YY1 knockout in pro-B cells impairs lineage commitment, enabling unusual hematopoietic lineage plasticity

In this study, Banerjee et al. describe a role for transcription factor YY1 in B-cell fate commitment. They show that a YY1 deficiency drives lineage plasticity in pro-B cells and promotes transcriptional programs associated with multiple hematopoietic lineages, including T cells, indicating its potentially ubiquitous role in immune cell differentiation.

Abstract

During B-cell development, cells progress through multiple developmental stages, with the pro-B-cell stage defining commitment to the B-cell lineage. YY1 is a ubiquitous transcription factor that is capable of both activation and repression functions. We found here that knockout of YY1 at the pro-B-cell stage eliminates B lineage commitment. YY1 knockout pro-B cells can generate T lineage cells in vitro using the OP9-DL4 feeder system and in vivo after injection into sublethally irradiated Rag1^−/− mice. These T lineage-like cells lose their B lineage transcript profile and gain a T-cell lineage profile. Single-cell RNA-seq experiments showed that as YY1 knockout pro-B cells transition into T lineage cells in vitro, various cell clusters adopt transcript profiles representing a multiplicity of hematopoietic lineages, indicating unusual lineage plasticity. In addition, YY1 KO pro-B cells in vivo can give rise to other hematopoietic lineages in vivo. Evaluation of RNA-seq, scRNA-seq, ChIP-seq, and scATAC-seq data indicates that YY1 controls numerous chromatin-modifying proteins leading to increased accessibility of alternative lineage genes in YY1 knockout pro-B cells. Given the ubiquitous nature of YY1 and its dual activation and repression functions, YY1 may regulate commitment in multiple cell lineages.

Development of the B lymphocyte lineage follows an ordered progression of cell stages, including common lymphoid progenitors (CLPs), pre-pro-B cells, pro-B cells, pre-B cells, immature B cells, and additional more mature and specialized B cells (Henderson and Calame 1998; Bartholdy and Matthias 2004; Shapiro-Shelef and Calame 2005; Hagman and Lukin 2006; Boller and Grosschedl 2014; Choukrallah and Matthias 2014; Pang et al. 2014; Rothenberg 2014; Miyai et al. 2018; Aubrey et al. 2022; Peña-Pérez et al. 2022). Precursor cells (pre-pro-B and earlier) are capable of adopting alternative lineage fates, but lineage commitment occurs at the pro-B-cell stage subsequent to somatic rearrangement of the immunoglobulin heavy chain (IgH) gene variable, diversity, and joining (VDJ) segments to produce a functional IgH gene (Nutt et al. 1999; Jung et al. 2006).

During the transition to committed pro-B cells, specific transcription factors function to initiate gene expression profiles necessary for development of B lineage cells while simultaneously repressing the expression of genes needed for the development of alternative lineages (Bain et al. 1994; Nutt et al. 1997; Henderson and Calame 1998; Su et al. 2003; Busslinger 2004; Hagman and Lukin 2006; Liu et al. 2007; Banerjee et al. 2016; Kleiman et al. 2016). Transcription factors EBF1 and E2A are crucial regulatory proteins that initiate early stages of B lineage development, enabling expression of key genes such as Igll1, VpreB1, Cd79a, Rag1, and Rag2, as well as expression of transcription factor Pax5. Pax5 further strengthens the B lineage pathway by (1) stabilizing the B lineage expression profile, (2) repressing the expression of alternative lineage genes, (3) inducing IgH V-to-DJ rearrangements, and (4) initiating locus contraction of the IgH locus (Kee and Murre 2001; Fuxa et al. 2004). Upon executing these regulatory networks at the pro-B-cell stage, the cells are exclusively committed to the B-cell lineage and cannot adopt other lineage fates.

Transcription factor Yin Yang 1 (YY1) is a ubiquitously expressed factor that plays significant roles in normal biological processes including cell differentiation, proliferation, replication, DNA repair, and embryogenesis. Embryonic YY1 knockout results in peri-implantation lethality, indicating its critical role during embryogenesis (Donohoe et al. 1999). YY1's name is based on its ability to both activate expression of some genes and repress expression of others (Hariharan et al. 1991; Park and Atchison 1991; Shi et al. 1991). YY1 can activate gene expression through recruitment of histone acetyltransferase (HAT) proteins and can positively regulate numerous genes (Bushmeyer et al. 1995; Galvin and Shi 1997). Moreover, YY1 can mediate long-distance enhancer–promoter DNA interactions, perhaps through self-dimerization (Beagan et al. 2017; Weintraub et al. 2017), and is required for Ig locus contraction needed for immunoglobulin gene rearrangement and enhancer interactions for Ig class switch recombination (Ebert et al. 2011; Medvedovic et al. 2013; Mehra et al. 2016). Alternatively, YY1 can repress transcription of many genes by transient repression mechanisms involving recruitment of histone deacetyltransferase (HDAC) proteins (Galvin and Shi 1997). In addition, our laboratory found that YY1 can mediate Polycomb group (PcG)-dependent repression that stably represses genes during development (Atchison et al. 2003; Wilkinson et al. 2006; Basu et al. 2010).

YY1 knockout early in the B lymphocyte lineage by action of Mb1-driven CRE results in developmental arrest at the pro-B-cell stage (Liu et al. 2007). These arrested pro-B cells generate functional immunoglobulin heavy chain (IgH) gene V(D)J rearrangements that are skewed proximally within the IgH locus (Liu et al. 2007). Kleiman et al. (2016) found that deletion of YY1 at all stages of B-cell development by action of CD19-CRE results in a negative impact at essentially all stages of B-cell development, and we showed that YY1 knockout driven by γ1-CRE activated in germinal center B cells results in complete loss of these cells and the absence of serum IgG1 (Banerjee et al. 2016). In addition to B cell lineage development, YY1 plays important roles in a multiplicity of developmental programs, including T cells, intestinal stem cells, iNKT cells, cardiac and striated muscle, myeloid cells, mesoderm cells, erythroid cells, and extended pluripotent stem cells (Walowitz et al. 1998; Erkeland et al. 2003; Sucharov et al. 2003; Blättler et al. 2012; Perekatt et al. 2014; Ou et al. 2019; Perreault et al. 2020; Dong et al. 2022).

While analyzing transcripts in wild-type pro-B cells compared with YY1 knockout pro-B cells, we observed reduced expression of several B lineage genes coincident with increased expression of some alternative hematopoietic lineage genes. This prompted us to explore whether YY1 knockout pro-B cells might be capable of differentiation into alternative lineages. We show here that YY1 knockout pro-B cells acquire the capacity to develop into T lineage cells both in vitro and in vivo. These cells gain a T-like transcript profile while losing the B lineage transcript profile. Forced expression of YY1 ablates this lineage plasticity. Strikingly, single-cell RNA-seq experiments indicate that during transition from the pro-B to T cell lineage, most cells transiently acquire transcript profiles representative of a multiplicity of hematopoietic lineages, and we show that YY1 KO pro-B cells can develop into T cells, monocytes, and dendritic cells in vivo. ChIP-seq data from pro-B cells indicate that YY1 directly binds to numerous chromatin-modifying genes. Single-cell ATAC-seq (scATAC-seq) experiments from wild-type and YY1 knockout pro-B cells from mice revealed that YY1 deletion reduces accessibility of some B lineage genes while simultaneously increasing accessibility at numerous alternative lineage genes, providing a model for lineage plasticity induced by knockout of YY1 in pro-B cells.

Results

YY1-null pro-B cells show lineage plasticity in vitro and can develop into T lineage cells

Evaluation of published RNA-seq data (Kleiman et al. 2016) revealed a number of B lineage-specific genes, including Igll1, Vpreb1, Vpreb2, and CD79b, that showed reduced expression when comparing yy1^f/f and yy1^f/f Mb1-CRE (YY1-null) pro-B cells (Supplemental Fig. S1A, left). However, the expression levels of key B lineage regulatory factors such as Pax5, Ebf1, E2A, Foxo1, and Ikaros were unchanged (Supplemental Fig. S1A, right). Interestingly, expression of several hematopoietic lineage-specific genes was elevated (Supplemental Fig. S1B). Based on these unusual expression patterns, we questioned whether YY1-null pro-B cells, which should be committed to the B lineage pathway, exhibit lineage plasticity. To address this, we used a feeder cell line (OP9-DL4) that expresses the Notch ligand Delta-like 4 and promotes T-cell development in the presence of IL7, SCF, and Flt3L (Mohtashami et al. 2013).

Pro-B cells from bone marrow (BM) were purified by fluorescence-activated cell sorting (FACS) from wild-type (yy1^f/f) C57BL/6 mice as well as from mice in which the yy1 gene was deleted in pro-B cells by the action of Mb1CRE (yy1^f/f Mb1-CRE). Isolated pro-B cells were cultured on OP9-DL4 feeder cells in the presence of IL-7, Flt3L, and SCF (Fig. 1A; Mohtashami et al. 2010) and, after 3 weeks, were harvested and analyzed for cell surface expression of T lineage marker proteins CD25 and Thy1.2. As anticipated, yy1^f/f (wild-type) pro-B cells failed to express CD25 or Thy1.2 (Fig. 1B, left top and right panels), indicating B lineage commitment. However, >90% of yy1^f/f Mb1-CRE (YY1-null) pro-B cells efficiently expressed T lineage marker proteins CD25 and Thy1.2 (Fig. 1B, left bottom and right panels). We further evaluated YY1-null pro-B-cell-derived CD25⁺ Thy1.2⁺ cells by FACS with anti-CD44 and anti-CD25, which distinguish double-negative T lineage cells into four stages (DN1, DN2, DN3, and DN4). This evaluation showed that YY1-null pro-B cells predominantly developed in vitro into DN3 T lineage cells (Fig. 1C). Coincident with this lineage transition within the OP9 DL4 culture system, yy1^f/f cells remained spherical in shape, typical of pro-B cells, whereas the yy1^f/f Mb1CRE cells took on a nonspherical appearance (Fig. 1D). These cells contained IgH Vh7183-DJh rearrangements, confirming their pro-B-cell origin (Fig. 1E, left panel), as well as T-cell receptor (TCR) Vβ4 rearrangements consistent with their development into T-like cells (Fig. 1E, right panel).

Figure 1.

YY1 knockout pro-B cells can adopt T lineage properties in vitro. (A) Diagram of in vitro strategy for testing lineage plasticity of either wild-type (yy1^f/f) or YY1 knockout (yy1^f/f Mb1-CRE) pro-B cells. (B) Comparison of yy1^f/f and yy1^f/f Mb1-CRE pro-B cells grown on OP9-DL4 feeder cells in the presence of IL-7, Flt3L, and SCF for 3 weeks. (Left panel) FACS plots of yy1^f/f (top) compared with yy1^f/f Mb1-CRE (bottom) pro-B cells with antibodies that detect T lineage cells (Thy1.2 and CD25). (Right panel) Quantitation of cell numbers in each sample. (C) CD44 and CD25 antibody FACS profile of T lineage DN stages of yy1^f/f Mb1-CRE pro-B cells grown on OP9-DL4 feeders for 3 weeks. (D) Morphological changes in cells after incubation on OP9-DL4 feeders for 18 days. Phase contrast images of either yy1^f/f (top) or yy1^f/f Mb1-CRE (bottom) pro-B cells incubated on OP9-DL4 feeders were captured using the EVOS cell imaging system by Life Technologies (Thermo Fisher Scientific). Scale bar, 100 μm. (E) YY1 knockout pro-B cells grown on OP9-DL4 feeders for 3 weeks contain both IgH and TCRβ somatic rearrangements. (Left panel) V(D)J rearrangement assays for VH7183 rearrangements with LSK-negative control and wild-type pro-B-positive control. (Right panel) TCRβ V4 rearrangements with pro-B-negative control, CD4/CD8 T-cell-positive control, and three samples of yy1^f/f Mb1-CRE pro-B cells grown on OP9-DL4 feeder cells in the presence of IL-7, Flt3L, and SCF for 3 weeks.

