All animal work was approved by the Leiden University Animal Ethics Committee and the animal experiments were performed conform the guidelines from Directive 2010/63/EU of the European Parliament on the protection of animals used for scientific purposes. Female C57BL/6, B6.SJL-PtprcaPepcb/BoyCrl (also known as CD45.1) and apolipoprotein E-deficient (apoE^−/−) mice were bred in house and kept under standard laboratory conditions. Diet and water were provided ad libitum. All injections were administered i.v. to the lateral tail vein in a total volume of 100 μl. During the experiments, mice were weighed, and blood samples were obtained by tail vein bleeding. At the end of experiments, mice were anesthetized by a subcutaneous injection of a cocktail containing ketamine (40 mg/ml), atropine (0.1 mg/ml), and xylazine (8 mg/ml). Mice were bled and perfused with phosphate-buffered saline (PBS) through the left cardiac ventricle.

B cells were isolated from splenocytes of C57BL/6 or apoE^−/− mice using CD19⁺ microbeads (Miltenyi Biotec) and cultured in complete RPMI medium. Isolated B cells were cultured with different concentrations of heat killed Staphylococcus aureus (Invivogen), B-cell activating factor (BAFF; R&D systems) or interferon-gamma (IFNγ; ThermoFisher) for 24 h. For co-culture experiments, CD4⁺ T cells were isolated from splenocytes of C57BL/6 mice using a CD4⁺ T cell isolation kit (Miltenyi Biotec). For some experiments, the CD4⁺ T cells were labeled using a CellTrace Violet Cell Proliferation kit (ThermoFisher) according to the instructions of the manufacturer. B and T cells were co-cultured and stimulated with an agonistic plate-bound CD3 antibody (5 μg/ml) for 72 h. For adoptive transfer experiments, B cells were isolated from splenocytes of apoE^−/− or CD45.1 mice, cultured in RPMI medium and stimulated for 24 h with 20 ng/ml of IFNγ. Subsequently, cells were washed, checked for purity using flow cytometry and resuspended for injections with PBS. For the injection of untouched B cells, B cells were freshly isolated from splenocytes of apoE^−/− or CD45.1 mice and directly used for adoptive transfer experiments.

RNA was extracted from cultured B cells by using Trizol reagent according to manufacturer's instructions (Invitrogen) after which cDNA was generated using RevertAid M-MuLV reverse transcriptase according to the instructions of the manufacturer (Thermo Scientific). Quantitative gene expression analysis was performed using Power SYBR Green Master Mix on a 7500 Fast Real-Time PCR system (Applied Biosystems). Gene expression was normalized to housekeeping genes. Primer sequences are available in Supplementary Table 1.

For all in vivo experiments, apoE^−/− mice were used. ApoE^−/− mice were fed a Western-type diet (WTD) containing 0.25% cholesterol and 15% cacao butter (SDS, Sussex, UK). For the pilot study, mice were pre-fed a WTD for 2 weeks and were subsequently treated with 2 × 10⁶ freshly isolated or cultured B cells and sacrificed after 3 days. To assess the impact of WTD on the percentage of PD-L1^hi B cells, apoE^−/− mice were fed a chow diet or a WTD for 2 or 7 weeks. For atherosclerosis experiments, carotid artery plaque formation was induced after 2 weeks of WTD feeding by perivascular collar placement in these mice as described previously (17). Mice continued to be fed a WTD for 5 weeks and during this period received 3 injections with either freshly isolated B cells (2 × 10⁶ cells/injection), B cells stimulated with 20 ng/ml IFNγ for 24 h (2 × 10⁶ cells/injection) or PBS. To discriminate between adoptively transferred cells and endogenous cells, B cells were isolated from CD45.1 mice for the last injection. Time between injections was 2 weeks. At the end of experiments, mice were sacrificed and relevant organs were harvested for analysis.

Isolated splenocytes from PBS, B cell or IFNγ-B cell treated mice were cultured in complete RPMI medium and stimulated for 72 h with anti-CD3 (1 μg/ml) and anti-CD28 (0.5 μg/ml). The levels of cytokines in culture supernatants were measured using a Luminex bead-based multiplex assay (ProcartaPlex, Thermo Fisher Scientific) on a Luminex Instrument (MAGPIX). Recombinant cytokine standards (Thermo Fisher Scientific) were used to calculate cytokine concentrations and data were analyzed using Bio-Rad software.

