Our current study has demonstrated that B cell lineage-specific ablation of the U2 snRNP component Phf5a led to a severe block at the transition from pre-pro-B to early pro-B cell stage of B cell development, underscoring its pivotal role in early B lymphopoiesis. Although previous studies have shown that Phf5a regulates cell proliferation and survival, with its upregulation correlated with tumor progression (Wang et al., 2019; Chang et al., 2021; Zheng et al., 2018), we observed that Phf5a ablation in B cells did not notably affect B cell proliferation and survival. Similarly, Usp39, a U4/U6.U5 tri-snRNP component, was also found to be dispensable for B cell proliferation and survival (Ruan et al., 2022). These findings collectively indicate that these spliceosome components are not crucial for B cell proliferation and survival or may have overlapped functions in these processes. Nevertheless, our findings clearly suggest these spliceosome components such as Phf5a play indispensable roles during early B cell development that extend beyond the regulation of proliferation and survival.

As a key spliceosomal component, the primary function of Phf5a is to orchestrate pre-mRNA splicing by facilitating U2 snRNP assembly and BPS recognition within introns. Interestingly, our study found that the absence of Phf5a did not affect splice site recognition or global basal pre-mRNA splicing in B cells, suggesting that Phf5a is dispensable for these processes during normal B cell development. This is consistent with recent findings in human brain tumor cells and mouse embryonic stem cells, where Phf5a silencing by siRNA or CRISPR did not affect exon recognition and splicing in normal cells, although defects were observed in malignant cells (Hubert et al., 2013; Strikoudis et al., 2016). Similarly, previous studies showed that the loss of Usp39 has no significant impact on splice site binding or basal RNA splicing (Ruan et al., 2022). These insights lead us to propose that spliceosome components, including Phf5a and Usp39, might be involved in B cell development through spliceosome-dependent mechanisms that are distinct from their conventional roles in RNA splicing, considering the complex and dynamic nature of spliceosomes.

Although the role of Phf5a in regulating cell proliferation, survival and pre-mRNA splicing appears to be non-essential in early B cells, our study has identified an essential role for Phf5a in regulating IgH V(D)J gene rearrangement. This process is initiated by the RAG recombinase, the activity of which is influenced by chromatin loops at the Igh locus, shaped by interactions among cis-elements (Dai et al., 2021). Our study demonstrated that the absence of Phf5a disrupted the interaction between the RC and other loops, resulting in compromised loop extrusions and impaired RAG scanning of the RC to the V_H segments at the Igh locus. The impaired chromatin loop formation and subsequent RAG scanning could lead to defective RAG recombinase activity, thereby undermining IgH V(D)J gene rearrangement and blocking B cell development at the transition from pre-pro-B to early pro-B cell stage. Further studies are required to elucidate the precise mechanisms by which spliceosome components, such as Phf5a, interact with the Igh locus and promote chromatin looping.

A recent study has highlighted the role of Phf5a in regulating IgH CSR, demonstrating its involvement in the regulation of NHEJ-dependent DNA repair by maintaining chromatin integrity for an optimal DNA damage response (Begum et al., 2021). In particular, Phf5a stabilizes the p400 histone chaperone complex, thereby promoting the incorporation of H2A variants that are essential for the DNA damage response and NHEJ in a U2 snRNP-dependent manner. Given that RAG recombinase also induces double-strand DNA breaks during V(D)J rearrangement, which are then rejoined by NHEJ-dependent DNA repair (Lieber, 2010), it is tempting to speculate that Phf5a could play a similar role during V(D)J rearrangement. This hypothesis is supported by our mCherry reporter assay, which indicated that Phf5a could modulate recombination through means beyond mere chromatin interaction regulation. Future studies should delve more deeply into the interactions of Phf5a with histones, which will likely shed more light on the intricate regulatory mechanisms for IgH V(D)J rearrangement.