Ectopic expression of YY1 in YY1-null pro-B cells ablates their T lineage potential

To prove that loss of YY1 expression in pro-B cells was the cause of their acquired lineage plasticity, we performed YY1 complementation experiments (Fig. 2A). Bone marrow (BM) cells isolated from yy1^f/f Mb1-CRE mice were transduced with either empty retroviral vector (MigR1) or vector expressing the YY1 cDNA (MigR1-YY1) (Fig. 2A). Transduced bone marrow cells were then injected into lethally irradiated C57BL/6 mice, and after 12 weeks the mice were evaluated. We anticipated that YY1-null pro-B cells expressing the MigR1 vector alone would fail to support B lineage development past the pro-B-cell stage, as we and others had previously demonstrated (Liu et al. 2007; Pan et al. 2013), but would be capable of T lineage development in vitro on OP9-DL4 feeders due to the absence of YY1. Alternatively, YY1-null pro-B cells transduced with MigR1-YY1 would support B lineage development past the pro-B-cell stage but would fail to develop into T lineage cells on DL4-OP9 feeders. As expected, in mice receiving MigR1 vector control, B lineage development was largely arrested at the pro-B-cell stage due to the absence of YY1 (Fig. 2B, top panel). B lineage cells failed to progress efficiently past the pro-B-cell stage, yielding very few pre-B cells, immature B cells, or recirculating B cells (Fig. 2B, top panel). However, YY1-null pro-B cells isolated from the BM of these mice were capable of generating CD25⁺ Thy1.2⁺ T lineage cells on OP9-DL4 feeders, consistent with their lack of YY1 (Fig. 2B, bottom panel). In contrast, mice injected with cells transduced with MigR1-YY1 fully recovered B lineage development due to the presence of retrovirally expressed YY1, efficiently generating pre-B, immature B, and recirculating B cells (Fig. 2C, top panel). Moreover, pro-B cells isolated from these mice were incapable of generating cells with T lineage surface markers CD25⁺ Thy1.2⁺ on OP9-DL4 feeders (Fig. 2C, bottom panel). Thus, we conclude that the absence of YY1 is the regulating factor enabling lineage plasticity of normally committed pro-B cells.

Figure 2.

YY1 complementation ablates lineage plasticity and restores B lineage commitment. (A) Diagram of the experimental strategy. (B) Empty retroviral vector MigR1 does not rescue B lineage arrest caused by YY1 knockout and allows T lineage plasticity of yy1 knockout pro-B cells. Bone marrow cells from yy1^f/f Mb1-CRE mice were isolated, transduced with MigR1 empty vector, and injected into lethally irradiated mice. Twelve weeks later, GFP⁺ pro-B cells were isolated and placed on OP9-DL4 feeder cells for 3 weeks. (Top panel) FACS analyses show development of pro-B cells but not of pre-B or recirculating B lineage cells. (Bottom panel) These cells generated T lineage cells on OP9-DL4 feeders as assessed by FACS with anti-Thy1.2, anti-CD25, and anti-CD44 antibodies, yielding patterns similar to DN2 and DN3 cells. (C) Retroviral provision of YY1 restores B lineage development but ablates lineage plasticity to the T-cell lineage. Bone marrow cells from yy1^f/f Mb1-CRE mice were isolated, transduced with YY1-expressing vector MigR1-YY1, and injected into lethally irradiated mice. Twelve weeks later, GFP⁺ pro-B cells were isolated and placed on OP9-DL4 feeder cells for 3 weeks. YY1 expression completely rescued B lineage development by formation of pro-B, pre-B, immature B, and recirculating B lineage cells in vivo (top panel), but these cells failed to generate T lineage cells on OP9-DL4 feeder cells (bottom panel).

YY1-null pro-B-cell-derived T-like cells exhibit transcript profiles resembling wild-type thymic T cells

To evaluate the similarity of the T-like cells developed from YY1-null pro-B cells to wild-type thymic T cells isolated from mice, we performed RNA-seq studies with RNA prepared from five independent experiments. RNA-seq profiles of YY1-null pro-B cells developed into T-like cells (T) and control wild-type thymic T cells (C) and were evaluated and compared with already available RNA-seq profiles of wild-type pro-B cells (B) and YY1-null pro-B cells (Y) (Kleiman et al. 2016). Principal component analyses (PCAs) of all samples indicated that the YY1-null pro-B-derived T-like lineage cells were very distinct from wild-type and YY1-null pro-B cells and clustered adjacent to wild-type thymocytes (Fig. 3A, left panel). Transcript correlation distances by cluster dendrogram also showed that YY1-null-derived T-like cells were much more similar to T cells than to pro-B cells (Fig. 3A, right panel).

Figure 3.

YY1 knockout pro-B cells grown on OP9-DL4 feeders for 3 weeks gain a T lineage transcript profile and ablate their B lineage profile. (A) Three-dimensional principal component analyses (PCAs) and dendograms of the RNA-seq data from control thymic T cells (C1–C5), T cells developed from YY1-null pro-B cells (T1–T5), normal pro-B cells (B1–B3), and YY1-null pro-B cells (Y1–Y3). (B) Three-dimensional PCAs and dendograms of RNA-seq data from 15 select T-cell and 11 select B-cell genes. (C) Gain of T lineage RNA-seq profiles of yy1^f/f Mb1-CRE pro-B cells grown on OP9-DL4 feeders for 3 weeks. RNA-seq expression levels are shown for select T lineage genes using RNA isolated from YY1 knockout pro-B cells (Y), control pro-B cells (B), YY1 knockout pro-B cells grown on OP9-DL4 feeders for 3 weeks (T), and control thymic T cells (C). (D) Loss of B lineage gene expression in yy1^f/f Mb1-CRE pro-B cells grown on OP9-DL4 feeders for 3 weeks. Gene expression profiles are shown for select B lineage genes using the same RNA-seq samples as in C. (E) Heat map expression profiles of the data in C and D. (F) Overlap in gene expression profiles in each sample. (Left) Control thymic T-cell transcripts show extensive overlap in RNA-seq profile (89.1%) with yy1^f/f Mb1-CRE pro-B cells grown on OP9-DL4 feeders (T). (Right) In contrast, yy1^f/f Mb1-CRE pro-B cells grown on OP9-DL4 feeders (T) show expression of only 9.2% of genes expressed in their original YY1-null pro-B-cell identify (Y).

We evaluated a set of 15 genes representative of the T-cell lineage and 11 genes of the B-cell lineage to determine whether T-like cells developed from YY1-null pro-B cells gained a T lineage gene expression profile while simultaneously extinguishing the B lineage gene expression profile. PCAs and cluster dendrograms of expression patterns of the 15 T lineage and 11 B lineage genes indicated that YY1-null pro-B-cell-derived T-like cells clustered with wild-type thymic T lineage cells and were distinct from wild-type pro-B or YY1-null pro-B cells (Fig. 3B). Transcript profiles of 13 of the 15 key T lineage genes—Gata3, Bcl11b, Notch1, Tcf7, Id3, Il2ra, Cd28, Ccr9, Lck, Ly6d, Runx1, Runx3, and Erg—all showed significantly increased expression in T-like cells developed from YY1-null pro-B cells (T) and closely matched the expression pattern in control wild-type thymic T cells (Fig. 3C, C and T data samples). These genes were nearly silent in wild-type pro-B cells (B) as well as YY1 knockout pro-B cells (Y) (Fig. 3C, B and Y samples). Only expression of Cd44 and Tbx1 showed insignificant differences between wild-type T cells and YY1-null pro-B-cell-derived T-like cells compared with pro-B-cell samples (Fig. 3C). Similarly, we found that critical B lineage genes Ebf1, Pax5, Cd79a, Bcl11a, Cd79b, Igll1, Mef2c, Irf4, Irf8, Vpreb1, and Vpreb2 all significantly dropped in expression in T-like cells derived from YY1-null pro-B cells, with expression patterns resembling wild-type thymic T cells (Fig. 3D). Heat map representation of the 15 T lineage and 11 B lineage genes in each RNA-seq sample is shown in Figure 3E.

We also evaluated genes that were differentially expressed between our RNA-seq samples. A comparative analysis of data sets of YY1-null pro-B cells developed into T-like cells (T) versus wild-type thymic T cells (C) or YY1-null pro-B cells (Y) identified a total of 3116 and 2552 significant differentially expressed genes (DEGs), respectively, with log₂ fold change in the range of ≥+1.0 to ≤ −1.0. Out of 3116 DEGs, 692 were upregulated (log₂FC ≥ +1.0 with P-value ≤ 0.05) and 2424 were downregulated (log₂FC ≤ −1.0 with P-value ≤ 0.05) in YY1-null pro-B cells developed into T-like cells (T) versus wild-type thymic T cells (C). In wild-type pro-B (B) versus YY1-null pro-B cells (Y) data, 993 genes were upregulated (log₂FC ≥ +1.0 with P-value ≤ 0.05) and 1559 were downregulated (log₂FC ≤ −1.0 with P-value ≤ 0.05) out of 2552 DEGs.

Expression of 89.1% of differentially expressed genes (2775 genes) was shared between wild-type thymic T cells and YY1 knock-out pro-B cells developed into T-like cells (Fig. 3F, left panel). A very small fraction of genes (83 genes; 2.7%) was uniquely expressed in YY1-null pro-B-cell-derived T-like cells, and 257 genes (8.3%) were unique to control thymic T cells (Fig. 3F, left panel). Gene ontology (GO) analyses showed that the unique genes in control T lineage cells compared with YY1 knockout-derived T-like cells grouped into a variety of pathways that were not specific to T lineage cells (Supplemental Fig. S1C, left panel, red lettering). Similarly, the pathways uniquely expressed in YY1-null pro-B cells developed into T-like cells that showed mainly non-T lineage-specific pathways (Supplemental Fig. S1C, right panel, red lettering). In contrast, T-like cells developed from YY1-null pro-B cells showed extensive GO differences compared with YY1-null pro-B cells (Supplemental Fig. S1D, left panel). Only 9.2% of transcripts overlapped between these samples, whereas 90.8% of transcripts were unique (Fig. 3F, right panel). GO analyses showed that T-like cells developed from YY1-null pro-B cells expressed pathways enriched in T lineage cells (Supplemental Fig. S1D, left panel, blue lettering). In contrast, as expected, YY1-null pro-B cells expressed pathways indicative of their B-cell lineage (Supplemental Fig. S1D, right panel, green lettering). Intriguingly, some T lineage and other lineage pathways were also expressed in YY1-null pro-B cells (Supplemental Fig. S1C, right panel blue lettering). These differences are consistent with YY1 knockout pro-B cells perhaps being capable of generating other lineages in addition to T cells.

YY1-null pro-B cells can differentiate into T-like cells in vivo

Our studies using OP9-DL4 feeder cells clearly showed that YY1-null pro-B cells can develop into cells that closely resemble thymic T lineage cells. We questioned whether YY1-null pro-B cells could differentiate into T-like cells in vivo and perhaps past the DN T cell stages. To test this, using FACS, we isolated wild-type pro-B cells from yy1^f/f mice and YY1-null pro-B cells from yy1^f/f mb1-CRE mice and injected these cells into sublethally irradiated Rag1^−/− mice, which fail to develop either mature T lineage or B lineage cells (Fig. 4A). Any developing T-like cells would therefore be due to the injected pro-B cells. After 7 months, Rag1^−/− mice injected with wild-type pro-B cells failed to develop detectable thymuses (Fig. 4B, top panel). In contrast, three out of 15 Rag1^−/− mice injected with YY1-null pro-B cells developed large thymuses, with the remaining developing smaller rudimentary thymuses (Fig. 4B, bottom panel). Thymocytes and splenocytes isolated from these mice with the larger thymuses showed the acquisition of T-like double-positive (DP) and single-positive CD4 and CD8 cells, respectively (Fig. 4C), and their new T-like phenotype was confirmed in splenic T cells by the expression of TCRβ protein on the cell surface (Fig. 4D). Consistent with their pro-B-cell origin, these cells also possessed IgH VhQ52 rearrangements (Fig. 4E). In addition, these in vivo developed splenic T cells adopted T-cell functions as they secreted IFNγ and TNFα proteins in response to CD3/CD28 activation, similar to wild-type T lineage cells (Fig. 4F).

Figure 4.

