B4galt5^fl/+ mice were obtained from Masahide Asano (23). B4galt5^fl/+ mice were bred with Mb1Cre mice (The Jackson Laboratory, Stock 020505) to obtain Mb1Cre;B4galt5^fl/fl mice. Nsd1^fl/+ mice were purchased from Gempharmatech (T018412), bred with Mb1Cre mice to obtain Mb1Cre;Nsd1^fl/fl mice. IgM^b-macroself mice were generated by David Nemazee´s laboratory (The Scripps Research Institute, La Jolla, CA) as previously described (20, 24). All mouse strains were housed and bred in pathogen-free conditions.

The immature B cell line WEHI-231 was originally obtained from David Nemazee (The Scripps Research Institute, La Jolla, CA) and cultured in DMEM (Gibco, Cat:10569-044, containing GlutaMAX and Sodium Pyruvate) supplemented with 10% FBS, 1% Pen/Strep, and 50 μM 2-mercaptoethanol. The HEK293T cell line was originally obtained from Jun Lu (Yale University, New Haven, CT) and cultured in DMEM (Gibco, Cat:10569-044, containing GlutaMAX and Sodium Pyruvate) supplemented with 10% FBS and 1% Pen/Strep.

The TransOMIC-shRNA sublibrary was obtained from the RNAi Core at La Jolla Institute for Allergy and Immunology (La Jolla, CA), amplified from a glycerol stock, and DNA prepared using an endotoxin-free DNA purification kit (K210001, ThermoFisher). Retrovirus preparations were obtained as previously described (20). Briefly, packaging vectors GAG-POL, VSV-G and individual shRNA-encoding or empty plasmids were co-transfected into HEK293T cells using TransIT-LT1 (Mirus) following the manufacturer´s instructions. Supernatants containing retrovirus were collected after 48 and 72 hours, filtered through a 0.45μm low protein retention filter, aliquoted and stored at −80°C until use.

For WEHI-231 apoptosis analysis, the active Caspase 3 kit (C92-605, BD Pharmingen) was used following the manufacturer’s instructions. Briefly, cells were fixed and permeabilized with Fix/Perm Buffer for 30 min at 4°C, stained with an antibody for active Caspase 3 labelled with APC for 30 min at 4°C, and analyzed by flow cytometry. For splenic B cell analysis of IgM^b-macroself mice reconstituted with retrovirus-transduced HPSCs or bone marrow cells from donor mice, single-cell suspensions were obtained by disaggregating spleens through a 0.70μm cell strainer, depleted of red blood cells by incubation for 5 min RT with ACK lysis buffer, stained with CD45.1-PerCP/Cy5.5 (A20; Biolegend), CD45.2-PE (104; Biolegend), IgM-APC (RMM-1; Biolegend) and CD19 PE/Cy7 (1D3, eBioscience) for 20 min at 4°C, and analyzed by flow cytometry. Data were acquired on FACS Calibur, Novocyte or Fortessa X-20 flow cytometers, and analyzed with Flowjo Version10.7.1.

To generate stable WEHI-231 cell lines, 0.5x10⁶ cells were transduced with retroviruses encoding individual shRNAs in 24-well plates overnight. Media was replaced with fresh one the next morning. After 24 hours of culture, 10µg/ml puromycin was added to medium to select transduced cells for 5 days. After selection, stable cell lines were maintained in the presence of 2µg/ml puromycin in culture medium. Stable WEHI-231 cells were stimulated with 2µg/ml anti-IgM (115-006-020, µ chain specific, Jackson) for 72 hours, stained for intracellular active caspase-3, and analyzed by flow cytometry as described above.