All mice were on a C57BL/6 background. Conditional knockout or transgenic mice were obtained by crossing the following strains: Phf5a^f/f were purchased from Cyagen Biosciences; and Mb1^Cre/+ (B6.C(Cg)-Cd79a^tm1(cre)Reth/EhobJ) and MD4^Tg/+ (MD4 transgenic) mice were obtained from the Jackson Laboratory. The Phf5a^f/f mice were crossed with Mb1^Cre/+ mice to generate Phf5a^f/fMb1^Cre/+ mice. All experiments were conducted using 6- to 14-week-old male or female mice, and all animal experiments were approved by the Institutional Animal Care and Use Committee of Southern University of Science and Technology.

Flow cytometry and cell sorting were performed as described previously (Xu and Lam, 2002). Cells were harvested from the BM or spleens, incubated with red blood cell lysis buffer (RBC, Sangon Biotech) on ice and filtered through a 70 µm nylon membrane. Cells were counted and stained with antibodies conjugated with various fluorescent dyes or biotin, including anti-B220 (RA3-6B2, BD Pharmingen), anti-CD43 (R2/60, eBiosciences), anti-CD24 (M1/69, BD Pharmingen), anti-CD249 (BP-1, BD Pharmingen), anti-IgM (RMM-1, BioLegend) and anti-IgD (11-26c; eBiosciences). The samples stained with the biotinylated antibodies were further stained with APC-streptavidin (Biolegend) or PE-streptavidin (eBioscience). All antibodies were used at 1:100 dilution in FACS buffer or as specified. Finally, the cells were analyzed using a FACS flow cytometer (Beckman) or sorted on a FACSAria cell sorter (BD Biosciences), with dead cells excluded electronically using DAPI staining. The detailed key sources are provided in Table S1.

Cell apoptosis was evaluated by detecting activated pan-caspases using a CaspGLOW Fluorescein Active Caspase Staining Kit (BioVision, K180) following the manufacturer's instruction as described previously (Xu et al., 2012). Briefly, 1×10⁷ BM cells were first incubated with FITC-VAD-FMK in RPMI medium for 1 h at 37°C and then stained with cell surface markers for flow cytometry analysis.

BrdU labelling experiment was performed as described previously (Ruan et al., 2022). Mice were first injected intraperitoneally with 1 mg of BrdU (eBioscience). BM cells were collected for surface staining 24 h later. Cells were then fixed and DNA was denatured using BrdU Staining Kit (eBioscience) following the manufacturer's instructions. Finally, cells were stained with BrdU-FITC (BD Pharmingen) and subjected to flow cytometric analysis.

Immunofluorescence microscopic analysis of the mouse spleen was performed as described previously (Li et al., 2014). Spleens were first embedded in Tissue-Tek (Sakura) and cut into 10 μm sections. The cryostat sections were then fixed with ice-cold acetone for 15 min. Next, the sections were blocked with 5% serum plus 2% BSA in PBS for 1 h at room temperature. After washing, the sections were stained with anti-CD4-PE (RM4-5, eBioscience), anti-IgD-Biotin (11-26c, eBioscience) and streptavidin-APC (Biolegend). Finally, sections were analyzed with a confocal microscope (Nikon A1R).

Total RNA was extracted from sorted pre-pro-B (Fr. A) and early pro-B (Fr. B) cells using RNAiso Plus (Takara) and reverse transcribed using 1st Strand cDNA Synthesis Kit (YEASEN). qPCR SYBR Green Master Mix (YEASEN) was used for quantitative RT-PCR and the qRT-PCR was performed on a qTOWER3 real-time PCR thermal cycler (Analytik Jena). The primers used in the assay are as follows: β-Actin forward, 5′-CGTGAAAAGATGACCCAGATCA-3′; β-Actin reverse, 5′-CACAGCCTGGATGGCTACGT-3′; Phf5a forward, 5′-TGTGGCTATCGGAAGACTGTGT-3′; Phf5a reverse, 5′-GGTGC-ACTCTTTACAGTAGTAGGC-3′. The primers used in this paper are listed in Table S2.