YY1 knockout pro-B cells can develop into T-like cells in vivo. (A) Diagram showing experimental strategy. (B) Sublethally irradiated Rag1^−/− mice were injected with either control pro-B cells (yy1^f/f) or YY1 knockout pro-B cells (yy1^f/f Mb1-CRE). Seven months later, mice were sacrificed and assessed for T lineage development. Pictures are shown of mouse thymuses from control Rag1^−/− mice injected with either wild-type (yy1^f/f) pro-B cells (top panels) or YY1 knockout pro-B cells (yy1^f/f Mb1-CRE; bottom panels). (C) FACS plots of thymic DP, CD4, and CD8 T lineage development of two mice injected with yy1^f/f Mb1-CRE pro-B cells (left panel), and CD4⁺ CD8⁺ expression of splenic T cells of two mice injected with yy1^f/f Mb1-CRE pro-B cells (right panel). (D) Splenic T lineage cells developed in vivo from yy1^f/f Mb1-CRE pro-B cells express TCRβ. (E) Two of three samples of T lineage cells developed in vivo from yy1^f/f Mb1-CRE pro-B cells exhibit IgH gene rearrangements (samples 5 and 6). (F) Splenic T cells developed from YY1 KO pro-B cells injected into Rag1^−/− mice express cytokines upon CD3/CD28 stimulation, similar to control T cells.

Alternative hematopoietic lineage genes are activated during the transition from the pro-B-cell to the T-cell lineage

The transformation of YY1-null pro-B cells into T-like cells on OP9-DL4 feeders required ∼3 weeks to generate a population that consisted of >90% CD25⁺ Thy1⁺ cells. We reasoned that harvesting cells midway during this process and subjecting them to single-cell RNA-seq (scRNA-seq) would enable us to decipher key mechanistic details of this alternative lineage differentiation process. Therefore, we isolated pro-B cells from yy1^f/f and yy1^f/f mb1-CRE mice and cultured them on OP9-DL4 cells with IL7, Flt3L, and SCF until 3%–5% of the yy1^f/f mb1-CRE pro-B population had become CD25⁺ Thy1⁺ (as expected, cells from wild-type yy1^f/f mice remained negative). Cells cultured on OP9-DL4 feeders were harvested and subjected to scRNA-seq in parallel with pro-B cells isolated directly from yy1^f/f and yy1^f/f mb1-CRE mice.

Merged scRNA-seq data from duplicate samples of the four conditions (pro-B cells from yy1^f/f and yy1^f/f Mb1-CRE mice, plus the two pro-B samples on OP9-DL4 feeders) generated 22 Seurat UMAP clusters (Supplemental Fig. S5A). Separating the overlapping merged UMAPs into the individual sample UMAPs showed very similar patterns when comparing yy1^f/f and yy1^f/f Mb1CRE pro-B cells directly from mice, as well as yy1^f/f pro-B cells cultured on OP9-DL4 feeders (Fig. 5A, second through fourth panels). However, the UMAPs of the yy1^f/f Mb1-CRE pro-B cells cultured on OP9-DL4 feeders showed considerable differences, with reductions of cell numbers in clusters 0, 1, 2, 4, 14, and 21 but substantial gains in clusters 3, 7, 8, 13, 18, and 20 (Fig. 5A, first panel). We used the SingleR package in R Studio for cell annotation using the Immgen database to define the different cell types present in each cluster. This analysis annotated nearly all the cells in the yy1^f/f and yy1^f/f Mb1-CRE pro-B cells directly from mice, as well as the yy1^f/f samples grown on OP9-DL4 feeders as B lineage cells (Fig. 5B, second through fourth panels, pink–orange color). In contrast, nearly all the cells in the yy1^f/f Mb1-CRE sample grown on OP9-DL4 feeders had lost their B lineage phenotype and expressed transcripts specific for numerous distinct hematopoietic cell types, as indicated by multiple non-B-cell lineage colors. (Fig. 5B, first panel). This panel is expanded in Figure 5C, with cell types identified by the SingleR program overlaid on the figure. Cell types annotated in this sample included T, Tγδ, ILC, NKT, NK, dendritic, monocyte, basophil, macrophage, mast, neutrophil, and stem cell phenotypes (Fig. 5C). The 3%–5% of cells identified as T lineage when the population was harvested for scRNA-seq analysis are indicated by the arrow in Figure 5C.

Figure 5.

yy1^f/f Mb1-CRE pro-B cells grown on OP9-DL4 feeders for 2 weeks develop into cells expressing a multiplicity of hematopoietic lineage transcripts and extinguish their B lineage transcript profile. Pro-B cells from yy1^f/f or yy1^f/f Mb1CRE mice were either subjected to scRNA-seq immediately or grown on OP9-DL4 feeders for 2 weeks when 3%–5% of cells from the yy1^f/f Mb1-CRE sample developed a Thy1.2⁺ CD25⁺ phenotype. (A) Patterns of the 22 Seurat clusters in each sample. (B) Assignment of cell types in each sample using the SingleR program. A cell identity key is shown in the right panel. (C) Pattern of the clusters and cell types in the yy1^f/f Mb1-CRE sample grown on OP9-DL4 feeders. Cell types identified by SingleR are indicated in the figure. The arrow points to the location of cells defined as either T or Tγ/d lineage. (D) Cell type percentages in each sample. The percentage of each sample that represents B lineage cells is shown above each column. The percentage of cells in the yy1/^f/f Mb1-CRE sample grown on OP9-DL4 feeders that are identified as either stem cell, monocyte, macrophage, or dendritic cells is indicated, and cell identities are indicated in the color key at the right. (E) UMAP profiles are shown for B lineage genes Pax5, Ebf1, Cd19, Cd79a, Cplx2, Vpreb2, and Vpreb3 from pro-B cells isolated from yy1^f/f and yy1^f/f Mb1-CRE mice or the same cells grown on OP9-DL4 feeders for 2 weeks until 3%–5% of the cells from yy1^f/f Mb1-CRE mice were Thy1.2⁺ CD25⁺. Gene names are at the left, and the cell types are indicated above each column. (yy1^f/f Mb1-CRE Diff) YY1 knockout pro-B cells developed until 3%–5% of the pro-B cells developed a Thy1.2⁺ CD25⁺ phenotype, (yy1^f/f Diff) wild-type pro-B cells incubated on feeders for the same period of time (2 weeks), (yy1^f/f Mb1-CRE and yy1^f/f) pro-B cells directly isolated from mice and subjected to scRNA-seq. (F) Box plot charts of the genes listed in E for expression level in various B-, T-, dendritic (DC), macrophage (M), monocyte (Mo), granulocyte (GN), basophil (Baso), or eosinophil (Eo) cell types. The red arrow points to pro-B fraction B/C cells.

Percentages of various cell types in the entire population are shown in Figure 5D. The percentage of cells assigned by SingleR that were identified as B cells in the yy1^f/f and yy1^f/f Mb1-CRE cells from mice and the yy1^f/f cells on OP9-DL4 feeders ranged from 97% to 100% (Fig. 5D). In contrast, only 0.3% of the cells (48 cells out of >16,000) were annotated as B lineage cells in the yy1^f/f Mb1-CRE pro-B cells grown on OP9-DL4 feeders (Fig. 5D). Forty-five percent of the cells were identified as monocytes, 40% were identified as dendritic cells, 7% were identified as stem cells, 4% were identified as neutrophils, 1% were identified as macrophages, and 1% were identified as ILC cells (Fig. 5D).

The remarkable differences between the yy1^f/f Mb1CRE pro-B cells on OP9-DL4 feeders and the other three samples indicate that the B lineage transcript profile in these cells is extinguished, and genes expressed in alternative lineages are elevated. Therefore, we set out to evaluate expression of genes closely associated with specific hematopoietic lineages.

Extinguished B lineage expression patterns

We evaluated expression of key genes indicative of the B cell lineage. As expected, we found that pro-B cells directly isolated from yy1^f/f and yy1^f/f mb1CRE mice as well as control yy1^f/f pro-B cells grown on OP9-DL4 feeders highly expressed critical B lineage genes Pax5, Ebf1, Cd19, Cd79a, Cplx2, Vpreb2, and Vpreb3 throughout their UMAP profiles (Fig. 5E, UMAP, second through fourth panels). Quite strikingly, however, expression of each of these B lineage genes was completely extinguished in the yy1^f/f mb1-CRE pro-B cells grown on OP9-DL4 feeders (Fig. 5E, UMAP, first panel). Box plot graphs from the Immgen MyGeneSet program against various B lineage cells (pro-B fractions B and C, splenic B, follicular B, marginal zone B, germinal center centroblasts, germinal center centrocytes, and splenic plasma cells) indicated high correlation of these transcripts (Pax5, Ebf1, Cd19, Cd79a, Cplx2, VpreB2, and VpreB3) with the B lineage, with the highest expression in pro-B cells (Fig. 5F, red arrow). In contrast, there was very low correlation of these genes in numerous T lineage, dendritic, monocyte, granulocyte, basophil, and eosinophil cell types (Fig. 5F). The dramatic loss of B lineage gene transcripts in the yy1^f/f Mb1-CRE pro-B cells grown on OP9-DL4 feeders (Fig. 5E, first panel) is consistent with the SingleR cell type annotation program indicating that only 0.3% of cells retained a B lineage transcript profile (Fig. 5C,D).

Gain of T-cell gene expression

The OP9-DL4 in vitro differentiation system is designed to generate T cells from precursor cells (Mohtashami et al. 2013). In our scRNA-seq experiments, we halted the culture system when only a small fraction of cells (<5%) had generated a Thy1.2⁺ CD25⁺ surface phenotype. If incubated to completion (3 weeks), we showed that most cells developed to the DN3 T-like cells (Fig. 1C). Therefore, we set out to determine the identities of genes that are enriched in DN3 cells compared with pro-B cells and then evaluate expression of those genes in our scRNA-seq data. First, we used the Immgen MyGeneset program to compare traditional RNA-seq data from pro-B fraction B/C cells with DN3 cells. The top 25 genes enriched in DN3 cells compared with pro-B cells are shown in Supplemental Table S1. This comparison identified Cd3g, Trbc1, Bcl11b, Cd3d, Tcf7, and Trbc2 within the top 10 enriched genes, whereas Trdc and Trgv2 were in the top 25 enriched genes. Thy1 was the 47th most enriched gene. We evaluated expression of each of these genes in UMAPs of our four scRNA-seq samples (yy1^f/f Mb1CRE pro-B cells differentiated on OP9-DL4 feeders until 5% of cells were Thy1.2⁺ CD25⁺, yy1^f/f pro-B cells differentiated for the same time on OP9-DL4 feeders, and yy1^f/f Mb1-CRE pro-B cells and yy1^f/f pro-B cells both isolated directly from mice). High gene expression levels were observed for each of the key DN3 T lineage genes in the yy1^f/f Mb1-CRE differentiated sample within the small cluster of cells identified as either T cells or Tγδ cells by SingleR (Fig. 6A [red circle], B [red arrows], respectively). Scattered expression was observed in other cells throughout the UMAPs in the differentiated yy1^f/f Mb1-CRE sample in Figure 6B, consistent with nearly all cells ultimately adopting a T lineage phenotype if incubated longer (3 weeks). Either no expression or very low expression of DN3 T lineage genes was observed in each of the control scRNA-seq samples (undifferentiated yy1^f/f Mb1-CRE pro-B cells or the yy1^f/f pro-B cells either directly from mice or incubated on OP9-DL4 feeders) (Fig. 6B, second though fourth panels). We also evaluated expression of Cd3g, Trbc1, Bcl11b, Cd3d, Tcf7, Trdc, Trgv2, and Thy1 in violin plots of each scRNA-seq cell cluster from the differentiated yy1^f/f Mb1-CRE sample. Consistent with our data in Figure 6B, the highest-level expression was observed in cells from Seurat cluster 0, which SingleR identified as T cells (Fig. 6C). Immgen MyGeneSet box plots confirmed that expression of our DN3-enriched genes is highly specific for various T-cell subsets (Fig. 6D, red arrow).

Figure 6.

T lineage genes are expressed in the small cell cluster identified as T cells. (A) UMAP profile of yy1^f/f Mb1-CRE pro-B cells grown on OP9-DL4 feeders for 2 weeks until 3%–5% of cells were Thy1.2⁺ CD25⁺. The red circle identifies the small cluster of cells identified by SingleR as T lineage cells in the yy1^f/f Mb1-CRE sample. (B) UMAP gene expression profiles of key T-cell lineage genes that are listed at the right. The red arrows show the location of the small cell cluster identified by SingleR as T lineage cells. (C) Violin plots of the RNA expression level of each T lineage gene in each Seurat cluster. The dashed red arrow shows that high expression of T lineage genes matches the small cluster of cells in A, identified as T cells (red circle). (D) Box plot charts of the genes listed in B for expression level in various B-, T-, NK, dendritic (DC), macrophage (Mc), granulocyte (GN), monocyte (Mo), basophil (Baso), or eosinophil (Eo) cell types.