For flow cytometry analysis, Fc receptors of single cell suspensions were blocked with an unconjugated antibody against CD16/CD32. Samples were then stained with a fixable viability marker (ThermoScientific) to select live cells. Next, cells were stained with anti-mouse fluorochrome-conjugated antibodies (Supplementary Table 2). Regular flow cytometry was performed on a Cytoflex S (Beckman Coulter) and the acquired data were analyzed using FlowJo software. Gates were set according to isotype or fluorescence minus one controls.

Serum was acquired by centrifugation and stored at −20°C until further use. Total serum titers of IgM, IgG1, IgG2c and oxidized LDL-specific antibodies were quantified by ELISA as previously described (5).

Carotid arteries and hearts were frozen in OCT compound (TissueTek) and stored at −80°C until further use. Transverse cryosections proximal to the collar were collected and mounted on Superfrost adhesion slides (ThermoFisher). To determine lesion size, cryosections were stained with hematoxylin and eosin (Sigma-Aldrich). Quantification of lesion size was assessed every 100 μm from the first section with visible lesion proximal to the collar until no lesion could be observed. Plaque volume was determined with lesion size and the total distance of the lesions in the carotid artery. Phenotypic analysis of the lesion was performed on sections containing the largest three lesions. Collagen content in the lesion was assessed with a Masson's trichrome staining according to the manufacturers protocol (Sigma-Aldrich). Necrotic core size was determined manually by selecting acellular areas in the Masson's trichrome stained sections and shown as absolute area and percentage of the total plaque area. Corresponding sections on separate slides were also stained for monocyte/macrophage content using a monoclonal rat IgG2b antibody (MOMA-2, 1:1000, AbD Serotec) followed by a goat anti-rat IgG-horseradish peroxidase antibody (1:100, Sigma-Aldrich) and color development using the ImmPACT NovaRED substrate (Vector Laboratories). For the detection of vascular smooth muscle cells, cryosections were stained with a monoclonal rat alpha-smooth muscle actin antibody conjugated to Alexa fluor 647 (1:1500, Novus Biologicals). Furthermore, cryosections were stained with a monoclonal rat CD4 antibody conjugated to FITC (1:150, eBioscience) and a monoclonal rat IgG2b isotype control conjugated to FITC (1:150, MBL) to determine CD4⁺ T cell infiltration. For all fluorescent images, cryosections were blocked with αCD16/32 Fc block (1:250, Biolegend) and nuclei were stained with DAPI. All slides were analyzed with a Leica DM-RE microscope and LeicaQwin software (Leica Imaging Systems).

All data are expressed as mean ± SEM. Data were tested for significance using a Student's t-test for two normally distributed groups. Data from three groups or more were analyzed by an ordinary one-way ANOVA test followed by Holm-Sidak post-hoc test. Probability values of p < 0.05 were considered significant. All statistical analyses were performed using GraphPad Prism.