Western blotting was conducted as described previously (Huang et al., 2022). Briefly, pre-pro-B (Fr. A) and early pro-B cells (Fr. B) were sorted and lysed in RIPA buffer (Sigma) containing protease inhibitor (Roche). Protein extracts were loaded on SDS-PAGE. After electrophoresis, separated total proteins were transferred to PVDF (Millipore) membranes, which were then blocked with 5% skim milk in TBST for 1 h at room temperature. Membranes were probed with anti-Phf5a (15554-1-AP, Proteintech, 1:1000 diluted), anti-beta-Actin (66009-1-Ig, Proteintech, 1:2000), anti-RAG1 (A12646, Abclonal, 1:1000), anti-RAG2 (A12488, Abclonal, 1:1000 diluted) or anti-GAPDH (60004-1-Ig, Proteintech, 1:5000) antibodies at 4°C overnight. After washing, the membranes were further incubated with goat anti-rabbit IgG conjugated with HRP (HS101-01, Transgene, 1:5000) or goat anti-mouse IgG conjugated with HRP (HS201-01, Transgene, 1:5000) at room temperature for 1 h. Finally, Blots were developed using a chemiluminescent HRP substrate (Millipore) and visualized using Tanon 5200 Image system (Tanon).

The total RNA of pre-pro-B (Fr. A) cells was extracted using an RNeasy mini kit (Qiagen), and poly(A) mRNA isolation was performed using Poly(A) mRNA Magnetic Isolation Module (NEB). Libraries were constructed and sequenced on an Illumina HiSeq instrument (Illumina) using a 150 bp paired-end configuration. Sequence reads were trimmed using Trimmomatic and then mapped to the mouse GRCm38 genome using Hisat2 (version 2.1.0). Gene expression analysis was performed using featureCounts (version 1.6.1). The differential gene expression analysis was conducted using DESeq2. The genes with an adjusted P-value<0.05 and a fold change>2 were considered as differentially expressed genes. Gene ontology enrichment analysis was performed by PANTHER (Mi et al., 2019).

MiXCR (version 3.0.12) and TRUST4 were used to search for reads aligning to the B cell receptor sequence (Bolotin et al., 2015; Song et al., 2021). The CDR3 sequences were used for downstream analysis using VDJtools (version 1.2.1), including clonotypes diversity, V(D)J segment usage and CDR3 properties (Shugay et al., 2015). Splice sites (SS) were predicted by Hisat2 by setting the parameter -novel-splice site-out file. 5′SS (exon region 3 bp, intron region 6 bp) and 3′SS (exon region 3 bp, intron region 20 bp) sequences were extracted using bedtools according to the mouse genome. Splice site motifs were visualized by WebLogo3 (version 3.7.4).

IgH chain V(D)J gene recombination was assessed with a previous protocol with modification (Xu et al., 2000). Briefly, genomic DNA was extracted from sorted pre-pro-B and early pro-B cells using MiniBEST Universal Genomic DNA Extraction Kit (TaKaRa). The genomic DNA was serially diluted four times, starting from 20 ng and subjected to semi-quantitative PCR using KOD-Plus Neo (TOYOBO). Primers used in the assay were identified from previous studies (Guo et al., 2011; Subrahmanyam et al., 2012). The Rosa26 gene was used as a loading control.

Germline transcript analysis was assessed by semi-quantitative PCR analysis. Total RNA was extracted from sorted pre-pro-B (Fr. A) cells and reverse transcribed using PrimeScript first Strand cDNA Synthesis Kit (TaKaRa) with mixed Oligo dT and Random 6mers primers. The synthesized cDNA was serially diluted four times and subjected to semi-quantitative PCR using KOD-Plus Neo (TOYOBO). Primers used in the assay were identified from previous studies (Bolland et al., 2004). λ5 transcript was used as a loading control and the quantitation was performed using Tanon 3500 Image System.