Gain of dendritic cell-like expression patterns

Forty percent of yy1^f/f Mb1-CRE pro-B cells grown on OP9-DL4 feeders were identified as dendritic cells in our scRNA-seq experiments (Fig. 5C,D). Using the Immgen MyGeneset program with the Immgen RNA-seq databases of DC4⁺ and DC8⁺ splenic dendritic cells compared with pro-B fraction B/C data, we identified Clec9a, Tlr11, Gm6377, Itgax, Slamf8, Gpr141b, Gpr35, Gpr34, Xcr1, Havcr2, Ifi205, Il1b, Cd83, Apol7c, and Cacnb3 genes as 15 of the top 25 genes highly enriched in DC4⁺ and DC8⁺ dendritic cells (Supplemental Table S1). SingleR identified Seurat clusters 9 and 10 in our yy1^f/f Mb1-CRE differentiated sample as dendritic cells (see red ovals in Fig. 7A). Indeed, dendritic genes Itgax, Apol7c, Cacnb3, Clec9a, Tlr11, Gpr141b, and Xcr1 were highly expressed in these clusters in yy1^f/f Mb1-CRE pro-B cells differentiated on OP9-DL4 feeders for 14 days (Fig. 7B, first panel, red arrows). Essentially no expression was observed in the control panels (Fig. 7B, second through fourth panels). The other DC4⁺ and DC8⁺ dendritic genes (Gpr35, Gpr34, Gm6377, Itgax, Slamf8, Havcr2, Ifi205, Il1b, and Cd83) were also highly expressed in Seurat clusters 9 and 10 (Supplemental Fig. S2B, red oval). These genes were also expressed in the region identified as monocytic cells (clusters 3 and 15), consistent with the ability of monocytes to develop into dendritic cells (Supplemental Fig. S2B,C, blue circle and arrows, respectively). Although the dendritic genes tested were highly expressed in YY1 knockout pro-B cells on OP9-DL4 feeders, they were silent or poorly expressed in the three control cell types in Figure 7B and Supplemental Figure S2C (second through fourth panels), indicating that YY1 is critical for repressing these dendritic genes in the OP9-DL4 differentiation system.

Figure 7.

Elevated DC4⁺ and DC8⁺ dendritic lineage gene expression in clusters identified by SingleR as dendritic. (A) UMAP profile of yy1^f/f Mb1-CRE pro-B cells grown on OP9-DL4 feeders for 2 weeks until 3%–5% of cells were Thy1.2⁺ CD25⁺. Cell clusters 9 and 10, identified by SingleR as dendritic cells, are indicated by red ovals. Cluster 21, also expressing dendritic genes, is indicated by the blue oval. (B) UMAP profiles of gene expression for dendritic genes Itgax, Apol7c, Cacnb3, Clec9a, Tlr11, Gpr141b, and Xcr1 show high-level expression in clusters 9 and/or 10 (red arrows). Expression of some dendritic genes in cluster 21 is shown by the blue arrows. (C) Violin plots of RNA expression show that expression of the seven dendritic genes in B as well as that of eight other dendritic genes match the dendritic clusters 9, 10 (red dashed arrows), and 21 (blue dashed arrows). (D) Box plot charts of the genes in B for expression level in various B-, T-, NK, dendritic (DC), macrophage (Mc), granulocyte (GN), monocyte (Mo), basophil (Baso), or eosinophil (Eo) cell types show strong specificity for DC4⁺ and DC8⁺ dendritic cells (red arrow).

Violin plots confirmed expression of the 15 DC4⁺ or DC8⁺ dendritic genes in clusters 9 and 10 (Fig. 7C). Some genes (Xcr1, Vavcr2, Ifi205, Il1b, and Cd83) were also expressed highly in cluster 21 directly adjacent to dendritic clusters 9 and 10 (Fig. 7A [blue circle], C [dashed blue arrow]). Immgen MyGeneSet box plots confirmed that expression of these genes is specific for DC4⁺ and DC8⁺ dendritic cells (Fig. 7D, red arrow). Thus, YY1 knockout pro-B cells on OP9-DL4 feeders yields a transient phenotype consistent with dendritic cells.

We also evaluated transcripts representative of the DCpDC dendritic cell type. Immgen MyGeneSet analyses of RNA-seq data from DCpDC dendritic cells compared with pro-B fraction B/C cells identified Ccr9, Siglech, Pacsin1, Klk1, Gm5547, Press30, Lag3, Pltp, and Runx2 as within the top 25 DCpDC-expressed genes compared with pro-B cells (Supplemental Table S1). Expression of these genes was highly elevated in cluster 18 in yy1^f/f Mb1-CRE samples grown on OP9-DL4 feeders, with very low-level expression in each of the control scRNA-seq samples (Supplemental Fig. S3A [red circle], B [first panel, red arrows]). In some cases (Pltp, Prss30, and Gm5547), high-level expression was also observed in cluster 3 (monocytic cells) (Supplemental Fig. S3B, first panel, blue arrows), consistent with monocytes giving rise to dendritic cells. Violin plots confirmed elevated expression of DCpDC genes in cluster 18 (Supplemental Fig. S3C, red dashed line), and the specificity of these genes for DCpDC cells is also confirmed by box plots (Supplemental Fig. S5D, red arrow).

Gain of monocyte cell-like expression patterns

SingleR identified the largest fraction of cells in the yy1^f/f Mb1-CRE pro-B cells grown on OP9-DL4 feeders as monocytes residing in clusters 3 and 15 (45% of cells) (Fig. 5C,D). To confirm this phenotype, we used the Immgen public database of transcripts expressed in monocyte populations (monocytes 6C⁻ II⁻ BI and 6C⁺ II⁻ BI) compared with those in pro-B cells (fraction B/C). We identified genes Klra2, Arhgef37, Slfn1, Cd300ld, Ptpro, Gda, Gm9733, Ccl9, and Clec4a3 as being within the top 25 genes expressed in monocytes relative to pro-B cells (Supplemental Table S1). These genes showed high transcript expression in clusters 3 and 15 (Fig. 8A [red circles], B [first panel, red arrows]). These genes were poorly expressed in the control scRNA-seq UMAPs (Fig. 8B, second through fourth panels). Violin plots also showed very high expression of these genes in clusters 3 and 15 (Fig. 8C, red dashed arrows). Immgen MyGeneSet box plots confirmed the strong correlation of Klra2, Arhgef37, Slfn1, Cd300ld, Ptpro, Gda, Gm9733, Ccl9, and Clec4a3 genes with the monocyte lineage (Fig. 8D, red arrow).

Figure 8.

Elevated expression of monocyte genes in yy1^f/f Mb1-CRE cells on OP9-DL4 feeders. (A) UMAP profile of yy1^f/f Mb1-CRE pro-B cells from OP9-DL4 feeders grown for 2 weeks until 3%–5% of the cells were Thy1.2⁺ Cd25⁺. The positions of clusters identified as monocytic cells (clusters 3 and 15) are indicated by the red oval and circle. (B) UMAP profiles of gene expression of Klra2, Arhgef37, Slfn1, Cd300ld, Ptpro, Gda, Gm9733, Ccl9, and Clec4a3 show high-level expression in clusters 3 and 15 (red arrows). (C) Violin plots also show high expression of these genes in clusters 3 and 15 (red dashed arrows). (D) Immgen box blots show that the genes presented in B and C are enriched for expression in monocytes (red arrow).

To assess functional activity, yy1^f/f and yy1^f/f Mb1-CRE pro-B cells cultured on OP9-DL4 cells for 13 days were assayed for secretion of proinflammatory cytokines CXCL1 and IL6 (Chomarat et al. 2000; De Filippo et al. 2013). Indeed, cells derived from yy1^f/f Mb1-CRE mice secreted 100 times more CXCL1 and 10 times more IL6 compared with yy1^f/f pro-B cells (Supplemental Fig. S4A). These data show that the cells identified as monocytes in Figure 6C express genes highly enriched in monocytes and also secrete cytokines that are involved in upregulating inflammatory responses, indicating functional activity.

In addition to the T-cell lineage, Notch signaling is known to impact monocyte, dendritic, and macrophage development (Gamrekelashvili et al. 2020; Makino et al. 2022; Ren et al. 2022; Kapanadze et al. 2023). To determine whether Notch signaling is required for the lineage plasticity that we observed here, we incubated either wild-type or YY1-null pro-B cells on either OP9 feeders or their Notch signaling derivative, OP9-DL4, and measured expression of the monocyte-specific surface marker Ly6C. Only YY1-null pro-B cells incubated on OP9-DL4 feeders expressed Ly6C, indicating that the conversion of pro-B cells to myeloid-like cells occurs only in the presence of Notch signals (Supplemental Fig. S4B).

Gain of precursor gene expression

Intriguingly, 7% of yy1^f/f Mb1-CRE pro-B cells on OP9-DL4 feeders were identified as stem cells (Fig. 5C,D), although the assignment did not appear to correlate with any specific cell cluster. We assayed for expression of the Cd34 gene, which is highly expressed in hematopoietic precursors. We found elevated expression of Cd34 throughout the UMAP expression profile in the yy1^f/f Mb1-CRE pro-B cells grown on OP9-DL4 feeders (Supplemental Figure S4C, first panel) and very low expression within the three control samples (Supplemental Figure S4C, second through fourth panels). Elevated expression of this gene suggests a possible dedifferentiation pathway of YY1-null pro-B cells into hematopoietic precursors.

In summary, our scRNA-seq results show that YY1 knockout pro-B cells grown on OP9-DL4 feeders dramatically lose expression of transcripts representative of the B-cell lineage profile and simultaneously gain clusters of cells expressing numerous key genes indicative of various alternative hematopoietic lineages. These results indicate a loss of B-cell lineage commitment as well as development of an unusual plasticity of gene expression caused by loss of transcription factor YY1. Given sufficient time (3 weeks) and under defined conditions (OP9-DL4 feeders with IL7, SCF, and Flt3L), these cells adopt a phenotype highly consistent with T lineage cells.

Pro-B cells in yy1^f/f Mb1-CRE mice exhibit lineage plasticity in vivo, giving rise to monocytes and dendritic cells

The OP9-DL4 feeder system clearly showed that YY1-null pro-B cells exhibit lineage plasticity in vitro. If similar plasticity occurs in vivo, hematopoietic lineages in yy1^f/f Mb1-CRE mice that are distinct from the B-cell lineage should contain cells that have V(D)J rearrangements as well as deletion of the yy1 gene by action of the pro-B-cell-specific Mb1-CRE gene. To test this, we FACS-isolated monocytes from bone marrow and dendritic cells from the spleens of yy1^f/f and yy1^f/f Mb1-CRE mice and assayed for IgH V(D)J rearrangements and deletion of the yy1 gene. To detect the status of the yy1 gene, primers 1 and 2 detected the upstream loxp sequence, primers 1 and 4 detected the deleted yy1 allele, and primers 3 and 4 served as a loading control, detecting both WT and deleted alleles (Fig. 9C). As expected, monocytes and dendritic cells from WT yy1^f/f pro-B cells failed to show V(D)J rearrangements (Vh7183 or VhQ52) and did not show deletion of the yy1^f/f alleles due to the absence of Mb1-CRE (Fig. 9A,B, lanes 2, 6). In contrast, we found that monocytes and dendritic cells from yy1^f/f Mb1-CRE mice contained V(D)J rearrangements (both Vh7183 and VhQ52 Vh gene families) (Fig. 9A lanes 3–5, 7–9, top and bottom panels). In addition, these cells contained the Mb1-CRE-mediated yy1 gene deletion, further confirming their pro-B-cell origin (Fig. 9B, lanes 3–5, 7–9, middle panel, primers 1 and 4). Primers 3 and 4 (Fig. 9B, bottom panel) indicate equal loading of DNA in each reaction. Thus, we conclude that YY1 KO pro-B cells in yy1^f/f Mb1-CRE mice are capable of lineage plasticity in vivo.

Figure 9.