For in vivo validation, plasmids encoding positive shRNA hits were modified to delete the puromycin resistance gene, amplified using an endotoxin-free DNA purification kit (12362, Qiagen), and packaged to generate retroviral supernatants as described above. For HPSC enrichment, wild type C57BL/6J mice were treated with 4 mg/mouse 5-fluorouracil (5-FU) (Sigma) by intravenous (iv) injection, followed by bone marrow cell collection 5 days later. HPSCs were stimulated with SCF (100ng/ml, Novoprotein), IL-3 (20ng/ml, Novoprotein) and IL-6 (25ng/ml, Novoprotein) for 36 hours before transduction. Retrovirus transduction of HPSCs was performed by spinoculation (2000rpm, 2 hours without brake) and media was replaced 4 hours later. Transduced cells were harvested after 20 hours for intravenous transfer into IgM^b-macroself mice. Alternatively, freshly isolated BM cells from mutant mice were depleted of red blood cells using ACK lysis buffer, counted, and resuspended in PBS. 1-10x10⁶ retrovirally transduced HPSCs or freshly isolated BM cells were iv injected into recipient mice. IgM^b-macroself mice were irradiated with 6 Gy by X-Ray before reconstitution. Recipient mice were euthanized and their splenocytes analyzed by flow cytometry 8 weeks after reconstitution.

Data were analyzed with an unpaired two-tailed Student’s t test or one-way ANOVA using the Prism software Version9 as indicated in the figure legends. Statistical significance was set at P<0.05.

WEHI-231 is a commonly used immature B cell line that undergoes cell cycle arrest at G0/G1 phase and apoptosis upon anti-IgM-induced BCR crosslinking (25). This mimics the recognition of self-antigens by and subsequent clonal deletion (apoptosis) of autoreactive B cells that occur in vivo under homeostatic conditions. We previously found that miR-148a protects WEHI-231 cells from BCR engagement-induced apoptosis triggered by anti-IgM stimulation. This recapitulated the role of miR-148a in B cell central tolerance, where it prevented clonal deletion of autoreactive B cells following self-antigen encounter (20). We further identified 119 protein-coding genes that contained predicted miR-148a binding sites in their mRNAs and showed at least 1.5-fold reduction in their mRNA levels upon miR-148a overexpression. Four of them were chosen for experimental validation. Among them, Bcl2l11 (encoding Bim), Pten and Gadd45a were validated as regulators of B cell tolerance, while Tnfrsf1b did not play any significant role in this process (20). In this study, we investigated the function of the other miR-148a target genes in B cell tolerance, first in the WEHI-231 system, followed by validation in animal models (Figure 1A). To this end, a retroviral TransOMIC shRNA library containing 288 shRNAs (Supplementary Table 1) covering 106 miR-148a target genes, with one to nine shRNAs for each gene, was generated. An empty vector (EV) and a vector encoding an shRNA against Bcl2l11 (shRNA-Bcl2l11) were included as negative and positive controls. The backbone vector used to generate this library contains a shRNA cassette followed by a puromycin resistance gene that enables puromycin selection of transduced cells (Figure 1B). Retroviruses were produced from individual constructs to generate the shRNA retroviral library as previously described (20).

The strategy for our in vitro functional screen is outlined in Figure 1A. WEHI-231 cells were transduced with retroviruses expressing individual shRNAs. After 48 hours, transduced cells were selected with puromycin for 5 days. Half of the cells were then stimulated with anti-IgM (2μg/ml) for 72 hours to induce BCR-engagement induced apoptosis, while the other half were left without stimulation and served as control. After 72 hours of stimulation, cells were harvested, stained for active Caspase 3, and analyzed by flow cytometry to quantify the percentage of apoptotic cells as a readout for clonal deletion. WEHI-231 cells transduced with empty retroviruses (WEHI-EV) and with retroviruses encoding shRNA-Bcl2l11 (WEHI-shRNA-Bcl2l11) were included in every independent experiment as controls.

The feasibility of this approach was first validated with WEHI-EV and WEHI-shRNA-Bcl2l11 controls. As expected, a significant fraction of non-transduced and WEHI-EV-transduced WEHI-231 cells died by apoptosis when stimulated with anti-IgM, and cell death was largely abrogated in shRNA-Bcl2l11-transduced cells (Figures 1C, D). This protection from clonal deletion by decreasing the expression levels of Bim recapitulated our previous results obtained in the IgM^b-macroself B cell tolerance mouse model (20).