Total RNA was extracted from mouse splenic naive B cells using RNAiso Plus (TaKaRa) and reverse transcribed using PrimeScript first-strand cDNA synthesis kit (Takara). Constructions were carried out using ClonExpress MultiS one Step Cloning Kit (Vazyme). All plasmids were constructed based on the MSCV internal IRES-GFP pseudo-type retroviral vector. The reporter plasmids MIGR1-CD8a, MIGR1-CD8a-i1, MIGR1-CD8a-δi1, MIGR1-23>/12>RSS-mCherry and MIGR1-12>/12>RSS-mCherry were constructed as described previously (Ruan et al., 2022; Trancoso et al., 2013). A Phf5a-D47A mutation was achieved by the amino acid substitution of aspartic acid (GAT) for alanine (GCT) at position 140. A Phf5a-C72A mutation was achieved by the amino acid substitution of aspartic acid (TGT) for alanine (GCT) at positions 214 and 215. All the inserted sequences were confirmed by Sanger sequencing.

The HEK293T cells were a gift from Li Yan's lab (School of Life Sciences, Southern University of Science and Technology, Shenzhen, China). The OP9 cell line was purchased from Cell Bank of Chinese Academy of Sciences. The cell lines were cultured at 37°C with 5% (v/v) CO₂ in DMEM medium (Hyclone) for HEK293T cells and α-MEM medium (Hyclone) for OP9 cell line supplemented with 10% fetal bovine serum (Gibco) and 1% penicillin/streptomycin (Hyclone).

Retrovirus transduction was carried out as described previously (Li et al., 2022). Briefly, virus particles were packaged using HEK293T cells. All vectors were co-transfected with pCL-ECO plasmid (1:1 w/w) using lipofectamine 3000 (2:1 w/w, Invitrogen). The viral supernatants were collected and filtered using a 0.45 µm membrane at 48 h and 72 h after transfection. For the analysis of B cell development in vitro, BM cells were cultured together with OP9 cells and transduced with retrovirus supernatants for 90 min at 700 g at 37°C and cultured with IL-7 (10 ng/ml, PeproTech, 217-17). Every other day, BM cells were harvested and re-cultured with fresh OP9 cells in DMEM medium supplemented with 10 ng/ml IL-7. After 7 days, cells were collected for flow cytometric analysis.

BM chimera assay was carried out as described previously (Ou et al., 2014). Mice were injected i.p. with 5-fluorouracil (0.15 mg/g). BM cells were harvested 3 days later and cultured in DMEM (HyClone) medium supplemented with 20% FBS (Gibco), 1% penicillin/streptomycin (HyClone), 1% nonessential amino acids (Life Technologies), IL-3 (20 ng/ml, R&D, 403-ML), IL-6 (50 ng/ml, R&D, 206-IL) and stem cell factor (50 ng/ml, R&D, 455-MC) for 24 h. Cells were then transduced twice with retroviruses packaged (48 h and 72 h) over 2 days. After transduction, cells were mixed with μMT BM cells (1:4) and injected i.v. into lethally irradiated wild-type mice (900 rads). The mice were examined 5 weeks later.

In situ Hi-C was performed based on the previously described protocol (Rao et al., 2014). Five biological replicates of wild-type pre-pro-B (Fr. A) cells and four biological replicates of Phf5a bKO pre-pro-B (Fr. A) cells were merged, respectively, for a Hi-C library. Two independent experiments were conducted for each genotype. Illumina NovaSeq was used for the sequencing with a read length of 150 bp in the paired-end mode. The sequencing data were first processed with HiCUP (version 0.7.4). Mus_musculus.GRCm38.dna.Primary assembly was used as the reference genome. Normalized contact matrices were then generated using Juicer tools (version 1.22.01). Next, the calculate map resolution script in the Juicer was used to calculate the resolution of the Hi-C data. Arrowhead and HiCCUPS algorithms were used to annotate contact domains and chromatin interactions, respectively, with enrichment statistical significance of FDR <0.01. Finally, contact matrices visualization was generated using Juicebox (version 1.11.08).

All statistical analyses were performed using GraphPad Software Prism (version 9.01). Statistical analysis was performed using an unpaired two-tailed Student's t-test. Data are mean±s.d. Statistical significance is indicated as follows in all figures: ns, not significant; *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001. The details of the experiments are listed in the corresponding figure legends.

We thank SUSTech Core Research Facilities and Laboratory Animal Research Center for their assistance with animal experiments.