YY1 knockout pro-B cells develop into monocytes and dendritic cells in vivo, and YY1 regulates chromatin-modifying genes and chromatin accessibility at lineage-specific genes. (A–C) Monocytes and dendritic cells from yy1^f/f Mb1-CRE mice, but not yy1^f/f mice, contain V(D)J rearrangements and have deleted the yy1 gene. (A) PCR reactions show that DNA isolated from FACS-purified bone marrow monocytes and splenic dendritic cells contain Vh7183 and VhQ52 rearrangements in yy1^f/f Mb1-CRE mice but not yy1^f/f mice. (B) Monocyte and dendritic cells from yy1^f/f Mb1-CRE mice, but not yy1^f/f mice, show deletion of the yy1 gene (shown in the middle panel). (C) Map of the PCR primers amplifying the yy1 gene. (D) Gene ontology analyses of biological processes and KEGG pathway analyses were performed on genes that contain YY1 binding sites defined by ChIP-seq. Nearly all genes are involved in chromatin remodeling (green) or differentiation pathways (blue). (E) scATAC-seq-enriched transcription factor DNA binding motifs that are distinct in WT YY1 pro-B cells (left column) or in YY1 KO pro-B cells are shown in the right column. (F) scATAC-seq accessibility is reduced in YY1KO pro-B cells at some B lineage genes, enhanced at many alternative lineage genes, and reduced at some chromatin remodeling genes. Regions of change are indicated by boxes in the figure. The top panels (KO and orange) are YY1 KO (yy1ff Mb1-CRE) scATAC-seq data, and the bottom panels (WT and blue) are YY1 WT (yy1^f/f) scATAC-seq data. (G) Model of chromatin and gene expression changes that enable YY1-null pro-B cells to develop into alternative lineage cells.

YY1 regulates genes encoding general chromatin modifiers but not lineage-specific transcription factors

To gain mechanistic insight into how YY1 loss in pro-B cells might enable lineage plasticity, we evaluated the genes that directly bind YY1 in pro-B cells. We retrieved the raw YY1 ChIP-seq data set for pro-B cells from the NCBI-SRA (accession no. PRJNA175047) and subsequently analyzed the data. An average of ∼1337 YY1 ChIP-seq peaks in pro-B cells was identified over the input data sets, with an FDR value of <0.0001, considering all significantly enriched peaks. Furthermore, a genome-wide analysis of YY1 peak distribution relative to gene structure revealed that ∼90.20% of YY1 peaks were in the promoter-proximal region (within ≤1 kb from the TSS), followed by promoter-distal intergenic regions (1–2 kb from the TSS), and intronic regions.

The genes identified in the promoter-proximal region (≤1 kb from the TSS) were subsequently subjected to functional enrichment analysis. The biological process (BP) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses of genes associated with promoter-proximal region YY1 binding revealed a preponderance of pathways involving general transcriptional regulation (protein–DNA complex assembly, transcriptional initiation, chromosome organization, histone modifications, Polycomb repression, chromatin remodeling, etc.) (see Fig. 9D, green) and a number of lineage differentiation pathways (skeletal muscle, myeloid progenitor, T cell, osteoblast stem cell, hematopoietic stem cell, etc.) (see Fig. 9D, blue). Inspection of the individual genes in the identified gene regulatory pathways (Supplemental Table S2) revealed general chromatin-modifying and other transcriptional proteins but very few DNA site-specific and lineage-specific transcription factors such as Pax5, Ebf1, E2a, Gata-3, Bcl11c, etc. The transcription factor genes involved in B-cell, T-cell, monocyte, dendritic, HSC, ESC, or multiple hematopoietic lineages failed to bind YY1 (see Supplemental Table S3). The YY1 binding genes in the various lineage or stem cell-like pathways identified by gene ontology analyses did contain several general transcription factors (Sp1, Sp3, and ATF4) (Supplemental Table S2), but these factors do not regulate specific lineage development. Instead, the lineage-specific pathways contained a number of general chromatin-modifying transcriptional proteins and proteins regulating general cellular processes (Supplemental Table S2). Many of the chromatin-modifying genes bound by YY1 (Supplemental Table S2) showed very prominent YY1 peaks in their promoter regions. Nine of these genes are shown in Supplemental Fig. S5A. In addition, the majority of the identified chromatin remodeling, Polycomb, and lineage development genes showed reduced expression in YY1 KO pro-B cells compared with wild-type pro-B cells (Supplemental Fig. S5B). These data indicate that YY1 loss results in reduced expression of many chromatin-modifying genes that can either stimulate (histone acetyltransferases, chromatin remodeling proteins, etc.) or repress (HDACs, PcG proteins, etc.) gene expression, and many of the same chromatin remodeling proteins regulated by YY1 are involved in numerous cell lineage or precursor cell pathways (Supplemental Fig. S5B; Supplemental Table S2). Based on these results, we sought to further explore chromatin structures in WT and YY1 KO pro-B cells.

scATAC-seq analyses of WT and YY1 KO pro-B cells reveal differing accessibilities in transcription factor binding motifs as well as at alternative lineage genes

We performed single-cell ATAC-seq (scATAC-seq) on WT and YY1 KO pro-B cells to delve further into the mechanism of lineage plasticity enabled by YY1 KO. Merged data from WT and YY1 KO pro-B cells generated UMAPs with 24 clusters of cells (Supplemental Fig. S6A). The separate UMAP profiles of the WT and YY1 KO data sets revealed that 74.44% of the cells mapped to clusters that were highly similar between YY1 KO and WT samples (Supplemental Fig. S6B). However, 25.56% of cells between the two samples generated 15 distinct clusters (boxed regions in Supplemental Fig. S6B). These differing cell clusters were subclustered from the common cell clusters in Supplemental Figure S6, C and D. We compared transcription factor DNA binding motifs with increased accessibility, comparing the WT and YY1 KO distinct clusters, and identified various distinct transcription factor binding site accessibilities. The YY1 KO-specific transcription factor binding motifs included a predominance of JUN/FOS and REL/NFκB binding sites (Fig. 9E). Interestingly, AP1 factors (Jun and Fos family transcription factors) and AP1 binding sites have been implicated in the development of dendritic and myeloid lineages and in directing hematopoietic cell fate in multipotential progenitors (Zhao et al. 2022). As anticipated, the unique binding sites identified in YY1 KO pro-B cells clustered predominantly within the YY1 KO-specific UMAP clusters (Supplemental Fig. S6E). Conversely, the wild-type-specific binding sites mapped mainly to the WT pro-B-specific UMAP clusters (Supplemental Fig. S6F). Our results indicate that many YY1 KO pro-B cells exhibit accessibilities at transcription factor binding motifs (AP1, FOS, JUN, and REL/NFKB) that cluster distinctly from WT pro-B cells, and some of these transcription factor binding motifs are important in dendritic and monocyte lineages.

We next evaluated chromatin accessibility at B lineage genes as well as the alternative lineage genes that we showed in our scRNA-seq data were activated in YY1 KO pro-B cells on OP9-DL4 feeders (Figs. 5–8; Supplemental Figs. S2, S3). These results revealed some striking differences. Some B lineage-specific genes showed reduced scATAC-seq accessibility in YY1 KO-specific pro-B-cell clusters either within the gene or at their promoters (Cplx2 and VpreB3) (Fig. 9F, boxed regions). In contrast, consistent with the potential plasticity of YY1 KO pro-B cells, numerous alternative lineage genes showed increased accessibility in YY1 KO pro-B cells compared with WT pro-B cells (Fig. 9F, boxed regions). These genes included T lineage genes (Thy1 and Cd3g), monocyte genes (Ccl9, Cd300ld, and Klra2), and dendritic genes (Ccr5, Ccl5, Apol7c, Cacnb3, Havcr2, Il1b, and Pltp), all of which were activated in YY1 KO pro-B cells grown on OP9-DL4 feeders. Interestingly, the hematopoietic precursor gene Cd34 also showed greatly increased accessibility in YY1 KO pro-B cells (Fig. 9F, boxed region), suggesting possible precursor-like properties in YY1 KO pro-B cells. It should be noted that none of the genes activated in YY1-null pro-B cells grown on OP9-DL4 cells (Figs. 6–8; Supplemental Figs. S2, S3) contain YY1 binding sites (Supplemental Table S4) apart from Itgax (see Supplemental Fig. S8; Supplemental Table S4). On the other hand, consistent with reduced expression in YY1 KO pro-B cells, the chromatin modifier genes Tet1 and Daxx, which do contain YY1 binding sites in their promoters (Supplemental Tables S2, S3), showed reduced accessibility in YY1 KO pro-B cells compared with WT (Fig. 9F). Thus, our data show multiple gene accessibility differences in YY1 KO pro-B cells that provide insight into the mechanism of lineage plasticity enabled by YY1 KO.

Some deregulated genes show persistent expression after YY1-null pro-B cells develop into T-like cells

Our scRNA-seq data demonstrate dramatically increased expression of alternative hematopoietic lineage genes 2 weeks after YY1 knockout pro-B cells were incubated on OP9-DL4 feeders. Although most of these alternative lineage genes are ultimately repressed as the cells transition to the T-like cell lineage (Fig. 3), we reasoned that some alternative lineage genes might persist in expression, particularly if they are directly repressed by YY1. We evaluated expression of the top 25 genes (Supplemental Table S1) from dendritic cells (DC4⁺, DC8⁺, and DCpDC) and monocytes (6C⁺ 11⁻ BI and 6C⁻ 11⁻ BI) in our traditional RNA-seq data from wild-type thymic T cells, YY1-null pro-B cells differentiated into T-like cells, wild-type pro-B cells, and YY1-null pro-B cells. This comparison showed elevated expression in YY1-null pro-B cells developed into T-like cells of dendritic genes Itgb7, Runx2, I830077J02Rik, Havcr2, Ccl5, and Ccr5, as well as monocytic genes Ifi204, Emilin2, and Tmem51 (Supplemental Figs. S7, S8A, first panel). These genes were not highly expressed in control T cells or either of the pro-B-cell populations (Supplemental Figs. S7, S8A, second through fourth panels) but are highly activated during the transition from pro-B cells to T-like cells (Supplemental Figs. S7, S8A, first panel).

The highest-level expression was observed for the Itgb7 gene (Supplemental Figs. S7, S8A, first panel 1). YY1 binds to this gene in pro-B cells at an intronic location (Supplemental Fig. S8B, red oval) with properties of a potential DNA loop anchor as it also binds CTCF and Rad2. This DNA region also binds MED1, BRG1, IRF4, PU.1, and p300, implicating a transcriptional regulatory site (Supplemental Fig. S8B). Thus, YY1 may directly repress Itgb7 in pro-B cells, but as YY1-null pro-B cells begin to develop into T-like cells and the Itbg7 gene becomes activated (Supplemental Fig. S8A), the gene is ultimately not rerepressed in the developed T-like cells, perhaps due to the absence of YY1 (Supplemental Figs. S7, S8B).

Monocyte genes Emililn2, Runx2, and Tmem51 also show elevated expression during the transition from YY1-null pro-B cells into T-like cells (Supplemental Figs. S7, S8A, first panel). These genes exhibit Ezh2 binding at their promoters in pro-B cells, implicating a PcG mechanism of repression (Supplemental Fig. S8B, red ovals). The ability of YY1 to initiate PcG repression has been demonstrated previously (Atchison et al. 2003), implicating a potential YY1-dependent mechanism for the inability to rerepress these genes in the YY1-null pro-B cells developed into T-like cells. Overall, the elevated expression of alternative lineage genes revealed in our scRNA-seq data is matched in some cases by persistent expression of these genes after the cells develop from YY1-null pro-B cells into T-like cells. Regulation of some of these genes by YY1 may be direct (Itgb7 and chromatin-modifying genes, for instance) or indirect.

Discussion

Our results show that knockout of transcription factor YY1 in pro-B cells results in loss of B lineage commitment and gain of the plasticity to adopt the T-like cell lineage and perhaps other hematopoietic cell lineages such as monocytes, dendrocytes, and stem cells. We found that development of YY1 knockout pro-B cells to the T-like cell lineage can occur in vitro using the OP9-DL4 system, which promotes T lineage development, or in vivo by injection into sublethally irradiated Rag1^−/− mice. Moreover, we found that dendritic and monocyte cells from yy1^f/f Mb1-CRE mice contain V(D)J rearrangements and deletion of the yy1 gene, implicating YY1 KO pro-B-cell plasticity in vivo. The cells grown for 3 weeks in vitro possessed transcript profiles that closely align with the thymic T-cell lineage and nearly completely extinguish B lineage markers. This novel lineage plasticity is clearly the result of YY1 knockout, as provision of YY1 expression by retroviral transduction eliminated this lineage plasticity. Interestingly, scRNA-seq experiments to assess this differentiation process midtransition revealed cell clusters expressing a multiplicity of alternative hematopoietic lineage genes, whereas the B lineage transcript profile was nearly completely extinguished. Although B-cell lineage plasticity has been observed following knockout of several lineage-specific transcription factors (Nutt et al. 1999; Horcher et al. 2001; Ivanova et al. 2002; Mikkola et al. 2002; Cobaleda et al. 2007; Nechanitzky et al. 2013; Somasundaram et al. 2015), YY1 is unique in being a ubiquitous transcription factor expressed in all cell types, suggesting a potentially universal mechanism of lineage commitment.