Previously, it has been shown that hypercholesterolemia promotes PD-L1 expression on B cells in ldlr^−/− mice, which can regulate T_FH cell accumulation, limiting an exacerbated adaptive immune response (13). In order to investigate if the administration of a Western-type diet (WTD) also affects PD-L1-expressing B cells in apoE^−/− mice, we compared PD-L1 expression on B cells from chow fed apoE^−/− mice and apoE^−/− mice fed a WTD for 2 and 7 weeks using flow cytometry (Figure 1A; Supplementary Figure 1A). Whereas we observed no significant differences in PD-L1^hi B cells between chow fed apoE^−/− mice and apoE^−/− mice fed a WTD for 2 weeks, longer administration of WTD increases the percentage of PD-L1^hi B cells, in line with previous findings. Nonetheless, this elevation in PD-L1 expressing B cells upon hypercholesterolemia is not sufficient to halt disease development (13), indicating the need to further stimulate the regulation of the T_FH–GC B cell axis through the PD-1/PD-L1 pathway. We therefore aimed to generate a population of PD-L1^hi B cells ex vivo, to adoptively transfer in WTD fed apoE^−/− mice to halt atherosclerosis. We explored several stimuli that previously have been shown to control PD-L1 expression (18, 19) (Supplementary Figures 1B,C; Figure 1) and found that IFNγ dose-dependently increased the number of PD-L1⁺ B cells, with 20.0 ng/ml of IFNγ leading to an almost pure population of PD-L1-expressing B cells (Figure 1B). Using qPCR, we also found an almost eight-fold induction of PD-L1 on mRNA level after IFNγ stimulation (Figure 1C). Furthermore, we observed that the majority of IFNγ-stimulated B cells expressed very high levels of PD-L1 (Figure 1D). Previously, it has been shown that IFNγ-signaling in B cells drives STAT-1 dependent expression of T-bet and BCL-6 and switches them toward a GC B cell phenotype with increased IFNγ and IL-6 production (20, 21). In line with this, our ex vivo IFNγ-stimulated B cells (IFNγ-B cells) showed a similar increase in STAT1, T-bet and BCL-6 (Figure 1E) gene expression. In contrast, we observed a trend toward less IL-6 and no difference in IFNγ expression (Supplementary Figure 1D). Interestingly, IFNγ stimulation resulted in a strong significant increase in TGF-β expression (Figure 1E). We also investigated the chemokine receptor profile of these B cells and found a dose-dependent increase in CCR7 expression (Figure 1E), while there was no effect on CXCR5 or Ebi-2 (Supplementary Figure 1D). This change in chemokine receptor expression is typical of B cells that migrate toward the T-B cell border in lymphoid tissues in response to the CCL21 gradient (22). Hence, these data indicate that IFNγ-stimulated B cells express high levels of PD-L1 and TGF-β and a chemokine profile that homes B cells to the T-B cell border. In line with these findings, we observed that IFNγ stimulation of B cells not only induced coinhibitory PD-L1 expression on all B cells (Figure 1C) but also increased the percentage of GC B cells and MZ B cells (Figure 1F). Together with a decrease in FO B cells, we thus show that IFNγ stimulation generates a B cell population with an enhanced anti-inflammatory phenotype.

Next, we explored the functional effects of IFNγ-stimulated B cells on T_FH cell development using a CD4⁺ T cell and B cell coculture. We stimulated wild-type CD4⁺ T cells for 72 h with anti-CD3 in the presence of unstimulated or IFNγ-stimulated apoE^−/− B cells. Although we did not observe any difference in proliferative capacity of CD4⁺ T cells (Figure 2A), we found a remarkable decrease of T_FH cells when CD4⁺ T cells were cocultured with IFNγ-stimulated B cells compared to unstimulated B cells (Figure 2B). These findings illustrate that IFNγ-B cells are able to curb T_FH cell development in vitro. We subsequently tested IFNγ-stimulated B cells in an in vivo setting by adoptively transferring either freshly isolated B cells or B cells stimulated with 20.0 ng/ml of IFNγ for 24 h into apoE^−/− mice. To induce initial T_FH cell accumulation, mice were fed a Western-type diet for 2 weeks before they received the adoptive transfers (Figure 2C) (13). We observed a remarkable reduction in effector CD4⁺ T cells in mice treated with IFNγ-B cells compared to mice that received PBS or B cells (Figure 2D). Contrary, the number of naïve CD4⁺ T cells was increased in IFNγ-B cells treated mice compared to mice receiving PBS (Figure 2D). Most importantly, treatment of mice with IFNγ-B cells resulted in a strong reduction in T_FH cells compared to mice that were administered with PBS or B cells (Figure 2E). These data demonstrate that IFNγ-B cells are also able to inhibit T_FH cells in vivo.