Once the in vitro screen platform had been validated, 288 WEHI-231 stable cell lines expressing individual shRNAs for miR-148a target genes were generated. WEHI-231 cells were transduced with retroviruses encoding individual shRNAs, selected with puromycin, stimulated with anti-IgM, and analyzed for apoptosis as described above (Figure 1A). The effect of shRNAs that significantly impaired BCR engagement-induced apoptosis were confirmed in three independent experiments.

Among the 106 miR-148a target genes tested, nearly half of them showed an effect in protecting WEHI-231 cells from BCR-induced apoptosis when knocked down by shRNAs, with protection ratios ranging from 5% to 65% (Figure 2A). Among those, 22 shRNAs, which correspond to 13 genes, showed an average protection effect greater than 30% (Figure 2B and Figure 3). Of note, knockdown of B4galt5, Phip and Nsd1 showed the most robust protection against BCR engagement-induced apoptosis, with all shRNAs showing consistent effect (Figures 2C, D).

We next investigated the function of those positive hits in regulating B cell tolerance in vivo, by retroviral transduction of HSPCs and reconstitution of IgM^b-macroself mice. Briefly, C57BL/6J mice were treated with 5-fluorouracil (5-FU) for 5 days. HSPCs from treated mice were enriched, transduced with shRNA-encoding retroviruses, and transferred to sublethally irradiated IgM^b-macroself mice. After 8 weeks, the presence of peripheral B cells was analyzed in the spleen of these reconstituted mice by flow cytometry (Figure 4A). We set the presence of 5% splenic B cells as the threshold to consider an effect of breaking B cell tolerance because irradiated IgM^b-macroself mice reconstituted with wild-type C57BL/6J bone marrow cells show some background of splenic B cells after 8 weeks, up to this percentage in some experiments.

We observed a significant number of splenic B cells in 6 out of 16 (37%) and 6 out of 19 (32%) IgM^b-macroself mice reconstituted with HPSCs transduced with shRNA-B4galt5 and shRNA-Nsd1, respectively. Of note, a combination of two different shRNAs for B4galt5 was required for this effect, as transduction of HSPCs with individual shRNAs failed to break B cell tolerance (Supplementary Figure 1). The remaining mice did not show compromised B cell tolerance (Figures 4B, C). Although shRNA-Phip showed a significant protection against BCR engagement-induced apoptosis in WEHI-231 cells, those shRNAs did not show much effect in IgM^b-macroself mice, suggesting that Phip does not play any significant role in regulating B cell tolerance (Figures 4B, C). All other hits from the screen failed to break B cell tolerance in vivo.

To confirm that B4galt5 and Nsd1 indeed regulate B cell tolerance, we obtained conditional knockout (KO) mice with specific deletion of these genes in B cells. B4galt5^fl/+ and Nsd1^fl/+ mice were bred with Mb1Cre mice to drive specific deletion in the B cell lineage at the early pro-B cell stage. Deletion of these genes was validated in the resulting Mb1Cre;B4galt5^fl/fl and Mb1Cre;Nsd1^fl/fl mice by qRT-PCR analysis of splenic B cells (Figures 5A, B). Bone marrow cells from Mb1Cre;B4galt5^fl/fl and Mb1Cre;Nsd1^fl/fl mice were used to reconstitute irradiated IgM^b-macroself mice. Splenic B cells of recipient mice were analyzed 8 weeks after reconstitution. No accumulation of splenic B cells was observed in IgM^b-macroself mice reconstituted with bone marrow cells from Mb1Cre;B4galt5^fl/fl, Mb1Cre;Nsd1^fl/fl or their littermate controls B4galt5^fl/fl and Nsd1^fl/fl at terminal analysis (Figures 5C, G). Concurrently, bone marrow cells from heterozygous Mb1Cre;B4galt5^fl/+ also failed to break B cell tolerance in IgM^b-macroself mice (Supplementary Figure 2). This indicates that B4galt5 and Nsd1 do not play significant roles in the regulation of B cell tolerance. Taken together, our previous and this studies show that Bcl2l11, Pten and Gadd45a play key roles in mediating miR-148a-regulation of B cell central tolerance, with minimal, if any, contribution from the other target genes (20).