B lineage development involves multiple steps of transcriptional priming to progress from multipotent progenitors to B lineage cells (Lin et al. 2010; Miyai et al. 2018). Initial priming and dynamic occupancy by EBF1 at B lymphoid enhancers, as well as functions by E2A and Foxo1, initiate the B lineage pathway (Lin et al. 2010; Lenaerts et al. 2022). Subsequently, Pax5 is required for continued B lineage commitment (Nutt et al. 1999). A variety of studies have shown that both EBF1 and Pax5 are critical for B lineage development and commitment, and their knockout can result in loss of B lineage identity and development into a variety of alternative hematopoietic lineages (Nutt et al. 1999; Horcher et al. 2001; Ivanova et al. 2002; Mikkola et al. 2002; Cobaleda et al. 2007; Nechanitzky et al. 2013; Somasundaram et al. 2015; Gruenbacher et al. 2023). Complete loss of Ebf1 and Pax5 expression in YY1-null cells on DL4 feeders indicates a critical role of YY1 in maintaining B-cell identity, at least partially due to an important role in maintaining Ebf1 and Pax5 expression.

A growing body of evidence indicates that lineage initiation and commitment require not only activation of lineage-specific genes but also repression of genes expressed in alternative lineages. For instance, early in B lineage development, EBF1 binds to the Gata3 promoter required for T lineage development and, in association with EZH2, represses Gata3 expression (Banerjee et al. 2013). Many alternative lineage genes and enhancers are repressed by EBF1 and Pax5 to maintain B lineage fidelity (Mikkola et al. 2002; Cobaleda et al. 2007; Nechanitzky et al. 2013; Lenaerts et al. 2022). Similarly, TCF1, a pioneer factor for T lineage development, not only initiates expression of T lineage genes (Yui and Rothenberg 2014; Johnson et al. 2018) but also represses expression of GATA2, needed for mast cell development (Goldman et al. 2023). Additionally, deletion of the repressive transcriptional cofactor histone deacetylase gene HDAC7 leads to lymphopenia and B lineage promiscuity (Azagra et al. 2016), consistent with the importance of transcriptional repression for lineage development. It is noteworthy that T lineage transcription factor Thy1 showed increased chromatin accessibility in YY1 KO pro-B cells (Fig. 9F).

YY1 is well known for its ability to both activate and repress transcription and derives its name from these dual functions (Park and Atchison 1991; Seto et al. 1991; Shi et al. 1991). Its activation properties map to the N-terminal regions, whereas repression functions map to histone deacetylase binding regions, as well as to a small 25 amino acid segment that supports Polycomb group repression (the REPO domain) (Bushmeyer et al. 1995; Austen et al. 1997; Galvin and Shi 1997; Thomas and Seto 1999; Satjin et al. 2001; Atchison et al. 2003; Srinivasan and Atchison 2004; Wilkinson et al. 2006; Basu et al. 2010). In addition, YY1 can self-associate, providing a mechanism to bridge promoters and enhancers via long-distance DNA interactions that have been observed in B-, T-, neural, erythroid, and stem cell systems (Hwang et al. 2013; Medvedovic et al. 2013; Mehra et al. 2016; Beagan et al. 2017; Dong et al. 2022). Indeed, YY1's function in long-distance DNA interactions is well documented in the B-cell lineage, where YY1 knockout leads to decontraction of Ig loci and reduced V(D)J rearrangement of distal Ig genes (Liu et al. 2007). Similarly, deletion of YY1 in splenic B cells results in reduced Ig class switch recombination caused by a decrease in DNA loop formation between the IgH Eμ and 3′RR enhancers (Mehra et al. 2016). As enhancer–promoter loops often increase at key genes during lineage commitment (Hu et al. 2018; van Schoonhoven et al. 2020), it is possible that reduction of YY1-mediated promoter–enhancer DNA loops upon YY1 knockout reduces stable expression of genes required for B-cell lineage commitment.

Our scRNA-seq data show that as YY1 knockout pro-B cells develop on OP9-DL4 feeders, they lose expression of B lineage transcripts, and these cells concomitantly exhibit an extraordinary increase in expression of genes characteristic of numerous alternative hematopoietic lineages. Rather than a gradual transition from a B-cell to a T-cell phenotype as might be anticipated, we observed a much more diverse expression of a multitude of genes indicative of many alternative hematopoietic lineages. The large fraction of cells identified as monocytic or dendritic (85%) compared with the small fraction identified as T cells (3%) at this early time point suggests that it is more facile to convert YY1-null pro-B cells into these hematopoietic lineages compared with T cells. Incubation of YY1-null pro-B cells on OP9-DL4 feeders for a more extended time (3 weeks) results in downregulation of most of these alternative lineage genes and increased expression of the T lineage genes.

The appearance of clusters of cells representative of many alternative lineages argues that YY1 is required to repress key genes that regulate numerous alternative hematopoietic lineages. The analyses of ChIP-seq data sets coupled with our scATAC-seq results provide mechanistic insight. Rather than binding to the promoters of tissue-specific transcription factor genes, YY1 binds to numerous more general chromatin-modifying genes, including both activators (HATs, chromatin modifiers, etc.) and repressors (HDACs, PcG, etc.). Thus, YY1 loss may lead to less stable transcriptional activation of lineage-appropriate genes due to lost recruitment of coactivators, as well as less stable repression of alternative lineage genes due to failure to recruit corepressors. This is bolstered by our scATAC-seq data, which show reduced accessibility at some B lineage genes and increased accessibility at numerous alternative lineage genes. In addition, increased accessibility at non-B lineage transcription factor binding motifs in YY1 KO pro-B cells suggests that these transcription factor binding sites may additionally participate in the development of alternative lineages in the absence of YY1. Finally, continued expression of some alternative lineage genes after YY1 knockout pro-B cells fully develop into T-like cells after 3 weeks on OP9-DL4 feeders (Supplemental Fig. S7) indicates that normal repression mechanisms are not fully functioning. The appearance of cells expressing a hematopoietic precursor gene also suggests that YY1 knockout pro-B cells may be capable of dedifferentiating to a precursor-like phenotype, thus enabling development of numerous alternative hematopoietic lineages.

The multiplicity of cell types representing distinct hematopoietic lineages in our scRNA-seq experiments suggests that YY1 repression of alternative lineage transcripts may be a common mechanism in regulating lineage-specific genes. This, coupled with the positive impact of YY1 on expression of lineage-appropriate genes, implies a mechanism of both YY1 transcriptional activation and transcriptional repression in lineage commitment (Fig. 9G). We propose that in wild-type pro-B cells, YY1 stably represses alternative lineage genes while simultaneously stably activating B lineage genes (Fig. 9G, left panel). In YY1 knockout pro-B cells, B lineage gene expression is no longer stably enforced, resulting in reduced expression of some B lineage genes, whereas alternative lineage genes are no longer stably repressed, resulting in leaky expression. Incubation of these cells in a T-cell-inducing environment (OP9-DL4 cells, for instance) thus leads to lost B lineage gene expression and gain of alternative linage gene expression (Fig. 9G, right panel). Additional studies will be required to more fully define the mechanism of lineage plasticity enabled by YY1 knockout in pro-B cells.

Materials and methods

Mice

C57BL/6 (CD45.1), Mb1CRE, C57BL/6 (CD45.2), and Rag1^−/− mice were purchased from the Jackson Laboratory. yy1^f/f mice on a C57BL/6 background were a gift from Y. Shi (Oxford University). yy1^f/f mice were crossed with Mb1-CRE mice to generate yy1^f/f Mb1-CRE mice, in which the endogenous yy1 gene is conditionally knocked out at the pro-B-cell stage by the action of Mb1-driven CRE recombinase. yy1^f/f Mb1-CRE mice and control mice between 7 and 9 weeks of age were used for all the experiments. All experiments involving animals were approved by the Institutional Animal Care and Use Committee of the University of Pennsylvania and conform to the appropriate regulatory standards.

Transplantation

For bone marrow transplantation, bone marrow cells were harvested from 8 week old yy1^f/f Mb1CRE mice 4 days after intravenous injection of 250 μg/kg 5-fluorouracil (5-FU). Cells were cultured overnight in DMEM with 10% FBS, 1% antibiotic, 1% L-glutamine, 10 ng/mL IL-3, 5 ng/mL IL-6, and 100 ng/mL SCF and then transduced with either empty retroviral vector (MigR1) or with vector expressing the YY1 cDNA (MigR1-YY1) by using 4 μg/mL polybrene and the same cytokine cocktail. At least 1 million cells were injected intravenously into lethally irradiated (9 Gy) recipient CD45.2 mice. Antibiotic-containing drinking water was provided for recipient mice for 2 weeks after transplantation. For pro-B-cell transplantation, 400,000–500,000 sorted pro-B cells were injected intravenously into sublethally irradiated (450 rad) Rag1^−/− recipient mice.

Recombinant vectors and virus packaging

FLAG-tagged YY1 cDNAs were cloned into the HpaI site of GFP-expressing MSCV-IRES-GFP vector (MigR1) by blunt end ligation. High-titer retroviral supernatants were prepared following transfection of HEK293T cells. 293T cells were maintained in DMEM supplemented with 10% FBS, 1% penicillin–streptomycin, and 2 mM L-glutamine. Retroviral supernatant was then used for spin infection at 2500 rpm for 90 min in the presence of 4 μg/mL polybrene. A second round of spin infection was performed 24 h following the first one.

In vitro T-cell differentiation

T-cell differentiation assays were performed using OP9-DL4 stromal cells as described previously (Mohtashami et al. 2013). Briefly, OP9-DL4 stromal cells were maintained in MEMα nucleosides containing 15% fetal bovine serum, 50 μM β-mercaptoethanol, and antibiotic. Prior to initiation of coculture, OP9-DL4 stromal cells were irradiated. Seven week old to 9 week old yy1^f/f or yy1^f/f Mb1-CRE mice were sacrificed, and bone marrow cells were collected. After red blood cell lysis, nucleated cells were blocked by Fc blocker, incubated with pro-B-cell marker antibodies (B220, CD19, CD93, IgM, CD43, and CD23), and sorted on a BD FACSAria cell sorter. Sorted pro-B cells (B220^low/⁺CD19⁺CD93⁺IgM⁻CD43⁺CD23⁻) from yy1^f/f or yy1^f/f Mb1-CRE mice were added to pre-established OP9-DL4 stromal layers and cultured in αMEM supplemented with 10% FBS, β-mercaptoethanol, and penicillin–streptomycin and with 5 ng/mL IL-7, 5 ng/mL Ft3L, and 10 ng/mL SCF, respectively. After 7 days, cells were replated on fresh irradiated OP9-DL4 cells. After approximately 3 weeks total in culture, nonadherent cells were harvested and analyzed by flow cytometry for the presence of Thy1.1 and CD25.

OP9 vs. OP9-DL4 differentiation

FACS-sorted pro-B cells from the bone marrow of yy1^f/f and yy1^f/f Mb1-CRE mice were seeded in 96 well plates on either OP9 or OP9-DL4 feeders for T-cell differentiation. Prior to initiation of coculture, OP9 and OP9-DL4 stromal cells were irradiated. These cells were cultured in αMEM supplemented with 10% FBS, β-mercaptoethanol, and penicillin–streptomycin and with 5 ng/mL IL-7, 5 ng/mL Ft3L, and 10 ng/mL SCF as mentioned above. At day 13, the cells were collected and stained for DAPI (BV421), CD19 (PECy5), B220 (APCCy7), Thy1.2 (BV785), CD25 (APC), F4/80 (PECy7), Ly6C (FITC), and MHCII (PE) to look for the presence of pro-B cells and differentiation to T and alternate lineage cells.