Given the in vitro and in vivo regulatory effects of IFNγ-B cells on T_FH cells, we further investigated whether these B cells would be able to restrict T_FH cell numbers during atherosclerosis development. We fed apoE^−/− mice a Western-type diet for 2 weeks after which we placed a perivascular collar and started treatment with either PBS, B cells or IFNγ-B cells (Figure 3A). After a total of 3 injections and 7 weeks of Western-type diet, we harvested the organs and analyzed the immune cells associated with the T_FH–GC B cell axis. IFNγ-B cell-treated mice showed a trend toward reduced FO B cells (Figure 3B) and a significant increase in MZ B cells compared to mice receiving PBS (Figure 3C). Moreover, mice that received IFNγ-B cells showed a significant increase in the number of GC B cells compared to mice that were administered PBS (Figures 3D,E). By using the CD45 congenic marker system, we showed that adoptively transferred IFNγ-B cells indeed reached the germinal center (Figure 3F) and locally increased the number of PD-L1^hi B cells (Figure 3G). As shown in Figures 3D,E, freshly isolated B cells also increased the number of GC B cells, but these were not derived from the adoptive transfer (Figure 3F) and were PD-L1⁻ (Figure 3H). At sacrifice, we did not observe a difference in T_FH cells in IFNγ-B cell-treated mice, while adoptive transfer of B cells resulted in an increase in T_FH cells compared to PBS-treated mice (Figure 3I).

Next, we assessed the number of plasmablasts and plasma cells and found a significant reduction in plasmablasts when mice received IFNγ-B cells compared to mice receiving PBS (Figure 4A). Mice that received B cells showed a similar trend toward less plasmablasts and also a significant increase in plasma cells compared to PBS- and IFNγ-B cells-treated mice. Since BLIMP-1 is the driving transcription factor for plasma cell generation (23), we also measured BLIMP-1 expression which revealed a significant increase in BLIMP-1⁺ cells in mice that received B cells compared to mice receiving PBS or IFNγ-B cells (Figure 4B). Since plasmablasts and plasma cells are responsible for the humoral immunity, we measured circulating antibodies. However, neither B cell treatments led to a significant difference in circulating antibodies (Figure 4C).

The CD4⁺ T cell response is also highly involved in the pathogenesis of atherosclerosis and PD-L1^hi B cells have previously shown to restrict CD4⁺ T cell differentiation (16). While we did not observe any differences in Th1 or Th2 CD4⁺ T cells between mice treated with IFNγ-B cells, B cells or PBS in our study (Supplementary Figure 2), we observed a significant decrease in Th17 cells (Figure 4D) and a significant increase in atheroprotective regulatory T cells (Figure 4E; Tregs) in mice treated with IFNγ-B cells. We next measured cytokine levels of ex vivo anti-CD3 and anti-CD28 stimulated splenocytes for 72 h with a multiplex assay. In line with the increase in Tregs, we observed a significant increase in IL-10 production by splenocytes from IFNγ-B cells treated mice compared to splenocytes from mice treated with PBS or B cells (Figure 4F).

Given the immune-regulating effects of IFNγ-B cells, we subsequently assessed whether adoptive transfer of these IFNγ-B cells was able to reduce collar-induced atherosclerosis in apoE^−/− mice. We quantified lesion development in the carotid arteries and found a significant reduction in total lesion volume when mice were treated with IFNγ-B cells compared to PBS and a similar trend was found when compared to B cell-treated mice (Figure 5A). Media size did not differ between the treatment groups (Figure 5B). Furthermore, there were no differences in weight gain and serum cholesterol levels (Supplementary Figures 3A,B). We also assessed lesion phenotype at the site of the maximal lesion. This revealed that mice treated with IFNγ-B cells showed an early lesion phenotype with relatively more macrophages than collagen compared to mice that received B cells or PBS (Figures 5C,D), suggesting that adoptive transfer of IFNγ-B cells inhibited atherosclerotic lesion progression toward more advanced lesions. Evaluation of vascular smooth muscle cells (VSMCs) using α-smooth muscle actin did not reveal any differences (Supplementary Figure 4A) and we also did not observe altered necrotic core formation within the atherosclerotic plaque in IFNγ-B cell treated mice compared to the other groups (Supplementary Figure 4B). Finally, we did not observe significant differences in CD4⁺ T cell infiltration between the treatment groups (Supplementary Figure 4C) and only very low numbers of CD4⁺ T cells were found, suggesting that the reduction in lesion size is caused by systemic anti-atherogenic effects of PD-L1^hi B cells, rather than a local effect.