Flow cytometric analysis and cell sorting

Single-cell suspensions were stained and analyzed on an 18-color LSR Fortessa flow cytometer (Becton Dickinson) equipped with four lasers for excitation of Blue, red, green, and violet excited dyes. Antibodies were tagged with FITC, PE, PE-Cy5, PE-Cy7, APC or AF647, APC-Cy7, BV421, or BV605 versions of purified antibodies against CD3 (17A2), CD4 (RM4-5), CD8a (53-6.7), CD25 (PC61), CD44 (IM7), TCRβ (H57-597), Thy1.2 (30-H12), B220 (RA3-6B2), CD19 (6D5), CD43 (1B11), IgM (11/41), CD93 (AA4.1), CD23 (B3B4), c-kit (2B8), Sca-1 (D7), CD69 (H1.2F3), CD45.2 (104), and CD45.1 (A20). All directly conjugated antibodies were purchased from eBiosciences, BioLegend, BD Pharmingen, or Invitrogen. DAPI and Zombie Aqua were used to stain dead cells. All files were analyzed with FlowJo software (Tree Star, Inc.).

PCR detection of YY1 deletion efficiency

The deletion efficiency of the Yy1 gene was detected from sorted monocyte and dendritic cells from bone marrow and spleen, respectively, of yy1^f/f and yy1^f/fMb1-CRE mice as discussed elsewhere (Liu et al. 2007). Briefly, cell lysate from 1 × 10⁵ sorted cells was resuspended in 80 μL of 50 mM NaOH, incubated for 5 min at 95°C, and vortexed thoroughly to dissolve the cell pellets. Twenty microliters of 1 M Tris-HCl (pH 6.8) was then added to neutralize the NaOH. Five nanograms of the prepared DNA solution was then used to detect the deletion efficiency of the floxed yy1 allele in monocyte and dendritic cells by using the primers as described previously (Liu et al. 2007). Two percent agarose gels were run followed by ethidium-bromide staining to separate the PCR products and visualization. For primer sequences, see Table 1.

Table 1.

Key resources

Reagent or resource	Source	Identifier
PE/cyanine5 antimouse CD3ε antibody	BioLegend	100309; RRID: AB_312674
PE/cyanine5 antimouse CD4	BioLegend	130312, RRID: AB_2075572
PE/cyanine7 antimouse CD8a	BioLegend	100721; RRID: AB_312760
PE antimouse CD25 antibody	BioLegend	102008; RRID: AB_312857
APC/cyanine7 antimouse/human CD44 antibody	BioLegend	103028; RRID: AB_830785
APC antimouse TCRβ chain antibody	BioLegend	109211; RRID: AB_313434
Brilliant Violet 421 antimouse CD90.2 (Thy1.2) antibody	BioLegend	105341; RRID: AB_2632888
PE/cyanine5 antimouse/human CD45R/B220 antibody	BioLegend	103210; RRID: AB_312995
FITC antimouse CD19 antibody	BioLegend	115505; RRID: AB_313640
APC/Cyanine7 antimouse CD43 antibody	BioLegend	121220; RRID: AB_2194192
APC IgM antimouse monoclonal antibody (II/41) ebioscience	Thermo Fisher Scientific	17-5790-82; RRID: AB_469458
PE/cyanine7 antimouse CD93 (AA4.1, early B lineage) antibody	BioLegend	136506; RRID: AB_2044012
APC antimouse CD23 antibody	BioLegend	101619; RRID: AB_2563438
APC antimouse CD117 (c-kit) antibody	BioLegend	105812; RRID: AB_313221
APCCy7 antimouse Ly-6A/E (Sca-1) antibody	BioLegend	108126; RRID: AB_10645327
PE antimouse monoclonal CD69 antibody (H1.2F3), ebioscience	Thermo Fisher Scientific	12-0691-82; RRID: AB_465732
PE antimouse CD45.2 antibody	BD Pharmigen	560695; RRID: AB_1727493
PE antimouse CD45.1 antibody	BioLegend	110707; RRID: AB_313496
Antimouse CD45.1		A20
FITC antimouse Ly-6C	BioLegend	128005; RRID_AB_1134213
Chemicals, peptides, recombinant proteins	Source	Identifier
Recombinant murine IL-7	Peprotech	217-17
Recombinant murine SCF	Peprotech	250-03
Recombinant murine Flt3 ligand	Peprotech	250-31L
Hexadimethrine bromide (polybrene)	Sigma-Aldrich	H9268-5G
Fluorouracil injection, USP 500 mg (50 mg/mL)	APP Pharmaceuticals	63323-0117-10
Recombinant murine Interferon-γ	Peprotech	315-05
LPS from salmonella enterica serotype typhimurium	Sigma-Aldrich	L6143
Commercial assays	Source	Identifier
RNeasy micro kit	Qiagen	74004
SMART-Seq HT Plus kit	Takara Bio	R400748
10X Genomics	Chromium single-cell 3′ reagent kit v3.1	1000121
LEGENDplex MU macrophage/microglia panel	BioLegend	740846, lot B407587
ELISA Max deluxe set mouse TNF-α	BioLegend	430904, lot B306271
ELISA Max deluxe set mouse IFN- γ	BioLegend	430804, lot B307222
Untouched mouse T-cell kit	Dynabeads	11413D
Mouse T-activator CD3/CD28	Dynabeads	11452D
Deposited data	Source	Identifier
RNA-seq	This study	PRJNA903893
scRNA-seq	This study	PRJNA1073566
scATAC-seq	This study	PRJNA1149282
RNA-seq pro-B cells	GEO (Kleiman et al. 2016)	GSE73532
YY1 ChIP-seq	GEO (Kleiman et al. 2016)	GSM1002560
EzH2 ChIP-seq	GEO (Hill et al. 2020)	GSM4350099
CTCF ChIP-seq	GEO (Choi et al. 2013)	GSM1156665
H3K4me1 ChIP-seq	GEO (Lin et al. 2010)	GSM546527
Med1 ChIP-seq	GEO (Whyte et al. 2013)	GSM1038263
Brg1 ChIP-seq	GEO (Bossen et al. 2015)	GSM1635413
IRF4 ChIP-seq	GEO (Schwickert et al. 2014)	GSM1296534
PU.1 ChIP-seq	GEO (Mullen et al. 2011)	GSM539537
p300 ChIP-seq	GEO (Lin et al. 2012)	GSM987808
H3K4me3 ChIP-seq	GEO (Hill et al. 2020)	GSM4350110
Cell lines	Source	Identifier
HEK293T	ATCC	CRL-3216
OP9-DL4	David Allman, University of Pennsylvania	Mohtashami et al. 2013
OP9	David Allman, University of Pennsylvania	Nakano et al. 1994
Experimental models: organisms/strains	Source	Identifier
C57BL/6J (CD45.1; B6.SJL-Ptprca Pepcb/BoyJ)	The Jackson Laboratory	Strain 002014 RRID: IMSR_JAX:002014
C57BL/6J (for CD45.2)	The Jackson Laboratory	Strain 000664 RRID: IMSR_JAX:000664
Mb1-Cre on C57BL/6J [B6.C(Cg)-Cd79atm1(cre)Reth/EhobJ]	The Jackson Laboratory	Strain 020505 RRID: IMSR_JAX:020505
yy1f/f on C57BL/6J	Liu et al. 2007
yy1f/f Mb1-CRE on C57BL/6J	This study	N/A
Rag1 KO on C57BL/6J (C57BL/6J-Rag1em10Lutzy/J)	The Jackson Laboratory	Strain 034159 RRID: IMSR_JAX:034159
Oligonucleotides	Source	Identifier
YY1 KO genotype	For: ACCTGGTCTATCGAAAGGAAGCAC	Rev: CCAAAGTTCGAAACCTGCTTTCCT
VH 7183	For: CGGTACCAAGAASAMCCTGTWCCTGCAAATGASC	JH4E: AGGCTCTGAGATCCCTAGACAG
VH Q52	For: GCGAAGCTTCTCACAGAGCCTGTCCATCAC	JH4E: AGGCTCTGAGATCCCTAGACAG
Nested JH4A		JHFA: GGGTCTAGACTCTCAGCCGGCTCCCTCAGGG
TCR Vβ4	GGACAATCAGACTGCCTCAAGT	Jβ2: TGAGAGCTGTCTCCTACTATCGATT
TCR Vβ8	GATGACATCATCAGGTTTTGTC	Jβ2: TGAGAGCTGTCTCCTACTATCGATT
Jβ2.7 nested		GGAAGCGAGAGATGTGAATC
YY1p1/2	ACCTGGTCTATCGAAAGGAAGCAC	GCTTCGCCTATTCCTCGCTCATAA
yy1p1/4	ACCTGGTCTATCGAAAGGAAGCAC	CCAAAGTTCGAAACCTGCTTTCCT
yy1p3/4	TAGAGAATAGGAACTTCGGCCGCCA	CCAAAGTTCGAAACCTGCTTTCCT
Software and algorithms	Source	Identifier
FlowJo	Tree Star	RRID: SCR_008520
FastQC	Andrews 2010	RRID: SCR_014583
Trimmomatic	Bolger et al. 2014	RRID: SCR_011848
STAR	Dobin et al. 2013	RRID: SCR_004463
StringTie	Pertea et al. 2015	RRID: SCR_016323
DESeq2	Love et al. 2014	RRID: SCR_015687
clusterProfiler	Yu et al. 2012	RRID: SCR_016884
CellRanger	10X Genomics	RRID: SCR_023221
R 4.2.2	R statistical computing environment	RRID: SCR_000432
Seurat	Hao et al. 2021	RRID: SCR_016341
SingleR	Aran et al. 2019	N/A
Cowplot	https://github.com/wilkelab/cowplot/tree/master	RRID: SCR_018081
ggplot2	Villanueva and Chen 2019	RRID: SCR_014601
tensorflow	Abadi et al. 2016	RRID: SCR_016345
celldex	Aran et al. 2019	10.18129/B9.bioc.celldex
Fiji	Schindelin et al. 2012	RRID: SCR_002285
Bowtie2	Langmead and Salzberg 2012	RRID: SCR_016368
MACS3	Zhang et al. 2008	RRID: SCR_013291
Samtools v1.17	Li et al. 2009	RRID: SCR_002105
ChIPseeker	Yu et al. 2015	RRID: SCR_021322
Signac	Stuart et al. 2021	RRID: SCR_021158

Ig heavy chain V(D)J rearrangements analysis

CD25⁺ Thy1.2⁺ cells from in vitro differentiation cultures, CD4⁺ CD8⁺ cells from reconstituted thymocytes from Rag1^−/− mice, and monocytes and dendritic cells from bone marrow and spleen, respectively, from yy1^f/f and yy1^f/fMb1-CRE mice were isolated and sorted using a BD FACSAria cell sorter. DNA was prepared by using 50 mM NaOH and 1 M Tris-HCl (pH 6.8), and seminested PCR was performed for detecting Ig heavy chain V(D)J rearrangements as mentioned elsewhere (Nechanitzky et al. 2013). Amplification processes were carried out in two rounds: Round 1 contained 5′ primers VH7183 or VHQ52, and JH4E primer as reverse primer. First-round amplification was performed for 30 cycles (1 min at 95°C, 1 min at 60°C, and 1.5 min at 72°C). For the second-round amplification, 1 μL of the first-round product was used as a template, and the PCR consisted of 25 cycles (1 min at 95°C, 1 min at 63°C, and 1.5 min at 72°C) with 5′ primers being the same but JH4A being a nested reverse primer. All the primers have been used previously and published (Nechanitzky et al. 2013), and the primer sequences are listed in Table 1.

TCRβ chain V(D)J rearrangement analysis

Sorted CD25⁺, Thy1.2⁺ cells or CD4⁺ CD8⁺ cells were used to detect the TCRβ V(D)J rearrangements. DNA was prepared by using 50 mM NaOH and 1 M Tris-HCl (pH 6.8) as previously described. Nested PCR was performed in two rounds using the following primers: Vβ4 or Vβ8, and jβ2 and jβ2.7 nested (Table 1). First-round amplification was for 30 cycles (1 min at 95°C, 1 min at 60°C, and 1.5 min at 72°C), and second-round amplification was for 20 cycles (1 min at 95°C, 1 min at 63°C, and 1.5 min at 72°C).

Bulk RNA-seq library preparation

The cDNA of sorted CD25⁺, Thy1.2⁺ cells either from in vitro differentiation cultures or from wild-type mouse thymocytes were used to prepare sequencing libraries using the SMART-Seq HT Plus kit (R400748). Briefly, the quantification of total RNA was performed using a Qubit instrument, and the integrity of the RNA was confirmed using a TapeStation. Libraries were dual-indexed and pooled accordingly at equal molecular concentrations. Subsequently, 100 base pair reads were sequenced on an Illumina HiSeq 2000 platform.

Processing of bulk RNA-sequencing data

Fastq files of raw sequence data for control T DN cells (C) and T-like cells developed from YY1-null pro-B-cell (T) samples were generated using Illumina bcl2fastq software. The quality of fastq raw reads was checked using FastQC (Andrews 2010). Totals of 125,547,313 and 157,418,878 filtered reads were generated from YY1-null pro-B cells incubated for 3 weeks on OP9-DL4 feeders and wild-type thymic T cells, respectively. The average filtered rate percentage was 85.49% and 83.41%, and GC content was 43 and 44 in YY1-null pro-B cells incubated for 3 weeks on OP9-DL4 feeders and wild-type thymic T cells, respectively.

To eliminate low-quality reads, paired-end reads were filtered using the Trimmomatic 0.36 (Bolger et al. 2014) tool with average base quality of <20. After quality filtration, filtered reads were used for reference mapping using STAR (v.2.7.10a) (Dobin et al. 2013). The GRCm38 (mm10) mouse genome (http://useast.ensembl.org/Mus_musculus/Info/Index) was used for mapping. Raw reads for control pro-B-cell (B) and YY1-null pro-B-cell (Y) samples were retrieved from the National Center for Biotechnology Information-Sequence Read Archive (NCBI-SRA) with accession number PRJNA297235. The raw reads generated in the current study for control T DN cells (C) and YY1-null T DN cells (T) were submitted to NCBI-SRA with accession number PRJNA903893. Unsupervised clustering analysis was performed using factextra (https://cran.r-project.org/web/packages/factoextra/index.html), and heat maps were plotted using the gplots package in R.

Differential gene expression analysis

The differential expression analysis of YY1-null T DN cells and Y1-null pro-B cells, with respect to their control samples, was done by estimating the total read count of assembled transcripts. The total read count was retrieved using StringTie (Pertea et al. 2015). Furthermore, these count values were used for differential expression analysis using the DESeq2 (Love et al. 2014) package in R Studio. The gene annotation of differentially expressed genes (DEGs) was carried out with the GRCm38 (mm10) mouse genome (http://useast.ensembl.org/Mus_musculus/Info/Index). The threshold for significant differential gene expression was set as log₂ fold change (FC) of ≥1 and ≤−1, with P-value of ≤0.05 for upregulated and downregulated genes.

Gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis

To study the putative functions of the identified differentially expressed genes, gene ontology (GO) analysis was done that classified genes into cellular component (CC), molecular function (MF), and biological process (BP) categories. GO enrichment analysis was performed using org.Mm.eg.db keytype and the enrichplot package in RStudio with P-value cutoff of <0.05. To understand the enriched pathways of significant differentially expressed genes, KEGG analysis was performed. KEGG pathways analysis was performed using the clusterProfiler package (Yu et al. 2012) with the number of permutations (nPerm) set to 10,000 and org.Mm.eg.db and pAjustMethod set to “none.” P-value ≤ 0.05 was set for significant KEGG pathways.

Splenic T-cell activation and measurement of cytokine production

Splenic T cells developed in vivo from yy1^f/f Mb1CRE pro-B cells injected into Rag1^−/− mice as well as control T cells from WT mice were purified using the Dynabeads Untouched mouse T-cell kit (11413D). For the T-lymphocyte activation assay, Dynabeads mouse T-activator CD3/CD28 for activation of mouse T-cell kit (11452D) was used. Purified T cells (8 × 10⁴ cells) were plated in 200 µL of RPMI 1640 medium containing 10% heat-inactivated FBS and 100 mM glutamine in a 96-well tissue culture plate. Additionally, 2 µL of prewashed Dynabeads magnetic beads was added to achieve a bead to cell ratio of 1:1. The plate was then incubated in a humidified CO₂ incubator for 5 days at 37°C. Following the incubation period, supernatants were collected and analyzed for IFN-γ and TNF-α cytokines using the ELISA Max Deuxe set mouse TNF-α and IFN-γ kits (BioLegend 430904 and 430804, respectively) according to the manufacturer's instructions. Briefly, the plate was coated with 100 μL of diluted capture antibody (IFN-γ or TNF-α) and incubated overnight at 4°C. After four washes with washing buffer (PBS + 0.05% Tween-20), the plate was incubated with 200 μL of 1× assay diluent A (provided by the kit) for 1 h. Subsequently, 100 μL of diluted standards and samples was added and incubated for 2 h. Following this, each well was incubated with 100 μL of diluted detection antibody (provided by the kit) for 1 h, followed by the addition of 100 μL of Avidin-HRP, with an incubation period of 30 min at room temperature. After 30 min, each well was treated with 100 μL of TMB substrate solution and incubated for 15 min at room temperature in the dark. The reaction was stopped by adding 100 μL of stop solution, and the absorbance was read at 450 and 570 nm. All incubations were performed with shaking at room temperature, and the plate was washed with washing buffer four times after each incubation.

scRNA-seq library preparation

Biological replicates for scRNA-seq from mouse bone marrow cells were isolated from yy1^f/f and yy1^f/f Mb1CRE mice and FACS-sorted for pro-B cells as described. Cells either were subjected to scRNA-seq directly or were cultured on OP9-DL4 feeders as described above for 12–14 days when the yy1^f/f Mb1CRE cells had become 3%–5% Thy1.2⁺ CD25⁺. Next-generation sequencing libraries were prepared using the 10X Genomics Chromium single-cell 3′ reagent kit v3 according to the manufacturer's instructions. Libraries were uniquely indexed using the dual-index kit, pooled, and sequenced on an Illumina NovaSeq 6000 sequencer in a paired-end, dual-indexing run. Sequencing for each library targeted 20,000 mean reads per cell.

scRNA-seq data analysis

Raw data were processed using the Cell Ranger pipeline (10X Genomics, v.6.1.2) for demultiplexing with the mkfastq command to generate Fastq files and for alignment of sequencing reads to the mm10 reference genome and creation of feature–barcode matrices. Secondary data analyses were performed in R (v.4.2.2) with Seurat (v4.3.0). The unique molecular identified (UMI) count tables were initially loaded into R (v4.2.2) using the Read10X_h5 function, and then Seurat objects were created for each sample. The samples were merged, and during the normalization process in Seurat, SCTransform was used to regress out the effects of library size. In Seurat, the cells were clustered using a clustering algorithm based on shared nearest-neighbor modularity optimization, and dimensional reduction was performed with the first 30 PCs. Cells were annotated with Immgen data using the SingleR (v1.8.1) package, and marker gene expressions were shown with FeaturePlot and stacked violin plot using Seurat (v4.3.0), cowplot (v1.1.1), and ggplot2 (v3.4.0) packages in RStudio. The count of each cell type was retrieved using the tensorflow (2.14.0) and celldex (v1.4.0) packages.

Mapping and peak calling of ChIP-seq data

The raw reads of YY1 ChIP-seq data for pro-B cells (accession no. PRJNA175047) were retrieved from NCBI-SRA. Trimmomatic 0.36 (Bolger et al. 2014) was used to process adapter and low-quality reads. The cleaned reads were then aligned to the GRCm38 (mm10) reference mouse genome using Bowtie2 (Langmead and Salzberg 2012). Samtools v1.17 (Li et al. 2009) was used to filter out low-quality mappings and duplicates. MACS3 (Zhang et al. 2008) performed peak calling on paired-end mapped bam inputs with a q-value threshold of 0.0001. The Chipseeker package in R (Yu et al. 2015) annotated the enriched peaks for genomic regions. Genome-wide visualization of enriched peaks was conducted using the UCSC genome browser (https://www.genome.ucsc.edu).

scATAC-seq sample processing and library preparation

Pro-B cells were FACS-sorted from mouse bone marrow from the yy1^f/f (wild-type) and yy1^f/f Mb1CRE (YY1 KO pro-B cells) mice on a C57/BL6 background. Duplicate samples of pro-B cells from each genotype were processed. Cells were washed and lysed, and nuclei were counted and immediately processed for scATAC-seq, targeting 10,000 nuclei per sample, using the Chromium Next GEM single-cell ATAC kit v2 (10X Genomics 1000406) according to the manufacturer’s instructions. In brief, nuclei were stained with AO/PI, and their concentration was determined using a Cellometer K2 fluorescent automated cell counter (Nexcelom Bioscience). The nucleus concentrations were then adjusted and subjected to transposition. The transposed nuclei were loaded onto a Chromium Next GEM Chip H (10X Genomics 1000162), partitioned into gel beads in emulsion (GEMs), and barcoded using the 10X Genomics chromium controller. Libraries were constructed by sample indexing PCR and quality-checked on an Agilent Bioanalyzer high-sensitivity DNA chip.

scATAC-seq data analysis

Sequenced reads from scATAC-seq libraries were processed using the 10X Genomics Cell Ranger ATAC pipeline. Initially, raw sequencing reads were demultiplexed and converted into FASTQ files with the cellranger-atac mkfastq pipeline. Subsequently, the sequencing reads were processed through the cellranger-atac count pipeline, which involved read filtering, alignment to the reference genome using STAR, barcode counting, identification of transposase cut sites, and accessible chromatin peaks. This step also included cell calling and the generation of count matrices for chromatin peaks and transcription factors. The estimated number of cells captured per sample ranged from 7974 to 13,842, with 7938–20,650 median fragments per cell and 156,554–166,774 total peaks detected. A total of 101,015 cells was selected for further analysis using the R packages Seurat and Signac (https://satijalab.org/signac/; Stuart et al. 2021). To normalize sequencing depth differences across cells and give higher weights to rare peaks, a term frequency–inverse document frequency (TF-IDF) normalization was performed using the RunTFIDF function. For dimensionality reduction, singular value decomposition (SVD) was applied using the RunSVD function after selecting the top 95% of peaks with the FindTopFeatures function (object, minimun cutoff = “q10”). Nonlinear dimensionality reduction was conducted using UMAP (Diaz-Papkovich et al. 2019) through the RunUMAP function on the first 30 dimensions for visualization. Subsequently, a shared nearest neighbor (SNN) graph was constructed using the FindNeighbors function, cell clustering was performed with the FindClusters function at a resolution of 0.6, and cluster differential accessibility was analyzed using the FindMarkers function. To quantify chromatin accessibility at the gene level, fragments from each cell intersecting the gene and promoter regions (1000 bp upstream) were counted. Gene coordinates were obtained using the EnsDb.Mmusculus.v79 R package. For motif analysis, known transcription factor binding motifs from the JASPAR 2020 database were scanned within the DNA sequence of each peak. The FindMotifs function was then used to identify overrepresented motifs in specific clusters.

Multianalyte flow assay to measure cytokine production from differentiated yy1^f/f Mb1CRE pro-B cells

Cell culture supernatants from yy1^f/f or yy1^f/f Mb1CRE pro-B cells grown on OP9-DL4 cells were harvested at 13 days of culture and analyzed for cytokine production using the LEGENDplex MU macrophage/microglia panel kit (BioLegend 740846, lot B407587) according to the manufacturer's directions and using supernatant from LPS and IFN-γ (at 100 ng/mL for 12 h) activated macrophages from yy1^f/f bone marrow as a positive control. Briefly, standards were prepared by reconstituting and performing serial dilutions with assay buffer, and standard curve was generated. Assay buffer (25 µL) was added to the sample and standard wells in a V-bottom plate at a 1:1 ratio followed by the addition of 25 µL of mixed beads. The plate was then sealed, covered with aluminum foil, and incubated on a plate shaker at 800 rpm for 2 h at room temperature. After centrifugation at 250g for 5 min, the beads were washed with 200 µL of 1× wash buffer, and 25 µL of detection antibodies was added to each well. The plate was then incubated for 1 h at room temperature following the same condition as before, and 25 µL of streptavidin-phycoerythrin (SA-PE) beads was directly added to the solution without washing. After incubating the plate with continuous agitation of 800 rpm for 30 min at room temperature, the bead was washed with 200 µL of 1× wash buffer and finally resuspended in 150 µL of the same. The samples were then vortexed well, and 300 beads per analyte were acquired using a flow cytometer (BD LSRFortessa, BD Biosciences H79300008). The generated FCS files from the cytometer were analyzed using BioLegend's LEGENDplex data analysis software exactly following the steps from the user manual.

Competing interest statement

The authors declare no competing interests.